{ "best_metric": 0.026185082271695137, "best_model_checkpoint": "wav2vec2-large-pem123-960h-la/checkpoint-128583", "epoch": 2.999988334519324, "global_step": 128583, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 9e-08, "loss": 2.3526, "step": 5 }, { "epoch": 0.0, "learning_rate": 2.1e-07, "loss": 3.4983, "step": 10 }, { "epoch": 0.0, "learning_rate": 3.6e-07, "loss": 3.224, "step": 15 }, { "epoch": 0.0, "learning_rate": 5.100000000000001e-07, "loss": 3.4897, "step": 20 }, { "epoch": 0.0, "learning_rate": 6.599999999999999e-07, "loss": 3.5388, "step": 25 }, { "epoch": 0.0, "learning_rate": 8.1e-07, "loss": 3.5133, "step": 30 }, { "epoch": 0.0, "learning_rate": 9.3e-07, "loss": 3.3261, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.08e-06, "loss": 3.3158, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.23e-06, "loss": 3.8749, "step": 45 }, { "epoch": 0.0, "learning_rate": 1.35e-06, "loss": 3.2334, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.5e-06, "loss": 2.2091, "step": 55 }, { "epoch": 0.0, "learning_rate": 1.65e-06, "loss": 2.549, "step": 60 }, { "epoch": 0.0, "learning_rate": 1.8e-06, "loss": 2.5895, "step": 65 }, { "epoch": 0.0, "learning_rate": 1.95e-06, "loss": 2.3792, "step": 70 }, { "epoch": 0.0, "learning_rate": 2.1000000000000002e-06, "loss": 2.4595, "step": 75 }, { "epoch": 0.0, "learning_rate": 2.25e-06, "loss": 2.4939, "step": 80 }, { "epoch": 0.0, "learning_rate": 2.4000000000000003e-06, "loss": 2.196, "step": 85 }, { "epoch": 0.0, "learning_rate": 2.55e-06, "loss": 2.1306, "step": 90 }, { "epoch": 0.0, "learning_rate": 2.7e-06, "loss": 2.0683, "step": 95 }, { "epoch": 0.0, "learning_rate": 2.8500000000000002e-06, "loss": 2.3429, "step": 100 }, { "epoch": 0.0, "learning_rate": 3e-06, "loss": 1.3874, "step": 105 }, { "epoch": 0.0, "learning_rate": 3.15e-06, "loss": 1.321, "step": 110 }, { "epoch": 0.0, "learning_rate": 3.3e-06, "loss": 1.685, "step": 115 }, { "epoch": 0.0, "learning_rate": 3.4500000000000004e-06, "loss": 1.4842, "step": 120 }, { "epoch": 0.0, "learning_rate": 3.6e-06, "loss": 1.5263, "step": 125 }, { "epoch": 0.0, "learning_rate": 3.75e-06, "loss": 1.6352, "step": 130 }, { "epoch": 0.0, "learning_rate": 3.9e-06, "loss": 1.7527, "step": 135 }, { "epoch": 0.0, "learning_rate": 4.05e-06, "loss": 1.9171, "step": 140 }, { "epoch": 0.0, "learning_rate": 4.2000000000000004e-06, "loss": 1.7436, "step": 145 }, { "epoch": 0.0, "learning_rate": 4.35e-06, "loss": 2.4431, "step": 150 }, { "epoch": 0.0, "learning_rate": 4.5e-06, "loss": 1.5179, "step": 155 }, { "epoch": 0.0, "learning_rate": 4.65e-06, "loss": 1.1841, "step": 160 }, { "epoch": 0.0, "learning_rate": 4.800000000000001e-06, "loss": 1.3317, "step": 165 }, { "epoch": 0.0, "learning_rate": 4.95e-06, "loss": 1.3995, "step": 170 }, { "epoch": 0.0, "learning_rate": 5.1e-06, "loss": 1.3694, "step": 175 }, { "epoch": 0.0, "learning_rate": 5.25e-06, "loss": 1.4572, "step": 180 }, { "epoch": 0.0, "learning_rate": 5.4e-06, "loss": 1.4076, "step": 185 }, { "epoch": 0.0, "learning_rate": 5.55e-06, "loss": 1.3024, "step": 190 }, { "epoch": 0.0, "learning_rate": 5.7000000000000005e-06, "loss": 1.5938, "step": 195 }, { "epoch": 0.0, "learning_rate": 5.850000000000001e-06, "loss": 1.8611, "step": 200 }, { "epoch": 0.0, "learning_rate": 6e-06, "loss": 1.0738, "step": 205 }, { "epoch": 0.0, "learning_rate": 6.1499999999999996e-06, "loss": 1.2641, "step": 210 }, { "epoch": 0.01, "learning_rate": 6.3e-06, "loss": 1.1609, "step": 215 }, { "epoch": 0.01, "learning_rate": 6.45e-06, "loss": 1.0245, "step": 220 }, { "epoch": 0.01, "learning_rate": 6.6e-06, "loss": 1.159, "step": 225 }, { "epoch": 0.01, "learning_rate": 6.750000000000001e-06, "loss": 1.2281, "step": 230 }, { "epoch": 0.01, "learning_rate": 6.900000000000001e-06, "loss": 1.1537, "step": 235 }, { "epoch": 0.01, "learning_rate": 7.049999999999999e-06, "loss": 1.1336, "step": 240 }, { "epoch": 0.01, "learning_rate": 7.2e-06, "loss": 1.2782, "step": 245 }, { "epoch": 0.01, "learning_rate": 7.35e-06, "loss": 1.9498, "step": 250 }, { "epoch": 0.01, "learning_rate": 7.5e-06, "loss": 1.2166, "step": 255 }, { "epoch": 0.01, "learning_rate": 7.65e-06, "loss": 0.9303, "step": 260 }, { "epoch": 0.01, "learning_rate": 7.8e-06, "loss": 1.1386, "step": 265 }, { "epoch": 0.01, "learning_rate": 7.95e-06, "loss": 1.1336, "step": 270 }, { "epoch": 0.01, "learning_rate": 8.1e-06, "loss": 1.4349, "step": 275 }, { "epoch": 0.01, "learning_rate": 8.25e-06, "loss": 1.148, "step": 280 }, { "epoch": 0.01, "learning_rate": 8.400000000000001e-06, "loss": 1.3652, "step": 285 }, { "epoch": 0.01, "learning_rate": 8.55e-06, "loss": 1.0239, "step": 290 }, { "epoch": 0.01, "learning_rate": 8.7e-06, "loss": 1.4222, "step": 295 }, { "epoch": 0.01, "learning_rate": 8.85e-06, "loss": 1.7492, "step": 300 }, { "epoch": 0.01, "learning_rate": 9e-06, "loss": 1.1198, "step": 305 }, { "epoch": 0.01, "learning_rate": 9.15e-06, "loss": 1.0716, "step": 310 }, { "epoch": 0.01, "learning_rate": 9.3e-06, "loss": 0.9887, "step": 315 }, { "epoch": 0.01, "learning_rate": 9.450000000000001e-06, "loss": 0.9788, "step": 320 }, { "epoch": 0.01, "learning_rate": 9.600000000000001e-06, "loss": 1.197, "step": 325 }, { "epoch": 0.01, "learning_rate": 9.75e-06, "loss": 1.1805, "step": 330 }, { "epoch": 0.01, "learning_rate": 9.9e-06, "loss": 1.4189, "step": 335 }, { "epoch": 0.01, "learning_rate": 1.005e-05, "loss": 1.266, "step": 340 }, { "epoch": 0.01, "learning_rate": 1.02e-05, "loss": 1.4751, "step": 345 }, { "epoch": 0.01, "learning_rate": 1.035e-05, "loss": 1.5086, "step": 350 }, { "epoch": 0.01, "learning_rate": 1.05e-05, "loss": 1.0306, "step": 355 }, { "epoch": 0.01, "learning_rate": 1.065e-05, "loss": 0.817, "step": 360 }, { "epoch": 0.01, "learning_rate": 1.08e-05, "loss": 0.9599, "step": 365 }, { "epoch": 0.01, "learning_rate": 1.095e-05, "loss": 1.1482, "step": 370 }, { "epoch": 0.01, "learning_rate": 1.11e-05, "loss": 1.0518, "step": 375 }, { "epoch": 0.01, "learning_rate": 1.125e-05, "loss": 1.0029, "step": 380 }, { "epoch": 0.01, "learning_rate": 1.1400000000000001e-05, "loss": 1.1013, "step": 385 }, { "epoch": 0.01, "learning_rate": 1.1550000000000001e-05, "loss": 1.2762, "step": 390 }, { "epoch": 0.01, "learning_rate": 1.1700000000000001e-05, "loss": 1.3373, "step": 395 }, { "epoch": 0.01, "learning_rate": 1.185e-05, "loss": 1.522, "step": 400 }, { "epoch": 0.01, "learning_rate": 1.2e-05, "loss": 1.0738, "step": 405 }, { "epoch": 0.01, "learning_rate": 1.215e-05, "loss": 0.6606, "step": 410 }, { "epoch": 0.01, "learning_rate": 1.2299999999999999e-05, "loss": 0.8249, "step": 415 }, { "epoch": 0.01, "learning_rate": 1.245e-05, "loss": 0.938, "step": 420 }, { "epoch": 0.01, "learning_rate": 1.26e-05, "loss": 0.6348, "step": 425 }, { "epoch": 0.01, "learning_rate": 1.275e-05, "loss": 1.1462, "step": 430 }, { "epoch": 0.01, "learning_rate": 1.29e-05, "loss": 0.9481, "step": 435 }, { "epoch": 0.01, "learning_rate": 1.305e-05, "loss": 0.9554, "step": 440 }, { "epoch": 0.01, "learning_rate": 1.32e-05, "loss": 1.2571, "step": 445 }, { "epoch": 0.01, "learning_rate": 1.3350000000000001e-05, "loss": 1.3792, "step": 450 }, { "epoch": 0.01, "learning_rate": 1.3500000000000001e-05, "loss": 0.9654, "step": 455 }, { "epoch": 0.01, "learning_rate": 1.3650000000000001e-05, "loss": 0.7258, "step": 460 }, { "epoch": 0.01, "learning_rate": 1.3800000000000002e-05, "loss": 0.8351, "step": 465 }, { "epoch": 0.01, "learning_rate": 1.395e-05, "loss": 0.6394, "step": 470 }, { "epoch": 0.01, "learning_rate": 1.4099999999999999e-05, "loss": 0.7512, "step": 475 }, { "epoch": 0.01, "learning_rate": 1.4249999999999999e-05, "loss": 0.9593, "step": 480 }, { "epoch": 0.01, "learning_rate": 1.44e-05, "loss": 1.0303, "step": 485 }, { "epoch": 0.01, "learning_rate": 1.455e-05, "loss": 1.112, "step": 490 }, { "epoch": 0.01, "learning_rate": 1.47e-05, "loss": 1.3478, "step": 495 }, { "epoch": 0.01, "learning_rate": 1.485e-05, "loss": 1.4505, "step": 500 }, { "epoch": 0.01, "learning_rate": 1.5e-05, "loss": 0.9837, "step": 505 }, { "epoch": 0.01, "learning_rate": 1.515e-05, "loss": 0.6547, "step": 510 }, { "epoch": 0.01, "learning_rate": 1.53e-05, "loss": 0.7369, "step": 515 }, { "epoch": 0.01, "learning_rate": 1.545e-05, "loss": 0.7089, "step": 520 }, { "epoch": 0.01, "learning_rate": 1.56e-05, "loss": 0.9379, "step": 525 }, { "epoch": 0.01, "learning_rate": 1.575e-05, "loss": 0.7101, "step": 530 }, { "epoch": 0.01, "learning_rate": 1.59e-05, "loss": 0.7551, "step": 535 }, { "epoch": 0.01, "learning_rate": 1.605e-05, "loss": 1.1333, "step": 540 }, { "epoch": 0.01, "learning_rate": 1.62e-05, "loss": 1.0296, "step": 545 }, { "epoch": 0.01, "learning_rate": 1.635e-05, "loss": 1.7001, "step": 550 }, { "epoch": 0.01, "learning_rate": 1.65e-05, "loss": 0.892, "step": 555 }, { "epoch": 0.01, "learning_rate": 1.665e-05, "loss": 0.4653, "step": 560 }, { "epoch": 0.01, "learning_rate": 1.6800000000000002e-05, "loss": 0.7662, "step": 565 }, { "epoch": 0.01, "learning_rate": 1.695e-05, "loss": 0.7403, "step": 570 }, { "epoch": 0.01, "learning_rate": 1.71e-05, "loss": 0.8761, "step": 575 }, { "epoch": 0.01, "learning_rate": 1.725e-05, "loss": 0.769, "step": 580 }, { "epoch": 0.01, "learning_rate": 1.74e-05, "loss": 1.0235, "step": 585 }, { "epoch": 0.01, "learning_rate": 1.755e-05, "loss": 0.9654, "step": 590 }, { "epoch": 0.01, "learning_rate": 1.77e-05, "loss": 1.1367, "step": 595 }, { "epoch": 0.01, "learning_rate": 1.785e-05, "loss": 1.4638, "step": 600 }, { "epoch": 0.01, "learning_rate": 1.8e-05, "loss": 0.8417, "step": 605 }, { "epoch": 0.01, "learning_rate": 1.815e-05, "loss": 0.5591, "step": 610 }, { "epoch": 0.01, "learning_rate": 1.83e-05, "loss": 0.836, "step": 615 }, { "epoch": 0.01, "learning_rate": 1.845e-05, "loss": 0.8557, "step": 620 }, { "epoch": 0.01, "learning_rate": 1.86e-05, "loss": 0.8181, "step": 625 }, { "epoch": 0.01, "learning_rate": 1.8750000000000002e-05, "loss": 0.8651, "step": 630 }, { "epoch": 0.01, "learning_rate": 1.8900000000000002e-05, "loss": 1.0221, "step": 635 }, { "epoch": 0.01, "learning_rate": 1.9050000000000002e-05, "loss": 1.2453, "step": 640 }, { "epoch": 0.02, "learning_rate": 1.9200000000000003e-05, "loss": 1.1206, "step": 645 }, { "epoch": 0.02, "learning_rate": 1.935e-05, "loss": 1.1604, "step": 650 }, { "epoch": 0.02, "learning_rate": 1.95e-05, "loss": 0.827, "step": 655 }, { "epoch": 0.02, "learning_rate": 1.965e-05, "loss": 0.6968, "step": 660 }, { "epoch": 0.02, "learning_rate": 1.98e-05, "loss": 0.6557, "step": 665 }, { "epoch": 0.02, "learning_rate": 1.995e-05, "loss": 0.8323, "step": 670 }, { "epoch": 0.02, "learning_rate": 2.01e-05, "loss": 0.8274, "step": 675 }, { "epoch": 0.02, "learning_rate": 2.025e-05, "loss": 0.9101, "step": 680 }, { "epoch": 0.02, "learning_rate": 2.04e-05, "loss": 0.9683, "step": 685 }, { "epoch": 0.02, "learning_rate": 2.055e-05, "loss": 0.8473, "step": 690 }, { "epoch": 0.02, "learning_rate": 2.07e-05, "loss": 1.1503, "step": 695 }, { "epoch": 0.02, "learning_rate": 2.085e-05, "loss": 1.5373, "step": 700 }, { "epoch": 0.02, "learning_rate": 2.1e-05, "loss": 0.7089, "step": 705 }, { "epoch": 0.02, "learning_rate": 2.115e-05, "loss": 0.5048, "step": 710 }, { "epoch": 0.02, "learning_rate": 2.13e-05, "loss": 0.6618, "step": 715 }, { "epoch": 0.02, "learning_rate": 2.145e-05, "loss": 0.7328, "step": 720 }, { "epoch": 0.02, "learning_rate": 2.16e-05, "loss": 0.9687, "step": 725 }, { "epoch": 0.02, "learning_rate": 2.175e-05, "loss": 0.8333, "step": 730 }, { "epoch": 0.02, "learning_rate": 2.19e-05, "loss": 0.8595, "step": 735 }, { "epoch": 0.02, "learning_rate": 2.205e-05, "loss": 0.9809, "step": 740 }, { "epoch": 0.02, "learning_rate": 2.22e-05, "loss": 0.8315, "step": 745 }, { "epoch": 0.02, "learning_rate": 2.235e-05, "loss": 1.5107, "step": 750 }, { "epoch": 0.02, "learning_rate": 2.25e-05, "loss": 0.8627, "step": 755 }, { "epoch": 0.02, "learning_rate": 2.265e-05, "loss": 0.6485, "step": 760 }, { "epoch": 0.02, "learning_rate": 2.2800000000000002e-05, "loss": 0.6932, "step": 765 }, { "epoch": 0.02, "learning_rate": 2.2950000000000002e-05, "loss": 0.7105, "step": 770 }, { "epoch": 0.02, "learning_rate": 2.3100000000000002e-05, "loss": 0.7804, "step": 775 }, { "epoch": 0.02, "learning_rate": 2.3250000000000003e-05, "loss": 0.548, "step": 780 }, { "epoch": 0.02, "learning_rate": 2.3400000000000003e-05, "loss": 0.867, "step": 785 }, { "epoch": 0.02, "learning_rate": 2.3550000000000003e-05, "loss": 0.9944, "step": 790 }, { "epoch": 0.02, "learning_rate": 2.37e-05, "loss": 1.3024, "step": 795 }, { "epoch": 0.02, "learning_rate": 2.385e-05, "loss": 1.3287, "step": 800 }, { "epoch": 0.02, "learning_rate": 2.4e-05, "loss": 0.8523, "step": 805 }, { "epoch": 0.02, "learning_rate": 2.415e-05, "loss": 0.5302, "step": 810 }, { "epoch": 0.02, "learning_rate": 2.43e-05, "loss": 0.606, "step": 815 }, { "epoch": 0.02, "learning_rate": 2.4449999999999998e-05, "loss": 0.7036, "step": 820 }, { "epoch": 0.02, "learning_rate": 2.4599999999999998e-05, "loss": 0.5578, "step": 825 }, { "epoch": 0.02, "learning_rate": 2.475e-05, "loss": 0.7601, "step": 830 }, { "epoch": 0.02, "learning_rate": 2.49e-05, "loss": 1.0306, "step": 835 }, { "epoch": 0.02, "learning_rate": 2.505e-05, "loss": 0.8348, "step": 840 }, { "epoch": 0.02, "learning_rate": 2.52e-05, "loss": 1.1384, "step": 845 }, { "epoch": 0.02, "learning_rate": 2.535e-05, "loss": 1.4237, "step": 850 }, { "epoch": 0.02, "learning_rate": 2.55e-05, "loss": 0.8078, "step": 855 }, { "epoch": 0.02, "learning_rate": 2.565e-05, "loss": 0.6075, "step": 860 }, { "epoch": 0.02, "learning_rate": 2.58e-05, "loss": 0.3842, "step": 865 }, { "epoch": 0.02, "learning_rate": 2.595e-05, "loss": 0.6575, "step": 870 }, { "epoch": 0.02, "learning_rate": 2.61e-05, "loss": 0.7133, "step": 875 }, { "epoch": 0.02, "learning_rate": 2.625e-05, "loss": 0.8264, "step": 880 }, { "epoch": 0.02, "learning_rate": 2.64e-05, "loss": 0.9861, "step": 885 }, { "epoch": 0.02, "learning_rate": 2.655e-05, "loss": 1.0248, "step": 890 }, { "epoch": 0.02, "learning_rate": 2.6700000000000002e-05, "loss": 1.0116, "step": 895 }, { "epoch": 0.02, "learning_rate": 2.6850000000000002e-05, "loss": 1.3839, "step": 900 }, { "epoch": 0.02, "learning_rate": 2.7000000000000002e-05, "loss": 0.7682, "step": 905 }, { "epoch": 0.02, "learning_rate": 2.7150000000000003e-05, "loss": 0.5031, "step": 910 }, { "epoch": 0.02, "learning_rate": 2.7300000000000003e-05, "loss": 0.5735, "step": 915 }, { "epoch": 0.02, "learning_rate": 2.7450000000000003e-05, "loss": 0.5872, "step": 920 }, { "epoch": 0.02, "learning_rate": 2.7600000000000003e-05, "loss": 0.7418, "step": 925 }, { "epoch": 0.02, "learning_rate": 2.7750000000000004e-05, "loss": 0.7089, "step": 930 }, { "epoch": 0.02, "learning_rate": 2.79e-05, "loss": 0.7709, "step": 935 }, { "epoch": 0.02, "learning_rate": 2.805e-05, "loss": 1.0289, "step": 940 }, { "epoch": 0.02, "learning_rate": 2.8199999999999998e-05, "loss": 1.0879, "step": 945 }, { "epoch": 0.02, "learning_rate": 2.8349999999999998e-05, "loss": 1.5505, "step": 950 }, { "epoch": 0.02, "learning_rate": 2.8499999999999998e-05, "loss": 0.7483, "step": 955 }, { "epoch": 0.02, "learning_rate": 2.865e-05, "loss": 0.4613, "step": 960 }, { "epoch": 0.02, "learning_rate": 2.88e-05, "loss": 0.4939, "step": 965 }, { "epoch": 0.02, "learning_rate": 2.895e-05, "loss": 0.66, "step": 970 }, { "epoch": 0.02, "learning_rate": 2.91e-05, "loss": 0.639, "step": 975 }, { "epoch": 0.02, "learning_rate": 2.925e-05, "loss": 0.5705, "step": 980 }, { "epoch": 0.02, "learning_rate": 2.94e-05, "loss": 0.9141, "step": 985 }, { "epoch": 0.02, "learning_rate": 2.955e-05, "loss": 0.9266, "step": 990 }, { "epoch": 0.02, "learning_rate": 2.97e-05, "loss": 0.9949, "step": 995 }, { "epoch": 0.02, "learning_rate": 2.985e-05, "loss": 1.1391, "step": 1000 }, { "epoch": 0.02, "learning_rate": 3e-05, "loss": 0.7045, "step": 1005 }, { "epoch": 0.02, "learning_rate": 2.9999119945553966e-05, "loss": 0.4402, "step": 1010 }, { "epoch": 0.02, "learning_rate": 2.999823989110793e-05, "loss": 0.5648, "step": 1015 }, { "epoch": 0.02, "learning_rate": 2.9997359836661896e-05, "loss": 0.4815, "step": 1020 }, { "epoch": 0.02, "learning_rate": 2.999647978221586e-05, "loss": 0.6272, "step": 1025 }, { "epoch": 0.02, "learning_rate": 2.9995599727769826e-05, "loss": 0.7295, "step": 1030 }, { "epoch": 0.02, "learning_rate": 2.999471967332379e-05, "loss": 0.8361, "step": 1035 }, { "epoch": 0.02, "learning_rate": 2.9993839618877756e-05, "loss": 0.7377, "step": 1040 }, { "epoch": 0.02, "learning_rate": 2.999295956443172e-05, "loss": 0.9675, "step": 1045 }, { "epoch": 0.02, "learning_rate": 2.9992079509985683e-05, "loss": 1.5494, "step": 1050 }, { "epoch": 0.02, "learning_rate": 2.999119945553965e-05, "loss": 0.7034, "step": 1055 }, { "epoch": 0.02, "learning_rate": 2.9990319401093616e-05, "loss": 0.4125, "step": 1060 }, { "epoch": 0.02, "learning_rate": 2.998943934664758e-05, "loss": 0.5159, "step": 1065 }, { "epoch": 0.02, "learning_rate": 2.9988559292201543e-05, "loss": 0.451, "step": 1070 }, { "epoch": 0.03, "learning_rate": 2.998767923775551e-05, "loss": 0.5905, "step": 1075 }, { "epoch": 0.03, "learning_rate": 2.9986799183309476e-05, "loss": 0.5791, "step": 1080 }, { "epoch": 0.03, "learning_rate": 2.9985919128863438e-05, "loss": 0.6041, "step": 1085 }, { "epoch": 0.03, "learning_rate": 2.9985039074417403e-05, "loss": 0.8075, "step": 1090 }, { "epoch": 0.03, "learning_rate": 2.998415901997137e-05, "loss": 1.1991, "step": 1095 }, { "epoch": 0.03, "learning_rate": 2.9983278965525336e-05, "loss": 1.8861, "step": 1100 }, { "epoch": 0.03, "learning_rate": 2.9982398911079298e-05, "loss": 0.694, "step": 1105 }, { "epoch": 0.03, "learning_rate": 2.9981518856633263e-05, "loss": 0.351, "step": 1110 }, { "epoch": 0.03, "learning_rate": 2.998063880218723e-05, "loss": 0.5766, "step": 1115 }, { "epoch": 0.03, "learning_rate": 2.9979758747741193e-05, "loss": 0.6231, "step": 1120 }, { "epoch": 0.03, "learning_rate": 2.9978878693295158e-05, "loss": 0.6312, "step": 1125 }, { "epoch": 0.03, "learning_rate": 2.9977998638849123e-05, "loss": 0.7563, "step": 1130 }, { "epoch": 0.03, "learning_rate": 2.997711858440309e-05, "loss": 0.6477, "step": 1135 }, { "epoch": 0.03, "learning_rate": 2.9976238529957053e-05, "loss": 0.6872, "step": 1140 }, { "epoch": 0.03, "learning_rate": 2.9975358475511018e-05, "loss": 1.0495, "step": 1145 }, { "epoch": 0.03, "learning_rate": 2.9974478421064986e-05, "loss": 1.7174, "step": 1150 }, { "epoch": 0.03, "learning_rate": 2.9973598366618948e-05, "loss": 0.6995, "step": 1155 }, { "epoch": 0.03, "learning_rate": 2.9972718312172913e-05, "loss": 0.3495, "step": 1160 }, { "epoch": 0.03, "learning_rate": 2.9971838257726878e-05, "loss": 0.5469, "step": 1165 }, { "epoch": 0.03, "learning_rate": 2.9970958203280846e-05, "loss": 0.5165, "step": 1170 }, { "epoch": 0.03, "learning_rate": 2.9970078148834808e-05, "loss": 0.7759, "step": 1175 }, { "epoch": 0.03, "learning_rate": 2.9969198094388773e-05, "loss": 0.6411, "step": 1180 }, { "epoch": 0.03, "learning_rate": 2.9968318039942738e-05, "loss": 0.8374, "step": 1185 }, { "epoch": 0.03, "learning_rate": 2.9967437985496707e-05, "loss": 0.8479, "step": 1190 }, { "epoch": 0.03, "learning_rate": 2.9966557931050668e-05, "loss": 0.8935, "step": 1195 }, { "epoch": 0.03, "learning_rate": 2.9965677876604633e-05, "loss": 1.1898, "step": 1200 }, { "epoch": 0.03, "learning_rate": 2.9964797822158598e-05, "loss": 0.8111, "step": 1205 }, { "epoch": 0.03, "learning_rate": 2.9963917767712563e-05, "loss": 0.4952, "step": 1210 }, { "epoch": 0.03, "learning_rate": 2.9963037713266528e-05, "loss": 0.4721, "step": 1215 }, { "epoch": 0.03, "learning_rate": 2.9962157658820493e-05, "loss": 0.578, "step": 1220 }, { "epoch": 0.03, "learning_rate": 2.9961277604374458e-05, "loss": 0.7031, "step": 1225 }, { "epoch": 0.03, "learning_rate": 2.9960397549928423e-05, "loss": 0.68, "step": 1230 }, { "epoch": 0.03, "learning_rate": 2.995951749548239e-05, "loss": 0.8138, "step": 1235 }, { "epoch": 0.03, "learning_rate": 2.9958637441036353e-05, "loss": 0.9367, "step": 1240 }, { "epoch": 0.03, "learning_rate": 2.9957757386590315e-05, "loss": 0.9518, "step": 1245 }, { "epoch": 0.03, "learning_rate": 2.9956877332144283e-05, "loss": 1.6519, "step": 1250 }, { "epoch": 0.03, "learning_rate": 2.995599727769825e-05, "loss": 0.7183, "step": 1255 }, { "epoch": 0.03, "learning_rate": 2.9955117223252213e-05, "loss": 0.4444, "step": 1260 }, { "epoch": 0.03, "learning_rate": 2.9954237168806175e-05, "loss": 0.3977, "step": 1265 }, { "epoch": 0.03, "learning_rate": 2.9953357114360143e-05, "loss": 0.6202, "step": 1270 }, { "epoch": 0.03, "learning_rate": 2.995247705991411e-05, "loss": 0.5986, "step": 1275 }, { "epoch": 0.03, "learning_rate": 2.995159700546807e-05, "loss": 0.4864, "step": 1280 }, { "epoch": 0.03, "learning_rate": 2.995071695102204e-05, "loss": 0.675, "step": 1285 }, { "epoch": 0.03, "learning_rate": 2.9949836896576004e-05, "loss": 0.8216, "step": 1290 }, { "epoch": 0.03, "learning_rate": 2.994895684212997e-05, "loss": 0.9698, "step": 1295 }, { "epoch": 0.03, "learning_rate": 2.994807678768393e-05, "loss": 1.4193, "step": 1300 }, { "epoch": 0.03, "learning_rate": 2.99471967332379e-05, "loss": 0.7927, "step": 1305 }, { "epoch": 0.03, "learning_rate": 2.9946316678791864e-05, "loss": 0.5693, "step": 1310 }, { "epoch": 0.03, "learning_rate": 2.9945436624345825e-05, "loss": 0.4667, "step": 1315 }, { "epoch": 0.03, "learning_rate": 2.994455656989979e-05, "loss": 0.527, "step": 1320 }, { "epoch": 0.03, "learning_rate": 2.994367651545376e-05, "loss": 0.5844, "step": 1325 }, { "epoch": 0.03, "learning_rate": 2.9942796461007724e-05, "loss": 0.5947, "step": 1330 }, { "epoch": 0.03, "learning_rate": 2.9941916406561685e-05, "loss": 0.7047, "step": 1335 }, { "epoch": 0.03, "learning_rate": 2.994103635211565e-05, "loss": 0.7821, "step": 1340 }, { "epoch": 0.03, "learning_rate": 2.994015629766962e-05, "loss": 0.7723, "step": 1345 }, { "epoch": 0.03, "learning_rate": 2.993927624322358e-05, "loss": 1.3213, "step": 1350 }, { "epoch": 0.03, "learning_rate": 2.9938396188777545e-05, "loss": 0.7449, "step": 1355 }, { "epoch": 0.03, "learning_rate": 2.993751613433151e-05, "loss": 0.3073, "step": 1360 }, { "epoch": 0.03, "learning_rate": 2.993663607988548e-05, "loss": 0.5344, "step": 1365 }, { "epoch": 0.03, "learning_rate": 2.993575602543944e-05, "loss": 0.5691, "step": 1370 }, { "epoch": 0.03, "learning_rate": 2.9934875970993406e-05, "loss": 0.5009, "step": 1375 }, { "epoch": 0.03, "learning_rate": 2.993399591654737e-05, "loss": 0.6358, "step": 1380 }, { "epoch": 0.03, "learning_rate": 2.9933115862101336e-05, "loss": 0.6367, "step": 1385 }, { "epoch": 0.03, "learning_rate": 2.99322358076553e-05, "loss": 0.8328, "step": 1390 }, { "epoch": 0.03, "learning_rate": 2.9931355753209266e-05, "loss": 0.8655, "step": 1395 }, { "epoch": 0.03, "learning_rate": 2.993047569876323e-05, "loss": 1.1673, "step": 1400 }, { "epoch": 0.03, "learning_rate": 2.9929595644317196e-05, "loss": 0.8164, "step": 1405 }, { "epoch": 0.03, "learning_rate": 2.992871558987116e-05, "loss": 0.4538, "step": 1410 }, { "epoch": 0.03, "learning_rate": 2.9927835535425126e-05, "loss": 0.4349, "step": 1415 }, { "epoch": 0.03, "learning_rate": 2.992695548097909e-05, "loss": 0.5661, "step": 1420 }, { "epoch": 0.03, "learning_rate": 2.9926075426533056e-05, "loss": 0.6059, "step": 1425 }, { "epoch": 0.03, "learning_rate": 2.992519537208702e-05, "loss": 0.6103, "step": 1430 }, { "epoch": 0.03, "learning_rate": 2.9924315317640986e-05, "loss": 0.6356, "step": 1435 }, { "epoch": 0.03, "learning_rate": 2.992343526319495e-05, "loss": 0.9613, "step": 1440 }, { "epoch": 0.03, "learning_rate": 2.9922555208748916e-05, "loss": 0.8891, "step": 1445 }, { "epoch": 0.03, "learning_rate": 2.992167515430288e-05, "loss": 1.209, "step": 1450 }, { "epoch": 0.03, "learning_rate": 2.9920795099856842e-05, "loss": 0.6586, "step": 1455 }, { "epoch": 0.03, "learning_rate": 2.991991504541081e-05, "loss": 0.5212, "step": 1460 }, { "epoch": 0.03, "learning_rate": 2.9919034990964776e-05, "loss": 0.4062, "step": 1465 }, { "epoch": 0.03, "learning_rate": 2.991815493651874e-05, "loss": 0.4997, "step": 1470 }, { "epoch": 0.03, "learning_rate": 2.9917274882072703e-05, "loss": 0.491, "step": 1475 }, { "epoch": 0.03, "learning_rate": 2.991639482762667e-05, "loss": 0.6823, "step": 1480 }, { "epoch": 0.03, "learning_rate": 2.9915514773180636e-05, "loss": 0.6227, "step": 1485 }, { "epoch": 0.03, "learning_rate": 2.99146347187346e-05, "loss": 0.8315, "step": 1490 }, { "epoch": 0.03, "learning_rate": 2.9913754664288563e-05, "loss": 0.7643, "step": 1495 }, { "epoch": 0.03, "learning_rate": 2.991287460984253e-05, "loss": 0.8734, "step": 1500 }, { "epoch": 0.04, "learning_rate": 2.9911994555396496e-05, "loss": 0.633, "step": 1505 }, { "epoch": 0.04, "learning_rate": 2.9911114500950458e-05, "loss": 0.4566, "step": 1510 }, { "epoch": 0.04, "learning_rate": 2.9910234446504423e-05, "loss": 0.4848, "step": 1515 }, { "epoch": 0.04, "learning_rate": 2.990935439205839e-05, "loss": 0.6267, "step": 1520 }, { "epoch": 0.04, "learning_rate": 2.9908474337612356e-05, "loss": 0.6268, "step": 1525 }, { "epoch": 0.04, "learning_rate": 2.9907594283166318e-05, "loss": 0.6684, "step": 1530 }, { "epoch": 0.04, "learning_rate": 2.9906714228720283e-05, "loss": 0.6747, "step": 1535 }, { "epoch": 0.04, "learning_rate": 2.9906010185163457e-05, "loss": 0.7817, "step": 1540 }, { "epoch": 0.04, "learning_rate": 2.990513013071742e-05, "loss": 0.99, "step": 1545 }, { "epoch": 0.04, "learning_rate": 2.9904250076271384e-05, "loss": 1.3126, "step": 1550 }, { "epoch": 0.04, "learning_rate": 2.9903370021825353e-05, "loss": 0.674, "step": 1555 }, { "epoch": 0.04, "learning_rate": 2.9902489967379318e-05, "loss": 0.4815, "step": 1560 }, { "epoch": 0.04, "learning_rate": 2.990160991293328e-05, "loss": 0.3393, "step": 1565 }, { "epoch": 0.04, "learning_rate": 2.9900729858487244e-05, "loss": 0.4135, "step": 1570 }, { "epoch": 0.04, "learning_rate": 2.9899849804041213e-05, "loss": 0.5487, "step": 1575 }, { "epoch": 0.04, "learning_rate": 2.9898969749595174e-05, "loss": 0.6153, "step": 1580 }, { "epoch": 0.04, "learning_rate": 2.989808969514914e-05, "loss": 0.6472, "step": 1585 }, { "epoch": 0.04, "learning_rate": 2.9897209640703104e-05, "loss": 0.5356, "step": 1590 }, { "epoch": 0.04, "learning_rate": 2.9896329586257073e-05, "loss": 0.612, "step": 1595 }, { "epoch": 0.04, "learning_rate": 2.9895449531811034e-05, "loss": 1.143, "step": 1600 }, { "epoch": 0.04, "learning_rate": 2.9894569477365e-05, "loss": 0.6987, "step": 1605 }, { "epoch": 0.04, "learning_rate": 2.9893689422918968e-05, "loss": 0.5671, "step": 1610 }, { "epoch": 0.04, "learning_rate": 2.989280936847293e-05, "loss": 0.3659, "step": 1615 }, { "epoch": 0.04, "learning_rate": 2.9891929314026894e-05, "loss": 0.4344, "step": 1620 }, { "epoch": 0.04, "learning_rate": 2.989104925958086e-05, "loss": 0.5351, "step": 1625 }, { "epoch": 0.04, "learning_rate": 2.9890169205134828e-05, "loss": 0.5789, "step": 1630 }, { "epoch": 0.04, "learning_rate": 2.988928915068879e-05, "loss": 0.5416, "step": 1635 }, { "epoch": 0.04, "learning_rate": 2.9888409096242755e-05, "loss": 0.806, "step": 1640 }, { "epoch": 0.04, "learning_rate": 2.9887705052685926e-05, "loss": 0.9601, "step": 1645 }, { "epoch": 0.04, "learning_rate": 2.9886824998239894e-05, "loss": 1.1031, "step": 1650 }, { "epoch": 0.04, "learning_rate": 2.9885944943793856e-05, "loss": 0.6534, "step": 1655 }, { "epoch": 0.04, "learning_rate": 2.988506488934782e-05, "loss": 0.485, "step": 1660 }, { "epoch": 0.04, "learning_rate": 2.988418483490179e-05, "loss": 0.5737, "step": 1665 }, { "epoch": 0.04, "learning_rate": 2.988330478045575e-05, "loss": 0.6349, "step": 1670 }, { "epoch": 0.04, "learning_rate": 2.9882424726009716e-05, "loss": 0.448, "step": 1675 }, { "epoch": 0.04, "learning_rate": 2.988154467156368e-05, "loss": 0.5128, "step": 1680 }, { "epoch": 0.04, "learning_rate": 2.988066461711765e-05, "loss": 0.6134, "step": 1685 }, { "epoch": 0.04, "learning_rate": 2.987978456267161e-05, "loss": 0.5839, "step": 1690 }, { "epoch": 0.04, "learning_rate": 2.9878904508225576e-05, "loss": 0.8238, "step": 1695 }, { "epoch": 0.04, "learning_rate": 2.987802445377954e-05, "loss": 1.1745, "step": 1700 }, { "epoch": 0.04, "learning_rate": 2.9877144399333506e-05, "loss": 0.5163, "step": 1705 }, { "epoch": 0.04, "learning_rate": 2.987626434488747e-05, "loss": 0.2333, "step": 1710 }, { "epoch": 0.04, "learning_rate": 2.9875384290441436e-05, "loss": 0.3696, "step": 1715 }, { "epoch": 0.04, "learning_rate": 2.98745042359954e-05, "loss": 0.5512, "step": 1720 }, { "epoch": 0.04, "learning_rate": 2.9873624181549366e-05, "loss": 0.515, "step": 1725 }, { "epoch": 0.04, "learning_rate": 2.987274412710333e-05, "loss": 0.5851, "step": 1730 }, { "epoch": 0.04, "learning_rate": 2.9871864072657296e-05, "loss": 0.7857, "step": 1735 }, { "epoch": 0.04, "learning_rate": 2.9870984018211258e-05, "loss": 0.5871, "step": 1740 }, { "epoch": 0.04, "learning_rate": 2.9870103963765226e-05, "loss": 0.8909, "step": 1745 }, { "epoch": 0.04, "learning_rate": 2.986922390931919e-05, "loss": 0.83, "step": 1750 }, { "epoch": 0.04, "learning_rate": 2.9868343854873156e-05, "loss": 0.5291, "step": 1755 }, { "epoch": 0.04, "learning_rate": 2.9867463800427118e-05, "loss": 0.4018, "step": 1760 }, { "epoch": 0.04, "learning_rate": 2.9866583745981086e-05, "loss": 0.4395, "step": 1765 }, { "epoch": 0.04, "learning_rate": 2.986570369153505e-05, "loss": 0.4486, "step": 1770 }, { "epoch": 0.04, "learning_rate": 2.9864823637089013e-05, "loss": 0.4359, "step": 1775 }, { "epoch": 0.04, "learning_rate": 2.9863943582642978e-05, "loss": 0.6623, "step": 1780 }, { "epoch": 0.04, "learning_rate": 2.9863063528196946e-05, "loss": 0.7832, "step": 1785 }, { "epoch": 0.04, "learning_rate": 2.986218347375091e-05, "loss": 0.6615, "step": 1790 }, { "epoch": 0.04, "learning_rate": 2.9861303419304873e-05, "loss": 1.0805, "step": 1795 }, { "epoch": 0.04, "learning_rate": 2.986042336485884e-05, "loss": 1.2396, "step": 1800 }, { "epoch": 0.04, "learning_rate": 2.9859543310412807e-05, "loss": 0.6837, "step": 1805 }, { "epoch": 0.04, "learning_rate": 2.9858663255966768e-05, "loss": 0.4966, "step": 1810 }, { "epoch": 0.04, "learning_rate": 2.9857783201520733e-05, "loss": 0.43, "step": 1815 }, { "epoch": 0.04, "learning_rate": 2.98569031470747e-05, "loss": 0.5027, "step": 1820 }, { "epoch": 0.04, "learning_rate": 2.9856023092628667e-05, "loss": 0.3955, "step": 1825 }, { "epoch": 0.04, "learning_rate": 2.9855143038182628e-05, "loss": 0.773, "step": 1830 }, { "epoch": 0.04, "learning_rate": 2.9854262983736593e-05, "loss": 0.4935, "step": 1835 }, { "epoch": 0.04, "learning_rate": 2.985338292929056e-05, "loss": 0.844, "step": 1840 }, { "epoch": 0.04, "learning_rate": 2.9852502874844527e-05, "loss": 0.8925, "step": 1845 }, { "epoch": 0.04, "learning_rate": 2.985162282039849e-05, "loss": 1.2296, "step": 1850 }, { "epoch": 0.04, "learning_rate": 2.9850742765952453e-05, "loss": 0.4934, "step": 1855 }, { "epoch": 0.04, "learning_rate": 2.9849862711506422e-05, "loss": 0.4226, "step": 1860 }, { "epoch": 0.04, "learning_rate": 2.9848982657060383e-05, "loss": 0.4587, "step": 1865 }, { "epoch": 0.04, "learning_rate": 2.984810260261435e-05, "loss": 0.4462, "step": 1870 }, { "epoch": 0.04, "learning_rate": 2.9847222548168313e-05, "loss": 0.4574, "step": 1875 }, { "epoch": 0.04, "learning_rate": 2.9846342493722282e-05, "loss": 0.5674, "step": 1880 }, { "epoch": 0.04, "learning_rate": 2.9845462439276243e-05, "loss": 0.4723, "step": 1885 }, { "epoch": 0.04, "learning_rate": 2.984458238483021e-05, "loss": 0.7094, "step": 1890 }, { "epoch": 0.04, "learning_rate": 2.9843702330384173e-05, "loss": 0.8405, "step": 1895 }, { "epoch": 0.04, "learning_rate": 2.984282227593814e-05, "loss": 1.3062, "step": 1900 }, { "epoch": 0.04, "learning_rate": 2.9841942221492104e-05, "loss": 0.6546, "step": 1905 }, { "epoch": 0.04, "learning_rate": 2.984106216704607e-05, "loss": 0.4569, "step": 1910 }, { "epoch": 0.04, "learning_rate": 2.9840182112600034e-05, "loss": 0.4194, "step": 1915 }, { "epoch": 0.04, "learning_rate": 2.9839302058154e-05, "loss": 0.4278, "step": 1920 }, { "epoch": 0.04, "learning_rate": 2.9838422003707964e-05, "loss": 0.3483, "step": 1925 }, { "epoch": 0.05, "learning_rate": 2.983754194926193e-05, "loss": 0.7463, "step": 1930 }, { "epoch": 0.05, "learning_rate": 2.9836661894815894e-05, "loss": 0.6238, "step": 1935 }, { "epoch": 0.05, "learning_rate": 2.983578184036986e-05, "loss": 0.6048, "step": 1940 }, { "epoch": 0.05, "learning_rate": 2.9834901785923824e-05, "loss": 0.617, "step": 1945 }, { "epoch": 0.05, "learning_rate": 2.983402173147779e-05, "loss": 1.2095, "step": 1950 }, { "epoch": 0.05, "learning_rate": 2.9833141677031754e-05, "loss": 0.5201, "step": 1955 }, { "epoch": 0.05, "learning_rate": 2.983226162258572e-05, "loss": 0.3932, "step": 1960 }, { "epoch": 0.05, "learning_rate": 2.9831381568139684e-05, "loss": 0.4054, "step": 1965 }, { "epoch": 0.05, "learning_rate": 2.9830501513693645e-05, "loss": 0.3928, "step": 1970 }, { "epoch": 0.05, "learning_rate": 2.9829621459247614e-05, "loss": 0.7572, "step": 1975 }, { "epoch": 0.05, "learning_rate": 2.982874140480158e-05, "loss": 0.4215, "step": 1980 }, { "epoch": 0.05, "learning_rate": 2.9827861350355544e-05, "loss": 0.5974, "step": 1985 }, { "epoch": 0.05, "learning_rate": 2.9826981295909505e-05, "loss": 0.5439, "step": 1990 }, { "epoch": 0.05, "learning_rate": 2.9826101241463474e-05, "loss": 0.9534, "step": 1995 }, { "epoch": 0.05, "learning_rate": 2.982522118701744e-05, "loss": 1.1828, "step": 2000 }, { "epoch": 0.05, "learning_rate": 2.98243411325714e-05, "loss": 0.4444, "step": 2005 }, { "epoch": 0.05, "learning_rate": 2.9823461078125366e-05, "loss": 0.3975, "step": 2010 }, { "epoch": 0.05, "learning_rate": 2.9822581023679334e-05, "loss": 0.305, "step": 2015 }, { "epoch": 0.05, "learning_rate": 2.98217009692333e-05, "loss": 0.5497, "step": 2020 }, { "epoch": 0.05, "learning_rate": 2.982082091478726e-05, "loss": 0.5973, "step": 2025 }, { "epoch": 0.05, "learning_rate": 2.9819940860341226e-05, "loss": 0.5217, "step": 2030 }, { "epoch": 0.05, "learning_rate": 2.9819060805895194e-05, "loss": 0.8004, "step": 2035 }, { "epoch": 0.05, "learning_rate": 2.9818180751449156e-05, "loss": 0.5447, "step": 2040 }, { "epoch": 0.05, "learning_rate": 2.981730069700312e-05, "loss": 0.6598, "step": 2045 }, { "epoch": 0.05, "learning_rate": 2.9816420642557086e-05, "loss": 1.1138, "step": 2050 }, { "epoch": 0.05, "learning_rate": 2.9815540588111054e-05, "loss": 0.5833, "step": 2055 }, { "epoch": 0.05, "learning_rate": 2.9814660533665016e-05, "loss": 0.4151, "step": 2060 }, { "epoch": 0.05, "learning_rate": 2.981378047921898e-05, "loss": 0.4162, "step": 2065 }, { "epoch": 0.05, "learning_rate": 2.981290042477295e-05, "loss": 0.5025, "step": 2070 }, { "epoch": 0.05, "learning_rate": 2.981202037032691e-05, "loss": 0.4553, "step": 2075 }, { "epoch": 0.05, "learning_rate": 2.9811140315880876e-05, "loss": 0.6125, "step": 2080 }, { "epoch": 0.05, "learning_rate": 2.981026026143484e-05, "loss": 0.6168, "step": 2085 }, { "epoch": 0.05, "learning_rate": 2.980938020698881e-05, "loss": 0.8276, "step": 2090 }, { "epoch": 0.05, "learning_rate": 2.980850015254277e-05, "loss": 0.7357, "step": 2095 }, { "epoch": 0.05, "learning_rate": 2.9807620098096736e-05, "loss": 1.4633, "step": 2100 }, { "epoch": 0.05, "learning_rate": 2.98067400436507e-05, "loss": 0.6518, "step": 2105 }, { "epoch": 0.05, "learning_rate": 2.9805859989204666e-05, "loss": 0.4319, "step": 2110 }, { "epoch": 0.05, "learning_rate": 2.980497993475863e-05, "loss": 0.4891, "step": 2115 }, { "epoch": 0.05, "learning_rate": 2.9804099880312596e-05, "loss": 0.4411, "step": 2120 }, { "epoch": 0.05, "learning_rate": 2.980321982586656e-05, "loss": 0.5406, "step": 2125 }, { "epoch": 0.05, "learning_rate": 2.9802339771420526e-05, "loss": 0.6112, "step": 2130 }, { "epoch": 0.05, "learning_rate": 2.980145971697449e-05, "loss": 0.5593, "step": 2135 }, { "epoch": 0.05, "learning_rate": 2.9800579662528456e-05, "loss": 0.6513, "step": 2140 }, { "epoch": 0.05, "learning_rate": 2.979969960808242e-05, "loss": 1.006, "step": 2145 }, { "epoch": 0.05, "learning_rate": 2.9798819553636386e-05, "loss": 1.2303, "step": 2150 }, { "epoch": 0.05, "learning_rate": 2.979793949919035e-05, "loss": 0.6855, "step": 2155 }, { "epoch": 0.05, "learning_rate": 2.9797059444744316e-05, "loss": 0.3378, "step": 2160 }, { "epoch": 0.05, "learning_rate": 2.9796179390298278e-05, "loss": 0.4294, "step": 2165 }, { "epoch": 0.05, "learning_rate": 2.9795299335852246e-05, "loss": 0.3161, "step": 2170 }, { "epoch": 0.05, "learning_rate": 2.979441928140621e-05, "loss": 0.3886, "step": 2175 }, { "epoch": 0.05, "learning_rate": 2.9793539226960176e-05, "loss": 0.688, "step": 2180 }, { "epoch": 0.05, "learning_rate": 2.9792659172514138e-05, "loss": 0.6029, "step": 2185 }, { "epoch": 0.05, "learning_rate": 2.9791779118068106e-05, "loss": 0.5461, "step": 2190 }, { "epoch": 0.05, "learning_rate": 2.979089906362207e-05, "loss": 0.8989, "step": 2195 }, { "epoch": 0.05, "learning_rate": 2.9790019009176033e-05, "loss": 1.0839, "step": 2200 }, { "epoch": 0.05, "learning_rate": 2.978913895473e-05, "loss": 0.6046, "step": 2205 }, { "epoch": 0.05, "learning_rate": 2.9788258900283966e-05, "loss": 0.4098, "step": 2210 }, { "epoch": 0.05, "learning_rate": 2.978737884583793e-05, "loss": 0.342, "step": 2215 }, { "epoch": 0.05, "learning_rate": 2.9786498791391893e-05, "loss": 0.3958, "step": 2220 }, { "epoch": 0.05, "learning_rate": 2.978561873694586e-05, "loss": 0.5301, "step": 2225 }, { "epoch": 0.05, "learning_rate": 2.9784738682499826e-05, "loss": 0.5511, "step": 2230 }, { "epoch": 0.05, "learning_rate": 2.9783858628053788e-05, "loss": 0.6677, "step": 2235 }, { "epoch": 0.05, "learning_rate": 2.9782978573607753e-05, "loss": 0.699, "step": 2240 }, { "epoch": 0.05, "learning_rate": 2.978209851916172e-05, "loss": 0.7925, "step": 2245 }, { "epoch": 0.05, "learning_rate": 2.9781218464715687e-05, "loss": 0.8872, "step": 2250 }, { "epoch": 0.05, "learning_rate": 2.9780338410269648e-05, "loss": 0.463, "step": 2255 }, { "epoch": 0.05, "learning_rate": 2.9779458355823613e-05, "loss": 0.3687, "step": 2260 }, { "epoch": 0.05, "learning_rate": 2.977857830137758e-05, "loss": 0.4357, "step": 2265 }, { "epoch": 0.05, "learning_rate": 2.9777698246931543e-05, "loss": 0.379, "step": 2270 }, { "epoch": 0.05, "learning_rate": 2.9776818192485508e-05, "loss": 0.5929, "step": 2275 }, { "epoch": 0.05, "learning_rate": 2.9775938138039473e-05, "loss": 0.5619, "step": 2280 }, { "epoch": 0.05, "learning_rate": 2.9775058083593442e-05, "loss": 0.6034, "step": 2285 }, { "epoch": 0.05, "learning_rate": 2.9774178029147403e-05, "loss": 0.5, "step": 2290 }, { "epoch": 0.05, "learning_rate": 2.977329797470137e-05, "loss": 0.8621, "step": 2295 }, { "epoch": 0.05, "learning_rate": 2.9772417920255333e-05, "loss": 0.8457, "step": 2300 }, { "epoch": 0.05, "learning_rate": 2.97715378658093e-05, "loss": 0.5579, "step": 2305 }, { "epoch": 0.05, "learning_rate": 2.9770657811363263e-05, "loss": 0.2333, "step": 2310 }, { "epoch": 0.05, "learning_rate": 2.976977775691723e-05, "loss": 0.3564, "step": 2315 }, { "epoch": 0.05, "learning_rate": 2.9768897702471193e-05, "loss": 0.5618, "step": 2320 }, { "epoch": 0.05, "learning_rate": 2.976801764802516e-05, "loss": 0.5089, "step": 2325 }, { "epoch": 0.05, "learning_rate": 2.9767137593579124e-05, "loss": 0.5469, "step": 2330 }, { "epoch": 0.05, "learning_rate": 2.976625753913309e-05, "loss": 0.4926, "step": 2335 }, { "epoch": 0.05, "learning_rate": 2.9765377484687054e-05, "loss": 0.6009, "step": 2340 }, { "epoch": 0.05, "learning_rate": 2.976449743024102e-05, "loss": 0.8134, "step": 2345 }, { "epoch": 0.05, "learning_rate": 2.9763617375794984e-05, "loss": 1.2618, "step": 2350 }, { "epoch": 0.05, "learning_rate": 2.976273732134895e-05, "loss": 0.6957, "step": 2355 }, { "epoch": 0.06, "learning_rate": 2.9761857266902914e-05, "loss": 0.3156, "step": 2360 }, { "epoch": 0.06, "learning_rate": 2.976097721245688e-05, "loss": 0.3622, "step": 2365 }, { "epoch": 0.06, "learning_rate": 2.9760097158010844e-05, "loss": 0.3776, "step": 2370 }, { "epoch": 0.06, "learning_rate": 2.9759217103564805e-05, "loss": 0.5017, "step": 2375 }, { "epoch": 0.06, "learning_rate": 2.9758337049118774e-05, "loss": 0.7514, "step": 2380 }, { "epoch": 0.06, "learning_rate": 2.975745699467274e-05, "loss": 0.5417, "step": 2385 }, { "epoch": 0.06, "learning_rate": 2.9756576940226704e-05, "loss": 0.6751, "step": 2390 }, { "epoch": 0.06, "learning_rate": 2.9755696885780665e-05, "loss": 1.0935, "step": 2395 }, { "epoch": 0.06, "learning_rate": 2.9754816831334634e-05, "loss": 1.1083, "step": 2400 }, { "epoch": 0.06, "learning_rate": 2.97539367768886e-05, "loss": 0.6396, "step": 2405 }, { "epoch": 0.06, "learning_rate": 2.975305672244256e-05, "loss": 0.3685, "step": 2410 }, { "epoch": 0.06, "learning_rate": 2.9752176667996525e-05, "loss": 0.3012, "step": 2415 }, { "epoch": 0.06, "learning_rate": 2.9751296613550494e-05, "loss": 0.3014, "step": 2420 }, { "epoch": 0.06, "learning_rate": 2.975041655910446e-05, "loss": 0.3825, "step": 2425 }, { "epoch": 0.06, "learning_rate": 2.974953650465842e-05, "loss": 0.5322, "step": 2430 }, { "epoch": 0.06, "learning_rate": 2.9748656450212386e-05, "loss": 0.5322, "step": 2435 }, { "epoch": 0.06, "learning_rate": 2.9747776395766354e-05, "loss": 0.5181, "step": 2440 }, { "epoch": 0.06, "learning_rate": 2.974689634132032e-05, "loss": 0.8782, "step": 2445 }, { "epoch": 0.06, "learning_rate": 2.974601628687428e-05, "loss": 1.5176, "step": 2450 }, { "epoch": 0.06, "learning_rate": 2.9745136232428246e-05, "loss": 0.4092, "step": 2455 }, { "epoch": 0.06, "learning_rate": 2.9744256177982214e-05, "loss": 0.3282, "step": 2460 }, { "epoch": 0.06, "learning_rate": 2.9743376123536176e-05, "loss": 0.4243, "step": 2465 }, { "epoch": 0.06, "learning_rate": 2.974249606909014e-05, "loss": 0.5626, "step": 2470 }, { "epoch": 0.06, "learning_rate": 2.9741616014644106e-05, "loss": 0.4128, "step": 2475 }, { "epoch": 0.06, "learning_rate": 2.9740735960198074e-05, "loss": 0.5344, "step": 2480 }, { "epoch": 0.06, "learning_rate": 2.9739855905752036e-05, "loss": 0.632, "step": 2485 }, { "epoch": 0.06, "learning_rate": 2.9738975851306e-05, "loss": 0.5391, "step": 2490 }, { "epoch": 0.06, "learning_rate": 2.973809579685997e-05, "loss": 0.7894, "step": 2495 }, { "epoch": 0.06, "learning_rate": 2.973721574241393e-05, "loss": 1.1207, "step": 2500 }, { "epoch": 0.06, "learning_rate": 2.9736335687967896e-05, "loss": 0.5612, "step": 2505 }, { "epoch": 0.06, "learning_rate": 2.973545563352186e-05, "loss": 0.3742, "step": 2510 }, { "epoch": 0.06, "learning_rate": 2.973457557907583e-05, "loss": 0.4468, "step": 2515 }, { "epoch": 0.06, "learning_rate": 2.973369552462979e-05, "loss": 0.5122, "step": 2520 }, { "epoch": 0.06, "learning_rate": 2.9732815470183756e-05, "loss": 0.5016, "step": 2525 }, { "epoch": 0.06, "learning_rate": 2.973193541573772e-05, "loss": 0.5856, "step": 2530 }, { "epoch": 0.06, "learning_rate": 2.9731055361291686e-05, "loss": 0.5558, "step": 2535 }, { "epoch": 0.06, "learning_rate": 2.973017530684565e-05, "loss": 0.5193, "step": 2540 }, { "epoch": 0.06, "learning_rate": 2.9729295252399616e-05, "loss": 0.6659, "step": 2545 }, { "epoch": 0.06, "learning_rate": 2.972841519795358e-05, "loss": 0.8666, "step": 2550 }, { "epoch": 0.06, "learning_rate": 2.9727535143507546e-05, "loss": 0.5339, "step": 2555 }, { "epoch": 0.06, "learning_rate": 2.972665508906151e-05, "loss": 0.3524, "step": 2560 }, { "epoch": 0.06, "learning_rate": 2.9725775034615476e-05, "loss": 0.4822, "step": 2565 }, { "epoch": 0.06, "learning_rate": 2.9724894980169438e-05, "loss": 0.4178, "step": 2570 }, { "epoch": 0.06, "learning_rate": 2.9724014925723406e-05, "loss": 0.4046, "step": 2575 }, { "epoch": 0.06, "learning_rate": 2.972313487127737e-05, "loss": 0.4793, "step": 2580 }, { "epoch": 0.06, "learning_rate": 2.9722254816831336e-05, "loss": 0.6553, "step": 2585 }, { "epoch": 0.06, "learning_rate": 2.9721374762385298e-05, "loss": 0.6437, "step": 2590 }, { "epoch": 0.06, "learning_rate": 2.9720494707939266e-05, "loss": 0.7556, "step": 2595 }, { "epoch": 0.06, "learning_rate": 2.971961465349323e-05, "loss": 0.9871, "step": 2600 }, { "epoch": 0.06, "learning_rate": 2.9718734599047193e-05, "loss": 0.5288, "step": 2605 }, { "epoch": 0.06, "learning_rate": 2.971785454460116e-05, "loss": 0.3723, "step": 2610 }, { "epoch": 0.06, "learning_rate": 2.9716974490155126e-05, "loss": 0.2543, "step": 2615 }, { "epoch": 0.06, "learning_rate": 2.971609443570909e-05, "loss": 0.4769, "step": 2620 }, { "epoch": 0.06, "learning_rate": 2.9715214381263053e-05, "loss": 0.3653, "step": 2625 }, { "epoch": 0.06, "learning_rate": 2.971433432681702e-05, "loss": 0.4366, "step": 2630 }, { "epoch": 0.06, "learning_rate": 2.9713454272370986e-05, "loss": 0.5139, "step": 2635 }, { "epoch": 0.06, "learning_rate": 2.9712574217924948e-05, "loss": 0.7794, "step": 2640 }, { "epoch": 0.06, "learning_rate": 2.9711694163478913e-05, "loss": 0.8894, "step": 2645 }, { "epoch": 0.06, "learning_rate": 2.971081410903288e-05, "loss": 0.9835, "step": 2650 }, { "epoch": 0.06, "learning_rate": 2.9709934054586846e-05, "loss": 0.556, "step": 2655 }, { "epoch": 0.06, "learning_rate": 2.9709054000140808e-05, "loss": 0.409, "step": 2660 }, { "epoch": 0.06, "learning_rate": 2.9708173945694773e-05, "loss": 0.306, "step": 2665 }, { "epoch": 0.06, "learning_rate": 2.970729389124874e-05, "loss": 0.5587, "step": 2670 }, { "epoch": 0.06, "learning_rate": 2.9706413836802703e-05, "loss": 0.5507, "step": 2675 }, { "epoch": 0.06, "learning_rate": 2.9705533782356668e-05, "loss": 0.6737, "step": 2680 }, { "epoch": 0.06, "learning_rate": 2.9704653727910633e-05, "loss": 0.7604, "step": 2685 }, { "epoch": 0.06, "learning_rate": 2.97037736734646e-05, "loss": 0.7324, "step": 2690 }, { "epoch": 0.06, "learning_rate": 2.9702893619018563e-05, "loss": 0.6952, "step": 2695 }, { "epoch": 0.06, "learning_rate": 2.9702013564572528e-05, "loss": 1.1381, "step": 2700 }, { "epoch": 0.06, "learning_rate": 2.9701133510126493e-05, "loss": 0.7038, "step": 2705 }, { "epoch": 0.06, "learning_rate": 2.9700253455680458e-05, "loss": 0.45, "step": 2710 }, { "epoch": 0.06, "learning_rate": 2.9699373401234423e-05, "loss": 0.4237, "step": 2715 }, { "epoch": 0.06, "learning_rate": 2.969849334678839e-05, "loss": 0.4471, "step": 2720 }, { "epoch": 0.06, "learning_rate": 2.9697613292342353e-05, "loss": 0.4392, "step": 2725 }, { "epoch": 0.06, "learning_rate": 2.969673323789632e-05, "loss": 0.491, "step": 2730 }, { "epoch": 0.06, "learning_rate": 2.9695853183450283e-05, "loss": 0.5856, "step": 2735 }, { "epoch": 0.06, "learning_rate": 2.969497312900425e-05, "loss": 0.7265, "step": 2740 }, { "epoch": 0.06, "learning_rate": 2.9694093074558213e-05, "loss": 0.9252, "step": 2745 }, { "epoch": 0.06, "learning_rate": 2.969321302011218e-05, "loss": 0.9911, "step": 2750 }, { "epoch": 0.06, "learning_rate": 2.9692332965666143e-05, "loss": 0.6204, "step": 2755 }, { "epoch": 0.06, "learning_rate": 2.969145291122011e-05, "loss": 0.3906, "step": 2760 }, { "epoch": 0.06, "learning_rate": 2.9690572856774074e-05, "loss": 0.396, "step": 2765 }, { "epoch": 0.06, "learning_rate": 2.968969280232804e-05, "loss": 0.3045, "step": 2770 }, { "epoch": 0.06, "learning_rate": 2.9688812747882004e-05, "loss": 0.5304, "step": 2775 }, { "epoch": 0.06, "learning_rate": 2.968793269343597e-05, "loss": 0.5222, "step": 2780 }, { "epoch": 0.06, "learning_rate": 2.9687052638989934e-05, "loss": 0.5323, "step": 2785 }, { "epoch": 0.07, "learning_rate": 2.96861725845439e-05, "loss": 0.7246, "step": 2790 }, { "epoch": 0.07, "learning_rate": 2.9685292530097864e-05, "loss": 0.7779, "step": 2795 }, { "epoch": 0.07, "learning_rate": 2.9684412475651825e-05, "loss": 0.8061, "step": 2800 }, { "epoch": 0.07, "learning_rate": 2.9683532421205794e-05, "loss": 0.4274, "step": 2805 }, { "epoch": 0.07, "learning_rate": 2.968265236675976e-05, "loss": 0.3721, "step": 2810 }, { "epoch": 0.07, "learning_rate": 2.9681772312313724e-05, "loss": 0.5428, "step": 2815 }, { "epoch": 0.07, "learning_rate": 2.9680892257867685e-05, "loss": 0.4662, "step": 2820 }, { "epoch": 0.07, "learning_rate": 2.9680012203421654e-05, "loss": 0.5242, "step": 2825 }, { "epoch": 0.07, "learning_rate": 2.967913214897562e-05, "loss": 0.5313, "step": 2830 }, { "epoch": 0.07, "learning_rate": 2.967825209452958e-05, "loss": 0.5399, "step": 2835 }, { "epoch": 0.07, "learning_rate": 2.9677372040083545e-05, "loss": 0.6944, "step": 2840 }, { "epoch": 0.07, "learning_rate": 2.9676491985637514e-05, "loss": 0.7684, "step": 2845 }, { "epoch": 0.07, "learning_rate": 2.967561193119148e-05, "loss": 1.0986, "step": 2850 }, { "epoch": 0.07, "learning_rate": 2.967473187674544e-05, "loss": 0.638, "step": 2855 }, { "epoch": 0.07, "learning_rate": 2.9673851822299406e-05, "loss": 0.2695, "step": 2860 }, { "epoch": 0.07, "learning_rate": 2.9672971767853374e-05, "loss": 0.3742, "step": 2865 }, { "epoch": 0.07, "learning_rate": 2.9672091713407336e-05, "loss": 0.3506, "step": 2870 }, { "epoch": 0.07, "learning_rate": 2.96712116589613e-05, "loss": 0.454, "step": 2875 }, { "epoch": 0.07, "learning_rate": 2.9670331604515266e-05, "loss": 0.486, "step": 2880 }, { "epoch": 0.07, "learning_rate": 2.9669451550069234e-05, "loss": 0.5305, "step": 2885 }, { "epoch": 0.07, "learning_rate": 2.9668571495623196e-05, "loss": 0.6959, "step": 2890 }, { "epoch": 0.07, "learning_rate": 2.966769144117716e-05, "loss": 0.7673, "step": 2895 }, { "epoch": 0.07, "learning_rate": 2.966681138673113e-05, "loss": 0.7831, "step": 2900 }, { "epoch": 0.07, "learning_rate": 2.966593133228509e-05, "loss": 0.5168, "step": 2905 }, { "epoch": 0.07, "learning_rate": 2.9665051277839056e-05, "loss": 0.299, "step": 2910 }, { "epoch": 0.07, "learning_rate": 2.966417122339302e-05, "loss": 0.4243, "step": 2915 }, { "epoch": 0.07, "learning_rate": 2.966329116894699e-05, "loss": 0.3194, "step": 2920 }, { "epoch": 0.07, "learning_rate": 2.966241111450095e-05, "loss": 0.3753, "step": 2925 }, { "epoch": 0.07, "learning_rate": 2.9661531060054916e-05, "loss": 0.6637, "step": 2930 }, { "epoch": 0.07, "learning_rate": 2.966065100560888e-05, "loss": 0.5913, "step": 2935 }, { "epoch": 0.07, "learning_rate": 2.9659770951162846e-05, "loss": 0.7964, "step": 2940 }, { "epoch": 0.07, "learning_rate": 2.965889089671681e-05, "loss": 0.6223, "step": 2945 }, { "epoch": 0.07, "learning_rate": 2.9658010842270776e-05, "loss": 1.143, "step": 2950 }, { "epoch": 0.07, "learning_rate": 2.965713078782474e-05, "loss": 0.6297, "step": 2955 }, { "epoch": 0.07, "learning_rate": 2.9656250733378706e-05, "loss": 0.3298, "step": 2960 }, { "epoch": 0.07, "learning_rate": 2.965537067893267e-05, "loss": 0.3769, "step": 2965 }, { "epoch": 0.07, "learning_rate": 2.9654490624486636e-05, "loss": 0.3107, "step": 2970 }, { "epoch": 0.07, "learning_rate": 2.9653610570040598e-05, "loss": 0.3657, "step": 2975 }, { "epoch": 0.07, "learning_rate": 2.9652730515594566e-05, "loss": 0.3901, "step": 2980 }, { "epoch": 0.07, "learning_rate": 2.965185046114853e-05, "loss": 0.6873, "step": 2985 }, { "epoch": 0.07, "learning_rate": 2.9650970406702496e-05, "loss": 0.7885, "step": 2990 }, { "epoch": 0.07, "learning_rate": 2.9650090352256458e-05, "loss": 0.7709, "step": 2995 }, { "epoch": 0.07, "learning_rate": 2.9649210297810426e-05, "loss": 1.0128, "step": 3000 }, { "epoch": 0.07, "learning_rate": 2.964833024336439e-05, "loss": 0.5135, "step": 3005 }, { "epoch": 0.07, "learning_rate": 2.9647450188918353e-05, "loss": 0.3342, "step": 3010 }, { "epoch": 0.07, "learning_rate": 2.9646570134472318e-05, "loss": 0.3163, "step": 3015 }, { "epoch": 0.07, "learning_rate": 2.9645690080026286e-05, "loss": 0.5185, "step": 3020 }, { "epoch": 0.07, "learning_rate": 2.964481002558025e-05, "loss": 0.3824, "step": 3025 }, { "epoch": 0.07, "learning_rate": 2.9643929971134213e-05, "loss": 0.4786, "step": 3030 }, { "epoch": 0.07, "learning_rate": 2.964304991668818e-05, "loss": 0.4763, "step": 3035 }, { "epoch": 0.07, "learning_rate": 2.9642169862242146e-05, "loss": 0.5349, "step": 3040 }, { "epoch": 0.07, "learning_rate": 2.9641289807796108e-05, "loss": 0.572, "step": 3045 }, { "epoch": 0.07, "learning_rate": 2.9640409753350073e-05, "loss": 0.9938, "step": 3050 }, { "epoch": 0.07, "learning_rate": 2.963952969890404e-05, "loss": 0.3419, "step": 3055 }, { "epoch": 0.07, "learning_rate": 2.9638649644458006e-05, "loss": 0.239, "step": 3060 }, { "epoch": 0.07, "learning_rate": 2.9637769590011968e-05, "loss": 0.3522, "step": 3065 }, { "epoch": 0.07, "learning_rate": 2.9636889535565933e-05, "loss": 0.466, "step": 3070 }, { "epoch": 0.07, "learning_rate": 2.96360094811199e-05, "loss": 0.3938, "step": 3075 }, { "epoch": 0.07, "learning_rate": 2.9635129426673866e-05, "loss": 0.5127, "step": 3080 }, { "epoch": 0.07, "learning_rate": 2.9634249372227828e-05, "loss": 0.5348, "step": 3085 }, { "epoch": 0.07, "learning_rate": 2.9633369317781793e-05, "loss": 0.5983, "step": 3090 }, { "epoch": 0.07, "learning_rate": 2.963248926333576e-05, "loss": 0.6213, "step": 3095 }, { "epoch": 0.07, "learning_rate": 2.9631609208889723e-05, "loss": 1.1537, "step": 3100 }, { "epoch": 0.07, "learning_rate": 2.9630729154443688e-05, "loss": 0.4704, "step": 3105 }, { "epoch": 0.07, "learning_rate": 2.9629849099997653e-05, "loss": 0.3682, "step": 3110 }, { "epoch": 0.07, "learning_rate": 2.962896904555162e-05, "loss": 0.2324, "step": 3115 }, { "epoch": 0.07, "learning_rate": 2.9628088991105583e-05, "loss": 0.3322, "step": 3120 }, { "epoch": 0.07, "learning_rate": 2.9627208936659548e-05, "loss": 0.4924, "step": 3125 }, { "epoch": 0.07, "learning_rate": 2.9626328882213513e-05, "loss": 0.6047, "step": 3130 }, { "epoch": 0.07, "learning_rate": 2.9625448827767478e-05, "loss": 0.7823, "step": 3135 }, { "epoch": 0.07, "learning_rate": 2.9624568773321443e-05, "loss": 0.438, "step": 3140 }, { "epoch": 0.07, "learning_rate": 2.9623688718875408e-05, "loss": 0.6154, "step": 3145 }, { "epoch": 0.07, "learning_rate": 2.9622808664429373e-05, "loss": 1.1602, "step": 3150 }, { "epoch": 0.07, "learning_rate": 2.962192860998334e-05, "loss": 0.5138, "step": 3155 }, { "epoch": 0.07, "learning_rate": 2.9621048555537303e-05, "loss": 0.3125, "step": 3160 }, { "epoch": 0.07, "learning_rate": 2.962016850109127e-05, "loss": 0.3867, "step": 3165 }, { "epoch": 0.07, "learning_rate": 2.9619288446645233e-05, "loss": 0.3647, "step": 3170 }, { "epoch": 0.07, "learning_rate": 2.96184083921992e-05, "loss": 0.481, "step": 3175 }, { "epoch": 0.07, "learning_rate": 2.9617528337753163e-05, "loss": 0.4923, "step": 3180 }, { "epoch": 0.07, "learning_rate": 2.961664828330713e-05, "loss": 0.4016, "step": 3185 }, { "epoch": 0.07, "learning_rate": 2.9615768228861093e-05, "loss": 0.6934, "step": 3190 }, { "epoch": 0.07, "learning_rate": 2.961488817441506e-05, "loss": 0.6834, "step": 3195 }, { "epoch": 0.07, "learning_rate": 2.9614008119969024e-05, "loss": 1.0687, "step": 3200 }, { "epoch": 0.07, "learning_rate": 2.9613128065522985e-05, "loss": 0.5036, "step": 3205 }, { "epoch": 0.07, "learning_rate": 2.9612248011076954e-05, "loss": 0.3098, "step": 3210 }, { "epoch": 0.08, "learning_rate": 2.961136795663092e-05, "loss": 0.4089, "step": 3215 }, { "epoch": 0.08, "learning_rate": 2.9610487902184884e-05, "loss": 0.3025, "step": 3220 }, { "epoch": 0.08, "learning_rate": 2.9609607847738845e-05, "loss": 0.5088, "step": 3225 }, { "epoch": 0.08, "learning_rate": 2.9608727793292814e-05, "loss": 0.49, "step": 3230 }, { "epoch": 0.08, "learning_rate": 2.960784773884678e-05, "loss": 0.5418, "step": 3235 }, { "epoch": 0.08, "learning_rate": 2.960696768440074e-05, "loss": 0.5742, "step": 3240 }, { "epoch": 0.08, "learning_rate": 2.9606087629954705e-05, "loss": 0.8092, "step": 3245 }, { "epoch": 0.08, "learning_rate": 2.9605207575508674e-05, "loss": 0.7185, "step": 3250 }, { "epoch": 0.08, "learning_rate": 2.960432752106264e-05, "loss": 0.478, "step": 3255 }, { "epoch": 0.08, "learning_rate": 2.96034474666166e-05, "loss": 0.3365, "step": 3260 }, { "epoch": 0.08, "learning_rate": 2.9602567412170565e-05, "loss": 0.385, "step": 3265 }, { "epoch": 0.08, "learning_rate": 2.9601687357724534e-05, "loss": 0.4401, "step": 3270 }, { "epoch": 0.08, "learning_rate": 2.9600807303278495e-05, "loss": 0.452, "step": 3275 }, { "epoch": 0.08, "learning_rate": 2.959992724883246e-05, "loss": 0.5798, "step": 3280 }, { "epoch": 0.08, "learning_rate": 2.9599047194386425e-05, "loss": 0.4476, "step": 3285 }, { "epoch": 0.08, "learning_rate": 2.9598167139940394e-05, "loss": 0.5154, "step": 3290 }, { "epoch": 0.08, "learning_rate": 2.9597287085494356e-05, "loss": 0.6509, "step": 3295 }, { "epoch": 0.08, "learning_rate": 2.959640703104832e-05, "loss": 1.1662, "step": 3300 }, { "epoch": 0.08, "learning_rate": 2.959552697660229e-05, "loss": 0.4014, "step": 3305 }, { "epoch": 0.08, "learning_rate": 2.959464692215625e-05, "loss": 0.3318, "step": 3310 }, { "epoch": 0.08, "learning_rate": 2.9593766867710216e-05, "loss": 0.3804, "step": 3315 }, { "epoch": 0.08, "learning_rate": 2.959288681326418e-05, "loss": 0.5241, "step": 3320 }, { "epoch": 0.08, "learning_rate": 2.959200675881815e-05, "loss": 0.5188, "step": 3325 }, { "epoch": 0.08, "learning_rate": 2.959112670437211e-05, "loss": 0.5421, "step": 3330 }, { "epoch": 0.08, "learning_rate": 2.9590246649926076e-05, "loss": 0.4022, "step": 3335 }, { "epoch": 0.08, "learning_rate": 2.958936659548004e-05, "loss": 0.7901, "step": 3340 }, { "epoch": 0.08, "learning_rate": 2.9588486541034006e-05, "loss": 0.5907, "step": 3345 }, { "epoch": 0.08, "learning_rate": 2.958760648658797e-05, "loss": 0.8375, "step": 3350 }, { "epoch": 0.08, "learning_rate": 2.9586726432141936e-05, "loss": 0.5426, "step": 3355 }, { "epoch": 0.08, "learning_rate": 2.95858463776959e-05, "loss": 0.3094, "step": 3360 }, { "epoch": 0.08, "learning_rate": 2.9584966323249866e-05, "loss": 0.3296, "step": 3365 }, { "epoch": 0.08, "learning_rate": 2.958408626880383e-05, "loss": 0.4911, "step": 3370 }, { "epoch": 0.08, "learning_rate": 2.9583206214357796e-05, "loss": 0.4666, "step": 3375 }, { "epoch": 0.08, "learning_rate": 2.958232615991176e-05, "loss": 0.479, "step": 3380 }, { "epoch": 0.08, "learning_rate": 2.9581446105465726e-05, "loss": 0.3815, "step": 3385 }, { "epoch": 0.08, "learning_rate": 2.958056605101969e-05, "loss": 0.653, "step": 3390 }, { "epoch": 0.08, "learning_rate": 2.9579685996573656e-05, "loss": 0.7388, "step": 3395 }, { "epoch": 0.08, "learning_rate": 2.9578805942127618e-05, "loss": 0.7467, "step": 3400 }, { "epoch": 0.08, "learning_rate": 2.9577925887681586e-05, "loss": 0.4354, "step": 3405 }, { "epoch": 0.08, "learning_rate": 2.957704583323555e-05, "loss": 0.3183, "step": 3410 }, { "epoch": 0.08, "learning_rate": 2.9576165778789516e-05, "loss": 0.3228, "step": 3415 }, { "epoch": 0.08, "learning_rate": 2.9575285724343478e-05, "loss": 0.3785, "step": 3420 }, { "epoch": 0.08, "learning_rate": 2.9574405669897446e-05, "loss": 0.4141, "step": 3425 }, { "epoch": 0.08, "learning_rate": 2.957352561545141e-05, "loss": 0.5246, "step": 3430 }, { "epoch": 0.08, "learning_rate": 2.9572645561005373e-05, "loss": 0.437, "step": 3435 }, { "epoch": 0.08, "learning_rate": 2.957176550655934e-05, "loss": 0.4311, "step": 3440 }, { "epoch": 0.08, "learning_rate": 2.9570885452113306e-05, "loss": 0.6085, "step": 3445 }, { "epoch": 0.08, "learning_rate": 2.957000539766727e-05, "loss": 1.0113, "step": 3450 }, { "epoch": 0.08, "learning_rate": 2.9569125343221233e-05, "loss": 0.4029, "step": 3455 }, { "epoch": 0.08, "learning_rate": 2.95682452887752e-05, "loss": 0.19, "step": 3460 }, { "epoch": 0.08, "learning_rate": 2.9567365234329166e-05, "loss": 0.3862, "step": 3465 }, { "epoch": 0.08, "learning_rate": 2.9566485179883128e-05, "loss": 0.3814, "step": 3470 }, { "epoch": 0.08, "learning_rate": 2.9565605125437093e-05, "loss": 0.4112, "step": 3475 }, { "epoch": 0.08, "learning_rate": 2.956472507099106e-05, "loss": 0.2933, "step": 3480 }, { "epoch": 0.08, "learning_rate": 2.9563845016545026e-05, "loss": 0.4265, "step": 3485 }, { "epoch": 0.08, "learning_rate": 2.9562964962098988e-05, "loss": 0.5938, "step": 3490 }, { "epoch": 0.08, "learning_rate": 2.9562084907652953e-05, "loss": 0.6771, "step": 3495 }, { "epoch": 0.08, "learning_rate": 2.956120485320692e-05, "loss": 1.0986, "step": 3500 }, { "epoch": 0.08, "learning_rate": 2.9560324798760883e-05, "loss": 0.5282, "step": 3505 }, { "epoch": 0.08, "learning_rate": 2.9559444744314848e-05, "loss": 0.2969, "step": 3510 }, { "epoch": 0.08, "learning_rate": 2.9558564689868813e-05, "loss": 0.4202, "step": 3515 }, { "epoch": 0.08, "learning_rate": 2.955768463542278e-05, "loss": 0.3553, "step": 3520 }, { "epoch": 0.08, "learning_rate": 2.9556804580976743e-05, "loss": 0.3953, "step": 3525 }, { "epoch": 0.08, "learning_rate": 2.9555924526530708e-05, "loss": 0.5238, "step": 3530 }, { "epoch": 0.08, "learning_rate": 2.9555044472084673e-05, "loss": 0.4655, "step": 3535 }, { "epoch": 0.08, "learning_rate": 2.9554164417638638e-05, "loss": 0.4697, "step": 3540 }, { "epoch": 0.08, "learning_rate": 2.9553284363192603e-05, "loss": 0.8047, "step": 3545 }, { "epoch": 0.08, "learning_rate": 2.9552404308746568e-05, "loss": 1.1329, "step": 3550 }, { "epoch": 0.08, "learning_rate": 2.9551524254300533e-05, "loss": 0.6628, "step": 3555 }, { "epoch": 0.08, "learning_rate": 2.9550644199854498e-05, "loss": 0.3034, "step": 3560 }, { "epoch": 0.08, "learning_rate": 2.9549764145408463e-05, "loss": 0.239, "step": 3565 }, { "epoch": 0.08, "learning_rate": 2.9548884090962428e-05, "loss": 0.4903, "step": 3570 }, { "epoch": 0.08, "learning_rate": 2.9548004036516393e-05, "loss": 0.4691, "step": 3575 }, { "epoch": 0.08, "learning_rate": 2.954712398207036e-05, "loss": 0.4487, "step": 3580 }, { "epoch": 0.08, "learning_rate": 2.9546243927624323e-05, "loss": 0.4613, "step": 3585 }, { "epoch": 0.08, "learning_rate": 2.954536387317829e-05, "loss": 0.6772, "step": 3590 }, { "epoch": 0.08, "learning_rate": 2.9544483818732253e-05, "loss": 0.6832, "step": 3595 }, { "epoch": 0.08, "learning_rate": 2.954360376428622e-05, "loss": 0.8381, "step": 3600 }, { "epoch": 0.08, "learning_rate": 2.9542723709840183e-05, "loss": 0.463, "step": 3605 }, { "epoch": 0.08, "learning_rate": 2.9541843655394145e-05, "loss": 0.3135, "step": 3610 }, { "epoch": 0.08, "learning_rate": 2.9540963600948113e-05, "loss": 0.3269, "step": 3615 }, { "epoch": 0.08, "learning_rate": 2.954008354650208e-05, "loss": 0.3637, "step": 3620 }, { "epoch": 0.08, "learning_rate": 2.9539203492056043e-05, "loss": 0.3625, "step": 3625 }, { "epoch": 0.08, "learning_rate": 2.9538323437610005e-05, "loss": 0.4127, "step": 3630 }, { "epoch": 0.08, "learning_rate": 2.9537443383163974e-05, "loss": 0.5758, "step": 3635 }, { "epoch": 0.08, "learning_rate": 2.953656332871794e-05, "loss": 0.5984, "step": 3640 }, { "epoch": 0.09, "learning_rate": 2.95356832742719e-05, "loss": 0.6167, "step": 3645 }, { "epoch": 0.09, "learning_rate": 2.9534803219825865e-05, "loss": 0.6149, "step": 3650 }, { "epoch": 0.09, "learning_rate": 2.9533923165379834e-05, "loss": 0.3855, "step": 3655 }, { "epoch": 0.09, "learning_rate": 2.95330431109338e-05, "loss": 0.1919, "step": 3660 }, { "epoch": 0.09, "learning_rate": 2.953216305648776e-05, "loss": 0.262, "step": 3665 }, { "epoch": 0.09, "learning_rate": 2.9531283002041725e-05, "loss": 0.2918, "step": 3670 }, { "epoch": 0.09, "learning_rate": 2.9530402947595694e-05, "loss": 0.5179, "step": 3675 }, { "epoch": 0.09, "learning_rate": 2.952952289314966e-05, "loss": 0.4685, "step": 3680 }, { "epoch": 0.09, "learning_rate": 2.952864283870362e-05, "loss": 0.3901, "step": 3685 }, { "epoch": 0.09, "learning_rate": 2.9527762784257585e-05, "loss": 0.5196, "step": 3690 }, { "epoch": 0.09, "learning_rate": 2.9526882729811554e-05, "loss": 0.5151, "step": 3695 }, { "epoch": 0.09, "learning_rate": 2.9526002675365515e-05, "loss": 0.9657, "step": 3700 }, { "epoch": 0.09, "learning_rate": 2.952512262091948e-05, "loss": 0.5474, "step": 3705 }, { "epoch": 0.09, "learning_rate": 2.952424256647345e-05, "loss": 0.3932, "step": 3710 }, { "epoch": 0.09, "learning_rate": 2.9523362512027414e-05, "loss": 0.2938, "step": 3715 }, { "epoch": 0.09, "learning_rate": 2.9522482457581375e-05, "loss": 0.425, "step": 3720 }, { "epoch": 0.09, "learning_rate": 2.952160240313534e-05, "loss": 0.374, "step": 3725 }, { "epoch": 0.09, "learning_rate": 2.952072234868931e-05, "loss": 0.4245, "step": 3730 }, { "epoch": 0.09, "learning_rate": 2.951984229424327e-05, "loss": 0.4701, "step": 3735 }, { "epoch": 0.09, "learning_rate": 2.9518962239797236e-05, "loss": 0.4536, "step": 3740 }, { "epoch": 0.09, "learning_rate": 2.95180821853512e-05, "loss": 0.8685, "step": 3745 }, { "epoch": 0.09, "learning_rate": 2.951720213090517e-05, "loss": 0.8403, "step": 3750 }, { "epoch": 0.09, "learning_rate": 2.951632207645913e-05, "loss": 0.4084, "step": 3755 }, { "epoch": 0.09, "learning_rate": 2.9515442022013096e-05, "loss": 0.2596, "step": 3760 }, { "epoch": 0.09, "learning_rate": 2.951456196756706e-05, "loss": 0.3881, "step": 3765 }, { "epoch": 0.09, "learning_rate": 2.9513681913121026e-05, "loss": 0.2884, "step": 3770 }, { "epoch": 0.09, "learning_rate": 2.951280185867499e-05, "loss": 0.2596, "step": 3775 }, { "epoch": 0.09, "learning_rate": 2.9511921804228956e-05, "loss": 0.5453, "step": 3780 }, { "epoch": 0.09, "learning_rate": 2.951104174978292e-05, "loss": 0.5136, "step": 3785 }, { "epoch": 0.09, "learning_rate": 2.9510161695336886e-05, "loss": 0.5131, "step": 3790 }, { "epoch": 0.09, "learning_rate": 2.950928164089085e-05, "loss": 0.4852, "step": 3795 }, { "epoch": 0.09, "learning_rate": 2.9508401586444816e-05, "loss": 0.9346, "step": 3800 }, { "epoch": 0.09, "learning_rate": 2.9507521531998777e-05, "loss": 0.5394, "step": 3805 }, { "epoch": 0.09, "learning_rate": 2.9506641477552746e-05, "loss": 0.3379, "step": 3810 }, { "epoch": 0.09, "learning_rate": 2.950576142310671e-05, "loss": 0.3157, "step": 3815 }, { "epoch": 0.09, "learning_rate": 2.9504881368660676e-05, "loss": 0.363, "step": 3820 }, { "epoch": 0.09, "learning_rate": 2.9504001314214638e-05, "loss": 0.3921, "step": 3825 }, { "epoch": 0.09, "learning_rate": 2.9503121259768606e-05, "loss": 0.5275, "step": 3830 }, { "epoch": 0.09, "learning_rate": 2.950224120532257e-05, "loss": 0.7336, "step": 3835 }, { "epoch": 0.09, "learning_rate": 2.9501361150876533e-05, "loss": 0.7644, "step": 3840 }, { "epoch": 0.09, "learning_rate": 2.95004810964305e-05, "loss": 0.7222, "step": 3845 }, { "epoch": 0.09, "learning_rate": 2.9499601041984466e-05, "loss": 1.0424, "step": 3850 }, { "epoch": 0.09, "learning_rate": 2.949872098753843e-05, "loss": 0.4294, "step": 3855 }, { "epoch": 0.09, "learning_rate": 2.9497840933092393e-05, "loss": 0.1644, "step": 3860 }, { "epoch": 0.09, "learning_rate": 2.949696087864636e-05, "loss": 0.2214, "step": 3865 }, { "epoch": 0.09, "learning_rate": 2.9496080824200326e-05, "loss": 0.249, "step": 3870 }, { "epoch": 0.09, "learning_rate": 2.9495200769754288e-05, "loss": 0.3429, "step": 3875 }, { "epoch": 0.09, "learning_rate": 2.9494320715308253e-05, "loss": 0.359, "step": 3880 }, { "epoch": 0.09, "learning_rate": 2.949344066086222e-05, "loss": 0.5307, "step": 3885 }, { "epoch": 0.09, "learning_rate": 2.9492560606416186e-05, "loss": 0.5926, "step": 3890 }, { "epoch": 0.09, "learning_rate": 2.9491680551970148e-05, "loss": 0.492, "step": 3895 }, { "epoch": 0.09, "learning_rate": 2.9490800497524113e-05, "loss": 1.0193, "step": 3900 }, { "epoch": 0.09, "learning_rate": 2.948992044307808e-05, "loss": 0.4091, "step": 3905 }, { "epoch": 0.09, "learning_rate": 2.9489040388632043e-05, "loss": 0.1795, "step": 3910 }, { "epoch": 0.09, "learning_rate": 2.9488160334186008e-05, "loss": 0.1518, "step": 3915 }, { "epoch": 0.09, "learning_rate": 2.9487280279739973e-05, "loss": 0.2042, "step": 3920 }, { "epoch": 0.09, "learning_rate": 2.948640022529394e-05, "loss": 0.2689, "step": 3925 }, { "epoch": 0.09, "learning_rate": 2.9485520170847903e-05, "loss": 0.4815, "step": 3930 }, { "epoch": 0.09, "learning_rate": 2.9484640116401868e-05, "loss": 0.6937, "step": 3935 }, { "epoch": 0.09, "learning_rate": 2.9483760061955833e-05, "loss": 0.403, "step": 3940 }, { "epoch": 0.09, "learning_rate": 2.9482880007509798e-05, "loss": 0.5992, "step": 3945 }, { "epoch": 0.09, "learning_rate": 2.9481999953063763e-05, "loss": 0.689, "step": 3950 }, { "epoch": 0.09, "learning_rate": 2.9481119898617728e-05, "loss": 0.379, "step": 3955 }, { "epoch": 0.09, "learning_rate": 2.9480239844171693e-05, "loss": 0.3576, "step": 3960 }, { "epoch": 0.09, "learning_rate": 2.9479359789725658e-05, "loss": 0.3089, "step": 3965 }, { "epoch": 0.09, "learning_rate": 2.9478479735279623e-05, "loss": 0.2888, "step": 3970 }, { "epoch": 0.09, "learning_rate": 2.9477599680833588e-05, "loss": 0.4207, "step": 3975 }, { "epoch": 0.09, "learning_rate": 2.9476719626387557e-05, "loss": 0.3784, "step": 3980 }, { "epoch": 0.09, "learning_rate": 2.9475839571941518e-05, "loss": 0.5078, "step": 3985 }, { "epoch": 0.09, "learning_rate": 2.9474959517495483e-05, "loss": 0.5769, "step": 3990 }, { "epoch": 0.09, "learning_rate": 2.9474079463049448e-05, "loss": 0.791, "step": 3995 }, { "epoch": 0.09, "learning_rate": 2.9473199408603413e-05, "loss": 1.0531, "step": 4000 }, { "epoch": 0.09, "learning_rate": 2.9472319354157378e-05, "loss": 0.4333, "step": 4005 }, { "epoch": 0.09, "learning_rate": 2.9471439299711343e-05, "loss": 0.4067, "step": 4010 }, { "epoch": 0.09, "learning_rate": 2.947055924526531e-05, "loss": 0.2963, "step": 4015 }, { "epoch": 0.09, "learning_rate": 2.9469679190819273e-05, "loss": 0.2796, "step": 4020 }, { "epoch": 0.09, "learning_rate": 2.946879913637324e-05, "loss": 0.2707, "step": 4025 }, { "epoch": 0.09, "learning_rate": 2.9467919081927203e-05, "loss": 0.364, "step": 4030 }, { "epoch": 0.09, "learning_rate": 2.9467039027481165e-05, "loss": 0.4557, "step": 4035 }, { "epoch": 0.09, "learning_rate": 2.9466158973035133e-05, "loss": 0.4924, "step": 4040 }, { "epoch": 0.09, "learning_rate": 2.94652789185891e-05, "loss": 0.461, "step": 4045 }, { "epoch": 0.09, "learning_rate": 2.9464398864143063e-05, "loss": 1.0254, "step": 4050 }, { "epoch": 0.09, "learning_rate": 2.9463518809697025e-05, "loss": 0.498, "step": 4055 }, { "epoch": 0.09, "learning_rate": 2.9462638755250994e-05, "loss": 0.266, "step": 4060 }, { "epoch": 0.09, "learning_rate": 2.946175870080496e-05, "loss": 0.2018, "step": 4065 }, { "epoch": 0.09, "learning_rate": 2.946087864635892e-05, "loss": 0.3339, "step": 4070 }, { "epoch": 0.1, "learning_rate": 2.9459998591912885e-05, "loss": 0.3278, "step": 4075 }, { "epoch": 0.1, "learning_rate": 2.9459118537466854e-05, "loss": 0.64, "step": 4080 }, { "epoch": 0.1, "learning_rate": 2.945823848302082e-05, "loss": 0.4281, "step": 4085 }, { "epoch": 0.1, "learning_rate": 2.945735842857478e-05, "loss": 0.5146, "step": 4090 }, { "epoch": 0.1, "learning_rate": 2.9456478374128745e-05, "loss": 0.5927, "step": 4095 }, { "epoch": 0.1, "learning_rate": 2.9455598319682714e-05, "loss": 0.7302, "step": 4100 }, { "epoch": 0.1, "learning_rate": 2.9454718265236675e-05, "loss": 0.3638, "step": 4105 }, { "epoch": 0.1, "learning_rate": 2.945383821079064e-05, "loss": 0.4383, "step": 4110 }, { "epoch": 0.1, "learning_rate": 2.945295815634461e-05, "loss": 0.2999, "step": 4115 }, { "epoch": 0.1, "learning_rate": 2.9452078101898574e-05, "loss": 0.4488, "step": 4120 }, { "epoch": 0.1, "learning_rate": 2.9451198047452535e-05, "loss": 0.4672, "step": 4125 }, { "epoch": 0.1, "learning_rate": 2.94503179930065e-05, "loss": 0.4768, "step": 4130 }, { "epoch": 0.1, "learning_rate": 2.944943793856047e-05, "loss": 0.4974, "step": 4135 }, { "epoch": 0.1, "learning_rate": 2.944855788411443e-05, "loss": 0.6371, "step": 4140 }, { "epoch": 0.1, "learning_rate": 2.9447677829668395e-05, "loss": 0.5563, "step": 4145 }, { "epoch": 0.1, "learning_rate": 2.944679777522236e-05, "loss": 0.8374, "step": 4150 }, { "epoch": 0.1, "learning_rate": 2.944591772077633e-05, "loss": 0.453, "step": 4155 }, { "epoch": 0.1, "learning_rate": 2.944503766633029e-05, "loss": 0.2109, "step": 4160 }, { "epoch": 0.1, "learning_rate": 2.9444157611884256e-05, "loss": 0.4079, "step": 4165 }, { "epoch": 0.1, "learning_rate": 2.944327755743822e-05, "loss": 0.3005, "step": 4170 }, { "epoch": 0.1, "learning_rate": 2.9442397502992186e-05, "loss": 0.3284, "step": 4175 }, { "epoch": 0.1, "learning_rate": 2.944151744854615e-05, "loss": 0.4696, "step": 4180 }, { "epoch": 0.1, "learning_rate": 2.9440637394100116e-05, "loss": 0.7267, "step": 4185 }, { "epoch": 0.1, "learning_rate": 2.943975733965408e-05, "loss": 0.5309, "step": 4190 }, { "epoch": 0.1, "learning_rate": 2.9438877285208046e-05, "loss": 0.6907, "step": 4195 }, { "epoch": 0.1, "learning_rate": 2.943799723076201e-05, "loss": 0.8047, "step": 4200 }, { "epoch": 0.1, "learning_rate": 2.9437117176315976e-05, "loss": 0.4946, "step": 4205 }, { "epoch": 0.1, "learning_rate": 2.9436237121869937e-05, "loss": 0.2883, "step": 4210 }, { "epoch": 0.1, "learning_rate": 2.9435357067423906e-05, "loss": 0.3926, "step": 4215 }, { "epoch": 0.1, "learning_rate": 2.943447701297787e-05, "loss": 0.3214, "step": 4220 }, { "epoch": 0.1, "learning_rate": 2.9433596958531836e-05, "loss": 0.28, "step": 4225 }, { "epoch": 0.1, "learning_rate": 2.9432716904085797e-05, "loss": 0.4675, "step": 4230 }, { "epoch": 0.1, "learning_rate": 2.9431836849639766e-05, "loss": 0.4672, "step": 4235 }, { "epoch": 0.1, "learning_rate": 2.943095679519373e-05, "loss": 0.4835, "step": 4240 }, { "epoch": 0.1, "learning_rate": 2.9430076740747692e-05, "loss": 0.5337, "step": 4245 }, { "epoch": 0.1, "learning_rate": 2.942919668630166e-05, "loss": 1.0831, "step": 4250 }, { "epoch": 0.1, "learning_rate": 2.9428316631855626e-05, "loss": 0.3382, "step": 4255 }, { "epoch": 0.1, "learning_rate": 2.942743657740959e-05, "loss": 0.1831, "step": 4260 }, { "epoch": 0.1, "learning_rate": 2.9426556522963553e-05, "loss": 0.2122, "step": 4265 }, { "epoch": 0.1, "learning_rate": 2.942567646851752e-05, "loss": 0.3262, "step": 4270 }, { "epoch": 0.1, "learning_rate": 2.9424796414071486e-05, "loss": 0.325, "step": 4275 }, { "epoch": 0.1, "learning_rate": 2.942391635962545e-05, "loss": 0.4603, "step": 4280 }, { "epoch": 0.1, "learning_rate": 2.9423036305179413e-05, "loss": 0.3305, "step": 4285 }, { "epoch": 0.1, "learning_rate": 2.942215625073338e-05, "loss": 0.5551, "step": 4290 }, { "epoch": 0.1, "learning_rate": 2.9421276196287346e-05, "loss": 0.8088, "step": 4295 }, { "epoch": 0.1, "learning_rate": 2.9420396141841308e-05, "loss": 0.7793, "step": 4300 }, { "epoch": 0.1, "learning_rate": 2.9419516087395273e-05, "loss": 0.3334, "step": 4305 }, { "epoch": 0.1, "learning_rate": 2.941863603294924e-05, "loss": 0.3209, "step": 4310 }, { "epoch": 0.1, "learning_rate": 2.9417755978503206e-05, "loss": 0.281, "step": 4315 }, { "epoch": 0.1, "learning_rate": 2.9416875924057168e-05, "loss": 0.2711, "step": 4320 }, { "epoch": 0.1, "learning_rate": 2.9415995869611133e-05, "loss": 0.4046, "step": 4325 }, { "epoch": 0.1, "learning_rate": 2.94151158151651e-05, "loss": 0.3689, "step": 4330 }, { "epoch": 0.1, "learning_rate": 2.9414235760719063e-05, "loss": 0.8847, "step": 4335 }, { "epoch": 0.1, "learning_rate": 2.9413355706273028e-05, "loss": 0.5625, "step": 4340 }, { "epoch": 0.1, "learning_rate": 2.9412475651826993e-05, "loss": 0.7816, "step": 4345 }, { "epoch": 0.1, "learning_rate": 2.941159559738096e-05, "loss": 0.7614, "step": 4350 }, { "epoch": 0.1, "learning_rate": 2.9410715542934923e-05, "loss": 0.4242, "step": 4355 }, { "epoch": 0.1, "learning_rate": 2.9409835488488888e-05, "loss": 0.2437, "step": 4360 }, { "epoch": 0.1, "learning_rate": 2.9408955434042853e-05, "loss": 0.2719, "step": 4365 }, { "epoch": 0.1, "learning_rate": 2.9408075379596818e-05, "loss": 0.3212, "step": 4370 }, { "epoch": 0.1, "learning_rate": 2.9407195325150783e-05, "loss": 0.3234, "step": 4375 }, { "epoch": 0.1, "learning_rate": 2.9406315270704748e-05, "loss": 0.4546, "step": 4380 }, { "epoch": 0.1, "learning_rate": 2.9405435216258716e-05, "loss": 0.4895, "step": 4385 }, { "epoch": 0.1, "learning_rate": 2.9404555161812678e-05, "loss": 0.7262, "step": 4390 }, { "epoch": 0.1, "learning_rate": 2.9403675107366643e-05, "loss": 0.6423, "step": 4395 }, { "epoch": 0.1, "learning_rate": 2.9402795052920608e-05, "loss": 1.191, "step": 4400 }, { "epoch": 0.1, "learning_rate": 2.9401914998474573e-05, "loss": 0.5131, "step": 4405 }, { "epoch": 0.1, "learning_rate": 2.9401034944028538e-05, "loss": 0.2867, "step": 4410 }, { "epoch": 0.1, "learning_rate": 2.9400154889582503e-05, "loss": 0.4176, "step": 4415 }, { "epoch": 0.1, "learning_rate": 2.9399274835136468e-05, "loss": 0.4367, "step": 4420 }, { "epoch": 0.1, "learning_rate": 2.9398394780690433e-05, "loss": 0.4725, "step": 4425 }, { "epoch": 0.1, "learning_rate": 2.9397514726244398e-05, "loss": 0.4101, "step": 4430 }, { "epoch": 0.1, "learning_rate": 2.9396634671798363e-05, "loss": 0.5488, "step": 4435 }, { "epoch": 0.1, "learning_rate": 2.9395754617352325e-05, "loss": 0.6391, "step": 4440 }, { "epoch": 0.1, "learning_rate": 2.9394874562906293e-05, "loss": 0.6833, "step": 4445 }, { "epoch": 0.1, "learning_rate": 2.939399450846026e-05, "loss": 0.7072, "step": 4450 }, { "epoch": 0.1, "learning_rate": 2.9393114454014223e-05, "loss": 0.3504, "step": 4455 }, { "epoch": 0.1, "learning_rate": 2.9392234399568185e-05, "loss": 0.3309, "step": 4460 }, { "epoch": 0.1, "learning_rate": 2.9391354345122153e-05, "loss": 0.2515, "step": 4465 }, { "epoch": 0.1, "learning_rate": 2.939047429067612e-05, "loss": 0.3689, "step": 4470 }, { "epoch": 0.1, "learning_rate": 2.938959423623008e-05, "loss": 0.4175, "step": 4475 }, { "epoch": 0.1, "learning_rate": 2.9388714181784045e-05, "loss": 0.5701, "step": 4480 }, { "epoch": 0.1, "learning_rate": 2.9387834127338013e-05, "loss": 0.4377, "step": 4485 }, { "epoch": 0.1, "learning_rate": 2.938695407289198e-05, "loss": 0.4365, "step": 4490 }, { "epoch": 0.1, "learning_rate": 2.938607401844594e-05, "loss": 0.7388, "step": 4495 }, { "epoch": 0.1, "learning_rate": 2.9385193963999905e-05, "loss": 0.9822, "step": 4500 }, { "epoch": 0.11, "learning_rate": 2.9384313909553874e-05, "loss": 0.3843, "step": 4505 }, { "epoch": 0.11, "learning_rate": 2.9383433855107835e-05, "loss": 0.2008, "step": 4510 }, { "epoch": 0.11, "learning_rate": 2.93825538006618e-05, "loss": 0.3024, "step": 4515 }, { "epoch": 0.11, "learning_rate": 2.938167374621577e-05, "loss": 0.3329, "step": 4520 }, { "epoch": 0.11, "learning_rate": 2.9380793691769734e-05, "loss": 0.3492, "step": 4525 }, { "epoch": 0.11, "learning_rate": 2.9379913637323695e-05, "loss": 0.3295, "step": 4530 }, { "epoch": 0.11, "learning_rate": 2.937903358287766e-05, "loss": 0.4998, "step": 4535 }, { "epoch": 0.11, "learning_rate": 2.937815352843163e-05, "loss": 0.4754, "step": 4540 }, { "epoch": 0.11, "learning_rate": 2.937727347398559e-05, "loss": 0.7311, "step": 4545 }, { "epoch": 0.11, "learning_rate": 2.9376393419539555e-05, "loss": 0.7343, "step": 4550 }, { "epoch": 0.11, "learning_rate": 2.937551336509352e-05, "loss": 0.4433, "step": 4555 }, { "epoch": 0.11, "learning_rate": 2.937463331064749e-05, "loss": 0.1979, "step": 4560 }, { "epoch": 0.11, "learning_rate": 2.937375325620145e-05, "loss": 0.2406, "step": 4565 }, { "epoch": 0.11, "learning_rate": 2.9372873201755415e-05, "loss": 0.2519, "step": 4570 }, { "epoch": 0.11, "learning_rate": 2.937199314730938e-05, "loss": 0.3183, "step": 4575 }, { "epoch": 0.11, "learning_rate": 2.9371113092863345e-05, "loss": 0.4703, "step": 4580 }, { "epoch": 0.11, "learning_rate": 2.937023303841731e-05, "loss": 0.3589, "step": 4585 }, { "epoch": 0.11, "learning_rate": 2.9369352983971276e-05, "loss": 0.4538, "step": 4590 }, { "epoch": 0.11, "learning_rate": 2.936847292952524e-05, "loss": 0.5111, "step": 4595 }, { "epoch": 0.11, "learning_rate": 2.9367592875079206e-05, "loss": 1.0645, "step": 4600 }, { "epoch": 0.11, "learning_rate": 2.936671282063317e-05, "loss": 0.3546, "step": 4605 }, { "epoch": 0.11, "learning_rate": 2.9365832766187136e-05, "loss": 0.2622, "step": 4610 }, { "epoch": 0.11, "learning_rate": 2.93649527117411e-05, "loss": 0.2935, "step": 4615 }, { "epoch": 0.11, "learning_rate": 2.9364072657295066e-05, "loss": 0.3134, "step": 4620 }, { "epoch": 0.11, "learning_rate": 2.936319260284903e-05, "loss": 0.3494, "step": 4625 }, { "epoch": 0.11, "learning_rate": 2.9362312548402996e-05, "loss": 0.4426, "step": 4630 }, { "epoch": 0.11, "learning_rate": 2.9361432493956957e-05, "loss": 0.4751, "step": 4635 }, { "epoch": 0.11, "learning_rate": 2.9360552439510926e-05, "loss": 0.504, "step": 4640 }, { "epoch": 0.11, "learning_rate": 2.935967238506489e-05, "loss": 0.6584, "step": 4645 }, { "epoch": 0.11, "learning_rate": 2.9358792330618856e-05, "loss": 0.8445, "step": 4650 }, { "epoch": 0.11, "learning_rate": 2.935791227617282e-05, "loss": 0.4142, "step": 4655 }, { "epoch": 0.11, "learning_rate": 2.9357032221726786e-05, "loss": 0.2061, "step": 4660 }, { "epoch": 0.11, "learning_rate": 2.935615216728075e-05, "loss": 0.2613, "step": 4665 }, { "epoch": 0.11, "learning_rate": 2.9355272112834712e-05, "loss": 0.3847, "step": 4670 }, { "epoch": 0.11, "learning_rate": 2.935439205838868e-05, "loss": 0.3025, "step": 4675 }, { "epoch": 0.11, "learning_rate": 2.9353512003942646e-05, "loss": 0.3876, "step": 4680 }, { "epoch": 0.11, "learning_rate": 2.935263194949661e-05, "loss": 0.3715, "step": 4685 }, { "epoch": 0.11, "learning_rate": 2.9351751895050573e-05, "loss": 0.5722, "step": 4690 }, { "epoch": 0.11, "learning_rate": 2.935087184060454e-05, "loss": 0.5848, "step": 4695 }, { "epoch": 0.11, "learning_rate": 2.9349991786158506e-05, "loss": 0.9042, "step": 4700 }, { "epoch": 0.11, "learning_rate": 2.9349111731712468e-05, "loss": 0.5485, "step": 4705 }, { "epoch": 0.11, "learning_rate": 2.9348231677266433e-05, "loss": 0.1736, "step": 4710 }, { "epoch": 0.11, "learning_rate": 2.93473516228204e-05, "loss": 0.3293, "step": 4715 }, { "epoch": 0.11, "learning_rate": 2.9346471568374366e-05, "loss": 0.4446, "step": 4720 }, { "epoch": 0.11, "learning_rate": 2.9345591513928328e-05, "loss": 0.4149, "step": 4725 }, { "epoch": 0.11, "learning_rate": 2.9344711459482293e-05, "loss": 0.3636, "step": 4730 }, { "epoch": 0.11, "learning_rate": 2.934383140503626e-05, "loss": 0.446, "step": 4735 }, { "epoch": 0.11, "learning_rate": 2.9342951350590223e-05, "loss": 0.3909, "step": 4740 }, { "epoch": 0.11, "learning_rate": 2.9342071296144188e-05, "loss": 0.5079, "step": 4745 }, { "epoch": 0.11, "learning_rate": 2.9341191241698153e-05, "loss": 0.8475, "step": 4750 }, { "epoch": 0.11, "learning_rate": 2.934031118725212e-05, "loss": 0.5314, "step": 4755 }, { "epoch": 0.11, "learning_rate": 2.9339431132806083e-05, "loss": 0.2108, "step": 4760 }, { "epoch": 0.11, "learning_rate": 2.9338551078360048e-05, "loss": 0.1875, "step": 4765 }, { "epoch": 0.11, "learning_rate": 2.9337671023914013e-05, "loss": 0.2498, "step": 4770 }, { "epoch": 0.11, "learning_rate": 2.9336790969467978e-05, "loss": 0.4319, "step": 4775 }, { "epoch": 0.11, "learning_rate": 2.9335910915021943e-05, "loss": 0.5195, "step": 4780 }, { "epoch": 0.11, "learning_rate": 2.9335030860575908e-05, "loss": 0.5469, "step": 4785 }, { "epoch": 0.11, "learning_rate": 2.9334150806129876e-05, "loss": 0.5491, "step": 4790 }, { "epoch": 0.11, "learning_rate": 2.9333270751683838e-05, "loss": 0.6427, "step": 4795 }, { "epoch": 0.11, "learning_rate": 2.9332390697237803e-05, "loss": 0.9694, "step": 4800 }, { "epoch": 0.11, "learning_rate": 2.9331510642791768e-05, "loss": 0.3621, "step": 4805 }, { "epoch": 0.11, "learning_rate": 2.9330630588345733e-05, "loss": 0.1766, "step": 4810 }, { "epoch": 0.11, "learning_rate": 2.9329750533899698e-05, "loss": 0.1434, "step": 4815 }, { "epoch": 0.11, "learning_rate": 2.9328870479453663e-05, "loss": 0.2951, "step": 4820 }, { "epoch": 0.11, "learning_rate": 2.9327990425007628e-05, "loss": 0.3314, "step": 4825 }, { "epoch": 0.11, "learning_rate": 2.9327110370561593e-05, "loss": 0.3358, "step": 4830 }, { "epoch": 0.11, "learning_rate": 2.9326230316115558e-05, "loss": 0.5118, "step": 4835 }, { "epoch": 0.11, "learning_rate": 2.9325350261669523e-05, "loss": 0.5368, "step": 4840 }, { "epoch": 0.11, "learning_rate": 2.9324470207223485e-05, "loss": 0.5611, "step": 4845 }, { "epoch": 0.11, "learning_rate": 2.9323590152777453e-05, "loss": 0.928, "step": 4850 }, { "epoch": 0.11, "learning_rate": 2.9322710098331418e-05, "loss": 0.4803, "step": 4855 }, { "epoch": 0.11, "learning_rate": 2.9321830043885383e-05, "loss": 0.1836, "step": 4860 }, { "epoch": 0.11, "learning_rate": 2.9320949989439345e-05, "loss": 0.3508, "step": 4865 }, { "epoch": 0.11, "learning_rate": 2.9320069934993313e-05, "loss": 0.2023, "step": 4870 }, { "epoch": 0.11, "learning_rate": 2.9319189880547278e-05, "loss": 0.4517, "step": 4875 }, { "epoch": 0.11, "learning_rate": 2.931830982610124e-05, "loss": 0.4725, "step": 4880 }, { "epoch": 0.11, "learning_rate": 2.9317429771655205e-05, "loss": 0.4528, "step": 4885 }, { "epoch": 0.11, "learning_rate": 2.9316549717209173e-05, "loss": 0.4455, "step": 4890 }, { "epoch": 0.11, "learning_rate": 2.931566966276314e-05, "loss": 0.6456, "step": 4895 }, { "epoch": 0.11, "learning_rate": 2.93147896083171e-05, "loss": 0.9289, "step": 4900 }, { "epoch": 0.11, "learning_rate": 2.9313909553871065e-05, "loss": 0.4893, "step": 4905 }, { "epoch": 0.11, "learning_rate": 2.9313029499425033e-05, "loss": 0.2805, "step": 4910 }, { "epoch": 0.11, "learning_rate": 2.9312149444979e-05, "loss": 0.2942, "step": 4915 }, { "epoch": 0.11, "learning_rate": 2.931126939053296e-05, "loss": 0.2671, "step": 4920 }, { "epoch": 0.11, "learning_rate": 2.931038933608693e-05, "loss": 0.2367, "step": 4925 }, { "epoch": 0.12, "learning_rate": 2.9309509281640894e-05, "loss": 0.3116, "step": 4930 }, { "epoch": 0.12, "learning_rate": 2.9308629227194855e-05, "loss": 0.3839, "step": 4935 }, { "epoch": 0.12, "learning_rate": 2.930774917274882e-05, "loss": 0.5472, "step": 4940 }, { "epoch": 0.12, "learning_rate": 2.930686911830279e-05, "loss": 0.6326, "step": 4945 }, { "epoch": 0.12, "learning_rate": 2.9305989063856754e-05, "loss": 0.843, "step": 4950 }, { "epoch": 0.12, "learning_rate": 2.9305109009410715e-05, "loss": 0.3709, "step": 4955 }, { "epoch": 0.12, "learning_rate": 2.930422895496468e-05, "loss": 0.2225, "step": 4960 }, { "epoch": 0.12, "learning_rate": 2.930334890051865e-05, "loss": 0.1332, "step": 4965 }, { "epoch": 0.12, "learning_rate": 2.930246884607261e-05, "loss": 0.2667, "step": 4970 }, { "epoch": 0.12, "learning_rate": 2.9301588791626575e-05, "loss": 0.4758, "step": 4975 }, { "epoch": 0.12, "learning_rate": 2.930070873718054e-05, "loss": 0.4617, "step": 4980 }, { "epoch": 0.12, "learning_rate": 2.929982868273451e-05, "loss": 0.6049, "step": 4985 }, { "epoch": 0.12, "learning_rate": 2.929894862828847e-05, "loss": 0.4674, "step": 4990 }, { "epoch": 0.12, "learning_rate": 2.9298068573842435e-05, "loss": 0.5146, "step": 4995 }, { "epoch": 0.12, "learning_rate": 2.92971885193964e-05, "loss": 0.9358, "step": 5000 }, { "epoch": 0.12, "learning_rate": 2.9296308464950365e-05, "loss": 0.4057, "step": 5005 }, { "epoch": 0.12, "learning_rate": 2.929542841050433e-05, "loss": 0.2197, "step": 5010 }, { "epoch": 0.12, "learning_rate": 2.9294548356058295e-05, "loss": 0.2378, "step": 5015 }, { "epoch": 0.12, "learning_rate": 2.929366830161226e-05, "loss": 0.31, "step": 5020 }, { "epoch": 0.12, "learning_rate": 2.9292788247166226e-05, "loss": 0.3598, "step": 5025 }, { "epoch": 0.12, "learning_rate": 2.929190819272019e-05, "loss": 0.416, "step": 5030 }, { "epoch": 0.12, "learning_rate": 2.9291028138274156e-05, "loss": 0.6173, "step": 5035 }, { "epoch": 0.12, "learning_rate": 2.9290148083828117e-05, "loss": 0.5344, "step": 5040 }, { "epoch": 0.12, "learning_rate": 2.9289268029382086e-05, "loss": 0.4437, "step": 5045 }, { "epoch": 0.12, "learning_rate": 2.928838797493605e-05, "loss": 0.8571, "step": 5050 }, { "epoch": 0.12, "learning_rate": 2.9287507920490016e-05, "loss": 0.2738, "step": 5055 }, { "epoch": 0.12, "learning_rate": 2.928662786604398e-05, "loss": 0.298, "step": 5060 }, { "epoch": 0.12, "learning_rate": 2.9285747811597946e-05, "loss": 0.258, "step": 5065 }, { "epoch": 0.12, "learning_rate": 2.928486775715191e-05, "loss": 0.2896, "step": 5070 }, { "epoch": 0.12, "learning_rate": 2.9283987702705872e-05, "loss": 0.3141, "step": 5075 }, { "epoch": 0.12, "learning_rate": 2.928310764825984e-05, "loss": 0.6347, "step": 5080 }, { "epoch": 0.12, "learning_rate": 2.9282227593813806e-05, "loss": 0.4315, "step": 5085 }, { "epoch": 0.12, "learning_rate": 2.928134753936777e-05, "loss": 0.7037, "step": 5090 }, { "epoch": 0.12, "learning_rate": 2.9280467484921732e-05, "loss": 0.8146, "step": 5095 }, { "epoch": 0.12, "learning_rate": 2.92795874304757e-05, "loss": 0.9553, "step": 5100 }, { "epoch": 0.12, "learning_rate": 2.9278707376029666e-05, "loss": 0.3104, "step": 5105 }, { "epoch": 0.12, "learning_rate": 2.9277827321583627e-05, "loss": 0.3223, "step": 5110 }, { "epoch": 0.12, "learning_rate": 2.9276947267137593e-05, "loss": 0.3399, "step": 5115 }, { "epoch": 0.12, "learning_rate": 2.927606721269156e-05, "loss": 0.3746, "step": 5120 }, { "epoch": 0.12, "learning_rate": 2.9275187158245526e-05, "loss": 0.5267, "step": 5125 }, { "epoch": 0.12, "learning_rate": 2.9274307103799488e-05, "loss": 0.5439, "step": 5130 }, { "epoch": 0.12, "learning_rate": 2.9273427049353453e-05, "loss": 0.5503, "step": 5135 }, { "epoch": 0.12, "learning_rate": 2.927254699490742e-05, "loss": 0.3995, "step": 5140 }, { "epoch": 0.12, "learning_rate": 2.9271666940461383e-05, "loss": 0.6346, "step": 5145 }, { "epoch": 0.12, "learning_rate": 2.9270786886015348e-05, "loss": 0.9787, "step": 5150 }, { "epoch": 0.12, "learning_rate": 2.9269906831569313e-05, "loss": 0.494, "step": 5155 }, { "epoch": 0.12, "learning_rate": 2.926902677712328e-05, "loss": 0.2516, "step": 5160 }, { "epoch": 0.12, "learning_rate": 2.9268146722677243e-05, "loss": 0.3142, "step": 5165 }, { "epoch": 0.12, "learning_rate": 2.9267266668231208e-05, "loss": 0.3402, "step": 5170 }, { "epoch": 0.12, "learning_rate": 2.9266386613785173e-05, "loss": 0.4781, "step": 5175 }, { "epoch": 0.12, "learning_rate": 2.9265506559339138e-05, "loss": 0.2854, "step": 5180 }, { "epoch": 0.12, "learning_rate": 2.9264626504893103e-05, "loss": 0.4102, "step": 5185 }, { "epoch": 0.12, "learning_rate": 2.9263746450447068e-05, "loss": 0.4626, "step": 5190 }, { "epoch": 0.12, "learning_rate": 2.9262866396001033e-05, "loss": 0.4936, "step": 5195 }, { "epoch": 0.12, "learning_rate": 2.9261986341554998e-05, "loss": 0.8179, "step": 5200 }, { "epoch": 0.12, "learning_rate": 2.9261106287108963e-05, "loss": 0.524, "step": 5205 }, { "epoch": 0.12, "learning_rate": 2.9260226232662928e-05, "loss": 0.1892, "step": 5210 }, { "epoch": 0.12, "learning_rate": 2.9259346178216896e-05, "loss": 0.2768, "step": 5215 }, { "epoch": 0.12, "learning_rate": 2.9258466123770858e-05, "loss": 0.214, "step": 5220 }, { "epoch": 0.12, "learning_rate": 2.9257586069324823e-05, "loss": 0.3607, "step": 5225 }, { "epoch": 0.12, "learning_rate": 2.9256706014878788e-05, "loss": 0.2319, "step": 5230 }, { "epoch": 0.12, "learning_rate": 2.9255825960432753e-05, "loss": 0.4049, "step": 5235 }, { "epoch": 0.12, "learning_rate": 2.9254945905986718e-05, "loss": 0.4589, "step": 5240 }, { "epoch": 0.12, "learning_rate": 2.9254065851540683e-05, "loss": 0.53, "step": 5245 }, { "epoch": 0.12, "learning_rate": 2.9253185797094648e-05, "loss": 0.7657, "step": 5250 }, { "epoch": 0.12, "learning_rate": 2.9252305742648613e-05, "loss": 0.3048, "step": 5255 }, { "epoch": 0.12, "learning_rate": 2.9251425688202578e-05, "loss": 0.2334, "step": 5260 }, { "epoch": 0.12, "learning_rate": 2.9250545633756543e-05, "loss": 0.391, "step": 5265 }, { "epoch": 0.12, "learning_rate": 2.9249665579310505e-05, "loss": 0.2999, "step": 5270 }, { "epoch": 0.12, "learning_rate": 2.9248785524864473e-05, "loss": 0.3477, "step": 5275 }, { "epoch": 0.12, "learning_rate": 2.9247905470418438e-05, "loss": 0.3967, "step": 5280 }, { "epoch": 0.12, "learning_rate": 2.9247025415972403e-05, "loss": 0.5547, "step": 5285 }, { "epoch": 0.12, "learning_rate": 2.9246145361526365e-05, "loss": 0.5384, "step": 5290 }, { "epoch": 0.12, "learning_rate": 2.9245265307080333e-05, "loss": 0.3743, "step": 5295 }, { "epoch": 0.12, "learning_rate": 2.9244385252634298e-05, "loss": 1.2073, "step": 5300 }, { "epoch": 0.12, "learning_rate": 2.924350519818826e-05, "loss": 0.4523, "step": 5305 }, { "epoch": 0.12, "learning_rate": 2.9242625143742225e-05, "loss": 0.2289, "step": 5310 }, { "epoch": 0.12, "learning_rate": 2.9241745089296193e-05, "loss": 0.2903, "step": 5315 }, { "epoch": 0.12, "learning_rate": 2.924086503485016e-05, "loss": 0.3922, "step": 5320 }, { "epoch": 0.12, "learning_rate": 2.923998498040412e-05, "loss": 0.4738, "step": 5325 }, { "epoch": 0.12, "learning_rate": 2.923910492595809e-05, "loss": 0.379, "step": 5330 }, { "epoch": 0.12, "learning_rate": 2.9238224871512053e-05, "loss": 0.4108, "step": 5335 }, { "epoch": 0.12, "learning_rate": 2.9237344817066015e-05, "loss": 0.3328, "step": 5340 }, { "epoch": 0.12, "learning_rate": 2.923646476261998e-05, "loss": 0.5297, "step": 5345 }, { "epoch": 0.12, "learning_rate": 2.923558470817395e-05, "loss": 0.7928, "step": 5350 }, { "epoch": 0.12, "learning_rate": 2.9234704653727913e-05, "loss": 0.3439, "step": 5355 }, { "epoch": 0.13, "learning_rate": 2.9233824599281875e-05, "loss": 0.2224, "step": 5360 }, { "epoch": 0.13, "learning_rate": 2.923294454483584e-05, "loss": 0.2071, "step": 5365 }, { "epoch": 0.13, "learning_rate": 2.923206449038981e-05, "loss": 0.38, "step": 5370 }, { "epoch": 0.13, "learning_rate": 2.923118443594377e-05, "loss": 0.4246, "step": 5375 }, { "epoch": 0.13, "learning_rate": 2.9230304381497735e-05, "loss": 0.4473, "step": 5380 }, { "epoch": 0.13, "learning_rate": 2.92294243270517e-05, "loss": 0.4833, "step": 5385 }, { "epoch": 0.13, "learning_rate": 2.922854427260567e-05, "loss": 0.5431, "step": 5390 }, { "epoch": 0.13, "learning_rate": 2.922766421815963e-05, "loss": 0.7253, "step": 5395 }, { "epoch": 0.13, "learning_rate": 2.9226784163713595e-05, "loss": 0.7148, "step": 5400 }, { "epoch": 0.13, "learning_rate": 2.922590410926756e-05, "loss": 0.3863, "step": 5405 }, { "epoch": 0.13, "learning_rate": 2.9225024054821525e-05, "loss": 0.165, "step": 5410 }, { "epoch": 0.13, "learning_rate": 2.922414400037549e-05, "loss": 0.264, "step": 5415 }, { "epoch": 0.13, "learning_rate": 2.9223263945929455e-05, "loss": 0.1824, "step": 5420 }, { "epoch": 0.13, "learning_rate": 2.922238389148342e-05, "loss": 0.4501, "step": 5425 }, { "epoch": 0.13, "learning_rate": 2.9221503837037385e-05, "loss": 0.4079, "step": 5430 }, { "epoch": 0.13, "learning_rate": 2.922062378259135e-05, "loss": 0.4551, "step": 5435 }, { "epoch": 0.13, "learning_rate": 2.9219743728145315e-05, "loss": 0.4273, "step": 5440 }, { "epoch": 0.13, "learning_rate": 2.9218863673699277e-05, "loss": 0.6627, "step": 5445 }, { "epoch": 0.13, "learning_rate": 2.9217983619253245e-05, "loss": 0.822, "step": 5450 }, { "epoch": 0.13, "learning_rate": 2.921710356480721e-05, "loss": 0.4206, "step": 5455 }, { "epoch": 0.13, "learning_rate": 2.9216223510361176e-05, "loss": 0.3612, "step": 5460 }, { "epoch": 0.13, "learning_rate": 2.921534345591514e-05, "loss": 0.5354, "step": 5465 }, { "epoch": 0.13, "learning_rate": 2.9214463401469106e-05, "loss": 0.2471, "step": 5470 }, { "epoch": 0.13, "learning_rate": 2.921358334702307e-05, "loss": 0.4321, "step": 5475 }, { "epoch": 0.13, "learning_rate": 2.9212703292577032e-05, "loss": 0.4164, "step": 5480 }, { "epoch": 0.13, "learning_rate": 2.9211823238131e-05, "loss": 0.5597, "step": 5485 }, { "epoch": 0.13, "learning_rate": 2.9210943183684966e-05, "loss": 0.5078, "step": 5490 }, { "epoch": 0.13, "learning_rate": 2.921006312923893e-05, "loss": 0.6254, "step": 5495 }, { "epoch": 0.13, "learning_rate": 2.9209183074792892e-05, "loss": 1.1515, "step": 5500 }, { "epoch": 0.13, "learning_rate": 2.920830302034686e-05, "loss": 0.3524, "step": 5505 }, { "epoch": 0.13, "learning_rate": 2.9207422965900826e-05, "loss": 0.193, "step": 5510 }, { "epoch": 0.13, "learning_rate": 2.920654291145479e-05, "loss": 0.3121, "step": 5515 }, { "epoch": 0.13, "learning_rate": 2.9205662857008752e-05, "loss": 0.2809, "step": 5520 }, { "epoch": 0.13, "learning_rate": 2.920478280256272e-05, "loss": 0.4493, "step": 5525 }, { "epoch": 0.13, "learning_rate": 2.9203902748116686e-05, "loss": 0.3105, "step": 5530 }, { "epoch": 0.13, "learning_rate": 2.9203022693670647e-05, "loss": 0.5638, "step": 5535 }, { "epoch": 0.13, "learning_rate": 2.9202142639224612e-05, "loss": 0.6576, "step": 5540 }, { "epoch": 0.13, "learning_rate": 2.920126258477858e-05, "loss": 0.648, "step": 5545 }, { "epoch": 0.13, "learning_rate": 2.9200382530332546e-05, "loss": 0.6859, "step": 5550 }, { "epoch": 0.13, "learning_rate": 2.9199502475886508e-05, "loss": 0.4888, "step": 5555 }, { "epoch": 0.13, "learning_rate": 2.9198622421440473e-05, "loss": 0.2035, "step": 5560 }, { "epoch": 0.13, "learning_rate": 2.919774236699444e-05, "loss": 0.3005, "step": 5565 }, { "epoch": 0.13, "learning_rate": 2.9196862312548403e-05, "loss": 0.2857, "step": 5570 }, { "epoch": 0.13, "learning_rate": 2.9195982258102368e-05, "loss": 0.2725, "step": 5575 }, { "epoch": 0.13, "learning_rate": 2.9195102203656333e-05, "loss": 0.3346, "step": 5580 }, { "epoch": 0.13, "learning_rate": 2.91942221492103e-05, "loss": 0.5167, "step": 5585 }, { "epoch": 0.13, "learning_rate": 2.9193342094764263e-05, "loss": 0.5005, "step": 5590 }, { "epoch": 0.13, "learning_rate": 2.9192462040318228e-05, "loss": 0.4596, "step": 5595 }, { "epoch": 0.13, "learning_rate": 2.9191581985872193e-05, "loss": 0.6731, "step": 5600 }, { "epoch": 0.13, "learning_rate": 2.9190701931426158e-05, "loss": 0.2786, "step": 5605 }, { "epoch": 0.13, "learning_rate": 2.9189821876980123e-05, "loss": 0.2898, "step": 5610 }, { "epoch": 0.13, "learning_rate": 2.9188941822534088e-05, "loss": 0.1198, "step": 5615 }, { "epoch": 0.13, "learning_rate": 2.9188061768088056e-05, "loss": 0.328, "step": 5620 }, { "epoch": 0.13, "learning_rate": 2.9187181713642018e-05, "loss": 0.246, "step": 5625 }, { "epoch": 0.13, "learning_rate": 2.9186301659195983e-05, "loss": 0.3451, "step": 5630 }, { "epoch": 0.13, "learning_rate": 2.9185421604749948e-05, "loss": 0.3305, "step": 5635 }, { "epoch": 0.13, "learning_rate": 2.9184541550303913e-05, "loss": 0.3954, "step": 5640 }, { "epoch": 0.13, "learning_rate": 2.9183661495857878e-05, "loss": 0.5994, "step": 5645 }, { "epoch": 0.13, "learning_rate": 2.9182781441411843e-05, "loss": 0.9544, "step": 5650 }, { "epoch": 0.13, "learning_rate": 2.9181901386965808e-05, "loss": 0.4609, "step": 5655 }, { "epoch": 0.13, "learning_rate": 2.9181021332519773e-05, "loss": 0.2223, "step": 5660 }, { "epoch": 0.13, "learning_rate": 2.9180141278073738e-05, "loss": 0.3322, "step": 5665 }, { "epoch": 0.13, "learning_rate": 2.9179261223627703e-05, "loss": 0.1826, "step": 5670 }, { "epoch": 0.13, "learning_rate": 2.9178381169181665e-05, "loss": 0.3104, "step": 5675 }, { "epoch": 0.13, "learning_rate": 2.9177501114735633e-05, "loss": 0.3448, "step": 5680 }, { "epoch": 0.13, "learning_rate": 2.9176621060289598e-05, "loss": 0.4235, "step": 5685 }, { "epoch": 0.13, "learning_rate": 2.9175741005843563e-05, "loss": 0.3709, "step": 5690 }, { "epoch": 0.13, "learning_rate": 2.9174860951397525e-05, "loss": 0.6475, "step": 5695 }, { "epoch": 0.13, "learning_rate": 2.9173980896951493e-05, "loss": 0.9096, "step": 5700 }, { "epoch": 0.13, "learning_rate": 2.9173100842505458e-05, "loss": 0.3934, "step": 5705 }, { "epoch": 0.13, "learning_rate": 2.917222078805942e-05, "loss": 0.128, "step": 5710 }, { "epoch": 0.13, "learning_rate": 2.9171340733613385e-05, "loss": 0.4078, "step": 5715 }, { "epoch": 0.13, "learning_rate": 2.9170460679167353e-05, "loss": 0.2417, "step": 5720 }, { "epoch": 0.13, "learning_rate": 2.9169580624721318e-05, "loss": 0.2507, "step": 5725 }, { "epoch": 0.13, "learning_rate": 2.916870057027528e-05, "loss": 0.2183, "step": 5730 }, { "epoch": 0.13, "learning_rate": 2.9167820515829245e-05, "loss": 0.4999, "step": 5735 }, { "epoch": 0.13, "learning_rate": 2.9166940461383213e-05, "loss": 0.5141, "step": 5740 }, { "epoch": 0.13, "learning_rate": 2.9166060406937175e-05, "loss": 0.7659, "step": 5745 }, { "epoch": 0.13, "learning_rate": 2.916518035249114e-05, "loss": 1.0446, "step": 5750 }, { "epoch": 0.13, "learning_rate": 2.916430029804511e-05, "loss": 0.4505, "step": 5755 }, { "epoch": 0.13, "learning_rate": 2.9163420243599073e-05, "loss": 0.2267, "step": 5760 }, { "epoch": 0.13, "learning_rate": 2.9162540189153035e-05, "loss": 0.1291, "step": 5765 }, { "epoch": 0.13, "learning_rate": 2.9161660134707e-05, "loss": 0.1815, "step": 5770 }, { "epoch": 0.13, "learning_rate": 2.916078008026097e-05, "loss": 0.4513, "step": 5775 }, { "epoch": 0.13, "learning_rate": 2.915990002581493e-05, "loss": 0.2883, "step": 5780 }, { "epoch": 0.13, "learning_rate": 2.9159019971368895e-05, "loss": 0.3012, "step": 5785 }, { "epoch": 0.14, "learning_rate": 2.915813991692286e-05, "loss": 0.3832, "step": 5790 }, { "epoch": 0.14, "learning_rate": 2.915725986247683e-05, "loss": 0.4978, "step": 5795 }, { "epoch": 0.14, "learning_rate": 2.915637980803079e-05, "loss": 0.7657, "step": 5800 }, { "epoch": 0.14, "learning_rate": 2.9155499753584755e-05, "loss": 0.4116, "step": 5805 }, { "epoch": 0.14, "learning_rate": 2.915461969913872e-05, "loss": 0.2178, "step": 5810 }, { "epoch": 0.14, "learning_rate": 2.915373964469269e-05, "loss": 0.3437, "step": 5815 }, { "epoch": 0.14, "learning_rate": 2.915285959024665e-05, "loss": 0.3989, "step": 5820 }, { "epoch": 0.14, "learning_rate": 2.9151979535800615e-05, "loss": 0.3568, "step": 5825 }, { "epoch": 0.14, "learning_rate": 2.915109948135458e-05, "loss": 0.4127, "step": 5830 }, { "epoch": 0.14, "learning_rate": 2.9150219426908545e-05, "loss": 0.3254, "step": 5835 }, { "epoch": 0.14, "learning_rate": 2.914933937246251e-05, "loss": 0.4184, "step": 5840 }, { "epoch": 0.14, "learning_rate": 2.9148459318016475e-05, "loss": 0.6575, "step": 5845 }, { "epoch": 0.14, "learning_rate": 2.914757926357044e-05, "loss": 0.8442, "step": 5850 }, { "epoch": 0.14, "learning_rate": 2.9146699209124405e-05, "loss": 0.3113, "step": 5855 }, { "epoch": 0.14, "learning_rate": 2.914581915467837e-05, "loss": 0.1861, "step": 5860 }, { "epoch": 0.14, "learning_rate": 2.9144939100232335e-05, "loss": 0.317, "step": 5865 }, { "epoch": 0.14, "learning_rate": 2.91440590457863e-05, "loss": 0.2375, "step": 5870 }, { "epoch": 0.14, "learning_rate": 2.9143178991340265e-05, "loss": 0.2634, "step": 5875 }, { "epoch": 0.14, "learning_rate": 2.914229893689423e-05, "loss": 0.2817, "step": 5880 }, { "epoch": 0.14, "learning_rate": 2.9141418882448196e-05, "loss": 0.5043, "step": 5885 }, { "epoch": 0.14, "learning_rate": 2.914053882800216e-05, "loss": 0.4063, "step": 5890 }, { "epoch": 0.14, "learning_rate": 2.9139658773556126e-05, "loss": 0.6614, "step": 5895 }, { "epoch": 0.14, "learning_rate": 2.913877871911009e-05, "loss": 0.5818, "step": 5900 }, { "epoch": 0.14, "learning_rate": 2.9137898664664052e-05, "loss": 0.4689, "step": 5905 }, { "epoch": 0.14, "learning_rate": 2.913701861021802e-05, "loss": 0.1645, "step": 5910 }, { "epoch": 0.14, "learning_rate": 2.9136138555771986e-05, "loss": 0.2472, "step": 5915 }, { "epoch": 0.14, "learning_rate": 2.913525850132595e-05, "loss": 0.3983, "step": 5920 }, { "epoch": 0.14, "learning_rate": 2.9134378446879912e-05, "loss": 0.4756, "step": 5925 }, { "epoch": 0.14, "learning_rate": 2.913349839243388e-05, "loss": 0.4263, "step": 5930 }, { "epoch": 0.14, "learning_rate": 2.9132618337987846e-05, "loss": 0.3849, "step": 5935 }, { "epoch": 0.14, "learning_rate": 2.9131738283541807e-05, "loss": 0.3353, "step": 5940 }, { "epoch": 0.14, "learning_rate": 2.9130858229095772e-05, "loss": 0.3457, "step": 5945 }, { "epoch": 0.14, "learning_rate": 2.912997817464974e-05, "loss": 0.6283, "step": 5950 }, { "epoch": 0.14, "learning_rate": 2.9129098120203706e-05, "loss": 0.3031, "step": 5955 }, { "epoch": 0.14, "learning_rate": 2.9128218065757667e-05, "loss": 0.2292, "step": 5960 }, { "epoch": 0.14, "learning_rate": 2.9127338011311632e-05, "loss": 0.2369, "step": 5965 }, { "epoch": 0.14, "learning_rate": 2.91264579568656e-05, "loss": 0.2671, "step": 5970 }, { "epoch": 0.14, "learning_rate": 2.9125577902419562e-05, "loss": 0.4478, "step": 5975 }, { "epoch": 0.14, "learning_rate": 2.9124697847973528e-05, "loss": 0.2746, "step": 5980 }, { "epoch": 0.14, "learning_rate": 2.9123817793527493e-05, "loss": 0.4355, "step": 5985 }, { "epoch": 0.14, "learning_rate": 2.912293773908146e-05, "loss": 0.5143, "step": 5990 }, { "epoch": 0.14, "learning_rate": 2.9122057684635423e-05, "loss": 0.5502, "step": 5995 }, { "epoch": 0.14, "learning_rate": 2.9121177630189388e-05, "loss": 0.8993, "step": 6000 }, { "epoch": 0.14, "learning_rate": 2.9120297575743353e-05, "loss": 0.4691, "step": 6005 }, { "epoch": 0.14, "learning_rate": 2.9119417521297318e-05, "loss": 0.176, "step": 6010 }, { "epoch": 0.14, "learning_rate": 2.9118537466851283e-05, "loss": 0.2264, "step": 6015 }, { "epoch": 0.14, "learning_rate": 2.9117657412405248e-05, "loss": 0.3147, "step": 6020 }, { "epoch": 0.14, "learning_rate": 2.9116777357959216e-05, "loss": 0.273, "step": 6025 }, { "epoch": 0.14, "learning_rate": 2.9115897303513178e-05, "loss": 0.2966, "step": 6030 }, { "epoch": 0.14, "learning_rate": 2.9115017249067143e-05, "loss": 0.3953, "step": 6035 }, { "epoch": 0.14, "learning_rate": 2.9114137194621108e-05, "loss": 0.4554, "step": 6040 }, { "epoch": 0.14, "learning_rate": 2.9113257140175073e-05, "loss": 0.5646, "step": 6045 }, { "epoch": 0.14, "learning_rate": 2.9112377085729038e-05, "loss": 0.9268, "step": 6050 }, { "epoch": 0.14, "learning_rate": 2.9111497031283003e-05, "loss": 0.3464, "step": 6055 }, { "epoch": 0.14, "learning_rate": 2.9110616976836968e-05, "loss": 0.2905, "step": 6060 }, { "epoch": 0.14, "learning_rate": 2.9109736922390933e-05, "loss": 0.2065, "step": 6065 }, { "epoch": 0.14, "learning_rate": 2.9108856867944898e-05, "loss": 0.3534, "step": 6070 }, { "epoch": 0.14, "learning_rate": 2.9107976813498863e-05, "loss": 0.3777, "step": 6075 }, { "epoch": 0.14, "learning_rate": 2.9107096759052825e-05, "loss": 0.4595, "step": 6080 }, { "epoch": 0.14, "learning_rate": 2.9106216704606793e-05, "loss": 0.3001, "step": 6085 }, { "epoch": 0.14, "learning_rate": 2.9105336650160758e-05, "loss": 0.396, "step": 6090 }, { "epoch": 0.14, "learning_rate": 2.9104456595714723e-05, "loss": 0.6303, "step": 6095 }, { "epoch": 0.14, "learning_rate": 2.9103576541268685e-05, "loss": 1.0388, "step": 6100 }, { "epoch": 0.14, "learning_rate": 2.9102696486822653e-05, "loss": 0.2936, "step": 6105 }, { "epoch": 0.14, "learning_rate": 2.9101816432376618e-05, "loss": 0.2194, "step": 6110 }, { "epoch": 0.14, "learning_rate": 2.9100936377930583e-05, "loss": 0.2588, "step": 6115 }, { "epoch": 0.14, "learning_rate": 2.9100056323484545e-05, "loss": 0.2589, "step": 6120 }, { "epoch": 0.14, "learning_rate": 2.9099176269038513e-05, "loss": 0.3171, "step": 6125 }, { "epoch": 0.14, "learning_rate": 2.9098296214592478e-05, "loss": 0.2481, "step": 6130 }, { "epoch": 0.14, "learning_rate": 2.909741616014644e-05, "loss": 0.4368, "step": 6135 }, { "epoch": 0.14, "learning_rate": 2.9096536105700405e-05, "loss": 0.4741, "step": 6140 }, { "epoch": 0.14, "learning_rate": 2.9095656051254373e-05, "loss": 0.5977, "step": 6145 }, { "epoch": 0.14, "learning_rate": 2.9094775996808338e-05, "loss": 0.9377, "step": 6150 }, { "epoch": 0.14, "learning_rate": 2.90938959423623e-05, "loss": 0.3671, "step": 6155 }, { "epoch": 0.14, "learning_rate": 2.9093015887916268e-05, "loss": 0.1649, "step": 6160 }, { "epoch": 0.14, "learning_rate": 2.9092135833470233e-05, "loss": 0.1499, "step": 6165 }, { "epoch": 0.14, "learning_rate": 2.9091255779024195e-05, "loss": 0.1752, "step": 6170 }, { "epoch": 0.14, "learning_rate": 2.909037572457816e-05, "loss": 0.2962, "step": 6175 }, { "epoch": 0.14, "learning_rate": 2.908949567013213e-05, "loss": 0.2942, "step": 6180 }, { "epoch": 0.14, "learning_rate": 2.9088615615686093e-05, "loss": 0.4653, "step": 6185 }, { "epoch": 0.14, "learning_rate": 2.9087735561240055e-05, "loss": 0.5281, "step": 6190 }, { "epoch": 0.14, "learning_rate": 2.908685550679402e-05, "loss": 0.6148, "step": 6195 }, { "epoch": 0.14, "learning_rate": 2.908597545234799e-05, "loss": 0.8229, "step": 6200 }, { "epoch": 0.14, "learning_rate": 2.908509539790195e-05, "loss": 0.4186, "step": 6205 }, { "epoch": 0.14, "learning_rate": 2.9084215343455915e-05, "loss": 0.2836, "step": 6210 }, { "epoch": 0.15, "learning_rate": 2.908333528900988e-05, "loss": 0.2396, "step": 6215 }, { "epoch": 0.15, "learning_rate": 2.908245523456385e-05, "loss": 0.2721, "step": 6220 }, { "epoch": 0.15, "learning_rate": 2.908157518011781e-05, "loss": 0.3037, "step": 6225 }, { "epoch": 0.15, "learning_rate": 2.9080695125671775e-05, "loss": 0.3026, "step": 6230 }, { "epoch": 0.15, "learning_rate": 2.907981507122574e-05, "loss": 0.4992, "step": 6235 }, { "epoch": 0.15, "learning_rate": 2.9078935016779705e-05, "loss": 0.4753, "step": 6240 }, { "epoch": 0.15, "learning_rate": 2.907805496233367e-05, "loss": 0.7627, "step": 6245 }, { "epoch": 0.15, "learning_rate": 2.9077174907887635e-05, "loss": 0.7303, "step": 6250 }, { "epoch": 0.15, "learning_rate": 2.90762948534416e-05, "loss": 0.3791, "step": 6255 }, { "epoch": 0.15, "learning_rate": 2.9075414798995565e-05, "loss": 0.2209, "step": 6260 }, { "epoch": 0.15, "learning_rate": 2.907453474454953e-05, "loss": 0.2905, "step": 6265 }, { "epoch": 0.15, "learning_rate": 2.9073654690103495e-05, "loss": 0.3049, "step": 6270 }, { "epoch": 0.15, "learning_rate": 2.9072774635657457e-05, "loss": 0.3421, "step": 6275 }, { "epoch": 0.15, "learning_rate": 2.9071894581211425e-05, "loss": 0.2622, "step": 6280 }, { "epoch": 0.15, "learning_rate": 2.907101452676539e-05, "loss": 0.4227, "step": 6285 }, { "epoch": 0.15, "learning_rate": 2.9070134472319355e-05, "loss": 0.3183, "step": 6290 }, { "epoch": 0.15, "learning_rate": 2.906925441787332e-05, "loss": 0.4834, "step": 6295 }, { "epoch": 0.15, "learning_rate": 2.9068374363427285e-05, "loss": 0.8153, "step": 6300 }, { "epoch": 0.15, "learning_rate": 2.906749430898125e-05, "loss": 0.3177, "step": 6305 }, { "epoch": 0.15, "learning_rate": 2.9066614254535212e-05, "loss": 0.2292, "step": 6310 }, { "epoch": 0.15, "learning_rate": 2.906573420008918e-05, "loss": 0.2222, "step": 6315 }, { "epoch": 0.15, "learning_rate": 2.9064854145643146e-05, "loss": 0.1521, "step": 6320 }, { "epoch": 0.15, "learning_rate": 2.906397409119711e-05, "loss": 0.1855, "step": 6325 }, { "epoch": 0.15, "learning_rate": 2.9063094036751072e-05, "loss": 0.4017, "step": 6330 }, { "epoch": 0.15, "learning_rate": 2.906221398230504e-05, "loss": 0.3847, "step": 6335 }, { "epoch": 0.15, "learning_rate": 2.9061333927859006e-05, "loss": 0.5455, "step": 6340 }, { "epoch": 0.15, "learning_rate": 2.9060453873412967e-05, "loss": 0.7908, "step": 6345 }, { "epoch": 0.15, "learning_rate": 2.9059573818966932e-05, "loss": 1.1634, "step": 6350 }, { "epoch": 0.15, "learning_rate": 2.90586937645209e-05, "loss": 0.352, "step": 6355 }, { "epoch": 0.15, "learning_rate": 2.9057813710074866e-05, "loss": 0.1704, "step": 6360 }, { "epoch": 0.15, "learning_rate": 2.9056933655628827e-05, "loss": 0.315, "step": 6365 }, { "epoch": 0.15, "learning_rate": 2.9056053601182792e-05, "loss": 0.2924, "step": 6370 }, { "epoch": 0.15, "learning_rate": 2.905517354673676e-05, "loss": 0.3227, "step": 6375 }, { "epoch": 0.15, "learning_rate": 2.9054293492290722e-05, "loss": 0.386, "step": 6380 }, { "epoch": 0.15, "learning_rate": 2.9053413437844687e-05, "loss": 0.5854, "step": 6385 }, { "epoch": 0.15, "learning_rate": 2.9052533383398652e-05, "loss": 0.5877, "step": 6390 }, { "epoch": 0.15, "learning_rate": 2.905165332895262e-05, "loss": 0.4867, "step": 6395 }, { "epoch": 0.15, "learning_rate": 2.9050773274506582e-05, "loss": 0.7082, "step": 6400 }, { "epoch": 0.15, "learning_rate": 2.9049893220060547e-05, "loss": 0.459, "step": 6405 }, { "epoch": 0.15, "learning_rate": 2.9049013165614512e-05, "loss": 0.1684, "step": 6410 }, { "epoch": 0.15, "learning_rate": 2.9048133111168478e-05, "loss": 0.3148, "step": 6415 }, { "epoch": 0.15, "learning_rate": 2.9047253056722443e-05, "loss": 0.4022, "step": 6420 }, { "epoch": 0.15, "learning_rate": 2.9046373002276408e-05, "loss": 0.2483, "step": 6425 }, { "epoch": 0.15, "learning_rate": 2.9045492947830376e-05, "loss": 0.3093, "step": 6430 }, { "epoch": 0.15, "learning_rate": 2.9044612893384338e-05, "loss": 0.3539, "step": 6435 }, { "epoch": 0.15, "learning_rate": 2.9043732838938303e-05, "loss": 0.5225, "step": 6440 }, { "epoch": 0.15, "learning_rate": 2.9042852784492268e-05, "loss": 0.4285, "step": 6445 }, { "epoch": 0.15, "learning_rate": 2.9041972730046236e-05, "loss": 0.7605, "step": 6450 }, { "epoch": 0.15, "learning_rate": 2.9041092675600198e-05, "loss": 0.3846, "step": 6455 }, { "epoch": 0.15, "learning_rate": 2.9040212621154163e-05, "loss": 0.2351, "step": 6460 }, { "epoch": 0.15, "learning_rate": 2.9039332566708128e-05, "loss": 0.3095, "step": 6465 }, { "epoch": 0.15, "learning_rate": 2.9038452512262093e-05, "loss": 0.2918, "step": 6470 }, { "epoch": 0.15, "learning_rate": 2.9037572457816058e-05, "loss": 0.3107, "step": 6475 }, { "epoch": 0.15, "learning_rate": 2.9036692403370023e-05, "loss": 0.2969, "step": 6480 }, { "epoch": 0.15, "learning_rate": 2.9035812348923988e-05, "loss": 0.4042, "step": 6485 }, { "epoch": 0.15, "learning_rate": 2.9034932294477953e-05, "loss": 0.488, "step": 6490 }, { "epoch": 0.15, "learning_rate": 2.9034052240031918e-05, "loss": 0.4975, "step": 6495 }, { "epoch": 0.15, "learning_rate": 2.9033172185585883e-05, "loss": 0.7583, "step": 6500 }, { "epoch": 0.15, "learning_rate": 2.9032292131139844e-05, "loss": 0.4411, "step": 6505 }, { "epoch": 0.15, "learning_rate": 2.9031412076693813e-05, "loss": 0.1735, "step": 6510 }, { "epoch": 0.15, "learning_rate": 2.9030532022247778e-05, "loss": 0.1567, "step": 6515 }, { "epoch": 0.15, "learning_rate": 2.9029651967801743e-05, "loss": 0.1683, "step": 6520 }, { "epoch": 0.15, "learning_rate": 2.9028771913355705e-05, "loss": 0.3822, "step": 6525 }, { "epoch": 0.15, "learning_rate": 2.9027891858909673e-05, "loss": 0.5757, "step": 6530 }, { "epoch": 0.15, "learning_rate": 2.9027011804463638e-05, "loss": 0.5013, "step": 6535 }, { "epoch": 0.15, "learning_rate": 2.90261317500176e-05, "loss": 0.3524, "step": 6540 }, { "epoch": 0.15, "learning_rate": 2.9025251695571565e-05, "loss": 0.5556, "step": 6545 }, { "epoch": 0.15, "learning_rate": 2.9024371641125533e-05, "loss": 0.5775, "step": 6550 }, { "epoch": 0.15, "learning_rate": 2.9023491586679498e-05, "loss": 0.3531, "step": 6555 }, { "epoch": 0.15, "learning_rate": 2.902261153223346e-05, "loss": 0.1707, "step": 6560 }, { "epoch": 0.15, "learning_rate": 2.9021731477787428e-05, "loss": 0.211, "step": 6565 }, { "epoch": 0.15, "learning_rate": 2.9020851423341393e-05, "loss": 0.3257, "step": 6570 }, { "epoch": 0.15, "learning_rate": 2.9019971368895355e-05, "loss": 0.3277, "step": 6575 }, { "epoch": 0.15, "learning_rate": 2.901909131444932e-05, "loss": 0.344, "step": 6580 }, { "epoch": 0.15, "learning_rate": 2.9018211260003288e-05, "loss": 0.422, "step": 6585 }, { "epoch": 0.15, "learning_rate": 2.9017331205557253e-05, "loss": 0.4581, "step": 6590 }, { "epoch": 0.15, "learning_rate": 2.9016451151111215e-05, "loss": 0.3822, "step": 6595 }, { "epoch": 0.15, "learning_rate": 2.901557109666518e-05, "loss": 0.6911, "step": 6600 }, { "epoch": 0.15, "learning_rate": 2.9014691042219148e-05, "loss": 0.4282, "step": 6605 }, { "epoch": 0.15, "learning_rate": 2.901381098777311e-05, "loss": 0.2824, "step": 6610 }, { "epoch": 0.15, "learning_rate": 2.9012930933327075e-05, "loss": 0.3284, "step": 6615 }, { "epoch": 0.15, "learning_rate": 2.901205087888104e-05, "loss": 0.2313, "step": 6620 }, { "epoch": 0.15, "learning_rate": 2.901117082443501e-05, "loss": 0.2443, "step": 6625 }, { "epoch": 0.15, "learning_rate": 2.901029076998897e-05, "loss": 0.4728, "step": 6630 }, { "epoch": 0.15, "learning_rate": 2.9009410715542935e-05, "loss": 0.4028, "step": 6635 }, { "epoch": 0.15, "learning_rate": 2.90085306610969e-05, "loss": 0.4416, "step": 6640 }, { "epoch": 0.16, "learning_rate": 2.9007650606650865e-05, "loss": 0.6738, "step": 6645 }, { "epoch": 0.16, "learning_rate": 2.900677055220483e-05, "loss": 0.5387, "step": 6650 }, { "epoch": 0.16, "learning_rate": 2.9005890497758795e-05, "loss": 0.3581, "step": 6655 }, { "epoch": 0.16, "learning_rate": 2.900501044331276e-05, "loss": 0.2738, "step": 6660 }, { "epoch": 0.16, "learning_rate": 2.9004130388866725e-05, "loss": 0.2652, "step": 6665 }, { "epoch": 0.16, "learning_rate": 2.900325033442069e-05, "loss": 0.1761, "step": 6670 }, { "epoch": 0.16, "learning_rate": 2.9002370279974655e-05, "loss": 0.3469, "step": 6675 }, { "epoch": 0.16, "learning_rate": 2.9001490225528617e-05, "loss": 0.4881, "step": 6680 }, { "epoch": 0.16, "learning_rate": 2.9000610171082585e-05, "loss": 0.5325, "step": 6685 }, { "epoch": 0.16, "learning_rate": 2.899973011663655e-05, "loss": 0.4617, "step": 6690 }, { "epoch": 0.16, "learning_rate": 2.8998850062190515e-05, "loss": 0.5265, "step": 6695 }, { "epoch": 0.16, "learning_rate": 2.899797000774448e-05, "loss": 0.5697, "step": 6700 }, { "epoch": 0.16, "learning_rate": 2.8997089953298445e-05, "loss": 0.3618, "step": 6705 }, { "epoch": 0.16, "learning_rate": 2.899620989885241e-05, "loss": 0.1451, "step": 6710 }, { "epoch": 0.16, "learning_rate": 2.8995329844406372e-05, "loss": 0.31, "step": 6715 }, { "epoch": 0.16, "learning_rate": 2.899444978996034e-05, "loss": 0.1722, "step": 6720 }, { "epoch": 0.16, "learning_rate": 2.8993569735514305e-05, "loss": 0.298, "step": 6725 }, { "epoch": 0.16, "learning_rate": 2.899268968106827e-05, "loss": 0.4093, "step": 6730 }, { "epoch": 0.16, "learning_rate": 2.8991809626622232e-05, "loss": 0.236, "step": 6735 }, { "epoch": 0.16, "learning_rate": 2.89909295721762e-05, "loss": 0.5449, "step": 6740 }, { "epoch": 0.16, "learning_rate": 2.8990049517730165e-05, "loss": 0.3364, "step": 6745 }, { "epoch": 0.16, "learning_rate": 2.898916946328413e-05, "loss": 0.9739, "step": 6750 }, { "epoch": 0.16, "learning_rate": 2.8988289408838092e-05, "loss": 0.4219, "step": 6755 }, { "epoch": 0.16, "learning_rate": 2.898740935439206e-05, "loss": 0.1691, "step": 6760 }, { "epoch": 0.16, "learning_rate": 2.8986529299946026e-05, "loss": 0.2992, "step": 6765 }, { "epoch": 0.16, "learning_rate": 2.8985649245499987e-05, "loss": 0.2402, "step": 6770 }, { "epoch": 0.16, "learning_rate": 2.8984769191053952e-05, "loss": 0.4319, "step": 6775 }, { "epoch": 0.16, "learning_rate": 2.898388913660792e-05, "loss": 0.3247, "step": 6780 }, { "epoch": 0.16, "learning_rate": 2.8983009082161886e-05, "loss": 0.3053, "step": 6785 }, { "epoch": 0.16, "learning_rate": 2.8982129027715847e-05, "loss": 0.3, "step": 6790 }, { "epoch": 0.16, "learning_rate": 2.8981248973269812e-05, "loss": 0.5269, "step": 6795 }, { "epoch": 0.16, "learning_rate": 2.898036891882378e-05, "loss": 0.6143, "step": 6800 }, { "epoch": 0.16, "learning_rate": 2.8979488864377742e-05, "loss": 0.4245, "step": 6805 }, { "epoch": 0.16, "learning_rate": 2.8978608809931707e-05, "loss": 0.2036, "step": 6810 }, { "epoch": 0.16, "learning_rate": 2.8977728755485672e-05, "loss": 0.2534, "step": 6815 }, { "epoch": 0.16, "learning_rate": 2.897684870103964e-05, "loss": 0.1707, "step": 6820 }, { "epoch": 0.16, "learning_rate": 2.8975968646593602e-05, "loss": 0.4339, "step": 6825 }, { "epoch": 0.16, "learning_rate": 2.8975088592147567e-05, "loss": 0.4178, "step": 6830 }, { "epoch": 0.16, "learning_rate": 2.8974208537701536e-05, "loss": 0.5314, "step": 6835 }, { "epoch": 0.16, "learning_rate": 2.8973328483255497e-05, "loss": 0.4271, "step": 6840 }, { "epoch": 0.16, "learning_rate": 2.8972448428809463e-05, "loss": 0.5296, "step": 6845 }, { "epoch": 0.16, "learning_rate": 2.8971568374363428e-05, "loss": 0.5788, "step": 6850 }, { "epoch": 0.16, "learning_rate": 2.8970688319917396e-05, "loss": 0.297, "step": 6855 }, { "epoch": 0.16, "learning_rate": 2.8969808265471358e-05, "loss": 0.1879, "step": 6860 }, { "epoch": 0.16, "learning_rate": 2.8968928211025323e-05, "loss": 0.2564, "step": 6865 }, { "epoch": 0.16, "learning_rate": 2.8968048156579288e-05, "loss": 0.248, "step": 6870 }, { "epoch": 0.16, "learning_rate": 2.8967168102133253e-05, "loss": 0.3852, "step": 6875 }, { "epoch": 0.16, "learning_rate": 2.8966288047687218e-05, "loss": 0.311, "step": 6880 }, { "epoch": 0.16, "learning_rate": 2.8965407993241183e-05, "loss": 0.4212, "step": 6885 }, { "epoch": 0.16, "learning_rate": 2.8964527938795148e-05, "loss": 0.5224, "step": 6890 }, { "epoch": 0.16, "learning_rate": 2.8963647884349113e-05, "loss": 0.5854, "step": 6895 }, { "epoch": 0.16, "learning_rate": 2.8962767829903078e-05, "loss": 0.5388, "step": 6900 }, { "epoch": 0.16, "learning_rate": 2.8961887775457043e-05, "loss": 0.3011, "step": 6905 }, { "epoch": 0.16, "learning_rate": 2.8961007721011004e-05, "loss": 0.1577, "step": 6910 }, { "epoch": 0.16, "learning_rate": 2.8960127666564973e-05, "loss": 0.2199, "step": 6915 }, { "epoch": 0.16, "learning_rate": 2.8959247612118938e-05, "loss": 0.3484, "step": 6920 }, { "epoch": 0.16, "learning_rate": 2.8958367557672903e-05, "loss": 0.5117, "step": 6925 }, { "epoch": 0.16, "learning_rate": 2.8957487503226864e-05, "loss": 0.3499, "step": 6930 }, { "epoch": 0.16, "learning_rate": 2.8956607448780833e-05, "loss": 0.4629, "step": 6935 }, { "epoch": 0.16, "learning_rate": 2.8955727394334798e-05, "loss": 0.4721, "step": 6940 }, { "epoch": 0.16, "learning_rate": 2.895484733988876e-05, "loss": 0.52, "step": 6945 }, { "epoch": 0.16, "learning_rate": 2.8953967285442725e-05, "loss": 0.5846, "step": 6950 }, { "epoch": 0.16, "learning_rate": 2.8953087230996693e-05, "loss": 0.4811, "step": 6955 }, { "epoch": 0.16, "learning_rate": 2.8952207176550658e-05, "loss": 0.2274, "step": 6960 }, { "epoch": 0.16, "learning_rate": 2.895132712210462e-05, "loss": 0.1096, "step": 6965 }, { "epoch": 0.16, "learning_rate": 2.8950447067658588e-05, "loss": 0.3297, "step": 6970 }, { "epoch": 0.16, "learning_rate": 2.8949567013212553e-05, "loss": 0.4113, "step": 6975 }, { "epoch": 0.16, "learning_rate": 2.8948686958766515e-05, "loss": 0.2605, "step": 6980 }, { "epoch": 0.16, "learning_rate": 2.894780690432048e-05, "loss": 0.4163, "step": 6985 }, { "epoch": 0.16, "learning_rate": 2.8946926849874448e-05, "loss": 0.4244, "step": 6990 }, { "epoch": 0.16, "learning_rate": 2.8946046795428413e-05, "loss": 0.3709, "step": 6995 }, { "epoch": 0.16, "learning_rate": 2.8945166740982375e-05, "loss": 0.679, "step": 7000 }, { "epoch": 0.16, "learning_rate": 2.894428668653634e-05, "loss": 0.3361, "step": 7005 }, { "epoch": 0.16, "learning_rate": 2.8943406632090308e-05, "loss": 0.1343, "step": 7010 }, { "epoch": 0.16, "learning_rate": 2.894252657764427e-05, "loss": 0.2151, "step": 7015 }, { "epoch": 0.16, "learning_rate": 2.8941646523198235e-05, "loss": 0.2826, "step": 7020 }, { "epoch": 0.16, "learning_rate": 2.89407664687522e-05, "loss": 0.3063, "step": 7025 }, { "epoch": 0.16, "learning_rate": 2.8939886414306168e-05, "loss": 0.3585, "step": 7030 }, { "epoch": 0.16, "learning_rate": 2.893900635986013e-05, "loss": 0.3749, "step": 7035 }, { "epoch": 0.16, "learning_rate": 2.8938126305414095e-05, "loss": 0.4872, "step": 7040 }, { "epoch": 0.16, "learning_rate": 2.893724625096806e-05, "loss": 0.4563, "step": 7045 }, { "epoch": 0.16, "learning_rate": 2.893636619652203e-05, "loss": 0.5223, "step": 7050 }, { "epoch": 0.16, "learning_rate": 2.893548614207599e-05, "loss": 0.3213, "step": 7055 }, { "epoch": 0.16, "learning_rate": 2.8934606087629955e-05, "loss": 0.1585, "step": 7060 }, { "epoch": 0.16, "learning_rate": 2.893372603318392e-05, "loss": 0.2524, "step": 7065 }, { "epoch": 0.16, "learning_rate": 2.8932845978737885e-05, "loss": 0.3309, "step": 7070 }, { "epoch": 0.17, "learning_rate": 2.893196592429185e-05, "loss": 0.4199, "step": 7075 }, { "epoch": 0.17, "learning_rate": 2.8931085869845815e-05, "loss": 0.2489, "step": 7080 }, { "epoch": 0.17, "learning_rate": 2.893020581539978e-05, "loss": 0.3263, "step": 7085 }, { "epoch": 0.17, "learning_rate": 2.8929325760953745e-05, "loss": 0.3912, "step": 7090 }, { "epoch": 0.17, "learning_rate": 2.892844570650771e-05, "loss": 0.5768, "step": 7095 }, { "epoch": 0.17, "learning_rate": 2.8927565652061675e-05, "loss": 0.5619, "step": 7100 }, { "epoch": 0.17, "learning_rate": 2.892668559761564e-05, "loss": 0.3496, "step": 7105 }, { "epoch": 0.17, "learning_rate": 2.8925805543169605e-05, "loss": 0.2143, "step": 7110 }, { "epoch": 0.17, "learning_rate": 2.892492548872357e-05, "loss": 0.2267, "step": 7115 }, { "epoch": 0.17, "learning_rate": 2.8924045434277535e-05, "loss": 0.3221, "step": 7120 }, { "epoch": 0.17, "learning_rate": 2.89231653798315e-05, "loss": 0.2932, "step": 7125 }, { "epoch": 0.17, "learning_rate": 2.8922285325385465e-05, "loss": 0.4092, "step": 7130 }, { "epoch": 0.17, "learning_rate": 2.892140527093943e-05, "loss": 0.2444, "step": 7135 }, { "epoch": 0.17, "learning_rate": 2.8920525216493392e-05, "loss": 0.4049, "step": 7140 }, { "epoch": 0.17, "learning_rate": 2.891964516204736e-05, "loss": 0.4149, "step": 7145 }, { "epoch": 0.17, "learning_rate": 2.8918765107601325e-05, "loss": 0.6999, "step": 7150 }, { "epoch": 0.17, "learning_rate": 2.891788505315529e-05, "loss": 0.3469, "step": 7155 }, { "epoch": 0.17, "learning_rate": 2.8917004998709252e-05, "loss": 0.1191, "step": 7160 }, { "epoch": 0.17, "learning_rate": 2.891612494426322e-05, "loss": 0.15, "step": 7165 }, { "epoch": 0.17, "learning_rate": 2.8915244889817185e-05, "loss": 0.3576, "step": 7170 }, { "epoch": 0.17, "learning_rate": 2.8914364835371147e-05, "loss": 0.3947, "step": 7175 }, { "epoch": 0.17, "learning_rate": 2.8913484780925112e-05, "loss": 0.4685, "step": 7180 }, { "epoch": 0.17, "learning_rate": 2.891260472647908e-05, "loss": 0.319, "step": 7185 }, { "epoch": 0.17, "learning_rate": 2.8911724672033046e-05, "loss": 0.51, "step": 7190 }, { "epoch": 0.17, "learning_rate": 2.8910844617587007e-05, "loss": 0.535, "step": 7195 }, { "epoch": 0.17, "learning_rate": 2.8909964563140972e-05, "loss": 0.5526, "step": 7200 }, { "epoch": 0.17, "learning_rate": 2.890908450869494e-05, "loss": 0.3954, "step": 7205 }, { "epoch": 0.17, "learning_rate": 2.8908204454248902e-05, "loss": 0.124, "step": 7210 }, { "epoch": 0.17, "learning_rate": 2.8907324399802867e-05, "loss": 0.2888, "step": 7215 }, { "epoch": 0.17, "learning_rate": 2.8906444345356832e-05, "loss": 0.2498, "step": 7220 }, { "epoch": 0.17, "learning_rate": 2.89055642909108e-05, "loss": 0.2455, "step": 7225 }, { "epoch": 0.17, "learning_rate": 2.8904684236464762e-05, "loss": 0.2134, "step": 7230 }, { "epoch": 0.17, "learning_rate": 2.8903804182018727e-05, "loss": 0.2924, "step": 7235 }, { "epoch": 0.17, "learning_rate": 2.8903100138461902e-05, "loss": 0.4881, "step": 7240 }, { "epoch": 0.17, "learning_rate": 2.8902220084015867e-05, "loss": 0.4126, "step": 7245 }, { "epoch": 0.17, "learning_rate": 2.890134002956983e-05, "loss": 0.883, "step": 7250 }, { "epoch": 0.17, "learning_rate": 2.8900459975123794e-05, "loss": 0.3759, "step": 7255 }, { "epoch": 0.17, "learning_rate": 2.8899579920677762e-05, "loss": 0.1397, "step": 7260 }, { "epoch": 0.17, "learning_rate": 2.8898699866231724e-05, "loss": 0.2435, "step": 7265 }, { "epoch": 0.17, "learning_rate": 2.889781981178569e-05, "loss": 0.2095, "step": 7270 }, { "epoch": 0.17, "learning_rate": 2.8896939757339654e-05, "loss": 0.2333, "step": 7275 }, { "epoch": 0.17, "learning_rate": 2.8896059702893622e-05, "loss": 0.2994, "step": 7280 }, { "epoch": 0.17, "learning_rate": 2.8895179648447584e-05, "loss": 0.4115, "step": 7285 }, { "epoch": 0.17, "learning_rate": 2.889429959400155e-05, "loss": 0.4315, "step": 7290 }, { "epoch": 0.17, "learning_rate": 2.8893419539555517e-05, "loss": 0.6321, "step": 7295 }, { "epoch": 0.17, "learning_rate": 2.889253948510948e-05, "loss": 0.7341, "step": 7300 }, { "epoch": 0.17, "learning_rate": 2.8891659430663444e-05, "loss": 0.5027, "step": 7305 }, { "epoch": 0.17, "learning_rate": 2.889077937621741e-05, "loss": 0.1673, "step": 7310 }, { "epoch": 0.17, "learning_rate": 2.8889899321771377e-05, "loss": 0.2437, "step": 7315 }, { "epoch": 0.17, "learning_rate": 2.888901926732534e-05, "loss": 0.2585, "step": 7320 }, { "epoch": 0.17, "learning_rate": 2.8888139212879304e-05, "loss": 0.2675, "step": 7325 }, { "epoch": 0.17, "learning_rate": 2.888725915843327e-05, "loss": 0.2261, "step": 7330 }, { "epoch": 0.17, "learning_rate": 2.8886379103987234e-05, "loss": 0.3268, "step": 7335 }, { "epoch": 0.17, "learning_rate": 2.88854990495412e-05, "loss": 0.3871, "step": 7340 }, { "epoch": 0.17, "learning_rate": 2.8884618995095164e-05, "loss": 0.4481, "step": 7345 }, { "epoch": 0.17, "learning_rate": 2.888373894064913e-05, "loss": 1.0351, "step": 7350 }, { "epoch": 0.17, "learning_rate": 2.8882858886203094e-05, "loss": 0.2345, "step": 7355 }, { "epoch": 0.17, "learning_rate": 2.888197883175706e-05, "loss": 0.296, "step": 7360 }, { "epoch": 0.17, "learning_rate": 2.8881098777311024e-05, "loss": 0.2227, "step": 7365 }, { "epoch": 0.17, "learning_rate": 2.8880218722864986e-05, "loss": 0.2788, "step": 7370 }, { "epoch": 0.17, "learning_rate": 2.8879338668418954e-05, "loss": 0.3534, "step": 7375 }, { "epoch": 0.17, "learning_rate": 2.887845861397292e-05, "loss": 0.4974, "step": 7380 }, { "epoch": 0.17, "learning_rate": 2.8877578559526884e-05, "loss": 0.3479, "step": 7385 }, { "epoch": 0.17, "learning_rate": 2.8876698505080846e-05, "loss": 0.3366, "step": 7390 }, { "epoch": 0.17, "learning_rate": 2.8875818450634814e-05, "loss": 0.7585, "step": 7395 }, { "epoch": 0.17, "learning_rate": 2.887493839618878e-05, "loss": 0.6275, "step": 7400 }, { "epoch": 0.17, "learning_rate": 2.887405834174274e-05, "loss": 0.3719, "step": 7405 }, { "epoch": 0.17, "learning_rate": 2.8873178287296706e-05, "loss": 0.1992, "step": 7410 }, { "epoch": 0.17, "learning_rate": 2.8872298232850674e-05, "loss": 0.1032, "step": 7415 }, { "epoch": 0.17, "learning_rate": 2.887141817840464e-05, "loss": 0.2661, "step": 7420 }, { "epoch": 0.17, "learning_rate": 2.88705381239586e-05, "loss": 0.1884, "step": 7425 }, { "epoch": 0.17, "learning_rate": 2.886965806951257e-05, "loss": 0.3462, "step": 7430 }, { "epoch": 0.17, "learning_rate": 2.8868778015066534e-05, "loss": 0.2648, "step": 7435 }, { "epoch": 0.17, "learning_rate": 2.8867897960620496e-05, "loss": 0.3085, "step": 7440 }, { "epoch": 0.17, "learning_rate": 2.886701790617446e-05, "loss": 0.4876, "step": 7445 }, { "epoch": 0.17, "learning_rate": 2.886613785172843e-05, "loss": 0.7575, "step": 7450 }, { "epoch": 0.17, "learning_rate": 2.8865257797282395e-05, "loss": 0.4086, "step": 7455 }, { "epoch": 0.17, "learning_rate": 2.8864377742836356e-05, "loss": 0.1666, "step": 7460 }, { "epoch": 0.17, "learning_rate": 2.886349768839032e-05, "loss": 0.3263, "step": 7465 }, { "epoch": 0.17, "learning_rate": 2.886261763394429e-05, "loss": 0.305, "step": 7470 }, { "epoch": 0.17, "learning_rate": 2.886173757949825e-05, "loss": 0.253, "step": 7475 }, { "epoch": 0.17, "learning_rate": 2.8860857525052216e-05, "loss": 0.3575, "step": 7480 }, { "epoch": 0.17, "learning_rate": 2.885997747060618e-05, "loss": 0.5598, "step": 7485 }, { "epoch": 0.17, "learning_rate": 2.885909741616015e-05, "loss": 0.3394, "step": 7490 }, { "epoch": 0.17, "learning_rate": 2.885821736171411e-05, "loss": 0.3777, "step": 7495 }, { "epoch": 0.17, "learning_rate": 2.8857337307268076e-05, "loss": 0.5826, "step": 7500 }, { "epoch": 0.18, "learning_rate": 2.885645725282204e-05, "loss": 0.3967, "step": 7505 }, { "epoch": 0.18, "learning_rate": 2.8855577198376006e-05, "loss": 0.2813, "step": 7510 }, { "epoch": 0.18, "learning_rate": 2.885469714392997e-05, "loss": 0.1707, "step": 7515 }, { "epoch": 0.18, "learning_rate": 2.8853817089483936e-05, "loss": 0.1546, "step": 7520 }, { "epoch": 0.18, "learning_rate": 2.88529370350379e-05, "loss": 0.3664, "step": 7525 }, { "epoch": 0.18, "learning_rate": 2.8852056980591866e-05, "loss": 0.3855, "step": 7530 }, { "epoch": 0.18, "learning_rate": 2.885117692614583e-05, "loss": 0.5186, "step": 7535 }, { "epoch": 0.18, "learning_rate": 2.8850296871699797e-05, "loss": 0.5377, "step": 7540 }, { "epoch": 0.18, "learning_rate": 2.884941681725376e-05, "loss": 0.4615, "step": 7545 }, { "epoch": 0.18, "learning_rate": 2.8848536762807727e-05, "loss": 0.8683, "step": 7550 }, { "epoch": 0.18, "learning_rate": 2.884765670836169e-05, "loss": 0.4761, "step": 7555 }, { "epoch": 0.18, "learning_rate": 2.8846776653915657e-05, "loss": 0.1782, "step": 7560 }, { "epoch": 0.18, "learning_rate": 2.884589659946962e-05, "loss": 0.2257, "step": 7565 }, { "epoch": 0.18, "learning_rate": 2.8845016545023587e-05, "loss": 0.2724, "step": 7570 }, { "epoch": 0.18, "learning_rate": 2.884413649057755e-05, "loss": 0.3765, "step": 7575 }, { "epoch": 0.18, "learning_rate": 2.8843256436131517e-05, "loss": 0.3085, "step": 7580 }, { "epoch": 0.18, "learning_rate": 2.8842376381685482e-05, "loss": 0.3637, "step": 7585 }, { "epoch": 0.18, "learning_rate": 2.8841496327239447e-05, "loss": 0.4292, "step": 7590 }, { "epoch": 0.18, "learning_rate": 2.8840616272793412e-05, "loss": 0.4293, "step": 7595 }, { "epoch": 0.18, "learning_rate": 2.8839736218347373e-05, "loss": 1.1, "step": 7600 }, { "epoch": 0.18, "learning_rate": 2.8838856163901342e-05, "loss": 0.3377, "step": 7605 }, { "epoch": 0.18, "learning_rate": 2.8837976109455307e-05, "loss": 0.1744, "step": 7610 }, { "epoch": 0.18, "learning_rate": 2.8837096055009272e-05, "loss": 0.1345, "step": 7615 }, { "epoch": 0.18, "learning_rate": 2.8836216000563233e-05, "loss": 0.1276, "step": 7620 }, { "epoch": 0.18, "learning_rate": 2.8835335946117202e-05, "loss": 0.2472, "step": 7625 }, { "epoch": 0.18, "learning_rate": 2.8834455891671167e-05, "loss": 0.2399, "step": 7630 }, { "epoch": 0.18, "learning_rate": 2.883357583722513e-05, "loss": 0.3345, "step": 7635 }, { "epoch": 0.18, "learning_rate": 2.8832695782779094e-05, "loss": 0.3952, "step": 7640 }, { "epoch": 0.18, "learning_rate": 2.8831815728333062e-05, "loss": 0.6301, "step": 7645 }, { "epoch": 0.18, "learning_rate": 2.8830935673887027e-05, "loss": 0.4979, "step": 7650 }, { "epoch": 0.18, "learning_rate": 2.883005561944099e-05, "loss": 0.3019, "step": 7655 }, { "epoch": 0.18, "learning_rate": 2.8829175564994954e-05, "loss": 0.3542, "step": 7660 }, { "epoch": 0.18, "learning_rate": 2.8828295510548922e-05, "loss": 0.2465, "step": 7665 }, { "epoch": 0.18, "learning_rate": 2.8827415456102884e-05, "loss": 0.2357, "step": 7670 }, { "epoch": 0.18, "learning_rate": 2.882653540165685e-05, "loss": 0.4287, "step": 7675 }, { "epoch": 0.18, "learning_rate": 2.8825655347210814e-05, "loss": 0.4347, "step": 7680 }, { "epoch": 0.18, "learning_rate": 2.8824775292764782e-05, "loss": 0.4421, "step": 7685 }, { "epoch": 0.18, "learning_rate": 2.8823895238318744e-05, "loss": 0.5263, "step": 7690 }, { "epoch": 0.18, "learning_rate": 2.882301518387271e-05, "loss": 0.698, "step": 7695 }, { "epoch": 0.18, "learning_rate": 2.8822135129426677e-05, "loss": 0.737, "step": 7700 }, { "epoch": 0.18, "learning_rate": 2.882125507498064e-05, "loss": 0.3011, "step": 7705 }, { "epoch": 0.18, "learning_rate": 2.8820375020534604e-05, "loss": 0.2666, "step": 7710 }, { "epoch": 0.18, "learning_rate": 2.881949496608857e-05, "loss": 0.3008, "step": 7715 }, { "epoch": 0.18, "learning_rate": 2.8818614911642537e-05, "loss": 0.244, "step": 7720 }, { "epoch": 0.18, "learning_rate": 2.88177348571965e-05, "loss": 0.2017, "step": 7725 }, { "epoch": 0.18, "learning_rate": 2.8816854802750464e-05, "loss": 0.2987, "step": 7730 }, { "epoch": 0.18, "learning_rate": 2.881597474830443e-05, "loss": 0.4226, "step": 7735 }, { "epoch": 0.18, "learning_rate": 2.8815094693858394e-05, "loss": 0.5962, "step": 7740 }, { "epoch": 0.18, "learning_rate": 2.881421463941236e-05, "loss": 0.4816, "step": 7745 }, { "epoch": 0.18, "learning_rate": 2.8813334584966324e-05, "loss": 0.7418, "step": 7750 }, { "epoch": 0.18, "learning_rate": 2.881245453052029e-05, "loss": 0.3677, "step": 7755 }, { "epoch": 0.18, "learning_rate": 2.8811574476074254e-05, "loss": 0.1755, "step": 7760 }, { "epoch": 0.18, "learning_rate": 2.881069442162822e-05, "loss": 0.1156, "step": 7765 }, { "epoch": 0.18, "learning_rate": 2.8809814367182184e-05, "loss": 0.3514, "step": 7770 }, { "epoch": 0.18, "learning_rate": 2.8808934312736146e-05, "loss": 0.2421, "step": 7775 }, { "epoch": 0.18, "learning_rate": 2.8808054258290114e-05, "loss": 0.3947, "step": 7780 }, { "epoch": 0.18, "learning_rate": 2.880717420384408e-05, "loss": 0.3647, "step": 7785 }, { "epoch": 0.18, "learning_rate": 2.8806294149398044e-05, "loss": 0.3566, "step": 7790 }, { "epoch": 0.18, "learning_rate": 2.8805414094952006e-05, "loss": 0.6717, "step": 7795 }, { "epoch": 0.18, "learning_rate": 2.8804534040505974e-05, "loss": 0.8482, "step": 7800 }, { "epoch": 0.18, "learning_rate": 2.880365398605994e-05, "loss": 0.4998, "step": 7805 }, { "epoch": 0.18, "learning_rate": 2.88027739316139e-05, "loss": 0.1632, "step": 7810 }, { "epoch": 0.18, "learning_rate": 2.8801893877167866e-05, "loss": 0.2797, "step": 7815 }, { "epoch": 0.18, "learning_rate": 2.8801013822721834e-05, "loss": 0.1865, "step": 7820 }, { "epoch": 0.18, "learning_rate": 2.88001337682758e-05, "loss": 0.2773, "step": 7825 }, { "epoch": 0.18, "learning_rate": 2.879925371382976e-05, "loss": 0.4942, "step": 7830 }, { "epoch": 0.18, "learning_rate": 2.879837365938373e-05, "loss": 0.2921, "step": 7835 }, { "epoch": 0.18, "learning_rate": 2.8797493604937694e-05, "loss": 0.551, "step": 7840 }, { "epoch": 0.18, "learning_rate": 2.879661355049166e-05, "loss": 0.4688, "step": 7845 }, { "epoch": 0.18, "learning_rate": 2.879573349604562e-05, "loss": 0.7619, "step": 7850 }, { "epoch": 0.18, "learning_rate": 2.879485344159959e-05, "loss": 0.3558, "step": 7855 }, { "epoch": 0.18, "learning_rate": 2.8793973387153554e-05, "loss": 0.1868, "step": 7860 }, { "epoch": 0.18, "learning_rate": 2.8793093332707516e-05, "loss": 0.1805, "step": 7865 }, { "epoch": 0.18, "learning_rate": 2.879221327826148e-05, "loss": 0.2113, "step": 7870 }, { "epoch": 0.18, "learning_rate": 2.879133322381545e-05, "loss": 0.2636, "step": 7875 }, { "epoch": 0.18, "learning_rate": 2.8790453169369415e-05, "loss": 0.4004, "step": 7880 }, { "epoch": 0.18, "learning_rate": 2.8789573114923376e-05, "loss": 0.4812, "step": 7885 }, { "epoch": 0.18, "learning_rate": 2.878869306047734e-05, "loss": 0.2949, "step": 7890 }, { "epoch": 0.18, "learning_rate": 2.878781300603131e-05, "loss": 0.5413, "step": 7895 }, { "epoch": 0.18, "learning_rate": 2.878693295158527e-05, "loss": 0.7284, "step": 7900 }, { "epoch": 0.18, "learning_rate": 2.8786052897139236e-05, "loss": 0.2792, "step": 7905 }, { "epoch": 0.18, "learning_rate": 2.87851728426932e-05, "loss": 0.2306, "step": 7910 }, { "epoch": 0.18, "learning_rate": 2.878429278824717e-05, "loss": 0.2398, "step": 7915 }, { "epoch": 0.18, "learning_rate": 2.878341273380113e-05, "loss": 0.269, "step": 7920 }, { "epoch": 0.18, "learning_rate": 2.8782532679355096e-05, "loss": 0.2446, "step": 7925 }, { "epoch": 0.19, "learning_rate": 2.878165262490906e-05, "loss": 0.3766, "step": 7930 }, { "epoch": 0.19, "learning_rate": 2.8780772570463026e-05, "loss": 0.3046, "step": 7935 }, { "epoch": 0.19, "learning_rate": 2.877989251601699e-05, "loss": 0.3851, "step": 7940 }, { "epoch": 0.19, "learning_rate": 2.8779012461570956e-05, "loss": 0.7176, "step": 7945 }, { "epoch": 0.19, "learning_rate": 2.877813240712492e-05, "loss": 0.4733, "step": 7950 }, { "epoch": 0.19, "learning_rate": 2.8777252352678886e-05, "loss": 0.3702, "step": 7955 }, { "epoch": 0.19, "learning_rate": 2.877637229823285e-05, "loss": 0.1659, "step": 7960 }, { "epoch": 0.19, "learning_rate": 2.8775492243786816e-05, "loss": 0.2927, "step": 7965 }, { "epoch": 0.19, "learning_rate": 2.877461218934078e-05, "loss": 0.4526, "step": 7970 }, { "epoch": 0.19, "learning_rate": 2.8773732134894747e-05, "loss": 0.3992, "step": 7975 }, { "epoch": 0.19, "learning_rate": 2.877285208044871e-05, "loss": 0.3524, "step": 7980 }, { "epoch": 0.19, "learning_rate": 2.8771972026002677e-05, "loss": 0.4235, "step": 7985 }, { "epoch": 0.19, "learning_rate": 2.877109197155664e-05, "loss": 0.4723, "step": 7990 }, { "epoch": 0.19, "learning_rate": 2.8770211917110607e-05, "loss": 0.5849, "step": 7995 }, { "epoch": 0.19, "learning_rate": 2.876933186266457e-05, "loss": 0.6152, "step": 8000 }, { "epoch": 0.19, "learning_rate": 2.8768451808218533e-05, "loss": 0.334, "step": 8005 }, { "epoch": 0.19, "learning_rate": 2.87675717537725e-05, "loss": 0.1424, "step": 8010 }, { "epoch": 0.19, "learning_rate": 2.8766691699326467e-05, "loss": 0.1184, "step": 8015 }, { "epoch": 0.19, "learning_rate": 2.8765811644880432e-05, "loss": 0.2609, "step": 8020 }, { "epoch": 0.19, "learning_rate": 2.8764931590434393e-05, "loss": 0.3052, "step": 8025 }, { "epoch": 0.19, "learning_rate": 2.8764051535988362e-05, "loss": 0.3251, "step": 8030 }, { "epoch": 0.19, "learning_rate": 2.8763171481542327e-05, "loss": 0.267, "step": 8035 }, { "epoch": 0.19, "learning_rate": 2.876229142709629e-05, "loss": 0.3947, "step": 8040 }, { "epoch": 0.19, "learning_rate": 2.8761411372650253e-05, "loss": 0.5264, "step": 8045 }, { "epoch": 0.19, "learning_rate": 2.8760531318204222e-05, "loss": 0.6026, "step": 8050 }, { "epoch": 0.19, "learning_rate": 2.8759651263758187e-05, "loss": 0.359, "step": 8055 }, { "epoch": 0.19, "learning_rate": 2.875877120931215e-05, "loss": 0.1941, "step": 8060 }, { "epoch": 0.19, "learning_rate": 2.8757891154866114e-05, "loss": 0.3357, "step": 8065 }, { "epoch": 0.19, "learning_rate": 2.8757011100420082e-05, "loss": 0.2425, "step": 8070 }, { "epoch": 0.19, "learning_rate": 2.8756131045974044e-05, "loss": 0.2236, "step": 8075 }, { "epoch": 0.19, "learning_rate": 2.875525099152801e-05, "loss": 0.2731, "step": 8080 }, { "epoch": 0.19, "learning_rate": 2.8754370937081974e-05, "loss": 0.3181, "step": 8085 }, { "epoch": 0.19, "learning_rate": 2.8753490882635942e-05, "loss": 0.4632, "step": 8090 }, { "epoch": 0.19, "learning_rate": 2.8752610828189904e-05, "loss": 0.3486, "step": 8095 }, { "epoch": 0.19, "learning_rate": 2.875173077374387e-05, "loss": 0.6905, "step": 8100 }, { "epoch": 0.19, "learning_rate": 2.8750850719297837e-05, "loss": 0.3986, "step": 8105 }, { "epoch": 0.19, "learning_rate": 2.87499706648518e-05, "loss": 0.2039, "step": 8110 }, { "epoch": 0.19, "learning_rate": 2.8749090610405764e-05, "loss": 0.1484, "step": 8115 }, { "epoch": 0.19, "learning_rate": 2.874821055595973e-05, "loss": 0.2117, "step": 8120 }, { "epoch": 0.19, "learning_rate": 2.8747330501513697e-05, "loss": 0.2586, "step": 8125 }, { "epoch": 0.19, "learning_rate": 2.874645044706766e-05, "loss": 0.2347, "step": 8130 }, { "epoch": 0.19, "learning_rate": 2.8745570392621624e-05, "loss": 0.463, "step": 8135 }, { "epoch": 0.19, "learning_rate": 2.874469033817559e-05, "loss": 0.565, "step": 8140 }, { "epoch": 0.19, "learning_rate": 2.8743810283729557e-05, "loss": 0.6461, "step": 8145 }, { "epoch": 0.19, "learning_rate": 2.874293022928352e-05, "loss": 0.637, "step": 8150 }, { "epoch": 0.19, "learning_rate": 2.8742050174837484e-05, "loss": 0.4174, "step": 8155 }, { "epoch": 0.19, "learning_rate": 2.874117012039145e-05, "loss": 0.2191, "step": 8160 }, { "epoch": 0.19, "learning_rate": 2.8740290065945414e-05, "loss": 0.2457, "step": 8165 }, { "epoch": 0.19, "learning_rate": 2.873941001149938e-05, "loss": 0.2393, "step": 8170 }, { "epoch": 0.19, "learning_rate": 2.8738529957053344e-05, "loss": 0.2471, "step": 8175 }, { "epoch": 0.19, "learning_rate": 2.873764990260731e-05, "loss": 0.3572, "step": 8180 }, { "epoch": 0.19, "learning_rate": 2.8736769848161274e-05, "loss": 0.4312, "step": 8185 }, { "epoch": 0.19, "learning_rate": 2.873588979371524e-05, "loss": 0.2429, "step": 8190 }, { "epoch": 0.19, "learning_rate": 2.8735009739269204e-05, "loss": 0.4353, "step": 8195 }, { "epoch": 0.19, "learning_rate": 2.8734129684823166e-05, "loss": 0.67, "step": 8200 }, { "epoch": 0.19, "learning_rate": 2.8733249630377134e-05, "loss": 0.2734, "step": 8205 }, { "epoch": 0.19, "learning_rate": 2.87323695759311e-05, "loss": 0.1687, "step": 8210 }, { "epoch": 0.19, "learning_rate": 2.8731489521485064e-05, "loss": 0.1638, "step": 8215 }, { "epoch": 0.19, "learning_rate": 2.8730609467039026e-05, "loss": 0.2174, "step": 8220 }, { "epoch": 0.19, "learning_rate": 2.8729729412592994e-05, "loss": 0.2065, "step": 8225 }, { "epoch": 0.19, "learning_rate": 2.872884935814696e-05, "loss": 0.4891, "step": 8230 }, { "epoch": 0.19, "learning_rate": 2.872796930370092e-05, "loss": 0.3173, "step": 8235 }, { "epoch": 0.19, "learning_rate": 2.872708924925489e-05, "loss": 0.6282, "step": 8240 }, { "epoch": 0.19, "learning_rate": 2.8726209194808854e-05, "loss": 0.3953, "step": 8245 }, { "epoch": 0.19, "learning_rate": 2.872532914036282e-05, "loss": 0.6517, "step": 8250 }, { "epoch": 0.19, "learning_rate": 2.872444908591678e-05, "loss": 0.4423, "step": 8255 }, { "epoch": 0.19, "learning_rate": 2.872356903147075e-05, "loss": 0.1846, "step": 8260 }, { "epoch": 0.19, "learning_rate": 2.8722688977024714e-05, "loss": 0.2379, "step": 8265 }, { "epoch": 0.19, "learning_rate": 2.8721808922578676e-05, "loss": 0.3152, "step": 8270 }, { "epoch": 0.19, "learning_rate": 2.872092886813264e-05, "loss": 0.3257, "step": 8275 }, { "epoch": 0.19, "learning_rate": 2.872004881368661e-05, "loss": 0.2704, "step": 8280 }, { "epoch": 0.19, "learning_rate": 2.8719168759240574e-05, "loss": 0.3943, "step": 8285 }, { "epoch": 0.19, "learning_rate": 2.8718288704794536e-05, "loss": 0.3879, "step": 8290 }, { "epoch": 0.19, "learning_rate": 2.87174086503485e-05, "loss": 0.2792, "step": 8295 }, { "epoch": 0.19, "learning_rate": 2.871652859590247e-05, "loss": 0.8818, "step": 8300 }, { "epoch": 0.19, "learning_rate": 2.871564854145643e-05, "loss": 0.4355, "step": 8305 }, { "epoch": 0.19, "learning_rate": 2.8714768487010396e-05, "loss": 0.1168, "step": 8310 }, { "epoch": 0.19, "learning_rate": 2.871388843256436e-05, "loss": 0.2507, "step": 8315 }, { "epoch": 0.19, "learning_rate": 2.871300837811833e-05, "loss": 0.3483, "step": 8320 }, { "epoch": 0.19, "learning_rate": 2.871212832367229e-05, "loss": 0.2731, "step": 8325 }, { "epoch": 0.19, "learning_rate": 2.8711248269226256e-05, "loss": 0.2658, "step": 8330 }, { "epoch": 0.19, "learning_rate": 2.871036821478022e-05, "loss": 0.4266, "step": 8335 }, { "epoch": 0.19, "learning_rate": 2.8709488160334186e-05, "loss": 0.491, "step": 8340 }, { "epoch": 0.19, "learning_rate": 2.870860810588815e-05, "loss": 0.6497, "step": 8345 }, { "epoch": 0.19, "learning_rate": 2.8707728051442116e-05, "loss": 0.9916, "step": 8350 }, { "epoch": 0.19, "learning_rate": 2.870684799699608e-05, "loss": 0.2371, "step": 8355 }, { "epoch": 0.2, "learning_rate": 2.8705967942550046e-05, "loss": 0.1958, "step": 8360 }, { "epoch": 0.2, "learning_rate": 2.870508788810401e-05, "loss": 0.2105, "step": 8365 }, { "epoch": 0.2, "learning_rate": 2.8704207833657976e-05, "loss": 0.2231, "step": 8370 }, { "epoch": 0.2, "learning_rate": 2.870332777921194e-05, "loss": 0.231, "step": 8375 }, { "epoch": 0.2, "learning_rate": 2.8702447724765906e-05, "loss": 0.3574, "step": 8380 }, { "epoch": 0.2, "learning_rate": 2.870156767031987e-05, "loss": 0.4619, "step": 8385 }, { "epoch": 0.2, "learning_rate": 2.8700687615873836e-05, "loss": 0.2682, "step": 8390 }, { "epoch": 0.2, "learning_rate": 2.86998075614278e-05, "loss": 0.3319, "step": 8395 }, { "epoch": 0.2, "learning_rate": 2.8698927506981766e-05, "loss": 0.9142, "step": 8400 }, { "epoch": 0.2, "learning_rate": 2.869804745253573e-05, "loss": 0.3028, "step": 8405 }, { "epoch": 0.2, "learning_rate": 2.8697167398089693e-05, "loss": 0.1297, "step": 8410 }, { "epoch": 0.2, "learning_rate": 2.869628734364366e-05, "loss": 0.2081, "step": 8415 }, { "epoch": 0.2, "learning_rate": 2.8695407289197627e-05, "loss": 0.1481, "step": 8420 }, { "epoch": 0.2, "learning_rate": 2.869452723475159e-05, "loss": 0.241, "step": 8425 }, { "epoch": 0.2, "learning_rate": 2.8693647180305553e-05, "loss": 0.2859, "step": 8430 }, { "epoch": 0.2, "learning_rate": 2.869276712585952e-05, "loss": 0.4589, "step": 8435 }, { "epoch": 0.2, "learning_rate": 2.8691887071413487e-05, "loss": 0.3065, "step": 8440 }, { "epoch": 0.2, "learning_rate": 2.869100701696745e-05, "loss": 0.3523, "step": 8445 }, { "epoch": 0.2, "learning_rate": 2.8690126962521413e-05, "loss": 1.0486, "step": 8450 }, { "epoch": 0.2, "learning_rate": 2.8689246908075382e-05, "loss": 0.3201, "step": 8455 }, { "epoch": 0.2, "learning_rate": 2.8688366853629347e-05, "loss": 0.2494, "step": 8460 }, { "epoch": 0.2, "learning_rate": 2.868748679918331e-05, "loss": 0.2547, "step": 8465 }, { "epoch": 0.2, "learning_rate": 2.8686606744737273e-05, "loss": 0.2265, "step": 8470 }, { "epoch": 0.2, "learning_rate": 2.8685726690291242e-05, "loss": 0.2511, "step": 8475 }, { "epoch": 0.2, "learning_rate": 2.8684846635845207e-05, "loss": 0.3004, "step": 8480 }, { "epoch": 0.2, "learning_rate": 2.868396658139917e-05, "loss": 0.3762, "step": 8485 }, { "epoch": 0.2, "learning_rate": 2.8683086526953133e-05, "loss": 0.2286, "step": 8490 }, { "epoch": 0.2, "learning_rate": 2.8682206472507102e-05, "loss": 0.5977, "step": 8495 }, { "epoch": 0.2, "learning_rate": 2.8681326418061064e-05, "loss": 0.5691, "step": 8500 }, { "epoch": 0.2, "learning_rate": 2.868044636361503e-05, "loss": 0.3363, "step": 8505 }, { "epoch": 0.2, "learning_rate": 2.8679566309168994e-05, "loss": 0.2911, "step": 8510 }, { "epoch": 0.2, "learning_rate": 2.8678686254722962e-05, "loss": 0.2567, "step": 8515 }, { "epoch": 0.2, "learning_rate": 2.8677806200276924e-05, "loss": 0.1937, "step": 8520 }, { "epoch": 0.2, "learning_rate": 2.867692614583089e-05, "loss": 0.3368, "step": 8525 }, { "epoch": 0.2, "learning_rate": 2.8676046091384857e-05, "loss": 0.2906, "step": 8530 }, { "epoch": 0.2, "learning_rate": 2.867516603693882e-05, "loss": 0.453, "step": 8535 }, { "epoch": 0.2, "learning_rate": 2.8674285982492784e-05, "loss": 0.5084, "step": 8540 }, { "epoch": 0.2, "learning_rate": 2.867340592804675e-05, "loss": 0.8822, "step": 8545 }, { "epoch": 0.2, "learning_rate": 2.8672525873600717e-05, "loss": 0.545, "step": 8550 }, { "epoch": 0.2, "learning_rate": 2.867164581915468e-05, "loss": 0.4569, "step": 8555 }, { "epoch": 0.2, "learning_rate": 2.8670765764708644e-05, "loss": 0.2299, "step": 8560 }, { "epoch": 0.2, "learning_rate": 2.866988571026261e-05, "loss": 0.2065, "step": 8565 }, { "epoch": 0.2, "learning_rate": 2.8669005655816574e-05, "loss": 0.2024, "step": 8570 }, { "epoch": 0.2, "learning_rate": 2.866812560137054e-05, "loss": 0.2628, "step": 8575 }, { "epoch": 0.2, "learning_rate": 2.8667245546924504e-05, "loss": 0.222, "step": 8580 }, { "epoch": 0.2, "learning_rate": 2.866636549247847e-05, "loss": 0.353, "step": 8585 }, { "epoch": 0.2, "learning_rate": 2.8665485438032434e-05, "loss": 0.5118, "step": 8590 }, { "epoch": 0.2, "learning_rate": 2.86646053835864e-05, "loss": 0.4463, "step": 8595 }, { "epoch": 0.2, "learning_rate": 2.8663725329140364e-05, "loss": 0.6394, "step": 8600 }, { "epoch": 0.2, "learning_rate": 2.8662845274694326e-05, "loss": 0.2157, "step": 8605 }, { "epoch": 0.2, "learning_rate": 2.8661965220248294e-05, "loss": 0.1076, "step": 8610 }, { "epoch": 0.2, "learning_rate": 2.866108516580226e-05, "loss": 0.2304, "step": 8615 }, { "epoch": 0.2, "learning_rate": 2.8660205111356224e-05, "loss": 0.3553, "step": 8620 }, { "epoch": 0.2, "learning_rate": 2.8659325056910186e-05, "loss": 0.3783, "step": 8625 }, { "epoch": 0.2, "learning_rate": 2.8658445002464154e-05, "loss": 0.2744, "step": 8630 }, { "epoch": 0.2, "learning_rate": 2.865756494801812e-05, "loss": 0.3668, "step": 8635 }, { "epoch": 0.2, "learning_rate": 2.865668489357208e-05, "loss": 0.4139, "step": 8640 }, { "epoch": 0.2, "learning_rate": 2.865580483912605e-05, "loss": 0.3601, "step": 8645 }, { "epoch": 0.2, "learning_rate": 2.8654924784680014e-05, "loss": 0.8572, "step": 8650 }, { "epoch": 0.2, "learning_rate": 2.865404473023398e-05, "loss": 0.3885, "step": 8655 }, { "epoch": 0.2, "learning_rate": 2.865316467578794e-05, "loss": 0.1424, "step": 8660 }, { "epoch": 0.2, "learning_rate": 2.865228462134191e-05, "loss": 0.2, "step": 8665 }, { "epoch": 0.2, "learning_rate": 2.8651404566895874e-05, "loss": 0.2135, "step": 8670 }, { "epoch": 0.2, "learning_rate": 2.8650524512449836e-05, "loss": 0.3693, "step": 8675 }, { "epoch": 0.2, "learning_rate": 2.86496444580038e-05, "loss": 0.3017, "step": 8680 }, { "epoch": 0.2, "learning_rate": 2.864876440355777e-05, "loss": 0.2234, "step": 8685 }, { "epoch": 0.2, "learning_rate": 2.8647884349111734e-05, "loss": 0.4461, "step": 8690 }, { "epoch": 0.2, "learning_rate": 2.8647004294665696e-05, "loss": 0.5233, "step": 8695 }, { "epoch": 0.2, "learning_rate": 2.864612424021966e-05, "loss": 0.621, "step": 8700 }, { "epoch": 0.2, "learning_rate": 2.864524418577363e-05, "loss": 0.363, "step": 8705 }, { "epoch": 0.2, "learning_rate": 2.864436413132759e-05, "loss": 0.1145, "step": 8710 }, { "epoch": 0.2, "learning_rate": 2.8643484076881556e-05, "loss": 0.2051, "step": 8715 }, { "epoch": 0.2, "learning_rate": 2.864260402243552e-05, "loss": 0.2533, "step": 8720 }, { "epoch": 0.2, "learning_rate": 2.864172396798949e-05, "loss": 0.3678, "step": 8725 }, { "epoch": 0.2, "learning_rate": 2.864084391354345e-05, "loss": 0.3549, "step": 8730 }, { "epoch": 0.2, "learning_rate": 2.8639963859097416e-05, "loss": 0.5014, "step": 8735 }, { "epoch": 0.2, "learning_rate": 2.863908380465138e-05, "loss": 0.3496, "step": 8740 }, { "epoch": 0.2, "learning_rate": 2.863820375020535e-05, "loss": 0.3657, "step": 8745 }, { "epoch": 0.2, "learning_rate": 2.863732369575931e-05, "loss": 0.5528, "step": 8750 }, { "epoch": 0.2, "learning_rate": 2.8636443641313276e-05, "loss": 0.3521, "step": 8755 }, { "epoch": 0.2, "learning_rate": 2.863556358686724e-05, "loss": 0.1606, "step": 8760 }, { "epoch": 0.2, "learning_rate": 2.8634683532421206e-05, "loss": 0.1677, "step": 8765 }, { "epoch": 0.2, "learning_rate": 2.863380347797517e-05, "loss": 0.1309, "step": 8770 }, { "epoch": 0.2, "learning_rate": 2.8632923423529136e-05, "loss": 0.3097, "step": 8775 }, { "epoch": 0.2, "learning_rate": 2.86320433690831e-05, "loss": 0.3174, "step": 8780 }, { "epoch": 0.2, "learning_rate": 2.8631163314637066e-05, "loss": 0.2955, "step": 8785 }, { "epoch": 0.21, "learning_rate": 2.863028326019103e-05, "loss": 0.3892, "step": 8790 }, { "epoch": 0.21, "learning_rate": 2.8629403205744996e-05, "loss": 0.4101, "step": 8795 }, { "epoch": 0.21, "learning_rate": 2.862852315129896e-05, "loss": 0.7524, "step": 8800 }, { "epoch": 0.21, "learning_rate": 2.8627643096852926e-05, "loss": 0.3373, "step": 8805 }, { "epoch": 0.21, "learning_rate": 2.862676304240689e-05, "loss": 0.1955, "step": 8810 }, { "epoch": 0.21, "learning_rate": 2.8625882987960856e-05, "loss": 0.2267, "step": 8815 }, { "epoch": 0.21, "learning_rate": 2.862500293351482e-05, "loss": 0.3109, "step": 8820 }, { "epoch": 0.21, "learning_rate": 2.8624122879068786e-05, "loss": 0.2435, "step": 8825 }, { "epoch": 0.21, "learning_rate": 2.862324282462275e-05, "loss": 0.3741, "step": 8830 }, { "epoch": 0.21, "learning_rate": 2.8622362770176713e-05, "loss": 0.3656, "step": 8835 }, { "epoch": 0.21, "learning_rate": 2.862148271573068e-05, "loss": 0.3214, "step": 8840 }, { "epoch": 0.21, "learning_rate": 2.8620602661284647e-05, "loss": 0.4084, "step": 8845 }, { "epoch": 0.21, "learning_rate": 2.861972260683861e-05, "loss": 1.0186, "step": 8850 }, { "epoch": 0.21, "learning_rate": 2.8618842552392573e-05, "loss": 0.5085, "step": 8855 }, { "epoch": 0.21, "learning_rate": 2.861796249794654e-05, "loss": 0.2107, "step": 8860 }, { "epoch": 0.21, "learning_rate": 2.8617082443500507e-05, "loss": 0.2605, "step": 8865 }, { "epoch": 0.21, "learning_rate": 2.8616202389054468e-05, "loss": 0.2647, "step": 8870 }, { "epoch": 0.21, "learning_rate": 2.8615322334608433e-05, "loss": 0.3274, "step": 8875 }, { "epoch": 0.21, "learning_rate": 2.86144422801624e-05, "loss": 0.3471, "step": 8880 }, { "epoch": 0.21, "learning_rate": 2.8613562225716367e-05, "loss": 0.2903, "step": 8885 }, { "epoch": 0.21, "learning_rate": 2.861268217127033e-05, "loss": 0.3901, "step": 8890 }, { "epoch": 0.21, "learning_rate": 2.8611802116824293e-05, "loss": 0.6301, "step": 8895 }, { "epoch": 0.21, "learning_rate": 2.8610922062378262e-05, "loss": 0.7889, "step": 8900 }, { "epoch": 0.21, "learning_rate": 2.8610042007932223e-05, "loss": 0.3984, "step": 8905 }, { "epoch": 0.21, "learning_rate": 2.860916195348619e-05, "loss": 0.3034, "step": 8910 }, { "epoch": 0.21, "learning_rate": 2.8608281899040153e-05, "loss": 0.2821, "step": 8915 }, { "epoch": 0.21, "learning_rate": 2.8607401844594122e-05, "loss": 0.2333, "step": 8920 }, { "epoch": 0.21, "learning_rate": 2.8606521790148083e-05, "loss": 0.273, "step": 8925 }, { "epoch": 0.21, "learning_rate": 2.860564173570205e-05, "loss": 0.2452, "step": 8930 }, { "epoch": 0.21, "learning_rate": 2.8604761681256017e-05, "loss": 0.4131, "step": 8935 }, { "epoch": 0.21, "learning_rate": 2.860388162680998e-05, "loss": 0.2558, "step": 8940 }, { "epoch": 0.21, "learning_rate": 2.8603001572363944e-05, "loss": 0.4228, "step": 8945 }, { "epoch": 0.21, "learning_rate": 2.860212151791791e-05, "loss": 0.7583, "step": 8950 }, { "epoch": 0.21, "learning_rate": 2.8601241463471877e-05, "loss": 0.305, "step": 8955 }, { "epoch": 0.21, "learning_rate": 2.860036140902584e-05, "loss": 0.1276, "step": 8960 }, { "epoch": 0.21, "learning_rate": 2.8599481354579804e-05, "loss": 0.1865, "step": 8965 }, { "epoch": 0.21, "learning_rate": 2.859860130013377e-05, "loss": 0.1911, "step": 8970 }, { "epoch": 0.21, "learning_rate": 2.8597721245687734e-05, "loss": 0.2196, "step": 8975 }, { "epoch": 0.21, "learning_rate": 2.85968411912417e-05, "loss": 0.2271, "step": 8980 }, { "epoch": 0.21, "learning_rate": 2.8595961136795664e-05, "loss": 0.3531, "step": 8985 }, { "epoch": 0.21, "learning_rate": 2.859508108234963e-05, "loss": 0.3905, "step": 8990 }, { "epoch": 0.21, "learning_rate": 2.8594201027903594e-05, "loss": 0.3622, "step": 8995 }, { "epoch": 0.21, "learning_rate": 2.859332097345756e-05, "loss": 0.5169, "step": 9000 }, { "epoch": 0.21, "learning_rate": 2.8592440919011524e-05, "loss": 0.4006, "step": 9005 }, { "epoch": 0.21, "learning_rate": 2.8591560864565485e-05, "loss": 0.1653, "step": 9010 }, { "epoch": 0.21, "learning_rate": 2.8590680810119454e-05, "loss": 0.2325, "step": 9015 }, { "epoch": 0.21, "learning_rate": 2.858980075567342e-05, "loss": 0.2366, "step": 9020 }, { "epoch": 0.21, "learning_rate": 2.8588920701227384e-05, "loss": 0.388, "step": 9025 }, { "epoch": 0.21, "learning_rate": 2.8588040646781346e-05, "loss": 0.321, "step": 9030 }, { "epoch": 0.21, "learning_rate": 2.8587160592335314e-05, "loss": 0.2476, "step": 9035 }, { "epoch": 0.21, "learning_rate": 2.858628053788928e-05, "loss": 0.3687, "step": 9040 }, { "epoch": 0.21, "learning_rate": 2.858540048344324e-05, "loss": 0.3963, "step": 9045 }, { "epoch": 0.21, "learning_rate": 2.858452042899721e-05, "loss": 0.9041, "step": 9050 }, { "epoch": 0.21, "learning_rate": 2.8583640374551174e-05, "loss": 0.2883, "step": 9055 }, { "epoch": 0.21, "learning_rate": 2.858276032010514e-05, "loss": 0.1362, "step": 9060 }, { "epoch": 0.21, "learning_rate": 2.85818802656591e-05, "loss": 0.212, "step": 9065 }, { "epoch": 0.21, "learning_rate": 2.858100021121307e-05, "loss": 0.2415, "step": 9070 }, { "epoch": 0.21, "learning_rate": 2.8580120156767034e-05, "loss": 0.2539, "step": 9075 }, { "epoch": 0.21, "learning_rate": 2.8579240102321e-05, "loss": 0.3506, "step": 9080 }, { "epoch": 0.21, "learning_rate": 2.857836004787496e-05, "loss": 0.2253, "step": 9085 }, { "epoch": 0.21, "learning_rate": 2.857747999342893e-05, "loss": 0.4342, "step": 9090 }, { "epoch": 0.21, "learning_rate": 2.8576599938982894e-05, "loss": 0.5014, "step": 9095 }, { "epoch": 0.21, "learning_rate": 2.8575719884536856e-05, "loss": 0.7555, "step": 9100 }, { "epoch": 0.21, "learning_rate": 2.857483983009082e-05, "loss": 0.2847, "step": 9105 }, { "epoch": 0.21, "learning_rate": 2.857395977564479e-05, "loss": 0.0715, "step": 9110 }, { "epoch": 0.21, "learning_rate": 2.8573079721198754e-05, "loss": 0.1755, "step": 9115 }, { "epoch": 0.21, "learning_rate": 2.8572199666752716e-05, "loss": 0.1859, "step": 9120 }, { "epoch": 0.21, "learning_rate": 2.857131961230668e-05, "loss": 0.2235, "step": 9125 }, { "epoch": 0.21, "learning_rate": 2.857043955786065e-05, "loss": 0.2349, "step": 9130 }, { "epoch": 0.21, "learning_rate": 2.856955950341461e-05, "loss": 0.2425, "step": 9135 }, { "epoch": 0.21, "learning_rate": 2.8568679448968576e-05, "loss": 0.4019, "step": 9140 }, { "epoch": 0.21, "learning_rate": 2.856779939452254e-05, "loss": 0.4464, "step": 9145 }, { "epoch": 0.21, "learning_rate": 2.856691934007651e-05, "loss": 0.6406, "step": 9150 }, { "epoch": 0.21, "learning_rate": 2.856603928563047e-05, "loss": 0.3269, "step": 9155 }, { "epoch": 0.21, "learning_rate": 2.8565159231184436e-05, "loss": 0.1601, "step": 9160 }, { "epoch": 0.21, "learning_rate": 2.85642791767384e-05, "loss": 0.1677, "step": 9165 }, { "epoch": 0.21, "learning_rate": 2.8563399122292366e-05, "loss": 0.3127, "step": 9170 }, { "epoch": 0.21, "learning_rate": 2.856251906784633e-05, "loss": 0.2425, "step": 9175 }, { "epoch": 0.21, "learning_rate": 2.8561639013400296e-05, "loss": 0.2728, "step": 9180 }, { "epoch": 0.21, "learning_rate": 2.856075895895426e-05, "loss": 0.3161, "step": 9185 }, { "epoch": 0.21, "learning_rate": 2.8559878904508226e-05, "loss": 0.6597, "step": 9190 }, { "epoch": 0.21, "learning_rate": 2.855899885006219e-05, "loss": 0.5318, "step": 9195 }, { "epoch": 0.21, "learning_rate": 2.8558118795616156e-05, "loss": 0.7502, "step": 9200 }, { "epoch": 0.21, "learning_rate": 2.855723874117012e-05, "loss": 0.2542, "step": 9205 }, { "epoch": 0.21, "learning_rate": 2.8556358686724086e-05, "loss": 0.1525, "step": 9210 }, { "epoch": 0.21, "learning_rate": 2.855547863227805e-05, "loss": 0.1678, "step": 9215 }, { "epoch": 0.22, "learning_rate": 2.8554598577832016e-05, "loss": 0.3163, "step": 9220 }, { "epoch": 0.22, "learning_rate": 2.855371852338598e-05, "loss": 0.2587, "step": 9225 }, { "epoch": 0.22, "learning_rate": 2.8552838468939946e-05, "loss": 0.3315, "step": 9230 }, { "epoch": 0.22, "learning_rate": 2.855195841449391e-05, "loss": 0.3994, "step": 9235 }, { "epoch": 0.22, "learning_rate": 2.8551078360047873e-05, "loss": 0.6105, "step": 9240 }, { "epoch": 0.22, "learning_rate": 2.855019830560184e-05, "loss": 0.52, "step": 9245 }, { "epoch": 0.22, "learning_rate": 2.8549318251155806e-05, "loss": 0.8011, "step": 9250 }, { "epoch": 0.22, "learning_rate": 2.854843819670977e-05, "loss": 0.4434, "step": 9255 }, { "epoch": 0.22, "learning_rate": 2.8547558142263733e-05, "loss": 0.1215, "step": 9260 }, { "epoch": 0.22, "learning_rate": 2.85466780878177e-05, "loss": 0.1905, "step": 9265 }, { "epoch": 0.22, "learning_rate": 2.8545798033371667e-05, "loss": 0.2887, "step": 9270 }, { "epoch": 0.22, "learning_rate": 2.8544917978925628e-05, "loss": 0.3101, "step": 9275 }, { "epoch": 0.22, "learning_rate": 2.8544037924479593e-05, "loss": 0.2305, "step": 9280 }, { "epoch": 0.22, "learning_rate": 2.854315787003356e-05, "loss": 0.4263, "step": 9285 }, { "epoch": 0.22, "learning_rate": 2.8542277815587527e-05, "loss": 0.3077, "step": 9290 }, { "epoch": 0.22, "learning_rate": 2.8541397761141488e-05, "loss": 0.3986, "step": 9295 }, { "epoch": 0.22, "learning_rate": 2.8540517706695453e-05, "loss": 0.5493, "step": 9300 }, { "epoch": 0.22, "learning_rate": 2.853963765224942e-05, "loss": 0.2731, "step": 9305 }, { "epoch": 0.22, "learning_rate": 2.8538757597803383e-05, "loss": 0.288, "step": 9310 }, { "epoch": 0.22, "learning_rate": 2.8537877543357348e-05, "loss": 0.1788, "step": 9315 }, { "epoch": 0.22, "learning_rate": 2.8536997488911313e-05, "loss": 0.2808, "step": 9320 }, { "epoch": 0.22, "learning_rate": 2.8536117434465282e-05, "loss": 0.273, "step": 9325 }, { "epoch": 0.22, "learning_rate": 2.8535237380019243e-05, "loss": 0.4623, "step": 9330 }, { "epoch": 0.22, "learning_rate": 2.853435732557321e-05, "loss": 0.4471, "step": 9335 }, { "epoch": 0.22, "learning_rate": 2.8533477271127177e-05, "loss": 0.4588, "step": 9340 }, { "epoch": 0.22, "learning_rate": 2.853259721668114e-05, "loss": 0.5298, "step": 9345 }, { "epoch": 0.22, "learning_rate": 2.8531717162235103e-05, "loss": 0.6504, "step": 9350 }, { "epoch": 0.22, "learning_rate": 2.853083710778907e-05, "loss": 0.4439, "step": 9355 }, { "epoch": 0.22, "learning_rate": 2.8529957053343037e-05, "loss": 0.1707, "step": 9360 }, { "epoch": 0.22, "learning_rate": 2.8529076998897e-05, "loss": 0.172, "step": 9365 }, { "epoch": 0.22, "learning_rate": 2.8528196944450964e-05, "loss": 0.2051, "step": 9370 }, { "epoch": 0.22, "learning_rate": 2.852731689000493e-05, "loss": 0.1468, "step": 9375 }, { "epoch": 0.22, "learning_rate": 2.8526436835558897e-05, "loss": 0.2887, "step": 9380 }, { "epoch": 0.22, "learning_rate": 2.852555678111286e-05, "loss": 0.5303, "step": 9385 }, { "epoch": 0.22, "learning_rate": 2.8524676726666824e-05, "loss": 0.5506, "step": 9390 }, { "epoch": 0.22, "learning_rate": 2.852379667222079e-05, "loss": 0.3631, "step": 9395 }, { "epoch": 0.22, "learning_rate": 2.8522916617774754e-05, "loss": 0.6084, "step": 9400 }, { "epoch": 0.22, "learning_rate": 2.852203656332872e-05, "loss": 0.3251, "step": 9405 }, { "epoch": 0.22, "learning_rate": 2.8521156508882684e-05, "loss": 0.1618, "step": 9410 }, { "epoch": 0.22, "learning_rate": 2.852027645443665e-05, "loss": 0.0754, "step": 9415 }, { "epoch": 0.22, "learning_rate": 2.8519396399990614e-05, "loss": 0.3048, "step": 9420 }, { "epoch": 0.22, "learning_rate": 2.851851634554458e-05, "loss": 0.2485, "step": 9425 }, { "epoch": 0.22, "learning_rate": 2.8517636291098544e-05, "loss": 0.1802, "step": 9430 }, { "epoch": 0.22, "learning_rate": 2.8516756236652505e-05, "loss": 0.55, "step": 9435 }, { "epoch": 0.22, "learning_rate": 2.8515876182206474e-05, "loss": 0.334, "step": 9440 }, { "epoch": 0.22, "learning_rate": 2.851499612776044e-05, "loss": 0.5152, "step": 9445 }, { "epoch": 0.22, "learning_rate": 2.8514116073314404e-05, "loss": 0.8276, "step": 9450 }, { "epoch": 0.22, "learning_rate": 2.8513236018868365e-05, "loss": 0.3374, "step": 9455 }, { "epoch": 0.22, "learning_rate": 2.8512355964422334e-05, "loss": 0.1052, "step": 9460 }, { "epoch": 0.22, "learning_rate": 2.85114759099763e-05, "loss": 0.1647, "step": 9465 }, { "epoch": 0.22, "learning_rate": 2.851059585553026e-05, "loss": 0.2481, "step": 9470 }, { "epoch": 0.22, "learning_rate": 2.850971580108423e-05, "loss": 0.1806, "step": 9475 }, { "epoch": 0.22, "learning_rate": 2.8508835746638194e-05, "loss": 0.1987, "step": 9480 }, { "epoch": 0.22, "learning_rate": 2.850795569219216e-05, "loss": 0.2689, "step": 9485 }, { "epoch": 0.22, "learning_rate": 2.850707563774612e-05, "loss": 0.3172, "step": 9490 }, { "epoch": 0.22, "learning_rate": 2.850619558330009e-05, "loss": 0.5856, "step": 9495 }, { "epoch": 0.22, "learning_rate": 2.8505315528854054e-05, "loss": 0.6307, "step": 9500 }, { "epoch": 0.22, "learning_rate": 2.8504435474408016e-05, "loss": 0.3994, "step": 9505 }, { "epoch": 0.22, "learning_rate": 2.850355541996198e-05, "loss": 0.2171, "step": 9510 }, { "epoch": 0.22, "learning_rate": 2.850267536551595e-05, "loss": 0.13, "step": 9515 }, { "epoch": 0.22, "learning_rate": 2.8501795311069914e-05, "loss": 0.2336, "step": 9520 }, { "epoch": 0.22, "learning_rate": 2.8500915256623876e-05, "loss": 0.1927, "step": 9525 }, { "epoch": 0.22, "learning_rate": 2.850003520217784e-05, "loss": 0.3315, "step": 9530 }, { "epoch": 0.22, "learning_rate": 2.849915514773181e-05, "loss": 0.1755, "step": 9535 }, { "epoch": 0.22, "learning_rate": 2.849827509328577e-05, "loss": 0.3137, "step": 9540 }, { "epoch": 0.22, "learning_rate": 2.8497395038839736e-05, "loss": 0.4314, "step": 9545 }, { "epoch": 0.22, "learning_rate": 2.84965149843937e-05, "loss": 0.6896, "step": 9550 }, { "epoch": 0.22, "learning_rate": 2.849563492994767e-05, "loss": 0.2763, "step": 9555 }, { "epoch": 0.22, "learning_rate": 2.849475487550163e-05, "loss": 0.1256, "step": 9560 }, { "epoch": 0.22, "learning_rate": 2.8493874821055596e-05, "loss": 0.1985, "step": 9565 }, { "epoch": 0.22, "learning_rate": 2.849299476660956e-05, "loss": 0.2654, "step": 9570 }, { "epoch": 0.22, "learning_rate": 2.8492114712163526e-05, "loss": 0.2243, "step": 9575 }, { "epoch": 0.22, "learning_rate": 2.849123465771749e-05, "loss": 0.3792, "step": 9580 }, { "epoch": 0.22, "learning_rate": 2.8490354603271456e-05, "loss": 0.4242, "step": 9585 }, { "epoch": 0.22, "learning_rate": 2.848947454882542e-05, "loss": 0.396, "step": 9590 }, { "epoch": 0.22, "learning_rate": 2.8488594494379386e-05, "loss": 0.3941, "step": 9595 }, { "epoch": 0.22, "learning_rate": 2.848771443993335e-05, "loss": 0.8239, "step": 9600 }, { "epoch": 0.22, "learning_rate": 2.8486834385487316e-05, "loss": 0.4185, "step": 9605 }, { "epoch": 0.22, "learning_rate": 2.848595433104128e-05, "loss": 0.1234, "step": 9610 }, { "epoch": 0.22, "learning_rate": 2.8485074276595246e-05, "loss": 0.1713, "step": 9615 }, { "epoch": 0.22, "learning_rate": 2.848419422214921e-05, "loss": 0.1898, "step": 9620 }, { "epoch": 0.22, "learning_rate": 2.8483314167703176e-05, "loss": 0.1252, "step": 9625 }, { "epoch": 0.22, "learning_rate": 2.848243411325714e-05, "loss": 0.2736, "step": 9630 }, { "epoch": 0.22, "learning_rate": 2.8481554058811106e-05, "loss": 0.4337, "step": 9635 }, { "epoch": 0.22, "learning_rate": 2.848067400436507e-05, "loss": 0.4305, "step": 9640 }, { "epoch": 0.23, "learning_rate": 2.8479793949919033e-05, "loss": 0.5038, "step": 9645 }, { "epoch": 0.23, "learning_rate": 2.8478913895473e-05, "loss": 0.8309, "step": 9650 }, { "epoch": 0.23, "learning_rate": 2.8478033841026966e-05, "loss": 0.3041, "step": 9655 }, { "epoch": 0.23, "learning_rate": 2.847715378658093e-05, "loss": 0.1149, "step": 9660 }, { "epoch": 0.23, "learning_rate": 2.8476273732134893e-05, "loss": 0.19, "step": 9665 }, { "epoch": 0.23, "learning_rate": 2.847539367768886e-05, "loss": 0.3101, "step": 9670 }, { "epoch": 0.23, "learning_rate": 2.8474513623242826e-05, "loss": 0.1929, "step": 9675 }, { "epoch": 0.23, "learning_rate": 2.847363356879679e-05, "loss": 0.3636, "step": 9680 }, { "epoch": 0.23, "learning_rate": 2.8472753514350753e-05, "loss": 0.2324, "step": 9685 }, { "epoch": 0.23, "learning_rate": 2.847187345990472e-05, "loss": 0.5133, "step": 9690 }, { "epoch": 0.23, "learning_rate": 2.8470993405458686e-05, "loss": 0.4251, "step": 9695 }, { "epoch": 0.23, "learning_rate": 2.8470113351012648e-05, "loss": 0.8432, "step": 9700 }, { "epoch": 0.23, "learning_rate": 2.8469233296566613e-05, "loss": 0.3864, "step": 9705 }, { "epoch": 0.23, "learning_rate": 2.846835324212058e-05, "loss": 0.1077, "step": 9710 }, { "epoch": 0.23, "learning_rate": 2.8467473187674547e-05, "loss": 0.288, "step": 9715 }, { "epoch": 0.23, "learning_rate": 2.8466593133228508e-05, "loss": 0.227, "step": 9720 }, { "epoch": 0.23, "learning_rate": 2.8465713078782473e-05, "loss": 0.1981, "step": 9725 }, { "epoch": 0.23, "learning_rate": 2.846483302433644e-05, "loss": 0.2953, "step": 9730 }, { "epoch": 0.23, "learning_rate": 2.8463952969890403e-05, "loss": 0.5526, "step": 9735 }, { "epoch": 0.23, "learning_rate": 2.8463072915444368e-05, "loss": 0.4031, "step": 9740 }, { "epoch": 0.23, "learning_rate": 2.8462192860998337e-05, "loss": 0.4515, "step": 9745 }, { "epoch": 0.23, "learning_rate": 2.8461312806552302e-05, "loss": 0.7157, "step": 9750 }, { "epoch": 0.23, "learning_rate": 2.8460432752106263e-05, "loss": 0.4174, "step": 9755 }, { "epoch": 0.23, "learning_rate": 2.845955269766023e-05, "loss": 0.1863, "step": 9760 }, { "epoch": 0.23, "learning_rate": 2.8458672643214197e-05, "loss": 0.2207, "step": 9765 }, { "epoch": 0.23, "learning_rate": 2.845779258876816e-05, "loss": 0.1611, "step": 9770 }, { "epoch": 0.23, "learning_rate": 2.8456912534322123e-05, "loss": 0.2295, "step": 9775 }, { "epoch": 0.23, "learning_rate": 2.845603247987609e-05, "loss": 0.3251, "step": 9780 }, { "epoch": 0.23, "learning_rate": 2.8455152425430057e-05, "loss": 0.4119, "step": 9785 }, { "epoch": 0.23, "learning_rate": 2.845427237098402e-05, "loss": 0.3725, "step": 9790 }, { "epoch": 0.23, "learning_rate": 2.8453392316537983e-05, "loss": 0.5099, "step": 9795 }, { "epoch": 0.23, "learning_rate": 2.845251226209195e-05, "loss": 0.7575, "step": 9800 }, { "epoch": 0.23, "learning_rate": 2.8451632207645914e-05, "loss": 0.4596, "step": 9805 }, { "epoch": 0.23, "learning_rate": 2.845075215319988e-05, "loss": 0.2963, "step": 9810 }, { "epoch": 0.23, "learning_rate": 2.8449872098753844e-05, "loss": 0.1625, "step": 9815 }, { "epoch": 0.23, "learning_rate": 2.844899204430781e-05, "loss": 0.2016, "step": 9820 }, { "epoch": 0.23, "learning_rate": 2.8448111989861774e-05, "loss": 0.2855, "step": 9825 }, { "epoch": 0.23, "learning_rate": 2.844723193541574e-05, "loss": 0.4706, "step": 9830 }, { "epoch": 0.23, "learning_rate": 2.8446351880969704e-05, "loss": 0.3848, "step": 9835 }, { "epoch": 0.23, "learning_rate": 2.8445471826523665e-05, "loss": 0.3688, "step": 9840 }, { "epoch": 0.23, "learning_rate": 2.8444591772077634e-05, "loss": 0.4236, "step": 9845 }, { "epoch": 0.23, "learning_rate": 2.84437117176316e-05, "loss": 0.7271, "step": 9850 }, { "epoch": 0.23, "learning_rate": 2.8442831663185564e-05, "loss": 0.2542, "step": 9855 }, { "epoch": 0.23, "learning_rate": 2.8441951608739525e-05, "loss": 0.2058, "step": 9860 }, { "epoch": 0.23, "learning_rate": 2.8441071554293494e-05, "loss": 0.2132, "step": 9865 }, { "epoch": 0.23, "learning_rate": 2.844019149984746e-05, "loss": 0.3669, "step": 9870 }, { "epoch": 0.23, "learning_rate": 2.843931144540142e-05, "loss": 0.3556, "step": 9875 }, { "epoch": 0.23, "learning_rate": 2.843843139095539e-05, "loss": 0.2639, "step": 9880 }, { "epoch": 0.23, "learning_rate": 2.8437551336509354e-05, "loss": 0.3902, "step": 9885 }, { "epoch": 0.23, "learning_rate": 2.843667128206332e-05, "loss": 0.4111, "step": 9890 }, { "epoch": 0.23, "learning_rate": 2.843579122761728e-05, "loss": 0.4983, "step": 9895 }, { "epoch": 0.23, "learning_rate": 2.843491117317125e-05, "loss": 0.8109, "step": 9900 }, { "epoch": 0.23, "learning_rate": 2.8434031118725214e-05, "loss": 0.3265, "step": 9905 }, { "epoch": 0.23, "learning_rate": 2.8433151064279176e-05, "loss": 0.1365, "step": 9910 }, { "epoch": 0.23, "learning_rate": 2.843227100983314e-05, "loss": 0.2153, "step": 9915 }, { "epoch": 0.23, "learning_rate": 2.843139095538711e-05, "loss": 0.2204, "step": 9920 }, { "epoch": 0.23, "learning_rate": 2.8430510900941074e-05, "loss": 0.3682, "step": 9925 }, { "epoch": 0.23, "learning_rate": 2.8429630846495036e-05, "loss": 0.2075, "step": 9930 }, { "epoch": 0.23, "learning_rate": 2.8428750792049e-05, "loss": 0.5942, "step": 9935 }, { "epoch": 0.23, "learning_rate": 2.842787073760297e-05, "loss": 0.3068, "step": 9940 }, { "epoch": 0.23, "learning_rate": 2.842699068315693e-05, "loss": 0.4505, "step": 9945 }, { "epoch": 0.23, "learning_rate": 2.8426110628710896e-05, "loss": 0.6627, "step": 9950 }, { "epoch": 0.23, "learning_rate": 2.842523057426486e-05, "loss": 0.3905, "step": 9955 }, { "epoch": 0.23, "learning_rate": 2.842435051981883e-05, "loss": 0.1735, "step": 9960 }, { "epoch": 0.23, "learning_rate": 2.842347046537279e-05, "loss": 0.1867, "step": 9965 }, { "epoch": 0.23, "learning_rate": 2.8422590410926756e-05, "loss": 0.1964, "step": 9970 }, { "epoch": 0.23, "learning_rate": 2.842171035648072e-05, "loss": 0.2284, "step": 9975 }, { "epoch": 0.23, "learning_rate": 2.842083030203469e-05, "loss": 0.4411, "step": 9980 }, { "epoch": 0.23, "learning_rate": 2.841995024758865e-05, "loss": 0.3328, "step": 9985 }, { "epoch": 0.23, "learning_rate": 2.8419070193142616e-05, "loss": 0.4592, "step": 9990 }, { "epoch": 0.23, "learning_rate": 2.841819013869658e-05, "loss": 0.7001, "step": 9995 }, { "epoch": 0.23, "learning_rate": 2.8417310084250546e-05, "loss": 0.7318, "step": 10000 }, { "epoch": 0.23, "learning_rate": 2.841643002980451e-05, "loss": 0.3034, "step": 10005 }, { "epoch": 0.23, "learning_rate": 2.8415549975358476e-05, "loss": 0.1011, "step": 10010 }, { "epoch": 0.23, "learning_rate": 2.8414669920912444e-05, "loss": 0.248, "step": 10015 }, { "epoch": 0.23, "learning_rate": 2.8413789866466406e-05, "loss": 0.2088, "step": 10020 }, { "epoch": 0.23, "learning_rate": 2.841290981202037e-05, "loss": 0.1879, "step": 10025 }, { "epoch": 0.23, "learning_rate": 2.8412029757574336e-05, "loss": 0.238, "step": 10030 }, { "epoch": 0.23, "learning_rate": 2.84111497031283e-05, "loss": 0.2017, "step": 10035 }, { "epoch": 0.23, "learning_rate": 2.8410269648682266e-05, "loss": 0.4194, "step": 10040 }, { "epoch": 0.23, "learning_rate": 2.840938959423623e-05, "loss": 0.7085, "step": 10045 }, { "epoch": 0.23, "learning_rate": 2.8408509539790196e-05, "loss": 0.7971, "step": 10050 }, { "epoch": 0.23, "learning_rate": 2.840762948534416e-05, "loss": 0.4, "step": 10055 }, { "epoch": 0.23, "learning_rate": 2.8406749430898126e-05, "loss": 0.2043, "step": 10060 }, { "epoch": 0.23, "learning_rate": 2.840586937645209e-05, "loss": 0.2599, "step": 10065 }, { "epoch": 0.23, "learning_rate": 2.8404989322006053e-05, "loss": 0.3027, "step": 10070 }, { "epoch": 0.24, "learning_rate": 2.840410926756002e-05, "loss": 0.2119, "step": 10075 }, { "epoch": 0.24, "learning_rate": 2.8403229213113986e-05, "loss": 0.3152, "step": 10080 }, { "epoch": 0.24, "learning_rate": 2.840234915866795e-05, "loss": 0.4142, "step": 10085 }, { "epoch": 0.24, "learning_rate": 2.8401469104221913e-05, "loss": 0.4226, "step": 10090 }, { "epoch": 0.24, "learning_rate": 2.840058904977588e-05, "loss": 0.4407, "step": 10095 }, { "epoch": 0.24, "learning_rate": 2.8399708995329846e-05, "loss": 0.6792, "step": 10100 }, { "epoch": 0.24, "learning_rate": 2.8398828940883808e-05, "loss": 0.3629, "step": 10105 }, { "epoch": 0.24, "learning_rate": 2.8397948886437773e-05, "loss": 0.1551, "step": 10110 }, { "epoch": 0.24, "learning_rate": 2.839706883199174e-05, "loss": 0.1723, "step": 10115 }, { "epoch": 0.24, "learning_rate": 2.8396188777545706e-05, "loss": 0.2472, "step": 10120 }, { "epoch": 0.24, "learning_rate": 2.8395308723099668e-05, "loss": 0.158, "step": 10125 }, { "epoch": 0.24, "learning_rate": 2.8394428668653633e-05, "loss": 0.3857, "step": 10130 }, { "epoch": 0.24, "learning_rate": 2.83935486142076e-05, "loss": 0.2764, "step": 10135 }, { "epoch": 0.24, "learning_rate": 2.8392668559761563e-05, "loss": 0.3449, "step": 10140 }, { "epoch": 0.24, "learning_rate": 2.8391788505315528e-05, "loss": 0.4258, "step": 10145 }, { "epoch": 0.24, "learning_rate": 2.8390908450869497e-05, "loss": 0.5885, "step": 10150 }, { "epoch": 0.24, "learning_rate": 2.839002839642346e-05, "loss": 0.3347, "step": 10155 }, { "epoch": 0.24, "learning_rate": 2.8389148341977423e-05, "loss": 0.2633, "step": 10160 }, { "epoch": 0.24, "learning_rate": 2.8388268287531388e-05, "loss": 0.1961, "step": 10165 }, { "epoch": 0.24, "learning_rate": 2.8387388233085357e-05, "loss": 0.4482, "step": 10170 }, { "epoch": 0.24, "learning_rate": 2.8386508178639318e-05, "loss": 0.2246, "step": 10175 }, { "epoch": 0.24, "learning_rate": 2.8385628124193283e-05, "loss": 0.2533, "step": 10180 }, { "epoch": 0.24, "learning_rate": 2.838474806974725e-05, "loss": 0.2623, "step": 10185 }, { "epoch": 0.24, "learning_rate": 2.8383868015301217e-05, "loss": 0.4843, "step": 10190 }, { "epoch": 0.24, "learning_rate": 2.838298796085518e-05, "loss": 0.3541, "step": 10195 }, { "epoch": 0.24, "learning_rate": 2.8382107906409143e-05, "loss": 0.6706, "step": 10200 }, { "epoch": 0.24, "learning_rate": 2.838122785196311e-05, "loss": 0.3523, "step": 10205 }, { "epoch": 0.24, "learning_rate": 2.8380347797517073e-05, "loss": 0.1298, "step": 10210 }, { "epoch": 0.24, "learning_rate": 2.837946774307104e-05, "loss": 0.1538, "step": 10215 }, { "epoch": 0.24, "learning_rate": 2.8378587688625003e-05, "loss": 0.1246, "step": 10220 }, { "epoch": 0.24, "learning_rate": 2.837770763417897e-05, "loss": 0.3012, "step": 10225 }, { "epoch": 0.24, "learning_rate": 2.8376827579732934e-05, "loss": 0.1598, "step": 10230 }, { "epoch": 0.24, "learning_rate": 2.83759475252869e-05, "loss": 0.3412, "step": 10235 }, { "epoch": 0.24, "learning_rate": 2.8375067470840864e-05, "loss": 0.3415, "step": 10240 }, { "epoch": 0.24, "learning_rate": 2.8374187416394825e-05, "loss": 0.5635, "step": 10245 }, { "epoch": 0.24, "learning_rate": 2.8373307361948794e-05, "loss": 0.3821, "step": 10250 }, { "epoch": 0.24, "learning_rate": 2.837242730750276e-05, "loss": 0.3162, "step": 10255 }, { "epoch": 0.24, "learning_rate": 2.8371547253056724e-05, "loss": 0.2349, "step": 10260 }, { "epoch": 0.24, "learning_rate": 2.8370667198610685e-05, "loss": 0.1624, "step": 10265 }, { "epoch": 0.24, "learning_rate": 2.8369787144164654e-05, "loss": 0.2337, "step": 10270 }, { "epoch": 0.24, "learning_rate": 2.836890708971862e-05, "loss": 0.2276, "step": 10275 }, { "epoch": 0.24, "learning_rate": 2.8368027035272584e-05, "loss": 0.2266, "step": 10280 }, { "epoch": 0.24, "learning_rate": 2.836714698082655e-05, "loss": 0.3687, "step": 10285 }, { "epoch": 0.24, "learning_rate": 2.8366266926380514e-05, "loss": 0.4755, "step": 10290 }, { "epoch": 0.24, "learning_rate": 2.836538687193448e-05, "loss": 0.438, "step": 10295 }, { "epoch": 0.24, "learning_rate": 2.836450681748844e-05, "loss": 0.8044, "step": 10300 }, { "epoch": 0.24, "learning_rate": 2.836362676304241e-05, "loss": 0.479, "step": 10305 }, { "epoch": 0.24, "learning_rate": 2.8362746708596374e-05, "loss": 0.179, "step": 10310 }, { "epoch": 0.24, "learning_rate": 2.836186665415034e-05, "loss": 0.0193, "step": 10315 }, { "epoch": 0.24, "learning_rate": 2.83609865997043e-05, "loss": 0.2638, "step": 10320 }, { "epoch": 0.24, "learning_rate": 2.836010654525827e-05, "loss": 0.2807, "step": 10325 }, { "epoch": 0.24, "learning_rate": 2.8359226490812234e-05, "loss": 0.3452, "step": 10330 }, { "epoch": 0.24, "learning_rate": 2.8358346436366196e-05, "loss": 0.3911, "step": 10335 }, { "epoch": 0.24, "learning_rate": 2.835746638192016e-05, "loss": 0.3029, "step": 10340 }, { "epoch": 0.24, "learning_rate": 2.835658632747413e-05, "loss": 0.4273, "step": 10345 }, { "epoch": 0.24, "learning_rate": 2.8355706273028094e-05, "loss": 0.5369, "step": 10350 }, { "epoch": 0.24, "learning_rate": 2.8354826218582056e-05, "loss": 0.3952, "step": 10355 }, { "epoch": 0.24, "learning_rate": 2.835394616413602e-05, "loss": 0.13, "step": 10360 }, { "epoch": 0.24, "learning_rate": 2.835306610968999e-05, "loss": 0.1963, "step": 10365 }, { "epoch": 0.24, "learning_rate": 2.835218605524395e-05, "loss": 0.228, "step": 10370 }, { "epoch": 0.24, "learning_rate": 2.8351306000797916e-05, "loss": 0.2665, "step": 10375 }, { "epoch": 0.24, "learning_rate": 2.835042594635188e-05, "loss": 0.2205, "step": 10380 }, { "epoch": 0.24, "learning_rate": 2.834954589190585e-05, "loss": 0.3127, "step": 10385 }, { "epoch": 0.24, "learning_rate": 2.834866583745981e-05, "loss": 0.5213, "step": 10390 }, { "epoch": 0.24, "learning_rate": 2.8347785783013776e-05, "loss": 0.4116, "step": 10395 }, { "epoch": 0.24, "learning_rate": 2.834690572856774e-05, "loss": 0.6647, "step": 10400 }, { "epoch": 0.24, "learning_rate": 2.8346025674121706e-05, "loss": 0.4557, "step": 10405 }, { "epoch": 0.24, "learning_rate": 2.834514561967567e-05, "loss": 0.1779, "step": 10410 }, { "epoch": 0.24, "learning_rate": 2.8344265565229636e-05, "loss": 0.2126, "step": 10415 }, { "epoch": 0.24, "learning_rate": 2.8343385510783604e-05, "loss": 0.2417, "step": 10420 }, { "epoch": 0.24, "learning_rate": 2.8342505456337566e-05, "loss": 0.164, "step": 10425 }, { "epoch": 0.24, "learning_rate": 2.834162540189153e-05, "loss": 0.4136, "step": 10430 }, { "epoch": 0.24, "learning_rate": 2.8340745347445496e-05, "loss": 0.2688, "step": 10435 }, { "epoch": 0.24, "learning_rate": 2.833986529299946e-05, "loss": 0.3397, "step": 10440 }, { "epoch": 0.24, "learning_rate": 2.8338985238553426e-05, "loss": 0.3428, "step": 10445 }, { "epoch": 0.24, "learning_rate": 2.833810518410739e-05, "loss": 0.6351, "step": 10450 }, { "epoch": 0.24, "learning_rate": 2.8337225129661356e-05, "loss": 0.2874, "step": 10455 }, { "epoch": 0.24, "learning_rate": 2.833634507521532e-05, "loss": 0.1957, "step": 10460 }, { "epoch": 0.24, "learning_rate": 2.8335465020769286e-05, "loss": 0.2699, "step": 10465 }, { "epoch": 0.24, "learning_rate": 2.833458496632325e-05, "loss": 0.1934, "step": 10470 }, { "epoch": 0.24, "learning_rate": 2.8333704911877213e-05, "loss": 0.1859, "step": 10475 }, { "epoch": 0.24, "learning_rate": 2.833282485743118e-05, "loss": 0.3422, "step": 10480 }, { "epoch": 0.24, "learning_rate": 2.8331944802985146e-05, "loss": 0.2327, "step": 10485 }, { "epoch": 0.24, "learning_rate": 2.833106474853911e-05, "loss": 0.6878, "step": 10490 }, { "epoch": 0.24, "learning_rate": 2.8330184694093073e-05, "loss": 0.3825, "step": 10495 }, { "epoch": 0.24, "learning_rate": 2.832930463964704e-05, "loss": 0.7885, "step": 10500 }, { "epoch": 0.25, "learning_rate": 2.8328424585201006e-05, "loss": 0.4602, "step": 10505 }, { "epoch": 0.25, "learning_rate": 2.8327544530754968e-05, "loss": 0.1272, "step": 10510 }, { "epoch": 0.25, "learning_rate": 2.8326664476308933e-05, "loss": 0.1285, "step": 10515 }, { "epoch": 0.25, "learning_rate": 2.83257844218629e-05, "loss": 0.1745, "step": 10520 }, { "epoch": 0.25, "learning_rate": 2.8324904367416866e-05, "loss": 0.2223, "step": 10525 }, { "epoch": 0.25, "learning_rate": 2.8324024312970828e-05, "loss": 0.3421, "step": 10530 }, { "epoch": 0.25, "learning_rate": 2.8323144258524793e-05, "loss": 0.3921, "step": 10535 }, { "epoch": 0.25, "learning_rate": 2.832226420407876e-05, "loss": 0.2704, "step": 10540 }, { "epoch": 0.25, "learning_rate": 2.8321384149632723e-05, "loss": 0.4882, "step": 10545 }, { "epoch": 0.25, "learning_rate": 2.8320504095186688e-05, "loss": 0.8581, "step": 10550 }, { "epoch": 0.25, "learning_rate": 2.8319624040740656e-05, "loss": 0.309, "step": 10555 }, { "epoch": 0.25, "learning_rate": 2.831874398629462e-05, "loss": 0.2051, "step": 10560 }, { "epoch": 0.25, "learning_rate": 2.8317863931848583e-05, "loss": 0.2163, "step": 10565 }, { "epoch": 0.25, "learning_rate": 2.8316983877402548e-05, "loss": 0.3404, "step": 10570 }, { "epoch": 0.25, "learning_rate": 2.8316103822956517e-05, "loss": 0.4369, "step": 10575 }, { "epoch": 0.25, "learning_rate": 2.831522376851048e-05, "loss": 0.3508, "step": 10580 }, { "epoch": 0.25, "learning_rate": 2.8314343714064443e-05, "loss": 0.2536, "step": 10585 }, { "epoch": 0.25, "learning_rate": 2.8313463659618408e-05, "loss": 0.3509, "step": 10590 }, { "epoch": 0.25, "learning_rate": 2.8312583605172377e-05, "loss": 0.3982, "step": 10595 }, { "epoch": 0.25, "learning_rate": 2.8311703550726338e-05, "loss": 0.6909, "step": 10600 }, { "epoch": 0.25, "learning_rate": 2.8310823496280303e-05, "loss": 0.2619, "step": 10605 }, { "epoch": 0.25, "learning_rate": 2.8309943441834268e-05, "loss": 0.1676, "step": 10610 }, { "epoch": 0.25, "learning_rate": 2.8309063387388237e-05, "loss": 0.2052, "step": 10615 }, { "epoch": 0.25, "learning_rate": 2.83081833329422e-05, "loss": 0.2185, "step": 10620 }, { "epoch": 0.25, "learning_rate": 2.8307303278496163e-05, "loss": 0.2483, "step": 10625 }, { "epoch": 0.25, "learning_rate": 2.830642322405013e-05, "loss": 0.2559, "step": 10630 }, { "epoch": 0.25, "learning_rate": 2.8305543169604093e-05, "loss": 0.2331, "step": 10635 }, { "epoch": 0.25, "learning_rate": 2.830466311515806e-05, "loss": 0.2569, "step": 10640 }, { "epoch": 0.25, "learning_rate": 2.8303783060712023e-05, "loss": 0.5153, "step": 10645 }, { "epoch": 0.25, "learning_rate": 2.830290300626599e-05, "loss": 0.7447, "step": 10650 }, { "epoch": 0.25, "learning_rate": 2.8302022951819953e-05, "loss": 0.3187, "step": 10655 }, { "epoch": 0.25, "learning_rate": 2.830114289737392e-05, "loss": 0.1013, "step": 10660 }, { "epoch": 0.25, "learning_rate": 2.8300262842927884e-05, "loss": 0.2474, "step": 10665 }, { "epoch": 0.25, "learning_rate": 2.8299382788481845e-05, "loss": 0.212, "step": 10670 }, { "epoch": 0.25, "learning_rate": 2.8298502734035814e-05, "loss": 0.336, "step": 10675 }, { "epoch": 0.25, "learning_rate": 2.829762267958978e-05, "loss": 0.268, "step": 10680 }, { "epoch": 0.25, "learning_rate": 2.8296742625143744e-05, "loss": 0.3764, "step": 10685 }, { "epoch": 0.25, "learning_rate": 2.829586257069771e-05, "loss": 0.4697, "step": 10690 }, { "epoch": 0.25, "learning_rate": 2.8294982516251674e-05, "loss": 0.6862, "step": 10695 }, { "epoch": 0.25, "learning_rate": 2.829410246180564e-05, "loss": 0.5329, "step": 10700 }, { "epoch": 0.25, "learning_rate": 2.82932224073596e-05, "loss": 0.3453, "step": 10705 }, { "epoch": 0.25, "learning_rate": 2.829234235291357e-05, "loss": 0.1952, "step": 10710 }, { "epoch": 0.25, "learning_rate": 2.8291462298467534e-05, "loss": 0.2843, "step": 10715 }, { "epoch": 0.25, "learning_rate": 2.82905822440215e-05, "loss": 0.2395, "step": 10720 }, { "epoch": 0.25, "learning_rate": 2.828970218957546e-05, "loss": 0.4369, "step": 10725 }, { "epoch": 0.25, "learning_rate": 2.828882213512943e-05, "loss": 0.2147, "step": 10730 }, { "epoch": 0.25, "learning_rate": 2.8287942080683394e-05, "loss": 0.4492, "step": 10735 }, { "epoch": 0.25, "learning_rate": 2.8287062026237355e-05, "loss": 0.4735, "step": 10740 }, { "epoch": 0.25, "learning_rate": 2.828618197179132e-05, "loss": 0.2995, "step": 10745 }, { "epoch": 0.25, "learning_rate": 2.828530191734529e-05, "loss": 0.9343, "step": 10750 }, { "epoch": 0.25, "learning_rate": 2.8284421862899254e-05, "loss": 0.3098, "step": 10755 }, { "epoch": 0.25, "learning_rate": 2.8283541808453216e-05, "loss": 0.1257, "step": 10760 }, { "epoch": 0.25, "learning_rate": 2.828266175400718e-05, "loss": 0.2345, "step": 10765 }, { "epoch": 0.25, "learning_rate": 2.828178169956115e-05, "loss": 0.1606, "step": 10770 }, { "epoch": 0.25, "learning_rate": 2.828090164511511e-05, "loss": 0.3139, "step": 10775 }, { "epoch": 0.25, "learning_rate": 2.8280021590669076e-05, "loss": 0.4785, "step": 10780 }, { "epoch": 0.25, "learning_rate": 2.827914153622304e-05, "loss": 0.2193, "step": 10785 }, { "epoch": 0.25, "learning_rate": 2.827826148177701e-05, "loss": 0.3668, "step": 10790 }, { "epoch": 0.25, "learning_rate": 2.827738142733097e-05, "loss": 0.4088, "step": 10795 }, { "epoch": 0.25, "learning_rate": 2.8276501372884936e-05, "loss": 0.4793, "step": 10800 }, { "epoch": 0.25, "learning_rate": 2.82756213184389e-05, "loss": 0.3346, "step": 10805 }, { "epoch": 0.25, "learning_rate": 2.8274741263992866e-05, "loss": 0.1686, "step": 10810 }, { "epoch": 0.25, "learning_rate": 2.827386120954683e-05, "loss": 0.2129, "step": 10815 }, { "epoch": 0.25, "learning_rate": 2.8272981155100796e-05, "loss": 0.2737, "step": 10820 }, { "epoch": 0.25, "learning_rate": 2.8272101100654764e-05, "loss": 0.2845, "step": 10825 }, { "epoch": 0.25, "learning_rate": 2.8271221046208726e-05, "loss": 0.2334, "step": 10830 }, { "epoch": 0.25, "learning_rate": 2.827034099176269e-05, "loss": 0.281, "step": 10835 }, { "epoch": 0.25, "learning_rate": 2.8269460937316656e-05, "loss": 0.4115, "step": 10840 }, { "epoch": 0.25, "learning_rate": 2.826858088287062e-05, "loss": 0.4157, "step": 10845 }, { "epoch": 0.25, "learning_rate": 2.8267700828424586e-05, "loss": 0.7213, "step": 10850 }, { "epoch": 0.25, "learning_rate": 2.826682077397855e-05, "loss": 0.1895, "step": 10855 }, { "epoch": 0.25, "learning_rate": 2.8265940719532516e-05, "loss": 0.2806, "step": 10860 }, { "epoch": 0.25, "learning_rate": 2.826506066508648e-05, "loss": 0.2474, "step": 10865 }, { "epoch": 0.25, "learning_rate": 2.8264180610640446e-05, "loss": 0.1497, "step": 10870 }, { "epoch": 0.25, "learning_rate": 2.826330055619441e-05, "loss": 0.2724, "step": 10875 }, { "epoch": 0.25, "learning_rate": 2.8262420501748373e-05, "loss": 0.2518, "step": 10880 }, { "epoch": 0.25, "learning_rate": 2.826154044730234e-05, "loss": 0.4084, "step": 10885 }, { "epoch": 0.25, "learning_rate": 2.8260660392856306e-05, "loss": 0.2823, "step": 10890 }, { "epoch": 0.25, "learning_rate": 2.825978033841027e-05, "loss": 0.3603, "step": 10895 }, { "epoch": 0.25, "learning_rate": 2.8258900283964233e-05, "loss": 0.663, "step": 10900 }, { "epoch": 0.25, "learning_rate": 2.82580202295182e-05, "loss": 0.3072, "step": 10905 }, { "epoch": 0.25, "learning_rate": 2.8257140175072166e-05, "loss": 0.261, "step": 10910 }, { "epoch": 0.25, "learning_rate": 2.825626012062613e-05, "loss": 0.1796, "step": 10915 }, { "epoch": 0.25, "learning_rate": 2.8255380066180093e-05, "loss": 0.1813, "step": 10920 }, { "epoch": 0.25, "learning_rate": 2.825450001173406e-05, "loss": 0.2429, "step": 10925 }, { "epoch": 0.26, "learning_rate": 2.8253619957288026e-05, "loss": 0.3398, "step": 10930 }, { "epoch": 0.26, "learning_rate": 2.8252739902841988e-05, "loss": 0.2169, "step": 10935 }, { "epoch": 0.26, "learning_rate": 2.8251859848395953e-05, "loss": 0.3048, "step": 10940 }, { "epoch": 0.26, "learning_rate": 2.825097979394992e-05, "loss": 0.4524, "step": 10945 }, { "epoch": 0.26, "learning_rate": 2.8250099739503886e-05, "loss": 0.6033, "step": 10950 }, { "epoch": 0.26, "learning_rate": 2.8249219685057848e-05, "loss": 0.2806, "step": 10955 }, { "epoch": 0.26, "learning_rate": 2.8248339630611816e-05, "loss": 0.1074, "step": 10960 }, { "epoch": 0.26, "learning_rate": 2.824745957616578e-05, "loss": 0.1974, "step": 10965 }, { "epoch": 0.26, "learning_rate": 2.8246579521719743e-05, "loss": 0.1198, "step": 10970 }, { "epoch": 0.26, "learning_rate": 2.8245699467273708e-05, "loss": 0.261, "step": 10975 }, { "epoch": 0.26, "learning_rate": 2.8244819412827676e-05, "loss": 0.3751, "step": 10980 }, { "epoch": 0.26, "learning_rate": 2.824393935838164e-05, "loss": 0.1442, "step": 10985 }, { "epoch": 0.26, "learning_rate": 2.8243059303935603e-05, "loss": 0.2794, "step": 10990 }, { "epoch": 0.26, "learning_rate": 2.8242179249489568e-05, "loss": 0.528, "step": 10995 }, { "epoch": 0.26, "learning_rate": 2.8241299195043537e-05, "loss": 0.8047, "step": 11000 }, { "epoch": 0.26, "learning_rate": 2.8240419140597498e-05, "loss": 0.2764, "step": 11005 }, { "epoch": 0.26, "learning_rate": 2.8239539086151463e-05, "loss": 0.1739, "step": 11010 }, { "epoch": 0.26, "learning_rate": 2.8238659031705428e-05, "loss": 0.2857, "step": 11015 }, { "epoch": 0.26, "learning_rate": 2.8237778977259397e-05, "loss": 0.2074, "step": 11020 }, { "epoch": 0.26, "learning_rate": 2.8236898922813358e-05, "loss": 0.3995, "step": 11025 }, { "epoch": 0.26, "learning_rate": 2.8236018868367323e-05, "loss": 0.2577, "step": 11030 }, { "epoch": 0.26, "learning_rate": 2.8235138813921288e-05, "loss": 0.2492, "step": 11035 }, { "epoch": 0.26, "learning_rate": 2.8234258759475253e-05, "loss": 0.3465, "step": 11040 }, { "epoch": 0.26, "learning_rate": 2.8233378705029218e-05, "loss": 0.4394, "step": 11045 }, { "epoch": 0.26, "learning_rate": 2.8232498650583183e-05, "loss": 0.6942, "step": 11050 }, { "epoch": 0.26, "learning_rate": 2.823161859613715e-05, "loss": 0.3759, "step": 11055 }, { "epoch": 0.26, "learning_rate": 2.8230738541691113e-05, "loss": 0.114, "step": 11060 }, { "epoch": 0.26, "learning_rate": 2.822985848724508e-05, "loss": 0.2052, "step": 11065 }, { "epoch": 0.26, "learning_rate": 2.8228978432799043e-05, "loss": 0.1894, "step": 11070 }, { "epoch": 0.26, "learning_rate": 2.8228098378353005e-05, "loss": 0.2495, "step": 11075 }, { "epoch": 0.26, "learning_rate": 2.8227218323906973e-05, "loss": 0.2641, "step": 11080 }, { "epoch": 0.26, "learning_rate": 2.822633826946094e-05, "loss": 0.1918, "step": 11085 }, { "epoch": 0.26, "learning_rate": 2.8225458215014903e-05, "loss": 0.4123, "step": 11090 }, { "epoch": 0.26, "learning_rate": 2.822457816056887e-05, "loss": 0.3797, "step": 11095 }, { "epoch": 0.26, "learning_rate": 2.8223698106122834e-05, "loss": 0.6931, "step": 11100 }, { "epoch": 0.26, "learning_rate": 2.82228180516768e-05, "loss": 0.3473, "step": 11105 }, { "epoch": 0.26, "learning_rate": 2.822193799723076e-05, "loss": 0.2206, "step": 11110 }, { "epoch": 0.26, "learning_rate": 2.822105794278473e-05, "loss": 0.148, "step": 11115 }, { "epoch": 0.26, "learning_rate": 2.8220177888338694e-05, "loss": 0.1633, "step": 11120 }, { "epoch": 0.26, "learning_rate": 2.821929783389266e-05, "loss": 0.3237, "step": 11125 }, { "epoch": 0.26, "learning_rate": 2.821841777944662e-05, "loss": 0.2189, "step": 11130 }, { "epoch": 0.26, "learning_rate": 2.821753772500059e-05, "loss": 0.376, "step": 11135 }, { "epoch": 0.26, "learning_rate": 2.8216657670554554e-05, "loss": 0.3067, "step": 11140 }, { "epoch": 0.26, "learning_rate": 2.8215777616108515e-05, "loss": 0.6555, "step": 11145 }, { "epoch": 0.26, "learning_rate": 2.821489756166248e-05, "loss": 0.6229, "step": 11150 }, { "epoch": 0.26, "learning_rate": 2.821401750721645e-05, "loss": 0.3123, "step": 11155 }, { "epoch": 0.26, "learning_rate": 2.8213137452770414e-05, "loss": 0.1671, "step": 11160 }, { "epoch": 0.26, "learning_rate": 2.8212257398324375e-05, "loss": 0.1763, "step": 11165 }, { "epoch": 0.26, "learning_rate": 2.821137734387834e-05, "loss": 0.2604, "step": 11170 }, { "epoch": 0.26, "learning_rate": 2.821049728943231e-05, "loss": 0.2918, "step": 11175 }, { "epoch": 0.26, "learning_rate": 2.820961723498627e-05, "loss": 0.2654, "step": 11180 }, { "epoch": 0.26, "learning_rate": 2.8208737180540235e-05, "loss": 0.2705, "step": 11185 }, { "epoch": 0.26, "learning_rate": 2.82078571260942e-05, "loss": 0.4101, "step": 11190 }, { "epoch": 0.26, "learning_rate": 2.820697707164817e-05, "loss": 0.6948, "step": 11195 }, { "epoch": 0.26, "learning_rate": 2.820609701720213e-05, "loss": 0.5025, "step": 11200 }, { "epoch": 0.26, "learning_rate": 2.8205216962756096e-05, "loss": 0.2678, "step": 11205 }, { "epoch": 0.26, "learning_rate": 2.820433690831006e-05, "loss": 0.1161, "step": 11210 }, { "epoch": 0.26, "learning_rate": 2.820345685386403e-05, "loss": 0.1483, "step": 11215 }, { "epoch": 0.26, "learning_rate": 2.820257679941799e-05, "loss": 0.2023, "step": 11220 }, { "epoch": 0.26, "learning_rate": 2.8201696744971956e-05, "loss": 0.328, "step": 11225 }, { "epoch": 0.26, "learning_rate": 2.8200816690525924e-05, "loss": 0.1201, "step": 11230 }, { "epoch": 0.26, "learning_rate": 2.8199936636079886e-05, "loss": 0.4183, "step": 11235 }, { "epoch": 0.26, "learning_rate": 2.819905658163385e-05, "loss": 0.4826, "step": 11240 }, { "epoch": 0.26, "learning_rate": 2.8198176527187816e-05, "loss": 0.3939, "step": 11245 }, { "epoch": 0.26, "learning_rate": 2.8197296472741784e-05, "loss": 0.6775, "step": 11250 }, { "epoch": 0.26, "learning_rate": 2.8196416418295746e-05, "loss": 0.3212, "step": 11255 }, { "epoch": 0.26, "learning_rate": 2.819553636384971e-05, "loss": 0.123, "step": 11260 }, { "epoch": 0.26, "learning_rate": 2.8194656309403676e-05, "loss": 0.1876, "step": 11265 }, { "epoch": 0.26, "learning_rate": 2.819377625495764e-05, "loss": 0.281, "step": 11270 }, { "epoch": 0.26, "learning_rate": 2.8192896200511606e-05, "loss": 0.1925, "step": 11275 }, { "epoch": 0.26, "learning_rate": 2.819201614606557e-05, "loss": 0.189, "step": 11280 }, { "epoch": 0.26, "learning_rate": 2.8191136091619536e-05, "loss": 0.4397, "step": 11285 }, { "epoch": 0.26, "learning_rate": 2.81902560371735e-05, "loss": 0.3553, "step": 11290 }, { "epoch": 0.26, "learning_rate": 2.8189375982727466e-05, "loss": 0.4583, "step": 11295 }, { "epoch": 0.26, "learning_rate": 2.818849592828143e-05, "loss": 0.553, "step": 11300 }, { "epoch": 0.26, "learning_rate": 2.8187615873835393e-05, "loss": 0.3859, "step": 11305 }, { "epoch": 0.26, "learning_rate": 2.818673581938936e-05, "loss": 0.3093, "step": 11310 }, { "epoch": 0.26, "learning_rate": 2.8185855764943326e-05, "loss": 0.214, "step": 11315 }, { "epoch": 0.26, "learning_rate": 2.818497571049729e-05, "loss": 0.2412, "step": 11320 }, { "epoch": 0.26, "learning_rate": 2.8184095656051253e-05, "loss": 0.2266, "step": 11325 }, { "epoch": 0.26, "learning_rate": 2.818321560160522e-05, "loss": 0.1674, "step": 11330 }, { "epoch": 0.26, "learning_rate": 2.8182335547159186e-05, "loss": 0.2985, "step": 11335 }, { "epoch": 0.26, "learning_rate": 2.8181455492713148e-05, "loss": 0.5052, "step": 11340 }, { "epoch": 0.26, "learning_rate": 2.8180575438267113e-05, "loss": 0.8356, "step": 11345 }, { "epoch": 0.26, "learning_rate": 2.817969538382108e-05, "loss": 0.8026, "step": 11350 }, { "epoch": 0.26, "learning_rate": 2.8178815329375046e-05, "loss": 0.3295, "step": 11355 }, { "epoch": 0.27, "learning_rate": 2.8177935274929008e-05, "loss": 0.2001, "step": 11360 }, { "epoch": 0.27, "learning_rate": 2.8177055220482976e-05, "loss": 0.2014, "step": 11365 }, { "epoch": 0.27, "learning_rate": 2.817617516603694e-05, "loss": 0.3035, "step": 11370 }, { "epoch": 0.27, "learning_rate": 2.8175295111590903e-05, "loss": 0.1727, "step": 11375 }, { "epoch": 0.27, "learning_rate": 2.8174415057144868e-05, "loss": 0.2302, "step": 11380 }, { "epoch": 0.27, "learning_rate": 2.8173535002698836e-05, "loss": 0.2606, "step": 11385 }, { "epoch": 0.27, "learning_rate": 2.81726549482528e-05, "loss": 0.3283, "step": 11390 }, { "epoch": 0.27, "learning_rate": 2.8171774893806763e-05, "loss": 0.5312, "step": 11395 }, { "epoch": 0.27, "learning_rate": 2.8170894839360728e-05, "loss": 0.8038, "step": 11400 }, { "epoch": 0.27, "learning_rate": 2.8170014784914696e-05, "loss": 0.3359, "step": 11405 }, { "epoch": 0.27, "learning_rate": 2.8169134730468658e-05, "loss": 0.161, "step": 11410 }, { "epoch": 0.27, "learning_rate": 2.8168254676022623e-05, "loss": 0.1319, "step": 11415 }, { "epoch": 0.27, "learning_rate": 2.8167374621576588e-05, "loss": 0.1742, "step": 11420 }, { "epoch": 0.27, "learning_rate": 2.8166494567130556e-05, "loss": 0.3764, "step": 11425 }, { "epoch": 0.27, "learning_rate": 2.8165614512684518e-05, "loss": 0.2724, "step": 11430 }, { "epoch": 0.27, "learning_rate": 2.8164734458238483e-05, "loss": 0.3272, "step": 11435 }, { "epoch": 0.27, "learning_rate": 2.8163854403792448e-05, "loss": 0.4479, "step": 11440 }, { "epoch": 0.27, "learning_rate": 2.8162974349346413e-05, "loss": 0.4866, "step": 11445 }, { "epoch": 0.27, "learning_rate": 2.8162094294900378e-05, "loss": 0.9267, "step": 11450 }, { "epoch": 0.27, "learning_rate": 2.8161214240454343e-05, "loss": 0.3962, "step": 11455 }, { "epoch": 0.27, "learning_rate": 2.8160334186008308e-05, "loss": 0.1418, "step": 11460 }, { "epoch": 0.27, "learning_rate": 2.8159454131562273e-05, "loss": 0.1017, "step": 11465 }, { "epoch": 0.27, "learning_rate": 2.8158574077116238e-05, "loss": 0.3292, "step": 11470 }, { "epoch": 0.27, "learning_rate": 2.8157694022670203e-05, "loss": 0.2934, "step": 11475 }, { "epoch": 0.27, "learning_rate": 2.8156813968224165e-05, "loss": 0.4173, "step": 11480 }, { "epoch": 0.27, "learning_rate": 2.8155933913778133e-05, "loss": 0.4389, "step": 11485 }, { "epoch": 0.27, "learning_rate": 2.81550538593321e-05, "loss": 0.4167, "step": 11490 }, { "epoch": 0.27, "learning_rate": 2.8154173804886063e-05, "loss": 0.4911, "step": 11495 }, { "epoch": 0.27, "learning_rate": 2.815329375044003e-05, "loss": 0.631, "step": 11500 }, { "epoch": 0.27, "learning_rate": 2.8152413695993993e-05, "loss": 0.3285, "step": 11505 }, { "epoch": 0.27, "learning_rate": 2.815153364154796e-05, "loss": 0.0979, "step": 11510 }, { "epoch": 0.27, "learning_rate": 2.8150653587101923e-05, "loss": 0.2614, "step": 11515 }, { "epoch": 0.27, "learning_rate": 2.814977353265589e-05, "loss": 0.2523, "step": 11520 }, { "epoch": 0.27, "learning_rate": 2.8148893478209853e-05, "loss": 0.1727, "step": 11525 }, { "epoch": 0.27, "learning_rate": 2.814801342376382e-05, "loss": 0.4298, "step": 11530 }, { "epoch": 0.27, "learning_rate": 2.814713336931778e-05, "loss": 0.6712, "step": 11535 }, { "epoch": 0.27, "learning_rate": 2.814625331487175e-05, "loss": 0.4399, "step": 11540 }, { "epoch": 0.27, "learning_rate": 2.8145373260425714e-05, "loss": 0.3639, "step": 11545 }, { "epoch": 0.27, "learning_rate": 2.814449320597968e-05, "loss": 1.0126, "step": 11550 }, { "epoch": 0.27, "learning_rate": 2.814361315153364e-05, "loss": 0.3456, "step": 11555 }, { "epoch": 0.27, "learning_rate": 2.814273309708761e-05, "loss": 0.1965, "step": 11560 }, { "epoch": 0.27, "learning_rate": 2.8141853042641574e-05, "loss": 0.1214, "step": 11565 }, { "epoch": 0.27, "learning_rate": 2.8140972988195535e-05, "loss": 0.2549, "step": 11570 }, { "epoch": 0.27, "learning_rate": 2.81400929337495e-05, "loss": 0.4904, "step": 11575 }, { "epoch": 0.27, "learning_rate": 2.813921287930347e-05, "loss": 0.2393, "step": 11580 }, { "epoch": 0.27, "learning_rate": 2.8138332824857434e-05, "loss": 0.2158, "step": 11585 }, { "epoch": 0.27, "learning_rate": 2.8137452770411395e-05, "loss": 0.3722, "step": 11590 }, { "epoch": 0.27, "learning_rate": 2.813657271596536e-05, "loss": 0.5494, "step": 11595 }, { "epoch": 0.27, "learning_rate": 2.813569266151933e-05, "loss": 0.7336, "step": 11600 }, { "epoch": 0.27, "learning_rate": 2.813481260707329e-05, "loss": 0.2306, "step": 11605 }, { "epoch": 0.27, "learning_rate": 2.8133932552627255e-05, "loss": 0.1449, "step": 11610 }, { "epoch": 0.27, "learning_rate": 2.813305249818122e-05, "loss": 0.1852, "step": 11615 }, { "epoch": 0.27, "learning_rate": 2.813217244373519e-05, "loss": 0.2669, "step": 11620 }, { "epoch": 0.27, "learning_rate": 2.813129238928915e-05, "loss": 0.1263, "step": 11625 }, { "epoch": 0.27, "learning_rate": 2.8130412334843116e-05, "loss": 0.3168, "step": 11630 }, { "epoch": 0.27, "learning_rate": 2.812953228039708e-05, "loss": 0.2014, "step": 11635 }, { "epoch": 0.27, "learning_rate": 2.8128652225951046e-05, "loss": 0.2182, "step": 11640 }, { "epoch": 0.27, "learning_rate": 2.8127948182394217e-05, "loss": 0.4558, "step": 11645 }, { "epoch": 0.27, "learning_rate": 2.8127068127948182e-05, "loss": 0.6697, "step": 11650 }, { "epoch": 0.27, "learning_rate": 2.812618807350215e-05, "loss": 0.384, "step": 11655 }, { "epoch": 0.27, "learning_rate": 2.8125308019056112e-05, "loss": 0.126, "step": 11660 }, { "epoch": 0.27, "learning_rate": 2.8124427964610077e-05, "loss": 0.247, "step": 11665 }, { "epoch": 0.27, "learning_rate": 2.8123547910164042e-05, "loss": 0.2952, "step": 11670 }, { "epoch": 0.27, "learning_rate": 2.8122667855718007e-05, "loss": 0.2522, "step": 11675 }, { "epoch": 0.27, "learning_rate": 2.8121787801271972e-05, "loss": 0.2819, "step": 11680 }, { "epoch": 0.27, "learning_rate": 2.8120907746825937e-05, "loss": 0.5169, "step": 11685 }, { "epoch": 0.27, "learning_rate": 2.8120027692379902e-05, "loss": 0.3414, "step": 11690 }, { "epoch": 0.27, "learning_rate": 2.8119147637933867e-05, "loss": 0.5153, "step": 11695 }, { "epoch": 0.27, "learning_rate": 2.8118267583487832e-05, "loss": 0.6789, "step": 11700 }, { "epoch": 0.27, "learning_rate": 2.8117387529041797e-05, "loss": 0.2429, "step": 11705 }, { "epoch": 0.27, "learning_rate": 2.8116507474595766e-05, "loss": 0.1226, "step": 11710 }, { "epoch": 0.27, "learning_rate": 2.8115627420149727e-05, "loss": 0.1331, "step": 11715 }, { "epoch": 0.27, "learning_rate": 2.8114747365703692e-05, "loss": 0.1872, "step": 11720 }, { "epoch": 0.27, "learning_rate": 2.8113867311257657e-05, "loss": 0.4518, "step": 11725 }, { "epoch": 0.27, "learning_rate": 2.8112987256811622e-05, "loss": 0.2977, "step": 11730 }, { "epoch": 0.27, "learning_rate": 2.8112107202365587e-05, "loss": 0.3335, "step": 11735 }, { "epoch": 0.27, "learning_rate": 2.8111227147919552e-05, "loss": 0.3665, "step": 11740 }, { "epoch": 0.27, "learning_rate": 2.8110347093473517e-05, "loss": 0.5785, "step": 11745 }, { "epoch": 0.27, "learning_rate": 2.8109467039027482e-05, "loss": 0.5173, "step": 11750 }, { "epoch": 0.27, "learning_rate": 2.8108586984581447e-05, "loss": 0.3541, "step": 11755 }, { "epoch": 0.27, "learning_rate": 2.8107706930135412e-05, "loss": 0.1274, "step": 11760 }, { "epoch": 0.27, "learning_rate": 2.8106826875689374e-05, "loss": 0.1483, "step": 11765 }, { "epoch": 0.27, "learning_rate": 2.8105946821243342e-05, "loss": 0.1444, "step": 11770 }, { "epoch": 0.27, "learning_rate": 2.8105066766797307e-05, "loss": 0.3923, "step": 11775 }, { "epoch": 0.27, "learning_rate": 2.8104186712351272e-05, "loss": 0.2987, "step": 11780 }, { "epoch": 0.27, "learning_rate": 2.8103306657905234e-05, "loss": 0.4317, "step": 11785 }, { "epoch": 0.28, "learning_rate": 2.8102426603459203e-05, "loss": 0.5272, "step": 11790 }, { "epoch": 0.28, "learning_rate": 2.8101546549013168e-05, "loss": 0.4763, "step": 11795 }, { "epoch": 0.28, "learning_rate": 2.810066649456713e-05, "loss": 0.5564, "step": 11800 }, { "epoch": 0.28, "learning_rate": 2.8099786440121094e-05, "loss": 0.2614, "step": 11805 }, { "epoch": 0.28, "learning_rate": 2.8098906385675063e-05, "loss": 0.127, "step": 11810 }, { "epoch": 0.28, "learning_rate": 2.8098026331229028e-05, "loss": 0.1488, "step": 11815 }, { "epoch": 0.28, "learning_rate": 2.809714627678299e-05, "loss": 0.1774, "step": 11820 }, { "epoch": 0.28, "learning_rate": 2.8096266222336958e-05, "loss": 0.2649, "step": 11825 }, { "epoch": 0.28, "learning_rate": 2.8095386167890923e-05, "loss": 0.2445, "step": 11830 }, { "epoch": 0.28, "learning_rate": 2.8094506113444884e-05, "loss": 0.2628, "step": 11835 }, { "epoch": 0.28, "learning_rate": 2.809362605899885e-05, "loss": 0.3684, "step": 11840 }, { "epoch": 0.28, "learning_rate": 2.8092746004552818e-05, "loss": 0.6456, "step": 11845 }, { "epoch": 0.28, "learning_rate": 2.8091865950106783e-05, "loss": 0.7004, "step": 11850 }, { "epoch": 0.28, "learning_rate": 2.8090985895660744e-05, "loss": 0.2967, "step": 11855 }, { "epoch": 0.28, "learning_rate": 2.809010584121471e-05, "loss": 0.1061, "step": 11860 }, { "epoch": 0.28, "learning_rate": 2.8089225786768678e-05, "loss": 0.1705, "step": 11865 }, { "epoch": 0.28, "learning_rate": 2.808834573232264e-05, "loss": 0.2192, "step": 11870 }, { "epoch": 0.28, "learning_rate": 2.8087465677876604e-05, "loss": 0.2447, "step": 11875 }, { "epoch": 0.28, "learning_rate": 2.808658562343057e-05, "loss": 0.2318, "step": 11880 }, { "epoch": 0.28, "learning_rate": 2.8085705568984538e-05, "loss": 0.1875, "step": 11885 }, { "epoch": 0.28, "learning_rate": 2.80848255145385e-05, "loss": 0.3931, "step": 11890 }, { "epoch": 0.28, "learning_rate": 2.8083945460092465e-05, "loss": 0.3453, "step": 11895 }, { "epoch": 0.28, "learning_rate": 2.808306540564643e-05, "loss": 0.5456, "step": 11900 }, { "epoch": 0.28, "learning_rate": 2.8082185351200395e-05, "loss": 0.4742, "step": 11905 }, { "epoch": 0.28, "learning_rate": 2.808130529675436e-05, "loss": 0.1697, "step": 11910 }, { "epoch": 0.28, "learning_rate": 2.8080425242308325e-05, "loss": 0.0819, "step": 11915 }, { "epoch": 0.28, "learning_rate": 2.807954518786229e-05, "loss": 0.3423, "step": 11920 }, { "epoch": 0.28, "learning_rate": 2.8078665133416255e-05, "loss": 0.3776, "step": 11925 }, { "epoch": 0.28, "learning_rate": 2.807778507897022e-05, "loss": 0.3336, "step": 11930 }, { "epoch": 0.28, "learning_rate": 2.8076905024524185e-05, "loss": 0.2741, "step": 11935 }, { "epoch": 0.28, "learning_rate": 2.8076024970078146e-05, "loss": 0.4332, "step": 11940 }, { "epoch": 0.28, "learning_rate": 2.8075144915632115e-05, "loss": 0.4083, "step": 11945 }, { "epoch": 0.28, "learning_rate": 2.807426486118608e-05, "loss": 0.9895, "step": 11950 }, { "epoch": 0.28, "learning_rate": 2.8073384806740045e-05, "loss": 0.2878, "step": 11955 }, { "epoch": 0.28, "learning_rate": 2.807250475229401e-05, "loss": 0.1117, "step": 11960 }, { "epoch": 0.28, "learning_rate": 2.8071624697847975e-05, "loss": 0.1917, "step": 11965 }, { "epoch": 0.28, "learning_rate": 2.807074464340194e-05, "loss": 0.0816, "step": 11970 }, { "epoch": 0.28, "learning_rate": 2.80698645889559e-05, "loss": 0.2014, "step": 11975 }, { "epoch": 0.28, "learning_rate": 2.806898453450987e-05, "loss": 0.2949, "step": 11980 }, { "epoch": 0.28, "learning_rate": 2.8068104480063835e-05, "loss": 0.3076, "step": 11985 }, { "epoch": 0.28, "learning_rate": 2.80672244256178e-05, "loss": 0.5798, "step": 11990 }, { "epoch": 0.28, "learning_rate": 2.806634437117176e-05, "loss": 0.566, "step": 11995 }, { "epoch": 0.28, "learning_rate": 2.806546431672573e-05, "loss": 0.8678, "step": 12000 }, { "epoch": 0.28, "learning_rate": 2.8064584262279695e-05, "loss": 0.3687, "step": 12005 }, { "epoch": 0.28, "learning_rate": 2.806370420783366e-05, "loss": 0.1226, "step": 12010 }, { "epoch": 0.28, "learning_rate": 2.806282415338762e-05, "loss": 0.1437, "step": 12015 }, { "epoch": 0.28, "learning_rate": 2.806194409894159e-05, "loss": 0.0646, "step": 12020 }, { "epoch": 0.28, "learning_rate": 2.8061064044495555e-05, "loss": 0.2343, "step": 12025 }, { "epoch": 0.28, "learning_rate": 2.8060183990049517e-05, "loss": 0.2134, "step": 12030 }, { "epoch": 0.28, "learning_rate": 2.8059303935603482e-05, "loss": 0.3358, "step": 12035 }, { "epoch": 0.28, "learning_rate": 2.805842388115745e-05, "loss": 0.228, "step": 12040 }, { "epoch": 0.28, "learning_rate": 2.8057543826711415e-05, "loss": 0.4318, "step": 12045 }, { "epoch": 0.28, "learning_rate": 2.8056663772265377e-05, "loss": 0.5977, "step": 12050 }, { "epoch": 0.28, "learning_rate": 2.8055783717819342e-05, "loss": 0.3234, "step": 12055 }, { "epoch": 0.28, "learning_rate": 2.805490366337331e-05, "loss": 0.2072, "step": 12060 }, { "epoch": 0.28, "learning_rate": 2.8054023608927272e-05, "loss": 0.1917, "step": 12065 }, { "epoch": 0.28, "learning_rate": 2.8053143554481237e-05, "loss": 0.213, "step": 12070 }, { "epoch": 0.28, "learning_rate": 2.8052263500035202e-05, "loss": 0.3324, "step": 12075 }, { "epoch": 0.28, "learning_rate": 2.805138344558917e-05, "loss": 0.4237, "step": 12080 }, { "epoch": 0.28, "learning_rate": 2.8050503391143132e-05, "loss": 0.257, "step": 12085 }, { "epoch": 0.28, "learning_rate": 2.8049623336697097e-05, "loss": 0.4676, "step": 12090 }, { "epoch": 0.28, "learning_rate": 2.8048743282251062e-05, "loss": 0.3292, "step": 12095 }, { "epoch": 0.28, "learning_rate": 2.8047863227805027e-05, "loss": 0.7279, "step": 12100 }, { "epoch": 0.28, "learning_rate": 2.8046983173358992e-05, "loss": 0.2759, "step": 12105 }, { "epoch": 0.28, "learning_rate": 2.8046103118912957e-05, "loss": 0.2368, "step": 12110 }, { "epoch": 0.28, "learning_rate": 2.8045223064466925e-05, "loss": 0.1143, "step": 12115 }, { "epoch": 0.28, "learning_rate": 2.8044343010020887e-05, "loss": 0.1322, "step": 12120 }, { "epoch": 0.28, "learning_rate": 2.8043462955574852e-05, "loss": 0.3937, "step": 12125 }, { "epoch": 0.28, "learning_rate": 2.8042582901128817e-05, "loss": 0.4146, "step": 12130 }, { "epoch": 0.28, "learning_rate": 2.8041702846682782e-05, "loss": 0.4685, "step": 12135 }, { "epoch": 0.28, "learning_rate": 2.8040822792236747e-05, "loss": 0.2635, "step": 12140 }, { "epoch": 0.28, "learning_rate": 2.8039942737790712e-05, "loss": 0.3652, "step": 12145 }, { "epoch": 0.28, "learning_rate": 2.8039062683344677e-05, "loss": 0.6901, "step": 12150 }, { "epoch": 0.28, "learning_rate": 2.8038182628898642e-05, "loss": 0.4101, "step": 12155 }, { "epoch": 0.28, "learning_rate": 2.8037302574452607e-05, "loss": 0.2339, "step": 12160 }, { "epoch": 0.28, "learning_rate": 2.8036422520006572e-05, "loss": 0.254, "step": 12165 }, { "epoch": 0.28, "learning_rate": 2.8035542465560534e-05, "loss": 0.3094, "step": 12170 }, { "epoch": 0.28, "learning_rate": 2.8034662411114502e-05, "loss": 0.221, "step": 12175 }, { "epoch": 0.28, "learning_rate": 2.8033782356668467e-05, "loss": 0.4012, "step": 12180 }, { "epoch": 0.28, "learning_rate": 2.8032902302222432e-05, "loss": 0.4541, "step": 12185 }, { "epoch": 0.28, "learning_rate": 2.8032022247776394e-05, "loss": 0.3517, "step": 12190 }, { "epoch": 0.28, "learning_rate": 2.8031142193330362e-05, "loss": 0.4945, "step": 12195 }, { "epoch": 0.28, "learning_rate": 2.8030262138884327e-05, "loss": 0.5707, "step": 12200 }, { "epoch": 0.28, "learning_rate": 2.802938208443829e-05, "loss": 0.3567, "step": 12205 }, { "epoch": 0.28, "learning_rate": 2.8028502029992254e-05, "loss": 0.1651, "step": 12210 }, { "epoch": 0.28, "learning_rate": 2.8027621975546222e-05, "loss": 0.0904, "step": 12215 }, { "epoch": 0.29, "learning_rate": 2.8026741921100188e-05, "loss": 0.5068, "step": 12220 }, { "epoch": 0.29, "learning_rate": 2.802586186665415e-05, "loss": 0.2113, "step": 12225 }, { "epoch": 0.29, "learning_rate": 2.8024981812208114e-05, "loss": 0.2907, "step": 12230 }, { "epoch": 0.29, "learning_rate": 2.8024101757762083e-05, "loss": 0.3312, "step": 12235 }, { "epoch": 0.29, "learning_rate": 2.8023221703316044e-05, "loss": 0.422, "step": 12240 }, { "epoch": 0.29, "learning_rate": 2.802234164887001e-05, "loss": 0.4641, "step": 12245 }, { "epoch": 0.29, "learning_rate": 2.8021461594423978e-05, "loss": 0.5882, "step": 12250 }, { "epoch": 0.29, "learning_rate": 2.8020581539977943e-05, "loss": 0.3248, "step": 12255 }, { "epoch": 0.29, "learning_rate": 2.8019701485531904e-05, "loss": 0.141, "step": 12260 }, { "epoch": 0.29, "learning_rate": 2.801882143108587e-05, "loss": 0.1138, "step": 12265 }, { "epoch": 0.29, "learning_rate": 2.8017941376639838e-05, "loss": 0.2028, "step": 12270 }, { "epoch": 0.29, "learning_rate": 2.80170613221938e-05, "loss": 0.2836, "step": 12275 }, { "epoch": 0.29, "learning_rate": 2.8016181267747764e-05, "loss": 0.2948, "step": 12280 }, { "epoch": 0.29, "learning_rate": 2.801530121330173e-05, "loss": 0.3185, "step": 12285 }, { "epoch": 0.29, "learning_rate": 2.8014421158855698e-05, "loss": 0.3586, "step": 12290 }, { "epoch": 0.29, "learning_rate": 2.801354110440966e-05, "loss": 0.4109, "step": 12295 }, { "epoch": 0.29, "learning_rate": 2.8012661049963624e-05, "loss": 0.4865, "step": 12300 }, { "epoch": 0.29, "learning_rate": 2.801178099551759e-05, "loss": 0.3359, "step": 12305 }, { "epoch": 0.29, "learning_rate": 2.8010900941071558e-05, "loss": 0.1762, "step": 12310 }, { "epoch": 0.29, "learning_rate": 2.801002088662552e-05, "loss": 0.2099, "step": 12315 }, { "epoch": 0.29, "learning_rate": 2.8009140832179485e-05, "loss": 0.2233, "step": 12320 }, { "epoch": 0.29, "learning_rate": 2.800826077773345e-05, "loss": 0.2217, "step": 12325 }, { "epoch": 0.29, "learning_rate": 2.8007380723287415e-05, "loss": 0.2988, "step": 12330 }, { "epoch": 0.29, "learning_rate": 2.800650066884138e-05, "loss": 0.544, "step": 12335 }, { "epoch": 0.29, "learning_rate": 2.8005620614395345e-05, "loss": 0.4894, "step": 12340 }, { "epoch": 0.29, "learning_rate": 2.800474055994931e-05, "loss": 0.4696, "step": 12345 }, { "epoch": 0.29, "learning_rate": 2.8003860505503275e-05, "loss": 0.7285, "step": 12350 }, { "epoch": 0.29, "learning_rate": 2.800298045105724e-05, "loss": 0.2423, "step": 12355 }, { "epoch": 0.29, "learning_rate": 2.8002100396611205e-05, "loss": 0.2053, "step": 12360 }, { "epoch": 0.29, "learning_rate": 2.800122034216517e-05, "loss": 0.1639, "step": 12365 }, { "epoch": 0.29, "learning_rate": 2.8000340287719135e-05, "loss": 0.2081, "step": 12370 }, { "epoch": 0.29, "learning_rate": 2.79994602332731e-05, "loss": 0.1491, "step": 12375 }, { "epoch": 0.29, "learning_rate": 2.7998580178827065e-05, "loss": 0.3475, "step": 12380 }, { "epoch": 0.29, "learning_rate": 2.799770012438103e-05, "loss": 0.3543, "step": 12385 }, { "epoch": 0.29, "learning_rate": 2.7996820069934995e-05, "loss": 0.4486, "step": 12390 }, { "epoch": 0.29, "learning_rate": 2.799594001548896e-05, "loss": 0.49, "step": 12395 }, { "epoch": 0.29, "learning_rate": 2.799505996104292e-05, "loss": 0.6565, "step": 12400 }, { "epoch": 0.29, "learning_rate": 2.799417990659689e-05, "loss": 0.2942, "step": 12405 }, { "epoch": 0.29, "learning_rate": 2.7993299852150855e-05, "loss": 0.1581, "step": 12410 }, { "epoch": 0.29, "learning_rate": 2.799241979770482e-05, "loss": 0.3436, "step": 12415 }, { "epoch": 0.29, "learning_rate": 2.799153974325878e-05, "loss": 0.1976, "step": 12420 }, { "epoch": 0.29, "learning_rate": 2.799065968881275e-05, "loss": 0.3135, "step": 12425 }, { "epoch": 0.29, "learning_rate": 2.7989779634366715e-05, "loss": 0.2219, "step": 12430 }, { "epoch": 0.29, "learning_rate": 2.7988899579920677e-05, "loss": 0.3493, "step": 12435 }, { "epoch": 0.29, "learning_rate": 2.798801952547464e-05, "loss": 0.4565, "step": 12440 }, { "epoch": 0.29, "learning_rate": 2.798713947102861e-05, "loss": 0.5992, "step": 12445 }, { "epoch": 0.29, "learning_rate": 2.7986259416582575e-05, "loss": 0.5755, "step": 12450 }, { "epoch": 0.29, "learning_rate": 2.7985379362136537e-05, "loss": 0.31, "step": 12455 }, { "epoch": 0.29, "learning_rate": 2.7984499307690502e-05, "loss": 0.1615, "step": 12460 }, { "epoch": 0.29, "learning_rate": 2.798361925324447e-05, "loss": 0.1304, "step": 12465 }, { "epoch": 0.29, "learning_rate": 2.7982739198798432e-05, "loss": 0.1832, "step": 12470 }, { "epoch": 0.29, "learning_rate": 2.7981859144352397e-05, "loss": 0.1382, "step": 12475 }, { "epoch": 0.29, "learning_rate": 2.7980979089906362e-05, "loss": 0.2659, "step": 12480 }, { "epoch": 0.29, "learning_rate": 2.798009903546033e-05, "loss": 0.2879, "step": 12485 }, { "epoch": 0.29, "learning_rate": 2.7979218981014292e-05, "loss": 0.3527, "step": 12490 }, { "epoch": 0.29, "learning_rate": 2.7978338926568257e-05, "loss": 0.4257, "step": 12495 }, { "epoch": 0.29, "learning_rate": 2.7977458872122222e-05, "loss": 0.6792, "step": 12500 }, { "epoch": 0.29, "learning_rate": 2.7976578817676187e-05, "loss": 0.4047, "step": 12505 }, { "epoch": 0.29, "learning_rate": 2.7975698763230152e-05, "loss": 0.1614, "step": 12510 }, { "epoch": 0.29, "learning_rate": 2.7974818708784117e-05, "loss": 0.164, "step": 12515 }, { "epoch": 0.29, "learning_rate": 2.7973938654338085e-05, "loss": 0.2052, "step": 12520 }, { "epoch": 0.29, "learning_rate": 2.7973058599892047e-05, "loss": 0.1812, "step": 12525 }, { "epoch": 0.29, "learning_rate": 2.7972178545446012e-05, "loss": 0.2007, "step": 12530 }, { "epoch": 0.29, "learning_rate": 2.7971298490999977e-05, "loss": 0.4333, "step": 12535 }, { "epoch": 0.29, "learning_rate": 2.7970418436553942e-05, "loss": 0.5193, "step": 12540 }, { "epoch": 0.29, "learning_rate": 2.7969538382107907e-05, "loss": 0.5229, "step": 12545 }, { "epoch": 0.29, "learning_rate": 2.7968658327661872e-05, "loss": 0.5298, "step": 12550 }, { "epoch": 0.29, "learning_rate": 2.7967778273215837e-05, "loss": 0.1781, "step": 12555 }, { "epoch": 0.29, "learning_rate": 2.7966898218769802e-05, "loss": 0.2062, "step": 12560 }, { "epoch": 0.29, "learning_rate": 2.7966018164323767e-05, "loss": 0.1232, "step": 12565 }, { "epoch": 0.29, "learning_rate": 2.7965138109877732e-05, "loss": 0.2825, "step": 12570 }, { "epoch": 0.29, "learning_rate": 2.7964258055431694e-05, "loss": 0.272, "step": 12575 }, { "epoch": 0.29, "learning_rate": 2.7963378000985662e-05, "loss": 0.2447, "step": 12580 }, { "epoch": 0.29, "learning_rate": 2.7962497946539627e-05, "loss": 0.491, "step": 12585 }, { "epoch": 0.29, "learning_rate": 2.7961617892093592e-05, "loss": 0.3482, "step": 12590 }, { "epoch": 0.29, "learning_rate": 2.7960737837647554e-05, "loss": 0.5131, "step": 12595 }, { "epoch": 0.29, "learning_rate": 2.7959857783201522e-05, "loss": 0.813, "step": 12600 }, { "epoch": 0.29, "learning_rate": 2.7958977728755487e-05, "loss": 0.2786, "step": 12605 }, { "epoch": 0.29, "learning_rate": 2.7958097674309452e-05, "loss": 0.1645, "step": 12610 }, { "epoch": 0.29, "learning_rate": 2.7957217619863414e-05, "loss": 0.1645, "step": 12615 }, { "epoch": 0.29, "learning_rate": 2.7956337565417382e-05, "loss": 0.1452, "step": 12620 }, { "epoch": 0.29, "learning_rate": 2.7955457510971347e-05, "loss": 0.2779, "step": 12625 }, { "epoch": 0.29, "learning_rate": 2.795457745652531e-05, "loss": 0.3474, "step": 12630 }, { "epoch": 0.29, "learning_rate": 2.7953697402079274e-05, "loss": 0.322, "step": 12635 }, { "epoch": 0.29, "learning_rate": 2.7952817347633242e-05, "loss": 0.495, "step": 12640 }, { "epoch": 0.3, "learning_rate": 2.7951937293187207e-05, "loss": 0.4553, "step": 12645 }, { "epoch": 0.3, "learning_rate": 2.795105723874117e-05, "loss": 0.5551, "step": 12650 }, { "epoch": 0.3, "learning_rate": 2.7950177184295138e-05, "loss": 0.2997, "step": 12655 }, { "epoch": 0.3, "learning_rate": 2.7949297129849103e-05, "loss": 0.1183, "step": 12660 }, { "epoch": 0.3, "learning_rate": 2.7948417075403064e-05, "loss": 0.1323, "step": 12665 }, { "epoch": 0.3, "learning_rate": 2.794753702095703e-05, "loss": 0.1668, "step": 12670 }, { "epoch": 0.3, "learning_rate": 2.7946656966510998e-05, "loss": 0.2065, "step": 12675 }, { "epoch": 0.3, "learning_rate": 2.7945776912064963e-05, "loss": 0.3635, "step": 12680 }, { "epoch": 0.3, "learning_rate": 2.7944896857618924e-05, "loss": 0.4076, "step": 12685 }, { "epoch": 0.3, "learning_rate": 2.794401680317289e-05, "loss": 0.3667, "step": 12690 }, { "epoch": 0.3, "learning_rate": 2.7943136748726858e-05, "loss": 0.2419, "step": 12695 }, { "epoch": 0.3, "learning_rate": 2.7942432705170026e-05, "loss": 0.5144, "step": 12700 }, { "epoch": 0.3, "learning_rate": 2.794155265072399e-05, "loss": 0.3234, "step": 12705 }, { "epoch": 0.3, "learning_rate": 2.794067259627796e-05, "loss": 0.1498, "step": 12710 }, { "epoch": 0.3, "learning_rate": 2.7939792541831924e-05, "loss": 0.2239, "step": 12715 }, { "epoch": 0.3, "learning_rate": 2.7938912487385886e-05, "loss": 0.2719, "step": 12720 }, { "epoch": 0.3, "learning_rate": 2.793803243293985e-05, "loss": 0.3102, "step": 12725 }, { "epoch": 0.3, "learning_rate": 2.793715237849382e-05, "loss": 0.3014, "step": 12730 }, { "epoch": 0.3, "learning_rate": 2.793627232404778e-05, "loss": 0.3461, "step": 12735 }, { "epoch": 0.3, "learning_rate": 2.7935392269601746e-05, "loss": 0.3001, "step": 12740 }, { "epoch": 0.3, "learning_rate": 2.793451221515571e-05, "loss": 0.3882, "step": 12745 }, { "epoch": 0.3, "learning_rate": 2.793363216070968e-05, "loss": 0.4646, "step": 12750 }, { "epoch": 0.3, "learning_rate": 2.793275210626364e-05, "loss": 0.2479, "step": 12755 }, { "epoch": 0.3, "learning_rate": 2.7931872051817606e-05, "loss": 0.1611, "step": 12760 }, { "epoch": 0.3, "learning_rate": 2.793099199737157e-05, "loss": 0.0752, "step": 12765 }, { "epoch": 0.3, "learning_rate": 2.7930111942925536e-05, "loss": 0.2161, "step": 12770 }, { "epoch": 0.3, "learning_rate": 2.79292318884795e-05, "loss": 0.1928, "step": 12775 }, { "epoch": 0.3, "learning_rate": 2.7928351834033466e-05, "loss": 0.3097, "step": 12780 }, { "epoch": 0.3, "learning_rate": 2.792747177958743e-05, "loss": 0.2177, "step": 12785 }, { "epoch": 0.3, "learning_rate": 2.7926591725141396e-05, "loss": 0.3652, "step": 12790 }, { "epoch": 0.3, "learning_rate": 2.792571167069536e-05, "loss": 0.7508, "step": 12795 }, { "epoch": 0.3, "learning_rate": 2.7924831616249326e-05, "loss": 0.7232, "step": 12800 }, { "epoch": 0.3, "learning_rate": 2.792395156180329e-05, "loss": 0.4334, "step": 12805 }, { "epoch": 0.3, "learning_rate": 2.7923071507357256e-05, "loss": 0.1375, "step": 12810 }, { "epoch": 0.3, "learning_rate": 2.792219145291122e-05, "loss": 0.2568, "step": 12815 }, { "epoch": 0.3, "learning_rate": 2.7921311398465186e-05, "loss": 0.3135, "step": 12820 }, { "epoch": 0.3, "learning_rate": 2.7920431344019148e-05, "loss": 0.2726, "step": 12825 }, { "epoch": 0.3, "learning_rate": 2.7919551289573116e-05, "loss": 0.1766, "step": 12830 }, { "epoch": 0.3, "learning_rate": 2.791867123512708e-05, "loss": 0.3048, "step": 12835 }, { "epoch": 0.3, "learning_rate": 2.7917791180681046e-05, "loss": 0.488, "step": 12840 }, { "epoch": 0.3, "learning_rate": 2.791691112623501e-05, "loss": 0.5205, "step": 12845 }, { "epoch": 0.3, "learning_rate": 2.7916031071788976e-05, "loss": 0.7749, "step": 12850 }, { "epoch": 0.3, "learning_rate": 2.791515101734294e-05, "loss": 0.2224, "step": 12855 }, { "epoch": 0.3, "learning_rate": 2.7914270962896903e-05, "loss": 0.0955, "step": 12860 }, { "epoch": 0.3, "learning_rate": 2.791339090845087e-05, "loss": 0.1324, "step": 12865 }, { "epoch": 0.3, "learning_rate": 2.7912510854004836e-05, "loss": 0.2891, "step": 12870 }, { "epoch": 0.3, "learning_rate": 2.79116307995588e-05, "loss": 0.2928, "step": 12875 }, { "epoch": 0.3, "learning_rate": 2.7910750745112763e-05, "loss": 0.1307, "step": 12880 }, { "epoch": 0.3, "learning_rate": 2.790987069066673e-05, "loss": 0.3757, "step": 12885 }, { "epoch": 0.3, "learning_rate": 2.7908990636220696e-05, "loss": 0.5778, "step": 12890 }, { "epoch": 0.3, "learning_rate": 2.7908110581774658e-05, "loss": 0.3364, "step": 12895 }, { "epoch": 0.3, "learning_rate": 2.7907230527328623e-05, "loss": 0.5263, "step": 12900 }, { "epoch": 0.3, "learning_rate": 2.790635047288259e-05, "loss": 0.3812, "step": 12905 }, { "epoch": 0.3, "learning_rate": 2.7905470418436556e-05, "loss": 0.102, "step": 12910 }, { "epoch": 0.3, "learning_rate": 2.7904590363990518e-05, "loss": 0.1179, "step": 12915 }, { "epoch": 0.3, "learning_rate": 2.7903710309544483e-05, "loss": 0.3065, "step": 12920 }, { "epoch": 0.3, "learning_rate": 2.790283025509845e-05, "loss": 0.2152, "step": 12925 }, { "epoch": 0.3, "learning_rate": 2.7901950200652413e-05, "loss": 0.2784, "step": 12930 }, { "epoch": 0.3, "learning_rate": 2.7901070146206378e-05, "loss": 0.3367, "step": 12935 }, { "epoch": 0.3, "learning_rate": 2.7900190091760343e-05, "loss": 0.3965, "step": 12940 }, { "epoch": 0.3, "learning_rate": 2.789931003731431e-05, "loss": 0.3167, "step": 12945 }, { "epoch": 0.3, "learning_rate": 2.7898429982868273e-05, "loss": 0.9852, "step": 12950 }, { "epoch": 0.3, "learning_rate": 2.7897549928422238e-05, "loss": 0.3741, "step": 12955 }, { "epoch": 0.3, "learning_rate": 2.7896669873976203e-05, "loss": 0.2856, "step": 12960 }, { "epoch": 0.3, "learning_rate": 2.789578981953017e-05, "loss": 0.1326, "step": 12965 }, { "epoch": 0.3, "learning_rate": 2.7894909765084133e-05, "loss": 0.1851, "step": 12970 }, { "epoch": 0.3, "learning_rate": 2.78940297106381e-05, "loss": 0.252, "step": 12975 }, { "epoch": 0.3, "learning_rate": 2.7893149656192067e-05, "loss": 0.2627, "step": 12980 }, { "epoch": 0.3, "learning_rate": 2.789226960174603e-05, "loss": 0.3421, "step": 12985 }, { "epoch": 0.3, "learning_rate": 2.7891389547299993e-05, "loss": 0.4035, "step": 12990 }, { "epoch": 0.3, "learning_rate": 2.789050949285396e-05, "loss": 0.5278, "step": 12995 }, { "epoch": 0.3, "learning_rate": 2.7889629438407923e-05, "loss": 0.608, "step": 13000 }, { "epoch": 0.3, "learning_rate": 2.788874938396189e-05, "loss": 0.3211, "step": 13005 }, { "epoch": 0.3, "learning_rate": 2.7887869329515854e-05, "loss": 0.1133, "step": 13010 }, { "epoch": 0.3, "learning_rate": 2.788698927506982e-05, "loss": 0.0782, "step": 13015 }, { "epoch": 0.3, "learning_rate": 2.7886109220623784e-05, "loss": 0.1852, "step": 13020 }, { "epoch": 0.3, "learning_rate": 2.788522916617775e-05, "loss": 0.1569, "step": 13025 }, { "epoch": 0.3, "learning_rate": 2.7884349111731714e-05, "loss": 0.1864, "step": 13030 }, { "epoch": 0.3, "learning_rate": 2.7883469057285675e-05, "loss": 0.3056, "step": 13035 }, { "epoch": 0.3, "learning_rate": 2.7882589002839644e-05, "loss": 0.3585, "step": 13040 }, { "epoch": 0.3, "learning_rate": 2.788170894839361e-05, "loss": 0.3884, "step": 13045 }, { "epoch": 0.3, "learning_rate": 2.7880828893947574e-05, "loss": 0.7577, "step": 13050 }, { "epoch": 0.3, "learning_rate": 2.7879948839501535e-05, "loss": 0.2693, "step": 13055 }, { "epoch": 0.3, "learning_rate": 2.7879068785055504e-05, "loss": 0.0804, "step": 13060 }, { "epoch": 0.3, "learning_rate": 2.787818873060947e-05, "loss": 0.2131, "step": 13065 }, { "epoch": 0.3, "learning_rate": 2.787730867616343e-05, "loss": 0.2262, "step": 13070 }, { "epoch": 0.31, "learning_rate": 2.7876428621717395e-05, "loss": 0.4461, "step": 13075 }, { "epoch": 0.31, "learning_rate": 2.7875548567271364e-05, "loss": 0.3198, "step": 13080 }, { "epoch": 0.31, "learning_rate": 2.787466851282533e-05, "loss": 0.3961, "step": 13085 }, { "epoch": 0.31, "learning_rate": 2.787378845837929e-05, "loss": 0.6131, "step": 13090 }, { "epoch": 0.31, "learning_rate": 2.7872908403933255e-05, "loss": 0.5524, "step": 13095 }, { "epoch": 0.31, "learning_rate": 2.7872028349487224e-05, "loss": 0.7786, "step": 13100 }, { "epoch": 0.31, "learning_rate": 2.787114829504119e-05, "loss": 0.38, "step": 13105 }, { "epoch": 0.31, "learning_rate": 2.787026824059515e-05, "loss": 0.1649, "step": 13110 }, { "epoch": 0.31, "learning_rate": 2.786938818614912e-05, "loss": 0.1509, "step": 13115 }, { "epoch": 0.31, "learning_rate": 2.7868508131703084e-05, "loss": 0.2165, "step": 13120 }, { "epoch": 0.31, "learning_rate": 2.7867628077257046e-05, "loss": 0.132, "step": 13125 }, { "epoch": 0.31, "learning_rate": 2.786674802281101e-05, "loss": 0.2282, "step": 13130 }, { "epoch": 0.31, "learning_rate": 2.786586796836498e-05, "loss": 0.4499, "step": 13135 }, { "epoch": 0.31, "learning_rate": 2.7864987913918944e-05, "loss": 0.4389, "step": 13140 }, { "epoch": 0.31, "learning_rate": 2.7864107859472906e-05, "loss": 0.4045, "step": 13145 }, { "epoch": 0.31, "learning_rate": 2.786322780502687e-05, "loss": 0.8634, "step": 13150 }, { "epoch": 0.31, "learning_rate": 2.786234775058084e-05, "loss": 0.3788, "step": 13155 }, { "epoch": 0.31, "learning_rate": 2.78614676961348e-05, "loss": 0.2077, "step": 13160 }, { "epoch": 0.31, "learning_rate": 2.7860587641688766e-05, "loss": 0.1781, "step": 13165 }, { "epoch": 0.31, "learning_rate": 2.785970758724273e-05, "loss": 0.219, "step": 13170 }, { "epoch": 0.31, "learning_rate": 2.78588275327967e-05, "loss": 0.2265, "step": 13175 }, { "epoch": 0.31, "learning_rate": 2.785794747835066e-05, "loss": 0.2055, "step": 13180 }, { "epoch": 0.31, "learning_rate": 2.7857067423904626e-05, "loss": 0.2377, "step": 13185 }, { "epoch": 0.31, "learning_rate": 2.785618736945859e-05, "loss": 0.3135, "step": 13190 }, { "epoch": 0.31, "learning_rate": 2.7855307315012556e-05, "loss": 0.4486, "step": 13195 }, { "epoch": 0.31, "learning_rate": 2.785442726056652e-05, "loss": 1.0171, "step": 13200 }, { "epoch": 0.31, "learning_rate": 2.7853547206120486e-05, "loss": 0.323, "step": 13205 }, { "epoch": 0.31, "learning_rate": 2.785266715167445e-05, "loss": 0.1057, "step": 13210 }, { "epoch": 0.31, "learning_rate": 2.7851787097228416e-05, "loss": 0.1419, "step": 13215 }, { "epoch": 0.31, "learning_rate": 2.785090704278238e-05, "loss": 0.3628, "step": 13220 }, { "epoch": 0.31, "learning_rate": 2.7850026988336346e-05, "loss": 0.2618, "step": 13225 }, { "epoch": 0.31, "learning_rate": 2.7849146933890308e-05, "loss": 0.311, "step": 13230 }, { "epoch": 0.31, "learning_rate": 2.7848266879444276e-05, "loss": 0.1999, "step": 13235 }, { "epoch": 0.31, "learning_rate": 2.784738682499824e-05, "loss": 0.3791, "step": 13240 }, { "epoch": 0.31, "learning_rate": 2.7846506770552206e-05, "loss": 0.3062, "step": 13245 }, { "epoch": 0.31, "learning_rate": 2.784562671610617e-05, "loss": 0.5375, "step": 13250 }, { "epoch": 0.31, "learning_rate": 2.7844746661660136e-05, "loss": 0.2174, "step": 13255 }, { "epoch": 0.31, "learning_rate": 2.78438666072141e-05, "loss": 0.0973, "step": 13260 }, { "epoch": 0.31, "learning_rate": 2.7842986552768063e-05, "loss": 0.1054, "step": 13265 }, { "epoch": 0.31, "learning_rate": 2.784210649832203e-05, "loss": 0.1389, "step": 13270 }, { "epoch": 0.31, "learning_rate": 2.7841226443875996e-05, "loss": 0.3601, "step": 13275 }, { "epoch": 0.31, "learning_rate": 2.784034638942996e-05, "loss": 0.2888, "step": 13280 }, { "epoch": 0.31, "learning_rate": 2.7839466334983923e-05, "loss": 0.198, "step": 13285 }, { "epoch": 0.31, "learning_rate": 2.783858628053789e-05, "loss": 0.3494, "step": 13290 }, { "epoch": 0.31, "learning_rate": 2.7837706226091856e-05, "loss": 0.4418, "step": 13295 }, { "epoch": 0.31, "learning_rate": 2.7836826171645818e-05, "loss": 0.4755, "step": 13300 }, { "epoch": 0.31, "learning_rate": 2.7835946117199783e-05, "loss": 0.3414, "step": 13305 }, { "epoch": 0.31, "learning_rate": 2.783506606275375e-05, "loss": 0.1251, "step": 13310 }, { "epoch": 0.31, "learning_rate": 2.7834186008307716e-05, "loss": 0.1592, "step": 13315 }, { "epoch": 0.31, "learning_rate": 2.7833305953861678e-05, "loss": 0.2218, "step": 13320 }, { "epoch": 0.31, "learning_rate": 2.7832425899415643e-05, "loss": 0.235, "step": 13325 }, { "epoch": 0.31, "learning_rate": 2.783154584496961e-05, "loss": 0.2697, "step": 13330 }, { "epoch": 0.31, "learning_rate": 2.7830665790523573e-05, "loss": 0.2731, "step": 13335 }, { "epoch": 0.31, "learning_rate": 2.7829785736077538e-05, "loss": 0.3905, "step": 13340 }, { "epoch": 0.31, "learning_rate": 2.7828905681631503e-05, "loss": 0.3769, "step": 13345 }, { "epoch": 0.31, "learning_rate": 2.782802562718547e-05, "loss": 0.9329, "step": 13350 }, { "epoch": 0.31, "learning_rate": 2.7827145572739433e-05, "loss": 0.3222, "step": 13355 }, { "epoch": 0.31, "learning_rate": 2.7826265518293398e-05, "loss": 0.0698, "step": 13360 }, { "epoch": 0.31, "learning_rate": 2.7825385463847363e-05, "loss": 0.156, "step": 13365 }, { "epoch": 0.31, "learning_rate": 2.7824505409401328e-05, "loss": 0.2297, "step": 13370 }, { "epoch": 0.31, "learning_rate": 2.7823625354955293e-05, "loss": 0.4599, "step": 13375 }, { "epoch": 0.31, "learning_rate": 2.7822745300509258e-05, "loss": 0.2552, "step": 13380 }, { "epoch": 0.31, "learning_rate": 2.7821865246063227e-05, "loss": 0.4086, "step": 13385 }, { "epoch": 0.31, "learning_rate": 2.7820985191617188e-05, "loss": 0.3773, "step": 13390 }, { "epoch": 0.31, "learning_rate": 2.7820105137171153e-05, "loss": 0.2524, "step": 13395 }, { "epoch": 0.31, "learning_rate": 2.781922508272512e-05, "loss": 0.4724, "step": 13400 }, { "epoch": 0.31, "learning_rate": 2.7818345028279087e-05, "loss": 0.2934, "step": 13405 }, { "epoch": 0.31, "learning_rate": 2.781746497383305e-05, "loss": 0.1043, "step": 13410 }, { "epoch": 0.31, "learning_rate": 2.7816584919387013e-05, "loss": 0.0571, "step": 13415 }, { "epoch": 0.31, "learning_rate": 2.781570486494098e-05, "loss": 0.1631, "step": 13420 }, { "epoch": 0.31, "learning_rate": 2.7814824810494943e-05, "loss": 0.2634, "step": 13425 }, { "epoch": 0.31, "learning_rate": 2.781394475604891e-05, "loss": 0.2653, "step": 13430 }, { "epoch": 0.31, "learning_rate": 2.7813064701602873e-05, "loss": 0.3546, "step": 13435 }, { "epoch": 0.31, "learning_rate": 2.781218464715684e-05, "loss": 0.4539, "step": 13440 }, { "epoch": 0.31, "learning_rate": 2.7811304592710804e-05, "loss": 0.4561, "step": 13445 }, { "epoch": 0.31, "learning_rate": 2.781042453826477e-05, "loss": 0.91, "step": 13450 }, { "epoch": 0.31, "learning_rate": 2.7809544483818734e-05, "loss": 0.3893, "step": 13455 }, { "epoch": 0.31, "learning_rate": 2.7808664429372695e-05, "loss": 0.1129, "step": 13460 }, { "epoch": 0.31, "learning_rate": 2.7807784374926664e-05, "loss": 0.1995, "step": 13465 }, { "epoch": 0.31, "learning_rate": 2.780690432048063e-05, "loss": 0.1764, "step": 13470 }, { "epoch": 0.31, "learning_rate": 2.7806024266034594e-05, "loss": 0.2148, "step": 13475 }, { "epoch": 0.31, "learning_rate": 2.7805144211588555e-05, "loss": 0.2094, "step": 13480 }, { "epoch": 0.31, "learning_rate": 2.7804264157142524e-05, "loss": 0.4122, "step": 13485 }, { "epoch": 0.31, "learning_rate": 2.780338410269649e-05, "loss": 0.4215, "step": 13490 }, { "epoch": 0.31, "learning_rate": 2.780250404825045e-05, "loss": 0.4385, "step": 13495 }, { "epoch": 0.31, "learning_rate": 2.7801623993804415e-05, "loss": 0.7161, "step": 13500 }, { "epoch": 0.32, "learning_rate": 2.7800743939358384e-05, "loss": 0.2145, "step": 13505 }, { "epoch": 0.32, "learning_rate": 2.779986388491235e-05, "loss": 0.2215, "step": 13510 }, { "epoch": 0.32, "learning_rate": 2.779898383046631e-05, "loss": 0.2703, "step": 13515 }, { "epoch": 0.32, "learning_rate": 2.779810377602028e-05, "loss": 0.1024, "step": 13520 }, { "epoch": 0.32, "learning_rate": 2.7797223721574244e-05, "loss": 0.31, "step": 13525 }, { "epoch": 0.32, "learning_rate": 2.7796343667128205e-05, "loss": 0.1845, "step": 13530 }, { "epoch": 0.32, "learning_rate": 2.779546361268217e-05, "loss": 0.4877, "step": 13535 }, { "epoch": 0.32, "learning_rate": 2.779458355823614e-05, "loss": 0.3374, "step": 13540 }, { "epoch": 0.32, "learning_rate": 2.7793703503790104e-05, "loss": 0.553, "step": 13545 }, { "epoch": 0.32, "learning_rate": 2.7792823449344066e-05, "loss": 0.922, "step": 13550 }, { "epoch": 0.32, "learning_rate": 2.779194339489803e-05, "loss": 0.3471, "step": 13555 }, { "epoch": 0.32, "learning_rate": 2.7791063340452e-05, "loss": 0.1223, "step": 13560 }, { "epoch": 0.32, "learning_rate": 2.779018328600596e-05, "loss": 0.1977, "step": 13565 }, { "epoch": 0.32, "learning_rate": 2.7789303231559926e-05, "loss": 0.3053, "step": 13570 }, { "epoch": 0.32, "learning_rate": 2.778842317711389e-05, "loss": 0.2223, "step": 13575 }, { "epoch": 0.32, "learning_rate": 2.778754312266786e-05, "loss": 0.228, "step": 13580 }, { "epoch": 0.32, "learning_rate": 2.778666306822182e-05, "loss": 0.3342, "step": 13585 }, { "epoch": 0.32, "learning_rate": 2.7785783013775786e-05, "loss": 0.5121, "step": 13590 }, { "epoch": 0.32, "learning_rate": 2.778490295932975e-05, "loss": 0.5148, "step": 13595 }, { "epoch": 0.32, "learning_rate": 2.7784022904883716e-05, "loss": 0.4754, "step": 13600 }, { "epoch": 0.32, "learning_rate": 2.778314285043768e-05, "loss": 0.3291, "step": 13605 }, { "epoch": 0.32, "learning_rate": 2.7782262795991646e-05, "loss": 0.1093, "step": 13610 }, { "epoch": 0.32, "learning_rate": 2.778138274154561e-05, "loss": 0.1759, "step": 13615 }, { "epoch": 0.32, "learning_rate": 2.7780502687099576e-05, "loss": 0.1358, "step": 13620 }, { "epoch": 0.32, "learning_rate": 2.777962263265354e-05, "loss": 0.1688, "step": 13625 }, { "epoch": 0.32, "learning_rate": 2.7778742578207506e-05, "loss": 0.2344, "step": 13630 }, { "epoch": 0.32, "learning_rate": 2.7777862523761468e-05, "loss": 0.258, "step": 13635 }, { "epoch": 0.32, "learning_rate": 2.7776982469315436e-05, "loss": 0.2527, "step": 13640 }, { "epoch": 0.32, "learning_rate": 2.77761024148694e-05, "loss": 0.2908, "step": 13645 }, { "epoch": 0.32, "learning_rate": 2.7775222360423366e-05, "loss": 0.6322, "step": 13650 }, { "epoch": 0.32, "learning_rate": 2.777434230597733e-05, "loss": 0.29, "step": 13655 }, { "epoch": 0.32, "learning_rate": 2.7773462251531296e-05, "loss": 0.2242, "step": 13660 }, { "epoch": 0.32, "learning_rate": 2.777258219708526e-05, "loss": 0.1728, "step": 13665 }, { "epoch": 0.32, "learning_rate": 2.7771702142639223e-05, "loss": 0.2581, "step": 13670 }, { "epoch": 0.32, "learning_rate": 2.777082208819319e-05, "loss": 0.2355, "step": 13675 }, { "epoch": 0.32, "learning_rate": 2.7769942033747156e-05, "loss": 0.3538, "step": 13680 }, { "epoch": 0.32, "learning_rate": 2.776906197930112e-05, "loss": 0.3329, "step": 13685 }, { "epoch": 0.32, "learning_rate": 2.7768181924855083e-05, "loss": 0.3935, "step": 13690 }, { "epoch": 0.32, "learning_rate": 2.776730187040905e-05, "loss": 0.5932, "step": 13695 }, { "epoch": 0.32, "learning_rate": 2.7766421815963016e-05, "loss": 0.7401, "step": 13700 }, { "epoch": 0.32, "learning_rate": 2.776554176151698e-05, "loss": 0.3232, "step": 13705 }, { "epoch": 0.32, "learning_rate": 2.7764661707070943e-05, "loss": 0.1578, "step": 13710 }, { "epoch": 0.32, "learning_rate": 2.776378165262491e-05, "loss": 0.174, "step": 13715 }, { "epoch": 0.32, "learning_rate": 2.7762901598178876e-05, "loss": 0.2773, "step": 13720 }, { "epoch": 0.32, "learning_rate": 2.7762021543732838e-05, "loss": 0.3029, "step": 13725 }, { "epoch": 0.32, "learning_rate": 2.7761141489286803e-05, "loss": 0.2379, "step": 13730 }, { "epoch": 0.32, "learning_rate": 2.776026143484077e-05, "loss": 0.2705, "step": 13735 }, { "epoch": 0.32, "learning_rate": 2.7759381380394736e-05, "loss": 0.5616, "step": 13740 }, { "epoch": 0.32, "learning_rate": 2.7758501325948698e-05, "loss": 0.579, "step": 13745 }, { "epoch": 0.32, "learning_rate": 2.7757621271502663e-05, "loss": 0.8582, "step": 13750 }, { "epoch": 0.32, "learning_rate": 2.775674121705663e-05, "loss": 0.3192, "step": 13755 }, { "epoch": 0.32, "learning_rate": 2.7755861162610593e-05, "loss": 0.1003, "step": 13760 }, { "epoch": 0.32, "learning_rate": 2.7754981108164558e-05, "loss": 0.1618, "step": 13765 }, { "epoch": 0.32, "learning_rate": 2.7754101053718523e-05, "loss": 0.2372, "step": 13770 }, { "epoch": 0.32, "learning_rate": 2.775322099927249e-05, "loss": 0.2772, "step": 13775 }, { "epoch": 0.32, "learning_rate": 2.7752340944826453e-05, "loss": 0.2733, "step": 13780 }, { "epoch": 0.32, "learning_rate": 2.7751460890380418e-05, "loss": 0.3074, "step": 13785 }, { "epoch": 0.32, "learning_rate": 2.7750580835934387e-05, "loss": 0.3586, "step": 13790 }, { "epoch": 0.32, "learning_rate": 2.7749700781488348e-05, "loss": 0.4007, "step": 13795 }, { "epoch": 0.32, "learning_rate": 2.7748820727042313e-05, "loss": 0.7982, "step": 13800 }, { "epoch": 0.32, "learning_rate": 2.7747940672596278e-05, "loss": 0.3044, "step": 13805 }, { "epoch": 0.32, "learning_rate": 2.7747060618150247e-05, "loss": 0.2306, "step": 13810 }, { "epoch": 0.32, "learning_rate": 2.7746180563704208e-05, "loss": 0.1953, "step": 13815 }, { "epoch": 0.32, "learning_rate": 2.7745300509258173e-05, "loss": 0.2604, "step": 13820 }, { "epoch": 0.32, "learning_rate": 2.774442045481214e-05, "loss": 0.1978, "step": 13825 }, { "epoch": 0.32, "learning_rate": 2.7743540400366103e-05, "loss": 0.2665, "step": 13830 }, { "epoch": 0.32, "learning_rate": 2.774266034592007e-05, "loss": 0.1739, "step": 13835 }, { "epoch": 0.32, "learning_rate": 2.7741780291474033e-05, "loss": 0.3908, "step": 13840 }, { "epoch": 0.32, "learning_rate": 2.7740900237028e-05, "loss": 0.3236, "step": 13845 }, { "epoch": 0.32, "learning_rate": 2.7740020182581963e-05, "loss": 0.7746, "step": 13850 }, { "epoch": 0.32, "learning_rate": 2.773914012813593e-05, "loss": 0.2428, "step": 13855 }, { "epoch": 0.32, "learning_rate": 2.7738260073689893e-05, "loss": 0.1178, "step": 13860 }, { "epoch": 0.32, "learning_rate": 2.7737380019243855e-05, "loss": 0.2111, "step": 13865 }, { "epoch": 0.32, "learning_rate": 2.7736499964797823e-05, "loss": 0.1073, "step": 13870 }, { "epoch": 0.32, "learning_rate": 2.773561991035179e-05, "loss": 0.3632, "step": 13875 }, { "epoch": 0.32, "learning_rate": 2.7734739855905754e-05, "loss": 0.1438, "step": 13880 }, { "epoch": 0.32, "learning_rate": 2.7733859801459715e-05, "loss": 0.3378, "step": 13885 }, { "epoch": 0.32, "learning_rate": 2.7732979747013684e-05, "loss": 0.3451, "step": 13890 }, { "epoch": 0.32, "learning_rate": 2.773209969256765e-05, "loss": 0.338, "step": 13895 }, { "epoch": 0.32, "learning_rate": 2.773121963812161e-05, "loss": 0.4549, "step": 13900 }, { "epoch": 0.32, "learning_rate": 2.7730339583675575e-05, "loss": 0.456, "step": 13905 }, { "epoch": 0.32, "learning_rate": 2.7729459529229544e-05, "loss": 0.1076, "step": 13910 }, { "epoch": 0.32, "learning_rate": 2.772857947478351e-05, "loss": 0.1295, "step": 13915 }, { "epoch": 0.32, "learning_rate": 2.772769942033747e-05, "loss": 0.2064, "step": 13920 }, { "epoch": 0.32, "learning_rate": 2.772681936589144e-05, "loss": 0.2172, "step": 13925 }, { "epoch": 0.33, "learning_rate": 2.7725939311445404e-05, "loss": 0.3592, "step": 13930 }, { "epoch": 0.33, "learning_rate": 2.7725059256999365e-05, "loss": 0.3445, "step": 13935 }, { "epoch": 0.33, "learning_rate": 2.772417920255333e-05, "loss": 0.2566, "step": 13940 }, { "epoch": 0.33, "learning_rate": 2.77232991481073e-05, "loss": 0.2503, "step": 13945 }, { "epoch": 0.33, "learning_rate": 2.7722419093661264e-05, "loss": 0.5287, "step": 13950 }, { "epoch": 0.33, "learning_rate": 2.7721539039215225e-05, "loss": 0.2516, "step": 13955 }, { "epoch": 0.33, "learning_rate": 2.772065898476919e-05, "loss": 0.1253, "step": 13960 }, { "epoch": 0.33, "learning_rate": 2.771977893032316e-05, "loss": 0.1907, "step": 13965 }, { "epoch": 0.33, "learning_rate": 2.771889887587712e-05, "loss": 0.1341, "step": 13970 }, { "epoch": 0.33, "learning_rate": 2.7718018821431086e-05, "loss": 0.3524, "step": 13975 }, { "epoch": 0.33, "learning_rate": 2.771713876698505e-05, "loss": 0.3145, "step": 13980 }, { "epoch": 0.33, "learning_rate": 2.771625871253902e-05, "loss": 0.456, "step": 13985 }, { "epoch": 0.33, "learning_rate": 2.771537865809298e-05, "loss": 0.4011, "step": 13990 }, { "epoch": 0.33, "learning_rate": 2.7714498603646946e-05, "loss": 0.7097, "step": 13995 }, { "epoch": 0.33, "learning_rate": 2.771361854920091e-05, "loss": 0.4175, "step": 14000 }, { "epoch": 0.33, "learning_rate": 2.771273849475488e-05, "loss": 0.2541, "step": 14005 }, { "epoch": 0.33, "learning_rate": 2.771185844030884e-05, "loss": 0.1999, "step": 14010 }, { "epoch": 0.33, "learning_rate": 2.7710978385862806e-05, "loss": 0.1238, "step": 14015 }, { "epoch": 0.33, "learning_rate": 2.771009833141677e-05, "loss": 0.2107, "step": 14020 }, { "epoch": 0.33, "learning_rate": 2.7709218276970736e-05, "loss": 0.1553, "step": 14025 }, { "epoch": 0.33, "learning_rate": 2.77083382225247e-05, "loss": 0.315, "step": 14030 }, { "epoch": 0.33, "learning_rate": 2.7707458168078666e-05, "loss": 0.3788, "step": 14035 }, { "epoch": 0.33, "learning_rate": 2.770657811363263e-05, "loss": 0.406, "step": 14040 }, { "epoch": 0.33, "learning_rate": 2.7705698059186596e-05, "loss": 0.5067, "step": 14045 }, { "epoch": 0.33, "learning_rate": 2.770481800474056e-05, "loss": 0.6822, "step": 14050 }, { "epoch": 0.33, "learning_rate": 2.7703937950294526e-05, "loss": 0.3246, "step": 14055 }, { "epoch": 0.33, "learning_rate": 2.770305789584849e-05, "loss": 0.1809, "step": 14060 }, { "epoch": 0.33, "learning_rate": 2.7702177841402456e-05, "loss": 0.125, "step": 14065 }, { "epoch": 0.33, "learning_rate": 2.770129778695642e-05, "loss": 0.17, "step": 14070 }, { "epoch": 0.33, "learning_rate": 2.7700417732510386e-05, "loss": 0.2651, "step": 14075 }, { "epoch": 0.33, "learning_rate": 2.769953767806435e-05, "loss": 0.2415, "step": 14080 }, { "epoch": 0.33, "learning_rate": 2.7698657623618316e-05, "loss": 0.2301, "step": 14085 }, { "epoch": 0.33, "learning_rate": 2.769777756917228e-05, "loss": 0.3658, "step": 14090 }, { "epoch": 0.33, "learning_rate": 2.7696897514726243e-05, "loss": 0.5153, "step": 14095 }, { "epoch": 0.33, "learning_rate": 2.769601746028021e-05, "loss": 0.9753, "step": 14100 }, { "epoch": 0.33, "learning_rate": 2.7695137405834176e-05, "loss": 0.3545, "step": 14105 }, { "epoch": 0.33, "learning_rate": 2.769425735138814e-05, "loss": 0.2748, "step": 14110 }, { "epoch": 0.33, "learning_rate": 2.7693377296942103e-05, "loss": 0.1484, "step": 14115 }, { "epoch": 0.33, "learning_rate": 2.769249724249607e-05, "loss": 0.1945, "step": 14120 }, { "epoch": 0.33, "learning_rate": 2.7691617188050036e-05, "loss": 0.2323, "step": 14125 }, { "epoch": 0.33, "learning_rate": 2.7690737133603998e-05, "loss": 0.3581, "step": 14130 }, { "epoch": 0.33, "learning_rate": 2.7689857079157963e-05, "loss": 0.3393, "step": 14135 }, { "epoch": 0.33, "learning_rate": 2.768897702471193e-05, "loss": 0.1962, "step": 14140 }, { "epoch": 0.33, "learning_rate": 2.7688096970265896e-05, "loss": 0.4445, "step": 14145 }, { "epoch": 0.33, "learning_rate": 2.7687216915819858e-05, "loss": 0.6982, "step": 14150 }, { "epoch": 0.33, "learning_rate": 2.7686336861373823e-05, "loss": 0.3917, "step": 14155 }, { "epoch": 0.33, "learning_rate": 2.768545680692779e-05, "loss": 0.0869, "step": 14160 }, { "epoch": 0.33, "learning_rate": 2.7684576752481753e-05, "loss": 0.3195, "step": 14165 }, { "epoch": 0.33, "learning_rate": 2.7683696698035718e-05, "loss": 0.2118, "step": 14170 }, { "epoch": 0.33, "learning_rate": 2.7682816643589683e-05, "loss": 0.1949, "step": 14175 }, { "epoch": 0.33, "learning_rate": 2.768193658914365e-05, "loss": 0.2188, "step": 14180 }, { "epoch": 0.33, "learning_rate": 2.7681056534697613e-05, "loss": 0.3185, "step": 14185 }, { "epoch": 0.33, "learning_rate": 2.7680176480251578e-05, "loss": 0.4398, "step": 14190 }, { "epoch": 0.33, "learning_rate": 2.7679296425805546e-05, "loss": 0.5882, "step": 14195 }, { "epoch": 0.33, "learning_rate": 2.7678416371359508e-05, "loss": 0.7873, "step": 14200 }, { "epoch": 0.33, "learning_rate": 2.7677536316913473e-05, "loss": 0.3825, "step": 14205 }, { "epoch": 0.33, "learning_rate": 2.7676656262467438e-05, "loss": 0.0549, "step": 14210 }, { "epoch": 0.33, "learning_rate": 2.7675776208021407e-05, "loss": 0.1147, "step": 14215 }, { "epoch": 0.33, "learning_rate": 2.7674896153575368e-05, "loss": 0.2934, "step": 14220 }, { "epoch": 0.33, "learning_rate": 2.7674016099129333e-05, "loss": 0.2085, "step": 14225 }, { "epoch": 0.33, "learning_rate": 2.7673136044683298e-05, "loss": 0.2205, "step": 14230 }, { "epoch": 0.33, "learning_rate": 2.7672255990237263e-05, "loss": 0.3592, "step": 14235 }, { "epoch": 0.33, "learning_rate": 2.7671375935791228e-05, "loss": 0.551, "step": 14240 }, { "epoch": 0.33, "learning_rate": 2.7670495881345193e-05, "loss": 0.4595, "step": 14245 }, { "epoch": 0.33, "learning_rate": 2.7669615826899158e-05, "loss": 1.104, "step": 14250 }, { "epoch": 0.33, "learning_rate": 2.7668735772453123e-05, "loss": 0.3188, "step": 14255 }, { "epoch": 0.33, "learning_rate": 2.766785571800709e-05, "loss": 0.1756, "step": 14260 }, { "epoch": 0.33, "learning_rate": 2.7666975663561053e-05, "loss": 0.1225, "step": 14265 }, { "epoch": 0.33, "learning_rate": 2.7666095609115015e-05, "loss": 0.2702, "step": 14270 }, { "epoch": 0.33, "learning_rate": 2.7665215554668983e-05, "loss": 0.3046, "step": 14275 }, { "epoch": 0.33, "learning_rate": 2.766433550022295e-05, "loss": 0.2201, "step": 14280 }, { "epoch": 0.33, "learning_rate": 2.7663455445776913e-05, "loss": 0.299, "step": 14285 }, { "epoch": 0.33, "learning_rate": 2.7662575391330875e-05, "loss": 0.238, "step": 14290 }, { "epoch": 0.33, "learning_rate": 2.7661695336884843e-05, "loss": 0.5421, "step": 14295 }, { "epoch": 0.33, "learning_rate": 2.766081528243881e-05, "loss": 0.6436, "step": 14300 }, { "epoch": 0.33, "learning_rate": 2.765993522799277e-05, "loss": 0.2091, "step": 14305 }, { "epoch": 0.33, "learning_rate": 2.7659055173546735e-05, "loss": 0.1562, "step": 14310 }, { "epoch": 0.33, "learning_rate": 2.7658175119100704e-05, "loss": 0.1683, "step": 14315 }, { "epoch": 0.33, "learning_rate": 2.765729506465467e-05, "loss": 0.1799, "step": 14320 }, { "epoch": 0.33, "learning_rate": 2.765641501020863e-05, "loss": 0.1751, "step": 14325 }, { "epoch": 0.33, "learning_rate": 2.76555349557626e-05, "loss": 0.3052, "step": 14330 }, { "epoch": 0.33, "learning_rate": 2.7654654901316564e-05, "loss": 0.5174, "step": 14335 }, { "epoch": 0.33, "learning_rate": 2.765377484687053e-05, "loss": 0.3899, "step": 14340 }, { "epoch": 0.33, "learning_rate": 2.765289479242449e-05, "loss": 0.4651, "step": 14345 }, { "epoch": 0.33, "learning_rate": 2.765201473797846e-05, "loss": 0.7602, "step": 14350 }, { "epoch": 0.33, "learning_rate": 2.7651134683532424e-05, "loss": 0.308, "step": 14355 }, { "epoch": 0.34, "learning_rate": 2.7650254629086385e-05, "loss": 0.1475, "step": 14360 }, { "epoch": 0.34, "learning_rate": 2.764937457464035e-05, "loss": 0.1953, "step": 14365 }, { "epoch": 0.34, "learning_rate": 2.764849452019432e-05, "loss": 0.1298, "step": 14370 }, { "epoch": 0.34, "learning_rate": 2.7647614465748284e-05, "loss": 0.2158, "step": 14375 }, { "epoch": 0.34, "learning_rate": 2.7646734411302245e-05, "loss": 0.2746, "step": 14380 }, { "epoch": 0.34, "learning_rate": 2.764585435685621e-05, "loss": 0.4102, "step": 14385 }, { "epoch": 0.34, "learning_rate": 2.764497430241018e-05, "loss": 0.4055, "step": 14390 }, { "epoch": 0.34, "learning_rate": 2.764409424796414e-05, "loss": 0.5981, "step": 14395 }, { "epoch": 0.34, "learning_rate": 2.7643214193518106e-05, "loss": 0.5583, "step": 14400 }, { "epoch": 0.34, "learning_rate": 2.764233413907207e-05, "loss": 0.2763, "step": 14405 }, { "epoch": 0.34, "learning_rate": 2.764145408462604e-05, "loss": 0.05, "step": 14410 }, { "epoch": 0.34, "learning_rate": 2.764057403018e-05, "loss": 0.143, "step": 14415 }, { "epoch": 0.34, "learning_rate": 2.7639693975733966e-05, "loss": 0.1999, "step": 14420 }, { "epoch": 0.34, "learning_rate": 2.763881392128793e-05, "loss": 0.235, "step": 14425 }, { "epoch": 0.34, "learning_rate": 2.7637933866841896e-05, "loss": 0.1928, "step": 14430 }, { "epoch": 0.34, "learning_rate": 2.763705381239586e-05, "loss": 0.2461, "step": 14435 }, { "epoch": 0.34, "learning_rate": 2.7636173757949826e-05, "loss": 0.3286, "step": 14440 }, { "epoch": 0.34, "learning_rate": 2.763529370350379e-05, "loss": 0.242, "step": 14445 }, { "epoch": 0.34, "learning_rate": 2.7634413649057756e-05, "loss": 0.7898, "step": 14450 }, { "epoch": 0.34, "learning_rate": 2.763353359461172e-05, "loss": 0.3025, "step": 14455 }, { "epoch": 0.34, "learning_rate": 2.7632653540165686e-05, "loss": 0.1263, "step": 14460 }, { "epoch": 0.34, "learning_rate": 2.763177348571965e-05, "loss": 0.2661, "step": 14465 }, { "epoch": 0.34, "learning_rate": 2.7630893431273616e-05, "loss": 0.2212, "step": 14470 }, { "epoch": 0.34, "learning_rate": 2.763001337682758e-05, "loss": 0.2051, "step": 14475 }, { "epoch": 0.34, "learning_rate": 2.7629133322381546e-05, "loss": 0.281, "step": 14480 }, { "epoch": 0.34, "learning_rate": 2.762825326793551e-05, "loss": 0.3155, "step": 14485 }, { "epoch": 0.34, "learning_rate": 2.7627373213489476e-05, "loss": 0.3444, "step": 14490 }, { "epoch": 0.34, "learning_rate": 2.762649315904344e-05, "loss": 0.2722, "step": 14495 }, { "epoch": 0.34, "learning_rate": 2.7625613104597403e-05, "loss": 0.458, "step": 14500 }, { "epoch": 0.34, "learning_rate": 2.762473305015137e-05, "loss": 0.3637, "step": 14505 }, { "epoch": 0.34, "learning_rate": 2.7623852995705336e-05, "loss": 0.1384, "step": 14510 }, { "epoch": 0.34, "learning_rate": 2.76229729412593e-05, "loss": 0.154, "step": 14515 }, { "epoch": 0.34, "learning_rate": 2.7622092886813263e-05, "loss": 0.1287, "step": 14520 }, { "epoch": 0.34, "learning_rate": 2.762121283236723e-05, "loss": 0.1415, "step": 14525 }, { "epoch": 0.34, "learning_rate": 2.7620332777921196e-05, "loss": 0.233, "step": 14530 }, { "epoch": 0.34, "learning_rate": 2.7619452723475158e-05, "loss": 0.269, "step": 14535 }, { "epoch": 0.34, "learning_rate": 2.7618572669029123e-05, "loss": 0.2904, "step": 14540 }, { "epoch": 0.34, "learning_rate": 2.761769261458309e-05, "loss": 0.5155, "step": 14545 }, { "epoch": 0.34, "learning_rate": 2.7616812560137056e-05, "loss": 0.501, "step": 14550 }, { "epoch": 0.34, "learning_rate": 2.7615932505691018e-05, "loss": 0.3416, "step": 14555 }, { "epoch": 0.34, "learning_rate": 2.7615052451244983e-05, "loss": 0.1551, "step": 14560 }, { "epoch": 0.34, "learning_rate": 2.761417239679895e-05, "loss": 0.201, "step": 14565 }, { "epoch": 0.34, "learning_rate": 2.7613292342352913e-05, "loss": 0.2248, "step": 14570 }, { "epoch": 0.34, "learning_rate": 2.7612412287906878e-05, "loss": 0.1501, "step": 14575 }, { "epoch": 0.34, "learning_rate": 2.7611532233460843e-05, "loss": 0.2871, "step": 14580 }, { "epoch": 0.34, "learning_rate": 2.761065217901481e-05, "loss": 0.2971, "step": 14585 }, { "epoch": 0.34, "learning_rate": 2.7609772124568773e-05, "loss": 0.3687, "step": 14590 }, { "epoch": 0.34, "learning_rate": 2.7608892070122738e-05, "loss": 0.3623, "step": 14595 }, { "epoch": 0.34, "learning_rate": 2.7608012015676706e-05, "loss": 0.6827, "step": 14600 }, { "epoch": 0.34, "learning_rate": 2.7607131961230668e-05, "loss": 0.2888, "step": 14605 }, { "epoch": 0.34, "learning_rate": 2.7606251906784633e-05, "loss": 0.0996, "step": 14610 }, { "epoch": 0.34, "learning_rate": 2.7605371852338598e-05, "loss": 0.1408, "step": 14615 }, { "epoch": 0.34, "learning_rate": 2.7604491797892566e-05, "loss": 0.1883, "step": 14620 }, { "epoch": 0.34, "learning_rate": 2.7603611743446528e-05, "loss": 0.326, "step": 14625 }, { "epoch": 0.34, "learning_rate": 2.7602731689000493e-05, "loss": 0.2062, "step": 14630 }, { "epoch": 0.34, "learning_rate": 2.7601851634554458e-05, "loss": 0.235, "step": 14635 }, { "epoch": 0.34, "learning_rate": 2.7600971580108426e-05, "loss": 0.4071, "step": 14640 }, { "epoch": 0.34, "learning_rate": 2.7600091525662388e-05, "loss": 0.3313, "step": 14645 }, { "epoch": 0.34, "learning_rate": 2.7599211471216353e-05, "loss": 0.8492, "step": 14650 }, { "epoch": 0.34, "learning_rate": 2.7598331416770318e-05, "loss": 0.2152, "step": 14655 }, { "epoch": 0.34, "learning_rate": 2.7597451362324283e-05, "loss": 0.0842, "step": 14660 }, { "epoch": 0.34, "learning_rate": 2.7596571307878248e-05, "loss": 0.2635, "step": 14665 }, { "epoch": 0.34, "learning_rate": 2.7595691253432213e-05, "loss": 0.2663, "step": 14670 }, { "epoch": 0.34, "learning_rate": 2.7594811198986178e-05, "loss": 0.2839, "step": 14675 }, { "epoch": 0.34, "learning_rate": 2.7593931144540143e-05, "loss": 0.1665, "step": 14680 }, { "epoch": 0.34, "learning_rate": 2.7593051090094108e-05, "loss": 0.2732, "step": 14685 }, { "epoch": 0.34, "learning_rate": 2.7592171035648073e-05, "loss": 0.3182, "step": 14690 }, { "epoch": 0.34, "learning_rate": 2.7591290981202035e-05, "loss": 0.4281, "step": 14695 }, { "epoch": 0.34, "learning_rate": 2.7590410926756003e-05, "loss": 0.6143, "step": 14700 }, { "epoch": 0.34, "learning_rate": 2.758953087230997e-05, "loss": 0.3539, "step": 14705 }, { "epoch": 0.34, "learning_rate": 2.7588650817863933e-05, "loss": 0.1537, "step": 14710 }, { "epoch": 0.34, "learning_rate": 2.7587770763417895e-05, "loss": 0.1286, "step": 14715 }, { "epoch": 0.34, "learning_rate": 2.7586890708971863e-05, "loss": 0.2152, "step": 14720 }, { "epoch": 0.34, "learning_rate": 2.758601065452583e-05, "loss": 0.1535, "step": 14725 }, { "epoch": 0.34, "learning_rate": 2.758513060007979e-05, "loss": 0.2131, "step": 14730 }, { "epoch": 0.34, "learning_rate": 2.758425054563376e-05, "loss": 0.2737, "step": 14735 }, { "epoch": 0.34, "learning_rate": 2.7583370491187724e-05, "loss": 0.284, "step": 14740 }, { "epoch": 0.34, "learning_rate": 2.758249043674169e-05, "loss": 0.6259, "step": 14745 }, { "epoch": 0.34, "learning_rate": 2.758161038229565e-05, "loss": 0.768, "step": 14750 }, { "epoch": 0.34, "learning_rate": 2.758073032784962e-05, "loss": 0.3979, "step": 14755 }, { "epoch": 0.34, "learning_rate": 2.7579850273403584e-05, "loss": 0.1876, "step": 14760 }, { "epoch": 0.34, "learning_rate": 2.7578970218957545e-05, "loss": 0.2376, "step": 14765 }, { "epoch": 0.34, "learning_rate": 2.757809016451151e-05, "loss": 0.1648, "step": 14770 }, { "epoch": 0.34, "learning_rate": 2.757721011006548e-05, "loss": 0.1154, "step": 14775 }, { "epoch": 0.34, "learning_rate": 2.7576330055619444e-05, "loss": 0.2155, "step": 14780 }, { "epoch": 0.34, "learning_rate": 2.7575450001173405e-05, "loss": 0.174, "step": 14785 }, { "epoch": 0.35, "learning_rate": 2.757456994672737e-05, "loss": 0.4185, "step": 14790 }, { "epoch": 0.35, "learning_rate": 2.757368989228134e-05, "loss": 0.3865, "step": 14795 }, { "epoch": 0.35, "learning_rate": 2.75728098378353e-05, "loss": 0.7976, "step": 14800 }, { "epoch": 0.35, "learning_rate": 2.7571929783389265e-05, "loss": 0.1978, "step": 14805 }, { "epoch": 0.35, "learning_rate": 2.757104972894323e-05, "loss": 0.1303, "step": 14810 }, { "epoch": 0.35, "learning_rate": 2.75701696744972e-05, "loss": 0.1154, "step": 14815 }, { "epoch": 0.35, "learning_rate": 2.756928962005116e-05, "loss": 0.1818, "step": 14820 }, { "epoch": 0.35, "learning_rate": 2.7568409565605125e-05, "loss": 0.145, "step": 14825 }, { "epoch": 0.35, "learning_rate": 2.756752951115909e-05, "loss": 0.3503, "step": 14830 }, { "epoch": 0.35, "learning_rate": 2.7566649456713056e-05, "loss": 0.2947, "step": 14835 }, { "epoch": 0.35, "learning_rate": 2.756576940226702e-05, "loss": 0.4012, "step": 14840 }, { "epoch": 0.35, "learning_rate": 2.7564889347820986e-05, "loss": 0.4056, "step": 14845 }, { "epoch": 0.35, "learning_rate": 2.756400929337495e-05, "loss": 0.4525, "step": 14850 }, { "epoch": 0.35, "learning_rate": 2.7563129238928916e-05, "loss": 0.2154, "step": 14855 }, { "epoch": 0.35, "learning_rate": 2.756224918448288e-05, "loss": 0.091, "step": 14860 }, { "epoch": 0.35, "learning_rate": 2.7561369130036846e-05, "loss": 0.2928, "step": 14865 }, { "epoch": 0.35, "learning_rate": 2.756048907559081e-05, "loss": 0.1235, "step": 14870 }, { "epoch": 0.35, "learning_rate": 2.7559609021144776e-05, "loss": 0.4614, "step": 14875 }, { "epoch": 0.35, "learning_rate": 2.755872896669874e-05, "loss": 0.4258, "step": 14880 }, { "epoch": 0.35, "learning_rate": 2.7557848912252706e-05, "loss": 0.144, "step": 14885 }, { "epoch": 0.35, "learning_rate": 2.755696885780667e-05, "loss": 0.3254, "step": 14890 }, { "epoch": 0.35, "learning_rate": 2.7556088803360636e-05, "loss": 0.4627, "step": 14895 }, { "epoch": 0.35, "learning_rate": 2.75552087489146e-05, "loss": 0.4144, "step": 14900 }, { "epoch": 0.35, "learning_rate": 2.7554328694468562e-05, "loss": 0.4187, "step": 14905 }, { "epoch": 0.35, "learning_rate": 2.755344864002253e-05, "loss": 0.0929, "step": 14910 }, { "epoch": 0.35, "learning_rate": 2.7552568585576496e-05, "loss": 0.0906, "step": 14915 }, { "epoch": 0.35, "learning_rate": 2.755168853113046e-05, "loss": 0.1706, "step": 14920 }, { "epoch": 0.35, "learning_rate": 2.7550808476684422e-05, "loss": 0.2007, "step": 14925 }, { "epoch": 0.35, "learning_rate": 2.754992842223839e-05, "loss": 0.225, "step": 14930 }, { "epoch": 0.35, "learning_rate": 2.7549048367792356e-05, "loss": 0.2842, "step": 14935 }, { "epoch": 0.35, "learning_rate": 2.754816831334632e-05, "loss": 0.3351, "step": 14940 }, { "epoch": 0.35, "learning_rate": 2.7547288258900283e-05, "loss": 0.5018, "step": 14945 }, { "epoch": 0.35, "learning_rate": 2.754640820445425e-05, "loss": 0.6362, "step": 14950 }, { "epoch": 0.35, "learning_rate": 2.7545528150008216e-05, "loss": 0.2241, "step": 14955 }, { "epoch": 0.35, "learning_rate": 2.7544648095562178e-05, "loss": 0.094, "step": 14960 }, { "epoch": 0.35, "learning_rate": 2.7543768041116143e-05, "loss": 0.1235, "step": 14965 }, { "epoch": 0.35, "learning_rate": 2.754288798667011e-05, "loss": 0.2512, "step": 14970 }, { "epoch": 0.35, "learning_rate": 2.7542007932224076e-05, "loss": 0.2322, "step": 14975 }, { "epoch": 0.35, "learning_rate": 2.7541127877778038e-05, "loss": 0.1691, "step": 14980 }, { "epoch": 0.35, "learning_rate": 2.7540247823332003e-05, "loss": 0.318, "step": 14985 }, { "epoch": 0.35, "learning_rate": 2.753936776888597e-05, "loss": 0.2502, "step": 14990 }, { "epoch": 0.35, "learning_rate": 2.7538487714439933e-05, "loss": 0.4975, "step": 14995 }, { "epoch": 0.35, "learning_rate": 2.7537607659993898e-05, "loss": 0.5823, "step": 15000 }, { "epoch": 0.35, "learning_rate": 2.7536727605547863e-05, "loss": 0.4019, "step": 15005 }, { "epoch": 0.35, "learning_rate": 2.753584755110183e-05, "loss": 0.1233, "step": 15010 }, { "epoch": 0.35, "learning_rate": 2.7534967496655793e-05, "loss": 0.0573, "step": 15015 }, { "epoch": 0.35, "learning_rate": 2.7534087442209758e-05, "loss": 0.1124, "step": 15020 }, { "epoch": 0.35, "learning_rate": 2.7533207387763726e-05, "loss": 0.3417, "step": 15025 }, { "epoch": 0.35, "learning_rate": 2.7532327333317688e-05, "loss": 0.2564, "step": 15030 }, { "epoch": 0.35, "learning_rate": 2.7531447278871653e-05, "loss": 0.2489, "step": 15035 }, { "epoch": 0.35, "learning_rate": 2.7530567224425618e-05, "loss": 0.2097, "step": 15040 }, { "epoch": 0.35, "learning_rate": 2.7529687169979586e-05, "loss": 0.4787, "step": 15045 }, { "epoch": 0.35, "learning_rate": 2.7528807115533548e-05, "loss": 0.5964, "step": 15050 }, { "epoch": 0.35, "learning_rate": 2.7527927061087513e-05, "loss": 0.2566, "step": 15055 }, { "epoch": 0.35, "learning_rate": 2.7527047006641478e-05, "loss": 0.097, "step": 15060 }, { "epoch": 0.35, "learning_rate": 2.7526166952195443e-05, "loss": 0.182, "step": 15065 }, { "epoch": 0.35, "learning_rate": 2.7525286897749408e-05, "loss": 0.1164, "step": 15070 }, { "epoch": 0.35, "learning_rate": 2.7524406843303373e-05, "loss": 0.1659, "step": 15075 }, { "epoch": 0.35, "learning_rate": 2.7523526788857338e-05, "loss": 0.258, "step": 15080 }, { "epoch": 0.35, "learning_rate": 2.7522646734411303e-05, "loss": 0.2892, "step": 15085 }, { "epoch": 0.35, "learning_rate": 2.7521766679965268e-05, "loss": 0.4548, "step": 15090 }, { "epoch": 0.35, "learning_rate": 2.7520886625519233e-05, "loss": 0.4524, "step": 15095 }, { "epoch": 0.35, "learning_rate": 2.7520006571073195e-05, "loss": 0.7174, "step": 15100 }, { "epoch": 0.35, "learning_rate": 2.7519126516627163e-05, "loss": 0.2813, "step": 15105 }, { "epoch": 0.35, "learning_rate": 2.7518246462181128e-05, "loss": 0.0802, "step": 15110 }, { "epoch": 0.35, "learning_rate": 2.7517366407735093e-05, "loss": 0.1428, "step": 15115 }, { "epoch": 0.35, "learning_rate": 2.7516486353289055e-05, "loss": 0.2802, "step": 15120 }, { "epoch": 0.35, "learning_rate": 2.7515606298843023e-05, "loss": 0.1747, "step": 15125 }, { "epoch": 0.35, "learning_rate": 2.751472624439699e-05, "loss": 0.5115, "step": 15130 }, { "epoch": 0.35, "learning_rate": 2.751384618995095e-05, "loss": 0.2136, "step": 15135 }, { "epoch": 0.35, "learning_rate": 2.751296613550492e-05, "loss": 0.4034, "step": 15140 }, { "epoch": 0.35, "learning_rate": 2.7512086081058883e-05, "loss": 0.4945, "step": 15145 }, { "epoch": 0.35, "learning_rate": 2.751120602661285e-05, "loss": 0.6784, "step": 15150 }, { "epoch": 0.35, "learning_rate": 2.751032597216681e-05, "loss": 0.3251, "step": 15155 }, { "epoch": 0.35, "learning_rate": 2.750944591772078e-05, "loss": 0.2125, "step": 15160 }, { "epoch": 0.35, "learning_rate": 2.7508565863274743e-05, "loss": 0.2561, "step": 15165 }, { "epoch": 0.35, "learning_rate": 2.7507685808828705e-05, "loss": 0.2066, "step": 15170 }, { "epoch": 0.35, "learning_rate": 2.750680575438267e-05, "loss": 0.2122, "step": 15175 }, { "epoch": 0.35, "learning_rate": 2.750592569993664e-05, "loss": 0.265, "step": 15180 }, { "epoch": 0.35, "learning_rate": 2.7505045645490604e-05, "loss": 0.1608, "step": 15185 }, { "epoch": 0.35, "learning_rate": 2.7504165591044565e-05, "loss": 0.3992, "step": 15190 }, { "epoch": 0.35, "learning_rate": 2.750328553659853e-05, "loss": 0.3684, "step": 15195 }, { "epoch": 0.35, "learning_rate": 2.75024054821525e-05, "loss": 1.071, "step": 15200 }, { "epoch": 0.35, "learning_rate": 2.750152542770646e-05, "loss": 0.2917, "step": 15205 }, { "epoch": 0.35, "learning_rate": 2.7500645373260425e-05, "loss": 0.0693, "step": 15210 }, { "epoch": 0.35, "learning_rate": 2.749976531881439e-05, "loss": 0.3727, "step": 15215 }, { "epoch": 0.36, "learning_rate": 2.749888526436836e-05, "loss": 0.1545, "step": 15220 }, { "epoch": 0.36, "learning_rate": 2.749800520992232e-05, "loss": 0.1254, "step": 15225 }, { "epoch": 0.36, "learning_rate": 2.7497125155476285e-05, "loss": 0.2017, "step": 15230 }, { "epoch": 0.36, "learning_rate": 2.749624510103025e-05, "loss": 0.2704, "step": 15235 }, { "epoch": 0.36, "learning_rate": 2.749536504658422e-05, "loss": 0.3564, "step": 15240 }, { "epoch": 0.36, "learning_rate": 2.749448499213818e-05, "loss": 0.4514, "step": 15245 }, { "epoch": 0.36, "learning_rate": 2.7493604937692145e-05, "loss": 0.6186, "step": 15250 }, { "epoch": 0.36, "learning_rate": 2.749272488324611e-05, "loss": 0.2669, "step": 15255 }, { "epoch": 0.36, "learning_rate": 2.7491844828800075e-05, "loss": 0.0799, "step": 15260 }, { "epoch": 0.36, "learning_rate": 2.749096477435404e-05, "loss": 0.2876, "step": 15265 }, { "epoch": 0.36, "learning_rate": 2.7490084719908006e-05, "loss": 0.1625, "step": 15270 }, { "epoch": 0.36, "learning_rate": 2.748920466546197e-05, "loss": 0.1769, "step": 15275 }, { "epoch": 0.36, "learning_rate": 2.7488324611015936e-05, "loss": 0.2787, "step": 15280 }, { "epoch": 0.36, "learning_rate": 2.74874445565699e-05, "loss": 0.2327, "step": 15285 }, { "epoch": 0.36, "learning_rate": 2.7486564502123866e-05, "loss": 0.2994, "step": 15290 }, { "epoch": 0.36, "learning_rate": 2.748568444767783e-05, "loss": 0.2452, "step": 15295 }, { "epoch": 0.36, "learning_rate": 2.7484804393231796e-05, "loss": 0.4371, "step": 15300 }, { "epoch": 0.36, "learning_rate": 2.748392433878576e-05, "loss": 0.3112, "step": 15305 }, { "epoch": 0.36, "learning_rate": 2.7483044284339726e-05, "loss": 0.0889, "step": 15310 }, { "epoch": 0.36, "learning_rate": 2.748216422989369e-05, "loss": 0.154, "step": 15315 }, { "epoch": 0.36, "learning_rate": 2.7481284175447656e-05, "loss": 0.2536, "step": 15320 }, { "epoch": 0.36, "learning_rate": 2.748040412100162e-05, "loss": 0.141, "step": 15325 }, { "epoch": 0.36, "learning_rate": 2.7479524066555582e-05, "loss": 0.2325, "step": 15330 }, { "epoch": 0.36, "learning_rate": 2.747864401210955e-05, "loss": 0.3547, "step": 15335 }, { "epoch": 0.36, "learning_rate": 2.7477763957663516e-05, "loss": 0.261, "step": 15340 }, { "epoch": 0.36, "learning_rate": 2.747688390321748e-05, "loss": 0.6055, "step": 15345 }, { "epoch": 0.36, "learning_rate": 2.7476003848771442e-05, "loss": 0.4976, "step": 15350 }, { "epoch": 0.36, "learning_rate": 2.747512379432541e-05, "loss": 0.3444, "step": 15355 }, { "epoch": 0.36, "learning_rate": 2.7474243739879376e-05, "loss": 0.1724, "step": 15360 }, { "epoch": 0.36, "learning_rate": 2.7473363685433338e-05, "loss": 0.1671, "step": 15365 }, { "epoch": 0.36, "learning_rate": 2.7472483630987303e-05, "loss": 0.2214, "step": 15370 }, { "epoch": 0.36, "learning_rate": 2.747160357654127e-05, "loss": 0.2128, "step": 15375 }, { "epoch": 0.36, "learning_rate": 2.7470723522095236e-05, "loss": 0.2269, "step": 15380 }, { "epoch": 0.36, "learning_rate": 2.7469843467649198e-05, "loss": 0.403, "step": 15385 }, { "epoch": 0.36, "learning_rate": 2.7468963413203163e-05, "loss": 0.4273, "step": 15390 }, { "epoch": 0.36, "learning_rate": 2.746808335875713e-05, "loss": 0.5754, "step": 15395 }, { "epoch": 0.36, "learning_rate": 2.7467203304311093e-05, "loss": 0.6293, "step": 15400 }, { "epoch": 0.36, "learning_rate": 2.7466323249865058e-05, "loss": 0.3693, "step": 15405 }, { "epoch": 0.36, "learning_rate": 2.7465443195419023e-05, "loss": 0.1409, "step": 15410 }, { "epoch": 0.36, "learning_rate": 2.746456314097299e-05, "loss": 0.1659, "step": 15415 }, { "epoch": 0.36, "learning_rate": 2.7463683086526953e-05, "loss": 0.1612, "step": 15420 }, { "epoch": 0.36, "learning_rate": 2.7462803032080918e-05, "loss": 0.2841, "step": 15425 }, { "epoch": 0.36, "learning_rate": 2.7461922977634886e-05, "loss": 0.3868, "step": 15430 }, { "epoch": 0.36, "learning_rate": 2.7461042923188848e-05, "loss": 0.2908, "step": 15435 }, { "epoch": 0.36, "learning_rate": 2.7460162868742813e-05, "loss": 0.3253, "step": 15440 }, { "epoch": 0.36, "learning_rate": 2.7459282814296778e-05, "loss": 0.5168, "step": 15445 }, { "epoch": 0.36, "learning_rate": 2.7458402759850746e-05, "loss": 0.6879, "step": 15450 }, { "epoch": 0.36, "learning_rate": 2.7457522705404708e-05, "loss": 0.2657, "step": 15455 }, { "epoch": 0.36, "learning_rate": 2.7456642650958673e-05, "loss": 0.1106, "step": 15460 }, { "epoch": 0.36, "learning_rate": 2.7455762596512638e-05, "loss": 0.0911, "step": 15465 }, { "epoch": 0.36, "learning_rate": 2.7454882542066603e-05, "loss": 0.2312, "step": 15470 }, { "epoch": 0.36, "learning_rate": 2.7454002487620568e-05, "loss": 0.2495, "step": 15475 }, { "epoch": 0.36, "learning_rate": 2.7453122433174533e-05, "loss": 0.2058, "step": 15480 }, { "epoch": 0.36, "learning_rate": 2.7452242378728498e-05, "loss": 0.3089, "step": 15485 }, { "epoch": 0.36, "learning_rate": 2.7451362324282463e-05, "loss": 0.2495, "step": 15490 }, { "epoch": 0.36, "learning_rate": 2.7450482269836428e-05, "loss": 0.5871, "step": 15495 }, { "epoch": 0.36, "learning_rate": 2.7449602215390393e-05, "loss": 0.828, "step": 15500 }, { "epoch": 0.36, "learning_rate": 2.7448722160944355e-05, "loss": 0.4423, "step": 15505 }, { "epoch": 0.36, "learning_rate": 2.7447842106498323e-05, "loss": 0.1014, "step": 15510 }, { "epoch": 0.36, "learning_rate": 2.7446962052052288e-05, "loss": 0.0815, "step": 15515 }, { "epoch": 0.36, "learning_rate": 2.7446081997606253e-05, "loss": 0.3095, "step": 15520 }, { "epoch": 0.36, "learning_rate": 2.7445201943160215e-05, "loss": 0.1698, "step": 15525 }, { "epoch": 0.36, "learning_rate": 2.7444321888714183e-05, "loss": 0.3004, "step": 15530 }, { "epoch": 0.36, "learning_rate": 2.7443441834268148e-05, "loss": 0.2806, "step": 15535 }, { "epoch": 0.36, "learning_rate": 2.7442561779822113e-05, "loss": 0.5232, "step": 15540 }, { "epoch": 0.36, "learning_rate": 2.7441681725376075e-05, "loss": 0.4832, "step": 15545 }, { "epoch": 0.36, "learning_rate": 2.7440801670930043e-05, "loss": 0.575, "step": 15550 }, { "epoch": 0.36, "learning_rate": 2.7439921616484008e-05, "loss": 0.3386, "step": 15555 }, { "epoch": 0.36, "learning_rate": 2.743904156203797e-05, "loss": 0.0816, "step": 15560 }, { "epoch": 0.36, "learning_rate": 2.743816150759194e-05, "loss": 0.2477, "step": 15565 }, { "epoch": 0.36, "learning_rate": 2.7437281453145903e-05, "loss": 0.1624, "step": 15570 }, { "epoch": 0.36, "learning_rate": 2.743640139869987e-05, "loss": 0.1884, "step": 15575 }, { "epoch": 0.36, "learning_rate": 2.743552134425383e-05, "loss": 0.3436, "step": 15580 }, { "epoch": 0.36, "learning_rate": 2.74346412898078e-05, "loss": 0.2759, "step": 15585 }, { "epoch": 0.36, "learning_rate": 2.7433761235361763e-05, "loss": 0.5323, "step": 15590 }, { "epoch": 0.36, "learning_rate": 2.7432881180915725e-05, "loss": 0.5776, "step": 15595 }, { "epoch": 0.36, "learning_rate": 2.743200112646969e-05, "loss": 0.7016, "step": 15600 }, { "epoch": 0.36, "learning_rate": 2.743112107202366e-05, "loss": 0.3121, "step": 15605 }, { "epoch": 0.36, "learning_rate": 2.7430241017577624e-05, "loss": 0.0864, "step": 15610 }, { "epoch": 0.36, "learning_rate": 2.7429360963131585e-05, "loss": 0.248, "step": 15615 }, { "epoch": 0.36, "learning_rate": 2.742848090868555e-05, "loss": 0.1444, "step": 15620 }, { "epoch": 0.36, "learning_rate": 2.742760085423952e-05, "loss": 0.289, "step": 15625 }, { "epoch": 0.36, "learning_rate": 2.742672079979348e-05, "loss": 0.272, "step": 15630 }, { "epoch": 0.36, "learning_rate": 2.7425840745347445e-05, "loss": 0.3019, "step": 15635 }, { "epoch": 0.36, "learning_rate": 2.742496069090141e-05, "loss": 0.279, "step": 15640 }, { "epoch": 0.37, "learning_rate": 2.742408063645538e-05, "loss": 0.5555, "step": 15645 }, { "epoch": 0.37, "learning_rate": 2.742320058200934e-05, "loss": 0.7704, "step": 15650 }, { "epoch": 0.37, "learning_rate": 2.7422320527563305e-05, "loss": 0.3108, "step": 15655 }, { "epoch": 0.37, "learning_rate": 2.742144047311727e-05, "loss": 0.1348, "step": 15660 }, { "epoch": 0.37, "learning_rate": 2.7420560418671235e-05, "loss": 0.1226, "step": 15665 }, { "epoch": 0.37, "learning_rate": 2.74196803642252e-05, "loss": 0.3231, "step": 15670 }, { "epoch": 0.37, "learning_rate": 2.7418800309779165e-05, "loss": 0.2446, "step": 15675 }, { "epoch": 0.37, "learning_rate": 2.741792025533313e-05, "loss": 0.2562, "step": 15680 }, { "epoch": 0.37, "learning_rate": 2.7417040200887095e-05, "loss": 0.2649, "step": 15685 }, { "epoch": 0.37, "learning_rate": 2.741616014644106e-05, "loss": 0.2129, "step": 15690 }, { "epoch": 0.37, "learning_rate": 2.7415280091995025e-05, "loss": 0.4033, "step": 15695 }, { "epoch": 0.37, "learning_rate": 2.741440003754899e-05, "loss": 0.5157, "step": 15700 }, { "epoch": 0.37, "learning_rate": 2.7413519983102956e-05, "loss": 0.123, "step": 15705 }, { "epoch": 0.37, "learning_rate": 2.741263992865692e-05, "loss": 0.0343, "step": 15710 }, { "epoch": 0.37, "learning_rate": 2.7411759874210886e-05, "loss": 0.1335, "step": 15715 }, { "epoch": 0.37, "learning_rate": 2.741087981976485e-05, "loss": 0.2342, "step": 15720 }, { "epoch": 0.37, "learning_rate": 2.7409999765318816e-05, "loss": 0.1467, "step": 15725 }, { "epoch": 0.37, "learning_rate": 2.740911971087278e-05, "loss": 0.2985, "step": 15730 }, { "epoch": 0.37, "learning_rate": 2.7408239656426742e-05, "loss": 0.2493, "step": 15735 }, { "epoch": 0.37, "learning_rate": 2.740735960198071e-05, "loss": 0.3198, "step": 15740 }, { "epoch": 0.37, "learning_rate": 2.7406479547534676e-05, "loss": 0.4605, "step": 15745 }, { "epoch": 0.37, "learning_rate": 2.740559949308864e-05, "loss": 0.7908, "step": 15750 }, { "epoch": 0.37, "learning_rate": 2.7404719438642602e-05, "loss": 0.2643, "step": 15755 }, { "epoch": 0.37, "learning_rate": 2.740383938419657e-05, "loss": 0.148, "step": 15760 }, { "epoch": 0.37, "learning_rate": 2.7402959329750536e-05, "loss": 0.1629, "step": 15765 }, { "epoch": 0.37, "learning_rate": 2.7402079275304497e-05, "loss": 0.2097, "step": 15770 }, { "epoch": 0.37, "learning_rate": 2.7401199220858462e-05, "loss": 0.269, "step": 15775 }, { "epoch": 0.37, "learning_rate": 2.740031916641243e-05, "loss": 0.2766, "step": 15780 }, { "epoch": 0.37, "learning_rate": 2.7399439111966396e-05, "loss": 0.3365, "step": 15785 }, { "epoch": 0.37, "learning_rate": 2.7398559057520357e-05, "loss": 0.1766, "step": 15790 }, { "epoch": 0.37, "learning_rate": 2.7397679003074323e-05, "loss": 0.5255, "step": 15795 }, { "epoch": 0.37, "learning_rate": 2.739679894862829e-05, "loss": 0.5616, "step": 15800 }, { "epoch": 0.37, "learning_rate": 2.7395918894182253e-05, "loss": 0.3481, "step": 15805 }, { "epoch": 0.37, "learning_rate": 2.7395038839736218e-05, "loss": 0.1287, "step": 15810 }, { "epoch": 0.37, "learning_rate": 2.7394158785290183e-05, "loss": 0.1352, "step": 15815 }, { "epoch": 0.37, "learning_rate": 2.739327873084415e-05, "loss": 0.2455, "step": 15820 }, { "epoch": 0.37, "learning_rate": 2.7392398676398113e-05, "loss": 0.2576, "step": 15825 }, { "epoch": 0.37, "learning_rate": 2.7391518621952078e-05, "loss": 0.3256, "step": 15830 }, { "epoch": 0.37, "learning_rate": 2.7390638567506046e-05, "loss": 0.3395, "step": 15835 }, { "epoch": 0.37, "learning_rate": 2.738975851306001e-05, "loss": 0.4491, "step": 15840 }, { "epoch": 0.37, "learning_rate": 2.7388878458613973e-05, "loss": 0.1987, "step": 15845 }, { "epoch": 0.37, "learning_rate": 2.7387998404167938e-05, "loss": 0.421, "step": 15850 }, { "epoch": 0.37, "learning_rate": 2.7387118349721906e-05, "loss": 0.2413, "step": 15855 }, { "epoch": 0.37, "learning_rate": 2.7386238295275868e-05, "loss": 0.1984, "step": 15860 }, { "epoch": 0.37, "learning_rate": 2.7385358240829833e-05, "loss": 0.2117, "step": 15865 }, { "epoch": 0.37, "learning_rate": 2.7384478186383798e-05, "loss": 0.0896, "step": 15870 }, { "epoch": 0.37, "learning_rate": 2.7383598131937766e-05, "loss": 0.1353, "step": 15875 }, { "epoch": 0.37, "learning_rate": 2.7382718077491728e-05, "loss": 0.2988, "step": 15880 }, { "epoch": 0.37, "learning_rate": 2.7381838023045693e-05, "loss": 0.257, "step": 15885 }, { "epoch": 0.37, "learning_rate": 2.7380957968599658e-05, "loss": 0.2644, "step": 15890 }, { "epoch": 0.37, "learning_rate": 2.7380077914153623e-05, "loss": 0.1796, "step": 15895 }, { "epoch": 0.37, "learning_rate": 2.7379197859707588e-05, "loss": 0.6983, "step": 15900 }, { "epoch": 0.37, "learning_rate": 2.7378317805261553e-05, "loss": 0.3278, "step": 15905 }, { "epoch": 0.37, "learning_rate": 2.7377437750815518e-05, "loss": 0.1259, "step": 15910 }, { "epoch": 0.37, "learning_rate": 2.7376557696369483e-05, "loss": 0.1201, "step": 15915 }, { "epoch": 0.37, "learning_rate": 2.7375677641923448e-05, "loss": 0.2573, "step": 15920 }, { "epoch": 0.37, "learning_rate": 2.7374797587477413e-05, "loss": 0.1622, "step": 15925 }, { "epoch": 0.37, "learning_rate": 2.7373917533031375e-05, "loss": 0.2813, "step": 15930 }, { "epoch": 0.37, "learning_rate": 2.7373037478585343e-05, "loss": 0.2571, "step": 15935 }, { "epoch": 0.37, "learning_rate": 2.7372157424139308e-05, "loss": 0.4536, "step": 15940 }, { "epoch": 0.37, "learning_rate": 2.7371277369693273e-05, "loss": 0.4058, "step": 15945 }, { "epoch": 0.37, "learning_rate": 2.7370397315247235e-05, "loss": 0.5341, "step": 15950 }, { "epoch": 0.37, "learning_rate": 2.7369517260801203e-05, "loss": 0.3078, "step": 15955 }, { "epoch": 0.37, "learning_rate": 2.7368637206355168e-05, "loss": 0.181, "step": 15960 }, { "epoch": 0.37, "learning_rate": 2.736775715190913e-05, "loss": 0.1398, "step": 15965 }, { "epoch": 0.37, "learning_rate": 2.7366877097463098e-05, "loss": 0.2568, "step": 15970 }, { "epoch": 0.37, "learning_rate": 2.7365997043017063e-05, "loss": 0.2987, "step": 15975 }, { "epoch": 0.37, "learning_rate": 2.7365116988571028e-05, "loss": 0.0922, "step": 15980 }, { "epoch": 0.37, "learning_rate": 2.736423693412499e-05, "loss": 0.2342, "step": 15985 }, { "epoch": 0.37, "learning_rate": 2.736335687967896e-05, "loss": 0.373, "step": 15990 }, { "epoch": 0.37, "learning_rate": 2.7362476825232923e-05, "loss": 0.6878, "step": 15995 }, { "epoch": 0.37, "learning_rate": 2.7361596770786885e-05, "loss": 0.4746, "step": 16000 }, { "epoch": 0.37, "learning_rate": 2.736071671634085e-05, "loss": 0.3031, "step": 16005 }, { "epoch": 0.37, "learning_rate": 2.735983666189482e-05, "loss": 0.09, "step": 16010 }, { "epoch": 0.37, "learning_rate": 2.7358956607448783e-05, "loss": 0.1269, "step": 16015 }, { "epoch": 0.37, "learning_rate": 2.7358076553002745e-05, "loss": 0.1254, "step": 16020 }, { "epoch": 0.37, "learning_rate": 2.735719649855671e-05, "loss": 0.2367, "step": 16025 }, { "epoch": 0.37, "learning_rate": 2.735631644411068e-05, "loss": 0.2942, "step": 16030 }, { "epoch": 0.37, "learning_rate": 2.735543638966464e-05, "loss": 0.3407, "step": 16035 }, { "epoch": 0.37, "learning_rate": 2.7354556335218605e-05, "loss": 0.3716, "step": 16040 }, { "epoch": 0.37, "learning_rate": 2.735367628077257e-05, "loss": 0.2904, "step": 16045 }, { "epoch": 0.37, "learning_rate": 2.735279622632654e-05, "loss": 0.5498, "step": 16050 }, { "epoch": 0.37, "learning_rate": 2.73519161718805e-05, "loss": 0.1914, "step": 16055 }, { "epoch": 0.37, "learning_rate": 2.7351036117434465e-05, "loss": 0.082, "step": 16060 }, { "epoch": 0.37, "learning_rate": 2.735015606298843e-05, "loss": 0.1697, "step": 16065 }, { "epoch": 0.37, "learning_rate": 2.7349276008542395e-05, "loss": 0.18, "step": 16070 }, { "epoch": 0.38, "learning_rate": 2.734839595409636e-05, "loss": 0.1991, "step": 16075 }, { "epoch": 0.38, "learning_rate": 2.7347515899650325e-05, "loss": 0.1825, "step": 16080 }, { "epoch": 0.38, "learning_rate": 2.734663584520429e-05, "loss": 0.2626, "step": 16085 }, { "epoch": 0.38, "learning_rate": 2.7345755790758255e-05, "loss": 0.3279, "step": 16090 }, { "epoch": 0.38, "learning_rate": 2.734487573631222e-05, "loss": 0.297, "step": 16095 }, { "epoch": 0.38, "learning_rate": 2.7343995681866185e-05, "loss": 0.6024, "step": 16100 }, { "epoch": 0.38, "learning_rate": 2.734311562742015e-05, "loss": 0.4147, "step": 16105 }, { "epoch": 0.38, "learning_rate": 2.7342235572974115e-05, "loss": 0.1083, "step": 16110 }, { "epoch": 0.38, "learning_rate": 2.734135551852808e-05, "loss": 0.0873, "step": 16115 }, { "epoch": 0.38, "learning_rate": 2.7340475464082045e-05, "loss": 0.1437, "step": 16120 }, { "epoch": 0.38, "learning_rate": 2.733959540963601e-05, "loss": 0.1494, "step": 16125 }, { "epoch": 0.38, "learning_rate": 2.7338715355189975e-05, "loss": 0.2021, "step": 16130 }, { "epoch": 0.38, "learning_rate": 2.733783530074394e-05, "loss": 0.2961, "step": 16135 }, { "epoch": 0.38, "learning_rate": 2.7336955246297902e-05, "loss": 0.2635, "step": 16140 }, { "epoch": 0.38, "learning_rate": 2.733607519185187e-05, "loss": 0.2952, "step": 16145 }, { "epoch": 0.38, "learning_rate": 2.7335195137405836e-05, "loss": 0.8806, "step": 16150 }, { "epoch": 0.38, "learning_rate": 2.73343150829598e-05, "loss": 0.2492, "step": 16155 }, { "epoch": 0.38, "learning_rate": 2.7333435028513762e-05, "loss": 0.2082, "step": 16160 }, { "epoch": 0.38, "learning_rate": 2.733255497406773e-05, "loss": 0.2683, "step": 16165 }, { "epoch": 0.38, "learning_rate": 2.7331674919621696e-05, "loss": 0.1451, "step": 16170 }, { "epoch": 0.38, "learning_rate": 2.733079486517566e-05, "loss": 0.1731, "step": 16175 }, { "epoch": 0.38, "learning_rate": 2.7329914810729622e-05, "loss": 0.1653, "step": 16180 }, { "epoch": 0.38, "learning_rate": 2.732903475628359e-05, "loss": 0.2838, "step": 16185 }, { "epoch": 0.38, "learning_rate": 2.7328154701837556e-05, "loss": 0.2966, "step": 16190 }, { "epoch": 0.38, "learning_rate": 2.7327274647391517e-05, "loss": 0.3066, "step": 16195 }, { "epoch": 0.38, "learning_rate": 2.7326394592945482e-05, "loss": 0.5428, "step": 16200 }, { "epoch": 0.38, "learning_rate": 2.732551453849945e-05, "loss": 0.2181, "step": 16205 }, { "epoch": 0.38, "learning_rate": 2.7324634484053416e-05, "loss": 0.1155, "step": 16210 }, { "epoch": 0.38, "learning_rate": 2.7323754429607377e-05, "loss": 0.1207, "step": 16215 }, { "epoch": 0.38, "learning_rate": 2.7322874375161342e-05, "loss": 0.2237, "step": 16220 }, { "epoch": 0.38, "learning_rate": 2.732199432071531e-05, "loss": 0.3095, "step": 16225 }, { "epoch": 0.38, "learning_rate": 2.7321114266269273e-05, "loss": 0.2639, "step": 16230 }, { "epoch": 0.38, "learning_rate": 2.7320234211823238e-05, "loss": 0.2832, "step": 16235 }, { "epoch": 0.38, "learning_rate": 2.7319354157377206e-05, "loss": 0.3657, "step": 16240 }, { "epoch": 0.38, "learning_rate": 2.731847410293117e-05, "loss": 0.3967, "step": 16245 }, { "epoch": 0.38, "learning_rate": 2.7317594048485133e-05, "loss": 0.5485, "step": 16250 }, { "epoch": 0.38, "learning_rate": 2.7316713994039098e-05, "loss": 0.2415, "step": 16255 }, { "epoch": 0.38, "learning_rate": 2.7315833939593066e-05, "loss": 0.1365, "step": 16260 }, { "epoch": 0.38, "learning_rate": 2.7314953885147028e-05, "loss": 0.1676, "step": 16265 }, { "epoch": 0.38, "learning_rate": 2.7314073830700993e-05, "loss": 0.304, "step": 16270 }, { "epoch": 0.38, "learning_rate": 2.7313193776254958e-05, "loss": 0.2193, "step": 16275 }, { "epoch": 0.38, "learning_rate": 2.7312313721808926e-05, "loss": 0.2467, "step": 16280 }, { "epoch": 0.38, "learning_rate": 2.7311433667362888e-05, "loss": 0.312, "step": 16285 }, { "epoch": 0.38, "learning_rate": 2.7310553612916853e-05, "loss": 0.2165, "step": 16290 }, { "epoch": 0.38, "learning_rate": 2.7309673558470818e-05, "loss": 0.3248, "step": 16295 }, { "epoch": 0.38, "learning_rate": 2.7308793504024783e-05, "loss": 0.3881, "step": 16300 }, { "epoch": 0.38, "learning_rate": 2.7307913449578748e-05, "loss": 0.2829, "step": 16305 }, { "epoch": 0.38, "learning_rate": 2.7307033395132713e-05, "loss": 0.1581, "step": 16310 }, { "epoch": 0.38, "learning_rate": 2.7306153340686678e-05, "loss": 0.2236, "step": 16315 }, { "epoch": 0.38, "learning_rate": 2.7305273286240643e-05, "loss": 0.2549, "step": 16320 }, { "epoch": 0.38, "learning_rate": 2.7304393231794608e-05, "loss": 0.1952, "step": 16325 }, { "epoch": 0.38, "learning_rate": 2.7303513177348573e-05, "loss": 0.2359, "step": 16330 }, { "epoch": 0.38, "learning_rate": 2.7302633122902535e-05, "loss": 0.3733, "step": 16335 }, { "epoch": 0.38, "learning_rate": 2.7301753068456503e-05, "loss": 0.3444, "step": 16340 }, { "epoch": 0.38, "learning_rate": 2.7300873014010468e-05, "loss": 0.4, "step": 16345 }, { "epoch": 0.38, "learning_rate": 2.7299992959564433e-05, "loss": 1.1132, "step": 16350 }, { "epoch": 0.38, "learning_rate": 2.7299112905118395e-05, "loss": 0.242, "step": 16355 }, { "epoch": 0.38, "learning_rate": 2.7298232850672363e-05, "loss": 0.0917, "step": 16360 }, { "epoch": 0.38, "learning_rate": 2.7297352796226328e-05, "loss": 0.1526, "step": 16365 }, { "epoch": 0.38, "learning_rate": 2.729647274178029e-05, "loss": 0.1037, "step": 16370 }, { "epoch": 0.38, "learning_rate": 2.7295592687334258e-05, "loss": 0.3039, "step": 16375 }, { "epoch": 0.38, "learning_rate": 2.729488864377743e-05, "loss": 0.1845, "step": 16380 }, { "epoch": 0.38, "learning_rate": 2.7294008589331394e-05, "loss": 0.3211, "step": 16385 }, { "epoch": 0.38, "learning_rate": 2.7293128534885356e-05, "loss": 0.3263, "step": 16390 }, { "epoch": 0.38, "learning_rate": 2.7292248480439325e-05, "loss": 0.4431, "step": 16395 }, { "epoch": 0.38, "learning_rate": 2.729136842599329e-05, "loss": 0.5681, "step": 16400 }, { "epoch": 0.38, "learning_rate": 2.7290488371547255e-05, "loss": 0.2846, "step": 16405 }, { "epoch": 0.38, "learning_rate": 2.7289608317101216e-05, "loss": 0.1768, "step": 16410 }, { "epoch": 0.38, "learning_rate": 2.7288728262655185e-05, "loss": 0.217, "step": 16415 }, { "epoch": 0.38, "learning_rate": 2.728784820820915e-05, "loss": 0.12, "step": 16420 }, { "epoch": 0.38, "learning_rate": 2.728696815376311e-05, "loss": 0.0907, "step": 16425 }, { "epoch": 0.38, "learning_rate": 2.728608809931708e-05, "loss": 0.3301, "step": 16430 }, { "epoch": 0.38, "learning_rate": 2.7285208044871045e-05, "loss": 0.2706, "step": 16435 }, { "epoch": 0.38, "learning_rate": 2.728432799042501e-05, "loss": 0.4011, "step": 16440 }, { "epoch": 0.38, "learning_rate": 2.728344793597897e-05, "loss": 0.5244, "step": 16445 }, { "epoch": 0.38, "learning_rate": 2.728256788153294e-05, "loss": 0.6937, "step": 16450 }, { "epoch": 0.38, "learning_rate": 2.7281687827086905e-05, "loss": 0.334, "step": 16455 }, { "epoch": 0.38, "learning_rate": 2.7280807772640866e-05, "loss": 0.0781, "step": 16460 }, { "epoch": 0.38, "learning_rate": 2.727992771819483e-05, "loss": 0.2075, "step": 16465 }, { "epoch": 0.38, "learning_rate": 2.72790476637488e-05, "loss": 0.2012, "step": 16470 }, { "epoch": 0.38, "learning_rate": 2.7278167609302765e-05, "loss": 0.2206, "step": 16475 }, { "epoch": 0.38, "learning_rate": 2.7277287554856726e-05, "loss": 0.255, "step": 16480 }, { "epoch": 0.38, "learning_rate": 2.727640750041069e-05, "loss": 0.3665, "step": 16485 }, { "epoch": 0.38, "learning_rate": 2.727552744596466e-05, "loss": 0.2974, "step": 16490 }, { "epoch": 0.38, "learning_rate": 2.727464739151862e-05, "loss": 0.3405, "step": 16495 }, { "epoch": 0.38, "learning_rate": 2.7273767337072587e-05, "loss": 0.6269, "step": 16500 }, { "epoch": 0.39, "learning_rate": 2.727288728262655e-05, "loss": 0.3266, "step": 16505 }, { "epoch": 0.39, "learning_rate": 2.727200722818052e-05, "loss": 0.2333, "step": 16510 }, { "epoch": 0.39, "learning_rate": 2.727112717373448e-05, "loss": 0.2267, "step": 16515 }, { "epoch": 0.39, "learning_rate": 2.7270247119288447e-05, "loss": 0.1713, "step": 16520 }, { "epoch": 0.39, "learning_rate": 2.726936706484241e-05, "loss": 0.2602, "step": 16525 }, { "epoch": 0.39, "learning_rate": 2.7268487010396377e-05, "loss": 0.2218, "step": 16530 }, { "epoch": 0.39, "learning_rate": 2.7267606955950342e-05, "loss": 0.36, "step": 16535 }, { "epoch": 0.39, "learning_rate": 2.7266726901504307e-05, "loss": 0.4008, "step": 16540 }, { "epoch": 0.39, "learning_rate": 2.7265846847058272e-05, "loss": 0.3662, "step": 16545 }, { "epoch": 0.39, "learning_rate": 2.7264966792612237e-05, "loss": 0.6261, "step": 16550 }, { "epoch": 0.39, "learning_rate": 2.7264086738166202e-05, "loss": 0.2626, "step": 16555 }, { "epoch": 0.39, "learning_rate": 2.7263206683720167e-05, "loss": 0.082, "step": 16560 }, { "epoch": 0.39, "learning_rate": 2.7262326629274132e-05, "loss": 0.1262, "step": 16565 }, { "epoch": 0.39, "learning_rate": 2.7261446574828097e-05, "loss": 0.1592, "step": 16570 }, { "epoch": 0.39, "learning_rate": 2.7260566520382062e-05, "loss": 0.1629, "step": 16575 }, { "epoch": 0.39, "learning_rate": 2.7259686465936027e-05, "loss": 0.3094, "step": 16580 }, { "epoch": 0.39, "learning_rate": 2.7258806411489992e-05, "loss": 0.3896, "step": 16585 }, { "epoch": 0.39, "learning_rate": 2.7257926357043957e-05, "loss": 0.2518, "step": 16590 }, { "epoch": 0.39, "learning_rate": 2.7257046302597922e-05, "loss": 0.3059, "step": 16595 }, { "epoch": 0.39, "learning_rate": 2.7256166248151884e-05, "loss": 0.6235, "step": 16600 }, { "epoch": 0.39, "learning_rate": 2.7255286193705852e-05, "loss": 0.1581, "step": 16605 }, { "epoch": 0.39, "learning_rate": 2.7254406139259817e-05, "loss": 0.1048, "step": 16610 }, { "epoch": 0.39, "learning_rate": 2.7253526084813782e-05, "loss": 0.131, "step": 16615 }, { "epoch": 0.39, "learning_rate": 2.7252646030367744e-05, "loss": 0.2126, "step": 16620 }, { "epoch": 0.39, "learning_rate": 2.7251765975921712e-05, "loss": 0.3245, "step": 16625 }, { "epoch": 0.39, "learning_rate": 2.7250885921475677e-05, "loss": 0.2089, "step": 16630 }, { "epoch": 0.39, "learning_rate": 2.7250005867029642e-05, "loss": 0.2456, "step": 16635 }, { "epoch": 0.39, "learning_rate": 2.7249125812583604e-05, "loss": 0.2128, "step": 16640 }, { "epoch": 0.39, "learning_rate": 2.7248245758137572e-05, "loss": 0.2881, "step": 16645 }, { "epoch": 0.39, "learning_rate": 2.7247365703691537e-05, "loss": 0.846, "step": 16650 }, { "epoch": 0.39, "learning_rate": 2.72464856492455e-05, "loss": 0.1823, "step": 16655 }, { "epoch": 0.39, "learning_rate": 2.7245605594799464e-05, "loss": 0.132, "step": 16660 }, { "epoch": 0.39, "learning_rate": 2.7244725540353432e-05, "loss": 0.1125, "step": 16665 }, { "epoch": 0.39, "learning_rate": 2.7243845485907397e-05, "loss": 0.1984, "step": 16670 }, { "epoch": 0.39, "learning_rate": 2.724296543146136e-05, "loss": 0.2653, "step": 16675 }, { "epoch": 0.39, "learning_rate": 2.7242085377015324e-05, "loss": 0.3283, "step": 16680 }, { "epoch": 0.39, "learning_rate": 2.7241205322569292e-05, "loss": 0.2002, "step": 16685 }, { "epoch": 0.39, "learning_rate": 2.7240325268123254e-05, "loss": 0.4155, "step": 16690 }, { "epoch": 0.39, "learning_rate": 2.723944521367722e-05, "loss": 0.661, "step": 16695 }, { "epoch": 0.39, "learning_rate": 2.7238565159231187e-05, "loss": 0.531, "step": 16700 }, { "epoch": 0.39, "learning_rate": 2.7237685104785152e-05, "loss": 0.3908, "step": 16705 }, { "epoch": 0.39, "learning_rate": 2.7236805050339114e-05, "loss": 0.122, "step": 16710 }, { "epoch": 0.39, "learning_rate": 2.723592499589308e-05, "loss": 0.1944, "step": 16715 }, { "epoch": 0.39, "learning_rate": 2.7235044941447047e-05, "loss": 0.2125, "step": 16720 }, { "epoch": 0.39, "learning_rate": 2.723416488700101e-05, "loss": 0.1669, "step": 16725 }, { "epoch": 0.39, "learning_rate": 2.7233284832554974e-05, "loss": 0.1772, "step": 16730 }, { "epoch": 0.39, "learning_rate": 2.723240477810894e-05, "loss": 0.1305, "step": 16735 }, { "epoch": 0.39, "learning_rate": 2.7231524723662908e-05, "loss": 0.3229, "step": 16740 }, { "epoch": 0.39, "learning_rate": 2.723064466921687e-05, "loss": 0.376, "step": 16745 }, { "epoch": 0.39, "learning_rate": 2.7229764614770834e-05, "loss": 0.8267, "step": 16750 }, { "epoch": 0.39, "learning_rate": 2.72288845603248e-05, "loss": 0.3385, "step": 16755 }, { "epoch": 0.39, "learning_rate": 2.7228004505878764e-05, "loss": 0.1282, "step": 16760 }, { "epoch": 0.39, "learning_rate": 2.722712445143273e-05, "loss": 0.1166, "step": 16765 }, { "epoch": 0.39, "learning_rate": 2.7226244396986694e-05, "loss": 0.2292, "step": 16770 }, { "epoch": 0.39, "learning_rate": 2.722536434254066e-05, "loss": 0.3773, "step": 16775 }, { "epoch": 0.39, "learning_rate": 2.7224484288094624e-05, "loss": 0.1937, "step": 16780 }, { "epoch": 0.39, "learning_rate": 2.722360423364859e-05, "loss": 0.3417, "step": 16785 }, { "epoch": 0.39, "learning_rate": 2.7222724179202554e-05, "loss": 0.3546, "step": 16790 }, { "epoch": 0.39, "learning_rate": 2.7221844124756516e-05, "loss": 0.3862, "step": 16795 }, { "epoch": 0.39, "learning_rate": 2.7220964070310484e-05, "loss": 0.7833, "step": 16800 }, { "epoch": 0.39, "learning_rate": 2.722008401586445e-05, "loss": 0.2569, "step": 16805 }, { "epoch": 0.39, "learning_rate": 2.7219203961418414e-05, "loss": 0.11, "step": 16810 }, { "epoch": 0.39, "learning_rate": 2.7218323906972376e-05, "loss": 0.1006, "step": 16815 }, { "epoch": 0.39, "learning_rate": 2.7217443852526344e-05, "loss": 0.2961, "step": 16820 }, { "epoch": 0.39, "learning_rate": 2.721656379808031e-05, "loss": 0.2501, "step": 16825 }, { "epoch": 0.39, "learning_rate": 2.721568374363427e-05, "loss": 0.2257, "step": 16830 }, { "epoch": 0.39, "learning_rate": 2.721480368918824e-05, "loss": 0.2274, "step": 16835 }, { "epoch": 0.39, "learning_rate": 2.7213923634742205e-05, "loss": 0.3328, "step": 16840 }, { "epoch": 0.39, "learning_rate": 2.721304358029617e-05, "loss": 0.3377, "step": 16845 }, { "epoch": 0.39, "learning_rate": 2.721216352585013e-05, "loss": 0.9903, "step": 16850 }, { "epoch": 0.39, "learning_rate": 2.72112834714041e-05, "loss": 0.2831, "step": 16855 }, { "epoch": 0.39, "learning_rate": 2.7210403416958065e-05, "loss": 0.2005, "step": 16860 }, { "epoch": 0.39, "learning_rate": 2.7209523362512026e-05, "loss": 0.0958, "step": 16865 }, { "epoch": 0.39, "learning_rate": 2.720864330806599e-05, "loss": 0.2129, "step": 16870 }, { "epoch": 0.39, "learning_rate": 2.720776325361996e-05, "loss": 0.1539, "step": 16875 }, { "epoch": 0.39, "learning_rate": 2.7206883199173925e-05, "loss": 0.263, "step": 16880 }, { "epoch": 0.39, "learning_rate": 2.7206003144727886e-05, "loss": 0.3781, "step": 16885 }, { "epoch": 0.39, "learning_rate": 2.720512309028185e-05, "loss": 0.1637, "step": 16890 }, { "epoch": 0.39, "learning_rate": 2.720424303583582e-05, "loss": 0.7287, "step": 16895 }, { "epoch": 0.39, "learning_rate": 2.720336298138978e-05, "loss": 0.7749, "step": 16900 }, { "epoch": 0.39, "learning_rate": 2.7202482926943746e-05, "loss": 0.2739, "step": 16905 }, { "epoch": 0.39, "learning_rate": 2.720160287249771e-05, "loss": 0.0993, "step": 16910 }, { "epoch": 0.39, "learning_rate": 2.720072281805168e-05, "loss": 0.1831, "step": 16915 }, { "epoch": 0.39, "learning_rate": 2.719984276360564e-05, "loss": 0.158, "step": 16920 }, { "epoch": 0.39, "learning_rate": 2.7198962709159607e-05, "loss": 0.1536, "step": 16925 }, { "epoch": 0.39, "learning_rate": 2.719808265471357e-05, "loss": 0.3177, "step": 16930 }, { "epoch": 0.4, "learning_rate": 2.7197202600267537e-05, "loss": 0.3187, "step": 16935 }, { "epoch": 0.4, "learning_rate": 2.71963225458215e-05, "loss": 0.261, "step": 16940 }, { "epoch": 0.4, "learning_rate": 2.7195442491375467e-05, "loss": 0.3254, "step": 16945 }, { "epoch": 0.4, "learning_rate": 2.719456243692943e-05, "loss": 0.3406, "step": 16950 }, { "epoch": 0.4, "learning_rate": 2.7193682382483397e-05, "loss": 0.3101, "step": 16955 }, { "epoch": 0.4, "learning_rate": 2.719280232803736e-05, "loss": 0.1239, "step": 16960 }, { "epoch": 0.4, "learning_rate": 2.7191922273591327e-05, "loss": 0.1255, "step": 16965 }, { "epoch": 0.4, "learning_rate": 2.7191042219145295e-05, "loss": 0.1505, "step": 16970 }, { "epoch": 0.4, "learning_rate": 2.7190162164699257e-05, "loss": 0.2297, "step": 16975 }, { "epoch": 0.4, "learning_rate": 2.7189282110253222e-05, "loss": 0.2663, "step": 16980 }, { "epoch": 0.4, "learning_rate": 2.7188402055807187e-05, "loss": 0.3483, "step": 16985 }, { "epoch": 0.4, "learning_rate": 2.7187522001361152e-05, "loss": 0.2922, "step": 16990 }, { "epoch": 0.4, "learning_rate": 2.7186641946915117e-05, "loss": 0.4816, "step": 16995 }, { "epoch": 0.4, "learning_rate": 2.7185761892469082e-05, "loss": 0.489, "step": 17000 }, { "epoch": 0.4, "learning_rate": 2.7184881838023047e-05, "loss": 0.243, "step": 17005 }, { "epoch": 0.4, "learning_rate": 2.7184001783577012e-05, "loss": 0.1226, "step": 17010 }, { "epoch": 0.4, "learning_rate": 2.7183121729130977e-05, "loss": 0.1656, "step": 17015 }, { "epoch": 0.4, "learning_rate": 2.7182241674684942e-05, "loss": 0.2575, "step": 17020 }, { "epoch": 0.4, "learning_rate": 2.7181361620238904e-05, "loss": 0.2233, "step": 17025 }, { "epoch": 0.4, "learning_rate": 2.7180481565792872e-05, "loss": 0.2737, "step": 17030 }, { "epoch": 0.4, "learning_rate": 2.7179601511346837e-05, "loss": 0.2678, "step": 17035 }, { "epoch": 0.4, "learning_rate": 2.7178721456900802e-05, "loss": 0.351, "step": 17040 }, { "epoch": 0.4, "learning_rate": 2.7177841402454764e-05, "loss": 0.3961, "step": 17045 }, { "epoch": 0.4, "learning_rate": 2.7176961348008732e-05, "loss": 0.2945, "step": 17050 }, { "epoch": 0.4, "learning_rate": 2.7176081293562697e-05, "loss": 0.3148, "step": 17055 }, { "epoch": 0.4, "learning_rate": 2.717520123911666e-05, "loss": 0.1521, "step": 17060 }, { "epoch": 0.4, "learning_rate": 2.7174321184670624e-05, "loss": 0.1478, "step": 17065 }, { "epoch": 0.4, "learning_rate": 2.7173441130224592e-05, "loss": 0.319, "step": 17070 }, { "epoch": 0.4, "learning_rate": 2.7172561075778557e-05, "loss": 0.139, "step": 17075 }, { "epoch": 0.4, "learning_rate": 2.717168102133252e-05, "loss": 0.2561, "step": 17080 }, { "epoch": 0.4, "learning_rate": 2.7170800966886484e-05, "loss": 0.2346, "step": 17085 }, { "epoch": 0.4, "learning_rate": 2.7169920912440452e-05, "loss": 0.3295, "step": 17090 }, { "epoch": 0.4, "learning_rate": 2.7169040857994414e-05, "loss": 0.3548, "step": 17095 }, { "epoch": 0.4, "learning_rate": 2.716816080354838e-05, "loss": 0.7333, "step": 17100 }, { "epoch": 0.4, "learning_rate": 2.7167280749102347e-05, "loss": 0.2238, "step": 17105 }, { "epoch": 0.4, "learning_rate": 2.7166400694656312e-05, "loss": 0.1628, "step": 17110 }, { "epoch": 0.4, "learning_rate": 2.7165520640210274e-05, "loss": 0.1346, "step": 17115 }, { "epoch": 0.4, "learning_rate": 2.716464058576424e-05, "loss": 0.2305, "step": 17120 }, { "epoch": 0.4, "learning_rate": 2.7163760531318207e-05, "loss": 0.2602, "step": 17125 }, { "epoch": 0.4, "learning_rate": 2.716288047687217e-05, "loss": 0.2282, "step": 17130 }, { "epoch": 0.4, "learning_rate": 2.7162000422426134e-05, "loss": 0.3637, "step": 17135 }, { "epoch": 0.4, "learning_rate": 2.71611203679801e-05, "loss": 0.2147, "step": 17140 }, { "epoch": 0.4, "learning_rate": 2.7160240313534067e-05, "loss": 0.228, "step": 17145 }, { "epoch": 0.4, "learning_rate": 2.715936025908803e-05, "loss": 0.7111, "step": 17150 }, { "epoch": 0.4, "learning_rate": 2.7158480204641994e-05, "loss": 0.2662, "step": 17155 }, { "epoch": 0.4, "learning_rate": 2.715760015019596e-05, "loss": 0.1208, "step": 17160 }, { "epoch": 0.4, "learning_rate": 2.7156720095749924e-05, "loss": 0.1544, "step": 17165 }, { "epoch": 0.4, "learning_rate": 2.715584004130389e-05, "loss": 0.2047, "step": 17170 }, { "epoch": 0.4, "learning_rate": 2.7154959986857854e-05, "loss": 0.093, "step": 17175 }, { "epoch": 0.4, "learning_rate": 2.715407993241182e-05, "loss": 0.3646, "step": 17180 }, { "epoch": 0.4, "learning_rate": 2.7153199877965784e-05, "loss": 0.2989, "step": 17185 }, { "epoch": 0.4, "learning_rate": 2.715231982351975e-05, "loss": 0.4107, "step": 17190 }, { "epoch": 0.4, "learning_rate": 2.7151439769073714e-05, "loss": 0.4011, "step": 17195 }, { "epoch": 0.4, "learning_rate": 2.7150559714627676e-05, "loss": 0.3621, "step": 17200 }, { "epoch": 0.4, "learning_rate": 2.7149679660181644e-05, "loss": 0.1828, "step": 17205 }, { "epoch": 0.4, "learning_rate": 2.714879960573561e-05, "loss": 0.1496, "step": 17210 }, { "epoch": 0.4, "learning_rate": 2.7147919551289574e-05, "loss": 0.1385, "step": 17215 }, { "epoch": 0.4, "learning_rate": 2.7147039496843536e-05, "loss": 0.1995, "step": 17220 }, { "epoch": 0.4, "learning_rate": 2.7146159442397504e-05, "loss": 0.1486, "step": 17225 }, { "epoch": 0.4, "learning_rate": 2.714527938795147e-05, "loss": 0.1398, "step": 17230 }, { "epoch": 0.4, "learning_rate": 2.714439933350543e-05, "loss": 0.3685, "step": 17235 }, { "epoch": 0.4, "learning_rate": 2.71435192790594e-05, "loss": 0.2906, "step": 17240 }, { "epoch": 0.4, "learning_rate": 2.7142639224613364e-05, "loss": 0.592, "step": 17245 }, { "epoch": 0.4, "learning_rate": 2.714175917016733e-05, "loss": 0.5928, "step": 17250 }, { "epoch": 0.4, "learning_rate": 2.714087911572129e-05, "loss": 0.2805, "step": 17255 }, { "epoch": 0.4, "learning_rate": 2.713999906127526e-05, "loss": 0.1741, "step": 17260 }, { "epoch": 0.4, "learning_rate": 2.7139119006829225e-05, "loss": 0.1993, "step": 17265 }, { "epoch": 0.4, "learning_rate": 2.713823895238319e-05, "loss": 0.2042, "step": 17270 }, { "epoch": 0.4, "learning_rate": 2.713735889793715e-05, "loss": 0.2994, "step": 17275 }, { "epoch": 0.4, "learning_rate": 2.713647884349112e-05, "loss": 0.2496, "step": 17280 }, { "epoch": 0.4, "learning_rate": 2.7135598789045085e-05, "loss": 0.365, "step": 17285 }, { "epoch": 0.4, "learning_rate": 2.7134718734599046e-05, "loss": 0.2283, "step": 17290 }, { "epoch": 0.4, "learning_rate": 2.713383868015301e-05, "loss": 0.2697, "step": 17295 }, { "epoch": 0.4, "learning_rate": 2.713295862570698e-05, "loss": 0.9212, "step": 17300 }, { "epoch": 0.4, "learning_rate": 2.7132078571260945e-05, "loss": 0.2572, "step": 17305 }, { "epoch": 0.4, "learning_rate": 2.7131198516814906e-05, "loss": 0.1324, "step": 17310 }, { "epoch": 0.4, "learning_rate": 2.713031846236887e-05, "loss": 0.127, "step": 17315 }, { "epoch": 0.4, "learning_rate": 2.712943840792284e-05, "loss": 0.1357, "step": 17320 }, { "epoch": 0.4, "learning_rate": 2.71285583534768e-05, "loss": 0.2383, "step": 17325 }, { "epoch": 0.4, "learning_rate": 2.7127678299030766e-05, "loss": 0.2243, "step": 17330 }, { "epoch": 0.4, "learning_rate": 2.712679824458473e-05, "loss": 0.2935, "step": 17335 }, { "epoch": 0.4, "learning_rate": 2.71259181901387e-05, "loss": 0.3298, "step": 17340 }, { "epoch": 0.4, "learning_rate": 2.712503813569266e-05, "loss": 0.3225, "step": 17345 }, { "epoch": 0.4, "learning_rate": 2.7124158081246626e-05, "loss": 0.7997, "step": 17350 }, { "epoch": 0.4, "learning_rate": 2.712327802680059e-05, "loss": 0.2446, "step": 17355 }, { "epoch": 0.41, "learning_rate": 2.7122397972354557e-05, "loss": 0.0651, "step": 17360 }, { "epoch": 0.41, "learning_rate": 2.712151791790852e-05, "loss": 0.1759, "step": 17365 }, { "epoch": 0.41, "learning_rate": 2.7120637863462487e-05, "loss": 0.1072, "step": 17370 }, { "epoch": 0.41, "learning_rate": 2.7119757809016455e-05, "loss": 0.2379, "step": 17375 }, { "epoch": 0.41, "learning_rate": 2.7118877754570417e-05, "loss": 0.3375, "step": 17380 }, { "epoch": 0.41, "learning_rate": 2.711799770012438e-05, "loss": 0.206, "step": 17385 }, { "epoch": 0.41, "learning_rate": 2.7117117645678347e-05, "loss": 0.394, "step": 17390 }, { "epoch": 0.41, "learning_rate": 2.711623759123231e-05, "loss": 0.3181, "step": 17395 }, { "epoch": 0.41, "learning_rate": 2.7115357536786277e-05, "loss": 0.3766, "step": 17400 }, { "epoch": 0.41, "learning_rate": 2.7114477482340242e-05, "loss": 0.221, "step": 17405 }, { "epoch": 0.41, "learning_rate": 2.7113597427894207e-05, "loss": 0.1662, "step": 17410 }, { "epoch": 0.41, "learning_rate": 2.7112717373448172e-05, "loss": 0.2406, "step": 17415 }, { "epoch": 0.41, "learning_rate": 2.7111837319002137e-05, "loss": 0.1494, "step": 17420 }, { "epoch": 0.41, "learning_rate": 2.7110957264556102e-05, "loss": 0.2295, "step": 17425 }, { "epoch": 0.41, "learning_rate": 2.7110077210110063e-05, "loss": 0.3281, "step": 17430 }, { "epoch": 0.41, "learning_rate": 2.7109197155664032e-05, "loss": 0.1909, "step": 17435 }, { "epoch": 0.41, "learning_rate": 2.7108317101217997e-05, "loss": 0.1734, "step": 17440 }, { "epoch": 0.41, "learning_rate": 2.7107437046771962e-05, "loss": 0.3234, "step": 17445 }, { "epoch": 0.41, "learning_rate": 2.7106556992325924e-05, "loss": 0.6082, "step": 17450 }, { "epoch": 0.41, "learning_rate": 2.7105676937879892e-05, "loss": 0.3674, "step": 17455 }, { "epoch": 0.41, "learning_rate": 2.7104796883433857e-05, "loss": 0.1006, "step": 17460 }, { "epoch": 0.41, "learning_rate": 2.710391682898782e-05, "loss": 0.0892, "step": 17465 }, { "epoch": 0.41, "learning_rate": 2.7103036774541784e-05, "loss": 0.1524, "step": 17470 }, { "epoch": 0.41, "learning_rate": 2.7102156720095752e-05, "loss": 0.2252, "step": 17475 }, { "epoch": 0.41, "learning_rate": 2.7101276665649717e-05, "loss": 0.1578, "step": 17480 }, { "epoch": 0.41, "learning_rate": 2.710039661120368e-05, "loss": 0.2648, "step": 17485 }, { "epoch": 0.41, "learning_rate": 2.7099516556757644e-05, "loss": 0.3085, "step": 17490 }, { "epoch": 0.41, "learning_rate": 2.7098636502311612e-05, "loss": 0.41, "step": 17495 }, { "epoch": 0.41, "learning_rate": 2.7097756447865574e-05, "loss": 0.5623, "step": 17500 }, { "epoch": 0.41, "learning_rate": 2.709687639341954e-05, "loss": 0.2652, "step": 17505 }, { "epoch": 0.41, "learning_rate": 2.7095996338973507e-05, "loss": 0.1019, "step": 17510 }, { "epoch": 0.41, "learning_rate": 2.7095116284527472e-05, "loss": 0.0897, "step": 17515 }, { "epoch": 0.41, "learning_rate": 2.7094236230081434e-05, "loss": 0.1095, "step": 17520 }, { "epoch": 0.41, "learning_rate": 2.70933561756354e-05, "loss": 0.1926, "step": 17525 }, { "epoch": 0.41, "learning_rate": 2.7092476121189367e-05, "loss": 0.2659, "step": 17530 }, { "epoch": 0.41, "learning_rate": 2.709159606674333e-05, "loss": 0.3819, "step": 17535 }, { "epoch": 0.41, "learning_rate": 2.7090716012297294e-05, "loss": 0.3989, "step": 17540 }, { "epoch": 0.41, "learning_rate": 2.708983595785126e-05, "loss": 0.4185, "step": 17545 }, { "epoch": 0.41, "learning_rate": 2.7088955903405227e-05, "loss": 0.8498, "step": 17550 }, { "epoch": 0.41, "learning_rate": 2.708807584895919e-05, "loss": 0.1941, "step": 17555 }, { "epoch": 0.41, "learning_rate": 2.7087195794513154e-05, "loss": 0.16, "step": 17560 }, { "epoch": 0.41, "learning_rate": 2.708631574006712e-05, "loss": 0.2687, "step": 17565 }, { "epoch": 0.41, "learning_rate": 2.7085435685621087e-05, "loss": 0.1134, "step": 17570 }, { "epoch": 0.41, "learning_rate": 2.708455563117505e-05, "loss": 0.2304, "step": 17575 }, { "epoch": 0.41, "learning_rate": 2.7083675576729014e-05, "loss": 0.2857, "step": 17580 }, { "epoch": 0.41, "learning_rate": 2.708279552228298e-05, "loss": 0.1768, "step": 17585 }, { "epoch": 0.41, "learning_rate": 2.7081915467836944e-05, "loss": 0.267, "step": 17590 }, { "epoch": 0.41, "learning_rate": 2.708103541339091e-05, "loss": 0.2727, "step": 17595 }, { "epoch": 0.41, "learning_rate": 2.7080155358944874e-05, "loss": 0.8044, "step": 17600 }, { "epoch": 0.41, "learning_rate": 2.707927530449884e-05, "loss": 0.298, "step": 17605 }, { "epoch": 0.41, "learning_rate": 2.7078395250052804e-05, "loss": 0.2053, "step": 17610 }, { "epoch": 0.41, "learning_rate": 2.707751519560677e-05, "loss": 0.1904, "step": 17615 }, { "epoch": 0.41, "learning_rate": 2.7076635141160734e-05, "loss": 0.1548, "step": 17620 }, { "epoch": 0.41, "learning_rate": 2.7075755086714696e-05, "loss": 0.28, "step": 17625 }, { "epoch": 0.41, "learning_rate": 2.7074875032268664e-05, "loss": 0.1948, "step": 17630 }, { "epoch": 0.41, "learning_rate": 2.707399497782263e-05, "loss": 0.2352, "step": 17635 }, { "epoch": 0.41, "learning_rate": 2.7073114923376594e-05, "loss": 0.3201, "step": 17640 }, { "epoch": 0.41, "learning_rate": 2.707223486893056e-05, "loss": 0.5344, "step": 17645 }, { "epoch": 0.41, "learning_rate": 2.7071354814484524e-05, "loss": 0.5856, "step": 17650 }, { "epoch": 0.41, "learning_rate": 2.707047476003849e-05, "loss": 0.2844, "step": 17655 }, { "epoch": 0.41, "learning_rate": 2.706959470559245e-05, "loss": 0.1077, "step": 17660 }, { "epoch": 0.41, "learning_rate": 2.706871465114642e-05, "loss": 0.1618, "step": 17665 }, { "epoch": 0.41, "learning_rate": 2.7067834596700384e-05, "loss": 0.179, "step": 17670 }, { "epoch": 0.41, "learning_rate": 2.706695454225435e-05, "loss": 0.2457, "step": 17675 }, { "epoch": 0.41, "learning_rate": 2.706607448780831e-05, "loss": 0.2183, "step": 17680 }, { "epoch": 0.41, "learning_rate": 2.706519443336228e-05, "loss": 0.4785, "step": 17685 }, { "epoch": 0.41, "learning_rate": 2.7064314378916245e-05, "loss": 0.3004, "step": 17690 }, { "epoch": 0.41, "learning_rate": 2.7063434324470206e-05, "loss": 0.3905, "step": 17695 }, { "epoch": 0.41, "learning_rate": 2.706255427002417e-05, "loss": 0.5261, "step": 17700 }, { "epoch": 0.41, "learning_rate": 2.706167421557814e-05, "loss": 0.2658, "step": 17705 }, { "epoch": 0.41, "learning_rate": 2.7060794161132105e-05, "loss": 0.0697, "step": 17710 }, { "epoch": 0.41, "learning_rate": 2.7059914106686066e-05, "loss": 0.1318, "step": 17715 }, { "epoch": 0.41, "learning_rate": 2.705903405224003e-05, "loss": 0.1568, "step": 17720 }, { "epoch": 0.41, "learning_rate": 2.7058153997794e-05, "loss": 0.132, "step": 17725 }, { "epoch": 0.41, "learning_rate": 2.705727394334796e-05, "loss": 0.1878, "step": 17730 }, { "epoch": 0.41, "learning_rate": 2.7056393888901926e-05, "loss": 0.3254, "step": 17735 }, { "epoch": 0.41, "learning_rate": 2.705551383445589e-05, "loss": 0.4009, "step": 17740 }, { "epoch": 0.41, "learning_rate": 2.705463378000986e-05, "loss": 0.3773, "step": 17745 }, { "epoch": 0.41, "learning_rate": 2.705375372556382e-05, "loss": 0.5996, "step": 17750 }, { "epoch": 0.41, "learning_rate": 2.7052873671117786e-05, "loss": 0.267, "step": 17755 }, { "epoch": 0.41, "learning_rate": 2.705199361667175e-05, "loss": 0.19, "step": 17760 }, { "epoch": 0.41, "learning_rate": 2.7051113562225716e-05, "loss": 0.0385, "step": 17765 }, { "epoch": 0.41, "learning_rate": 2.705023350777968e-05, "loss": 0.1935, "step": 17770 }, { "epoch": 0.41, "learning_rate": 2.7049353453333646e-05, "loss": 0.3502, "step": 17775 }, { "epoch": 0.41, "learning_rate": 2.704847339888761e-05, "loss": 0.1679, "step": 17780 }, { "epoch": 0.41, "learning_rate": 2.7047593344441577e-05, "loss": 0.2568, "step": 17785 }, { "epoch": 0.42, "learning_rate": 2.704671328999554e-05, "loss": 0.3608, "step": 17790 }, { "epoch": 0.42, "learning_rate": 2.7045833235549507e-05, "loss": 0.4754, "step": 17795 }, { "epoch": 0.42, "learning_rate": 2.704495318110347e-05, "loss": 0.5364, "step": 17800 }, { "epoch": 0.42, "learning_rate": 2.7044073126657437e-05, "loss": 0.3352, "step": 17805 }, { "epoch": 0.42, "learning_rate": 2.70431930722114e-05, "loss": 0.0372, "step": 17810 }, { "epoch": 0.42, "learning_rate": 2.7042313017765367e-05, "loss": 0.1164, "step": 17815 }, { "epoch": 0.42, "learning_rate": 2.704143296331933e-05, "loss": 0.1503, "step": 17820 }, { "epoch": 0.42, "learning_rate": 2.7040552908873297e-05, "loss": 0.1258, "step": 17825 }, { "epoch": 0.42, "learning_rate": 2.703967285442726e-05, "loss": 0.2184, "step": 17830 }, { "epoch": 0.42, "learning_rate": 2.7038792799981223e-05, "loss": 0.2714, "step": 17835 }, { "epoch": 0.42, "learning_rate": 2.7037912745535192e-05, "loss": 0.3235, "step": 17840 }, { "epoch": 0.42, "learning_rate": 2.7037032691089157e-05, "loss": 0.2566, "step": 17845 }, { "epoch": 0.42, "learning_rate": 2.7036152636643122e-05, "loss": 0.6491, "step": 17850 }, { "epoch": 0.42, "learning_rate": 2.7035272582197083e-05, "loss": 0.2771, "step": 17855 }, { "epoch": 0.42, "learning_rate": 2.7034392527751052e-05, "loss": 0.1517, "step": 17860 }, { "epoch": 0.42, "learning_rate": 2.7033512473305017e-05, "loss": 0.204, "step": 17865 }, { "epoch": 0.42, "learning_rate": 2.7032632418858982e-05, "loss": 0.181, "step": 17870 }, { "epoch": 0.42, "learning_rate": 2.7031752364412943e-05, "loss": 0.2256, "step": 17875 }, { "epoch": 0.42, "learning_rate": 2.7030872309966912e-05, "loss": 0.1508, "step": 17880 }, { "epoch": 0.42, "learning_rate": 2.7029992255520877e-05, "loss": 0.2208, "step": 17885 }, { "epoch": 0.42, "learning_rate": 2.702911220107484e-05, "loss": 0.2789, "step": 17890 }, { "epoch": 0.42, "learning_rate": 2.7028232146628804e-05, "loss": 0.4062, "step": 17895 }, { "epoch": 0.42, "learning_rate": 2.7027352092182772e-05, "loss": 0.4671, "step": 17900 }, { "epoch": 0.42, "learning_rate": 2.7026472037736737e-05, "loss": 0.2963, "step": 17905 }, { "epoch": 0.42, "learning_rate": 2.70255919832907e-05, "loss": 0.1387, "step": 17910 }, { "epoch": 0.42, "learning_rate": 2.7024711928844667e-05, "loss": 0.0638, "step": 17915 }, { "epoch": 0.42, "learning_rate": 2.7023831874398632e-05, "loss": 0.2176, "step": 17920 }, { "epoch": 0.42, "learning_rate": 2.7022951819952594e-05, "loss": 0.2722, "step": 17925 }, { "epoch": 0.42, "learning_rate": 2.702207176550656e-05, "loss": 0.1933, "step": 17930 }, { "epoch": 0.42, "learning_rate": 2.7021191711060527e-05, "loss": 0.1906, "step": 17935 }, { "epoch": 0.42, "learning_rate": 2.7020311656614492e-05, "loss": 0.5103, "step": 17940 }, { "epoch": 0.42, "learning_rate": 2.7019431602168454e-05, "loss": 0.5182, "step": 17945 }, { "epoch": 0.42, "learning_rate": 2.701855154772242e-05, "loss": 0.6114, "step": 17950 }, { "epoch": 0.42, "learning_rate": 2.7017671493276387e-05, "loss": 0.3838, "step": 17955 }, { "epoch": 0.42, "learning_rate": 2.701679143883035e-05, "loss": 0.1972, "step": 17960 }, { "epoch": 0.42, "learning_rate": 2.7015911384384314e-05, "loss": 0.0926, "step": 17965 }, { "epoch": 0.42, "learning_rate": 2.701503132993828e-05, "loss": 0.1408, "step": 17970 }, { "epoch": 0.42, "learning_rate": 2.7014151275492247e-05, "loss": 0.3497, "step": 17975 }, { "epoch": 0.42, "learning_rate": 2.701327122104621e-05, "loss": 0.2446, "step": 17980 }, { "epoch": 0.42, "learning_rate": 2.7012391166600174e-05, "loss": 0.316, "step": 17985 }, { "epoch": 0.42, "learning_rate": 2.701151111215414e-05, "loss": 0.3938, "step": 17990 }, { "epoch": 0.42, "learning_rate": 2.7010631057708104e-05, "loss": 0.4565, "step": 17995 }, { "epoch": 0.42, "learning_rate": 2.700975100326207e-05, "loss": 0.6955, "step": 18000 }, { "epoch": 0.42, "learning_rate": 2.7008870948816034e-05, "loss": 0.2319, "step": 18005 }, { "epoch": 0.42, "learning_rate": 2.700799089437e-05, "loss": 0.12, "step": 18010 }, { "epoch": 0.42, "learning_rate": 2.7007110839923964e-05, "loss": 0.1594, "step": 18015 }, { "epoch": 0.42, "learning_rate": 2.700623078547793e-05, "loss": 0.1821, "step": 18020 }, { "epoch": 0.42, "learning_rate": 2.7005350731031894e-05, "loss": 0.1441, "step": 18025 }, { "epoch": 0.42, "learning_rate": 2.7004470676585856e-05, "loss": 0.2853, "step": 18030 }, { "epoch": 0.42, "learning_rate": 2.7003590622139824e-05, "loss": 0.1993, "step": 18035 }, { "epoch": 0.42, "learning_rate": 2.700271056769379e-05, "loss": 0.2863, "step": 18040 }, { "epoch": 0.42, "learning_rate": 2.7001830513247754e-05, "loss": 0.4884, "step": 18045 }, { "epoch": 0.42, "learning_rate": 2.700095045880172e-05, "loss": 0.6162, "step": 18050 }, { "epoch": 0.42, "learning_rate": 2.7000070404355684e-05, "loss": 0.2689, "step": 18055 }, { "epoch": 0.42, "learning_rate": 2.699919034990965e-05, "loss": 0.126, "step": 18060 }, { "epoch": 0.42, "learning_rate": 2.699831029546361e-05, "loss": 0.2037, "step": 18065 }, { "epoch": 0.42, "learning_rate": 2.699743024101758e-05, "loss": 0.1037, "step": 18070 }, { "epoch": 0.42, "learning_rate": 2.6996550186571544e-05, "loss": 0.3225, "step": 18075 }, { "epoch": 0.42, "learning_rate": 2.699567013212551e-05, "loss": 0.2529, "step": 18080 }, { "epoch": 0.42, "learning_rate": 2.699479007767947e-05, "loss": 0.5518, "step": 18085 }, { "epoch": 0.42, "learning_rate": 2.699391002323344e-05, "loss": 0.3619, "step": 18090 }, { "epoch": 0.42, "learning_rate": 2.6993029968787404e-05, "loss": 0.3606, "step": 18095 }, { "epoch": 0.42, "learning_rate": 2.6992149914341366e-05, "loss": 0.4761, "step": 18100 }, { "epoch": 0.42, "learning_rate": 2.699126985989533e-05, "loss": 0.1527, "step": 18105 }, { "epoch": 0.42, "learning_rate": 2.69903898054493e-05, "loss": 0.0715, "step": 18110 }, { "epoch": 0.42, "learning_rate": 2.6989509751003264e-05, "loss": 0.1079, "step": 18115 }, { "epoch": 0.42, "learning_rate": 2.6988629696557226e-05, "loss": 0.129, "step": 18120 }, { "epoch": 0.42, "learning_rate": 2.698774964211119e-05, "loss": 0.1582, "step": 18125 }, { "epoch": 0.42, "learning_rate": 2.698686958766516e-05, "loss": 0.263, "step": 18130 }, { "epoch": 0.42, "learning_rate": 2.698598953321912e-05, "loss": 0.2561, "step": 18135 }, { "epoch": 0.42, "learning_rate": 2.6985109478773086e-05, "loss": 0.2997, "step": 18140 }, { "epoch": 0.42, "learning_rate": 2.698422942432705e-05, "loss": 0.4754, "step": 18145 }, { "epoch": 0.42, "learning_rate": 2.698334936988102e-05, "loss": 0.6432, "step": 18150 }, { "epoch": 0.42, "learning_rate": 2.698246931543498e-05, "loss": 0.2778, "step": 18155 }, { "epoch": 0.42, "learning_rate": 2.6981589260988946e-05, "loss": 0.1583, "step": 18160 }, { "epoch": 0.42, "learning_rate": 2.698070920654291e-05, "loss": 0.0924, "step": 18165 }, { "epoch": 0.42, "learning_rate": 2.697982915209688e-05, "loss": 0.1408, "step": 18170 }, { "epoch": 0.42, "learning_rate": 2.697894909765084e-05, "loss": 0.2624, "step": 18175 }, { "epoch": 0.42, "learning_rate": 2.6978069043204806e-05, "loss": 0.2139, "step": 18180 }, { "epoch": 0.42, "learning_rate": 2.697718898875877e-05, "loss": 0.2714, "step": 18185 }, { "epoch": 0.42, "learning_rate": 2.6976308934312736e-05, "loss": 0.2689, "step": 18190 }, { "epoch": 0.42, "learning_rate": 2.69754288798667e-05, "loss": 0.5474, "step": 18195 }, { "epoch": 0.42, "learning_rate": 2.6974548825420666e-05, "loss": 0.8021, "step": 18200 }, { "epoch": 0.42, "learning_rate": 2.6973668770974635e-05, "loss": 0.3436, "step": 18205 }, { "epoch": 0.42, "learning_rate": 2.6972788716528596e-05, "loss": 0.2143, "step": 18210 }, { "epoch": 0.42, "learning_rate": 2.697190866208256e-05, "loss": 0.1671, "step": 18215 }, { "epoch": 0.43, "learning_rate": 2.6971028607636527e-05, "loss": 0.1102, "step": 18220 }, { "epoch": 0.43, "learning_rate": 2.697014855319049e-05, "loss": 0.2147, "step": 18225 }, { "epoch": 0.43, "learning_rate": 2.6969268498744457e-05, "loss": 0.1501, "step": 18230 }, { "epoch": 0.43, "learning_rate": 2.696838844429842e-05, "loss": 0.1883, "step": 18235 }, { "epoch": 0.43, "learning_rate": 2.6967508389852387e-05, "loss": 0.2441, "step": 18240 }, { "epoch": 0.43, "learning_rate": 2.696662833540635e-05, "loss": 0.2824, "step": 18245 }, { "epoch": 0.43, "learning_rate": 2.6965748280960317e-05, "loss": 0.4125, "step": 18250 }, { "epoch": 0.43, "learning_rate": 2.696486822651428e-05, "loss": 0.2872, "step": 18255 }, { "epoch": 0.43, "learning_rate": 2.6963988172068243e-05, "loss": 0.0447, "step": 18260 }, { "epoch": 0.43, "learning_rate": 2.6963108117622212e-05, "loss": 0.1254, "step": 18265 }, { "epoch": 0.43, "learning_rate": 2.6962228063176177e-05, "loss": 0.2663, "step": 18270 }, { "epoch": 0.43, "learning_rate": 2.6961348008730142e-05, "loss": 0.3185, "step": 18275 }, { "epoch": 0.43, "learning_rate": 2.6960467954284103e-05, "loss": 0.1412, "step": 18280 }, { "epoch": 0.43, "learning_rate": 2.6959587899838072e-05, "loss": 0.1871, "step": 18285 }, { "epoch": 0.43, "learning_rate": 2.6958707845392037e-05, "loss": 0.1649, "step": 18290 }, { "epoch": 0.43, "learning_rate": 2.6957827790946e-05, "loss": 0.3983, "step": 18295 }, { "epoch": 0.43, "learning_rate": 2.6956947736499963e-05, "loss": 0.6683, "step": 18300 }, { "epoch": 0.43, "learning_rate": 2.6956067682053932e-05, "loss": 0.1832, "step": 18305 }, { "epoch": 0.43, "learning_rate": 2.6955187627607897e-05, "loss": 0.1089, "step": 18310 }, { "epoch": 0.43, "learning_rate": 2.695430757316186e-05, "loss": 0.1716, "step": 18315 }, { "epoch": 0.43, "learning_rate": 2.6953427518715824e-05, "loss": 0.146, "step": 18320 }, { "epoch": 0.43, "learning_rate": 2.6952547464269792e-05, "loss": 0.2263, "step": 18325 }, { "epoch": 0.43, "learning_rate": 2.6951667409823754e-05, "loss": 0.2615, "step": 18330 }, { "epoch": 0.43, "learning_rate": 2.695078735537772e-05, "loss": 0.2322, "step": 18335 }, { "epoch": 0.43, "learning_rate": 2.6949907300931687e-05, "loss": 0.2573, "step": 18340 }, { "epoch": 0.43, "learning_rate": 2.6949027246485652e-05, "loss": 0.3667, "step": 18345 }, { "epoch": 0.43, "learning_rate": 2.6948147192039614e-05, "loss": 0.6089, "step": 18350 }, { "epoch": 0.43, "learning_rate": 2.694726713759358e-05, "loss": 0.2522, "step": 18355 }, { "epoch": 0.43, "learning_rate": 2.6946387083147547e-05, "loss": 0.1135, "step": 18360 }, { "epoch": 0.43, "learning_rate": 2.694550702870151e-05, "loss": 0.1099, "step": 18365 }, { "epoch": 0.43, "learning_rate": 2.6944626974255474e-05, "loss": 0.2695, "step": 18370 }, { "epoch": 0.43, "learning_rate": 2.694374691980944e-05, "loss": 0.1541, "step": 18375 }, { "epoch": 0.43, "learning_rate": 2.6942866865363407e-05, "loss": 0.255, "step": 18380 }, { "epoch": 0.43, "learning_rate": 2.694198681091737e-05, "loss": 0.4014, "step": 18385 }, { "epoch": 0.43, "learning_rate": 2.6941106756471334e-05, "loss": 0.3279, "step": 18390 }, { "epoch": 0.43, "learning_rate": 2.69402267020253e-05, "loss": 0.4063, "step": 18395 }, { "epoch": 0.43, "learning_rate": 2.6939346647579264e-05, "loss": 0.6448, "step": 18400 }, { "epoch": 0.43, "learning_rate": 2.693846659313323e-05, "loss": 0.2334, "step": 18405 }, { "epoch": 0.43, "learning_rate": 2.6937586538687194e-05, "loss": 0.1198, "step": 18410 }, { "epoch": 0.43, "learning_rate": 2.693670648424116e-05, "loss": 0.1271, "step": 18415 }, { "epoch": 0.43, "learning_rate": 2.6935826429795124e-05, "loss": 0.2079, "step": 18420 }, { "epoch": 0.43, "learning_rate": 2.693494637534909e-05, "loss": 0.0919, "step": 18425 }, { "epoch": 0.43, "learning_rate": 2.6934066320903054e-05, "loss": 0.1388, "step": 18430 }, { "epoch": 0.43, "learning_rate": 2.6933186266457016e-05, "loss": 0.2774, "step": 18435 }, { "epoch": 0.43, "learning_rate": 2.693248222290019e-05, "loss": 0.2868, "step": 18440 }, { "epoch": 0.43, "learning_rate": 2.6931602168454155e-05, "loss": 0.4267, "step": 18445 }, { "epoch": 0.43, "learning_rate": 2.693072211400812e-05, "loss": 0.6971, "step": 18450 }, { "epoch": 0.43, "learning_rate": 2.6929842059562085e-05, "loss": 0.2816, "step": 18455 }, { "epoch": 0.43, "learning_rate": 2.692896200511605e-05, "loss": 0.1367, "step": 18460 }, { "epoch": 0.43, "learning_rate": 2.6928081950670015e-05, "loss": 0.1549, "step": 18465 }, { "epoch": 0.43, "learning_rate": 2.692720189622398e-05, "loss": 0.2124, "step": 18470 }, { "epoch": 0.43, "learning_rate": 2.6926321841777945e-05, "loss": 0.2265, "step": 18475 }, { "epoch": 0.43, "learning_rate": 2.692544178733191e-05, "loss": 0.1954, "step": 18480 }, { "epoch": 0.43, "learning_rate": 2.6924561732885876e-05, "loss": 0.2535, "step": 18485 }, { "epoch": 0.43, "learning_rate": 2.6923681678439837e-05, "loss": 0.2563, "step": 18490 }, { "epoch": 0.43, "learning_rate": 2.6922801623993806e-05, "loss": 0.3442, "step": 18495 }, { "epoch": 0.43, "learning_rate": 2.692192156954777e-05, "loss": 0.6223, "step": 18500 }, { "epoch": 0.43, "learning_rate": 2.6921041515101736e-05, "loss": 0.4213, "step": 18505 }, { "epoch": 0.43, "learning_rate": 2.6920161460655697e-05, "loss": 0.1454, "step": 18510 }, { "epoch": 0.43, "learning_rate": 2.6919281406209666e-05, "loss": 0.1006, "step": 18515 }, { "epoch": 0.43, "learning_rate": 2.691840135176363e-05, "loss": 0.0958, "step": 18520 }, { "epoch": 0.43, "learning_rate": 2.6917521297317592e-05, "loss": 0.2614, "step": 18525 }, { "epoch": 0.43, "learning_rate": 2.691664124287156e-05, "loss": 0.2091, "step": 18530 }, { "epoch": 0.43, "learning_rate": 2.6915761188425526e-05, "loss": 0.1849, "step": 18535 }, { "epoch": 0.43, "learning_rate": 2.691488113397949e-05, "loss": 0.2184, "step": 18540 }, { "epoch": 0.43, "learning_rate": 2.6914001079533452e-05, "loss": 0.3666, "step": 18545 }, { "epoch": 0.43, "learning_rate": 2.691312102508742e-05, "loss": 0.4932, "step": 18550 }, { "epoch": 0.43, "learning_rate": 2.6912240970641386e-05, "loss": 0.381, "step": 18555 }, { "epoch": 0.43, "learning_rate": 2.6911360916195347e-05, "loss": 0.1105, "step": 18560 }, { "epoch": 0.43, "learning_rate": 2.6910480861749312e-05, "loss": 0.0593, "step": 18565 }, { "epoch": 0.43, "learning_rate": 2.690960080730328e-05, "loss": 0.1499, "step": 18570 }, { "epoch": 0.43, "learning_rate": 2.6908720752857246e-05, "loss": 0.1508, "step": 18575 }, { "epoch": 0.43, "learning_rate": 2.6907840698411208e-05, "loss": 0.3568, "step": 18580 }, { "epoch": 0.43, "learning_rate": 2.6906960643965173e-05, "loss": 0.2201, "step": 18585 }, { "epoch": 0.43, "learning_rate": 2.690608058951914e-05, "loss": 0.4301, "step": 18590 }, { "epoch": 0.43, "learning_rate": 2.6905200535073103e-05, "loss": 0.2685, "step": 18595 }, { "epoch": 0.43, "learning_rate": 2.6904320480627068e-05, "loss": 0.6324, "step": 18600 }, { "epoch": 0.43, "learning_rate": 2.6903440426181033e-05, "loss": 0.2367, "step": 18605 }, { "epoch": 0.43, "learning_rate": 2.6902560371735e-05, "loss": 0.2244, "step": 18610 }, { "epoch": 0.43, "learning_rate": 2.6901680317288963e-05, "loss": 0.1555, "step": 18615 }, { "epoch": 0.43, "learning_rate": 2.6900800262842928e-05, "loss": 0.0968, "step": 18620 }, { "epoch": 0.43, "learning_rate": 2.6899920208396893e-05, "loss": 0.1406, "step": 18625 }, { "epoch": 0.43, "learning_rate": 2.6899040153950858e-05, "loss": 0.1487, "step": 18630 }, { "epoch": 0.43, "learning_rate": 2.6898160099504823e-05, "loss": 0.2964, "step": 18635 }, { "epoch": 0.43, "learning_rate": 2.6897280045058788e-05, "loss": 0.483, "step": 18640 }, { "epoch": 0.44, "learning_rate": 2.6896399990612753e-05, "loss": 0.3915, "step": 18645 }, { "epoch": 0.44, "learning_rate": 2.6895519936166718e-05, "loss": 0.5271, "step": 18650 }, { "epoch": 0.44, "learning_rate": 2.6894639881720683e-05, "loss": 0.2305, "step": 18655 }, { "epoch": 0.44, "learning_rate": 2.6893759827274648e-05, "loss": 0.1436, "step": 18660 }, { "epoch": 0.44, "learning_rate": 2.6892879772828616e-05, "loss": 0.1713, "step": 18665 }, { "epoch": 0.44, "learning_rate": 2.6891999718382578e-05, "loss": 0.114, "step": 18670 }, { "epoch": 0.44, "learning_rate": 2.6891119663936543e-05, "loss": 0.1752, "step": 18675 }, { "epoch": 0.44, "learning_rate": 2.6890239609490508e-05, "loss": 0.2475, "step": 18680 }, { "epoch": 0.44, "learning_rate": 2.6889359555044473e-05, "loss": 0.2192, "step": 18685 }, { "epoch": 0.44, "learning_rate": 2.6888479500598438e-05, "loss": 0.1786, "step": 18690 }, { "epoch": 0.44, "learning_rate": 2.6887599446152403e-05, "loss": 0.3275, "step": 18695 }, { "epoch": 0.44, "learning_rate": 2.6886719391706368e-05, "loss": 0.6413, "step": 18700 }, { "epoch": 0.44, "learning_rate": 2.6885839337260333e-05, "loss": 0.2528, "step": 18705 }, { "epoch": 0.44, "learning_rate": 2.6884959282814298e-05, "loss": 0.1551, "step": 18710 }, { "epoch": 0.44, "learning_rate": 2.6884079228368263e-05, "loss": 0.1701, "step": 18715 }, { "epoch": 0.44, "learning_rate": 2.6883199173922225e-05, "loss": 0.1474, "step": 18720 }, { "epoch": 0.44, "learning_rate": 2.6882319119476193e-05, "loss": 0.1424, "step": 18725 }, { "epoch": 0.44, "learning_rate": 2.6881439065030158e-05, "loss": 0.333, "step": 18730 }, { "epoch": 0.44, "learning_rate": 2.6880559010584123e-05, "loss": 0.1931, "step": 18735 }, { "epoch": 0.44, "learning_rate": 2.6879678956138085e-05, "loss": 0.3847, "step": 18740 }, { "epoch": 0.44, "learning_rate": 2.6878798901692053e-05, "loss": 0.3612, "step": 18745 }, { "epoch": 0.44, "learning_rate": 2.6877918847246018e-05, "loss": 0.7876, "step": 18750 }, { "epoch": 0.44, "learning_rate": 2.687703879279998e-05, "loss": 0.2412, "step": 18755 }, { "epoch": 0.44, "learning_rate": 2.6876158738353945e-05, "loss": 0.1489, "step": 18760 }, { "epoch": 0.44, "learning_rate": 2.6875278683907913e-05, "loss": 0.0997, "step": 18765 }, { "epoch": 0.44, "learning_rate": 2.687439862946188e-05, "loss": 0.2771, "step": 18770 }, { "epoch": 0.44, "learning_rate": 2.687351857501584e-05, "loss": 0.2375, "step": 18775 }, { "epoch": 0.44, "learning_rate": 2.6872638520569805e-05, "loss": 0.2253, "step": 18780 }, { "epoch": 0.44, "learning_rate": 2.6871758466123773e-05, "loss": 0.2599, "step": 18785 }, { "epoch": 0.44, "learning_rate": 2.6870878411677735e-05, "loss": 0.4404, "step": 18790 }, { "epoch": 0.44, "learning_rate": 2.68699983572317e-05, "loss": 0.2599, "step": 18795 }, { "epoch": 0.44, "learning_rate": 2.686911830278567e-05, "loss": 0.7216, "step": 18800 }, { "epoch": 0.44, "learning_rate": 2.6868238248339633e-05, "loss": 0.3877, "step": 18805 }, { "epoch": 0.44, "learning_rate": 2.6867358193893595e-05, "loss": 0.2458, "step": 18810 }, { "epoch": 0.44, "learning_rate": 2.686647813944756e-05, "loss": 0.0853, "step": 18815 }, { "epoch": 0.44, "learning_rate": 2.686559808500153e-05, "loss": 0.1498, "step": 18820 }, { "epoch": 0.44, "learning_rate": 2.686471803055549e-05, "loss": 0.2121, "step": 18825 }, { "epoch": 0.44, "learning_rate": 2.6863837976109455e-05, "loss": 0.1486, "step": 18830 }, { "epoch": 0.44, "learning_rate": 2.686295792166342e-05, "loss": 0.4181, "step": 18835 }, { "epoch": 0.44, "learning_rate": 2.686207786721739e-05, "loss": 0.265, "step": 18840 }, { "epoch": 0.44, "learning_rate": 2.686119781277135e-05, "loss": 0.2995, "step": 18845 }, { "epoch": 0.44, "learning_rate": 2.6860317758325315e-05, "loss": 0.624, "step": 18850 }, { "epoch": 0.44, "learning_rate": 2.685943770387928e-05, "loss": 0.2354, "step": 18855 }, { "epoch": 0.44, "learning_rate": 2.6858557649433245e-05, "loss": 0.159, "step": 18860 }, { "epoch": 0.44, "learning_rate": 2.685767759498721e-05, "loss": 0.2108, "step": 18865 }, { "epoch": 0.44, "learning_rate": 2.6856797540541175e-05, "loss": 0.2256, "step": 18870 }, { "epoch": 0.44, "learning_rate": 2.685591748609514e-05, "loss": 0.2194, "step": 18875 }, { "epoch": 0.44, "learning_rate": 2.6855037431649105e-05, "loss": 0.1653, "step": 18880 }, { "epoch": 0.44, "learning_rate": 2.685415737720307e-05, "loss": 0.1515, "step": 18885 }, { "epoch": 0.44, "learning_rate": 2.6853277322757035e-05, "loss": 0.4276, "step": 18890 }, { "epoch": 0.44, "learning_rate": 2.6852397268310997e-05, "loss": 0.3775, "step": 18895 }, { "epoch": 0.44, "learning_rate": 2.6851517213864965e-05, "loss": 0.5506, "step": 18900 }, { "epoch": 0.44, "learning_rate": 2.685063715941893e-05, "loss": 0.3496, "step": 18905 }, { "epoch": 0.44, "learning_rate": 2.6849757104972896e-05, "loss": 0.099, "step": 18910 }, { "epoch": 0.44, "learning_rate": 2.6848877050526857e-05, "loss": 0.1318, "step": 18915 }, { "epoch": 0.44, "learning_rate": 2.6847996996080826e-05, "loss": 0.1489, "step": 18920 }, { "epoch": 0.44, "learning_rate": 2.684711694163479e-05, "loss": 0.1601, "step": 18925 }, { "epoch": 0.44, "learning_rate": 2.6846236887188752e-05, "loss": 0.2674, "step": 18930 }, { "epoch": 0.44, "learning_rate": 2.684535683274272e-05, "loss": 0.2729, "step": 18935 }, { "epoch": 0.44, "learning_rate": 2.6844476778296686e-05, "loss": 0.3707, "step": 18940 }, { "epoch": 0.44, "learning_rate": 2.684359672385065e-05, "loss": 0.2904, "step": 18945 }, { "epoch": 0.44, "learning_rate": 2.6842716669404612e-05, "loss": 0.6174, "step": 18950 }, { "epoch": 0.44, "learning_rate": 2.684183661495858e-05, "loss": 0.2862, "step": 18955 }, { "epoch": 0.44, "learning_rate": 2.6840956560512546e-05, "loss": 0.1323, "step": 18960 }, { "epoch": 0.44, "learning_rate": 2.684007650606651e-05, "loss": 0.0606, "step": 18965 }, { "epoch": 0.44, "learning_rate": 2.6839196451620472e-05, "loss": 0.1209, "step": 18970 }, { "epoch": 0.44, "learning_rate": 2.683831639717444e-05, "loss": 0.2862, "step": 18975 }, { "epoch": 0.44, "learning_rate": 2.6837436342728406e-05, "loss": 0.1302, "step": 18980 }, { "epoch": 0.44, "learning_rate": 2.6836556288282367e-05, "loss": 0.4079, "step": 18985 }, { "epoch": 0.44, "learning_rate": 2.6835676233836332e-05, "loss": 0.2648, "step": 18990 }, { "epoch": 0.44, "learning_rate": 2.68347961793903e-05, "loss": 0.3285, "step": 18995 }, { "epoch": 0.44, "learning_rate": 2.6833916124944266e-05, "loss": 0.5398, "step": 19000 }, { "epoch": 0.44, "learning_rate": 2.6833036070498228e-05, "loss": 0.1712, "step": 19005 }, { "epoch": 0.44, "learning_rate": 2.6832156016052193e-05, "loss": 0.1142, "step": 19010 }, { "epoch": 0.44, "learning_rate": 2.683127596160616e-05, "loss": 0.1683, "step": 19015 }, { "epoch": 0.44, "learning_rate": 2.6830395907160123e-05, "loss": 0.1942, "step": 19020 }, { "epoch": 0.44, "learning_rate": 2.6829515852714088e-05, "loss": 0.2422, "step": 19025 }, { "epoch": 0.44, "learning_rate": 2.6828635798268053e-05, "loss": 0.3857, "step": 19030 }, { "epoch": 0.44, "learning_rate": 2.682775574382202e-05, "loss": 0.2582, "step": 19035 }, { "epoch": 0.44, "learning_rate": 2.6826875689375983e-05, "loss": 0.3832, "step": 19040 }, { "epoch": 0.44, "learning_rate": 2.6825995634929948e-05, "loss": 0.3951, "step": 19045 }, { "epoch": 0.44, "learning_rate": 2.6825115580483913e-05, "loss": 0.7767, "step": 19050 }, { "epoch": 0.44, "learning_rate": 2.6824235526037878e-05, "loss": 0.1783, "step": 19055 }, { "epoch": 0.44, "learning_rate": 2.6823355471591843e-05, "loss": 0.1273, "step": 19060 }, { "epoch": 0.44, "learning_rate": 2.6822475417145808e-05, "loss": 0.1994, "step": 19065 }, { "epoch": 0.44, "learning_rate": 2.6821595362699776e-05, "loss": 0.1625, "step": 19070 }, { "epoch": 0.45, "learning_rate": 2.6820715308253738e-05, "loss": 0.1298, "step": 19075 }, { "epoch": 0.45, "learning_rate": 2.6819835253807703e-05, "loss": 0.2304, "step": 19080 }, { "epoch": 0.45, "learning_rate": 2.6818955199361668e-05, "loss": 0.2567, "step": 19085 }, { "epoch": 0.45, "learning_rate": 2.6818075144915633e-05, "loss": 0.5195, "step": 19090 }, { "epoch": 0.45, "learning_rate": 2.6817195090469598e-05, "loss": 0.4362, "step": 19095 }, { "epoch": 0.45, "learning_rate": 2.6816315036023563e-05, "loss": 0.5998, "step": 19100 }, { "epoch": 0.45, "learning_rate": 2.6815434981577528e-05, "loss": 0.2016, "step": 19105 }, { "epoch": 0.45, "learning_rate": 2.6814554927131493e-05, "loss": 0.0832, "step": 19110 }, { "epoch": 0.45, "learning_rate": 2.6813674872685458e-05, "loss": 0.1327, "step": 19115 }, { "epoch": 0.45, "learning_rate": 2.6812794818239423e-05, "loss": 0.2731, "step": 19120 }, { "epoch": 0.45, "learning_rate": 2.6811914763793385e-05, "loss": 0.1332, "step": 19125 }, { "epoch": 0.45, "learning_rate": 2.6811034709347353e-05, "loss": 0.2363, "step": 19130 }, { "epoch": 0.45, "learning_rate": 2.6810154654901318e-05, "loss": 0.2138, "step": 19135 }, { "epoch": 0.45, "learning_rate": 2.6809274600455283e-05, "loss": 0.187, "step": 19140 }, { "epoch": 0.45, "learning_rate": 2.6808394546009245e-05, "loss": 0.416, "step": 19145 }, { "epoch": 0.45, "learning_rate": 2.6807514491563213e-05, "loss": 0.4437, "step": 19150 }, { "epoch": 0.45, "learning_rate": 2.6806634437117178e-05, "loss": 0.3756, "step": 19155 }, { "epoch": 0.45, "learning_rate": 2.680575438267114e-05, "loss": 0.1061, "step": 19160 }, { "epoch": 0.45, "learning_rate": 2.6804874328225105e-05, "loss": 0.1134, "step": 19165 }, { "epoch": 0.45, "learning_rate": 2.6803994273779073e-05, "loss": 0.2017, "step": 19170 }, { "epoch": 0.45, "learning_rate": 2.6803114219333038e-05, "loss": 0.2922, "step": 19175 }, { "epoch": 0.45, "learning_rate": 2.6802234164887e-05, "loss": 0.2861, "step": 19180 }, { "epoch": 0.45, "learning_rate": 2.6801354110440965e-05, "loss": 0.1768, "step": 19185 }, { "epoch": 0.45, "learning_rate": 2.6800474055994933e-05, "loss": 0.4428, "step": 19190 }, { "epoch": 0.45, "learning_rate": 2.6799594001548895e-05, "loss": 0.5692, "step": 19195 }, { "epoch": 0.45, "learning_rate": 2.679871394710286e-05, "loss": 0.5965, "step": 19200 }, { "epoch": 0.45, "learning_rate": 2.679783389265683e-05, "loss": 0.2241, "step": 19205 }, { "epoch": 0.45, "learning_rate": 2.6796953838210793e-05, "loss": 0.1938, "step": 19210 }, { "epoch": 0.45, "learning_rate": 2.6796073783764755e-05, "loss": 0.1322, "step": 19215 }, { "epoch": 0.45, "learning_rate": 2.679519372931872e-05, "loss": 0.243, "step": 19220 }, { "epoch": 0.45, "learning_rate": 2.679431367487269e-05, "loss": 0.2007, "step": 19225 }, { "epoch": 0.45, "learning_rate": 2.679343362042665e-05, "loss": 0.1992, "step": 19230 }, { "epoch": 0.45, "learning_rate": 2.6792553565980615e-05, "loss": 0.322, "step": 19235 }, { "epoch": 0.45, "learning_rate": 2.679167351153458e-05, "loss": 0.3981, "step": 19240 }, { "epoch": 0.45, "learning_rate": 2.679079345708855e-05, "loss": 0.4799, "step": 19245 }, { "epoch": 0.45, "learning_rate": 2.678991340264251e-05, "loss": 0.7518, "step": 19250 }, { "epoch": 0.45, "learning_rate": 2.6789033348196475e-05, "loss": 0.2489, "step": 19255 }, { "epoch": 0.45, "learning_rate": 2.678815329375044e-05, "loss": 0.1961, "step": 19260 }, { "epoch": 0.45, "learning_rate": 2.678727323930441e-05, "loss": 0.1212, "step": 19265 }, { "epoch": 0.45, "learning_rate": 2.678639318485837e-05, "loss": 0.1682, "step": 19270 }, { "epoch": 0.45, "learning_rate": 2.6785513130412335e-05, "loss": 0.2331, "step": 19275 }, { "epoch": 0.45, "learning_rate": 2.67846330759663e-05, "loss": 0.3747, "step": 19280 }, { "epoch": 0.45, "learning_rate": 2.6783753021520265e-05, "loss": 0.265, "step": 19285 }, { "epoch": 0.45, "learning_rate": 2.678287296707423e-05, "loss": 0.3795, "step": 19290 }, { "epoch": 0.45, "learning_rate": 2.6781992912628195e-05, "loss": 0.3403, "step": 19295 }, { "epoch": 0.45, "learning_rate": 2.678111285818216e-05, "loss": 0.5147, "step": 19300 }, { "epoch": 0.45, "learning_rate": 2.6780232803736125e-05, "loss": 0.2272, "step": 19305 }, { "epoch": 0.45, "learning_rate": 2.677935274929009e-05, "loss": 0.1249, "step": 19310 }, { "epoch": 0.45, "learning_rate": 2.6778472694844055e-05, "loss": 0.1506, "step": 19315 }, { "epoch": 0.45, "learning_rate": 2.6777592640398017e-05, "loss": 0.1723, "step": 19320 }, { "epoch": 0.45, "learning_rate": 2.6776712585951985e-05, "loss": 0.2831, "step": 19325 }, { "epoch": 0.45, "learning_rate": 2.677583253150595e-05, "loss": 0.275, "step": 19330 }, { "epoch": 0.45, "learning_rate": 2.6774952477059915e-05, "loss": 0.3106, "step": 19335 }, { "epoch": 0.45, "learning_rate": 2.677407242261388e-05, "loss": 0.3803, "step": 19340 }, { "epoch": 0.45, "learning_rate": 2.6773192368167846e-05, "loss": 0.5567, "step": 19345 }, { "epoch": 0.45, "learning_rate": 2.677231231372181e-05, "loss": 1.1133, "step": 19350 }, { "epoch": 0.45, "learning_rate": 2.6771432259275772e-05, "loss": 0.242, "step": 19355 }, { "epoch": 0.45, "learning_rate": 2.677055220482974e-05, "loss": 0.1263, "step": 19360 }, { "epoch": 0.45, "learning_rate": 2.6769672150383706e-05, "loss": 0.0973, "step": 19365 }, { "epoch": 0.45, "learning_rate": 2.676879209593767e-05, "loss": 0.3194, "step": 19370 }, { "epoch": 0.45, "learning_rate": 2.6767912041491632e-05, "loss": 0.165, "step": 19375 }, { "epoch": 0.45, "learning_rate": 2.67670319870456e-05, "loss": 0.2259, "step": 19380 }, { "epoch": 0.45, "learning_rate": 2.6766151932599566e-05, "loss": 0.2255, "step": 19385 }, { "epoch": 0.45, "learning_rate": 2.6765271878153527e-05, "loss": 0.3557, "step": 19390 }, { "epoch": 0.45, "learning_rate": 2.6764391823707492e-05, "loss": 0.3789, "step": 19395 }, { "epoch": 0.45, "learning_rate": 2.676351176926146e-05, "loss": 0.6432, "step": 19400 }, { "epoch": 0.45, "learning_rate": 2.6762631714815426e-05, "loss": 0.2128, "step": 19405 }, { "epoch": 0.45, "learning_rate": 2.6761751660369387e-05, "loss": 0.1298, "step": 19410 }, { "epoch": 0.45, "learning_rate": 2.6760871605923352e-05, "loss": 0.1014, "step": 19415 }, { "epoch": 0.45, "learning_rate": 2.675999155147732e-05, "loss": 0.1118, "step": 19420 }, { "epoch": 0.45, "learning_rate": 2.6759111497031282e-05, "loss": 0.2325, "step": 19425 }, { "epoch": 0.45, "learning_rate": 2.6758231442585247e-05, "loss": 0.3632, "step": 19430 }, { "epoch": 0.45, "learning_rate": 2.6757351388139212e-05, "loss": 0.2755, "step": 19435 }, { "epoch": 0.45, "learning_rate": 2.675647133369318e-05, "loss": 0.4224, "step": 19440 }, { "epoch": 0.45, "learning_rate": 2.6755591279247143e-05, "loss": 0.4625, "step": 19445 }, { "epoch": 0.45, "learning_rate": 2.6754711224801108e-05, "loss": 0.8556, "step": 19450 }, { "epoch": 0.45, "learning_rate": 2.6753831170355073e-05, "loss": 0.2749, "step": 19455 }, { "epoch": 0.45, "learning_rate": 2.6752951115909038e-05, "loss": 0.0867, "step": 19460 }, { "epoch": 0.45, "learning_rate": 2.6752071061463003e-05, "loss": 0.1869, "step": 19465 }, { "epoch": 0.45, "learning_rate": 2.6751191007016968e-05, "loss": 0.1983, "step": 19470 }, { "epoch": 0.45, "learning_rate": 2.6750310952570936e-05, "loss": 0.2626, "step": 19475 }, { "epoch": 0.45, "learning_rate": 2.6749430898124898e-05, "loss": 0.217, "step": 19480 }, { "epoch": 0.45, "learning_rate": 2.6748550843678863e-05, "loss": 0.2973, "step": 19485 }, { "epoch": 0.45, "learning_rate": 2.6747670789232828e-05, "loss": 0.2135, "step": 19490 }, { "epoch": 0.45, "learning_rate": 2.6746790734786793e-05, "loss": 0.6211, "step": 19495 }, { "epoch": 0.45, "learning_rate": 2.6745910680340758e-05, "loss": 0.6901, "step": 19500 }, { "epoch": 0.46, "learning_rate": 2.6745030625894723e-05, "loss": 0.2861, "step": 19505 }, { "epoch": 0.46, "learning_rate": 2.6744150571448688e-05, "loss": 0.168, "step": 19510 }, { "epoch": 0.46, "learning_rate": 2.6743270517002653e-05, "loss": 0.1795, "step": 19515 }, { "epoch": 0.46, "learning_rate": 2.6742390462556618e-05, "loss": 0.2479, "step": 19520 }, { "epoch": 0.46, "learning_rate": 2.6741510408110583e-05, "loss": 0.1596, "step": 19525 }, { "epoch": 0.46, "learning_rate": 2.6740630353664544e-05, "loss": 0.1532, "step": 19530 }, { "epoch": 0.46, "learning_rate": 2.6739750299218513e-05, "loss": 0.2667, "step": 19535 }, { "epoch": 0.46, "learning_rate": 2.6738870244772478e-05, "loss": 0.441, "step": 19540 }, { "epoch": 0.46, "learning_rate": 2.6737990190326443e-05, "loss": 0.2993, "step": 19545 }, { "epoch": 0.46, "learning_rate": 2.6737110135880405e-05, "loss": 0.8235, "step": 19550 }, { "epoch": 0.46, "learning_rate": 2.6736230081434373e-05, "loss": 0.1824, "step": 19555 }, { "epoch": 0.46, "learning_rate": 2.6735350026988338e-05, "loss": 0.0475, "step": 19560 }, { "epoch": 0.46, "learning_rate": 2.67344699725423e-05, "loss": 0.2444, "step": 19565 }, { "epoch": 0.46, "learning_rate": 2.6733589918096265e-05, "loss": 0.141, "step": 19570 }, { "epoch": 0.46, "learning_rate": 2.6732709863650233e-05, "loss": 0.1181, "step": 19575 }, { "epoch": 0.46, "learning_rate": 2.6731829809204198e-05, "loss": 0.248, "step": 19580 }, { "epoch": 0.46, "learning_rate": 2.673094975475816e-05, "loss": 0.0907, "step": 19585 }, { "epoch": 0.46, "learning_rate": 2.6730069700312125e-05, "loss": 0.2532, "step": 19590 }, { "epoch": 0.46, "learning_rate": 2.6729189645866093e-05, "loss": 0.3843, "step": 19595 }, { "epoch": 0.46, "learning_rate": 2.6728309591420058e-05, "loss": 0.5012, "step": 19600 }, { "epoch": 0.46, "learning_rate": 2.672742953697402e-05, "loss": 0.2306, "step": 19605 }, { "epoch": 0.46, "learning_rate": 2.6726549482527988e-05, "loss": 0.1466, "step": 19610 }, { "epoch": 0.46, "learning_rate": 2.6725669428081953e-05, "loss": 0.1075, "step": 19615 }, { "epoch": 0.46, "learning_rate": 2.6724789373635915e-05, "loss": 0.0621, "step": 19620 }, { "epoch": 0.46, "learning_rate": 2.672390931918988e-05, "loss": 0.2987, "step": 19625 }, { "epoch": 0.46, "learning_rate": 2.6723029264743848e-05, "loss": 0.2222, "step": 19630 }, { "epoch": 0.46, "learning_rate": 2.6722149210297813e-05, "loss": 0.3289, "step": 19635 }, { "epoch": 0.46, "learning_rate": 2.6721269155851775e-05, "loss": 0.1625, "step": 19640 }, { "epoch": 0.46, "learning_rate": 2.672038910140574e-05, "loss": 0.3905, "step": 19645 }, { "epoch": 0.46, "learning_rate": 2.671950904695971e-05, "loss": 0.4511, "step": 19650 }, { "epoch": 0.46, "learning_rate": 2.671862899251367e-05, "loss": 0.3855, "step": 19655 }, { "epoch": 0.46, "learning_rate": 2.6717748938067635e-05, "loss": 0.1751, "step": 19660 }, { "epoch": 0.46, "learning_rate": 2.67168688836216e-05, "loss": 0.1693, "step": 19665 }, { "epoch": 0.46, "learning_rate": 2.671598882917557e-05, "loss": 0.1501, "step": 19670 }, { "epoch": 0.46, "learning_rate": 2.671510877472953e-05, "loss": 0.1204, "step": 19675 }, { "epoch": 0.46, "learning_rate": 2.6714228720283495e-05, "loss": 0.3289, "step": 19680 }, { "epoch": 0.46, "learning_rate": 2.671334866583746e-05, "loss": 0.2853, "step": 19685 }, { "epoch": 0.46, "learning_rate": 2.6712468611391425e-05, "loss": 0.3563, "step": 19690 }, { "epoch": 0.46, "learning_rate": 2.671158855694539e-05, "loss": 0.4088, "step": 19695 }, { "epoch": 0.46, "learning_rate": 2.6710708502499355e-05, "loss": 0.7068, "step": 19700 }, { "epoch": 0.46, "learning_rate": 2.670982844805332e-05, "loss": 0.3862, "step": 19705 }, { "epoch": 0.46, "learning_rate": 2.6708948393607285e-05, "loss": 0.1416, "step": 19710 }, { "epoch": 0.46, "learning_rate": 2.670806833916125e-05, "loss": 0.2103, "step": 19715 }, { "epoch": 0.46, "learning_rate": 2.6707188284715215e-05, "loss": 0.2716, "step": 19720 }, { "epoch": 0.46, "learning_rate": 2.6706308230269177e-05, "loss": 0.1786, "step": 19725 }, { "epoch": 0.46, "learning_rate": 2.6705428175823145e-05, "loss": 0.241, "step": 19730 }, { "epoch": 0.46, "learning_rate": 2.670454812137711e-05, "loss": 0.4164, "step": 19735 }, { "epoch": 0.46, "learning_rate": 2.6703668066931075e-05, "loss": 0.5177, "step": 19740 }, { "epoch": 0.46, "learning_rate": 2.670278801248504e-05, "loss": 0.354, "step": 19745 }, { "epoch": 0.46, "learning_rate": 2.6701907958039005e-05, "loss": 0.6165, "step": 19750 }, { "epoch": 0.46, "learning_rate": 2.670102790359297e-05, "loss": 0.3211, "step": 19755 }, { "epoch": 0.46, "learning_rate": 2.6700147849146932e-05, "loss": 0.0952, "step": 19760 }, { "epoch": 0.46, "learning_rate": 2.66992677947009e-05, "loss": 0.2082, "step": 19765 }, { "epoch": 0.46, "learning_rate": 2.6698387740254865e-05, "loss": 0.1211, "step": 19770 }, { "epoch": 0.46, "learning_rate": 2.669750768580883e-05, "loss": 0.205, "step": 19775 }, { "epoch": 0.46, "learning_rate": 2.6696627631362792e-05, "loss": 0.2824, "step": 19780 }, { "epoch": 0.46, "learning_rate": 2.669574757691676e-05, "loss": 0.3088, "step": 19785 }, { "epoch": 0.46, "learning_rate": 2.6694867522470726e-05, "loss": 0.292, "step": 19790 }, { "epoch": 0.46, "learning_rate": 2.6693987468024687e-05, "loss": 0.4102, "step": 19795 }, { "epoch": 0.46, "learning_rate": 2.6693107413578652e-05, "loss": 0.7074, "step": 19800 }, { "epoch": 0.46, "learning_rate": 2.669222735913262e-05, "loss": 0.2827, "step": 19805 }, { "epoch": 0.46, "learning_rate": 2.6691347304686586e-05, "loss": 0.1466, "step": 19810 }, { "epoch": 0.46, "learning_rate": 2.6690467250240547e-05, "loss": 0.1659, "step": 19815 }, { "epoch": 0.46, "learning_rate": 2.6689587195794512e-05, "loss": 0.0761, "step": 19820 }, { "epoch": 0.46, "learning_rate": 2.668870714134848e-05, "loss": 0.1837, "step": 19825 }, { "epoch": 0.46, "learning_rate": 2.6687827086902442e-05, "loss": 0.2221, "step": 19830 }, { "epoch": 0.46, "learning_rate": 2.6686947032456407e-05, "loss": 0.338, "step": 19835 }, { "epoch": 0.46, "learning_rate": 2.6686066978010372e-05, "loss": 0.3471, "step": 19840 }, { "epoch": 0.46, "learning_rate": 2.668518692356434e-05, "loss": 0.3545, "step": 19845 }, { "epoch": 0.46, "learning_rate": 2.6684306869118302e-05, "loss": 0.5924, "step": 19850 }, { "epoch": 0.46, "learning_rate": 2.6683426814672267e-05, "loss": 0.2682, "step": 19855 }, { "epoch": 0.46, "learning_rate": 2.6682546760226232e-05, "loss": 0.0943, "step": 19860 }, { "epoch": 0.46, "learning_rate": 2.6681666705780197e-05, "loss": 0.0874, "step": 19865 }, { "epoch": 0.46, "learning_rate": 2.6680786651334163e-05, "loss": 0.2286, "step": 19870 }, { "epoch": 0.46, "learning_rate": 2.6679906596888128e-05, "loss": 0.2754, "step": 19875 }, { "epoch": 0.46, "learning_rate": 2.6679026542442096e-05, "loss": 0.2601, "step": 19880 }, { "epoch": 0.46, "learning_rate": 2.6678146487996058e-05, "loss": 0.3461, "step": 19885 }, { "epoch": 0.46, "learning_rate": 2.6677266433550023e-05, "loss": 0.234, "step": 19890 }, { "epoch": 0.46, "learning_rate": 2.6676386379103988e-05, "loss": 0.3915, "step": 19895 }, { "epoch": 0.46, "learning_rate": 2.6675506324657956e-05, "loss": 0.7868, "step": 19900 }, { "epoch": 0.46, "learning_rate": 2.6674626270211918e-05, "loss": 0.3212, "step": 19905 }, { "epoch": 0.46, "learning_rate": 2.6673746215765883e-05, "loss": 0.0944, "step": 19910 }, { "epoch": 0.46, "learning_rate": 2.6672866161319848e-05, "loss": 0.1849, "step": 19915 }, { "epoch": 0.46, "learning_rate": 2.6671986106873813e-05, "loss": 0.2355, "step": 19920 }, { "epoch": 0.46, "learning_rate": 2.6671106052427778e-05, "loss": 0.2451, "step": 19925 }, { "epoch": 0.46, "learning_rate": 2.6670225997981743e-05, "loss": 0.2539, "step": 19930 }, { "epoch": 0.47, "learning_rate": 2.6669345943535708e-05, "loss": 0.1368, "step": 19935 }, { "epoch": 0.47, "learning_rate": 2.6668465889089673e-05, "loss": 0.2406, "step": 19940 }, { "epoch": 0.47, "learning_rate": 2.6667585834643638e-05, "loss": 0.4476, "step": 19945 }, { "epoch": 0.47, "learning_rate": 2.6666705780197603e-05, "loss": 0.6135, "step": 19950 }, { "epoch": 0.47, "learning_rate": 2.6665825725751564e-05, "loss": 0.3348, "step": 19955 }, { "epoch": 0.47, "learning_rate": 2.6664945671305533e-05, "loss": 0.077, "step": 19960 }, { "epoch": 0.47, "learning_rate": 2.6664065616859498e-05, "loss": 0.1357, "step": 19965 }, { "epoch": 0.47, "learning_rate": 2.6663185562413463e-05, "loss": 0.1154, "step": 19970 }, { "epoch": 0.47, "learning_rate": 2.6662305507967425e-05, "loss": 0.1779, "step": 19975 }, { "epoch": 0.47, "learning_rate": 2.6661425453521393e-05, "loss": 0.2171, "step": 19980 }, { "epoch": 0.47, "learning_rate": 2.6660545399075358e-05, "loss": 0.1438, "step": 19985 }, { "epoch": 0.47, "learning_rate": 2.665966534462932e-05, "loss": 0.2862, "step": 19990 }, { "epoch": 0.47, "learning_rate": 2.6658785290183285e-05, "loss": 0.4149, "step": 19995 }, { "epoch": 0.47, "learning_rate": 2.6657905235737253e-05, "loss": 0.4176, "step": 20000 }, { "epoch": 0.47, "learning_rate": 2.6657025181291218e-05, "loss": 0.2769, "step": 20005 }, { "epoch": 0.47, "learning_rate": 2.665614512684518e-05, "loss": 0.0676, "step": 20010 }, { "epoch": 0.47, "learning_rate": 2.6655265072399148e-05, "loss": 0.2302, "step": 20015 }, { "epoch": 0.47, "learning_rate": 2.6654385017953113e-05, "loss": 0.0791, "step": 20020 }, { "epoch": 0.47, "learning_rate": 2.6653504963507075e-05, "loss": 0.2956, "step": 20025 }, { "epoch": 0.47, "learning_rate": 2.665262490906104e-05, "loss": 0.1893, "step": 20030 }, { "epoch": 0.47, "learning_rate": 2.6651744854615008e-05, "loss": 0.3063, "step": 20035 }, { "epoch": 0.47, "learning_rate": 2.6650864800168973e-05, "loss": 0.2506, "step": 20040 }, { "epoch": 0.47, "learning_rate": 2.6649984745722935e-05, "loss": 0.2606, "step": 20045 }, { "epoch": 0.47, "learning_rate": 2.66491046912769e-05, "loss": 0.434, "step": 20050 }, { "epoch": 0.47, "learning_rate": 2.6648224636830868e-05, "loss": 0.3071, "step": 20055 }, { "epoch": 0.47, "learning_rate": 2.664734458238483e-05, "loss": 0.0266, "step": 20060 }, { "epoch": 0.47, "learning_rate": 2.6646464527938795e-05, "loss": 0.2579, "step": 20065 }, { "epoch": 0.47, "learning_rate": 2.664558447349276e-05, "loss": 0.2636, "step": 20070 }, { "epoch": 0.47, "learning_rate": 2.664470441904673e-05, "loss": 0.2246, "step": 20075 }, { "epoch": 0.47, "learning_rate": 2.664382436460069e-05, "loss": 0.1937, "step": 20080 }, { "epoch": 0.47, "learning_rate": 2.6642944310154655e-05, "loss": 0.2926, "step": 20085 }, { "epoch": 0.47, "learning_rate": 2.664206425570862e-05, "loss": 0.1883, "step": 20090 }, { "epoch": 0.47, "learning_rate": 2.6641184201262585e-05, "loss": 0.2764, "step": 20095 }, { "epoch": 0.47, "learning_rate": 2.664030414681655e-05, "loss": 0.6541, "step": 20100 }, { "epoch": 0.47, "learning_rate": 2.6639424092370515e-05, "loss": 0.2293, "step": 20105 }, { "epoch": 0.47, "learning_rate": 2.663854403792448e-05, "loss": 0.1393, "step": 20110 }, { "epoch": 0.47, "learning_rate": 2.6637663983478445e-05, "loss": 0.1635, "step": 20115 }, { "epoch": 0.47, "learning_rate": 2.663678392903241e-05, "loss": 0.2666, "step": 20120 }, { "epoch": 0.47, "learning_rate": 2.6635903874586375e-05, "loss": 0.2453, "step": 20125 }, { "epoch": 0.47, "learning_rate": 2.6635023820140337e-05, "loss": 0.1687, "step": 20130 }, { "epoch": 0.47, "learning_rate": 2.6634143765694305e-05, "loss": 0.1798, "step": 20135 }, { "epoch": 0.47, "learning_rate": 2.663326371124827e-05, "loss": 0.278, "step": 20140 }, { "epoch": 0.47, "learning_rate": 2.6632383656802235e-05, "loss": 0.4455, "step": 20145 }, { "epoch": 0.47, "learning_rate": 2.66315036023562e-05, "loss": 0.2841, "step": 20150 }, { "epoch": 0.47, "learning_rate": 2.6630623547910165e-05, "loss": 0.3363, "step": 20155 }, { "epoch": 0.47, "learning_rate": 2.662974349346413e-05, "loss": 0.1267, "step": 20160 }, { "epoch": 0.47, "learning_rate": 2.6628863439018092e-05, "loss": 0.1467, "step": 20165 }, { "epoch": 0.47, "learning_rate": 2.662798338457206e-05, "loss": 0.1177, "step": 20170 }, { "epoch": 0.47, "learning_rate": 2.6627103330126025e-05, "loss": 0.3679, "step": 20175 }, { "epoch": 0.47, "learning_rate": 2.662622327567999e-05, "loss": 0.2486, "step": 20180 }, { "epoch": 0.47, "learning_rate": 2.6625343221233952e-05, "loss": 0.3331, "step": 20185 }, { "epoch": 0.47, "learning_rate": 2.662446316678792e-05, "loss": 0.3203, "step": 20190 }, { "epoch": 0.47, "learning_rate": 2.6623583112341885e-05, "loss": 0.3457, "step": 20195 }, { "epoch": 0.47, "learning_rate": 2.662270305789585e-05, "loss": 0.8872, "step": 20200 }, { "epoch": 0.47, "learning_rate": 2.6621823003449812e-05, "loss": 0.2734, "step": 20205 }, { "epoch": 0.47, "learning_rate": 2.662094294900378e-05, "loss": 0.0752, "step": 20210 }, { "epoch": 0.47, "learning_rate": 2.6620062894557746e-05, "loss": 0.1699, "step": 20215 }, { "epoch": 0.47, "learning_rate": 2.6619182840111707e-05, "loss": 0.1408, "step": 20220 }, { "epoch": 0.47, "learning_rate": 2.6618302785665672e-05, "loss": 0.2624, "step": 20225 }, { "epoch": 0.47, "learning_rate": 2.661742273121964e-05, "loss": 0.1797, "step": 20230 }, { "epoch": 0.47, "learning_rate": 2.6616542676773606e-05, "loss": 0.4077, "step": 20235 }, { "epoch": 0.47, "learning_rate": 2.6615662622327567e-05, "loss": 0.3395, "step": 20240 }, { "epoch": 0.47, "learning_rate": 2.6614782567881532e-05, "loss": 0.4528, "step": 20245 }, { "epoch": 0.47, "learning_rate": 2.66139025134355e-05, "loss": 0.5676, "step": 20250 }, { "epoch": 0.47, "learning_rate": 2.6613022458989462e-05, "loss": 0.2758, "step": 20255 }, { "epoch": 0.47, "learning_rate": 2.6612142404543427e-05, "loss": 0.1107, "step": 20260 }, { "epoch": 0.47, "learning_rate": 2.6611262350097392e-05, "loss": 0.1344, "step": 20265 }, { "epoch": 0.47, "learning_rate": 2.661038229565136e-05, "loss": 0.1665, "step": 20270 }, { "epoch": 0.47, "learning_rate": 2.6609502241205322e-05, "loss": 0.2834, "step": 20275 }, { "epoch": 0.47, "learning_rate": 2.6608622186759287e-05, "loss": 0.2257, "step": 20280 }, { "epoch": 0.47, "learning_rate": 2.6607742132313256e-05, "loss": 0.3565, "step": 20285 }, { "epoch": 0.47, "learning_rate": 2.6606862077867217e-05, "loss": 0.2708, "step": 20290 }, { "epoch": 0.47, "learning_rate": 2.6605982023421182e-05, "loss": 0.3329, "step": 20295 }, { "epoch": 0.47, "learning_rate": 2.6605101968975147e-05, "loss": 1.1224, "step": 20300 }, { "epoch": 0.47, "learning_rate": 2.6604221914529116e-05, "loss": 0.3233, "step": 20305 }, { "epoch": 0.47, "learning_rate": 2.6603341860083078e-05, "loss": 0.0885, "step": 20310 }, { "epoch": 0.47, "learning_rate": 2.6602461805637043e-05, "loss": 0.1642, "step": 20315 }, { "epoch": 0.47, "learning_rate": 2.6601581751191008e-05, "loss": 0.201, "step": 20320 }, { "epoch": 0.47, "learning_rate": 2.6600701696744973e-05, "loss": 0.3006, "step": 20325 }, { "epoch": 0.47, "learning_rate": 2.6599821642298938e-05, "loss": 0.2008, "step": 20330 }, { "epoch": 0.47, "learning_rate": 2.6598941587852903e-05, "loss": 0.2118, "step": 20335 }, { "epoch": 0.47, "learning_rate": 2.6598061533406868e-05, "loss": 0.3458, "step": 20340 }, { "epoch": 0.47, "learning_rate": 2.6597181478960833e-05, "loss": 0.426, "step": 20345 }, { "epoch": 0.47, "learning_rate": 2.6596301424514798e-05, "loss": 0.3988, "step": 20350 }, { "epoch": 0.47, "learning_rate": 2.6595421370068763e-05, "loss": 0.1842, "step": 20355 }, { "epoch": 0.48, "learning_rate": 2.6594541315622724e-05, "loss": 0.0979, "step": 20360 }, { "epoch": 0.48, "learning_rate": 2.6593661261176693e-05, "loss": 0.144, "step": 20365 }, { "epoch": 0.48, "learning_rate": 2.6592781206730658e-05, "loss": 0.2321, "step": 20370 }, { "epoch": 0.48, "learning_rate": 2.6591901152284623e-05, "loss": 0.3078, "step": 20375 }, { "epoch": 0.48, "learning_rate": 2.6591021097838584e-05, "loss": 0.1885, "step": 20380 }, { "epoch": 0.48, "learning_rate": 2.6590141043392553e-05, "loss": 0.2879, "step": 20385 }, { "epoch": 0.48, "learning_rate": 2.6589260988946518e-05, "loss": 0.2738, "step": 20390 }, { "epoch": 0.48, "learning_rate": 2.658838093450048e-05, "loss": 0.3832, "step": 20395 }, { "epoch": 0.48, "learning_rate": 2.6587500880054445e-05, "loss": 0.7397, "step": 20400 }, { "epoch": 0.48, "learning_rate": 2.6586620825608413e-05, "loss": 0.2912, "step": 20405 }, { "epoch": 0.48, "learning_rate": 2.6585740771162378e-05, "loss": 0.1254, "step": 20410 }, { "epoch": 0.48, "learning_rate": 2.658486071671634e-05, "loss": 0.2372, "step": 20415 }, { "epoch": 0.48, "learning_rate": 2.6583980662270308e-05, "loss": 0.2033, "step": 20420 }, { "epoch": 0.48, "learning_rate": 2.6583100607824273e-05, "loss": 0.1216, "step": 20425 }, { "epoch": 0.48, "learning_rate": 2.6582220553378235e-05, "loss": 0.2211, "step": 20430 }, { "epoch": 0.48, "learning_rate": 2.65813404989322e-05, "loss": 0.2325, "step": 20435 }, { "epoch": 0.48, "learning_rate": 2.6580460444486168e-05, "loss": 0.4911, "step": 20440 }, { "epoch": 0.48, "learning_rate": 2.6579580390040133e-05, "loss": 0.4687, "step": 20445 }, { "epoch": 0.48, "learning_rate": 2.6578700335594095e-05, "loss": 0.622, "step": 20450 }, { "epoch": 0.48, "learning_rate": 2.657782028114806e-05, "loss": 0.2513, "step": 20455 }, { "epoch": 0.48, "learning_rate": 2.6576940226702028e-05, "loss": 0.0757, "step": 20460 }, { "epoch": 0.48, "learning_rate": 2.657606017225599e-05, "loss": 0.2029, "step": 20465 }, { "epoch": 0.48, "learning_rate": 2.6575180117809955e-05, "loss": 0.2015, "step": 20470 }, { "epoch": 0.48, "learning_rate": 2.657430006336392e-05, "loss": 0.2616, "step": 20475 }, { "epoch": 0.48, "learning_rate": 2.6573420008917888e-05, "loss": 0.237, "step": 20480 }, { "epoch": 0.48, "learning_rate": 2.657253995447185e-05, "loss": 0.1315, "step": 20485 }, { "epoch": 0.48, "learning_rate": 2.6571659900025815e-05, "loss": 0.3532, "step": 20490 }, { "epoch": 0.48, "learning_rate": 2.657077984557978e-05, "loss": 0.4593, "step": 20495 }, { "epoch": 0.48, "learning_rate": 2.656989979113375e-05, "loss": 0.4284, "step": 20500 }, { "epoch": 0.48, "learning_rate": 2.656901973668771e-05, "loss": 0.2188, "step": 20505 }, { "epoch": 0.48, "learning_rate": 2.6568139682241675e-05, "loss": 0.055, "step": 20510 }, { "epoch": 0.48, "learning_rate": 2.656725962779564e-05, "loss": 0.1078, "step": 20515 }, { "epoch": 0.48, "learning_rate": 2.6566379573349605e-05, "loss": 0.2924, "step": 20520 }, { "epoch": 0.48, "learning_rate": 2.656549951890357e-05, "loss": 0.2049, "step": 20525 }, { "epoch": 0.48, "learning_rate": 2.6564619464457535e-05, "loss": 0.1884, "step": 20530 }, { "epoch": 0.48, "learning_rate": 2.65637394100115e-05, "loss": 0.2339, "step": 20535 }, { "epoch": 0.48, "learning_rate": 2.6562859355565465e-05, "loss": 0.2869, "step": 20540 }, { "epoch": 0.48, "learning_rate": 2.656197930111943e-05, "loss": 0.2908, "step": 20545 }, { "epoch": 0.48, "learning_rate": 2.6561099246673395e-05, "loss": 0.7591, "step": 20550 }, { "epoch": 0.48, "learning_rate": 2.656021919222736e-05, "loss": 0.3381, "step": 20555 }, { "epoch": 0.48, "learning_rate": 2.6559339137781325e-05, "loss": 0.0845, "step": 20560 }, { "epoch": 0.48, "learning_rate": 2.655845908333529e-05, "loss": 0.2069, "step": 20565 }, { "epoch": 0.48, "learning_rate": 2.6557579028889255e-05, "loss": 0.1479, "step": 20570 }, { "epoch": 0.48, "learning_rate": 2.655669897444322e-05, "loss": 0.1953, "step": 20575 }, { "epoch": 0.48, "learning_rate": 2.6555818919997185e-05, "loss": 0.1565, "step": 20580 }, { "epoch": 0.48, "learning_rate": 2.655493886555115e-05, "loss": 0.2803, "step": 20585 }, { "epoch": 0.48, "learning_rate": 2.6554058811105112e-05, "loss": 0.1866, "step": 20590 }, { "epoch": 0.48, "learning_rate": 2.655317875665908e-05, "loss": 0.5521, "step": 20595 }, { "epoch": 0.48, "learning_rate": 2.6552298702213045e-05, "loss": 0.6031, "step": 20600 }, { "epoch": 0.48, "learning_rate": 2.655141864776701e-05, "loss": 0.2037, "step": 20605 }, { "epoch": 0.48, "learning_rate": 2.6550538593320972e-05, "loss": 0.1941, "step": 20610 }, { "epoch": 0.48, "learning_rate": 2.654965853887494e-05, "loss": 0.1209, "step": 20615 }, { "epoch": 0.48, "learning_rate": 2.6548778484428905e-05, "loss": 0.1553, "step": 20620 }, { "epoch": 0.48, "learning_rate": 2.6547898429982867e-05, "loss": 0.1929, "step": 20625 }, { "epoch": 0.48, "learning_rate": 2.6547018375536832e-05, "loss": 0.3662, "step": 20630 }, { "epoch": 0.48, "learning_rate": 2.65461383210908e-05, "loss": 0.2908, "step": 20635 }, { "epoch": 0.48, "learning_rate": 2.6545258266644766e-05, "loss": 0.4369, "step": 20640 }, { "epoch": 0.48, "learning_rate": 2.6544378212198727e-05, "loss": 0.3771, "step": 20645 }, { "epoch": 0.48, "learning_rate": 2.6543498157752692e-05, "loss": 0.8361, "step": 20650 }, { "epoch": 0.48, "learning_rate": 2.654261810330666e-05, "loss": 0.2917, "step": 20655 }, { "epoch": 0.48, "learning_rate": 2.6541738048860622e-05, "loss": 0.0592, "step": 20660 }, { "epoch": 0.48, "learning_rate": 2.6540857994414587e-05, "loss": 0.0662, "step": 20665 }, { "epoch": 0.48, "learning_rate": 2.6539977939968552e-05, "loss": 0.1667, "step": 20670 }, { "epoch": 0.48, "learning_rate": 2.653909788552252e-05, "loss": 0.2761, "step": 20675 }, { "epoch": 0.48, "learning_rate": 2.6538217831076482e-05, "loss": 0.2951, "step": 20680 }, { "epoch": 0.48, "learning_rate": 2.6537337776630447e-05, "loss": 0.2009, "step": 20685 }, { "epoch": 0.48, "learning_rate": 2.6536457722184416e-05, "loss": 0.2988, "step": 20690 }, { "epoch": 0.48, "learning_rate": 2.6535577667738377e-05, "loss": 0.4795, "step": 20695 }, { "epoch": 0.48, "learning_rate": 2.6534697613292342e-05, "loss": 0.5673, "step": 20700 }, { "epoch": 0.48, "learning_rate": 2.6533817558846307e-05, "loss": 0.2995, "step": 20705 }, { "epoch": 0.48, "learning_rate": 2.6532937504400276e-05, "loss": 0.0562, "step": 20710 }, { "epoch": 0.48, "learning_rate": 2.6532057449954237e-05, "loss": 0.1747, "step": 20715 }, { "epoch": 0.48, "learning_rate": 2.6531177395508202e-05, "loss": 0.1329, "step": 20720 }, { "epoch": 0.48, "learning_rate": 2.6530297341062167e-05, "loss": 0.2066, "step": 20725 }, { "epoch": 0.48, "learning_rate": 2.6529417286616132e-05, "loss": 0.2983, "step": 20730 }, { "epoch": 0.48, "learning_rate": 2.6528537232170098e-05, "loss": 0.347, "step": 20735 }, { "epoch": 0.48, "learning_rate": 2.6527657177724063e-05, "loss": 0.2066, "step": 20740 }, { "epoch": 0.48, "learning_rate": 2.6526777123278028e-05, "loss": 0.3012, "step": 20745 }, { "epoch": 0.48, "learning_rate": 2.6525897068831993e-05, "loss": 0.8186, "step": 20750 }, { "epoch": 0.48, "learning_rate": 2.6525017014385958e-05, "loss": 0.3731, "step": 20755 }, { "epoch": 0.48, "learning_rate": 2.6524136959939923e-05, "loss": 0.1553, "step": 20760 }, { "epoch": 0.48, "learning_rate": 2.6523256905493884e-05, "loss": 0.1051, "step": 20765 }, { "epoch": 0.48, "learning_rate": 2.6522376851047853e-05, "loss": 0.1852, "step": 20770 }, { "epoch": 0.48, "learning_rate": 2.6521496796601818e-05, "loss": 0.1529, "step": 20775 }, { "epoch": 0.48, "learning_rate": 2.6520616742155783e-05, "loss": 0.1348, "step": 20780 }, { "epoch": 0.48, "learning_rate": 2.6519736687709744e-05, "loss": 0.2832, "step": 20785 }, { "epoch": 0.49, "learning_rate": 2.6518856633263713e-05, "loss": 0.2136, "step": 20790 }, { "epoch": 0.49, "learning_rate": 2.6517976578817678e-05, "loss": 0.4489, "step": 20795 }, { "epoch": 0.49, "learning_rate": 2.6517096524371643e-05, "loss": 0.5807, "step": 20800 }, { "epoch": 0.49, "learning_rate": 2.6516216469925604e-05, "loss": 0.2121, "step": 20805 }, { "epoch": 0.49, "learning_rate": 2.6515336415479573e-05, "loss": 0.0749, "step": 20810 }, { "epoch": 0.49, "learning_rate": 2.6514456361033538e-05, "loss": 0.0668, "step": 20815 }, { "epoch": 0.49, "learning_rate": 2.65135763065875e-05, "loss": 0.1227, "step": 20820 }, { "epoch": 0.49, "learning_rate": 2.6512696252141468e-05, "loss": 0.1558, "step": 20825 }, { "epoch": 0.49, "learning_rate": 2.6511816197695433e-05, "loss": 0.2312, "step": 20830 }, { "epoch": 0.49, "learning_rate": 2.6510936143249398e-05, "loss": 0.2238, "step": 20835 }, { "epoch": 0.49, "learning_rate": 2.651005608880336e-05, "loss": 0.3381, "step": 20840 }, { "epoch": 0.49, "learning_rate": 2.6509176034357328e-05, "loss": 0.1607, "step": 20845 }, { "epoch": 0.49, "learning_rate": 2.6508295979911293e-05, "loss": 0.3377, "step": 20850 }, { "epoch": 0.49, "learning_rate": 2.6507415925465255e-05, "loss": 0.1539, "step": 20855 }, { "epoch": 0.49, "learning_rate": 2.650653587101922e-05, "loss": 0.1657, "step": 20860 }, { "epoch": 0.49, "learning_rate": 2.6505655816573188e-05, "loss": 0.2728, "step": 20865 }, { "epoch": 0.49, "learning_rate": 2.6504775762127153e-05, "loss": 0.074, "step": 20870 }, { "epoch": 0.49, "learning_rate": 2.6503895707681115e-05, "loss": 0.1053, "step": 20875 }, { "epoch": 0.49, "learning_rate": 2.650301565323508e-05, "loss": 0.2359, "step": 20880 }, { "epoch": 0.49, "learning_rate": 2.6502135598789048e-05, "loss": 0.2512, "step": 20885 }, { "epoch": 0.49, "learning_rate": 2.650125554434301e-05, "loss": 0.451, "step": 20890 }, { "epoch": 0.49, "learning_rate": 2.6500375489896975e-05, "loss": 0.3662, "step": 20895 }, { "epoch": 0.49, "learning_rate": 2.649949543545094e-05, "loss": 0.5671, "step": 20900 }, { "epoch": 0.49, "learning_rate": 2.6498615381004908e-05, "loss": 0.375, "step": 20905 }, { "epoch": 0.49, "learning_rate": 2.649773532655887e-05, "loss": 0.1518, "step": 20910 }, { "epoch": 0.49, "learning_rate": 2.6496855272112835e-05, "loss": 0.0758, "step": 20915 }, { "epoch": 0.49, "learning_rate": 2.64959752176668e-05, "loss": 0.09, "step": 20920 }, { "epoch": 0.49, "learning_rate": 2.6495095163220765e-05, "loss": 0.1975, "step": 20925 }, { "epoch": 0.49, "learning_rate": 2.649421510877473e-05, "loss": 0.2068, "step": 20930 }, { "epoch": 0.49, "learning_rate": 2.6493335054328695e-05, "loss": 0.1537, "step": 20935 }, { "epoch": 0.49, "learning_rate": 2.649245499988266e-05, "loss": 0.1453, "step": 20940 }, { "epoch": 0.49, "learning_rate": 2.6491574945436625e-05, "loss": 0.4272, "step": 20945 }, { "epoch": 0.49, "learning_rate": 2.649069489099059e-05, "loss": 0.5208, "step": 20950 }, { "epoch": 0.49, "learning_rate": 2.6489814836544555e-05, "loss": 0.3189, "step": 20955 }, { "epoch": 0.49, "learning_rate": 2.648893478209852e-05, "loss": 0.0669, "step": 20960 }, { "epoch": 0.49, "learning_rate": 2.6488054727652485e-05, "loss": 0.0768, "step": 20965 }, { "epoch": 0.49, "learning_rate": 2.648717467320645e-05, "loss": 0.1094, "step": 20970 }, { "epoch": 0.49, "learning_rate": 2.6486294618760415e-05, "loss": 0.2413, "step": 20975 }, { "epoch": 0.49, "learning_rate": 2.648541456431438e-05, "loss": 0.202, "step": 20980 }, { "epoch": 0.49, "learning_rate": 2.6484534509868345e-05, "loss": 0.3105, "step": 20985 }, { "epoch": 0.49, "learning_rate": 2.648365445542231e-05, "loss": 0.2852, "step": 20990 }, { "epoch": 0.49, "learning_rate": 2.6482774400976272e-05, "loss": 0.3022, "step": 20995 }, { "epoch": 0.49, "learning_rate": 2.648189434653024e-05, "loss": 0.7002, "step": 21000 }, { "epoch": 0.49, "learning_rate": 2.6481014292084205e-05, "loss": 0.3558, "step": 21005 }, { "epoch": 0.49, "learning_rate": 2.648013423763817e-05, "loss": 0.1341, "step": 21010 }, { "epoch": 0.49, "learning_rate": 2.6479254183192132e-05, "loss": 0.171, "step": 21015 }, { "epoch": 0.49, "learning_rate": 2.64783741287461e-05, "loss": 0.1816, "step": 21020 }, { "epoch": 0.49, "learning_rate": 2.6477494074300065e-05, "loss": 0.1314, "step": 21025 }, { "epoch": 0.49, "learning_rate": 2.6476614019854027e-05, "loss": 0.236, "step": 21030 }, { "epoch": 0.49, "learning_rate": 2.6475733965407992e-05, "loss": 0.2626, "step": 21035 }, { "epoch": 0.49, "learning_rate": 2.647485391096196e-05, "loss": 0.1529, "step": 21040 }, { "epoch": 0.49, "learning_rate": 2.6473973856515925e-05, "loss": 0.2467, "step": 21045 }, { "epoch": 0.49, "learning_rate": 2.6473093802069887e-05, "loss": 0.6172, "step": 21050 }, { "epoch": 0.49, "learning_rate": 2.6472213747623852e-05, "loss": 0.317, "step": 21055 }, { "epoch": 0.49, "learning_rate": 2.647133369317782e-05, "loss": 0.079, "step": 21060 }, { "epoch": 0.49, "learning_rate": 2.6470453638731782e-05, "loss": 0.0734, "step": 21065 }, { "epoch": 0.49, "learning_rate": 2.6469573584285747e-05, "loss": 0.2465, "step": 21070 }, { "epoch": 0.49, "learning_rate": 2.6468693529839712e-05, "loss": 0.1513, "step": 21075 }, { "epoch": 0.49, "learning_rate": 2.646781347539368e-05, "loss": 0.2078, "step": 21080 }, { "epoch": 0.49, "learning_rate": 2.6466933420947642e-05, "loss": 0.2225, "step": 21085 }, { "epoch": 0.49, "learning_rate": 2.6466053366501607e-05, "loss": 0.3203, "step": 21090 }, { "epoch": 0.49, "learning_rate": 2.6465173312055572e-05, "loss": 0.4989, "step": 21095 }, { "epoch": 0.49, "learning_rate": 2.646429325760954e-05, "loss": 0.6847, "step": 21100 }, { "epoch": 0.49, "learning_rate": 2.6463413203163502e-05, "loss": 0.2036, "step": 21105 }, { "epoch": 0.49, "learning_rate": 2.6462533148717467e-05, "loss": 0.151, "step": 21110 }, { "epoch": 0.49, "learning_rate": 2.6461653094271436e-05, "loss": 0.2215, "step": 21115 }, { "epoch": 0.49, "learning_rate": 2.6460773039825397e-05, "loss": 0.1787, "step": 21120 }, { "epoch": 0.49, "learning_rate": 2.6459892985379362e-05, "loss": 0.1534, "step": 21125 }, { "epoch": 0.49, "learning_rate": 2.6459012930933327e-05, "loss": 0.1729, "step": 21130 }, { "epoch": 0.49, "learning_rate": 2.6458132876487296e-05, "loss": 0.2706, "step": 21135 }, { "epoch": 0.49, "learning_rate": 2.6457252822041257e-05, "loss": 0.4266, "step": 21140 }, { "epoch": 0.49, "learning_rate": 2.6456372767595222e-05, "loss": 0.3462, "step": 21145 }, { "epoch": 0.49, "learning_rate": 2.6455492713149187e-05, "loss": 0.9728, "step": 21150 }, { "epoch": 0.49, "learning_rate": 2.6454612658703152e-05, "loss": 0.2589, "step": 21155 }, { "epoch": 0.49, "learning_rate": 2.6453732604257117e-05, "loss": 0.1293, "step": 21160 }, { "epoch": 0.49, "learning_rate": 2.6452852549811082e-05, "loss": 0.1168, "step": 21165 }, { "epoch": 0.49, "learning_rate": 2.6451972495365048e-05, "loss": 0.1176, "step": 21170 }, { "epoch": 0.49, "learning_rate": 2.6451092440919013e-05, "loss": 0.2488, "step": 21175 }, { "epoch": 0.49, "learning_rate": 2.6450212386472978e-05, "loss": 0.2292, "step": 21180 }, { "epoch": 0.49, "learning_rate": 2.6449332332026943e-05, "loss": 0.2906, "step": 21185 }, { "epoch": 0.49, "learning_rate": 2.6448452277580904e-05, "loss": 0.224, "step": 21190 }, { "epoch": 0.49, "learning_rate": 2.6447572223134873e-05, "loss": 0.4922, "step": 21195 }, { "epoch": 0.49, "learning_rate": 2.6446692168688838e-05, "loss": 0.5379, "step": 21200 }, { "epoch": 0.49, "learning_rate": 2.6445812114242803e-05, "loss": 0.2491, "step": 21205 }, { "epoch": 0.49, "learning_rate": 2.6444932059796764e-05, "loss": 0.0909, "step": 21210 }, { "epoch": 0.49, "learning_rate": 2.6444052005350733e-05, "loss": 0.1216, "step": 21215 }, { "epoch": 0.5, "learning_rate": 2.6443171950904698e-05, "loss": 0.0615, "step": 21220 }, { "epoch": 0.5, "learning_rate": 2.644229189645866e-05, "loss": 0.185, "step": 21225 }, { "epoch": 0.5, "learning_rate": 2.6441411842012628e-05, "loss": 0.1435, "step": 21230 }, { "epoch": 0.5, "learning_rate": 2.6440531787566593e-05, "loss": 0.2432, "step": 21235 }, { "epoch": 0.5, "learning_rate": 2.6439651733120558e-05, "loss": 0.3206, "step": 21240 }, { "epoch": 0.5, "learning_rate": 2.643877167867452e-05, "loss": 0.4883, "step": 21245 }, { "epoch": 0.5, "learning_rate": 2.6437891624228488e-05, "loss": 0.3839, "step": 21250 }, { "epoch": 0.5, "learning_rate": 2.6437011569782453e-05, "loss": 0.2946, "step": 21255 }, { "epoch": 0.5, "learning_rate": 2.6436131515336414e-05, "loss": 0.1666, "step": 21260 }, { "epoch": 0.5, "learning_rate": 2.643525146089038e-05, "loss": 0.1904, "step": 21265 }, { "epoch": 0.5, "learning_rate": 2.6434371406444348e-05, "loss": 0.1107, "step": 21270 }, { "epoch": 0.5, "learning_rate": 2.6433491351998313e-05, "loss": 0.1925, "step": 21275 }, { "epoch": 0.5, "learning_rate": 2.6432611297552275e-05, "loss": 0.205, "step": 21280 }, { "epoch": 0.5, "learning_rate": 2.643173124310624e-05, "loss": 0.2958, "step": 21285 }, { "epoch": 0.5, "learning_rate": 2.6430851188660208e-05, "loss": 0.3017, "step": 21290 }, { "epoch": 0.5, "learning_rate": 2.642997113421417e-05, "loss": 0.2221, "step": 21295 }, { "epoch": 0.5, "learning_rate": 2.6429091079768135e-05, "loss": 0.6624, "step": 21300 }, { "epoch": 0.5, "learning_rate": 2.64282110253221e-05, "loss": 0.3094, "step": 21305 }, { "epoch": 0.5, "learning_rate": 2.6427330970876068e-05, "loss": 0.1691, "step": 21310 }, { "epoch": 0.5, "learning_rate": 2.642645091643003e-05, "loss": 0.0583, "step": 21315 }, { "epoch": 0.5, "learning_rate": 2.6425570861983995e-05, "loss": 0.108, "step": 21320 }, { "epoch": 0.5, "learning_rate": 2.642469080753796e-05, "loss": 0.1653, "step": 21325 }, { "epoch": 0.5, "learning_rate": 2.6423810753091925e-05, "loss": 0.1217, "step": 21330 }, { "epoch": 0.5, "learning_rate": 2.642293069864589e-05, "loss": 0.2321, "step": 21335 }, { "epoch": 0.5, "learning_rate": 2.6422050644199855e-05, "loss": 0.3367, "step": 21340 }, { "epoch": 0.5, "learning_rate": 2.642117058975382e-05, "loss": 0.2959, "step": 21345 }, { "epoch": 0.5, "learning_rate": 2.6420290535307785e-05, "loss": 0.652, "step": 21350 }, { "epoch": 0.5, "learning_rate": 2.641941048086175e-05, "loss": 0.1601, "step": 21355 }, { "epoch": 0.5, "learning_rate": 2.6418530426415715e-05, "loss": 0.2023, "step": 21360 }, { "epoch": 0.5, "learning_rate": 2.641765037196968e-05, "loss": 0.1173, "step": 21365 }, { "epoch": 0.5, "learning_rate": 2.6416770317523645e-05, "loss": 0.2469, "step": 21370 }, { "epoch": 0.5, "learning_rate": 2.641589026307761e-05, "loss": 0.2446, "step": 21375 }, { "epoch": 0.5, "learning_rate": 2.6415010208631575e-05, "loss": 0.2143, "step": 21380 }, { "epoch": 0.5, "learning_rate": 2.641413015418554e-05, "loss": 0.1459, "step": 21385 }, { "epoch": 0.5, "learning_rate": 2.6413250099739505e-05, "loss": 0.3183, "step": 21390 }, { "epoch": 0.5, "learning_rate": 2.641237004529347e-05, "loss": 0.2837, "step": 21395 }, { "epoch": 0.5, "learning_rate": 2.641148999084743e-05, "loss": 0.4681, "step": 21400 }, { "epoch": 0.5, "learning_rate": 2.64106099364014e-05, "loss": 0.2451, "step": 21405 }, { "epoch": 0.5, "learning_rate": 2.6409729881955365e-05, "loss": 0.1673, "step": 21410 }, { "epoch": 0.5, "learning_rate": 2.640884982750933e-05, "loss": 0.0782, "step": 21415 }, { "epoch": 0.5, "learning_rate": 2.6407969773063292e-05, "loss": 0.2075, "step": 21420 }, { "epoch": 0.5, "learning_rate": 2.640708971861726e-05, "loss": 0.1639, "step": 21425 }, { "epoch": 0.5, "learning_rate": 2.6406209664171225e-05, "loss": 0.2253, "step": 21430 }, { "epoch": 0.5, "learning_rate": 2.640532960972519e-05, "loss": 0.2415, "step": 21435 }, { "epoch": 0.5, "learning_rate": 2.6404449555279152e-05, "loss": 0.202, "step": 21440 }, { "epoch": 0.5, "learning_rate": 2.640356950083312e-05, "loss": 0.4449, "step": 21445 }, { "epoch": 0.5, "learning_rate": 2.6402689446387085e-05, "loss": 0.479, "step": 21450 }, { "epoch": 0.5, "learning_rate": 2.6401809391941047e-05, "loss": 0.3004, "step": 21455 }, { "epoch": 0.5, "learning_rate": 2.6400929337495012e-05, "loss": 0.2069, "step": 21460 }, { "epoch": 0.5, "learning_rate": 2.640004928304898e-05, "loss": 0.0732, "step": 21465 }, { "epoch": 0.5, "learning_rate": 2.6399169228602945e-05, "loss": 0.1429, "step": 21470 }, { "epoch": 0.5, "learning_rate": 2.6398289174156907e-05, "loss": 0.1365, "step": 21475 }, { "epoch": 0.5, "learning_rate": 2.6397409119710872e-05, "loss": 0.1498, "step": 21480 }, { "epoch": 0.5, "learning_rate": 2.639652906526484e-05, "loss": 0.3507, "step": 21485 }, { "epoch": 0.5, "learning_rate": 2.6395649010818802e-05, "loss": 0.1549, "step": 21490 }, { "epoch": 0.5, "learning_rate": 2.6394768956372767e-05, "loss": 0.4316, "step": 21495 }, { "epoch": 0.5, "learning_rate": 2.6393888901926732e-05, "loss": 0.4991, "step": 21500 }, { "epoch": 0.5, "learning_rate": 2.63930088474807e-05, "loss": 0.2015, "step": 21505 }, { "epoch": 0.5, "learning_rate": 2.6392128793034662e-05, "loss": 0.0994, "step": 21510 }, { "epoch": 0.5, "learning_rate": 2.6391248738588627e-05, "loss": 0.1876, "step": 21515 }, { "epoch": 0.5, "learning_rate": 2.6390368684142596e-05, "loss": 0.1227, "step": 21520 }, { "epoch": 0.5, "learning_rate": 2.6389488629696557e-05, "loss": 0.2756, "step": 21525 }, { "epoch": 0.5, "learning_rate": 2.6388608575250522e-05, "loss": 0.1664, "step": 21530 }, { "epoch": 0.5, "learning_rate": 2.6387728520804487e-05, "loss": 0.1678, "step": 21535 }, { "epoch": 0.5, "learning_rate": 2.6386848466358456e-05, "loss": 0.3611, "step": 21540 }, { "epoch": 0.5, "learning_rate": 2.6385968411912417e-05, "loss": 0.6095, "step": 21545 }, { "epoch": 0.5, "learning_rate": 2.6385088357466382e-05, "loss": 0.6301, "step": 21550 }, { "epoch": 0.5, "learning_rate": 2.6384208303020347e-05, "loss": 0.321, "step": 21555 }, { "epoch": 0.5, "learning_rate": 2.6383328248574312e-05, "loss": 0.0548, "step": 21560 }, { "epoch": 0.5, "learning_rate": 2.6382448194128277e-05, "loss": 0.172, "step": 21565 }, { "epoch": 0.5, "learning_rate": 2.6381568139682242e-05, "loss": 0.1797, "step": 21570 }, { "epoch": 0.5, "learning_rate": 2.6380688085236207e-05, "loss": 0.3631, "step": 21575 }, { "epoch": 0.5, "learning_rate": 2.6379808030790172e-05, "loss": 0.2053, "step": 21580 }, { "epoch": 0.5, "learning_rate": 2.6378927976344137e-05, "loss": 0.2797, "step": 21585 }, { "epoch": 0.5, "learning_rate": 2.6378047921898102e-05, "loss": 0.2354, "step": 21590 }, { "epoch": 0.5, "learning_rate": 2.6377167867452064e-05, "loss": 0.4117, "step": 21595 }, { "epoch": 0.5, "learning_rate": 2.6376287813006033e-05, "loss": 0.7828, "step": 21600 }, { "epoch": 0.5, "learning_rate": 2.6375407758559998e-05, "loss": 0.3369, "step": 21605 }, { "epoch": 0.5, "learning_rate": 2.6374527704113963e-05, "loss": 0.1106, "step": 21610 }, { "epoch": 0.5, "learning_rate": 2.6373647649667924e-05, "loss": 0.0857, "step": 21615 }, { "epoch": 0.5, "learning_rate": 2.6372767595221893e-05, "loss": 0.2152, "step": 21620 }, { "epoch": 0.5, "learning_rate": 2.6371887540775858e-05, "loss": 0.1744, "step": 21625 }, { "epoch": 0.5, "learning_rate": 2.637100748632982e-05, "loss": 0.2218, "step": 21630 }, { "epoch": 0.5, "learning_rate": 2.6370127431883784e-05, "loss": 0.3294, "step": 21635 }, { "epoch": 0.5, "learning_rate": 2.6369247377437753e-05, "loss": 0.2023, "step": 21640 }, { "epoch": 0.5, "learning_rate": 2.6368367322991718e-05, "loss": 0.0854, "step": 21645 }, { "epoch": 0.51, "learning_rate": 2.636748726854568e-05, "loss": 0.5257, "step": 21650 }, { "epoch": 0.51, "learning_rate": 2.6366607214099648e-05, "loss": 0.3537, "step": 21655 }, { "epoch": 0.51, "learning_rate": 2.6365727159653613e-05, "loss": 0.1641, "step": 21660 }, { "epoch": 0.51, "learning_rate": 2.6364847105207574e-05, "loss": 0.1186, "step": 21665 }, { "epoch": 0.51, "learning_rate": 2.636396705076154e-05, "loss": 0.1173, "step": 21670 }, { "epoch": 0.51, "learning_rate": 2.6363086996315508e-05, "loss": 0.2238, "step": 21675 }, { "epoch": 0.51, "learning_rate": 2.6362206941869473e-05, "loss": 0.2318, "step": 21680 }, { "epoch": 0.51, "learning_rate": 2.6361326887423434e-05, "loss": 0.2805, "step": 21685 }, { "epoch": 0.51, "learning_rate": 2.63604468329774e-05, "loss": 0.2701, "step": 21690 }, { "epoch": 0.51, "learning_rate": 2.6359566778531368e-05, "loss": 0.314, "step": 21695 }, { "epoch": 0.51, "learning_rate": 2.635868672408533e-05, "loss": 0.4998, "step": 21700 }, { "epoch": 0.51, "learning_rate": 2.6357806669639295e-05, "loss": 0.1138, "step": 21705 }, { "epoch": 0.51, "learning_rate": 2.635692661519326e-05, "loss": 0.422, "step": 21710 }, { "epoch": 0.51, "learning_rate": 2.6356046560747228e-05, "loss": 0.1415, "step": 21715 }, { "epoch": 0.51, "learning_rate": 2.635516650630119e-05, "loss": 0.1791, "step": 21720 }, { "epoch": 0.51, "learning_rate": 2.6354286451855155e-05, "loss": 0.1975, "step": 21725 }, { "epoch": 0.51, "learning_rate": 2.635340639740912e-05, "loss": 0.1394, "step": 21730 }, { "epoch": 0.51, "learning_rate": 2.6352526342963088e-05, "loss": 0.2677, "step": 21735 }, { "epoch": 0.51, "learning_rate": 2.635164628851705e-05, "loss": 0.3087, "step": 21740 }, { "epoch": 0.51, "learning_rate": 2.6350766234071015e-05, "loss": 0.3425, "step": 21745 }, { "epoch": 0.51, "learning_rate": 2.634988617962498e-05, "loss": 0.4949, "step": 21750 }, { "epoch": 0.51, "learning_rate": 2.6349006125178945e-05, "loss": 0.2799, "step": 21755 }, { "epoch": 0.51, "learning_rate": 2.634812607073291e-05, "loss": 0.0747, "step": 21760 }, { "epoch": 0.51, "learning_rate": 2.6347246016286875e-05, "loss": 0.119, "step": 21765 }, { "epoch": 0.51, "learning_rate": 2.634636596184084e-05, "loss": 0.1806, "step": 21770 }, { "epoch": 0.51, "learning_rate": 2.6345485907394805e-05, "loss": 0.1481, "step": 21775 }, { "epoch": 0.51, "learning_rate": 2.634460585294877e-05, "loss": 0.2057, "step": 21780 }, { "epoch": 0.51, "learning_rate": 2.6343725798502735e-05, "loss": 0.1996, "step": 21785 }, { "epoch": 0.51, "learning_rate": 2.63428457440567e-05, "loss": 0.25, "step": 21790 }, { "epoch": 0.51, "learning_rate": 2.6341965689610665e-05, "loss": 0.4593, "step": 21795 }, { "epoch": 0.51, "learning_rate": 2.634108563516463e-05, "loss": 0.6915, "step": 21800 }, { "epoch": 0.51, "learning_rate": 2.6340205580718595e-05, "loss": 0.1832, "step": 21805 }, { "epoch": 0.51, "learning_rate": 2.633932552627256e-05, "loss": 0.1428, "step": 21810 }, { "epoch": 0.51, "learning_rate": 2.6338445471826525e-05, "loss": 0.1495, "step": 21815 }, { "epoch": 0.51, "learning_rate": 2.633756541738049e-05, "loss": 0.1507, "step": 21820 }, { "epoch": 0.51, "learning_rate": 2.633668536293445e-05, "loss": 0.1268, "step": 21825 }, { "epoch": 0.51, "learning_rate": 2.633580530848842e-05, "loss": 0.1755, "step": 21830 }, { "epoch": 0.51, "learning_rate": 2.6334925254042385e-05, "loss": 0.3657, "step": 21835 }, { "epoch": 0.51, "learning_rate": 2.633404519959635e-05, "loss": 0.2466, "step": 21840 }, { "epoch": 0.51, "learning_rate": 2.6333165145150312e-05, "loss": 0.3376, "step": 21845 }, { "epoch": 0.51, "learning_rate": 2.633228509070428e-05, "loss": 0.5322, "step": 21850 }, { "epoch": 0.51, "learning_rate": 2.6331405036258245e-05, "loss": 0.3319, "step": 21855 }, { "epoch": 0.51, "learning_rate": 2.6330524981812207e-05, "loss": 0.1094, "step": 21860 }, { "epoch": 0.51, "learning_rate": 2.6329644927366172e-05, "loss": 0.1125, "step": 21865 }, { "epoch": 0.51, "learning_rate": 2.632876487292014e-05, "loss": 0.1118, "step": 21870 }, { "epoch": 0.51, "learning_rate": 2.6327884818474105e-05, "loss": 0.2555, "step": 21875 }, { "epoch": 0.51, "learning_rate": 2.6327004764028067e-05, "loss": 0.2082, "step": 21880 }, { "epoch": 0.51, "learning_rate": 2.6326124709582032e-05, "loss": 0.2101, "step": 21885 }, { "epoch": 0.51, "learning_rate": 2.6325244655136e-05, "loss": 0.3864, "step": 21890 }, { "epoch": 0.51, "learning_rate": 2.6324364600689962e-05, "loss": 0.3705, "step": 21895 }, { "epoch": 0.51, "learning_rate": 2.6323484546243927e-05, "loss": 0.3525, "step": 21900 }, { "epoch": 0.51, "learning_rate": 2.6322604491797892e-05, "loss": 0.3311, "step": 21905 }, { "epoch": 0.51, "learning_rate": 2.632172443735186e-05, "loss": 0.1563, "step": 21910 }, { "epoch": 0.51, "learning_rate": 2.6320844382905822e-05, "loss": 0.0557, "step": 21915 }, { "epoch": 0.51, "learning_rate": 2.6319964328459787e-05, "loss": 0.1966, "step": 21920 }, { "epoch": 0.51, "learning_rate": 2.6319084274013755e-05, "loss": 0.2853, "step": 21925 }, { "epoch": 0.51, "learning_rate": 2.6318204219567717e-05, "loss": 0.2326, "step": 21930 }, { "epoch": 0.51, "learning_rate": 2.6317324165121682e-05, "loss": 0.3716, "step": 21935 }, { "epoch": 0.51, "learning_rate": 2.6316444110675647e-05, "loss": 0.2766, "step": 21940 }, { "epoch": 0.51, "learning_rate": 2.6315564056229616e-05, "loss": 0.4135, "step": 21945 }, { "epoch": 0.51, "learning_rate": 2.6314684001783577e-05, "loss": 0.5958, "step": 21950 }, { "epoch": 0.51, "learning_rate": 2.6313803947337542e-05, "loss": 0.2772, "step": 21955 }, { "epoch": 0.51, "learning_rate": 2.6312923892891507e-05, "loss": 0.2057, "step": 21960 }, { "epoch": 0.51, "learning_rate": 2.6312043838445472e-05, "loss": 0.1, "step": 21965 }, { "epoch": 0.51, "learning_rate": 2.6311163783999437e-05, "loss": 0.123, "step": 21970 }, { "epoch": 0.51, "learning_rate": 2.6310283729553402e-05, "loss": 0.1691, "step": 21975 }, { "epoch": 0.51, "learning_rate": 2.6309403675107367e-05, "loss": 0.2336, "step": 21980 }, { "epoch": 0.51, "learning_rate": 2.6308523620661332e-05, "loss": 0.1514, "step": 21985 }, { "epoch": 0.51, "learning_rate": 2.6307643566215297e-05, "loss": 0.3866, "step": 21990 }, { "epoch": 0.51, "learning_rate": 2.6306763511769262e-05, "loss": 0.381, "step": 21995 }, { "epoch": 0.51, "learning_rate": 2.6305883457323224e-05, "loss": 0.6599, "step": 22000 }, { "epoch": 0.51, "learning_rate": 2.6305003402877192e-05, "loss": 0.2758, "step": 22005 }, { "epoch": 0.51, "learning_rate": 2.6304123348431157e-05, "loss": 0.1253, "step": 22010 }, { "epoch": 0.51, "learning_rate": 2.6303243293985122e-05, "loss": 0.189, "step": 22015 }, { "epoch": 0.51, "learning_rate": 2.6302363239539084e-05, "loss": 0.1877, "step": 22020 }, { "epoch": 0.51, "learning_rate": 2.6301483185093052e-05, "loss": 0.2559, "step": 22025 }, { "epoch": 0.51, "learning_rate": 2.6300603130647017e-05, "loss": 0.167, "step": 22030 }, { "epoch": 0.51, "learning_rate": 2.6299723076200983e-05, "loss": 0.109, "step": 22035 }, { "epoch": 0.51, "learning_rate": 2.6298843021754944e-05, "loss": 0.3327, "step": 22040 }, { "epoch": 0.51, "learning_rate": 2.6297962967308913e-05, "loss": 0.2888, "step": 22045 }, { "epoch": 0.51, "learning_rate": 2.6297082912862878e-05, "loss": 0.5, "step": 22050 }, { "epoch": 0.51, "learning_rate": 2.629620285841684e-05, "loss": 0.2224, "step": 22055 }, { "epoch": 0.51, "learning_rate": 2.6295322803970808e-05, "loss": 0.0668, "step": 22060 }, { "epoch": 0.51, "learning_rate": 2.6294442749524773e-05, "loss": 0.1013, "step": 22065 }, { "epoch": 0.51, "learning_rate": 2.6293562695078738e-05, "loss": 0.1321, "step": 22070 }, { "epoch": 0.52, "learning_rate": 2.62926826406327e-05, "loss": 0.3296, "step": 22075 }, { "epoch": 0.52, "learning_rate": 2.6291802586186668e-05, "loss": 0.1753, "step": 22080 }, { "epoch": 0.52, "learning_rate": 2.6290922531740633e-05, "loss": 0.3048, "step": 22085 }, { "epoch": 0.52, "learning_rate": 2.6290042477294594e-05, "loss": 0.322, "step": 22090 }, { "epoch": 0.52, "learning_rate": 2.628916242284856e-05, "loss": 0.4543, "step": 22095 }, { "epoch": 0.52, "learning_rate": 2.6288282368402528e-05, "loss": 0.5222, "step": 22100 }, { "epoch": 0.52, "learning_rate": 2.6287402313956493e-05, "loss": 0.2935, "step": 22105 }, { "epoch": 0.52, "learning_rate": 2.6286522259510454e-05, "loss": 0.0699, "step": 22110 }, { "epoch": 0.52, "learning_rate": 2.628564220506442e-05, "loss": 0.131, "step": 22115 }, { "epoch": 0.52, "learning_rate": 2.6284762150618388e-05, "loss": 0.2225, "step": 22120 }, { "epoch": 0.52, "learning_rate": 2.628388209617235e-05, "loss": 0.1734, "step": 22125 }, { "epoch": 0.52, "learning_rate": 2.6283002041726315e-05, "loss": 0.2581, "step": 22130 }, { "epoch": 0.52, "learning_rate": 2.628212198728028e-05, "loss": 0.1047, "step": 22135 }, { "epoch": 0.52, "learning_rate": 2.6281241932834248e-05, "loss": 0.253, "step": 22140 }, { "epoch": 0.52, "learning_rate": 2.628036187838821e-05, "loss": 0.3869, "step": 22145 }, { "epoch": 0.52, "learning_rate": 2.6279481823942175e-05, "loss": 0.5181, "step": 22150 }, { "epoch": 0.52, "learning_rate": 2.627860176949614e-05, "loss": 0.2711, "step": 22155 }, { "epoch": 0.52, "learning_rate": 2.6277721715050105e-05, "loss": 0.0444, "step": 22160 }, { "epoch": 0.52, "learning_rate": 2.627684166060407e-05, "loss": 0.0645, "step": 22165 }, { "epoch": 0.52, "learning_rate": 2.6275961606158035e-05, "loss": 0.2167, "step": 22170 }, { "epoch": 0.52, "learning_rate": 2.6275081551712e-05, "loss": 0.1711, "step": 22175 }, { "epoch": 0.52, "learning_rate": 2.6274201497265965e-05, "loss": 0.145, "step": 22180 }, { "epoch": 0.52, "learning_rate": 2.627332144281993e-05, "loss": 0.1045, "step": 22185 }, { "epoch": 0.52, "learning_rate": 2.6272441388373895e-05, "loss": 0.2192, "step": 22190 }, { "epoch": 0.52, "learning_rate": 2.627156133392786e-05, "loss": 0.4287, "step": 22195 }, { "epoch": 0.52, "learning_rate": 2.6270681279481825e-05, "loss": 0.6397, "step": 22200 }, { "epoch": 0.52, "learning_rate": 2.626980122503579e-05, "loss": 0.2718, "step": 22205 }, { "epoch": 0.52, "learning_rate": 2.6268921170589755e-05, "loss": 0.1405, "step": 22210 }, { "epoch": 0.52, "learning_rate": 2.626804111614372e-05, "loss": 0.1618, "step": 22215 }, { "epoch": 0.52, "learning_rate": 2.6267161061697685e-05, "loss": 0.2482, "step": 22220 }, { "epoch": 0.52, "learning_rate": 2.626628100725165e-05, "loss": 0.0634, "step": 22225 }, { "epoch": 0.52, "learning_rate": 2.626540095280561e-05, "loss": 0.3767, "step": 22230 }, { "epoch": 0.52, "learning_rate": 2.626452089835958e-05, "loss": 0.224, "step": 22235 }, { "epoch": 0.52, "learning_rate": 2.6263640843913545e-05, "loss": 0.422, "step": 22240 }, { "epoch": 0.52, "learning_rate": 2.626276078946751e-05, "loss": 0.2346, "step": 22245 }, { "epoch": 0.52, "learning_rate": 2.626188073502147e-05, "loss": 0.4941, "step": 22250 }, { "epoch": 0.52, "learning_rate": 2.626100068057544e-05, "loss": 0.2815, "step": 22255 }, { "epoch": 0.52, "learning_rate": 2.6260120626129405e-05, "loss": 0.1947, "step": 22260 }, { "epoch": 0.52, "learning_rate": 2.6259240571683367e-05, "loss": 0.103, "step": 22265 }, { "epoch": 0.52, "learning_rate": 2.6258360517237332e-05, "loss": 0.2059, "step": 22270 }, { "epoch": 0.52, "learning_rate": 2.62574804627913e-05, "loss": 0.1811, "step": 22275 }, { "epoch": 0.52, "learning_rate": 2.6256600408345265e-05, "loss": 0.2215, "step": 22280 }, { "epoch": 0.52, "learning_rate": 2.6255720353899227e-05, "loss": 0.3047, "step": 22285 }, { "epoch": 0.52, "learning_rate": 2.6254840299453192e-05, "loss": 0.395, "step": 22290 }, { "epoch": 0.52, "learning_rate": 2.625396024500716e-05, "loss": 0.2684, "step": 22295 }, { "epoch": 0.52, "learning_rate": 2.6253080190561122e-05, "loss": 0.6539, "step": 22300 }, { "epoch": 0.52, "learning_rate": 2.6252200136115087e-05, "loss": 0.2084, "step": 22305 }, { "epoch": 0.52, "learning_rate": 2.6251320081669052e-05, "loss": 0.1824, "step": 22310 }, { "epoch": 0.52, "learning_rate": 2.625044002722302e-05, "loss": 0.1145, "step": 22315 }, { "epoch": 0.52, "learning_rate": 2.6249559972776982e-05, "loss": 0.0998, "step": 22320 }, { "epoch": 0.52, "learning_rate": 2.6248679918330947e-05, "loss": 0.177, "step": 22325 }, { "epoch": 0.52, "learning_rate": 2.6247799863884915e-05, "loss": 0.2468, "step": 22330 }, { "epoch": 0.52, "learning_rate": 2.624691980943888e-05, "loss": 0.2945, "step": 22335 }, { "epoch": 0.52, "learning_rate": 2.6246039754992842e-05, "loss": 0.4621, "step": 22340 }, { "epoch": 0.52, "learning_rate": 2.6245159700546807e-05, "loss": 0.4699, "step": 22345 }, { "epoch": 0.52, "learning_rate": 2.6244279646100775e-05, "loss": 0.622, "step": 22350 }, { "epoch": 0.52, "learning_rate": 2.6243399591654737e-05, "loss": 0.2491, "step": 22355 }, { "epoch": 0.52, "learning_rate": 2.6242519537208702e-05, "loss": 0.0971, "step": 22360 }, { "epoch": 0.52, "learning_rate": 2.6241639482762667e-05, "loss": 0.1645, "step": 22365 }, { "epoch": 0.52, "learning_rate": 2.6240759428316635e-05, "loss": 0.156, "step": 22370 }, { "epoch": 0.52, "learning_rate": 2.6239879373870597e-05, "loss": 0.1649, "step": 22375 }, { "epoch": 0.52, "learning_rate": 2.6238999319424562e-05, "loss": 0.15, "step": 22380 }, { "epoch": 0.52, "learning_rate": 2.6238119264978527e-05, "loss": 0.2512, "step": 22385 }, { "epoch": 0.52, "learning_rate": 2.6237239210532492e-05, "loss": 0.3395, "step": 22390 }, { "epoch": 0.52, "learning_rate": 2.6236359156086457e-05, "loss": 0.3834, "step": 22395 }, { "epoch": 0.52, "learning_rate": 2.6235479101640422e-05, "loss": 0.8219, "step": 22400 }, { "epoch": 0.52, "learning_rate": 2.6234599047194387e-05, "loss": 0.3257, "step": 22405 }, { "epoch": 0.52, "learning_rate": 2.6233718992748352e-05, "loss": 0.1207, "step": 22410 }, { "epoch": 0.52, "learning_rate": 2.6232838938302317e-05, "loss": 0.1011, "step": 22415 }, { "epoch": 0.52, "learning_rate": 2.6231958883856282e-05, "loss": 0.1237, "step": 22420 }, { "epoch": 0.52, "learning_rate": 2.6231078829410244e-05, "loss": 0.1406, "step": 22425 }, { "epoch": 0.52, "learning_rate": 2.6230198774964212e-05, "loss": 0.1246, "step": 22430 }, { "epoch": 0.52, "learning_rate": 2.6229318720518177e-05, "loss": 0.3782, "step": 22435 }, { "epoch": 0.52, "learning_rate": 2.6228438666072142e-05, "loss": 0.2558, "step": 22440 }, { "epoch": 0.52, "learning_rate": 2.6227558611626104e-05, "loss": 0.386, "step": 22445 }, { "epoch": 0.52, "learning_rate": 2.6226678557180072e-05, "loss": 0.8218, "step": 22450 }, { "epoch": 0.52, "learning_rate": 2.6225798502734037e-05, "loss": 0.3877, "step": 22455 }, { "epoch": 0.52, "learning_rate": 2.6224918448288e-05, "loss": 0.1082, "step": 22460 }, { "epoch": 0.52, "learning_rate": 2.6224038393841968e-05, "loss": 0.0639, "step": 22465 }, { "epoch": 0.52, "learning_rate": 2.6223158339395933e-05, "loss": 0.2415, "step": 22470 }, { "epoch": 0.52, "learning_rate": 2.6222278284949898e-05, "loss": 0.1713, "step": 22475 }, { "epoch": 0.52, "learning_rate": 2.622139823050386e-05, "loss": 0.1743, "step": 22480 }, { "epoch": 0.52, "learning_rate": 2.6220518176057828e-05, "loss": 0.2874, "step": 22485 }, { "epoch": 0.52, "learning_rate": 2.6219638121611793e-05, "loss": 0.2796, "step": 22490 }, { "epoch": 0.52, "learning_rate": 2.6218758067165754e-05, "loss": 0.299, "step": 22495 }, { "epoch": 0.52, "learning_rate": 2.621787801271972e-05, "loss": 0.6539, "step": 22500 }, { "epoch": 0.53, "learning_rate": 2.6216997958273688e-05, "loss": 0.2174, "step": 22505 }, { "epoch": 0.53, "learning_rate": 2.6216117903827653e-05, "loss": 0.068, "step": 22510 }, { "epoch": 0.53, "learning_rate": 2.6215237849381614e-05, "loss": 0.1783, "step": 22515 }, { "epoch": 0.53, "learning_rate": 2.621435779493558e-05, "loss": 0.2243, "step": 22520 }, { "epoch": 0.53, "learning_rate": 2.6213477740489548e-05, "loss": 0.2049, "step": 22525 }, { "epoch": 0.53, "learning_rate": 2.621259768604351e-05, "loss": 0.1679, "step": 22530 }, { "epoch": 0.53, "learning_rate": 2.6211717631597474e-05, "loss": 0.1952, "step": 22535 }, { "epoch": 0.53, "learning_rate": 2.621083757715144e-05, "loss": 0.2295, "step": 22540 }, { "epoch": 0.53, "learning_rate": 2.6209957522705408e-05, "loss": 0.4451, "step": 22545 }, { "epoch": 0.53, "learning_rate": 2.620907746825937e-05, "loss": 0.3598, "step": 22550 }, { "epoch": 0.53, "learning_rate": 2.6208197413813334e-05, "loss": 0.2281, "step": 22555 }, { "epoch": 0.53, "learning_rate": 2.62073173593673e-05, "loss": 0.2382, "step": 22560 }, { "epoch": 0.53, "learning_rate": 2.6206437304921265e-05, "loss": 0.0684, "step": 22565 }, { "epoch": 0.53, "learning_rate": 2.620555725047523e-05, "loss": 0.2537, "step": 22570 }, { "epoch": 0.53, "learning_rate": 2.6204677196029195e-05, "loss": 0.1546, "step": 22575 }, { "epoch": 0.53, "learning_rate": 2.620379714158316e-05, "loss": 0.0672, "step": 22580 }, { "epoch": 0.53, "learning_rate": 2.6202917087137125e-05, "loss": 0.227, "step": 22585 }, { "epoch": 0.53, "learning_rate": 2.620203703269109e-05, "loss": 0.3289, "step": 22590 }, { "epoch": 0.53, "learning_rate": 2.6201156978245055e-05, "loss": 0.5059, "step": 22595 }, { "epoch": 0.53, "learning_rate": 2.620027692379902e-05, "loss": 0.3121, "step": 22600 }, { "epoch": 0.53, "learning_rate": 2.6199396869352985e-05, "loss": 0.1767, "step": 22605 }, { "epoch": 0.53, "learning_rate": 2.619851681490695e-05, "loss": 0.1433, "step": 22610 }, { "epoch": 0.53, "learning_rate": 2.6197636760460915e-05, "loss": 0.2547, "step": 22615 }, { "epoch": 0.53, "learning_rate": 2.619675670601488e-05, "loss": 0.1822, "step": 22620 }, { "epoch": 0.53, "learning_rate": 2.6195876651568845e-05, "loss": 0.2671, "step": 22625 }, { "epoch": 0.53, "learning_rate": 2.619499659712281e-05, "loss": 0.294, "step": 22630 }, { "epoch": 0.53, "learning_rate": 2.6194116542676775e-05, "loss": 0.184, "step": 22635 }, { "epoch": 0.53, "learning_rate": 2.619323648823074e-05, "loss": 0.3595, "step": 22640 }, { "epoch": 0.53, "learning_rate": 2.6192356433784705e-05, "loss": 0.7812, "step": 22645 }, { "epoch": 0.53, "learning_rate": 2.619147637933867e-05, "loss": 0.6475, "step": 22650 }, { "epoch": 0.53, "learning_rate": 2.619059632489263e-05, "loss": 0.3591, "step": 22655 }, { "epoch": 0.53, "learning_rate": 2.61897162704466e-05, "loss": 0.0844, "step": 22660 }, { "epoch": 0.53, "learning_rate": 2.6188836216000565e-05, "loss": 0.1121, "step": 22665 }, { "epoch": 0.53, "learning_rate": 2.618795616155453e-05, "loss": 0.1693, "step": 22670 }, { "epoch": 0.53, "learning_rate": 2.618707610710849e-05, "loss": 0.1783, "step": 22675 }, { "epoch": 0.53, "learning_rate": 2.618619605266246e-05, "loss": 0.1746, "step": 22680 }, { "epoch": 0.53, "learning_rate": 2.6185315998216425e-05, "loss": 0.1596, "step": 22685 }, { "epoch": 0.53, "learning_rate": 2.6184435943770387e-05, "loss": 0.2039, "step": 22690 }, { "epoch": 0.53, "learning_rate": 2.618355588932435e-05, "loss": 0.4559, "step": 22695 }, { "epoch": 0.53, "learning_rate": 2.618267583487832e-05, "loss": 0.7865, "step": 22700 }, { "epoch": 0.53, "learning_rate": 2.6181795780432285e-05, "loss": 0.2307, "step": 22705 }, { "epoch": 0.53, "learning_rate": 2.6180915725986247e-05, "loss": 0.1132, "step": 22710 }, { "epoch": 0.53, "learning_rate": 2.6180035671540212e-05, "loss": 0.0671, "step": 22715 }, { "epoch": 0.53, "learning_rate": 2.617915561709418e-05, "loss": 0.1782, "step": 22720 }, { "epoch": 0.53, "learning_rate": 2.6178275562648142e-05, "loss": 0.1347, "step": 22725 }, { "epoch": 0.53, "learning_rate": 2.6177395508202107e-05, "loss": 0.1861, "step": 22730 }, { "epoch": 0.53, "learning_rate": 2.6176515453756075e-05, "loss": 0.2007, "step": 22735 }, { "epoch": 0.53, "learning_rate": 2.617563539931004e-05, "loss": 0.3757, "step": 22740 }, { "epoch": 0.53, "learning_rate": 2.6174755344864002e-05, "loss": 0.2618, "step": 22745 }, { "epoch": 0.53, "learning_rate": 2.6173875290417967e-05, "loss": 0.3278, "step": 22750 }, { "epoch": 0.53, "learning_rate": 2.6172995235971935e-05, "loss": 0.2237, "step": 22755 }, { "epoch": 0.53, "learning_rate": 2.6172115181525897e-05, "loss": 0.0534, "step": 22760 }, { "epoch": 0.53, "learning_rate": 2.6171235127079862e-05, "loss": 0.2873, "step": 22765 }, { "epoch": 0.53, "learning_rate": 2.6170355072633827e-05, "loss": 0.1352, "step": 22770 }, { "epoch": 0.53, "learning_rate": 2.6169475018187795e-05, "loss": 0.2928, "step": 22775 }, { "epoch": 0.53, "learning_rate": 2.6168594963741757e-05, "loss": 0.2064, "step": 22780 }, { "epoch": 0.53, "learning_rate": 2.6167714909295722e-05, "loss": 0.1948, "step": 22785 }, { "epoch": 0.53, "learning_rate": 2.6166834854849687e-05, "loss": 0.3766, "step": 22790 }, { "epoch": 0.53, "learning_rate": 2.6165954800403652e-05, "loss": 0.4063, "step": 22795 }, { "epoch": 0.53, "learning_rate": 2.6165074745957617e-05, "loss": 0.464, "step": 22800 }, { "epoch": 0.53, "learning_rate": 2.6164194691511582e-05, "loss": 0.26, "step": 22805 }, { "epoch": 0.53, "learning_rate": 2.6163314637065547e-05, "loss": 0.1562, "step": 22810 }, { "epoch": 0.53, "learning_rate": 2.6162434582619512e-05, "loss": 0.1174, "step": 22815 }, { "epoch": 0.53, "learning_rate": 2.6161554528173477e-05, "loss": 0.1103, "step": 22820 }, { "epoch": 0.53, "learning_rate": 2.6160674473727442e-05, "loss": 0.2157, "step": 22825 }, { "epoch": 0.53, "learning_rate": 2.6159794419281404e-05, "loss": 0.1236, "step": 22830 }, { "epoch": 0.53, "learning_rate": 2.6158914364835372e-05, "loss": 0.236, "step": 22835 }, { "epoch": 0.53, "learning_rate": 2.6158034310389337e-05, "loss": 0.2627, "step": 22840 }, { "epoch": 0.53, "learning_rate": 2.6157154255943302e-05, "loss": 0.4303, "step": 22845 }, { "epoch": 0.53, "learning_rate": 2.6156274201497264e-05, "loss": 0.3756, "step": 22850 }, { "epoch": 0.53, "learning_rate": 2.6155394147051232e-05, "loss": 0.2742, "step": 22855 }, { "epoch": 0.53, "learning_rate": 2.6154514092605197e-05, "loss": 0.0669, "step": 22860 }, { "epoch": 0.53, "learning_rate": 2.615363403815916e-05, "loss": 0.1659, "step": 22865 }, { "epoch": 0.53, "learning_rate": 2.6152753983713127e-05, "loss": 0.1813, "step": 22870 }, { "epoch": 0.53, "learning_rate": 2.6151873929267092e-05, "loss": 0.2737, "step": 22875 }, { "epoch": 0.53, "learning_rate": 2.6150993874821057e-05, "loss": 0.162, "step": 22880 }, { "epoch": 0.53, "learning_rate": 2.615011382037502e-05, "loss": 0.3525, "step": 22885 }, { "epoch": 0.53, "learning_rate": 2.6149233765928987e-05, "loss": 0.3782, "step": 22890 }, { "epoch": 0.53, "learning_rate": 2.6148353711482952e-05, "loss": 0.2127, "step": 22895 }, { "epoch": 0.53, "learning_rate": 2.6147473657036914e-05, "loss": 0.4685, "step": 22900 }, { "epoch": 0.53, "learning_rate": 2.614659360259088e-05, "loss": 0.2618, "step": 22905 }, { "epoch": 0.53, "learning_rate": 2.6145713548144848e-05, "loss": 0.0865, "step": 22910 }, { "epoch": 0.53, "learning_rate": 2.6144833493698813e-05, "loss": 0.1542, "step": 22915 }, { "epoch": 0.53, "learning_rate": 2.6143953439252774e-05, "loss": 0.1105, "step": 22920 }, { "epoch": 0.53, "learning_rate": 2.614307338480674e-05, "loss": 0.1998, "step": 22925 }, { "epoch": 0.53, "learning_rate": 2.6142193330360708e-05, "loss": 0.2125, "step": 22930 }, { "epoch": 0.54, "learning_rate": 2.6141313275914673e-05, "loss": 0.2491, "step": 22935 }, { "epoch": 0.54, "learning_rate": 2.6140433221468634e-05, "loss": 0.466, "step": 22940 }, { "epoch": 0.54, "learning_rate": 2.61395531670226e-05, "loss": 0.3237, "step": 22945 }, { "epoch": 0.54, "learning_rate": 2.6138673112576568e-05, "loss": 0.9216, "step": 22950 }, { "epoch": 0.54, "learning_rate": 2.613779305813053e-05, "loss": 0.2235, "step": 22955 }, { "epoch": 0.54, "learning_rate": 2.6136913003684494e-05, "loss": 0.092, "step": 22960 }, { "epoch": 0.54, "learning_rate": 2.613603294923846e-05, "loss": 0.1782, "step": 22965 }, { "epoch": 0.54, "learning_rate": 2.6135152894792428e-05, "loss": 0.1708, "step": 22970 }, { "epoch": 0.54, "learning_rate": 2.613427284034639e-05, "loss": 0.1683, "step": 22975 }, { "epoch": 0.54, "learning_rate": 2.6133392785900354e-05, "loss": 0.1658, "step": 22980 }, { "epoch": 0.54, "learning_rate": 2.613251273145432e-05, "loss": 0.2315, "step": 22985 }, { "epoch": 0.54, "learning_rate": 2.6131632677008284e-05, "loss": 0.414, "step": 22990 }, { "epoch": 0.54, "learning_rate": 2.613075262256225e-05, "loss": 0.3472, "step": 22995 }, { "epoch": 0.54, "learning_rate": 2.6129872568116215e-05, "loss": 0.6911, "step": 23000 }, { "epoch": 0.54, "learning_rate": 2.6128992513670183e-05, "loss": 0.2511, "step": 23005 }, { "epoch": 0.54, "learning_rate": 2.6128112459224145e-05, "loss": 0.0604, "step": 23010 }, { "epoch": 0.54, "learning_rate": 2.612723240477811e-05, "loss": 0.1208, "step": 23015 }, { "epoch": 0.54, "learning_rate": 2.6126352350332075e-05, "loss": 0.1684, "step": 23020 }, { "epoch": 0.54, "learning_rate": 2.612547229588604e-05, "loss": 0.0842, "step": 23025 }, { "epoch": 0.54, "learning_rate": 2.6124592241440005e-05, "loss": 0.1817, "step": 23030 }, { "epoch": 0.54, "learning_rate": 2.612371218699397e-05, "loss": 0.215, "step": 23035 }, { "epoch": 0.54, "learning_rate": 2.6122832132547935e-05, "loss": 0.2076, "step": 23040 }, { "epoch": 0.54, "learning_rate": 2.61219520781019e-05, "loss": 0.5234, "step": 23045 }, { "epoch": 0.54, "learning_rate": 2.6121072023655865e-05, "loss": 0.6801, "step": 23050 }, { "epoch": 0.54, "learning_rate": 2.612019196920983e-05, "loss": 0.1782, "step": 23055 }, { "epoch": 0.54, "learning_rate": 2.611931191476379e-05, "loss": 0.0767, "step": 23060 }, { "epoch": 0.54, "learning_rate": 2.611843186031776e-05, "loss": 0.1117, "step": 23065 }, { "epoch": 0.54, "learning_rate": 2.6117551805871725e-05, "loss": 0.2096, "step": 23070 }, { "epoch": 0.54, "learning_rate": 2.611667175142569e-05, "loss": 0.1388, "step": 23075 }, { "epoch": 0.54, "learning_rate": 2.611579169697965e-05, "loss": 0.0806, "step": 23080 }, { "epoch": 0.54, "learning_rate": 2.611491164253362e-05, "loss": 0.2859, "step": 23085 }, { "epoch": 0.54, "learning_rate": 2.6114031588087585e-05, "loss": 0.4294, "step": 23090 }, { "epoch": 0.54, "learning_rate": 2.6113151533641547e-05, "loss": 0.2425, "step": 23095 }, { "epoch": 0.54, "learning_rate": 2.611227147919551e-05, "loss": 0.3734, "step": 23100 }, { "epoch": 0.54, "learning_rate": 2.611139142474948e-05, "loss": 0.2542, "step": 23105 }, { "epoch": 0.54, "learning_rate": 2.6110511370303445e-05, "loss": 0.1302, "step": 23110 }, { "epoch": 0.54, "learning_rate": 2.6109631315857407e-05, "loss": 0.1796, "step": 23115 }, { "epoch": 0.54, "learning_rate": 2.610875126141137e-05, "loss": 0.2141, "step": 23120 }, { "epoch": 0.54, "learning_rate": 2.610787120696534e-05, "loss": 0.2556, "step": 23125 }, { "epoch": 0.54, "learning_rate": 2.61069911525193e-05, "loss": 0.2103, "step": 23130 }, { "epoch": 0.54, "learning_rate": 2.6106111098073267e-05, "loss": 0.3067, "step": 23135 }, { "epoch": 0.54, "learning_rate": 2.6105231043627235e-05, "loss": 0.0789, "step": 23140 }, { "epoch": 0.54, "learning_rate": 2.61043509891812e-05, "loss": 0.4749, "step": 23145 }, { "epoch": 0.54, "learning_rate": 2.6103470934735162e-05, "loss": 0.5633, "step": 23150 }, { "epoch": 0.54, "learning_rate": 2.6102590880289127e-05, "loss": 0.2449, "step": 23155 }, { "epoch": 0.54, "learning_rate": 2.6101710825843095e-05, "loss": 0.0944, "step": 23160 }, { "epoch": 0.54, "learning_rate": 2.6100830771397057e-05, "loss": 0.0607, "step": 23165 }, { "epoch": 0.54, "learning_rate": 2.6099950716951022e-05, "loss": 0.1254, "step": 23170 }, { "epoch": 0.54, "learning_rate": 2.6099070662504987e-05, "loss": 0.1623, "step": 23175 }, { "epoch": 0.54, "learning_rate": 2.6098190608058955e-05, "loss": 0.2208, "step": 23180 }, { "epoch": 0.54, "learning_rate": 2.6097310553612917e-05, "loss": 0.1628, "step": 23185 }, { "epoch": 0.54, "learning_rate": 2.6096430499166882e-05, "loss": 0.3108, "step": 23190 }, { "epoch": 0.54, "learning_rate": 2.6095550444720847e-05, "loss": 0.3434, "step": 23195 }, { "epoch": 0.54, "learning_rate": 2.6094670390274812e-05, "loss": 0.6925, "step": 23200 }, { "epoch": 0.54, "learning_rate": 2.6093790335828777e-05, "loss": 0.1685, "step": 23205 }, { "epoch": 0.54, "learning_rate": 2.6092910281382742e-05, "loss": 0.1108, "step": 23210 }, { "epoch": 0.54, "learning_rate": 2.6092030226936707e-05, "loss": 0.1191, "step": 23215 }, { "epoch": 0.54, "learning_rate": 2.6091150172490672e-05, "loss": 0.1258, "step": 23220 }, { "epoch": 0.54, "learning_rate": 2.6090270118044637e-05, "loss": 0.1486, "step": 23225 }, { "epoch": 0.54, "learning_rate": 2.6089390063598602e-05, "loss": 0.2017, "step": 23230 }, { "epoch": 0.54, "learning_rate": 2.6088510009152567e-05, "loss": 0.2096, "step": 23235 }, { "epoch": 0.54, "learning_rate": 2.6087629954706532e-05, "loss": 0.3818, "step": 23240 }, { "epoch": 0.54, "learning_rate": 2.6086749900260497e-05, "loss": 0.3623, "step": 23245 }, { "epoch": 0.54, "learning_rate": 2.6085869845814462e-05, "loss": 0.4953, "step": 23250 }, { "epoch": 0.54, "learning_rate": 2.6084989791368424e-05, "loss": 0.2192, "step": 23255 }, { "epoch": 0.54, "learning_rate": 2.6084109736922392e-05, "loss": 0.1404, "step": 23260 }, { "epoch": 0.54, "learning_rate": 2.6083229682476357e-05, "loss": 0.0952, "step": 23265 }, { "epoch": 0.54, "learning_rate": 2.6082349628030322e-05, "loss": 0.203, "step": 23270 }, { "epoch": 0.54, "learning_rate": 2.6081469573584287e-05, "loss": 0.2247, "step": 23275 }, { "epoch": 0.54, "learning_rate": 2.6080589519138252e-05, "loss": 0.1116, "step": 23280 }, { "epoch": 0.54, "learning_rate": 2.6079709464692217e-05, "loss": 0.2197, "step": 23285 }, { "epoch": 0.54, "learning_rate": 2.607882941024618e-05, "loss": 0.2446, "step": 23290 }, { "epoch": 0.54, "learning_rate": 2.6077949355800147e-05, "loss": 0.2222, "step": 23295 }, { "epoch": 0.54, "learning_rate": 2.6077069301354112e-05, "loss": 0.4343, "step": 23300 }, { "epoch": 0.54, "learning_rate": 2.6076189246908077e-05, "loss": 0.2485, "step": 23305 }, { "epoch": 0.54, "learning_rate": 2.607530919246204e-05, "loss": 0.1084, "step": 23310 }, { "epoch": 0.54, "learning_rate": 2.6074429138016007e-05, "loss": 0.0939, "step": 23315 }, { "epoch": 0.54, "learning_rate": 2.6073549083569972e-05, "loss": 0.1025, "step": 23320 }, { "epoch": 0.54, "learning_rate": 2.6072669029123934e-05, "loss": 0.0973, "step": 23325 }, { "epoch": 0.54, "learning_rate": 2.60717889746779e-05, "loss": 0.2465, "step": 23330 }, { "epoch": 0.54, "learning_rate": 2.6070908920231868e-05, "loss": 0.6017, "step": 23335 }, { "epoch": 0.54, "learning_rate": 2.6070028865785833e-05, "loss": 0.2061, "step": 23340 }, { "epoch": 0.54, "learning_rate": 2.6069148811339794e-05, "loss": 0.2862, "step": 23345 }, { "epoch": 0.54, "learning_rate": 2.606826875689376e-05, "loss": 0.5728, "step": 23350 }, { "epoch": 0.54, "learning_rate": 2.6067388702447728e-05, "loss": 0.2078, "step": 23355 }, { "epoch": 0.55, "learning_rate": 2.606650864800169e-05, "loss": 0.0994, "step": 23360 }, { "epoch": 0.55, "learning_rate": 2.6065628593555654e-05, "loss": 0.1822, "step": 23365 }, { "epoch": 0.55, "learning_rate": 2.606474853910962e-05, "loss": 0.1422, "step": 23370 }, { "epoch": 0.55, "learning_rate": 2.6063868484663588e-05, "loss": 0.1504, "step": 23375 }, { "epoch": 0.55, "learning_rate": 2.606298843021755e-05, "loss": 0.111, "step": 23380 }, { "epoch": 0.55, "learning_rate": 2.6062108375771514e-05, "loss": 0.2023, "step": 23385 }, { "epoch": 0.55, "learning_rate": 2.606122832132548e-05, "loss": 0.1959, "step": 23390 }, { "epoch": 0.55, "learning_rate": 2.6060348266879444e-05, "loss": 0.3092, "step": 23395 }, { "epoch": 0.55, "learning_rate": 2.605946821243341e-05, "loss": 0.5602, "step": 23400 }, { "epoch": 0.55, "learning_rate": 2.6058588157987374e-05, "loss": 0.2217, "step": 23405 }, { "epoch": 0.55, "learning_rate": 2.6057708103541343e-05, "loss": 0.0739, "step": 23410 }, { "epoch": 0.55, "learning_rate": 2.6056828049095304e-05, "loss": 0.0602, "step": 23415 }, { "epoch": 0.55, "learning_rate": 2.605594799464927e-05, "loss": 0.1935, "step": 23420 }, { "epoch": 0.55, "learning_rate": 2.6055067940203235e-05, "loss": 0.3781, "step": 23425 }, { "epoch": 0.55, "learning_rate": 2.60541878857572e-05, "loss": 0.1813, "step": 23430 }, { "epoch": 0.55, "learning_rate": 2.6053307831311165e-05, "loss": 0.2111, "step": 23435 }, { "epoch": 0.55, "learning_rate": 2.605242777686513e-05, "loss": 0.2254, "step": 23440 }, { "epoch": 0.55, "learning_rate": 2.6051547722419095e-05, "loss": 0.4239, "step": 23445 }, { "epoch": 0.55, "learning_rate": 2.605066766797306e-05, "loss": 0.4542, "step": 23450 }, { "epoch": 0.55, "learning_rate": 2.6049787613527025e-05, "loss": 0.2701, "step": 23455 }, { "epoch": 0.55, "learning_rate": 2.604890755908099e-05, "loss": 0.1793, "step": 23460 }, { "epoch": 0.55, "learning_rate": 2.604802750463495e-05, "loss": 0.1235, "step": 23465 }, { "epoch": 0.55, "learning_rate": 2.604714745018892e-05, "loss": 0.148, "step": 23470 }, { "epoch": 0.55, "learning_rate": 2.6046267395742885e-05, "loss": 0.1443, "step": 23475 }, { "epoch": 0.55, "learning_rate": 2.604538734129685e-05, "loss": 0.224, "step": 23480 }, { "epoch": 0.55, "learning_rate": 2.604450728685081e-05, "loss": 0.1713, "step": 23485 }, { "epoch": 0.55, "learning_rate": 2.604362723240478e-05, "loss": 0.4366, "step": 23490 }, { "epoch": 0.55, "learning_rate": 2.6042747177958745e-05, "loss": 0.4002, "step": 23495 }, { "epoch": 0.55, "learning_rate": 2.6041867123512706e-05, "loss": 0.5718, "step": 23500 }, { "epoch": 0.55, "learning_rate": 2.604098706906667e-05, "loss": 0.2244, "step": 23505 }, { "epoch": 0.55, "learning_rate": 2.604010701462064e-05, "loss": 0.065, "step": 23510 }, { "epoch": 0.55, "learning_rate": 2.6039226960174605e-05, "loss": 0.1658, "step": 23515 }, { "epoch": 0.55, "learning_rate": 2.6038346905728567e-05, "loss": 0.2159, "step": 23520 }, { "epoch": 0.55, "learning_rate": 2.603746685128253e-05, "loss": 0.2646, "step": 23525 }, { "epoch": 0.55, "learning_rate": 2.60365867968365e-05, "loss": 0.2444, "step": 23530 }, { "epoch": 0.55, "learning_rate": 2.603570674239046e-05, "loss": 0.2827, "step": 23535 }, { "epoch": 0.55, "learning_rate": 2.6034826687944427e-05, "loss": 0.3409, "step": 23540 }, { "epoch": 0.55, "learning_rate": 2.6033946633498395e-05, "loss": 0.4397, "step": 23545 }, { "epoch": 0.55, "learning_rate": 2.603306657905236e-05, "loss": 0.5895, "step": 23550 }, { "epoch": 0.55, "learning_rate": 2.603218652460632e-05, "loss": 0.3063, "step": 23555 }, { "epoch": 0.55, "learning_rate": 2.6031306470160287e-05, "loss": 0.2131, "step": 23560 }, { "epoch": 0.55, "learning_rate": 2.6030426415714255e-05, "loss": 0.047, "step": 23565 }, { "epoch": 0.55, "learning_rate": 2.602954636126822e-05, "loss": 0.188, "step": 23570 }, { "epoch": 0.55, "learning_rate": 2.6028666306822182e-05, "loss": 0.1778, "step": 23575 }, { "epoch": 0.55, "learning_rate": 2.6027786252376147e-05, "loss": 0.2844, "step": 23580 }, { "epoch": 0.55, "learning_rate": 2.6026906197930115e-05, "loss": 0.3481, "step": 23585 }, { "epoch": 0.55, "learning_rate": 2.6026026143484077e-05, "loss": 0.0943, "step": 23590 }, { "epoch": 0.55, "learning_rate": 2.6025146089038042e-05, "loss": 0.6128, "step": 23595 }, { "epoch": 0.55, "learning_rate": 2.6024266034592007e-05, "loss": 0.3466, "step": 23600 }, { "epoch": 0.55, "learning_rate": 2.6023385980145975e-05, "loss": 0.1841, "step": 23605 }, { "epoch": 0.55, "learning_rate": 2.6022505925699937e-05, "loss": 0.1377, "step": 23610 }, { "epoch": 0.55, "learning_rate": 2.6021625871253902e-05, "loss": 0.2216, "step": 23615 }, { "epoch": 0.55, "learning_rate": 2.6020745816807867e-05, "loss": 0.1689, "step": 23620 }, { "epoch": 0.55, "learning_rate": 2.6019865762361832e-05, "loss": 0.1418, "step": 23625 }, { "epoch": 0.55, "learning_rate": 2.6018985707915797e-05, "loss": 0.1826, "step": 23630 }, { "epoch": 0.55, "learning_rate": 2.6018105653469762e-05, "loss": 0.1555, "step": 23635 }, { "epoch": 0.55, "learning_rate": 2.6017225599023727e-05, "loss": 0.2385, "step": 23640 }, { "epoch": 0.55, "learning_rate": 2.6016345544577692e-05, "loss": 0.3627, "step": 23645 }, { "epoch": 0.55, "learning_rate": 2.6015465490131657e-05, "loss": 0.4817, "step": 23650 }, { "epoch": 0.55, "learning_rate": 2.6014585435685622e-05, "loss": 0.2743, "step": 23655 }, { "epoch": 0.55, "learning_rate": 2.6013705381239584e-05, "loss": 0.1378, "step": 23660 }, { "epoch": 0.55, "learning_rate": 2.6012825326793552e-05, "loss": 0.1801, "step": 23665 }, { "epoch": 0.55, "learning_rate": 2.6011945272347517e-05, "loss": 0.1044, "step": 23670 }, { "epoch": 0.55, "learning_rate": 2.6011065217901482e-05, "loss": 0.1691, "step": 23675 }, { "epoch": 0.55, "learning_rate": 2.6010185163455447e-05, "loss": 0.263, "step": 23680 }, { "epoch": 0.55, "learning_rate": 2.6009305109009412e-05, "loss": 0.2544, "step": 23685 }, { "epoch": 0.55, "learning_rate": 2.6008425054563377e-05, "loss": 0.2695, "step": 23690 }, { "epoch": 0.55, "learning_rate": 2.600754500011734e-05, "loss": 0.3404, "step": 23695 }, { "epoch": 0.55, "learning_rate": 2.6006664945671307e-05, "loss": 0.4565, "step": 23700 }, { "epoch": 0.55, "learning_rate": 2.6005784891225272e-05, "loss": 0.2353, "step": 23705 }, { "epoch": 0.55, "learning_rate": 2.6004904836779237e-05, "loss": 0.0909, "step": 23710 }, { "epoch": 0.55, "learning_rate": 2.60040247823332e-05, "loss": 0.1333, "step": 23715 }, { "epoch": 0.55, "learning_rate": 2.6003144727887167e-05, "loss": 0.0717, "step": 23720 }, { "epoch": 0.55, "learning_rate": 2.6002264673441132e-05, "loss": 0.1029, "step": 23725 }, { "epoch": 0.55, "learning_rate": 2.6001384618995094e-05, "loss": 0.1256, "step": 23730 }, { "epoch": 0.55, "learning_rate": 2.600050456454906e-05, "loss": 0.2693, "step": 23735 }, { "epoch": 0.55, "learning_rate": 2.5999624510103027e-05, "loss": 0.2518, "step": 23740 }, { "epoch": 0.55, "learning_rate": 2.5998744455656992e-05, "loss": 0.3253, "step": 23745 }, { "epoch": 0.55, "learning_rate": 2.5997864401210954e-05, "loss": 0.8365, "step": 23750 }, { "epoch": 0.55, "learning_rate": 2.599698434676492e-05, "loss": 0.1767, "step": 23755 }, { "epoch": 0.55, "learning_rate": 2.5996104292318887e-05, "loss": 0.0553, "step": 23760 }, { "epoch": 0.55, "learning_rate": 2.599522423787285e-05, "loss": 0.0973, "step": 23765 }, { "epoch": 0.55, "learning_rate": 2.5994344183426814e-05, "loss": 0.1991, "step": 23770 }, { "epoch": 0.55, "learning_rate": 2.599346412898078e-05, "loss": 0.0551, "step": 23775 }, { "epoch": 0.55, "learning_rate": 2.5992584074534748e-05, "loss": 0.2884, "step": 23780 }, { "epoch": 0.55, "learning_rate": 2.599170402008871e-05, "loss": 0.16, "step": 23785 }, { "epoch": 0.56, "learning_rate": 2.5990823965642674e-05, "loss": 0.2933, "step": 23790 }, { "epoch": 0.56, "learning_rate": 2.598994391119664e-05, "loss": 0.1992, "step": 23795 }, { "epoch": 0.56, "learning_rate": 2.5989063856750604e-05, "loss": 0.578, "step": 23800 }, { "epoch": 0.56, "learning_rate": 2.598818380230457e-05, "loss": 0.1634, "step": 23805 }, { "epoch": 0.56, "learning_rate": 2.5987303747858534e-05, "loss": 0.0596, "step": 23810 }, { "epoch": 0.56, "learning_rate": 2.59864236934125e-05, "loss": 0.0527, "step": 23815 }, { "epoch": 0.56, "learning_rate": 2.5985543638966464e-05, "loss": 0.1927, "step": 23820 }, { "epoch": 0.56, "learning_rate": 2.598466358452043e-05, "loss": 0.1967, "step": 23825 }, { "epoch": 0.56, "learning_rate": 2.5983783530074394e-05, "loss": 0.2249, "step": 23830 }, { "epoch": 0.56, "learning_rate": 2.598290347562836e-05, "loss": 0.1467, "step": 23835 }, { "epoch": 0.56, "learning_rate": 2.5982023421182324e-05, "loss": 0.1827, "step": 23840 }, { "epoch": 0.56, "learning_rate": 2.598114336673629e-05, "loss": 0.2972, "step": 23845 }, { "epoch": 0.56, "learning_rate": 2.5980263312290254e-05, "loss": 0.7812, "step": 23850 }, { "epoch": 0.56, "learning_rate": 2.597938325784422e-05, "loss": 0.3074, "step": 23855 }, { "epoch": 0.56, "learning_rate": 2.5978503203398185e-05, "loss": 0.112, "step": 23860 }, { "epoch": 0.56, "learning_rate": 2.597762314895215e-05, "loss": 0.0577, "step": 23865 }, { "epoch": 0.56, "learning_rate": 2.5976743094506115e-05, "loss": 0.2003, "step": 23870 }, { "epoch": 0.56, "learning_rate": 2.597586304006008e-05, "loss": 0.1844, "step": 23875 }, { "epoch": 0.56, "learning_rate": 2.5974982985614045e-05, "loss": 0.3859, "step": 23880 }, { "epoch": 0.56, "learning_rate": 2.597410293116801e-05, "loss": 0.2585, "step": 23885 }, { "epoch": 0.56, "learning_rate": 2.597322287672197e-05, "loss": 0.1678, "step": 23890 }, { "epoch": 0.56, "learning_rate": 2.597234282227594e-05, "loss": 0.4049, "step": 23895 }, { "epoch": 0.56, "learning_rate": 2.5971462767829905e-05, "loss": 0.5833, "step": 23900 }, { "epoch": 0.56, "learning_rate": 2.597058271338387e-05, "loss": 0.228, "step": 23905 }, { "epoch": 0.56, "learning_rate": 2.596970265893783e-05, "loss": 0.0422, "step": 23910 }, { "epoch": 0.56, "learning_rate": 2.59688226044918e-05, "loss": 0.1054, "step": 23915 }, { "epoch": 0.56, "learning_rate": 2.5967942550045765e-05, "loss": 0.0961, "step": 23920 }, { "epoch": 0.56, "learning_rate": 2.5967062495599726e-05, "loss": 0.2136, "step": 23925 }, { "epoch": 0.56, "learning_rate": 2.596618244115369e-05, "loss": 0.1779, "step": 23930 }, { "epoch": 0.56, "learning_rate": 2.596530238670766e-05, "loss": 0.2595, "step": 23935 }, { "epoch": 0.56, "learning_rate": 2.5964422332261625e-05, "loss": 0.3027, "step": 23940 }, { "epoch": 0.56, "learning_rate": 2.5963542277815586e-05, "loss": 0.278, "step": 23945 }, { "epoch": 0.56, "learning_rate": 2.5962662223369555e-05, "loss": 0.4358, "step": 23950 }, { "epoch": 0.56, "learning_rate": 2.596178216892352e-05, "loss": 0.1963, "step": 23955 }, { "epoch": 0.56, "learning_rate": 2.596090211447748e-05, "loss": 0.1452, "step": 23960 }, { "epoch": 0.56, "learning_rate": 2.5960022060031447e-05, "loss": 0.1059, "step": 23965 }, { "epoch": 0.56, "learning_rate": 2.5959142005585415e-05, "loss": 0.3599, "step": 23970 }, { "epoch": 0.56, "learning_rate": 2.595826195113938e-05, "loss": 0.2074, "step": 23975 }, { "epoch": 0.56, "learning_rate": 2.595738189669334e-05, "loss": 0.1846, "step": 23980 }, { "epoch": 0.56, "learning_rate": 2.5956501842247307e-05, "loss": 0.1564, "step": 23985 }, { "epoch": 0.56, "learning_rate": 2.5955621787801275e-05, "loss": 0.339, "step": 23990 }, { "epoch": 0.56, "learning_rate": 2.5954741733355237e-05, "loss": 0.3704, "step": 23995 }, { "epoch": 0.56, "learning_rate": 2.5953861678909202e-05, "loss": 0.392, "step": 24000 }, { "epoch": 0.56, "learning_rate": 2.5952981624463167e-05, "loss": 0.3416, "step": 24005 }, { "epoch": 0.56, "learning_rate": 2.5952101570017135e-05, "loss": 0.0676, "step": 24010 }, { "epoch": 0.56, "learning_rate": 2.5951221515571097e-05, "loss": 0.1566, "step": 24015 }, { "epoch": 0.56, "learning_rate": 2.5950341461125062e-05, "loss": 0.0487, "step": 24020 }, { "epoch": 0.56, "learning_rate": 2.5949461406679027e-05, "loss": 0.1398, "step": 24025 }, { "epoch": 0.56, "learning_rate": 2.5948581352232992e-05, "loss": 0.2118, "step": 24030 }, { "epoch": 0.56, "learning_rate": 2.5947701297786957e-05, "loss": 0.314, "step": 24035 }, { "epoch": 0.56, "learning_rate": 2.5946821243340922e-05, "loss": 0.2648, "step": 24040 }, { "epoch": 0.56, "learning_rate": 2.5945941188894887e-05, "loss": 0.347, "step": 24045 }, { "epoch": 0.56, "learning_rate": 2.5945061134448852e-05, "loss": 0.3082, "step": 24050 }, { "epoch": 0.56, "learning_rate": 2.5944181080002817e-05, "loss": 0.3024, "step": 24055 }, { "epoch": 0.56, "learning_rate": 2.5943301025556782e-05, "loss": 0.1158, "step": 24060 }, { "epoch": 0.56, "learning_rate": 2.5942420971110744e-05, "loss": 0.1069, "step": 24065 }, { "epoch": 0.56, "learning_rate": 2.5941540916664712e-05, "loss": 0.1938, "step": 24070 }, { "epoch": 0.56, "learning_rate": 2.5940660862218677e-05, "loss": 0.1703, "step": 24075 }, { "epoch": 0.56, "learning_rate": 2.5939780807772642e-05, "loss": 0.1247, "step": 24080 }, { "epoch": 0.56, "learning_rate": 2.5938900753326607e-05, "loss": 0.4362, "step": 24085 }, { "epoch": 0.56, "learning_rate": 2.5938020698880572e-05, "loss": 0.2245, "step": 24090 }, { "epoch": 0.56, "learning_rate": 2.5937140644434537e-05, "loss": 0.6776, "step": 24095 }, { "epoch": 0.56, "learning_rate": 2.59362605899885e-05, "loss": 0.5653, "step": 24100 }, { "epoch": 0.56, "learning_rate": 2.5935380535542467e-05, "loss": 0.2009, "step": 24105 }, { "epoch": 0.56, "learning_rate": 2.5934500481096432e-05, "loss": 0.1651, "step": 24110 }, { "epoch": 0.56, "learning_rate": 2.5933620426650397e-05, "loss": 0.1209, "step": 24115 }, { "epoch": 0.56, "learning_rate": 2.593274037220436e-05, "loss": 0.0523, "step": 24120 }, { "epoch": 0.56, "learning_rate": 2.5931860317758327e-05, "loss": 0.1588, "step": 24125 }, { "epoch": 0.56, "learning_rate": 2.5930980263312292e-05, "loss": 0.1748, "step": 24130 }, { "epoch": 0.56, "learning_rate": 2.5930100208866254e-05, "loss": 0.1765, "step": 24135 }, { "epoch": 0.56, "learning_rate": 2.592922015442022e-05, "loss": 0.2066, "step": 24140 }, { "epoch": 0.56, "learning_rate": 2.5928340099974187e-05, "loss": 0.4352, "step": 24145 }, { "epoch": 0.56, "learning_rate": 2.5927460045528152e-05, "loss": 0.369, "step": 24150 }, { "epoch": 0.56, "learning_rate": 2.5926579991082114e-05, "loss": 0.2257, "step": 24155 }, { "epoch": 0.56, "learning_rate": 2.592569993663608e-05, "loss": 0.0707, "step": 24160 }, { "epoch": 0.56, "learning_rate": 2.5924819882190047e-05, "loss": 0.1956, "step": 24165 }, { "epoch": 0.56, "learning_rate": 2.5923939827744012e-05, "loss": 0.0998, "step": 24170 }, { "epoch": 0.56, "learning_rate": 2.5923059773297974e-05, "loss": 0.0933, "step": 24175 }, { "epoch": 0.56, "learning_rate": 2.592217971885194e-05, "loss": 0.3149, "step": 24180 }, { "epoch": 0.56, "learning_rate": 2.5921299664405907e-05, "loss": 0.4111, "step": 24185 }, { "epoch": 0.56, "learning_rate": 2.592041960995987e-05, "loss": 0.0866, "step": 24190 }, { "epoch": 0.56, "learning_rate": 2.5919539555513834e-05, "loss": 0.305, "step": 24195 }, { "epoch": 0.56, "learning_rate": 2.59186595010678e-05, "loss": 0.4083, "step": 24200 }, { "epoch": 0.56, "learning_rate": 2.5917779446621768e-05, "loss": 0.2837, "step": 24205 }, { "epoch": 0.56, "learning_rate": 2.591689939217573e-05, "loss": 0.1373, "step": 24210 }, { "epoch": 0.56, "learning_rate": 2.5916019337729694e-05, "loss": 0.0723, "step": 24215 }, { "epoch": 0.57, "learning_rate": 2.591513928328366e-05, "loss": 0.15, "step": 24220 }, { "epoch": 0.57, "learning_rate": 2.5914259228837624e-05, "loss": 0.2148, "step": 24225 }, { "epoch": 0.57, "learning_rate": 2.591337917439159e-05, "loss": 0.2847, "step": 24230 }, { "epoch": 0.57, "learning_rate": 2.5912499119945554e-05, "loss": 0.2033, "step": 24235 }, { "epoch": 0.57, "learning_rate": 2.5911619065499523e-05, "loss": 0.4579, "step": 24240 }, { "epoch": 0.57, "learning_rate": 2.5910739011053484e-05, "loss": 0.5922, "step": 24245 }, { "epoch": 0.57, "learning_rate": 2.590985895660745e-05, "loss": 0.4544, "step": 24250 }, { "epoch": 0.57, "learning_rate": 2.5908978902161414e-05, "loss": 0.2719, "step": 24255 }, { "epoch": 0.57, "learning_rate": 2.590809884771538e-05, "loss": 0.0343, "step": 24260 }, { "epoch": 0.57, "learning_rate": 2.5907218793269344e-05, "loss": 0.1566, "step": 24265 }, { "epoch": 0.57, "learning_rate": 2.590633873882331e-05, "loss": 0.1534, "step": 24270 }, { "epoch": 0.57, "learning_rate": 2.5905458684377274e-05, "loss": 0.132, "step": 24275 }, { "epoch": 0.57, "learning_rate": 2.590457862993124e-05, "loss": 0.2452, "step": 24280 }, { "epoch": 0.57, "learning_rate": 2.5903698575485204e-05, "loss": 0.1882, "step": 24285 }, { "epoch": 0.57, "learning_rate": 2.590281852103917e-05, "loss": 0.2477, "step": 24290 }, { "epoch": 0.57, "learning_rate": 2.590193846659313e-05, "loss": 0.4372, "step": 24295 }, { "epoch": 0.57, "learning_rate": 2.59010584121471e-05, "loss": 0.6643, "step": 24300 }, { "epoch": 0.57, "learning_rate": 2.5900178357701065e-05, "loss": 0.2417, "step": 24305 }, { "epoch": 0.57, "learning_rate": 2.589929830325503e-05, "loss": 0.1027, "step": 24310 }, { "epoch": 0.57, "learning_rate": 2.589841824880899e-05, "loss": 0.0822, "step": 24315 }, { "epoch": 0.57, "learning_rate": 2.589753819436296e-05, "loss": 0.1586, "step": 24320 }, { "epoch": 0.57, "learning_rate": 2.5896658139916925e-05, "loss": 0.1332, "step": 24325 }, { "epoch": 0.57, "learning_rate": 2.5895778085470886e-05, "loss": 0.1513, "step": 24330 }, { "epoch": 0.57, "learning_rate": 2.589489803102485e-05, "loss": 0.2659, "step": 24335 }, { "epoch": 0.57, "learning_rate": 2.589401797657882e-05, "loss": 0.4118, "step": 24340 }, { "epoch": 0.57, "learning_rate": 2.5893137922132785e-05, "loss": 0.3001, "step": 24345 }, { "epoch": 0.57, "learning_rate": 2.5892257867686746e-05, "loss": 0.7022, "step": 24350 }, { "epoch": 0.57, "learning_rate": 2.589137781324071e-05, "loss": 0.318, "step": 24355 }, { "epoch": 0.57, "learning_rate": 2.589049775879468e-05, "loss": 0.0452, "step": 24360 }, { "epoch": 0.57, "learning_rate": 2.588961770434864e-05, "loss": 0.2664, "step": 24365 }, { "epoch": 0.57, "learning_rate": 2.5888737649902606e-05, "loss": 0.1607, "step": 24370 }, { "epoch": 0.57, "learning_rate": 2.5887857595456575e-05, "loss": 0.1994, "step": 24375 }, { "epoch": 0.57, "learning_rate": 2.588697754101054e-05, "loss": 0.3666, "step": 24380 }, { "epoch": 0.57, "learning_rate": 2.58860974865645e-05, "loss": 0.2442, "step": 24385 }, { "epoch": 0.57, "learning_rate": 2.5885217432118467e-05, "loss": 0.3868, "step": 24390 }, { "epoch": 0.57, "learning_rate": 2.5884337377672435e-05, "loss": 0.442, "step": 24395 }, { "epoch": 0.57, "learning_rate": 2.5883457323226397e-05, "loss": 0.7006, "step": 24400 }, { "epoch": 0.57, "learning_rate": 2.588257726878036e-05, "loss": 0.3516, "step": 24405 }, { "epoch": 0.57, "learning_rate": 2.5881697214334327e-05, "loss": 0.0928, "step": 24410 }, { "epoch": 0.57, "learning_rate": 2.5880817159888295e-05, "loss": 0.0825, "step": 24415 }, { "epoch": 0.57, "learning_rate": 2.5879937105442257e-05, "loss": 0.3513, "step": 24420 }, { "epoch": 0.57, "learning_rate": 2.587905705099622e-05, "loss": 0.1739, "step": 24425 }, { "epoch": 0.57, "learning_rate": 2.5878176996550187e-05, "loss": 0.1818, "step": 24430 }, { "epoch": 0.57, "learning_rate": 2.5877296942104152e-05, "loss": 0.2207, "step": 24435 }, { "epoch": 0.57, "learning_rate": 2.5876416887658117e-05, "loss": 0.3692, "step": 24440 }, { "epoch": 0.57, "learning_rate": 2.5875536833212082e-05, "loss": 0.3627, "step": 24445 }, { "epoch": 0.57, "learning_rate": 2.5874656778766047e-05, "loss": 0.5704, "step": 24450 }, { "epoch": 0.57, "learning_rate": 2.5873776724320012e-05, "loss": 0.3498, "step": 24455 }, { "epoch": 0.57, "learning_rate": 2.5872896669873977e-05, "loss": 0.1153, "step": 24460 }, { "epoch": 0.57, "learning_rate": 2.5872016615427942e-05, "loss": 0.1754, "step": 24465 }, { "epoch": 0.57, "learning_rate": 2.5871136560981907e-05, "loss": 0.1074, "step": 24470 }, { "epoch": 0.57, "learning_rate": 2.5870256506535872e-05, "loss": 0.1849, "step": 24475 }, { "epoch": 0.57, "learning_rate": 2.5869376452089837e-05, "loss": 0.203, "step": 24480 }, { "epoch": 0.57, "learning_rate": 2.5868496397643802e-05, "loss": 0.2067, "step": 24485 }, { "epoch": 0.57, "learning_rate": 2.5867616343197767e-05, "loss": 0.273, "step": 24490 }, { "epoch": 0.57, "learning_rate": 2.5866736288751732e-05, "loss": 0.4133, "step": 24495 }, { "epoch": 0.57, "learning_rate": 2.5865856234305697e-05, "loss": 0.4573, "step": 24500 }, { "epoch": 0.57, "learning_rate": 2.5864976179859662e-05, "loss": 0.2513, "step": 24505 }, { "epoch": 0.57, "learning_rate": 2.5864096125413627e-05, "loss": 0.1753, "step": 24510 }, { "epoch": 0.57, "learning_rate": 2.5863216070967592e-05, "loss": 0.1775, "step": 24515 }, { "epoch": 0.57, "learning_rate": 2.5862336016521557e-05, "loss": 0.1809, "step": 24520 }, { "epoch": 0.57, "learning_rate": 2.586145596207552e-05, "loss": 0.2163, "step": 24525 }, { "epoch": 0.57, "learning_rate": 2.5860575907629487e-05, "loss": 0.2139, "step": 24530 }, { "epoch": 0.57, "learning_rate": 2.5859695853183452e-05, "loss": 0.1888, "step": 24535 }, { "epoch": 0.57, "learning_rate": 2.5858815798737417e-05, "loss": 0.2844, "step": 24540 }, { "epoch": 0.57, "learning_rate": 2.585793574429138e-05, "loss": 0.2297, "step": 24545 }, { "epoch": 0.57, "learning_rate": 2.5857055689845347e-05, "loss": 0.6657, "step": 24550 }, { "epoch": 0.57, "learning_rate": 2.5856175635399312e-05, "loss": 0.3181, "step": 24555 }, { "epoch": 0.57, "learning_rate": 2.5855295580953274e-05, "loss": 0.0959, "step": 24560 }, { "epoch": 0.57, "learning_rate": 2.585441552650724e-05, "loss": 0.1432, "step": 24565 }, { "epoch": 0.57, "learning_rate": 2.5853535472061207e-05, "loss": 0.0991, "step": 24570 }, { "epoch": 0.57, "learning_rate": 2.5852655417615172e-05, "loss": 0.158, "step": 24575 }, { "epoch": 0.57, "learning_rate": 2.5851775363169134e-05, "loss": 0.1152, "step": 24580 }, { "epoch": 0.57, "learning_rate": 2.58508953087231e-05, "loss": 0.2953, "step": 24585 }, { "epoch": 0.57, "learning_rate": 2.5850015254277067e-05, "loss": 0.2833, "step": 24590 }, { "epoch": 0.57, "learning_rate": 2.584913519983103e-05, "loss": 0.474, "step": 24595 }, { "epoch": 0.57, "learning_rate": 2.5848255145384994e-05, "loss": 0.5726, "step": 24600 }, { "epoch": 0.57, "learning_rate": 2.584737509093896e-05, "loss": 0.2596, "step": 24605 }, { "epoch": 0.57, "learning_rate": 2.5846495036492927e-05, "loss": 0.0715, "step": 24610 }, { "epoch": 0.57, "learning_rate": 2.584561498204689e-05, "loss": 0.1706, "step": 24615 }, { "epoch": 0.57, "learning_rate": 2.5844734927600854e-05, "loss": 0.0925, "step": 24620 }, { "epoch": 0.57, "learning_rate": 2.584385487315482e-05, "loss": 0.1506, "step": 24625 }, { "epoch": 0.57, "learning_rate": 2.5842974818708784e-05, "loss": 0.2199, "step": 24630 }, { "epoch": 0.57, "learning_rate": 2.584209476426275e-05, "loss": 0.2233, "step": 24635 }, { "epoch": 0.57, "learning_rate": 2.5841214709816714e-05, "loss": 0.3159, "step": 24640 }, { "epoch": 0.57, "learning_rate": 2.5840334655370683e-05, "loss": 0.4421, "step": 24645 }, { "epoch": 0.58, "learning_rate": 2.5839454600924644e-05, "loss": 0.6839, "step": 24650 }, { "epoch": 0.58, "learning_rate": 2.583857454647861e-05, "loss": 0.2469, "step": 24655 }, { "epoch": 0.58, "learning_rate": 2.5837694492032574e-05, "loss": 0.0692, "step": 24660 }, { "epoch": 0.58, "learning_rate": 2.583681443758654e-05, "loss": 0.1039, "step": 24665 }, { "epoch": 0.58, "learning_rate": 2.5835934383140504e-05, "loss": 0.1597, "step": 24670 }, { "epoch": 0.58, "learning_rate": 2.583505432869447e-05, "loss": 0.1824, "step": 24675 }, { "epoch": 0.58, "learning_rate": 2.5834174274248434e-05, "loss": 0.1036, "step": 24680 }, { "epoch": 0.58, "learning_rate": 2.58332942198024e-05, "loss": 0.2107, "step": 24685 }, { "epoch": 0.58, "learning_rate": 2.5832414165356364e-05, "loss": 0.2249, "step": 24690 }, { "epoch": 0.58, "learning_rate": 2.583153411091033e-05, "loss": 0.5493, "step": 24695 }, { "epoch": 0.58, "learning_rate": 2.583065405646429e-05, "loss": 0.709, "step": 24700 }, { "epoch": 0.58, "learning_rate": 2.582977400201826e-05, "loss": 0.1681, "step": 24705 }, { "epoch": 0.58, "learning_rate": 2.5828893947572224e-05, "loss": 0.1288, "step": 24710 }, { "epoch": 0.58, "learning_rate": 2.582801389312619e-05, "loss": 0.1854, "step": 24715 }, { "epoch": 0.58, "learning_rate": 2.582713383868015e-05, "loss": 0.0679, "step": 24720 }, { "epoch": 0.58, "learning_rate": 2.582625378423412e-05, "loss": 0.2591, "step": 24725 }, { "epoch": 0.58, "learning_rate": 2.5825373729788085e-05, "loss": 0.1964, "step": 24730 }, { "epoch": 0.58, "learning_rate": 2.5824493675342046e-05, "loss": 0.3027, "step": 24735 }, { "epoch": 0.58, "learning_rate": 2.582361362089601e-05, "loss": 0.2306, "step": 24740 }, { "epoch": 0.58, "learning_rate": 2.5822909577339186e-05, "loss": 0.4484, "step": 24745 }, { "epoch": 0.58, "learning_rate": 2.582202952289315e-05, "loss": 0.4466, "step": 24750 }, { "epoch": 0.58, "learning_rate": 2.5821149468447113e-05, "loss": 0.219, "step": 24755 }, { "epoch": 0.58, "learning_rate": 2.582026941400108e-05, "loss": 0.0474, "step": 24760 }, { "epoch": 0.58, "learning_rate": 2.5819389359555046e-05, "loss": 0.1203, "step": 24765 }, { "epoch": 0.58, "learning_rate": 2.581850930510901e-05, "loss": 0.0397, "step": 24770 }, { "epoch": 0.58, "learning_rate": 2.5817629250662973e-05, "loss": 0.194, "step": 24775 }, { "epoch": 0.58, "learning_rate": 2.581674919621694e-05, "loss": 0.2314, "step": 24780 }, { "epoch": 0.58, "learning_rate": 2.5815869141770906e-05, "loss": 0.3571, "step": 24785 }, { "epoch": 0.58, "learning_rate": 2.5814989087324868e-05, "loss": 0.2221, "step": 24790 }, { "epoch": 0.58, "learning_rate": 2.5814109032878833e-05, "loss": 0.2494, "step": 24795 }, { "epoch": 0.58, "learning_rate": 2.58132289784328e-05, "loss": 0.5422, "step": 24800 }, { "epoch": 0.58, "learning_rate": 2.5812348923986766e-05, "loss": 0.2402, "step": 24805 }, { "epoch": 0.58, "learning_rate": 2.5811468869540728e-05, "loss": 0.084, "step": 24810 }, { "epoch": 0.58, "learning_rate": 2.5810588815094693e-05, "loss": 0.1277, "step": 24815 }, { "epoch": 0.58, "learning_rate": 2.580970876064866e-05, "loss": 0.2288, "step": 24820 }, { "epoch": 0.58, "learning_rate": 2.5808828706202623e-05, "loss": 0.2926, "step": 24825 }, { "epoch": 0.58, "learning_rate": 2.5807948651756588e-05, "loss": 0.2546, "step": 24830 }, { "epoch": 0.58, "learning_rate": 2.5807068597310556e-05, "loss": 0.2618, "step": 24835 }, { "epoch": 0.58, "learning_rate": 2.580618854286452e-05, "loss": 0.264, "step": 24840 }, { "epoch": 0.58, "learning_rate": 2.5805308488418483e-05, "loss": 0.4359, "step": 24845 }, { "epoch": 0.58, "learning_rate": 2.5804428433972448e-05, "loss": 0.5411, "step": 24850 }, { "epoch": 0.58, "learning_rate": 2.5803548379526416e-05, "loss": 0.3118, "step": 24855 }, { "epoch": 0.58, "learning_rate": 2.5802668325080378e-05, "loss": 0.1454, "step": 24860 }, { "epoch": 0.58, "learning_rate": 2.5801788270634343e-05, "loss": 0.1074, "step": 24865 }, { "epoch": 0.58, "learning_rate": 2.5800908216188308e-05, "loss": 0.1342, "step": 24870 }, { "epoch": 0.58, "learning_rate": 2.5800028161742276e-05, "loss": 0.198, "step": 24875 }, { "epoch": 0.58, "learning_rate": 2.5799148107296238e-05, "loss": 0.1342, "step": 24880 }, { "epoch": 0.58, "learning_rate": 2.5798268052850203e-05, "loss": 0.2789, "step": 24885 }, { "epoch": 0.58, "learning_rate": 2.5797387998404168e-05, "loss": 0.241, "step": 24890 }, { "epoch": 0.58, "learning_rate": 2.5796507943958133e-05, "loss": 0.4162, "step": 24895 }, { "epoch": 0.58, "learning_rate": 2.5795627889512098e-05, "loss": 0.3839, "step": 24900 }, { "epoch": 0.58, "learning_rate": 2.5794747835066063e-05, "loss": 0.1911, "step": 24905 }, { "epoch": 0.58, "learning_rate": 2.5793867780620028e-05, "loss": 0.1079, "step": 24910 }, { "epoch": 0.58, "learning_rate": 2.5792987726173993e-05, "loss": 0.1428, "step": 24915 }, { "epoch": 0.58, "learning_rate": 2.5792107671727958e-05, "loss": 0.1555, "step": 24920 }, { "epoch": 0.58, "learning_rate": 2.5791227617281923e-05, "loss": 0.2306, "step": 24925 }, { "epoch": 0.58, "learning_rate": 2.5790347562835885e-05, "loss": 0.1276, "step": 24930 }, { "epoch": 0.58, "learning_rate": 2.5789467508389853e-05, "loss": 0.2801, "step": 24935 }, { "epoch": 0.58, "learning_rate": 2.578858745394382e-05, "loss": 0.3626, "step": 24940 }, { "epoch": 0.58, "learning_rate": 2.5787707399497783e-05, "loss": 0.507, "step": 24945 }, { "epoch": 0.58, "learning_rate": 2.5786827345051745e-05, "loss": 0.5877, "step": 24950 }, { "epoch": 0.58, "learning_rate": 2.5785947290605713e-05, "loss": 0.2834, "step": 24955 }, { "epoch": 0.58, "learning_rate": 2.578506723615968e-05, "loss": 0.0245, "step": 24960 }, { "epoch": 0.58, "learning_rate": 2.5784187181713643e-05, "loss": 0.1027, "step": 24965 }, { "epoch": 0.58, "learning_rate": 2.578330712726761e-05, "loss": 0.1238, "step": 24970 }, { "epoch": 0.58, "learning_rate": 2.5782427072821573e-05, "loss": 0.1072, "step": 24975 }, { "epoch": 0.58, "learning_rate": 2.578154701837554e-05, "loss": 0.2191, "step": 24980 }, { "epoch": 0.58, "learning_rate": 2.57806669639295e-05, "loss": 0.2692, "step": 24985 }, { "epoch": 0.58, "learning_rate": 2.577978690948347e-05, "loss": 0.2752, "step": 24990 }, { "epoch": 0.58, "learning_rate": 2.5778906855037434e-05, "loss": 0.4354, "step": 24995 }, { "epoch": 0.58, "learning_rate": 2.57780268005914e-05, "loss": 0.6298, "step": 25000 }, { "epoch": 0.58, "learning_rate": 2.577714674614536e-05, "loss": 0.3365, "step": 25005 }, { "epoch": 0.58, "learning_rate": 2.577626669169933e-05, "loss": 0.0839, "step": 25010 }, { "epoch": 0.58, "learning_rate": 2.5775386637253294e-05, "loss": 0.2072, "step": 25015 }, { "epoch": 0.58, "learning_rate": 2.5774506582807255e-05, "loss": 0.0704, "step": 25020 }, { "epoch": 0.58, "learning_rate": 2.577362652836122e-05, "loss": 0.1672, "step": 25025 }, { "epoch": 0.58, "learning_rate": 2.577274647391519e-05, "loss": 0.1843, "step": 25030 }, { "epoch": 0.58, "learning_rate": 2.5771866419469154e-05, "loss": 0.2651, "step": 25035 }, { "epoch": 0.58, "learning_rate": 2.5770986365023115e-05, "loss": 0.2604, "step": 25040 }, { "epoch": 0.58, "learning_rate": 2.577010631057708e-05, "loss": 0.3946, "step": 25045 }, { "epoch": 0.58, "learning_rate": 2.576922625613105e-05, "loss": 0.67, "step": 25050 }, { "epoch": 0.58, "learning_rate": 2.576834620168501e-05, "loss": 0.3648, "step": 25055 }, { "epoch": 0.58, "learning_rate": 2.5767466147238975e-05, "loss": 0.1011, "step": 25060 }, { "epoch": 0.58, "learning_rate": 2.576658609279294e-05, "loss": 0.1966, "step": 25065 }, { "epoch": 0.58, "learning_rate": 2.576570603834691e-05, "loss": 0.1261, "step": 25070 }, { "epoch": 0.59, "learning_rate": 2.576482598390087e-05, "loss": 0.1461, "step": 25075 }, { "epoch": 0.59, "learning_rate": 2.5763945929454836e-05, "loss": 0.2033, "step": 25080 }, { "epoch": 0.59, "learning_rate": 2.57630658750088e-05, "loss": 0.2964, "step": 25085 }, { "epoch": 0.59, "learning_rate": 2.5762185820562766e-05, "loss": 0.2602, "step": 25090 }, { "epoch": 0.59, "learning_rate": 2.576130576611673e-05, "loss": 0.3051, "step": 25095 }, { "epoch": 0.59, "learning_rate": 2.5760425711670696e-05, "loss": 0.4699, "step": 25100 }, { "epoch": 0.59, "learning_rate": 2.5759545657224664e-05, "loss": 0.1013, "step": 25105 }, { "epoch": 0.59, "learning_rate": 2.5758665602778626e-05, "loss": 0.1551, "step": 25110 }, { "epoch": 0.59, "learning_rate": 2.575778554833259e-05, "loss": 0.1285, "step": 25115 }, { "epoch": 0.59, "learning_rate": 2.5756905493886556e-05, "loss": 0.1879, "step": 25120 }, { "epoch": 0.59, "learning_rate": 2.575602543944052e-05, "loss": 0.2415, "step": 25125 }, { "epoch": 0.59, "learning_rate": 2.5755145384994486e-05, "loss": 0.1959, "step": 25130 }, { "epoch": 0.59, "learning_rate": 2.575426533054845e-05, "loss": 0.4133, "step": 25135 }, { "epoch": 0.59, "learning_rate": 2.5753385276102416e-05, "loss": 0.29, "step": 25140 }, { "epoch": 0.59, "learning_rate": 2.575250522165638e-05, "loss": 0.2858, "step": 25145 }, { "epoch": 0.59, "learning_rate": 2.5751625167210346e-05, "loss": 0.2968, "step": 25150 }, { "epoch": 0.59, "learning_rate": 2.575074511276431e-05, "loss": 0.1272, "step": 25155 }, { "epoch": 0.59, "learning_rate": 2.5749865058318272e-05, "loss": 0.1179, "step": 25160 }, { "epoch": 0.59, "learning_rate": 2.574898500387224e-05, "loss": 0.0666, "step": 25165 }, { "epoch": 0.59, "learning_rate": 2.5748104949426206e-05, "loss": 0.1648, "step": 25170 }, { "epoch": 0.59, "learning_rate": 2.574722489498017e-05, "loss": 0.1244, "step": 25175 }, { "epoch": 0.59, "learning_rate": 2.5746344840534133e-05, "loss": 0.1498, "step": 25180 }, { "epoch": 0.59, "learning_rate": 2.57454647860881e-05, "loss": 0.2362, "step": 25185 }, { "epoch": 0.59, "learning_rate": 2.5744584731642066e-05, "loss": 0.1523, "step": 25190 }, { "epoch": 0.59, "learning_rate": 2.5743704677196028e-05, "loss": 0.3563, "step": 25195 }, { "epoch": 0.59, "learning_rate": 2.5742824622749993e-05, "loss": 0.3501, "step": 25200 }, { "epoch": 0.59, "learning_rate": 2.574194456830396e-05, "loss": 0.2522, "step": 25205 }, { "epoch": 0.59, "learning_rate": 2.5741064513857926e-05, "loss": 0.1017, "step": 25210 }, { "epoch": 0.59, "learning_rate": 2.5740184459411888e-05, "loss": 0.1108, "step": 25215 }, { "epoch": 0.59, "learning_rate": 2.5739304404965853e-05, "loss": 0.0968, "step": 25220 }, { "epoch": 0.59, "learning_rate": 2.573842435051982e-05, "loss": 0.155, "step": 25225 }, { "epoch": 0.59, "learning_rate": 2.5737544296073783e-05, "loss": 0.1696, "step": 25230 }, { "epoch": 0.59, "learning_rate": 2.5736664241627748e-05, "loss": 0.3376, "step": 25235 }, { "epoch": 0.59, "learning_rate": 2.5735784187181716e-05, "loss": 0.3803, "step": 25240 }, { "epoch": 0.59, "learning_rate": 2.573490413273568e-05, "loss": 0.4867, "step": 25245 }, { "epoch": 0.59, "learning_rate": 2.5734024078289643e-05, "loss": 0.6163, "step": 25250 }, { "epoch": 0.59, "learning_rate": 2.5733144023843608e-05, "loss": 0.1604, "step": 25255 }, { "epoch": 0.59, "learning_rate": 2.5732263969397576e-05, "loss": 0.1573, "step": 25260 }, { "epoch": 0.59, "learning_rate": 2.573138391495154e-05, "loss": 0.0829, "step": 25265 }, { "epoch": 0.59, "learning_rate": 2.5730503860505503e-05, "loss": 0.1026, "step": 25270 }, { "epoch": 0.59, "learning_rate": 2.5729623806059468e-05, "loss": 0.1174, "step": 25275 }, { "epoch": 0.59, "learning_rate": 2.5728743751613436e-05, "loss": 0.1535, "step": 25280 }, { "epoch": 0.59, "learning_rate": 2.5727863697167398e-05, "loss": 0.2601, "step": 25285 }, { "epoch": 0.59, "learning_rate": 2.5726983642721363e-05, "loss": 0.2954, "step": 25290 }, { "epoch": 0.59, "learning_rate": 2.5726103588275328e-05, "loss": 0.3656, "step": 25295 }, { "epoch": 0.59, "learning_rate": 2.5725223533829296e-05, "loss": 0.5345, "step": 25300 }, { "epoch": 0.59, "learning_rate": 2.5724343479383258e-05, "loss": 0.1331, "step": 25305 }, { "epoch": 0.59, "learning_rate": 2.5723463424937223e-05, "loss": 0.0735, "step": 25310 }, { "epoch": 0.59, "learning_rate": 2.5722583370491188e-05, "loss": 0.306, "step": 25315 }, { "epoch": 0.59, "learning_rate": 2.5721703316045153e-05, "loss": 0.0895, "step": 25320 }, { "epoch": 0.59, "learning_rate": 2.5720823261599118e-05, "loss": 0.412, "step": 25325 }, { "epoch": 0.59, "learning_rate": 2.5719943207153083e-05, "loss": 0.1549, "step": 25330 }, { "epoch": 0.59, "learning_rate": 2.5719063152707048e-05, "loss": 0.2575, "step": 25335 }, { "epoch": 0.59, "learning_rate": 2.5718183098261013e-05, "loss": 0.2, "step": 25340 }, { "epoch": 0.59, "learning_rate": 2.5717303043814978e-05, "loss": 0.4852, "step": 25345 }, { "epoch": 0.59, "learning_rate": 2.5716422989368943e-05, "loss": 0.6174, "step": 25350 }, { "epoch": 0.59, "learning_rate": 2.5715542934922905e-05, "loss": 0.3066, "step": 25355 }, { "epoch": 0.59, "learning_rate": 2.5714662880476873e-05, "loss": 0.1316, "step": 25360 }, { "epoch": 0.59, "learning_rate": 2.5713782826030838e-05, "loss": 0.144, "step": 25365 }, { "epoch": 0.59, "learning_rate": 2.5712902771584803e-05, "loss": 0.2674, "step": 25370 }, { "epoch": 0.59, "learning_rate": 2.571202271713877e-05, "loss": 0.1009, "step": 25375 }, { "epoch": 0.59, "learning_rate": 2.5711142662692733e-05, "loss": 0.2775, "step": 25380 }, { "epoch": 0.59, "learning_rate": 2.57102626082467e-05, "loss": 0.2139, "step": 25385 }, { "epoch": 0.59, "learning_rate": 2.570938255380066e-05, "loss": 0.2266, "step": 25390 }, { "epoch": 0.59, "learning_rate": 2.570850249935463e-05, "loss": 0.3109, "step": 25395 }, { "epoch": 0.59, "learning_rate": 2.5707622444908593e-05, "loss": 0.3192, "step": 25400 }, { "epoch": 0.59, "learning_rate": 2.570674239046256e-05, "loss": 0.3013, "step": 25405 }, { "epoch": 0.59, "learning_rate": 2.570586233601652e-05, "loss": 0.1462, "step": 25410 }, { "epoch": 0.59, "learning_rate": 2.570498228157049e-05, "loss": 0.1293, "step": 25415 }, { "epoch": 0.59, "learning_rate": 2.5704102227124454e-05, "loss": 0.2431, "step": 25420 }, { "epoch": 0.59, "learning_rate": 2.5703222172678415e-05, "loss": 0.199, "step": 25425 }, { "epoch": 0.59, "learning_rate": 2.570234211823238e-05, "loss": 0.2834, "step": 25430 }, { "epoch": 0.59, "learning_rate": 2.570146206378635e-05, "loss": 0.1728, "step": 25435 }, { "epoch": 0.59, "learning_rate": 2.5700582009340314e-05, "loss": 0.4968, "step": 25440 }, { "epoch": 0.59, "learning_rate": 2.5699701954894275e-05, "loss": 0.2974, "step": 25445 }, { "epoch": 0.59, "learning_rate": 2.569882190044824e-05, "loss": 0.4382, "step": 25450 }, { "epoch": 0.59, "learning_rate": 2.569794184600221e-05, "loss": 0.2322, "step": 25455 }, { "epoch": 0.59, "learning_rate": 2.569706179155617e-05, "loss": 0.0951, "step": 25460 }, { "epoch": 0.59, "learning_rate": 2.5696181737110135e-05, "loss": 0.119, "step": 25465 }, { "epoch": 0.59, "learning_rate": 2.56953016826641e-05, "loss": 0.1574, "step": 25470 }, { "epoch": 0.59, "learning_rate": 2.569442162821807e-05, "loss": 0.1782, "step": 25475 }, { "epoch": 0.59, "learning_rate": 2.569354157377203e-05, "loss": 0.2572, "step": 25480 }, { "epoch": 0.59, "learning_rate": 2.5692661519325995e-05, "loss": 0.2812, "step": 25485 }, { "epoch": 0.59, "learning_rate": 2.569178146487996e-05, "loss": 0.2472, "step": 25490 }, { "epoch": 0.59, "learning_rate": 2.5690901410433925e-05, "loss": 0.4015, "step": 25495 }, { "epoch": 0.59, "learning_rate": 2.569002135598789e-05, "loss": 0.5159, "step": 25500 }, { "epoch": 0.6, "learning_rate": 2.5689141301541855e-05, "loss": 0.2866, "step": 25505 }, { "epoch": 0.6, "learning_rate": 2.5688261247095824e-05, "loss": 0.115, "step": 25510 }, { "epoch": 0.6, "learning_rate": 2.5687381192649786e-05, "loss": 0.1712, "step": 25515 }, { "epoch": 0.6, "learning_rate": 2.568650113820375e-05, "loss": 0.1287, "step": 25520 }, { "epoch": 0.6, "learning_rate": 2.5685621083757716e-05, "loss": 0.203, "step": 25525 }, { "epoch": 0.6, "learning_rate": 2.568474102931168e-05, "loss": 0.2496, "step": 25530 }, { "epoch": 0.6, "learning_rate": 2.5683860974865646e-05, "loss": 0.2533, "step": 25535 }, { "epoch": 0.6, "learning_rate": 2.568298092041961e-05, "loss": 0.3372, "step": 25540 }, { "epoch": 0.6, "learning_rate": 2.5682100865973576e-05, "loss": 0.2836, "step": 25545 }, { "epoch": 0.6, "learning_rate": 2.568122081152754e-05, "loss": 0.6303, "step": 25550 }, { "epoch": 0.6, "learning_rate": 2.5680340757081506e-05, "loss": 0.2521, "step": 25555 }, { "epoch": 0.6, "learning_rate": 2.567946070263547e-05, "loss": 0.1743, "step": 25560 }, { "epoch": 0.6, "learning_rate": 2.5678580648189436e-05, "loss": 0.0943, "step": 25565 }, { "epoch": 0.6, "learning_rate": 2.56777005937434e-05, "loss": 0.1089, "step": 25570 }, { "epoch": 0.6, "learning_rate": 2.5676820539297366e-05, "loss": 0.1464, "step": 25575 }, { "epoch": 0.6, "learning_rate": 2.567594048485133e-05, "loss": 0.1535, "step": 25580 }, { "epoch": 0.6, "learning_rate": 2.5675060430405292e-05, "loss": 0.2347, "step": 25585 }, { "epoch": 0.6, "learning_rate": 2.567418037595926e-05, "loss": 0.5172, "step": 25590 }, { "epoch": 0.6, "learning_rate": 2.5673300321513226e-05, "loss": 0.2499, "step": 25595 }, { "epoch": 0.6, "learning_rate": 2.567242026706719e-05, "loss": 1.0518, "step": 25600 }, { "epoch": 0.6, "learning_rate": 2.5671540212621152e-05, "loss": 0.2293, "step": 25605 }, { "epoch": 0.6, "learning_rate": 2.567066015817512e-05, "loss": 0.1027, "step": 25610 }, { "epoch": 0.6, "learning_rate": 2.5669780103729086e-05, "loss": 0.1185, "step": 25615 }, { "epoch": 0.6, "learning_rate": 2.5668900049283048e-05, "loss": 0.172, "step": 25620 }, { "epoch": 0.6, "learning_rate": 2.5668019994837013e-05, "loss": 0.1958, "step": 25625 }, { "epoch": 0.6, "learning_rate": 2.566713994039098e-05, "loss": 0.117, "step": 25630 }, { "epoch": 0.6, "learning_rate": 2.5666259885944946e-05, "loss": 0.4191, "step": 25635 }, { "epoch": 0.6, "learning_rate": 2.5665379831498908e-05, "loss": 0.2673, "step": 25640 }, { "epoch": 0.6, "learning_rate": 2.5664499777052876e-05, "loss": 0.5506, "step": 25645 }, { "epoch": 0.6, "learning_rate": 2.566361972260684e-05, "loss": 0.5789, "step": 25650 }, { "epoch": 0.6, "learning_rate": 2.5662739668160803e-05, "loss": 0.1906, "step": 25655 }, { "epoch": 0.6, "learning_rate": 2.5661859613714768e-05, "loss": 0.1339, "step": 25660 }, { "epoch": 0.6, "learning_rate": 2.5660979559268736e-05, "loss": 0.0873, "step": 25665 }, { "epoch": 0.6, "learning_rate": 2.56600995048227e-05, "loss": 0.0683, "step": 25670 }, { "epoch": 0.6, "learning_rate": 2.5659219450376663e-05, "loss": 0.2135, "step": 25675 }, { "epoch": 0.6, "learning_rate": 2.5658339395930628e-05, "loss": 0.13, "step": 25680 }, { "epoch": 0.6, "learning_rate": 2.5657459341484596e-05, "loss": 0.3055, "step": 25685 }, { "epoch": 0.6, "learning_rate": 2.5656579287038558e-05, "loss": 0.2751, "step": 25690 }, { "epoch": 0.6, "learning_rate": 2.5655699232592523e-05, "loss": 0.2838, "step": 25695 }, { "epoch": 0.6, "learning_rate": 2.5654819178146488e-05, "loss": 0.3905, "step": 25700 }, { "epoch": 0.6, "learning_rate": 2.5653939123700456e-05, "loss": 0.3494, "step": 25705 }, { "epoch": 0.6, "learning_rate": 2.5653059069254418e-05, "loss": 0.1301, "step": 25710 }, { "epoch": 0.6, "learning_rate": 2.5652179014808383e-05, "loss": 0.0784, "step": 25715 }, { "epoch": 0.6, "learning_rate": 2.5651298960362348e-05, "loss": 0.229, "step": 25720 }, { "epoch": 0.6, "learning_rate": 2.5650418905916313e-05, "loss": 0.2491, "step": 25725 }, { "epoch": 0.6, "learning_rate": 2.5649538851470278e-05, "loss": 0.2165, "step": 25730 }, { "epoch": 0.6, "learning_rate": 2.5648658797024243e-05, "loss": 0.242, "step": 25735 }, { "epoch": 0.6, "learning_rate": 2.5647778742578208e-05, "loss": 0.3224, "step": 25740 }, { "epoch": 0.6, "learning_rate": 2.5646898688132173e-05, "loss": 0.3286, "step": 25745 }, { "epoch": 0.6, "learning_rate": 2.5646018633686138e-05, "loss": 0.3097, "step": 25750 }, { "epoch": 0.6, "learning_rate": 2.5645138579240103e-05, "loss": 0.2442, "step": 25755 }, { "epoch": 0.6, "learning_rate": 2.5644258524794065e-05, "loss": 0.0928, "step": 25760 }, { "epoch": 0.6, "learning_rate": 2.5643378470348033e-05, "loss": 0.1962, "step": 25765 }, { "epoch": 0.6, "learning_rate": 2.5642498415901998e-05, "loss": 0.115, "step": 25770 }, { "epoch": 0.6, "learning_rate": 2.5641618361455963e-05, "loss": 0.1555, "step": 25775 }, { "epoch": 0.6, "learning_rate": 2.5640738307009928e-05, "loss": 0.2135, "step": 25780 }, { "epoch": 0.6, "learning_rate": 2.5639858252563893e-05, "loss": 0.2689, "step": 25785 }, { "epoch": 0.6, "learning_rate": 2.5638978198117858e-05, "loss": 0.1973, "step": 25790 }, { "epoch": 0.6, "learning_rate": 2.563809814367182e-05, "loss": 0.2675, "step": 25795 }, { "epoch": 0.6, "learning_rate": 2.5637218089225788e-05, "loss": 1.0073, "step": 25800 }, { "epoch": 0.6, "learning_rate": 2.5636338034779753e-05, "loss": 0.2535, "step": 25805 }, { "epoch": 0.6, "learning_rate": 2.563545798033372e-05, "loss": 0.0974, "step": 25810 }, { "epoch": 0.6, "learning_rate": 2.563457792588768e-05, "loss": 0.1069, "step": 25815 }, { "epoch": 0.6, "learning_rate": 2.563369787144165e-05, "loss": 0.1962, "step": 25820 }, { "epoch": 0.6, "learning_rate": 2.5632817816995613e-05, "loss": 0.2316, "step": 25825 }, { "epoch": 0.6, "learning_rate": 2.5631937762549575e-05, "loss": 0.1597, "step": 25830 }, { "epoch": 0.6, "learning_rate": 2.563105770810354e-05, "loss": 0.291, "step": 25835 }, { "epoch": 0.6, "learning_rate": 2.563017765365751e-05, "loss": 0.3037, "step": 25840 }, { "epoch": 0.6, "learning_rate": 2.5629297599211473e-05, "loss": 0.4037, "step": 25845 }, { "epoch": 0.6, "learning_rate": 2.5628417544765435e-05, "loss": 0.8024, "step": 25850 }, { "epoch": 0.6, "learning_rate": 2.56275374903194e-05, "loss": 0.2129, "step": 25855 }, { "epoch": 0.6, "learning_rate": 2.562665743587337e-05, "loss": 0.1376, "step": 25860 }, { "epoch": 0.6, "learning_rate": 2.5625777381427334e-05, "loss": 0.134, "step": 25865 }, { "epoch": 0.6, "learning_rate": 2.5624897326981295e-05, "loss": 0.2244, "step": 25870 }, { "epoch": 0.6, "learning_rate": 2.562401727253526e-05, "loss": 0.1618, "step": 25875 }, { "epoch": 0.6, "learning_rate": 2.562313721808923e-05, "loss": 0.2012, "step": 25880 }, { "epoch": 0.6, "learning_rate": 2.562225716364319e-05, "loss": 0.1602, "step": 25885 }, { "epoch": 0.6, "learning_rate": 2.5621377109197155e-05, "loss": 0.2866, "step": 25890 }, { "epoch": 0.6, "learning_rate": 2.562049705475112e-05, "loss": 0.4362, "step": 25895 }, { "epoch": 0.6, "learning_rate": 2.561961700030509e-05, "loss": 0.5573, "step": 25900 }, { "epoch": 0.6, "learning_rate": 2.561873694585905e-05, "loss": 0.1796, "step": 25905 }, { "epoch": 0.6, "learning_rate": 2.5617856891413015e-05, "loss": 0.0765, "step": 25910 }, { "epoch": 0.6, "learning_rate": 2.5616976836966984e-05, "loss": 0.2804, "step": 25915 }, { "epoch": 0.6, "learning_rate": 2.5616096782520945e-05, "loss": 0.057, "step": 25920 }, { "epoch": 0.6, "learning_rate": 2.561521672807491e-05, "loss": 0.0901, "step": 25925 }, { "epoch": 0.6, "learning_rate": 2.5614336673628875e-05, "loss": 0.2712, "step": 25930 }, { "epoch": 0.61, "learning_rate": 2.5613456619182844e-05, "loss": 0.2664, "step": 25935 }, { "epoch": 0.61, "learning_rate": 2.5612576564736805e-05, "loss": 0.1563, "step": 25940 }, { "epoch": 0.61, "learning_rate": 2.561169651029077e-05, "loss": 0.4021, "step": 25945 }, { "epoch": 0.61, "learning_rate": 2.5610816455844736e-05, "loss": 0.7023, "step": 25950 }, { "epoch": 0.61, "learning_rate": 2.56099364013987e-05, "loss": 0.2665, "step": 25955 }, { "epoch": 0.61, "learning_rate": 2.5609056346952666e-05, "loss": 0.1541, "step": 25960 }, { "epoch": 0.61, "learning_rate": 2.560817629250663e-05, "loss": 0.1411, "step": 25965 }, { "epoch": 0.61, "learning_rate": 2.5607296238060596e-05, "loss": 0.0653, "step": 25970 }, { "epoch": 0.61, "learning_rate": 2.560641618361456e-05, "loss": 0.0816, "step": 25975 }, { "epoch": 0.61, "learning_rate": 2.5605536129168526e-05, "loss": 0.2391, "step": 25980 }, { "epoch": 0.61, "learning_rate": 2.560465607472249e-05, "loss": 0.3782, "step": 25985 }, { "epoch": 0.61, "learning_rate": 2.5603776020276452e-05, "loss": 0.2558, "step": 25990 }, { "epoch": 0.61, "learning_rate": 2.560289596583042e-05, "loss": 0.2696, "step": 25995 }, { "epoch": 0.61, "learning_rate": 2.5602015911384386e-05, "loss": 0.5442, "step": 26000 }, { "epoch": 0.61, "learning_rate": 2.560113585693835e-05, "loss": 0.2488, "step": 26005 }, { "epoch": 0.61, "learning_rate": 2.5600255802492312e-05, "loss": 0.1703, "step": 26010 }, { "epoch": 0.61, "learning_rate": 2.559937574804628e-05, "loss": 0.1596, "step": 26015 }, { "epoch": 0.61, "learning_rate": 2.5598495693600246e-05, "loss": 0.1339, "step": 26020 }, { "epoch": 0.61, "learning_rate": 2.5597615639154207e-05, "loss": 0.1338, "step": 26025 }, { "epoch": 0.61, "learning_rate": 2.5596735584708172e-05, "loss": 0.0652, "step": 26030 }, { "epoch": 0.61, "learning_rate": 2.559585553026214e-05, "loss": 0.1862, "step": 26035 }, { "epoch": 0.61, "learning_rate": 2.5594975475816106e-05, "loss": 0.268, "step": 26040 }, { "epoch": 0.61, "learning_rate": 2.5594095421370068e-05, "loss": 0.3124, "step": 26045 }, { "epoch": 0.61, "learning_rate": 2.5593215366924036e-05, "loss": 0.43, "step": 26050 }, { "epoch": 0.61, "learning_rate": 2.5592335312478e-05, "loss": 0.3415, "step": 26055 }, { "epoch": 0.61, "learning_rate": 2.5591455258031963e-05, "loss": 0.0529, "step": 26060 }, { "epoch": 0.61, "learning_rate": 2.5590575203585928e-05, "loss": 0.1851, "step": 26065 }, { "epoch": 0.61, "learning_rate": 2.5589695149139896e-05, "loss": 0.0506, "step": 26070 }, { "epoch": 0.61, "learning_rate": 2.558881509469386e-05, "loss": 0.1921, "step": 26075 }, { "epoch": 0.61, "learning_rate": 2.5587935040247823e-05, "loss": 0.1254, "step": 26080 }, { "epoch": 0.61, "learning_rate": 2.5587054985801788e-05, "loss": 0.3445, "step": 26085 }, { "epoch": 0.61, "learning_rate": 2.5586174931355756e-05, "loss": 0.2669, "step": 26090 }, { "epoch": 0.61, "learning_rate": 2.5585294876909718e-05, "loss": 0.4049, "step": 26095 }, { "epoch": 0.61, "learning_rate": 2.5584414822463683e-05, "loss": 0.4844, "step": 26100 }, { "epoch": 0.61, "learning_rate": 2.5583534768017648e-05, "loss": 0.1996, "step": 26105 }, { "epoch": 0.61, "learning_rate": 2.5582654713571616e-05, "loss": 0.0802, "step": 26110 }, { "epoch": 0.61, "learning_rate": 2.5581774659125578e-05, "loss": 0.2169, "step": 26115 }, { "epoch": 0.61, "learning_rate": 2.5580894604679543e-05, "loss": 0.145, "step": 26120 }, { "epoch": 0.61, "learning_rate": 2.5580014550233508e-05, "loss": 0.2198, "step": 26125 }, { "epoch": 0.61, "learning_rate": 2.5579134495787473e-05, "loss": 0.1319, "step": 26130 }, { "epoch": 0.61, "learning_rate": 2.5578254441341438e-05, "loss": 0.2251, "step": 26135 }, { "epoch": 0.61, "learning_rate": 2.5577374386895403e-05, "loss": 0.1937, "step": 26140 }, { "epoch": 0.61, "learning_rate": 2.5576494332449368e-05, "loss": 0.3729, "step": 26145 }, { "epoch": 0.61, "learning_rate": 2.5575614278003333e-05, "loss": 0.6456, "step": 26150 }, { "epoch": 0.61, "learning_rate": 2.5574734223557298e-05, "loss": 0.2623, "step": 26155 }, { "epoch": 0.61, "learning_rate": 2.5573854169111263e-05, "loss": 0.1104, "step": 26160 }, { "epoch": 0.61, "learning_rate": 2.5572974114665225e-05, "loss": 0.0985, "step": 26165 }, { "epoch": 0.61, "learning_rate": 2.5572094060219193e-05, "loss": 0.0707, "step": 26170 }, { "epoch": 0.61, "learning_rate": 2.5571214005773158e-05, "loss": 0.081, "step": 26175 }, { "epoch": 0.61, "learning_rate": 2.5570333951327123e-05, "loss": 0.1495, "step": 26180 }, { "epoch": 0.61, "learning_rate": 2.5569453896881088e-05, "loss": 0.1116, "step": 26185 }, { "epoch": 0.61, "learning_rate": 2.5568573842435053e-05, "loss": 0.1909, "step": 26190 }, { "epoch": 0.61, "learning_rate": 2.5567693787989018e-05, "loss": 0.3603, "step": 26195 }, { "epoch": 0.61, "learning_rate": 2.5566813733542983e-05, "loss": 0.5437, "step": 26200 }, { "epoch": 0.61, "learning_rate": 2.5565933679096948e-05, "loss": 0.2998, "step": 26205 }, { "epoch": 0.61, "learning_rate": 2.5565053624650913e-05, "loss": 0.0687, "step": 26210 }, { "epoch": 0.61, "learning_rate": 2.5564173570204878e-05, "loss": 0.0871, "step": 26215 }, { "epoch": 0.61, "learning_rate": 2.556329351575884e-05, "loss": 0.1284, "step": 26220 }, { "epoch": 0.61, "learning_rate": 2.5562413461312808e-05, "loss": 0.1857, "step": 26225 }, { "epoch": 0.61, "learning_rate": 2.5561533406866773e-05, "loss": 0.1082, "step": 26230 }, { "epoch": 0.61, "learning_rate": 2.556065335242074e-05, "loss": 0.2527, "step": 26235 }, { "epoch": 0.61, "learning_rate": 2.55597732979747e-05, "loss": 0.4088, "step": 26240 }, { "epoch": 0.61, "learning_rate": 2.555889324352867e-05, "loss": 0.3394, "step": 26245 }, { "epoch": 0.61, "learning_rate": 2.5558013189082633e-05, "loss": 0.7346, "step": 26250 }, { "epoch": 0.61, "learning_rate": 2.5557133134636595e-05, "loss": 0.1863, "step": 26255 }, { "epoch": 0.61, "learning_rate": 2.555625308019056e-05, "loss": 0.0615, "step": 26260 }, { "epoch": 0.61, "learning_rate": 2.555537302574453e-05, "loss": 0.0948, "step": 26265 }, { "epoch": 0.61, "learning_rate": 2.5554492971298493e-05, "loss": 0.2368, "step": 26270 }, { "epoch": 0.61, "learning_rate": 2.5553612916852455e-05, "loss": 0.1386, "step": 26275 }, { "epoch": 0.61, "learning_rate": 2.555273286240642e-05, "loss": 0.1367, "step": 26280 }, { "epoch": 0.61, "learning_rate": 2.555185280796039e-05, "loss": 0.1526, "step": 26285 }, { "epoch": 0.61, "learning_rate": 2.555097275351435e-05, "loss": 0.3577, "step": 26290 }, { "epoch": 0.61, "learning_rate": 2.5550092699068315e-05, "loss": 0.2669, "step": 26295 }, { "epoch": 0.61, "learning_rate": 2.554921264462228e-05, "loss": 0.5131, "step": 26300 }, { "epoch": 0.61, "learning_rate": 2.554833259017625e-05, "loss": 0.2392, "step": 26305 }, { "epoch": 0.61, "learning_rate": 2.554745253573021e-05, "loss": 0.0895, "step": 26310 }, { "epoch": 0.61, "learning_rate": 2.5546572481284175e-05, "loss": 0.2309, "step": 26315 }, { "epoch": 0.61, "learning_rate": 2.5545692426838144e-05, "loss": 0.1856, "step": 26320 }, { "epoch": 0.61, "learning_rate": 2.5544812372392105e-05, "loss": 0.1456, "step": 26325 }, { "epoch": 0.61, "learning_rate": 2.554393231794607e-05, "loss": 0.1864, "step": 26330 }, { "epoch": 0.61, "learning_rate": 2.5543052263500035e-05, "loss": 0.2433, "step": 26335 }, { "epoch": 0.61, "learning_rate": 2.5542172209054004e-05, "loss": 0.3256, "step": 26340 }, { "epoch": 0.61, "learning_rate": 2.5541292154607965e-05, "loss": 0.3478, "step": 26345 }, { "epoch": 0.61, "learning_rate": 2.554041210016193e-05, "loss": 0.484, "step": 26350 }, { "epoch": 0.61, "learning_rate": 2.5539532045715895e-05, "loss": 0.1353, "step": 26355 }, { "epoch": 0.62, "learning_rate": 2.553865199126986e-05, "loss": 0.3335, "step": 26360 }, { "epoch": 0.62, "learning_rate": 2.5537771936823825e-05, "loss": 0.1764, "step": 26365 }, { "epoch": 0.62, "learning_rate": 2.553689188237779e-05, "loss": 0.1751, "step": 26370 }, { "epoch": 0.62, "learning_rate": 2.5536011827931755e-05, "loss": 0.2917, "step": 26375 }, { "epoch": 0.62, "learning_rate": 2.553513177348572e-05, "loss": 0.1298, "step": 26380 }, { "epoch": 0.62, "learning_rate": 2.5534251719039686e-05, "loss": 0.1459, "step": 26385 }, { "epoch": 0.62, "learning_rate": 2.553337166459365e-05, "loss": 0.2033, "step": 26390 }, { "epoch": 0.62, "learning_rate": 2.5532491610147612e-05, "loss": 0.3957, "step": 26395 }, { "epoch": 0.62, "learning_rate": 2.553161155570158e-05, "loss": 0.5228, "step": 26400 }, { "epoch": 0.62, "learning_rate": 2.5530731501255546e-05, "loss": 0.1677, "step": 26405 }, { "epoch": 0.62, "learning_rate": 2.552985144680951e-05, "loss": 0.055, "step": 26410 }, { "epoch": 0.62, "learning_rate": 2.5528971392363472e-05, "loss": 0.1565, "step": 26415 }, { "epoch": 0.62, "learning_rate": 2.552809133791744e-05, "loss": 0.1692, "step": 26420 }, { "epoch": 0.62, "learning_rate": 2.5527211283471406e-05, "loss": 0.1312, "step": 26425 }, { "epoch": 0.62, "learning_rate": 2.5526331229025367e-05, "loss": 0.2934, "step": 26430 }, { "epoch": 0.62, "learning_rate": 2.5525451174579332e-05, "loss": 0.264, "step": 26435 }, { "epoch": 0.62, "learning_rate": 2.55245711201333e-05, "loss": 0.4485, "step": 26440 }, { "epoch": 0.62, "learning_rate": 2.5523691065687266e-05, "loss": 0.3406, "step": 26445 }, { "epoch": 0.62, "learning_rate": 2.5522811011241227e-05, "loss": 0.4867, "step": 26450 }, { "epoch": 0.62, "learning_rate": 2.5521930956795196e-05, "loss": 0.3077, "step": 26455 }, { "epoch": 0.62, "learning_rate": 2.552105090234916e-05, "loss": 0.0418, "step": 26460 }, { "epoch": 0.62, "learning_rate": 2.5520170847903122e-05, "loss": 0.1145, "step": 26465 }, { "epoch": 0.62, "learning_rate": 2.5519290793457087e-05, "loss": 0.0999, "step": 26470 }, { "epoch": 0.62, "learning_rate": 2.5518410739011056e-05, "loss": 0.0765, "step": 26475 }, { "epoch": 0.62, "learning_rate": 2.551753068456502e-05, "loss": 0.1315, "step": 26480 }, { "epoch": 0.62, "learning_rate": 2.5516650630118983e-05, "loss": 0.2451, "step": 26485 }, { "epoch": 0.62, "learning_rate": 2.5515770575672948e-05, "loss": 0.2367, "step": 26490 }, { "epoch": 0.62, "learning_rate": 2.5514890521226916e-05, "loss": 0.3576, "step": 26495 }, { "epoch": 0.62, "learning_rate": 2.551401046678088e-05, "loss": 0.5294, "step": 26500 }, { "epoch": 0.62, "learning_rate": 2.5513130412334843e-05, "loss": 0.2614, "step": 26505 }, { "epoch": 0.62, "learning_rate": 2.5512250357888808e-05, "loss": 0.0813, "step": 26510 }, { "epoch": 0.62, "learning_rate": 2.5511370303442776e-05, "loss": 0.1315, "step": 26515 }, { "epoch": 0.62, "learning_rate": 2.5510490248996738e-05, "loss": 0.191, "step": 26520 }, { "epoch": 0.62, "learning_rate": 2.5509610194550703e-05, "loss": 0.1902, "step": 26525 }, { "epoch": 0.62, "learning_rate": 2.5508730140104668e-05, "loss": 0.2337, "step": 26530 }, { "epoch": 0.62, "learning_rate": 2.5507850085658636e-05, "loss": 0.2323, "step": 26535 }, { "epoch": 0.62, "learning_rate": 2.5506970031212598e-05, "loss": 0.3818, "step": 26540 }, { "epoch": 0.62, "learning_rate": 2.5506089976766563e-05, "loss": 0.4075, "step": 26545 }, { "epoch": 0.62, "learning_rate": 2.5505209922320528e-05, "loss": 0.6772, "step": 26550 }, { "epoch": 0.62, "learning_rate": 2.5504329867874493e-05, "loss": 0.3902, "step": 26555 }, { "epoch": 0.62, "learning_rate": 2.5503449813428458e-05, "loss": 0.115, "step": 26560 }, { "epoch": 0.62, "learning_rate": 2.5502569758982423e-05, "loss": 0.1331, "step": 26565 }, { "epoch": 0.62, "learning_rate": 2.5501689704536388e-05, "loss": 0.1779, "step": 26570 }, { "epoch": 0.62, "learning_rate": 2.5500809650090353e-05, "loss": 0.198, "step": 26575 }, { "epoch": 0.62, "learning_rate": 2.5499929595644318e-05, "loss": 0.2037, "step": 26580 }, { "epoch": 0.62, "learning_rate": 2.5499049541198283e-05, "loss": 0.2476, "step": 26585 }, { "epoch": 0.62, "learning_rate": 2.5498169486752248e-05, "loss": 0.2166, "step": 26590 }, { "epoch": 0.62, "learning_rate": 2.5497289432306213e-05, "loss": 0.4212, "step": 26595 }, { "epoch": 0.62, "learning_rate": 2.5496409377860178e-05, "loss": 0.5278, "step": 26600 }, { "epoch": 0.62, "learning_rate": 2.5495529323414143e-05, "loss": 0.1896, "step": 26605 }, { "epoch": 0.62, "learning_rate": 2.5494649268968108e-05, "loss": 0.0713, "step": 26610 }, { "epoch": 0.62, "learning_rate": 2.5493769214522073e-05, "loss": 0.1203, "step": 26615 }, { "epoch": 0.62, "learning_rate": 2.5492889160076038e-05, "loss": 0.2557, "step": 26620 }, { "epoch": 0.62, "learning_rate": 2.549200910563e-05, "loss": 0.1554, "step": 26625 }, { "epoch": 0.62, "learning_rate": 2.5491129051183968e-05, "loss": 0.2044, "step": 26630 }, { "epoch": 0.62, "learning_rate": 2.5490248996737933e-05, "loss": 0.3018, "step": 26635 }, { "epoch": 0.62, "learning_rate": 2.5489368942291898e-05, "loss": 0.2177, "step": 26640 }, { "epoch": 0.62, "learning_rate": 2.548848888784586e-05, "loss": 0.2936, "step": 26645 }, { "epoch": 0.62, "learning_rate": 2.5487608833399828e-05, "loss": 0.7004, "step": 26650 }, { "epoch": 0.62, "learning_rate": 2.5486728778953793e-05, "loss": 0.1117, "step": 26655 }, { "epoch": 0.62, "learning_rate": 2.548620074628617e-05, "loss": 0.1419, "step": 26660 }, { "epoch": 0.62, "learning_rate": 2.5485320691840136e-05, "loss": 0.1025, "step": 26665 }, { "epoch": 0.62, "learning_rate": 2.54844406373941e-05, "loss": 0.1085, "step": 26670 }, { "epoch": 0.62, "learning_rate": 2.5483560582948066e-05, "loss": 0.2243, "step": 26675 }, { "epoch": 0.62, "learning_rate": 2.5482680528502028e-05, "loss": 0.3368, "step": 26680 }, { "epoch": 0.62, "learning_rate": 2.5481800474055996e-05, "loss": 0.3092, "step": 26685 }, { "epoch": 0.62, "learning_rate": 2.548092041960996e-05, "loss": 0.3377, "step": 26690 }, { "epoch": 0.62, "learning_rate": 2.5480040365163926e-05, "loss": 0.1951, "step": 26695 }, { "epoch": 0.62, "learning_rate": 2.547916031071789e-05, "loss": 0.5893, "step": 26700 }, { "epoch": 0.62, "learning_rate": 2.5478280256271856e-05, "loss": 0.2535, "step": 26705 }, { "epoch": 0.62, "learning_rate": 2.547740020182582e-05, "loss": 0.0999, "step": 26710 }, { "epoch": 0.62, "learning_rate": 2.5476520147379783e-05, "loss": 0.1117, "step": 26715 }, { "epoch": 0.62, "learning_rate": 2.547564009293375e-05, "loss": 0.0993, "step": 26720 }, { "epoch": 0.62, "learning_rate": 2.5474760038487716e-05, "loss": 0.1107, "step": 26725 }, { "epoch": 0.62, "learning_rate": 2.547387998404168e-05, "loss": 0.1929, "step": 26730 }, { "epoch": 0.62, "learning_rate": 2.5472999929595643e-05, "loss": 0.2219, "step": 26735 }, { "epoch": 0.62, "learning_rate": 2.547211987514961e-05, "loss": 0.2474, "step": 26740 }, { "epoch": 0.62, "learning_rate": 2.5471239820703576e-05, "loss": 0.4096, "step": 26745 }, { "epoch": 0.62, "learning_rate": 2.5470359766257538e-05, "loss": 0.5289, "step": 26750 }, { "epoch": 0.62, "learning_rate": 2.5469479711811503e-05, "loss": 0.213, "step": 26755 }, { "epoch": 0.62, "learning_rate": 2.546859965736547e-05, "loss": 0.1131, "step": 26760 }, { "epoch": 0.62, "learning_rate": 2.5467719602919436e-05, "loss": 0.2132, "step": 26765 }, { "epoch": 0.62, "learning_rate": 2.5466839548473398e-05, "loss": 0.1847, "step": 26770 }, { "epoch": 0.62, "learning_rate": 2.5465959494027363e-05, "loss": 0.1886, "step": 26775 }, { "epoch": 0.62, "learning_rate": 2.546507943958133e-05, "loss": 0.1942, "step": 26780 }, { "epoch": 0.62, "learning_rate": 2.5464199385135293e-05, "loss": 0.3155, "step": 26785 }, { "epoch": 0.63, "learning_rate": 2.5463319330689258e-05, "loss": 0.329, "step": 26790 }, { "epoch": 0.63, "learning_rate": 2.5462439276243223e-05, "loss": 0.3436, "step": 26795 }, { "epoch": 0.63, "learning_rate": 2.546155922179719e-05, "loss": 0.5414, "step": 26800 }, { "epoch": 0.63, "learning_rate": 2.5460679167351153e-05, "loss": 0.3176, "step": 26805 }, { "epoch": 0.63, "learning_rate": 2.5459799112905118e-05, "loss": 0.2165, "step": 26810 }, { "epoch": 0.63, "learning_rate": 2.5458919058459083e-05, "loss": 0.1283, "step": 26815 }, { "epoch": 0.63, "learning_rate": 2.5458039004013048e-05, "loss": 0.1377, "step": 26820 }, { "epoch": 0.63, "learning_rate": 2.5457158949567013e-05, "loss": 0.2574, "step": 26825 }, { "epoch": 0.63, "learning_rate": 2.5456278895120978e-05, "loss": 0.2192, "step": 26830 }, { "epoch": 0.63, "learning_rate": 2.5455398840674947e-05, "loss": 0.3248, "step": 26835 }, { "epoch": 0.63, "learning_rate": 2.5454518786228908e-05, "loss": 0.2936, "step": 26840 }, { "epoch": 0.63, "learning_rate": 2.5453638731782873e-05, "loss": 0.2529, "step": 26845 }, { "epoch": 0.63, "learning_rate": 2.545275867733684e-05, "loss": 0.6769, "step": 26850 }, { "epoch": 0.63, "learning_rate": 2.5451878622890807e-05, "loss": 0.3266, "step": 26855 }, { "epoch": 0.63, "learning_rate": 2.545099856844477e-05, "loss": 0.131, "step": 26860 }, { "epoch": 0.63, "learning_rate": 2.5450118513998733e-05, "loss": 0.1157, "step": 26865 }, { "epoch": 0.63, "learning_rate": 2.54492384595527e-05, "loss": 0.1888, "step": 26870 }, { "epoch": 0.63, "learning_rate": 2.5448358405106663e-05, "loss": 0.128, "step": 26875 }, { "epoch": 0.63, "learning_rate": 2.544747835066063e-05, "loss": 0.1476, "step": 26880 }, { "epoch": 0.63, "learning_rate": 2.5446598296214593e-05, "loss": 0.138, "step": 26885 }, { "epoch": 0.63, "learning_rate": 2.544571824176856e-05, "loss": 0.486, "step": 26890 }, { "epoch": 0.63, "learning_rate": 2.5444838187322523e-05, "loss": 0.3961, "step": 26895 }, { "epoch": 0.63, "learning_rate": 2.544395813287649e-05, "loss": 0.4998, "step": 26900 }, { "epoch": 0.63, "learning_rate": 2.5443078078430454e-05, "loss": 0.3647, "step": 26905 }, { "epoch": 0.63, "learning_rate": 2.5442198023984415e-05, "loss": 0.044, "step": 26910 }, { "epoch": 0.63, "learning_rate": 2.5441317969538384e-05, "loss": 0.1632, "step": 26915 }, { "epoch": 0.63, "learning_rate": 2.544043791509235e-05, "loss": 0.1769, "step": 26920 }, { "epoch": 0.63, "learning_rate": 2.5439557860646314e-05, "loss": 0.2369, "step": 26925 }, { "epoch": 0.63, "learning_rate": 2.5438677806200275e-05, "loss": 0.3628, "step": 26930 }, { "epoch": 0.63, "learning_rate": 2.5437797751754244e-05, "loss": 0.318, "step": 26935 }, { "epoch": 0.63, "learning_rate": 2.543691769730821e-05, "loss": 0.2862, "step": 26940 }, { "epoch": 0.63, "learning_rate": 2.543603764286217e-05, "loss": 0.2106, "step": 26945 }, { "epoch": 0.63, "learning_rate": 2.5435157588416135e-05, "loss": 0.5897, "step": 26950 }, { "epoch": 0.63, "learning_rate": 2.5434277533970104e-05, "loss": 0.2573, "step": 26955 }, { "epoch": 0.63, "learning_rate": 2.543339747952407e-05, "loss": 0.0334, "step": 26960 }, { "epoch": 0.63, "learning_rate": 2.543251742507803e-05, "loss": 0.2611, "step": 26965 }, { "epoch": 0.63, "learning_rate": 2.5431637370632e-05, "loss": 0.1529, "step": 26970 }, { "epoch": 0.63, "learning_rate": 2.5430757316185964e-05, "loss": 0.0744, "step": 26975 }, { "epoch": 0.63, "learning_rate": 2.5429877261739925e-05, "loss": 0.2139, "step": 26980 }, { "epoch": 0.63, "learning_rate": 2.542899720729389e-05, "loss": 0.1645, "step": 26985 }, { "epoch": 0.63, "learning_rate": 2.542811715284786e-05, "loss": 0.2892, "step": 26990 }, { "epoch": 0.63, "learning_rate": 2.5427237098401824e-05, "loss": 0.4188, "step": 26995 }, { "epoch": 0.63, "learning_rate": 2.5426357043955786e-05, "loss": 0.5578, "step": 27000 }, { "epoch": 0.63, "learning_rate": 2.542547698950975e-05, "loss": 0.1943, "step": 27005 }, { "epoch": 0.63, "learning_rate": 2.542459693506372e-05, "loss": 0.0979, "step": 27010 }, { "epoch": 0.63, "learning_rate": 2.542371688061768e-05, "loss": 0.149, "step": 27015 }, { "epoch": 0.63, "learning_rate": 2.5422836826171646e-05, "loss": 0.2044, "step": 27020 }, { "epoch": 0.63, "learning_rate": 2.542195677172561e-05, "loss": 0.2825, "step": 27025 }, { "epoch": 0.63, "learning_rate": 2.542107671727958e-05, "loss": 0.1689, "step": 27030 }, { "epoch": 0.63, "learning_rate": 2.542019666283354e-05, "loss": 0.3277, "step": 27035 }, { "epoch": 0.63, "learning_rate": 2.5419316608387506e-05, "loss": 0.2568, "step": 27040 }, { "epoch": 0.63, "learning_rate": 2.541843655394147e-05, "loss": 0.4261, "step": 27045 }, { "epoch": 0.63, "learning_rate": 2.5417556499495436e-05, "loss": 0.4242, "step": 27050 }, { "epoch": 0.63, "learning_rate": 2.54166764450494e-05, "loss": 0.3255, "step": 27055 }, { "epoch": 0.63, "learning_rate": 2.5415796390603366e-05, "loss": 0.078, "step": 27060 }, { "epoch": 0.63, "learning_rate": 2.541491633615733e-05, "loss": 0.1105, "step": 27065 }, { "epoch": 0.63, "learning_rate": 2.5414036281711296e-05, "loss": 0.1591, "step": 27070 }, { "epoch": 0.63, "learning_rate": 2.541315622726526e-05, "loss": 0.1161, "step": 27075 }, { "epoch": 0.63, "learning_rate": 2.5412276172819226e-05, "loss": 0.1585, "step": 27080 }, { "epoch": 0.63, "learning_rate": 2.5411396118373187e-05, "loss": 0.2245, "step": 27085 }, { "epoch": 0.63, "learning_rate": 2.5410516063927156e-05, "loss": 0.2143, "step": 27090 }, { "epoch": 0.63, "learning_rate": 2.540963600948112e-05, "loss": 0.3287, "step": 27095 }, { "epoch": 0.63, "learning_rate": 2.5408755955035086e-05, "loss": 0.4874, "step": 27100 }, { "epoch": 0.63, "learning_rate": 2.540787590058905e-05, "loss": 0.2046, "step": 27105 }, { "epoch": 0.63, "learning_rate": 2.5406995846143016e-05, "loss": 0.0926, "step": 27110 }, { "epoch": 0.63, "learning_rate": 2.540611579169698e-05, "loss": 0.0618, "step": 27115 }, { "epoch": 0.63, "learning_rate": 2.5405235737250943e-05, "loss": 0.1037, "step": 27120 }, { "epoch": 0.63, "learning_rate": 2.540435568280491e-05, "loss": 0.2307, "step": 27125 }, { "epoch": 0.63, "learning_rate": 2.5403475628358876e-05, "loss": 0.2645, "step": 27130 }, { "epoch": 0.63, "learning_rate": 2.540259557391284e-05, "loss": 0.374, "step": 27135 }, { "epoch": 0.63, "learning_rate": 2.5401715519466803e-05, "loss": 0.2492, "step": 27140 }, { "epoch": 0.63, "learning_rate": 2.540083546502077e-05, "loss": 0.2728, "step": 27145 }, { "epoch": 0.63, "learning_rate": 2.5399955410574736e-05, "loss": 0.6255, "step": 27150 }, { "epoch": 0.63, "learning_rate": 2.53990753561287e-05, "loss": 0.3815, "step": 27155 }, { "epoch": 0.63, "learning_rate": 2.5398195301682663e-05, "loss": 0.0545, "step": 27160 }, { "epoch": 0.63, "learning_rate": 2.539731524723663e-05, "loss": 0.103, "step": 27165 }, { "epoch": 0.63, "learning_rate": 2.5396435192790596e-05, "loss": 0.0951, "step": 27170 }, { "epoch": 0.63, "learning_rate": 2.5395555138344558e-05, "loss": 0.1582, "step": 27175 }, { "epoch": 0.63, "learning_rate": 2.5394675083898523e-05, "loss": 0.1689, "step": 27180 }, { "epoch": 0.63, "learning_rate": 2.539379502945249e-05, "loss": 0.1398, "step": 27185 }, { "epoch": 0.63, "learning_rate": 2.5392914975006456e-05, "loss": 0.3711, "step": 27190 }, { "epoch": 0.63, "learning_rate": 2.5392034920560418e-05, "loss": 0.544, "step": 27195 }, { "epoch": 0.63, "learning_rate": 2.5391154866114383e-05, "loss": 0.7649, "step": 27200 }, { "epoch": 0.63, "learning_rate": 2.539027481166835e-05, "loss": 0.218, "step": 27205 }, { "epoch": 0.63, "learning_rate": 2.5389394757222313e-05, "loss": 0.1397, "step": 27210 }, { "epoch": 0.63, "learning_rate": 2.5388514702776278e-05, "loss": 0.135, "step": 27215 }, { "epoch": 0.64, "learning_rate": 2.5387634648330243e-05, "loss": 0.1933, "step": 27220 }, { "epoch": 0.64, "learning_rate": 2.538675459388421e-05, "loss": 0.1738, "step": 27225 }, { "epoch": 0.64, "learning_rate": 2.5385874539438173e-05, "loss": 0.1034, "step": 27230 }, { "epoch": 0.64, "learning_rate": 2.5384994484992138e-05, "loss": 0.1376, "step": 27235 }, { "epoch": 0.64, "learning_rate": 2.5384114430546103e-05, "loss": 0.1566, "step": 27240 }, { "epoch": 0.64, "learning_rate": 2.5383234376100068e-05, "loss": 0.4209, "step": 27245 }, { "epoch": 0.64, "learning_rate": 2.5382354321654033e-05, "loss": 0.5799, "step": 27250 }, { "epoch": 0.64, "learning_rate": 2.5381474267207998e-05, "loss": 0.3084, "step": 27255 }, { "epoch": 0.64, "learning_rate": 2.5380594212761967e-05, "loss": 0.0784, "step": 27260 }, { "epoch": 0.64, "learning_rate": 2.5379714158315928e-05, "loss": 0.2605, "step": 27265 }, { "epoch": 0.64, "learning_rate": 2.5378834103869893e-05, "loss": 0.1081, "step": 27270 }, { "epoch": 0.64, "learning_rate": 2.5377954049423858e-05, "loss": 0.1499, "step": 27275 }, { "epoch": 0.64, "learning_rate": 2.5377073994977823e-05, "loss": 0.137, "step": 27280 }, { "epoch": 0.64, "learning_rate": 2.537619394053179e-05, "loss": 0.1811, "step": 27285 }, { "epoch": 0.64, "learning_rate": 2.5375313886085753e-05, "loss": 0.2572, "step": 27290 }, { "epoch": 0.64, "learning_rate": 2.537443383163972e-05, "loss": 0.2851, "step": 27295 }, { "epoch": 0.64, "learning_rate": 2.5373553777193683e-05, "loss": 0.2397, "step": 27300 }, { "epoch": 0.64, "learning_rate": 2.537267372274765e-05, "loss": 0.2327, "step": 27305 }, { "epoch": 0.64, "learning_rate": 2.5371793668301613e-05, "loss": 0.1027, "step": 27310 }, { "epoch": 0.64, "learning_rate": 2.5370913613855575e-05, "loss": 0.1259, "step": 27315 }, { "epoch": 0.64, "learning_rate": 2.5370033559409543e-05, "loss": 0.1901, "step": 27320 }, { "epoch": 0.64, "learning_rate": 2.536915350496351e-05, "loss": 0.1677, "step": 27325 }, { "epoch": 0.64, "learning_rate": 2.5368273450517474e-05, "loss": 0.1827, "step": 27330 }, { "epoch": 0.64, "learning_rate": 2.5367393396071435e-05, "loss": 0.1402, "step": 27335 }, { "epoch": 0.64, "learning_rate": 2.5366513341625404e-05, "loss": 0.3817, "step": 27340 }, { "epoch": 0.64, "learning_rate": 2.536563328717937e-05, "loss": 0.3232, "step": 27345 }, { "epoch": 0.64, "learning_rate": 2.536475323273333e-05, "loss": 0.8524, "step": 27350 }, { "epoch": 0.64, "learning_rate": 2.5363873178287295e-05, "loss": 0.275, "step": 27355 }, { "epoch": 0.64, "learning_rate": 2.5362993123841264e-05, "loss": 0.1182, "step": 27360 }, { "epoch": 0.64, "learning_rate": 2.536211306939523e-05, "loss": 0.1676, "step": 27365 }, { "epoch": 0.64, "learning_rate": 2.536123301494919e-05, "loss": 0.0956, "step": 27370 }, { "epoch": 0.64, "learning_rate": 2.536035296050316e-05, "loss": 0.0859, "step": 27375 }, { "epoch": 0.64, "learning_rate": 2.5359472906057124e-05, "loss": 0.1649, "step": 27380 }, { "epoch": 0.64, "learning_rate": 2.5358592851611085e-05, "loss": 0.186, "step": 27385 }, { "epoch": 0.64, "learning_rate": 2.535771279716505e-05, "loss": 0.272, "step": 27390 }, { "epoch": 0.64, "learning_rate": 2.535683274271902e-05, "loss": 0.4892, "step": 27395 }, { "epoch": 0.64, "learning_rate": 2.5355952688272984e-05, "loss": 0.2801, "step": 27400 }, { "epoch": 0.64, "learning_rate": 2.5355072633826945e-05, "loss": 0.1477, "step": 27405 }, { "epoch": 0.64, "learning_rate": 2.535419257938091e-05, "loss": 0.0594, "step": 27410 }, { "epoch": 0.64, "learning_rate": 2.535331252493488e-05, "loss": 0.1195, "step": 27415 }, { "epoch": 0.64, "learning_rate": 2.535243247048884e-05, "loss": 0.1331, "step": 27420 }, { "epoch": 0.64, "learning_rate": 2.5351552416042806e-05, "loss": 0.1861, "step": 27425 }, { "epoch": 0.64, "learning_rate": 2.535067236159677e-05, "loss": 0.2682, "step": 27430 }, { "epoch": 0.64, "learning_rate": 2.534979230715074e-05, "loss": 0.3074, "step": 27435 }, { "epoch": 0.64, "learning_rate": 2.53489122527047e-05, "loss": 0.2653, "step": 27440 }, { "epoch": 0.64, "learning_rate": 2.5348032198258666e-05, "loss": 0.2835, "step": 27445 }, { "epoch": 0.64, "learning_rate": 2.534715214381263e-05, "loss": 0.5273, "step": 27450 }, { "epoch": 0.64, "learning_rate": 2.5346272089366596e-05, "loss": 0.2595, "step": 27455 }, { "epoch": 0.64, "learning_rate": 2.534539203492056e-05, "loss": 0.137, "step": 27460 }, { "epoch": 0.64, "learning_rate": 2.5344511980474526e-05, "loss": 0.1595, "step": 27465 }, { "epoch": 0.64, "learning_rate": 2.534363192602849e-05, "loss": 0.082, "step": 27470 }, { "epoch": 0.64, "learning_rate": 2.5342751871582456e-05, "loss": 0.2363, "step": 27475 }, { "epoch": 0.64, "learning_rate": 2.534187181713642e-05, "loss": 0.0941, "step": 27480 }, { "epoch": 0.64, "learning_rate": 2.5340991762690386e-05, "loss": 0.2487, "step": 27485 }, { "epoch": 0.64, "learning_rate": 2.534011170824435e-05, "loss": 0.3856, "step": 27490 }, { "epoch": 0.64, "learning_rate": 2.5339231653798316e-05, "loss": 0.4527, "step": 27495 }, { "epoch": 0.64, "learning_rate": 2.533835159935228e-05, "loss": 0.6185, "step": 27500 }, { "epoch": 0.64, "learning_rate": 2.5337471544906246e-05, "loss": 0.1275, "step": 27505 }, { "epoch": 0.64, "learning_rate": 2.533659149046021e-05, "loss": 0.1186, "step": 27510 }, { "epoch": 0.64, "learning_rate": 2.5335711436014176e-05, "loss": 0.0718, "step": 27515 }, { "epoch": 0.64, "learning_rate": 2.533483138156814e-05, "loss": 0.1206, "step": 27520 }, { "epoch": 0.64, "learning_rate": 2.5333951327122106e-05, "loss": 0.1909, "step": 27525 }, { "epoch": 0.64, "learning_rate": 2.533307127267607e-05, "loss": 0.299, "step": 27530 }, { "epoch": 0.64, "learning_rate": 2.5332191218230036e-05, "loss": 0.2565, "step": 27535 }, { "epoch": 0.64, "learning_rate": 2.5331311163784e-05, "loss": 0.373, "step": 27540 }, { "epoch": 0.64, "learning_rate": 2.5330431109337963e-05, "loss": 0.472, "step": 27545 }, { "epoch": 0.64, "learning_rate": 2.532955105489193e-05, "loss": 0.4302, "step": 27550 }, { "epoch": 0.64, "learning_rate": 2.5328671000445896e-05, "loss": 0.2945, "step": 27555 }, { "epoch": 0.64, "learning_rate": 2.532779094599986e-05, "loss": 0.1237, "step": 27560 }, { "epoch": 0.64, "learning_rate": 2.5326910891553823e-05, "loss": 0.1503, "step": 27565 }, { "epoch": 0.64, "learning_rate": 2.532603083710779e-05, "loss": 0.1437, "step": 27570 }, { "epoch": 0.64, "learning_rate": 2.5325150782661756e-05, "loss": 0.0798, "step": 27575 }, { "epoch": 0.64, "learning_rate": 2.5324270728215718e-05, "loss": 0.101, "step": 27580 }, { "epoch": 0.64, "learning_rate": 2.5323390673769683e-05, "loss": 0.2122, "step": 27585 }, { "epoch": 0.64, "learning_rate": 2.532251061932365e-05, "loss": 0.3781, "step": 27590 }, { "epoch": 0.64, "learning_rate": 2.5321630564877616e-05, "loss": 0.5484, "step": 27595 }, { "epoch": 0.64, "learning_rate": 2.5320750510431578e-05, "loss": 0.5217, "step": 27600 }, { "epoch": 0.64, "learning_rate": 2.5319870455985543e-05, "loss": 0.2633, "step": 27605 }, { "epoch": 0.64, "learning_rate": 2.531899040153951e-05, "loss": 0.1565, "step": 27610 }, { "epoch": 0.64, "learning_rate": 2.5318110347093473e-05, "loss": 0.0737, "step": 27615 }, { "epoch": 0.64, "learning_rate": 2.5317230292647438e-05, "loss": 0.095, "step": 27620 }, { "epoch": 0.64, "learning_rate": 2.5316350238201403e-05, "loss": 0.2725, "step": 27625 }, { "epoch": 0.64, "learning_rate": 2.531547018375537e-05, "loss": 0.1469, "step": 27630 }, { "epoch": 0.64, "learning_rate": 2.5314590129309333e-05, "loss": 0.2197, "step": 27635 }, { "epoch": 0.64, "learning_rate": 2.5313710074863298e-05, "loss": 0.1407, "step": 27640 }, { "epoch": 0.64, "learning_rate": 2.5312830020417263e-05, "loss": 0.3594, "step": 27645 }, { "epoch": 0.65, "learning_rate": 2.5311949965971228e-05, "loss": 0.6237, "step": 27650 }, { "epoch": 0.65, "learning_rate": 2.5311069911525193e-05, "loss": 0.1959, "step": 27655 }, { "epoch": 0.65, "learning_rate": 2.5310189857079158e-05, "loss": 0.0457, "step": 27660 }, { "epoch": 0.65, "learning_rate": 2.5309309802633126e-05, "loss": 0.1056, "step": 27665 }, { "epoch": 0.65, "learning_rate": 2.5308429748187088e-05, "loss": 0.1895, "step": 27670 }, { "epoch": 0.65, "learning_rate": 2.5307549693741053e-05, "loss": 0.1643, "step": 27675 }, { "epoch": 0.65, "learning_rate": 2.5306669639295018e-05, "loss": 0.1823, "step": 27680 }, { "epoch": 0.65, "learning_rate": 2.5305789584848983e-05, "loss": 0.2082, "step": 27685 }, { "epoch": 0.65, "learning_rate": 2.5304909530402948e-05, "loss": 0.2651, "step": 27690 }, { "epoch": 0.65, "learning_rate": 2.5304029475956913e-05, "loss": 0.5811, "step": 27695 }, { "epoch": 0.65, "learning_rate": 2.5303149421510878e-05, "loss": 0.3984, "step": 27700 }, { "epoch": 0.65, "learning_rate": 2.5302269367064843e-05, "loss": 0.2996, "step": 27705 }, { "epoch": 0.65, "learning_rate": 2.5301389312618808e-05, "loss": 0.0899, "step": 27710 }, { "epoch": 0.65, "learning_rate": 2.5300509258172773e-05, "loss": 0.1221, "step": 27715 }, { "epoch": 0.65, "learning_rate": 2.5299629203726735e-05, "loss": 0.0774, "step": 27720 }, { "epoch": 0.65, "learning_rate": 2.5298749149280703e-05, "loss": 0.1428, "step": 27725 }, { "epoch": 0.65, "learning_rate": 2.529786909483467e-05, "loss": 0.16, "step": 27730 }, { "epoch": 0.65, "learning_rate": 2.5296989040388633e-05, "loss": 0.5114, "step": 27735 }, { "epoch": 0.65, "learning_rate": 2.5296108985942595e-05, "loss": 0.341, "step": 27740 }, { "epoch": 0.65, "learning_rate": 2.5295228931496563e-05, "loss": 0.3154, "step": 27745 }, { "epoch": 0.65, "learning_rate": 2.529434887705053e-05, "loss": 0.526, "step": 27750 }, { "epoch": 0.65, "learning_rate": 2.529346882260449e-05, "loss": 0.2516, "step": 27755 }, { "epoch": 0.65, "learning_rate": 2.5292588768158455e-05, "loss": 0.0467, "step": 27760 }, { "epoch": 0.65, "learning_rate": 2.5291708713712424e-05, "loss": 0.095, "step": 27765 }, { "epoch": 0.65, "learning_rate": 2.529082865926639e-05, "loss": 0.124, "step": 27770 }, { "epoch": 0.65, "learning_rate": 2.528994860482035e-05, "loss": 0.1709, "step": 27775 }, { "epoch": 0.65, "learning_rate": 2.5289068550374315e-05, "loss": 0.2192, "step": 27780 }, { "epoch": 0.65, "learning_rate": 2.5288188495928284e-05, "loss": 0.2557, "step": 27785 }, { "epoch": 0.65, "learning_rate": 2.528730844148225e-05, "loss": 0.1376, "step": 27790 }, { "epoch": 0.65, "learning_rate": 2.528642838703621e-05, "loss": 0.4109, "step": 27795 }, { "epoch": 0.65, "learning_rate": 2.528554833259018e-05, "loss": 0.4651, "step": 27800 }, { "epoch": 0.65, "learning_rate": 2.5284668278144144e-05, "loss": 0.2873, "step": 27805 }, { "epoch": 0.65, "learning_rate": 2.5283788223698105e-05, "loss": 0.0647, "step": 27810 }, { "epoch": 0.65, "learning_rate": 2.528290816925207e-05, "loss": 0.187, "step": 27815 }, { "epoch": 0.65, "learning_rate": 2.528202811480604e-05, "loss": 0.1316, "step": 27820 }, { "epoch": 0.65, "learning_rate": 2.5281148060360004e-05, "loss": 0.2783, "step": 27825 }, { "epoch": 0.65, "learning_rate": 2.5280268005913965e-05, "loss": 0.2721, "step": 27830 }, { "epoch": 0.65, "learning_rate": 2.527938795146793e-05, "loss": 0.1804, "step": 27835 }, { "epoch": 0.65, "learning_rate": 2.52785078970219e-05, "loss": 0.2628, "step": 27840 }, { "epoch": 0.65, "learning_rate": 2.527762784257586e-05, "loss": 0.4007, "step": 27845 }, { "epoch": 0.65, "learning_rate": 2.5276747788129825e-05, "loss": 0.3348, "step": 27850 }, { "epoch": 0.65, "learning_rate": 2.527586773368379e-05, "loss": 0.3473, "step": 27855 }, { "epoch": 0.65, "learning_rate": 2.527498767923776e-05, "loss": 0.1028, "step": 27860 }, { "epoch": 0.65, "learning_rate": 2.527410762479172e-05, "loss": 0.0703, "step": 27865 }, { "epoch": 0.65, "learning_rate": 2.5273227570345686e-05, "loss": 0.0678, "step": 27870 }, { "epoch": 0.65, "learning_rate": 2.527234751589965e-05, "loss": 0.1349, "step": 27875 }, { "epoch": 0.65, "learning_rate": 2.5271467461453616e-05, "loss": 0.2797, "step": 27880 }, { "epoch": 0.65, "learning_rate": 2.527058740700758e-05, "loss": 0.4111, "step": 27885 }, { "epoch": 0.65, "learning_rate": 2.5269707352561546e-05, "loss": 0.2858, "step": 27890 }, { "epoch": 0.65, "learning_rate": 2.526882729811551e-05, "loss": 0.375, "step": 27895 }, { "epoch": 0.65, "learning_rate": 2.5267947243669476e-05, "loss": 0.2839, "step": 27900 }, { "epoch": 0.65, "learning_rate": 2.526706718922344e-05, "loss": 0.1804, "step": 27905 }, { "epoch": 0.65, "learning_rate": 2.5266187134777406e-05, "loss": 0.1002, "step": 27910 }, { "epoch": 0.65, "learning_rate": 2.526530708033137e-05, "loss": 0.1353, "step": 27915 }, { "epoch": 0.65, "learning_rate": 2.5264427025885336e-05, "loss": 0.0962, "step": 27920 }, { "epoch": 0.65, "learning_rate": 2.52635469714393e-05, "loss": 0.2556, "step": 27925 }, { "epoch": 0.65, "learning_rate": 2.5262666916993266e-05, "loss": 0.3236, "step": 27930 }, { "epoch": 0.65, "learning_rate": 2.526178686254723e-05, "loss": 0.0874, "step": 27935 }, { "epoch": 0.65, "learning_rate": 2.5260906808101196e-05, "loss": 0.2264, "step": 27940 }, { "epoch": 0.65, "learning_rate": 2.526002675365516e-05, "loss": 0.3106, "step": 27945 }, { "epoch": 0.65, "learning_rate": 2.5259146699209122e-05, "loss": 0.5805, "step": 27950 }, { "epoch": 0.65, "learning_rate": 2.525826664476309e-05, "loss": 0.3224, "step": 27955 }, { "epoch": 0.65, "learning_rate": 2.5257386590317056e-05, "loss": 0.1086, "step": 27960 }, { "epoch": 0.65, "learning_rate": 2.525650653587102e-05, "loss": 0.1385, "step": 27965 }, { "epoch": 0.65, "learning_rate": 2.5255626481424983e-05, "loss": 0.1398, "step": 27970 }, { "epoch": 0.65, "learning_rate": 2.525474642697895e-05, "loss": 0.2519, "step": 27975 }, { "epoch": 0.65, "learning_rate": 2.5253866372532916e-05, "loss": 0.2138, "step": 27980 }, { "epoch": 0.65, "learning_rate": 2.5252986318086878e-05, "loss": 0.2035, "step": 27985 }, { "epoch": 0.65, "learning_rate": 2.5252106263640843e-05, "loss": 0.4682, "step": 27990 }, { "epoch": 0.65, "learning_rate": 2.525122620919481e-05, "loss": 0.1876, "step": 27995 }, { "epoch": 0.65, "learning_rate": 2.5250346154748776e-05, "loss": 0.4013, "step": 28000 }, { "epoch": 0.65, "learning_rate": 2.5249466100302738e-05, "loss": 0.3147, "step": 28005 }, { "epoch": 0.65, "learning_rate": 2.5248586045856703e-05, "loss": 0.1545, "step": 28010 }, { "epoch": 0.65, "learning_rate": 2.524770599141067e-05, "loss": 0.146, "step": 28015 }, { "epoch": 0.65, "learning_rate": 2.5246825936964633e-05, "loss": 0.104, "step": 28020 }, { "epoch": 0.65, "learning_rate": 2.5245945882518598e-05, "loss": 0.1654, "step": 28025 }, { "epoch": 0.65, "learning_rate": 2.5245065828072563e-05, "loss": 0.1532, "step": 28030 }, { "epoch": 0.65, "learning_rate": 2.524418577362653e-05, "loss": 0.1632, "step": 28035 }, { "epoch": 0.65, "learning_rate": 2.5243305719180493e-05, "loss": 0.2273, "step": 28040 }, { "epoch": 0.65, "learning_rate": 2.5242425664734458e-05, "loss": 0.3839, "step": 28045 }, { "epoch": 0.65, "learning_rate": 2.5241545610288423e-05, "loss": 0.453, "step": 28050 }, { "epoch": 0.65, "learning_rate": 2.5240665555842388e-05, "loss": 0.2285, "step": 28055 }, { "epoch": 0.65, "learning_rate": 2.5239785501396353e-05, "loss": 0.0699, "step": 28060 }, { "epoch": 0.65, "learning_rate": 2.5238905446950318e-05, "loss": 0.1987, "step": 28065 }, { "epoch": 0.65, "learning_rate": 2.5238025392504286e-05, "loss": 0.1851, "step": 28070 }, { "epoch": 0.66, "learning_rate": 2.5237145338058248e-05, "loss": 0.3159, "step": 28075 }, { "epoch": 0.66, "learning_rate": 2.5236265283612213e-05, "loss": 0.2771, "step": 28080 }, { "epoch": 0.66, "learning_rate": 2.5235385229166178e-05, "loss": 0.1429, "step": 28085 }, { "epoch": 0.66, "learning_rate": 2.5234505174720146e-05, "loss": 0.3998, "step": 28090 }, { "epoch": 0.66, "learning_rate": 2.5233625120274108e-05, "loss": 0.4587, "step": 28095 }, { "epoch": 0.66, "learning_rate": 2.5232745065828073e-05, "loss": 0.3742, "step": 28100 }, { "epoch": 0.66, "learning_rate": 2.5231865011382038e-05, "loss": 0.2468, "step": 28105 }, { "epoch": 0.66, "learning_rate": 2.5230984956936003e-05, "loss": 0.1836, "step": 28110 }, { "epoch": 0.66, "learning_rate": 2.5230104902489968e-05, "loss": 0.1531, "step": 28115 }, { "epoch": 0.66, "learning_rate": 2.5229224848043933e-05, "loss": 0.2117, "step": 28120 }, { "epoch": 0.66, "learning_rate": 2.5228344793597898e-05, "loss": 0.2031, "step": 28125 }, { "epoch": 0.66, "learning_rate": 2.5227464739151863e-05, "loss": 0.2505, "step": 28130 }, { "epoch": 0.66, "learning_rate": 2.5226584684705828e-05, "loss": 0.3888, "step": 28135 }, { "epoch": 0.66, "learning_rate": 2.5225704630259793e-05, "loss": 0.1811, "step": 28140 }, { "epoch": 0.66, "learning_rate": 2.5224824575813755e-05, "loss": 0.1971, "step": 28145 }, { "epoch": 0.66, "learning_rate": 2.5223944521367723e-05, "loss": 0.3167, "step": 28150 }, { "epoch": 0.66, "learning_rate": 2.522306446692169e-05, "loss": 0.2703, "step": 28155 }, { "epoch": 0.66, "learning_rate": 2.5222184412475653e-05, "loss": 0.1014, "step": 28160 }, { "epoch": 0.66, "learning_rate": 2.5221304358029615e-05, "loss": 0.0555, "step": 28165 }, { "epoch": 0.66, "learning_rate": 2.5220424303583583e-05, "loss": 0.1818, "step": 28170 }, { "epoch": 0.66, "learning_rate": 2.521954424913755e-05, "loss": 0.0332, "step": 28175 }, { "epoch": 0.66, "learning_rate": 2.521866419469151e-05, "loss": 0.1021, "step": 28180 }, { "epoch": 0.66, "learning_rate": 2.5217784140245475e-05, "loss": 0.1489, "step": 28185 }, { "epoch": 0.66, "learning_rate": 2.5216904085799443e-05, "loss": 0.2584, "step": 28190 }, { "epoch": 0.66, "learning_rate": 2.521602403135341e-05, "loss": 0.1991, "step": 28195 }, { "epoch": 0.66, "learning_rate": 2.521514397690737e-05, "loss": 0.4206, "step": 28200 }, { "epoch": 0.66, "learning_rate": 2.521426392246134e-05, "loss": 0.1701, "step": 28205 }, { "epoch": 0.66, "learning_rate": 2.5213383868015304e-05, "loss": 0.0741, "step": 28210 }, { "epoch": 0.66, "learning_rate": 2.5212503813569265e-05, "loss": 0.0526, "step": 28215 }, { "epoch": 0.66, "learning_rate": 2.521162375912323e-05, "loss": 0.0947, "step": 28220 }, { "epoch": 0.66, "learning_rate": 2.52107437046772e-05, "loss": 0.1408, "step": 28225 }, { "epoch": 0.66, "learning_rate": 2.5209863650231164e-05, "loss": 0.1743, "step": 28230 }, { "epoch": 0.66, "learning_rate": 2.5208983595785125e-05, "loss": 0.3016, "step": 28235 }, { "epoch": 0.66, "learning_rate": 2.520810354133909e-05, "loss": 0.4071, "step": 28240 }, { "epoch": 0.66, "learning_rate": 2.520722348689306e-05, "loss": 0.362, "step": 28245 }, { "epoch": 0.66, "learning_rate": 2.520634343244702e-05, "loss": 0.5806, "step": 28250 }, { "epoch": 0.66, "learning_rate": 2.5205463378000985e-05, "loss": 0.1611, "step": 28255 }, { "epoch": 0.66, "learning_rate": 2.520458332355495e-05, "loss": 0.0481, "step": 28260 }, { "epoch": 0.66, "learning_rate": 2.520370326910892e-05, "loss": 0.1948, "step": 28265 }, { "epoch": 0.66, "learning_rate": 2.520282321466288e-05, "loss": 0.0967, "step": 28270 }, { "epoch": 0.66, "learning_rate": 2.5201943160216845e-05, "loss": 0.218, "step": 28275 }, { "epoch": 0.66, "learning_rate": 2.520106310577081e-05, "loss": 0.1273, "step": 28280 }, { "epoch": 0.66, "learning_rate": 2.5200183051324775e-05, "loss": 0.3099, "step": 28285 }, { "epoch": 0.66, "learning_rate": 2.519930299687874e-05, "loss": 0.2393, "step": 28290 }, { "epoch": 0.66, "learning_rate": 2.5198422942432706e-05, "loss": 0.4024, "step": 28295 }, { "epoch": 0.66, "learning_rate": 2.519754288798667e-05, "loss": 0.4376, "step": 28300 }, { "epoch": 0.66, "learning_rate": 2.5196662833540636e-05, "loss": 0.3293, "step": 28305 }, { "epoch": 0.66, "learning_rate": 2.51957827790946e-05, "loss": 0.0653, "step": 28310 }, { "epoch": 0.66, "learning_rate": 2.5194902724648566e-05, "loss": 0.1332, "step": 28315 }, { "epoch": 0.66, "learning_rate": 2.5194022670202527e-05, "loss": 0.1628, "step": 28320 }, { "epoch": 0.66, "learning_rate": 2.5193142615756496e-05, "loss": 0.1547, "step": 28325 }, { "epoch": 0.66, "learning_rate": 2.519226256131046e-05, "loss": 0.2014, "step": 28330 }, { "epoch": 0.66, "learning_rate": 2.5191382506864426e-05, "loss": 0.1967, "step": 28335 }, { "epoch": 0.66, "learning_rate": 2.519050245241839e-05, "loss": 0.2367, "step": 28340 }, { "epoch": 0.66, "learning_rate": 2.5189622397972356e-05, "loss": 0.3466, "step": 28345 }, { "epoch": 0.66, "learning_rate": 2.518874234352632e-05, "loss": 0.4376, "step": 28350 }, { "epoch": 0.66, "learning_rate": 2.5187862289080282e-05, "loss": 0.2045, "step": 28355 }, { "epoch": 0.66, "learning_rate": 2.518698223463425e-05, "loss": 0.1036, "step": 28360 }, { "epoch": 0.66, "learning_rate": 2.5186102180188216e-05, "loss": 0.1749, "step": 28365 }, { "epoch": 0.66, "learning_rate": 2.518522212574218e-05, "loss": 0.2834, "step": 28370 }, { "epoch": 0.66, "learning_rate": 2.5184342071296142e-05, "loss": 0.1389, "step": 28375 }, { "epoch": 0.66, "learning_rate": 2.518346201685011e-05, "loss": 0.2044, "step": 28380 }, { "epoch": 0.66, "learning_rate": 2.5182581962404076e-05, "loss": 0.1484, "step": 28385 }, { "epoch": 0.66, "learning_rate": 2.518170190795804e-05, "loss": 0.4207, "step": 28390 }, { "epoch": 0.66, "learning_rate": 2.5180821853512003e-05, "loss": 0.2912, "step": 28395 }, { "epoch": 0.66, "learning_rate": 2.517994179906597e-05, "loss": 0.5532, "step": 28400 }, { "epoch": 0.66, "learning_rate": 2.5179061744619936e-05, "loss": 0.2988, "step": 28405 }, { "epoch": 0.66, "learning_rate": 2.5178181690173898e-05, "loss": 0.0911, "step": 28410 }, { "epoch": 0.66, "learning_rate": 2.5177301635727863e-05, "loss": 0.1359, "step": 28415 }, { "epoch": 0.66, "learning_rate": 2.517642158128183e-05, "loss": 0.1515, "step": 28420 }, { "epoch": 0.66, "learning_rate": 2.5175541526835796e-05, "loss": 0.1337, "step": 28425 }, { "epoch": 0.66, "learning_rate": 2.5174661472389758e-05, "loss": 0.1118, "step": 28430 }, { "epoch": 0.66, "learning_rate": 2.5173781417943723e-05, "loss": 0.2616, "step": 28435 }, { "epoch": 0.66, "learning_rate": 2.517290136349769e-05, "loss": 0.2884, "step": 28440 }, { "epoch": 0.66, "learning_rate": 2.5172021309051653e-05, "loss": 0.5631, "step": 28445 }, { "epoch": 0.66, "learning_rate": 2.5171141254605618e-05, "loss": 0.5089, "step": 28450 }, { "epoch": 0.66, "learning_rate": 2.5170261200159583e-05, "loss": 0.3116, "step": 28455 }, { "epoch": 0.66, "learning_rate": 2.516938114571355e-05, "loss": 0.0778, "step": 28460 }, { "epoch": 0.66, "learning_rate": 2.5168501091267513e-05, "loss": 0.1193, "step": 28465 }, { "epoch": 0.66, "learning_rate": 2.5167621036821478e-05, "loss": 0.0742, "step": 28470 }, { "epoch": 0.66, "learning_rate": 2.5166740982375446e-05, "loss": 0.1978, "step": 28475 }, { "epoch": 0.66, "learning_rate": 2.5165860927929408e-05, "loss": 0.1747, "step": 28480 }, { "epoch": 0.66, "learning_rate": 2.5164980873483373e-05, "loss": 0.2328, "step": 28485 }, { "epoch": 0.66, "learning_rate": 2.5164100819037338e-05, "loss": 0.3655, "step": 28490 }, { "epoch": 0.66, "learning_rate": 2.5163220764591306e-05, "loss": 0.3411, "step": 28495 }, { "epoch": 0.66, "learning_rate": 2.5162340710145268e-05, "loss": 0.6508, "step": 28500 }, { "epoch": 0.67, "learning_rate": 2.5161460655699233e-05, "loss": 0.2051, "step": 28505 }, { "epoch": 0.67, "learning_rate": 2.5160580601253198e-05, "loss": 0.1818, "step": 28510 }, { "epoch": 0.67, "learning_rate": 2.5159700546807163e-05, "loss": 0.0716, "step": 28515 }, { "epoch": 0.67, "learning_rate": 2.5158820492361128e-05, "loss": 0.0682, "step": 28520 }, { "epoch": 0.67, "learning_rate": 2.5157940437915093e-05, "loss": 0.1872, "step": 28525 }, { "epoch": 0.67, "learning_rate": 2.5157060383469058e-05, "loss": 0.1705, "step": 28530 }, { "epoch": 0.67, "learning_rate": 2.5156180329023023e-05, "loss": 0.2954, "step": 28535 }, { "epoch": 0.67, "learning_rate": 2.5155300274576988e-05, "loss": 0.2459, "step": 28540 }, { "epoch": 0.67, "learning_rate": 2.5154420220130953e-05, "loss": 0.3289, "step": 28545 }, { "epoch": 0.67, "learning_rate": 2.5153540165684915e-05, "loss": 0.3434, "step": 28550 }, { "epoch": 0.67, "learning_rate": 2.5152660111238883e-05, "loss": 0.2134, "step": 28555 }, { "epoch": 0.67, "learning_rate": 2.5151780056792848e-05, "loss": 0.089, "step": 28560 }, { "epoch": 0.67, "learning_rate": 2.5150900002346813e-05, "loss": 0.1751, "step": 28565 }, { "epoch": 0.67, "learning_rate": 2.5150019947900775e-05, "loss": 0.2115, "step": 28570 }, { "epoch": 0.67, "learning_rate": 2.5149139893454743e-05, "loss": 0.1691, "step": 28575 }, { "epoch": 0.67, "learning_rate": 2.5148259839008708e-05, "loss": 0.1877, "step": 28580 }, { "epoch": 0.67, "learning_rate": 2.514737978456267e-05, "loss": 0.1813, "step": 28585 }, { "epoch": 0.67, "learning_rate": 2.5146499730116635e-05, "loss": 0.2898, "step": 28590 }, { "epoch": 0.67, "learning_rate": 2.5145619675670603e-05, "loss": 0.4381, "step": 28595 }, { "epoch": 0.67, "learning_rate": 2.514473962122457e-05, "loss": 0.2901, "step": 28600 }, { "epoch": 0.67, "learning_rate": 2.514385956677853e-05, "loss": 0.2655, "step": 28605 }, { "epoch": 0.67, "learning_rate": 2.51429795123325e-05, "loss": 0.1876, "step": 28610 }, { "epoch": 0.67, "learning_rate": 2.5142099457886463e-05, "loss": 0.1499, "step": 28615 }, { "epoch": 0.67, "learning_rate": 2.5141219403440425e-05, "loss": 0.0986, "step": 28620 }, { "epoch": 0.67, "learning_rate": 2.514033934899439e-05, "loss": 0.1274, "step": 28625 }, { "epoch": 0.67, "learning_rate": 2.513945929454836e-05, "loss": 0.1948, "step": 28630 }, { "epoch": 0.67, "learning_rate": 2.5138579240102324e-05, "loss": 0.188, "step": 28635 }, { "epoch": 0.67, "learning_rate": 2.5137699185656285e-05, "loss": 0.3528, "step": 28640 }, { "epoch": 0.67, "learning_rate": 2.513681913121025e-05, "loss": 0.3079, "step": 28645 }, { "epoch": 0.67, "learning_rate": 2.513593907676422e-05, "loss": 0.6319, "step": 28650 }, { "epoch": 0.67, "learning_rate": 2.513505902231818e-05, "loss": 0.2473, "step": 28655 }, { "epoch": 0.67, "learning_rate": 2.5134178967872145e-05, "loss": 0.1303, "step": 28660 }, { "epoch": 0.67, "learning_rate": 2.513329891342611e-05, "loss": 0.185, "step": 28665 }, { "epoch": 0.67, "learning_rate": 2.513241885898008e-05, "loss": 0.1514, "step": 28670 }, { "epoch": 0.67, "learning_rate": 2.513153880453404e-05, "loss": 0.1874, "step": 28675 }, { "epoch": 0.67, "learning_rate": 2.5130658750088005e-05, "loss": 0.1968, "step": 28680 }, { "epoch": 0.67, "learning_rate": 2.512977869564197e-05, "loss": 0.2069, "step": 28685 }, { "epoch": 0.67, "learning_rate": 2.512889864119594e-05, "loss": 0.1775, "step": 28690 }, { "epoch": 0.67, "learning_rate": 2.51280185867499e-05, "loss": 0.4247, "step": 28695 }, { "epoch": 0.67, "learning_rate": 2.5127138532303865e-05, "loss": 0.5238, "step": 28700 }, { "epoch": 0.67, "learning_rate": 2.512625847785783e-05, "loss": 0.212, "step": 28705 }, { "epoch": 0.67, "learning_rate": 2.5125378423411795e-05, "loss": 0.0928, "step": 28710 }, { "epoch": 0.67, "learning_rate": 2.512449836896576e-05, "loss": 0.068, "step": 28715 }, { "epoch": 0.67, "learning_rate": 2.5123618314519725e-05, "loss": 0.1194, "step": 28720 }, { "epoch": 0.67, "learning_rate": 2.512273826007369e-05, "loss": 0.109, "step": 28725 }, { "epoch": 0.67, "learning_rate": 2.5121858205627656e-05, "loss": 0.2958, "step": 28730 }, { "epoch": 0.67, "learning_rate": 2.512097815118162e-05, "loss": 0.2194, "step": 28735 }, { "epoch": 0.67, "learning_rate": 2.5120098096735586e-05, "loss": 0.177, "step": 28740 }, { "epoch": 0.67, "learning_rate": 2.511921804228955e-05, "loss": 0.2067, "step": 28745 }, { "epoch": 0.67, "learning_rate": 2.5118337987843516e-05, "loss": 0.2798, "step": 28750 }, { "epoch": 0.67, "learning_rate": 2.511745793339748e-05, "loss": 0.1922, "step": 28755 }, { "epoch": 0.67, "learning_rate": 2.5116577878951446e-05, "loss": 0.0772, "step": 28760 }, { "epoch": 0.67, "learning_rate": 2.511569782450541e-05, "loss": 0.1307, "step": 28765 }, { "epoch": 0.67, "learning_rate": 2.5114817770059376e-05, "loss": 0.1947, "step": 28770 }, { "epoch": 0.67, "learning_rate": 2.511393771561334e-05, "loss": 0.1162, "step": 28775 }, { "epoch": 0.67, "learning_rate": 2.5113057661167302e-05, "loss": 0.1843, "step": 28780 }, { "epoch": 0.67, "learning_rate": 2.511217760672127e-05, "loss": 0.1968, "step": 28785 }, { "epoch": 0.67, "learning_rate": 2.5111297552275236e-05, "loss": 0.1873, "step": 28790 }, { "epoch": 0.67, "learning_rate": 2.51104174978292e-05, "loss": 0.2818, "step": 28795 }, { "epoch": 0.67, "learning_rate": 2.5109537443383162e-05, "loss": 0.6431, "step": 28800 }, { "epoch": 0.67, "learning_rate": 2.510865738893713e-05, "loss": 0.2476, "step": 28805 }, { "epoch": 0.67, "learning_rate": 2.5107777334491096e-05, "loss": 0.1271, "step": 28810 }, { "epoch": 0.67, "learning_rate": 2.5106897280045057e-05, "loss": 0.0552, "step": 28815 }, { "epoch": 0.67, "learning_rate": 2.5106017225599023e-05, "loss": 0.1664, "step": 28820 }, { "epoch": 0.67, "learning_rate": 2.510513717115299e-05, "loss": 0.1945, "step": 28825 }, { "epoch": 0.67, "learning_rate": 2.5104257116706956e-05, "loss": 0.1467, "step": 28830 }, { "epoch": 0.67, "learning_rate": 2.5103377062260918e-05, "loss": 0.481, "step": 28835 }, { "epoch": 0.67, "learning_rate": 2.5102497007814883e-05, "loss": 0.2041, "step": 28840 }, { "epoch": 0.67, "learning_rate": 2.510161695336885e-05, "loss": 0.3265, "step": 28845 }, { "epoch": 0.67, "learning_rate": 2.5100736898922813e-05, "loss": 0.4221, "step": 28850 }, { "epoch": 0.67, "learning_rate": 2.5099856844476778e-05, "loss": 0.3022, "step": 28855 }, { "epoch": 0.67, "learning_rate": 2.5098976790030743e-05, "loss": 0.0475, "step": 28860 }, { "epoch": 0.67, "learning_rate": 2.509809673558471e-05, "loss": 0.1279, "step": 28865 }, { "epoch": 0.67, "learning_rate": 2.5097216681138673e-05, "loss": 0.1694, "step": 28870 }, { "epoch": 0.67, "learning_rate": 2.5096336626692638e-05, "loss": 0.2751, "step": 28875 }, { "epoch": 0.67, "learning_rate": 2.5095456572246606e-05, "loss": 0.0859, "step": 28880 }, { "epoch": 0.67, "learning_rate": 2.5094576517800568e-05, "loss": 0.0745, "step": 28885 }, { "epoch": 0.67, "learning_rate": 2.5093696463354533e-05, "loss": 0.324, "step": 28890 }, { "epoch": 0.67, "learning_rate": 2.5092816408908498e-05, "loss": 0.3837, "step": 28895 }, { "epoch": 0.67, "learning_rate": 2.5091936354462466e-05, "loss": 0.4654, "step": 28900 }, { "epoch": 0.67, "learning_rate": 2.5091056300016428e-05, "loss": 0.2941, "step": 28905 }, { "epoch": 0.67, "learning_rate": 2.5090176245570393e-05, "loss": 0.1279, "step": 28910 }, { "epoch": 0.67, "learning_rate": 2.5089296191124358e-05, "loss": 0.0797, "step": 28915 }, { "epoch": 0.67, "learning_rate": 2.5088416136678323e-05, "loss": 0.1966, "step": 28920 }, { "epoch": 0.67, "learning_rate": 2.5087536082232288e-05, "loss": 0.2015, "step": 28925 }, { "epoch": 0.67, "learning_rate": 2.5086656027786253e-05, "loss": 0.2176, "step": 28930 }, { "epoch": 0.68, "learning_rate": 2.5085775973340218e-05, "loss": 0.1312, "step": 28935 }, { "epoch": 0.68, "learning_rate": 2.5084895918894183e-05, "loss": 0.3094, "step": 28940 }, { "epoch": 0.68, "learning_rate": 2.5084015864448148e-05, "loss": 0.2302, "step": 28945 }, { "epoch": 0.68, "learning_rate": 2.5083135810002113e-05, "loss": 0.6662, "step": 28950 }, { "epoch": 0.68, "learning_rate": 2.5082255755556075e-05, "loss": 0.1839, "step": 28955 }, { "epoch": 0.68, "learning_rate": 2.5081375701110043e-05, "loss": 0.1074, "step": 28960 }, { "epoch": 0.68, "learning_rate": 2.5080495646664008e-05, "loss": 0.215, "step": 28965 }, { "epoch": 0.68, "learning_rate": 2.5079615592217973e-05, "loss": 0.1373, "step": 28970 }, { "epoch": 0.68, "learning_rate": 2.5078735537771935e-05, "loss": 0.1668, "step": 28975 }, { "epoch": 0.68, "learning_rate": 2.5077855483325903e-05, "loss": 0.2188, "step": 28980 }, { "epoch": 0.68, "learning_rate": 2.5076975428879868e-05, "loss": 0.2054, "step": 28985 }, { "epoch": 0.68, "learning_rate": 2.5076095374433833e-05, "loss": 0.216, "step": 28990 }, { "epoch": 0.68, "learning_rate": 2.5075215319987795e-05, "loss": 0.3068, "step": 28995 }, { "epoch": 0.68, "learning_rate": 2.5074335265541763e-05, "loss": 0.4025, "step": 29000 }, { "epoch": 0.68, "learning_rate": 2.5073455211095728e-05, "loss": 0.1934, "step": 29005 }, { "epoch": 0.68, "learning_rate": 2.507257515664969e-05, "loss": 0.1289, "step": 29010 }, { "epoch": 0.68, "learning_rate": 2.507169510220366e-05, "loss": 0.1166, "step": 29015 }, { "epoch": 0.68, "learning_rate": 2.5070815047757623e-05, "loss": 0.2451, "step": 29020 }, { "epoch": 0.68, "learning_rate": 2.506993499331159e-05, "loss": 0.0921, "step": 29025 }, { "epoch": 0.68, "learning_rate": 2.506905493886555e-05, "loss": 0.106, "step": 29030 }, { "epoch": 0.68, "learning_rate": 2.506817488441952e-05, "loss": 0.4135, "step": 29035 }, { "epoch": 0.68, "learning_rate": 2.5067294829973483e-05, "loss": 0.3113, "step": 29040 }, { "epoch": 0.68, "learning_rate": 2.5066414775527445e-05, "loss": 0.4294, "step": 29045 }, { "epoch": 0.68, "learning_rate": 2.506553472108141e-05, "loss": 0.7999, "step": 29050 }, { "epoch": 0.68, "learning_rate": 2.506465466663538e-05, "loss": 0.209, "step": 29055 }, { "epoch": 0.68, "learning_rate": 2.5063774612189343e-05, "loss": 0.1545, "step": 29060 }, { "epoch": 0.68, "learning_rate": 2.5062894557743305e-05, "loss": 0.0817, "step": 29065 }, { "epoch": 0.68, "learning_rate": 2.506201450329727e-05, "loss": 0.1149, "step": 29070 }, { "epoch": 0.68, "learning_rate": 2.506113444885124e-05, "loss": 0.2108, "step": 29075 }, { "epoch": 0.68, "learning_rate": 2.50602543944052e-05, "loss": 0.145, "step": 29080 }, { "epoch": 0.68, "learning_rate": 2.5059374339959165e-05, "loss": 0.2485, "step": 29085 }, { "epoch": 0.68, "learning_rate": 2.505849428551313e-05, "loss": 0.2491, "step": 29090 }, { "epoch": 0.68, "learning_rate": 2.50576142310671e-05, "loss": 0.5114, "step": 29095 }, { "epoch": 0.68, "learning_rate": 2.505673417662106e-05, "loss": 0.6242, "step": 29100 }, { "epoch": 0.68, "learning_rate": 2.5055854122175025e-05, "loss": 0.296, "step": 29105 }, { "epoch": 0.68, "learning_rate": 2.505497406772899e-05, "loss": 0.1031, "step": 29110 }, { "epoch": 0.68, "learning_rate": 2.5054094013282955e-05, "loss": 0.1552, "step": 29115 }, { "epoch": 0.68, "learning_rate": 2.505321395883692e-05, "loss": 0.1006, "step": 29120 }, { "epoch": 0.68, "learning_rate": 2.5052333904390885e-05, "loss": 0.1023, "step": 29125 }, { "epoch": 0.68, "learning_rate": 2.505145384994485e-05, "loss": 0.2351, "step": 29130 }, { "epoch": 0.68, "learning_rate": 2.5050573795498815e-05, "loss": 0.1798, "step": 29135 }, { "epoch": 0.68, "learning_rate": 2.504969374105278e-05, "loss": 0.2296, "step": 29140 }, { "epoch": 0.68, "learning_rate": 2.5048813686606745e-05, "loss": 0.3408, "step": 29145 }, { "epoch": 0.68, "learning_rate": 2.504793363216071e-05, "loss": 0.7008, "step": 29150 }, { "epoch": 0.68, "learning_rate": 2.5047053577714675e-05, "loss": 0.3233, "step": 29155 }, { "epoch": 0.68, "learning_rate": 2.504617352326864e-05, "loss": 0.1077, "step": 29160 }, { "epoch": 0.68, "learning_rate": 2.5045293468822606e-05, "loss": 0.1328, "step": 29165 }, { "epoch": 0.68, "learning_rate": 2.504441341437657e-05, "loss": 0.1355, "step": 29170 }, { "epoch": 0.68, "learning_rate": 2.5043533359930536e-05, "loss": 0.1134, "step": 29175 }, { "epoch": 0.68, "learning_rate": 2.50426533054845e-05, "loss": 0.2788, "step": 29180 }, { "epoch": 0.68, "learning_rate": 2.5041773251038462e-05, "loss": 0.2449, "step": 29185 }, { "epoch": 0.68, "learning_rate": 2.504089319659243e-05, "loss": 0.3415, "step": 29190 }, { "epoch": 0.68, "learning_rate": 2.5040013142146396e-05, "loss": 0.3519, "step": 29195 }, { "epoch": 0.68, "learning_rate": 2.503913308770036e-05, "loss": 0.7382, "step": 29200 }, { "epoch": 0.68, "learning_rate": 2.5038253033254322e-05, "loss": 0.1198, "step": 29205 }, { "epoch": 0.68, "learning_rate": 2.503737297880829e-05, "loss": 0.0934, "step": 29210 }, { "epoch": 0.68, "learning_rate": 2.5036492924362256e-05, "loss": 0.058, "step": 29215 }, { "epoch": 0.68, "learning_rate": 2.5035612869916217e-05, "loss": 0.1726, "step": 29220 }, { "epoch": 0.68, "learning_rate": 2.5034732815470182e-05, "loss": 0.2293, "step": 29225 }, { "epoch": 0.68, "learning_rate": 2.503385276102415e-05, "loss": 0.2113, "step": 29230 }, { "epoch": 0.68, "learning_rate": 2.5032972706578116e-05, "loss": 0.3182, "step": 29235 }, { "epoch": 0.68, "learning_rate": 2.5032092652132077e-05, "loss": 0.4035, "step": 29240 }, { "epoch": 0.68, "learning_rate": 2.5031212597686042e-05, "loss": 0.3819, "step": 29245 }, { "epoch": 0.68, "learning_rate": 2.503033254324001e-05, "loss": 0.4912, "step": 29250 }, { "epoch": 0.68, "learning_rate": 2.5029452488793973e-05, "loss": 0.1959, "step": 29255 }, { "epoch": 0.68, "learning_rate": 2.5028572434347938e-05, "loss": 0.1432, "step": 29260 }, { "epoch": 0.68, "learning_rate": 2.5027692379901903e-05, "loss": 0.1496, "step": 29265 }, { "epoch": 0.68, "learning_rate": 2.502681232545587e-05, "loss": 0.1886, "step": 29270 }, { "epoch": 0.68, "learning_rate": 2.5025932271009833e-05, "loss": 0.2208, "step": 29275 }, { "epoch": 0.68, "learning_rate": 2.5025052216563798e-05, "loss": 0.2051, "step": 29280 }, { "epoch": 0.68, "learning_rate": 2.5024172162117766e-05, "loss": 0.1205, "step": 29285 }, { "epoch": 0.68, "learning_rate": 2.502329210767173e-05, "loss": 0.1672, "step": 29290 }, { "epoch": 0.68, "learning_rate": 2.5022412053225693e-05, "loss": 0.445, "step": 29295 }, { "epoch": 0.68, "learning_rate": 2.5021531998779658e-05, "loss": 0.5772, "step": 29300 }, { "epoch": 0.68, "learning_rate": 2.5020651944333626e-05, "loss": 0.0894, "step": 29305 }, { "epoch": 0.68, "learning_rate": 2.5019771889887588e-05, "loss": 0.0542, "step": 29310 }, { "epoch": 0.68, "learning_rate": 2.5018891835441553e-05, "loss": 0.1555, "step": 29315 }, { "epoch": 0.68, "learning_rate": 2.5018011780995518e-05, "loss": 0.1527, "step": 29320 }, { "epoch": 0.68, "learning_rate": 2.5017131726549486e-05, "loss": 0.098, "step": 29325 }, { "epoch": 0.68, "learning_rate": 2.5016251672103448e-05, "loss": 0.2431, "step": 29330 }, { "epoch": 0.68, "learning_rate": 2.5015371617657413e-05, "loss": 0.2407, "step": 29335 }, { "epoch": 0.68, "learning_rate": 2.5014491563211378e-05, "loss": 0.1672, "step": 29340 }, { "epoch": 0.68, "learning_rate": 2.5013611508765343e-05, "loss": 0.4333, "step": 29345 }, { "epoch": 0.68, "learning_rate": 2.5012731454319308e-05, "loss": 0.4488, "step": 29350 }, { "epoch": 0.68, "learning_rate": 2.5011851399873273e-05, "loss": 0.2442, "step": 29355 }, { "epoch": 0.68, "learning_rate": 2.5010971345427238e-05, "loss": 0.1068, "step": 29360 }, { "epoch": 0.69, "learning_rate": 2.5010091290981203e-05, "loss": 0.0435, "step": 29365 }, { "epoch": 0.69, "learning_rate": 2.5009211236535168e-05, "loss": 0.1978, "step": 29370 }, { "epoch": 0.69, "learning_rate": 2.5008331182089133e-05, "loss": 0.1689, "step": 29375 }, { "epoch": 0.69, "learning_rate": 2.5007451127643095e-05, "loss": 0.1545, "step": 29380 }, { "epoch": 0.69, "learning_rate": 2.5006571073197063e-05, "loss": 0.2813, "step": 29385 }, { "epoch": 0.69, "learning_rate": 2.5005691018751028e-05, "loss": 0.3314, "step": 29390 }, { "epoch": 0.69, "learning_rate": 2.5004810964304993e-05, "loss": 0.2827, "step": 29395 }, { "epoch": 0.69, "learning_rate": 2.5003930909858955e-05, "loss": 0.428, "step": 29400 }, { "epoch": 0.69, "learning_rate": 2.5003050855412923e-05, "loss": 0.158, "step": 29405 }, { "epoch": 0.69, "learning_rate": 2.5002170800966888e-05, "loss": 0.0908, "step": 29410 }, { "epoch": 0.69, "learning_rate": 2.500129074652085e-05, "loss": 0.0946, "step": 29415 }, { "epoch": 0.69, "learning_rate": 2.5000410692074818e-05, "loss": 0.1669, "step": 29420 }, { "epoch": 0.69, "learning_rate": 2.4999530637628783e-05, "loss": 0.1357, "step": 29425 }, { "epoch": 0.69, "learning_rate": 2.4998650583182748e-05, "loss": 0.1458, "step": 29430 }, { "epoch": 0.69, "learning_rate": 2.499777052873671e-05, "loss": 0.1369, "step": 29435 }, { "epoch": 0.69, "learning_rate": 2.4996890474290678e-05, "loss": 0.1543, "step": 29440 }, { "epoch": 0.69, "learning_rate": 2.4996010419844643e-05, "loss": 0.4146, "step": 29445 }, { "epoch": 0.69, "learning_rate": 2.4995130365398605e-05, "loss": 0.6137, "step": 29450 }, { "epoch": 0.69, "learning_rate": 2.499425031095257e-05, "loss": 0.2757, "step": 29455 }, { "epoch": 0.69, "learning_rate": 2.499337025650654e-05, "loss": 0.0669, "step": 29460 }, { "epoch": 0.69, "learning_rate": 2.4992490202060503e-05, "loss": 0.1661, "step": 29465 }, { "epoch": 0.69, "learning_rate": 2.4991610147614465e-05, "loss": 0.0765, "step": 29470 }, { "epoch": 0.69, "learning_rate": 2.499073009316843e-05, "loss": 0.1014, "step": 29475 }, { "epoch": 0.69, "learning_rate": 2.49898500387224e-05, "loss": 0.1307, "step": 29480 }, { "epoch": 0.69, "learning_rate": 2.498896998427636e-05, "loss": 0.369, "step": 29485 }, { "epoch": 0.69, "learning_rate": 2.4988089929830325e-05, "loss": 0.2399, "step": 29490 }, { "epoch": 0.69, "learning_rate": 2.498720987538429e-05, "loss": 0.2819, "step": 29495 }, { "epoch": 0.69, "learning_rate": 2.498632982093826e-05, "loss": 0.5956, "step": 29500 }, { "epoch": 0.69, "learning_rate": 2.498544976649222e-05, "loss": 0.1914, "step": 29505 }, { "epoch": 0.69, "learning_rate": 2.4984569712046185e-05, "loss": 0.1163, "step": 29510 }, { "epoch": 0.69, "learning_rate": 2.498368965760015e-05, "loss": 0.0602, "step": 29515 }, { "epoch": 0.69, "learning_rate": 2.4982809603154115e-05, "loss": 0.1386, "step": 29520 }, { "epoch": 0.69, "learning_rate": 2.498192954870808e-05, "loss": 0.1481, "step": 29525 }, { "epoch": 0.69, "learning_rate": 2.4981049494262045e-05, "loss": 0.2363, "step": 29530 }, { "epoch": 0.69, "learning_rate": 2.498016943981601e-05, "loss": 0.2544, "step": 29535 }, { "epoch": 0.69, "learning_rate": 2.4979289385369975e-05, "loss": 0.2763, "step": 29540 }, { "epoch": 0.69, "learning_rate": 2.497840933092394e-05, "loss": 0.2113, "step": 29545 }, { "epoch": 0.69, "learning_rate": 2.4977529276477905e-05, "loss": 0.558, "step": 29550 }, { "epoch": 0.69, "learning_rate": 2.497664922203187e-05, "loss": 0.3093, "step": 29555 }, { "epoch": 0.69, "learning_rate": 2.4975769167585835e-05, "loss": 0.1036, "step": 29560 }, { "epoch": 0.69, "learning_rate": 2.49748891131398e-05, "loss": 0.1353, "step": 29565 }, { "epoch": 0.69, "learning_rate": 2.4974009058693765e-05, "loss": 0.0806, "step": 29570 }, { "epoch": 0.69, "learning_rate": 2.497312900424773e-05, "loss": 0.1156, "step": 29575 }, { "epoch": 0.69, "learning_rate": 2.4972248949801695e-05, "loss": 0.1462, "step": 29580 }, { "epoch": 0.69, "learning_rate": 2.497136889535566e-05, "loss": 0.2664, "step": 29585 }, { "epoch": 0.69, "learning_rate": 2.4970488840909622e-05, "loss": 0.1827, "step": 29590 }, { "epoch": 0.69, "learning_rate": 2.496960878646359e-05, "loss": 0.241, "step": 29595 }, { "epoch": 0.69, "learning_rate": 2.4968728732017556e-05, "loss": 0.4571, "step": 29600 }, { "epoch": 0.69, "learning_rate": 2.496784867757152e-05, "loss": 0.2322, "step": 29605 }, { "epoch": 0.69, "learning_rate": 2.4966968623125482e-05, "loss": 0.0678, "step": 29610 }, { "epoch": 0.69, "learning_rate": 2.496608856867945e-05, "loss": 0.1321, "step": 29615 }, { "epoch": 0.69, "learning_rate": 2.4965208514233416e-05, "loss": 0.1061, "step": 29620 }, { "epoch": 0.69, "learning_rate": 2.496432845978738e-05, "loss": 0.1421, "step": 29625 }, { "epoch": 0.69, "learning_rate": 2.4963448405341342e-05, "loss": 0.2464, "step": 29630 }, { "epoch": 0.69, "learning_rate": 2.496256835089531e-05, "loss": 0.258, "step": 29635 }, { "epoch": 0.69, "learning_rate": 2.4961688296449276e-05, "loss": 0.2022, "step": 29640 }, { "epoch": 0.69, "learning_rate": 2.4960808242003237e-05, "loss": 0.2158, "step": 29645 }, { "epoch": 0.69, "learning_rate": 2.4959928187557202e-05, "loss": 0.3969, "step": 29650 }, { "epoch": 0.69, "learning_rate": 2.495904813311117e-05, "loss": 0.2724, "step": 29655 }, { "epoch": 0.69, "learning_rate": 2.4958168078665136e-05, "loss": 0.0882, "step": 29660 }, { "epoch": 0.69, "learning_rate": 2.4957288024219097e-05, "loss": 0.0817, "step": 29665 }, { "epoch": 0.69, "learning_rate": 2.4956407969773062e-05, "loss": 0.0884, "step": 29670 }, { "epoch": 0.69, "learning_rate": 2.495552791532703e-05, "loss": 0.1974, "step": 29675 }, { "epoch": 0.69, "learning_rate": 2.4954647860880992e-05, "loss": 0.1221, "step": 29680 }, { "epoch": 0.69, "learning_rate": 2.4953767806434958e-05, "loss": 0.2776, "step": 29685 }, { "epoch": 0.69, "learning_rate": 2.4952887751988926e-05, "loss": 0.3529, "step": 29690 }, { "epoch": 0.69, "learning_rate": 2.495200769754289e-05, "loss": 0.2811, "step": 29695 }, { "epoch": 0.69, "learning_rate": 2.4951127643096853e-05, "loss": 0.6524, "step": 29700 }, { "epoch": 0.69, "learning_rate": 2.4950247588650818e-05, "loss": 0.2417, "step": 29705 }, { "epoch": 0.69, "learning_rate": 2.4949367534204786e-05, "loss": 0.0576, "step": 29710 }, { "epoch": 0.69, "learning_rate": 2.4948487479758748e-05, "loss": 0.1241, "step": 29715 }, { "epoch": 0.69, "learning_rate": 2.4947607425312713e-05, "loss": 0.1972, "step": 29720 }, { "epoch": 0.69, "learning_rate": 2.4946727370866678e-05, "loss": 0.1236, "step": 29725 }, { "epoch": 0.69, "learning_rate": 2.4945847316420646e-05, "loss": 0.1933, "step": 29730 }, { "epoch": 0.69, "learning_rate": 2.4944967261974608e-05, "loss": 0.2097, "step": 29735 }, { "epoch": 0.69, "learning_rate": 2.4944087207528573e-05, "loss": 0.095, "step": 29740 }, { "epoch": 0.69, "learning_rate": 2.4943207153082538e-05, "loss": 0.3438, "step": 29745 }, { "epoch": 0.69, "learning_rate": 2.4942327098636503e-05, "loss": 0.6862, "step": 29750 }, { "epoch": 0.69, "learning_rate": 2.4941447044190468e-05, "loss": 0.2116, "step": 29755 }, { "epoch": 0.69, "learning_rate": 2.4940566989744433e-05, "loss": 0.0656, "step": 29760 }, { "epoch": 0.69, "learning_rate": 2.4939686935298398e-05, "loss": 0.0667, "step": 29765 }, { "epoch": 0.69, "learning_rate": 2.4938806880852363e-05, "loss": 0.1272, "step": 29770 }, { "epoch": 0.69, "learning_rate": 2.4937926826406328e-05, "loss": 0.1584, "step": 29775 }, { "epoch": 0.69, "learning_rate": 2.4937046771960293e-05, "loss": 0.222, "step": 29780 }, { "epoch": 0.69, "learning_rate": 2.4936166717514255e-05, "loss": 0.1624, "step": 29785 }, { "epoch": 0.7, "learning_rate": 2.4935286663068223e-05, "loss": 0.1308, "step": 29790 }, { "epoch": 0.7, "learning_rate": 2.4934406608622188e-05, "loss": 0.2716, "step": 29795 }, { "epoch": 0.7, "learning_rate": 2.4933526554176153e-05, "loss": 0.482, "step": 29800 }, { "epoch": 0.7, "learning_rate": 2.4932646499730115e-05, "loss": 0.2035, "step": 29805 }, { "epoch": 0.7, "learning_rate": 2.4931766445284083e-05, "loss": 0.1546, "step": 29810 }, { "epoch": 0.7, "learning_rate": 2.4930886390838048e-05, "loss": 0.0651, "step": 29815 }, { "epoch": 0.7, "learning_rate": 2.493000633639201e-05, "loss": 0.0596, "step": 29820 }, { "epoch": 0.7, "learning_rate": 2.4929126281945978e-05, "loss": 0.1972, "step": 29825 }, { "epoch": 0.7, "learning_rate": 2.4928246227499943e-05, "loss": 0.341, "step": 29830 }, { "epoch": 0.7, "learning_rate": 2.4927366173053908e-05, "loss": 0.2488, "step": 29835 }, { "epoch": 0.7, "learning_rate": 2.492648611860787e-05, "loss": 0.1762, "step": 29840 }, { "epoch": 0.7, "learning_rate": 2.4925606064161838e-05, "loss": 0.2627, "step": 29845 }, { "epoch": 0.7, "learning_rate": 2.4924726009715803e-05, "loss": 0.4141, "step": 29850 }, { "epoch": 0.7, "learning_rate": 2.4923845955269765e-05, "loss": 0.1535, "step": 29855 }, { "epoch": 0.7, "learning_rate": 2.492296590082373e-05, "loss": 0.0908, "step": 29860 }, { "epoch": 0.7, "learning_rate": 2.4922085846377698e-05, "loss": 0.0908, "step": 29865 }, { "epoch": 0.7, "learning_rate": 2.4921205791931663e-05, "loss": 0.1961, "step": 29870 }, { "epoch": 0.7, "learning_rate": 2.4920325737485625e-05, "loss": 0.1797, "step": 29875 }, { "epoch": 0.7, "learning_rate": 2.491944568303959e-05, "loss": 0.2071, "step": 29880 }, { "epoch": 0.7, "learning_rate": 2.491856562859356e-05, "loss": 0.2176, "step": 29885 }, { "epoch": 0.7, "learning_rate": 2.491768557414752e-05, "loss": 0.3287, "step": 29890 }, { "epoch": 0.7, "learning_rate": 2.4916805519701485e-05, "loss": 0.5702, "step": 29895 }, { "epoch": 0.7, "learning_rate": 2.491592546525545e-05, "loss": 0.5532, "step": 29900 }, { "epoch": 0.7, "learning_rate": 2.491504541080942e-05, "loss": 0.2418, "step": 29905 }, { "epoch": 0.7, "learning_rate": 2.491416535636338e-05, "loss": 0.0485, "step": 29910 }, { "epoch": 0.7, "learning_rate": 2.4913285301917345e-05, "loss": 0.1577, "step": 29915 }, { "epoch": 0.7, "learning_rate": 2.491240524747131e-05, "loss": 0.0959, "step": 29920 }, { "epoch": 0.7, "learning_rate": 2.491152519302528e-05, "loss": 0.1177, "step": 29925 }, { "epoch": 0.7, "learning_rate": 2.491064513857924e-05, "loss": 0.2355, "step": 29930 }, { "epoch": 0.7, "learning_rate": 2.4909765084133205e-05, "loss": 0.17, "step": 29935 }, { "epoch": 0.7, "learning_rate": 2.490888502968717e-05, "loss": 0.3821, "step": 29940 }, { "epoch": 0.7, "learning_rate": 2.4908004975241135e-05, "loss": 0.2716, "step": 29945 }, { "epoch": 0.7, "learning_rate": 2.49071249207951e-05, "loss": 0.475, "step": 29950 }, { "epoch": 0.7, "learning_rate": 2.4906244866349065e-05, "loss": 0.3011, "step": 29955 }, { "epoch": 0.7, "learning_rate": 2.490536481190303e-05, "loss": 0.0767, "step": 29960 }, { "epoch": 0.7, "learning_rate": 2.4904484757456995e-05, "loss": 0.0817, "step": 29965 }, { "epoch": 0.7, "learning_rate": 2.490360470301096e-05, "loss": 0.1099, "step": 29970 }, { "epoch": 0.7, "learning_rate": 2.4902724648564925e-05, "loss": 0.1303, "step": 29975 }, { "epoch": 0.7, "learning_rate": 2.490184459411889e-05, "loss": 0.22, "step": 29980 }, { "epoch": 0.7, "learning_rate": 2.4900964539672855e-05, "loss": 0.2955, "step": 29985 }, { "epoch": 0.7, "learning_rate": 2.490008448522682e-05, "loss": 0.2131, "step": 29990 }, { "epoch": 0.7, "learning_rate": 2.4899204430780785e-05, "loss": 0.4813, "step": 29995 }, { "epoch": 0.7, "learning_rate": 2.489832437633475e-05, "loss": 0.2988, "step": 30000 }, { "epoch": 0.7, "learning_rate": 2.4897444321888715e-05, "loss": 0.2256, "step": 30005 }, { "epoch": 0.7, "learning_rate": 2.489656426744268e-05, "loss": 0.0878, "step": 30010 }, { "epoch": 0.7, "learning_rate": 2.4895684212996642e-05, "loss": 0.0265, "step": 30015 }, { "epoch": 0.7, "learning_rate": 2.489480415855061e-05, "loss": 0.0921, "step": 30020 }, { "epoch": 0.7, "learning_rate": 2.4893924104104576e-05, "loss": 0.1541, "step": 30025 }, { "epoch": 0.7, "learning_rate": 2.489304404965854e-05, "loss": 0.2447, "step": 30030 }, { "epoch": 0.7, "learning_rate": 2.4892163995212502e-05, "loss": 0.1218, "step": 30035 }, { "epoch": 0.7, "learning_rate": 2.489128394076647e-05, "loss": 0.4072, "step": 30040 }, { "epoch": 0.7, "learning_rate": 2.4890403886320436e-05, "loss": 0.2261, "step": 30045 }, { "epoch": 0.7, "learning_rate": 2.4889523831874397e-05, "loss": 0.4453, "step": 30050 }, { "epoch": 0.7, "learning_rate": 2.4888643777428362e-05, "loss": 0.2603, "step": 30055 }, { "epoch": 0.7, "learning_rate": 2.488776372298233e-05, "loss": 0.097, "step": 30060 }, { "epoch": 0.7, "learning_rate": 2.4886883668536296e-05, "loss": 0.076, "step": 30065 }, { "epoch": 0.7, "learning_rate": 2.4886003614090257e-05, "loss": 0.1487, "step": 30070 }, { "epoch": 0.7, "learning_rate": 2.4885123559644222e-05, "loss": 0.2239, "step": 30075 }, { "epoch": 0.7, "learning_rate": 2.488424350519819e-05, "loss": 0.3869, "step": 30080 }, { "epoch": 0.7, "learning_rate": 2.4883363450752152e-05, "loss": 0.2061, "step": 30085 }, { "epoch": 0.7, "learning_rate": 2.4882483396306117e-05, "loss": 0.1941, "step": 30090 }, { "epoch": 0.7, "learning_rate": 2.4881603341860086e-05, "loss": 0.3543, "step": 30095 }, { "epoch": 0.7, "learning_rate": 2.488072328741405e-05, "loss": 0.5418, "step": 30100 }, { "epoch": 0.7, "learning_rate": 2.4879843232968012e-05, "loss": 0.168, "step": 30105 }, { "epoch": 0.7, "learning_rate": 2.4878963178521977e-05, "loss": 0.1292, "step": 30110 }, { "epoch": 0.7, "learning_rate": 2.4878083124075946e-05, "loss": 0.1322, "step": 30115 }, { "epoch": 0.7, "learning_rate": 2.4877203069629908e-05, "loss": 0.0764, "step": 30120 }, { "epoch": 0.7, "learning_rate": 2.4876323015183873e-05, "loss": 0.2196, "step": 30125 }, { "epoch": 0.7, "learning_rate": 2.4875442960737838e-05, "loss": 0.14, "step": 30130 }, { "epoch": 0.7, "learning_rate": 2.4874562906291806e-05, "loss": 0.1224, "step": 30135 }, { "epoch": 0.7, "learning_rate": 2.4873682851845768e-05, "loss": 0.3261, "step": 30140 }, { "epoch": 0.7, "learning_rate": 2.4872802797399733e-05, "loss": 0.3097, "step": 30145 }, { "epoch": 0.7, "learning_rate": 2.4871922742953698e-05, "loss": 0.4637, "step": 30150 }, { "epoch": 0.7, "learning_rate": 2.4871042688507663e-05, "loss": 0.1803, "step": 30155 }, { "epoch": 0.7, "learning_rate": 2.4870162634061628e-05, "loss": 0.0595, "step": 30160 }, { "epoch": 0.7, "learning_rate": 2.4869282579615593e-05, "loss": 0.1129, "step": 30165 }, { "epoch": 0.7, "learning_rate": 2.4868402525169558e-05, "loss": 0.0585, "step": 30170 }, { "epoch": 0.7, "learning_rate": 2.4867522470723523e-05, "loss": 0.1917, "step": 30175 }, { "epoch": 0.7, "learning_rate": 2.4866642416277488e-05, "loss": 0.1626, "step": 30180 }, { "epoch": 0.7, "learning_rate": 2.4865762361831453e-05, "loss": 0.2318, "step": 30185 }, { "epoch": 0.7, "learning_rate": 2.4864882307385414e-05, "loss": 0.2363, "step": 30190 }, { "epoch": 0.7, "learning_rate": 2.4864002252939383e-05, "loss": 0.4898, "step": 30195 }, { "epoch": 0.7, "learning_rate": 2.4863122198493348e-05, "loss": 0.5963, "step": 30200 }, { "epoch": 0.7, "learning_rate": 2.4862242144047313e-05, "loss": 0.2685, "step": 30205 }, { "epoch": 0.7, "learning_rate": 2.4861362089601275e-05, "loss": 0.0772, "step": 30210 }, { "epoch": 0.7, "learning_rate": 2.4860482035155243e-05, "loss": 0.114, "step": 30215 }, { "epoch": 0.71, "learning_rate": 2.4859601980709208e-05, "loss": 0.1969, "step": 30220 }, { "epoch": 0.71, "learning_rate": 2.4858721926263173e-05, "loss": 0.1503, "step": 30225 }, { "epoch": 0.71, "learning_rate": 2.4857841871817138e-05, "loss": 0.2744, "step": 30230 }, { "epoch": 0.71, "learning_rate": 2.4856961817371103e-05, "loss": 0.2891, "step": 30235 }, { "epoch": 0.71, "learning_rate": 2.4856081762925068e-05, "loss": 0.1205, "step": 30240 }, { "epoch": 0.71, "learning_rate": 2.485520170847903e-05, "loss": 0.3723, "step": 30245 }, { "epoch": 0.71, "learning_rate": 2.4854321654032998e-05, "loss": 0.4018, "step": 30250 }, { "epoch": 0.71, "learning_rate": 2.4853441599586963e-05, "loss": 0.3187, "step": 30255 }, { "epoch": 0.71, "learning_rate": 2.4852561545140928e-05, "loss": 0.0577, "step": 30260 }, { "epoch": 0.71, "learning_rate": 2.485168149069489e-05, "loss": 0.0852, "step": 30265 }, { "epoch": 0.71, "learning_rate": 2.4850801436248858e-05, "loss": 0.1017, "step": 30270 }, { "epoch": 0.71, "learning_rate": 2.4849921381802823e-05, "loss": 0.0726, "step": 30275 }, { "epoch": 0.71, "learning_rate": 2.4849041327356785e-05, "loss": 0.233, "step": 30280 }, { "epoch": 0.71, "learning_rate": 2.484816127291075e-05, "loss": 0.1702, "step": 30285 }, { "epoch": 0.71, "learning_rate": 2.4847281218464718e-05, "loss": 0.3215, "step": 30290 }, { "epoch": 0.71, "learning_rate": 2.4846401164018683e-05, "loss": 0.1843, "step": 30295 }, { "epoch": 0.71, "learning_rate": 2.4845521109572645e-05, "loss": 0.4998, "step": 30300 }, { "epoch": 0.71, "learning_rate": 2.484464105512661e-05, "loss": 0.213, "step": 30305 }, { "epoch": 0.71, "learning_rate": 2.4843761000680578e-05, "loss": 0.0954, "step": 30310 }, { "epoch": 0.71, "learning_rate": 2.484288094623454e-05, "loss": 0.1391, "step": 30315 }, { "epoch": 0.71, "learning_rate": 2.4842000891788505e-05, "loss": 0.0771, "step": 30320 }, { "epoch": 0.71, "learning_rate": 2.484112083734247e-05, "loss": 0.1055, "step": 30325 }, { "epoch": 0.71, "learning_rate": 2.484024078289644e-05, "loss": 0.1572, "step": 30330 }, { "epoch": 0.71, "learning_rate": 2.48393607284504e-05, "loss": 0.1578, "step": 30335 }, { "epoch": 0.71, "learning_rate": 2.4838480674004365e-05, "loss": 0.3117, "step": 30340 }, { "epoch": 0.71, "learning_rate": 2.483760061955833e-05, "loss": 0.053, "step": 30345 }, { "epoch": 0.71, "learning_rate": 2.4836720565112295e-05, "loss": 0.5864, "step": 30350 }, { "epoch": 0.71, "learning_rate": 2.483584051066626e-05, "loss": 0.1931, "step": 30355 }, { "epoch": 0.71, "learning_rate": 2.4834960456220225e-05, "loss": 0.0294, "step": 30360 }, { "epoch": 0.71, "learning_rate": 2.483408040177419e-05, "loss": 0.1868, "step": 30365 }, { "epoch": 0.71, "learning_rate": 2.4833200347328155e-05, "loss": 0.1711, "step": 30370 }, { "epoch": 0.71, "learning_rate": 2.483232029288212e-05, "loss": 0.2312, "step": 30375 }, { "epoch": 0.71, "learning_rate": 2.4831440238436085e-05, "loss": 0.2954, "step": 30380 }, { "epoch": 0.71, "learning_rate": 2.483056018399005e-05, "loss": 0.2451, "step": 30385 }, { "epoch": 0.71, "learning_rate": 2.4829680129544015e-05, "loss": 0.1825, "step": 30390 }, { "epoch": 0.71, "learning_rate": 2.482880007509798e-05, "loss": 0.2359, "step": 30395 }, { "epoch": 0.71, "learning_rate": 2.4827920020651945e-05, "loss": 0.5644, "step": 30400 }, { "epoch": 0.71, "learning_rate": 2.482703996620591e-05, "loss": 0.1347, "step": 30405 }, { "epoch": 0.71, "learning_rate": 2.4826159911759875e-05, "loss": 0.1237, "step": 30410 }, { "epoch": 0.71, "learning_rate": 2.482527985731384e-05, "loss": 0.1156, "step": 30415 }, { "epoch": 0.71, "learning_rate": 2.4824399802867802e-05, "loss": 0.1466, "step": 30420 }, { "epoch": 0.71, "learning_rate": 2.482351974842177e-05, "loss": 0.2163, "step": 30425 }, { "epoch": 0.71, "learning_rate": 2.4822639693975735e-05, "loss": 0.2077, "step": 30430 }, { "epoch": 0.71, "learning_rate": 2.48217596395297e-05, "loss": 0.1453, "step": 30435 }, { "epoch": 0.71, "learning_rate": 2.4820879585083662e-05, "loss": 0.3272, "step": 30440 }, { "epoch": 0.71, "learning_rate": 2.481999953063763e-05, "loss": 0.4248, "step": 30445 }, { "epoch": 0.71, "learning_rate": 2.4819119476191595e-05, "loss": 0.6534, "step": 30450 }, { "epoch": 0.71, "learning_rate": 2.4818239421745557e-05, "loss": 0.2566, "step": 30455 }, { "epoch": 0.71, "learning_rate": 2.4817359367299522e-05, "loss": 0.0859, "step": 30460 }, { "epoch": 0.71, "learning_rate": 2.481647931285349e-05, "loss": 0.1808, "step": 30465 }, { "epoch": 0.71, "learning_rate": 2.4815599258407456e-05, "loss": 0.0503, "step": 30470 }, { "epoch": 0.71, "learning_rate": 2.4814719203961417e-05, "loss": 0.1157, "step": 30475 }, { "epoch": 0.71, "learning_rate": 2.4813839149515382e-05, "loss": 0.0999, "step": 30480 }, { "epoch": 0.71, "learning_rate": 2.481295909506935e-05, "loss": 0.2599, "step": 30485 }, { "epoch": 0.71, "learning_rate": 2.4812079040623312e-05, "loss": 0.231, "step": 30490 }, { "epoch": 0.71, "learning_rate": 2.4811198986177277e-05, "loss": 0.3623, "step": 30495 }, { "epoch": 0.71, "learning_rate": 2.4810318931731242e-05, "loss": 0.4466, "step": 30500 }, { "epoch": 0.71, "learning_rate": 2.480943887728521e-05, "loss": 0.1933, "step": 30505 }, { "epoch": 0.71, "learning_rate": 2.4808558822839172e-05, "loss": 0.0358, "step": 30510 }, { "epoch": 0.71, "learning_rate": 2.4807678768393137e-05, "loss": 0.1067, "step": 30515 }, { "epoch": 0.71, "learning_rate": 2.4806798713947106e-05, "loss": 0.0288, "step": 30520 }, { "epoch": 0.71, "learning_rate": 2.480591865950107e-05, "loss": 0.1232, "step": 30525 }, { "epoch": 0.71, "learning_rate": 2.4805038605055032e-05, "loss": 0.1119, "step": 30530 }, { "epoch": 0.71, "learning_rate": 2.4804158550608997e-05, "loss": 0.3514, "step": 30535 }, { "epoch": 0.71, "learning_rate": 2.4803278496162966e-05, "loss": 0.3358, "step": 30540 }, { "epoch": 0.71, "learning_rate": 2.4802398441716927e-05, "loss": 0.4083, "step": 30545 }, { "epoch": 0.71, "learning_rate": 2.4801518387270893e-05, "loss": 0.7451, "step": 30550 }, { "epoch": 0.71, "learning_rate": 2.4800638332824858e-05, "loss": 0.3048, "step": 30555 }, { "epoch": 0.71, "learning_rate": 2.4799758278378826e-05, "loss": 0.1316, "step": 30560 }, { "epoch": 0.71, "learning_rate": 2.4798878223932788e-05, "loss": 0.1288, "step": 30565 }, { "epoch": 0.71, "learning_rate": 2.4797998169486753e-05, "loss": 0.0943, "step": 30570 }, { "epoch": 0.71, "learning_rate": 2.4797118115040718e-05, "loss": 0.1669, "step": 30575 }, { "epoch": 0.71, "learning_rate": 2.4796238060594683e-05, "loss": 0.1942, "step": 30580 }, { "epoch": 0.71, "learning_rate": 2.4795358006148648e-05, "loss": 0.128, "step": 30585 }, { "epoch": 0.71, "learning_rate": 2.4794477951702613e-05, "loss": 0.1689, "step": 30590 }, { "epoch": 0.71, "learning_rate": 2.4793597897256578e-05, "loss": 0.3701, "step": 30595 }, { "epoch": 0.71, "learning_rate": 2.4792717842810543e-05, "loss": 0.4191, "step": 30600 }, { "epoch": 0.71, "learning_rate": 2.4791837788364508e-05, "loss": 0.1671, "step": 30605 }, { "epoch": 0.71, "learning_rate": 2.4790957733918473e-05, "loss": 0.168, "step": 30610 }, { "epoch": 0.71, "learning_rate": 2.4790077679472434e-05, "loss": 0.0396, "step": 30615 }, { "epoch": 0.71, "learning_rate": 2.4789197625026403e-05, "loss": 0.0985, "step": 30620 }, { "epoch": 0.71, "learning_rate": 2.4788317570580368e-05, "loss": 0.0644, "step": 30625 }, { "epoch": 0.71, "learning_rate": 2.4787437516134333e-05, "loss": 0.2996, "step": 30630 }, { "epoch": 0.71, "learning_rate": 2.4786557461688298e-05, "loss": 0.2302, "step": 30635 }, { "epoch": 0.71, "learning_rate": 2.4785677407242263e-05, "loss": 0.1686, "step": 30640 }, { "epoch": 0.71, "learning_rate": 2.4784797352796228e-05, "loss": 0.3588, "step": 30645 }, { "epoch": 0.72, "learning_rate": 2.478391729835019e-05, "loss": 0.3774, "step": 30650 }, { "epoch": 0.72, "learning_rate": 2.4783037243904158e-05, "loss": 0.1993, "step": 30655 }, { "epoch": 0.72, "learning_rate": 2.4782157189458123e-05, "loss": 0.0406, "step": 30660 }, { "epoch": 0.72, "learning_rate": 2.4781277135012088e-05, "loss": 0.0518, "step": 30665 }, { "epoch": 0.72, "learning_rate": 2.478039708056605e-05, "loss": 0.3207, "step": 30670 }, { "epoch": 0.72, "learning_rate": 2.4779517026120018e-05, "loss": 0.1258, "step": 30675 }, { "epoch": 0.72, "learning_rate": 2.4778636971673983e-05, "loss": 0.3466, "step": 30680 }, { "epoch": 0.72, "learning_rate": 2.4777756917227945e-05, "loss": 0.1972, "step": 30685 }, { "epoch": 0.72, "learning_rate": 2.477687686278191e-05, "loss": 0.1773, "step": 30690 }, { "epoch": 0.72, "learning_rate": 2.4775996808335878e-05, "loss": 0.4297, "step": 30695 }, { "epoch": 0.72, "learning_rate": 2.4775116753889843e-05, "loss": 0.6673, "step": 30700 }, { "epoch": 0.72, "learning_rate": 2.4774236699443805e-05, "loss": 0.3159, "step": 30705 }, { "epoch": 0.72, "learning_rate": 2.477335664499777e-05, "loss": 0.0941, "step": 30710 }, { "epoch": 0.72, "learning_rate": 2.4772476590551738e-05, "loss": 0.0982, "step": 30715 }, { "epoch": 0.72, "learning_rate": 2.47715965361057e-05, "loss": 0.2025, "step": 30720 }, { "epoch": 0.72, "learning_rate": 2.4770716481659665e-05, "loss": 0.1773, "step": 30725 }, { "epoch": 0.72, "learning_rate": 2.476983642721363e-05, "loss": 0.1528, "step": 30730 }, { "epoch": 0.72, "learning_rate": 2.4768956372767598e-05, "loss": 0.3506, "step": 30735 }, { "epoch": 0.72, "learning_rate": 2.476807631832156e-05, "loss": 0.2819, "step": 30740 }, { "epoch": 0.72, "learning_rate": 2.4767196263875525e-05, "loss": 0.3687, "step": 30745 }, { "epoch": 0.72, "learning_rate": 2.476631620942949e-05, "loss": 0.5964, "step": 30750 }, { "epoch": 0.72, "learning_rate": 2.4765436154983455e-05, "loss": 0.259, "step": 30755 }, { "epoch": 0.72, "learning_rate": 2.476455610053742e-05, "loss": 0.0533, "step": 30760 }, { "epoch": 0.72, "learning_rate": 2.4763676046091385e-05, "loss": 0.0298, "step": 30765 }, { "epoch": 0.72, "learning_rate": 2.476279599164535e-05, "loss": 0.1386, "step": 30770 }, { "epoch": 0.72, "learning_rate": 2.4761915937199315e-05, "loss": 0.2099, "step": 30775 }, { "epoch": 0.72, "learning_rate": 2.476103588275328e-05, "loss": 0.2589, "step": 30780 }, { "epoch": 0.72, "learning_rate": 2.4760155828307245e-05, "loss": 0.1689, "step": 30785 }, { "epoch": 0.72, "learning_rate": 2.475927577386121e-05, "loss": 0.1835, "step": 30790 }, { "epoch": 0.72, "learning_rate": 2.4758395719415175e-05, "loss": 0.1514, "step": 30795 }, { "epoch": 0.72, "learning_rate": 2.475751566496914e-05, "loss": 0.4182, "step": 30800 }, { "epoch": 0.72, "learning_rate": 2.4756635610523105e-05, "loss": 0.2466, "step": 30805 }, { "epoch": 0.72, "learning_rate": 2.475575555607707e-05, "loss": 0.052, "step": 30810 }, { "epoch": 0.72, "learning_rate": 2.4754875501631035e-05, "loss": 0.1174, "step": 30815 }, { "epoch": 0.72, "learning_rate": 2.4753995447185e-05, "loss": 0.1313, "step": 30820 }, { "epoch": 0.72, "learning_rate": 2.4753115392738965e-05, "loss": 0.095, "step": 30825 }, { "epoch": 0.72, "learning_rate": 2.475223533829293e-05, "loss": 0.2821, "step": 30830 }, { "epoch": 0.72, "learning_rate": 2.4751355283846895e-05, "loss": 0.1437, "step": 30835 }, { "epoch": 0.72, "learning_rate": 2.475047522940086e-05, "loss": 0.3566, "step": 30840 }, { "epoch": 0.72, "learning_rate": 2.4749595174954822e-05, "loss": 0.4034, "step": 30845 }, { "epoch": 0.72, "learning_rate": 2.474871512050879e-05, "loss": 0.5444, "step": 30850 }, { "epoch": 0.72, "learning_rate": 2.4747835066062755e-05, "loss": 0.1858, "step": 30855 }, { "epoch": 0.72, "learning_rate": 2.474695501161672e-05, "loss": 0.2193, "step": 30860 }, { "epoch": 0.72, "learning_rate": 2.4746074957170682e-05, "loss": 0.0854, "step": 30865 }, { "epoch": 0.72, "learning_rate": 2.474519490272465e-05, "loss": 0.0538, "step": 30870 }, { "epoch": 0.72, "learning_rate": 2.4744314848278615e-05, "loss": 0.2896, "step": 30875 }, { "epoch": 0.72, "learning_rate": 2.4743434793832577e-05, "loss": 0.2118, "step": 30880 }, { "epoch": 0.72, "learning_rate": 2.4742554739386542e-05, "loss": 0.2206, "step": 30885 }, { "epoch": 0.72, "learning_rate": 2.474167468494051e-05, "loss": 0.0574, "step": 30890 }, { "epoch": 0.72, "learning_rate": 2.4740794630494476e-05, "loss": 0.5163, "step": 30895 }, { "epoch": 0.72, "learning_rate": 2.4739914576048437e-05, "loss": 0.4795, "step": 30900 }, { "epoch": 0.72, "learning_rate": 2.4739034521602402e-05, "loss": 0.2453, "step": 30905 }, { "epoch": 0.72, "learning_rate": 2.473815446715637e-05, "loss": 0.0111, "step": 30910 }, { "epoch": 0.72, "learning_rate": 2.4737274412710332e-05, "loss": 0.0394, "step": 30915 }, { "epoch": 0.72, "learning_rate": 2.4736394358264297e-05, "loss": 0.0653, "step": 30920 }, { "epoch": 0.72, "learning_rate": 2.4735514303818266e-05, "loss": 0.0903, "step": 30925 }, { "epoch": 0.72, "learning_rate": 2.473463424937223e-05, "loss": 0.3889, "step": 30930 }, { "epoch": 0.72, "learning_rate": 2.4733754194926192e-05, "loss": 0.2164, "step": 30935 }, { "epoch": 0.72, "learning_rate": 2.4732874140480157e-05, "loss": 0.2988, "step": 30940 }, { "epoch": 0.72, "learning_rate": 2.4731994086034126e-05, "loss": 0.5208, "step": 30945 }, { "epoch": 0.72, "learning_rate": 2.4731114031588087e-05, "loss": 0.8854, "step": 30950 }, { "epoch": 0.72, "learning_rate": 2.4730233977142052e-05, "loss": 0.2616, "step": 30955 }, { "epoch": 0.72, "learning_rate": 2.4729353922696017e-05, "loss": 0.0902, "step": 30960 }, { "epoch": 0.72, "learning_rate": 2.4728473868249986e-05, "loss": 0.1851, "step": 30965 }, { "epoch": 0.72, "learning_rate": 2.4727593813803947e-05, "loss": 0.1436, "step": 30970 }, { "epoch": 0.72, "learning_rate": 2.4726713759357912e-05, "loss": 0.0842, "step": 30975 }, { "epoch": 0.72, "learning_rate": 2.4725833704911877e-05, "loss": 0.3198, "step": 30980 }, { "epoch": 0.72, "learning_rate": 2.4724953650465843e-05, "loss": 0.4876, "step": 30985 }, { "epoch": 0.72, "learning_rate": 2.4724073596019808e-05, "loss": 0.2254, "step": 30990 }, { "epoch": 0.72, "learning_rate": 2.4723193541573773e-05, "loss": 0.339, "step": 30995 }, { "epoch": 0.72, "learning_rate": 2.4722313487127738e-05, "loss": 0.5059, "step": 31000 }, { "epoch": 0.72, "learning_rate": 2.4721433432681703e-05, "loss": 0.2371, "step": 31005 }, { "epoch": 0.72, "learning_rate": 2.4720553378235668e-05, "loss": 0.0696, "step": 31010 }, { "epoch": 0.72, "learning_rate": 2.4719673323789633e-05, "loss": 0.0367, "step": 31015 }, { "epoch": 0.72, "learning_rate": 2.4718793269343594e-05, "loss": 0.2219, "step": 31020 }, { "epoch": 0.72, "learning_rate": 2.4717913214897563e-05, "loss": 0.1222, "step": 31025 }, { "epoch": 0.72, "learning_rate": 2.4717033160451528e-05, "loss": 0.1455, "step": 31030 }, { "epoch": 0.72, "learning_rate": 2.4716153106005493e-05, "loss": 0.3037, "step": 31035 }, { "epoch": 0.72, "learning_rate": 2.4715273051559454e-05, "loss": 0.2225, "step": 31040 }, { "epoch": 0.72, "learning_rate": 2.4714392997113423e-05, "loss": 0.2213, "step": 31045 }, { "epoch": 0.72, "learning_rate": 2.4713512942667388e-05, "loss": 0.6611, "step": 31050 }, { "epoch": 0.72, "learning_rate": 2.471263288822135e-05, "loss": 0.22, "step": 31055 }, { "epoch": 0.72, "learning_rate": 2.4711752833775318e-05, "loss": 0.0625, "step": 31060 }, { "epoch": 0.72, "learning_rate": 2.4710872779329283e-05, "loss": 0.1213, "step": 31065 }, { "epoch": 0.72, "learning_rate": 2.4709992724883248e-05, "loss": 0.1751, "step": 31070 }, { "epoch": 0.73, "learning_rate": 2.470911267043721e-05, "loss": 0.1343, "step": 31075 }, { "epoch": 0.73, "learning_rate": 2.4708232615991178e-05, "loss": 0.2414, "step": 31080 }, { "epoch": 0.73, "learning_rate": 2.4707352561545143e-05, "loss": 0.1093, "step": 31085 }, { "epoch": 0.73, "learning_rate": 2.4706472507099105e-05, "loss": 0.2025, "step": 31090 }, { "epoch": 0.73, "learning_rate": 2.470559245265307e-05, "loss": 0.3297, "step": 31095 }, { "epoch": 0.73, "learning_rate": 2.4704712398207038e-05, "loss": 0.4458, "step": 31100 }, { "epoch": 0.73, "learning_rate": 2.4703832343761003e-05, "loss": 0.2174, "step": 31105 }, { "epoch": 0.73, "learning_rate": 2.4702952289314965e-05, "loss": 0.0851, "step": 31110 }, { "epoch": 0.73, "learning_rate": 2.470207223486893e-05, "loss": 0.208, "step": 31115 }, { "epoch": 0.73, "learning_rate": 2.4701192180422898e-05, "loss": 0.0871, "step": 31120 }, { "epoch": 0.73, "learning_rate": 2.4700312125976863e-05, "loss": 0.0992, "step": 31125 }, { "epoch": 0.73, "learning_rate": 2.4699432071530825e-05, "loss": 0.0946, "step": 31130 }, { "epoch": 0.73, "learning_rate": 2.469855201708479e-05, "loss": 0.2964, "step": 31135 }, { "epoch": 0.73, "learning_rate": 2.4697847973527964e-05, "loss": 0.4057, "step": 31140 }, { "epoch": 0.73, "learning_rate": 2.4696967919081926e-05, "loss": 0.2776, "step": 31145 }, { "epoch": 0.73, "learning_rate": 2.469608786463589e-05, "loss": 0.5112, "step": 31150 }, { "epoch": 0.73, "learning_rate": 2.469520781018986e-05, "loss": 0.1484, "step": 31155 }, { "epoch": 0.73, "learning_rate": 2.4694327755743825e-05, "loss": 0.0269, "step": 31160 }, { "epoch": 0.73, "learning_rate": 2.4693447701297786e-05, "loss": 0.1189, "step": 31165 }, { "epoch": 0.73, "learning_rate": 2.469256764685175e-05, "loss": 0.1288, "step": 31170 }, { "epoch": 0.73, "learning_rate": 2.469168759240572e-05, "loss": 0.1166, "step": 31175 }, { "epoch": 0.73, "learning_rate": 2.469080753795968e-05, "loss": 0.1595, "step": 31180 }, { "epoch": 0.73, "learning_rate": 2.4689927483513646e-05, "loss": 0.244, "step": 31185 }, { "epoch": 0.73, "learning_rate": 2.468904742906761e-05, "loss": 0.2156, "step": 31190 }, { "epoch": 0.73, "learning_rate": 2.468816737462158e-05, "loss": 0.1792, "step": 31195 }, { "epoch": 0.73, "learning_rate": 2.468728732017554e-05, "loss": 0.3334, "step": 31200 }, { "epoch": 0.73, "learning_rate": 2.4686407265729506e-05, "loss": 0.3028, "step": 31205 }, { "epoch": 0.73, "learning_rate": 2.468552721128347e-05, "loss": 0.0858, "step": 31210 }, { "epoch": 0.73, "learning_rate": 2.4684647156837436e-05, "loss": 0.2263, "step": 31215 }, { "epoch": 0.73, "learning_rate": 2.46837671023914e-05, "loss": 0.165, "step": 31220 }, { "epoch": 0.73, "learning_rate": 2.4682887047945366e-05, "loss": 0.037, "step": 31225 }, { "epoch": 0.73, "learning_rate": 2.468200699349933e-05, "loss": 0.2004, "step": 31230 }, { "epoch": 0.73, "learning_rate": 2.4681126939053296e-05, "loss": 0.273, "step": 31235 }, { "epoch": 0.73, "learning_rate": 2.468024688460726e-05, "loss": 0.3332, "step": 31240 }, { "epoch": 0.73, "learning_rate": 2.4679366830161227e-05, "loss": 0.3632, "step": 31245 }, { "epoch": 0.73, "learning_rate": 2.467848677571519e-05, "loss": 0.5276, "step": 31250 }, { "epoch": 0.73, "learning_rate": 2.4677606721269157e-05, "loss": 0.3124, "step": 31255 }, { "epoch": 0.73, "learning_rate": 2.467672666682312e-05, "loss": 0.1391, "step": 31260 }, { "epoch": 0.73, "learning_rate": 2.4675846612377087e-05, "loss": 0.0853, "step": 31265 }, { "epoch": 0.73, "learning_rate": 2.467496655793105e-05, "loss": 0.0992, "step": 31270 }, { "epoch": 0.73, "learning_rate": 2.4674086503485017e-05, "loss": 0.1444, "step": 31275 }, { "epoch": 0.73, "learning_rate": 2.467320644903898e-05, "loss": 0.1482, "step": 31280 }, { "epoch": 0.73, "learning_rate": 2.4672326394592943e-05, "loss": 0.2799, "step": 31285 }, { "epoch": 0.73, "learning_rate": 2.4671446340146912e-05, "loss": 0.3078, "step": 31290 }, { "epoch": 0.73, "learning_rate": 2.4670566285700877e-05, "loss": 0.4941, "step": 31295 }, { "epoch": 0.73, "learning_rate": 2.4669686231254842e-05, "loss": 0.5106, "step": 31300 }, { "epoch": 0.73, "learning_rate": 2.4668806176808803e-05, "loss": 0.292, "step": 31305 }, { "epoch": 0.73, "learning_rate": 2.4667926122362772e-05, "loss": 0.1299, "step": 31310 }, { "epoch": 0.73, "learning_rate": 2.4667046067916737e-05, "loss": 0.0911, "step": 31315 }, { "epoch": 0.73, "learning_rate": 2.4666166013470702e-05, "loss": 0.0962, "step": 31320 }, { "epoch": 0.73, "learning_rate": 2.4665285959024663e-05, "loss": 0.1601, "step": 31325 }, { "epoch": 0.73, "learning_rate": 2.4664405904578632e-05, "loss": 0.1201, "step": 31330 }, { "epoch": 0.73, "learning_rate": 2.4663525850132597e-05, "loss": 0.1435, "step": 31335 }, { "epoch": 0.73, "learning_rate": 2.466264579568656e-05, "loss": 0.1785, "step": 31340 }, { "epoch": 0.73, "learning_rate": 2.4661765741240524e-05, "loss": 0.4422, "step": 31345 }, { "epoch": 0.73, "learning_rate": 2.4660885686794492e-05, "loss": 0.4001, "step": 31350 }, { "epoch": 0.73, "learning_rate": 2.4660005632348457e-05, "loss": 0.2893, "step": 31355 }, { "epoch": 0.73, "learning_rate": 2.465912557790242e-05, "loss": 0.0968, "step": 31360 }, { "epoch": 0.73, "learning_rate": 2.4658245523456384e-05, "loss": 0.1099, "step": 31365 }, { "epoch": 0.73, "learning_rate": 2.4657365469010352e-05, "loss": 0.1968, "step": 31370 }, { "epoch": 0.73, "learning_rate": 2.4656485414564314e-05, "loss": 0.097, "step": 31375 }, { "epoch": 0.73, "learning_rate": 2.465560536011828e-05, "loss": 0.2012, "step": 31380 }, { "epoch": 0.73, "learning_rate": 2.4654725305672247e-05, "loss": 0.2387, "step": 31385 }, { "epoch": 0.73, "learning_rate": 2.4653845251226212e-05, "loss": 0.2918, "step": 31390 }, { "epoch": 0.73, "learning_rate": 2.4652965196780174e-05, "loss": 0.3232, "step": 31395 }, { "epoch": 0.73, "learning_rate": 2.465208514233414e-05, "loss": 0.5369, "step": 31400 }, { "epoch": 0.73, "learning_rate": 2.4651205087888107e-05, "loss": 0.1619, "step": 31405 }, { "epoch": 0.73, "learning_rate": 2.465032503344207e-05, "loss": 0.1674, "step": 31410 }, { "epoch": 0.73, "learning_rate": 2.4649444978996034e-05, "loss": 0.0985, "step": 31415 }, { "epoch": 0.73, "learning_rate": 2.464856492455e-05, "loss": 0.161, "step": 31420 }, { "epoch": 0.73, "learning_rate": 2.4647684870103967e-05, "loss": 0.1765, "step": 31425 }, { "epoch": 0.73, "learning_rate": 2.464680481565793e-05, "loss": 0.3609, "step": 31430 }, { "epoch": 0.73, "learning_rate": 2.4645924761211894e-05, "loss": 0.3146, "step": 31435 }, { "epoch": 0.73, "learning_rate": 2.464504470676586e-05, "loss": 0.3114, "step": 31440 }, { "epoch": 0.73, "learning_rate": 2.4644164652319824e-05, "loss": 0.377, "step": 31445 }, { "epoch": 0.73, "learning_rate": 2.464328459787379e-05, "loss": 0.2166, "step": 31450 }, { "epoch": 0.73, "learning_rate": 2.4642404543427754e-05, "loss": 0.2993, "step": 31455 }, { "epoch": 0.73, "learning_rate": 2.464152448898172e-05, "loss": 0.1048, "step": 31460 }, { "epoch": 0.73, "learning_rate": 2.4640644434535684e-05, "loss": 0.0328, "step": 31465 }, { "epoch": 0.73, "learning_rate": 2.463976438008965e-05, "loss": 0.1222, "step": 31470 }, { "epoch": 0.73, "learning_rate": 2.4638884325643614e-05, "loss": 0.2039, "step": 31475 }, { "epoch": 0.73, "learning_rate": 2.4638004271197576e-05, "loss": 0.223, "step": 31480 }, { "epoch": 0.73, "learning_rate": 2.4637124216751544e-05, "loss": 0.1369, "step": 31485 }, { "epoch": 0.73, "learning_rate": 2.463624416230551e-05, "loss": 0.2787, "step": 31490 }, { "epoch": 0.73, "learning_rate": 2.4635364107859474e-05, "loss": 0.2514, "step": 31495 }, { "epoch": 0.73, "learning_rate": 2.4634484053413436e-05, "loss": 0.433, "step": 31500 }, { "epoch": 0.74, "learning_rate": 2.4633603998967404e-05, "loss": 0.2393, "step": 31505 }, { "epoch": 0.74, "learning_rate": 2.463272394452137e-05, "loss": 0.0757, "step": 31510 }, { "epoch": 0.74, "learning_rate": 2.463184389007533e-05, "loss": 0.1144, "step": 31515 }, { "epoch": 0.74, "learning_rate": 2.46309638356293e-05, "loss": 0.1254, "step": 31520 }, { "epoch": 0.74, "learning_rate": 2.4630083781183264e-05, "loss": 0.147, "step": 31525 }, { "epoch": 0.74, "learning_rate": 2.462920372673723e-05, "loss": 0.1807, "step": 31530 }, { "epoch": 0.74, "learning_rate": 2.462832367229119e-05, "loss": 0.197, "step": 31535 }, { "epoch": 0.74, "learning_rate": 2.462744361784516e-05, "loss": 0.1688, "step": 31540 }, { "epoch": 0.74, "learning_rate": 2.4626563563399124e-05, "loss": 0.238, "step": 31545 }, { "epoch": 0.74, "learning_rate": 2.4625683508953086e-05, "loss": 0.7433, "step": 31550 }, { "epoch": 0.74, "learning_rate": 2.462480345450705e-05, "loss": 0.214, "step": 31555 }, { "epoch": 0.74, "learning_rate": 2.462392340006102e-05, "loss": 0.1296, "step": 31560 }, { "epoch": 0.74, "learning_rate": 2.4623043345614984e-05, "loss": 0.1235, "step": 31565 }, { "epoch": 0.74, "learning_rate": 2.4622163291168946e-05, "loss": 0.1099, "step": 31570 }, { "epoch": 0.74, "learning_rate": 2.462128323672291e-05, "loss": 0.3672, "step": 31575 }, { "epoch": 0.74, "learning_rate": 2.462040318227688e-05, "loss": 0.0914, "step": 31580 }, { "epoch": 0.74, "learning_rate": 2.461952312783084e-05, "loss": 0.2661, "step": 31585 }, { "epoch": 0.74, "learning_rate": 2.4618643073384806e-05, "loss": 0.3974, "step": 31590 }, { "epoch": 0.74, "learning_rate": 2.461776301893877e-05, "loss": 0.2513, "step": 31595 }, { "epoch": 0.74, "learning_rate": 2.461688296449274e-05, "loss": 0.4649, "step": 31600 }, { "epoch": 0.74, "learning_rate": 2.46160029100467e-05, "loss": 0.1763, "step": 31605 }, { "epoch": 0.74, "learning_rate": 2.4615122855600666e-05, "loss": 0.0832, "step": 31610 }, { "epoch": 0.74, "learning_rate": 2.461424280115463e-05, "loss": 0.1383, "step": 31615 }, { "epoch": 0.74, "learning_rate": 2.46133627467086e-05, "loss": 0.0613, "step": 31620 }, { "epoch": 0.74, "learning_rate": 2.461248269226256e-05, "loss": 0.1212, "step": 31625 }, { "epoch": 0.74, "learning_rate": 2.4611602637816526e-05, "loss": 0.1158, "step": 31630 }, { "epoch": 0.74, "learning_rate": 2.461072258337049e-05, "loss": 0.2344, "step": 31635 }, { "epoch": 0.74, "learning_rate": 2.4609842528924456e-05, "loss": 0.1028, "step": 31640 }, { "epoch": 0.74, "learning_rate": 2.460896247447842e-05, "loss": 0.3442, "step": 31645 }, { "epoch": 0.74, "learning_rate": 2.4608082420032386e-05, "loss": 0.4261, "step": 31650 }, { "epoch": 0.74, "learning_rate": 2.4607202365586355e-05, "loss": 0.1674, "step": 31655 }, { "epoch": 0.74, "learning_rate": 2.4606322311140316e-05, "loss": 0.1053, "step": 31660 }, { "epoch": 0.74, "learning_rate": 2.460544225669428e-05, "loss": 0.1332, "step": 31665 }, { "epoch": 0.74, "learning_rate": 2.4604562202248246e-05, "loss": 0.191, "step": 31670 }, { "epoch": 0.74, "learning_rate": 2.460368214780221e-05, "loss": 0.1546, "step": 31675 }, { "epoch": 0.74, "learning_rate": 2.4602802093356177e-05, "loss": 0.1881, "step": 31680 }, { "epoch": 0.74, "learning_rate": 2.460192203891014e-05, "loss": 0.2039, "step": 31685 }, { "epoch": 0.74, "learning_rate": 2.4601041984464107e-05, "loss": 0.2876, "step": 31690 }, { "epoch": 0.74, "learning_rate": 2.460016193001807e-05, "loss": 0.4425, "step": 31695 }, { "epoch": 0.74, "learning_rate": 2.4599281875572037e-05, "loss": 0.6215, "step": 31700 }, { "epoch": 0.74, "learning_rate": 2.4598401821126e-05, "loss": 0.2614, "step": 31705 }, { "epoch": 0.74, "learning_rate": 2.4597521766679963e-05, "loss": 0.048, "step": 31710 }, { "epoch": 0.74, "learning_rate": 2.459664171223393e-05, "loss": 0.0391, "step": 31715 }, { "epoch": 0.74, "learning_rate": 2.4595761657787897e-05, "loss": 0.0754, "step": 31720 }, { "epoch": 0.74, "learning_rate": 2.4594881603341862e-05, "loss": 0.285, "step": 31725 }, { "epoch": 0.74, "learning_rate": 2.4594001548895823e-05, "loss": 0.2052, "step": 31730 }, { "epoch": 0.74, "learning_rate": 2.4593121494449792e-05, "loss": 0.2291, "step": 31735 }, { "epoch": 0.74, "learning_rate": 2.4592241440003757e-05, "loss": 0.3813, "step": 31740 }, { "epoch": 0.74, "learning_rate": 2.459136138555772e-05, "loss": 0.3539, "step": 31745 }, { "epoch": 0.74, "learning_rate": 2.4590481331111683e-05, "loss": 0.4541, "step": 31750 }, { "epoch": 0.74, "learning_rate": 2.4589601276665652e-05, "loss": 0.2161, "step": 31755 }, { "epoch": 0.74, "learning_rate": 2.4588721222219617e-05, "loss": 0.1071, "step": 31760 }, { "epoch": 0.74, "learning_rate": 2.458784116777358e-05, "loss": 0.06, "step": 31765 }, { "epoch": 0.74, "learning_rate": 2.4586961113327544e-05, "loss": 0.1564, "step": 31770 }, { "epoch": 0.74, "learning_rate": 2.4586081058881512e-05, "loss": 0.1736, "step": 31775 }, { "epoch": 0.74, "learning_rate": 2.4585201004435474e-05, "loss": 0.1711, "step": 31780 }, { "epoch": 0.74, "learning_rate": 2.458432094998944e-05, "loss": 0.1864, "step": 31785 }, { "epoch": 0.74, "learning_rate": 2.4583440895543407e-05, "loss": 0.3526, "step": 31790 }, { "epoch": 0.74, "learning_rate": 2.4582560841097372e-05, "loss": 0.3281, "step": 31795 }, { "epoch": 0.74, "learning_rate": 2.4581680786651334e-05, "loss": 0.7607, "step": 31800 }, { "epoch": 0.74, "learning_rate": 2.45808007322053e-05, "loss": 0.135, "step": 31805 }, { "epoch": 0.74, "learning_rate": 2.4579920677759267e-05, "loss": 0.0345, "step": 31810 }, { "epoch": 0.74, "learning_rate": 2.457904062331323e-05, "loss": 0.1241, "step": 31815 }, { "epoch": 0.74, "learning_rate": 2.4578160568867194e-05, "loss": 0.0697, "step": 31820 }, { "epoch": 0.74, "learning_rate": 2.457728051442116e-05, "loss": 0.1078, "step": 31825 }, { "epoch": 0.74, "learning_rate": 2.4576400459975127e-05, "loss": 0.3541, "step": 31830 }, { "epoch": 0.74, "learning_rate": 2.457552040552909e-05, "loss": 0.1918, "step": 31835 }, { "epoch": 0.74, "learning_rate": 2.4574640351083054e-05, "loss": 0.2032, "step": 31840 }, { "epoch": 0.74, "learning_rate": 2.457376029663702e-05, "loss": 0.2763, "step": 31845 }, { "epoch": 0.74, "learning_rate": 2.4572880242190984e-05, "loss": 0.4349, "step": 31850 }, { "epoch": 0.74, "learning_rate": 2.457200018774495e-05, "loss": 0.1519, "step": 31855 }, { "epoch": 0.74, "learning_rate": 2.4571120133298914e-05, "loss": 0.1236, "step": 31860 }, { "epoch": 0.74, "learning_rate": 2.457024007885288e-05, "loss": 0.1792, "step": 31865 }, { "epoch": 0.74, "learning_rate": 2.4569360024406844e-05, "loss": 0.1063, "step": 31870 }, { "epoch": 0.74, "learning_rate": 2.456847996996081e-05, "loss": 0.2219, "step": 31875 }, { "epoch": 0.74, "learning_rate": 2.4567599915514774e-05, "loss": 0.1433, "step": 31880 }, { "epoch": 0.74, "learning_rate": 2.4566719861068736e-05, "loss": 0.1978, "step": 31885 }, { "epoch": 0.74, "learning_rate": 2.4565839806622704e-05, "loss": 0.3224, "step": 31890 }, { "epoch": 0.74, "learning_rate": 2.456495975217667e-05, "loss": 0.4391, "step": 31895 }, { "epoch": 0.74, "learning_rate": 2.4564079697730634e-05, "loss": 0.8095, "step": 31900 }, { "epoch": 0.74, "learning_rate": 2.4563199643284596e-05, "loss": 0.1416, "step": 31905 }, { "epoch": 0.74, "learning_rate": 2.4562319588838564e-05, "loss": 0.043, "step": 31910 }, { "epoch": 0.74, "learning_rate": 2.456143953439253e-05, "loss": 0.1146, "step": 31915 }, { "epoch": 0.74, "learning_rate": 2.4560559479946494e-05, "loss": 0.111, "step": 31920 }, { "epoch": 0.74, "learning_rate": 2.455967942550046e-05, "loss": 0.2343, "step": 31925 }, { "epoch": 0.74, "learning_rate": 2.4558799371054424e-05, "loss": 0.0505, "step": 31930 }, { "epoch": 0.75, "learning_rate": 2.455791931660839e-05, "loss": 0.2192, "step": 31935 }, { "epoch": 0.75, "learning_rate": 2.455703926216235e-05, "loss": 0.4314, "step": 31940 }, { "epoch": 0.75, "learning_rate": 2.455615920771632e-05, "loss": 0.2694, "step": 31945 }, { "epoch": 0.75, "learning_rate": 2.4555279153270284e-05, "loss": 0.5316, "step": 31950 }, { "epoch": 0.75, "learning_rate": 2.455439909882425e-05, "loss": 0.2718, "step": 31955 }, { "epoch": 0.75, "learning_rate": 2.455351904437821e-05, "loss": 0.0724, "step": 31960 }, { "epoch": 0.75, "learning_rate": 2.455263898993218e-05, "loss": 0.1426, "step": 31965 }, { "epoch": 0.75, "learning_rate": 2.4551758935486144e-05, "loss": 0.1097, "step": 31970 }, { "epoch": 0.75, "learning_rate": 2.4550878881040106e-05, "loss": 0.0639, "step": 31975 }, { "epoch": 0.75, "learning_rate": 2.454999882659407e-05, "loss": 0.2624, "step": 31980 }, { "epoch": 0.75, "learning_rate": 2.454911877214804e-05, "loss": 0.1938, "step": 31985 }, { "epoch": 0.75, "learning_rate": 2.4548238717702004e-05, "loss": 0.1541, "step": 31990 }, { "epoch": 0.75, "learning_rate": 2.4547358663255966e-05, "loss": 0.4746, "step": 31995 }, { "epoch": 0.75, "learning_rate": 2.454647860880993e-05, "loss": 0.5303, "step": 32000 }, { "epoch": 0.75, "learning_rate": 2.45455985543639e-05, "loss": 0.2698, "step": 32005 }, { "epoch": 0.75, "learning_rate": 2.454471849991786e-05, "loss": 0.1471, "step": 32010 }, { "epoch": 0.75, "learning_rate": 2.4543838445471826e-05, "loss": 0.1326, "step": 32015 }, { "epoch": 0.75, "learning_rate": 2.454295839102579e-05, "loss": 0.1507, "step": 32020 }, { "epoch": 0.75, "learning_rate": 2.454207833657976e-05, "loss": 0.1549, "step": 32025 }, { "epoch": 0.75, "learning_rate": 2.454119828213372e-05, "loss": 0.2016, "step": 32030 }, { "epoch": 0.75, "learning_rate": 2.4540318227687686e-05, "loss": 0.3402, "step": 32035 }, { "epoch": 0.75, "learning_rate": 2.453943817324165e-05, "loss": 0.2191, "step": 32040 }, { "epoch": 0.75, "learning_rate": 2.4538558118795616e-05, "loss": 0.4642, "step": 32045 }, { "epoch": 0.75, "learning_rate": 2.453767806434958e-05, "loss": 0.3907, "step": 32050 }, { "epoch": 0.75, "learning_rate": 2.4536798009903546e-05, "loss": 0.253, "step": 32055 }, { "epoch": 0.75, "learning_rate": 2.4535917955457515e-05, "loss": 0.083, "step": 32060 }, { "epoch": 0.75, "learning_rate": 2.4535037901011476e-05, "loss": 0.0598, "step": 32065 }, { "epoch": 0.75, "learning_rate": 2.453415784656544e-05, "loss": 0.1512, "step": 32070 }, { "epoch": 0.75, "learning_rate": 2.4533277792119406e-05, "loss": 0.094, "step": 32075 }, { "epoch": 0.75, "learning_rate": 2.453239773767337e-05, "loss": 0.1532, "step": 32080 }, { "epoch": 0.75, "learning_rate": 2.4531517683227336e-05, "loss": 0.1096, "step": 32085 }, { "epoch": 0.75, "learning_rate": 2.45306376287813e-05, "loss": 0.3501, "step": 32090 }, { "epoch": 0.75, "learning_rate": 2.4529757574335266e-05, "loss": 0.3492, "step": 32095 }, { "epoch": 0.75, "learning_rate": 2.452887751988923e-05, "loss": 0.6321, "step": 32100 }, { "epoch": 0.75, "learning_rate": 2.4527997465443196e-05, "loss": 0.2082, "step": 32105 }, { "epoch": 0.75, "learning_rate": 2.452711741099716e-05, "loss": 0.1338, "step": 32110 }, { "epoch": 0.75, "learning_rate": 2.4526237356551123e-05, "loss": 0.1247, "step": 32115 }, { "epoch": 0.75, "learning_rate": 2.452535730210509e-05, "loss": 0.1546, "step": 32120 }, { "epoch": 0.75, "learning_rate": 2.4524477247659057e-05, "loss": 0.0719, "step": 32125 }, { "epoch": 0.75, "learning_rate": 2.452359719321302e-05, "loss": 0.2447, "step": 32130 }, { "epoch": 0.75, "learning_rate": 2.4522717138766983e-05, "loss": 0.2489, "step": 32135 }, { "epoch": 0.75, "learning_rate": 2.452183708432095e-05, "loss": 0.1678, "step": 32140 }, { "epoch": 0.75, "learning_rate": 2.4520957029874917e-05, "loss": 0.4593, "step": 32145 }, { "epoch": 0.75, "learning_rate": 2.4520076975428878e-05, "loss": 0.5765, "step": 32150 }, { "epoch": 0.75, "learning_rate": 2.4519196920982843e-05, "loss": 0.2317, "step": 32155 }, { "epoch": 0.75, "learning_rate": 2.4518316866536812e-05, "loss": 0.0839, "step": 32160 }, { "epoch": 0.75, "learning_rate": 2.4517436812090777e-05, "loss": 0.1077, "step": 32165 }, { "epoch": 0.75, "learning_rate": 2.451655675764474e-05, "loss": 0.2356, "step": 32170 }, { "epoch": 0.75, "learning_rate": 2.4515676703198703e-05, "loss": 0.0843, "step": 32175 }, { "epoch": 0.75, "learning_rate": 2.4514796648752672e-05, "loss": 0.0962, "step": 32180 }, { "epoch": 0.75, "learning_rate": 2.4513916594306633e-05, "loss": 0.097, "step": 32185 }, { "epoch": 0.75, "learning_rate": 2.45130365398606e-05, "loss": 0.36, "step": 32190 }, { "epoch": 0.75, "learning_rate": 2.4512156485414567e-05, "loss": 0.4568, "step": 32195 }, { "epoch": 0.75, "learning_rate": 2.4511276430968532e-05, "loss": 0.5855, "step": 32200 }, { "epoch": 0.75, "learning_rate": 2.4510396376522494e-05, "loss": 0.224, "step": 32205 }, { "epoch": 0.75, "learning_rate": 2.450951632207646e-05, "loss": 0.1346, "step": 32210 }, { "epoch": 0.75, "learning_rate": 2.4508636267630427e-05, "loss": 0.0702, "step": 32215 }, { "epoch": 0.75, "learning_rate": 2.450775621318439e-05, "loss": 0.1999, "step": 32220 }, { "epoch": 0.75, "learning_rate": 2.4506876158738354e-05, "loss": 0.145, "step": 32225 }, { "epoch": 0.75, "learning_rate": 2.450599610429232e-05, "loss": 0.2117, "step": 32230 }, { "epoch": 0.75, "learning_rate": 2.4505116049846287e-05, "loss": 0.3239, "step": 32235 }, { "epoch": 0.75, "learning_rate": 2.450423599540025e-05, "loss": 0.196, "step": 32240 }, { "epoch": 0.75, "learning_rate": 2.4503355940954214e-05, "loss": 0.2528, "step": 32245 }, { "epoch": 0.75, "learning_rate": 2.450247588650818e-05, "loss": 0.4887, "step": 32250 }, { "epoch": 0.75, "learning_rate": 2.4501595832062147e-05, "loss": 0.242, "step": 32255 }, { "epoch": 0.75, "learning_rate": 2.450071577761611e-05, "loss": 0.1104, "step": 32260 }, { "epoch": 0.75, "learning_rate": 2.4499835723170074e-05, "loss": 0.2849, "step": 32265 }, { "epoch": 0.75, "learning_rate": 2.449895566872404e-05, "loss": 0.0603, "step": 32270 }, { "epoch": 0.75, "learning_rate": 2.4498075614278004e-05, "loss": 0.2371, "step": 32275 }, { "epoch": 0.75, "learning_rate": 2.449719555983197e-05, "loss": 0.2225, "step": 32280 }, { "epoch": 0.75, "learning_rate": 2.4496315505385934e-05, "loss": 0.2173, "step": 32285 }, { "epoch": 0.75, "learning_rate": 2.44954354509399e-05, "loss": 0.2616, "step": 32290 }, { "epoch": 0.75, "learning_rate": 2.4494555396493864e-05, "loss": 0.2073, "step": 32295 }, { "epoch": 0.75, "learning_rate": 2.449367534204783e-05, "loss": 0.5491, "step": 32300 }, { "epoch": 0.75, "learning_rate": 2.4492795287601794e-05, "loss": 0.144, "step": 32305 }, { "epoch": 0.75, "learning_rate": 2.4491915233155756e-05, "loss": 0.0894, "step": 32310 }, { "epoch": 0.75, "learning_rate": 2.4491035178709724e-05, "loss": 0.1844, "step": 32315 }, { "epoch": 0.75, "learning_rate": 2.449015512426369e-05, "loss": 0.1284, "step": 32320 }, { "epoch": 0.75, "learning_rate": 2.4489275069817654e-05, "loss": 0.1436, "step": 32325 }, { "epoch": 0.75, "learning_rate": 2.448839501537162e-05, "loss": 0.1518, "step": 32330 }, { "epoch": 0.75, "learning_rate": 2.4487514960925584e-05, "loss": 0.1266, "step": 32335 }, { "epoch": 0.75, "learning_rate": 2.448663490647955e-05, "loss": 0.2931, "step": 32340 }, { "epoch": 0.75, "learning_rate": 2.448575485203351e-05, "loss": 0.345, "step": 32345 }, { "epoch": 0.75, "learning_rate": 2.448487479758748e-05, "loss": 0.5616, "step": 32350 }, { "epoch": 0.75, "learning_rate": 2.4483994743141444e-05, "loss": 0.1872, "step": 32355 }, { "epoch": 0.75, "learning_rate": 2.448311468869541e-05, "loss": 0.0376, "step": 32360 }, { "epoch": 0.76, "learning_rate": 2.448223463424937e-05, "loss": 0.201, "step": 32365 }, { "epoch": 0.76, "learning_rate": 2.448135457980334e-05, "loss": 0.1602, "step": 32370 }, { "epoch": 0.76, "learning_rate": 2.4480474525357304e-05, "loss": 0.1462, "step": 32375 }, { "epoch": 0.76, "learning_rate": 2.4479594470911266e-05, "loss": 0.2648, "step": 32380 }, { "epoch": 0.76, "learning_rate": 2.447871441646523e-05, "loss": 0.2132, "step": 32385 }, { "epoch": 0.76, "learning_rate": 2.44778343620192e-05, "loss": 0.2775, "step": 32390 }, { "epoch": 0.76, "learning_rate": 2.4476954307573164e-05, "loss": 0.2604, "step": 32395 }, { "epoch": 0.76, "learning_rate": 2.4476074253127126e-05, "loss": 0.4934, "step": 32400 }, { "epoch": 0.76, "learning_rate": 2.447519419868109e-05, "loss": 0.2516, "step": 32405 }, { "epoch": 0.76, "learning_rate": 2.447431414423506e-05, "loss": 0.0401, "step": 32410 }, { "epoch": 0.76, "learning_rate": 2.447343408978902e-05, "loss": 0.1488, "step": 32415 }, { "epoch": 0.76, "learning_rate": 2.4472554035342986e-05, "loss": 0.0977, "step": 32420 }, { "epoch": 0.76, "learning_rate": 2.447167398089695e-05, "loss": 0.2365, "step": 32425 }, { "epoch": 0.76, "learning_rate": 2.447079392645092e-05, "loss": 0.0929, "step": 32430 }, { "epoch": 0.76, "learning_rate": 2.446991387200488e-05, "loss": 0.1215, "step": 32435 }, { "epoch": 0.76, "learning_rate": 2.4469033817558846e-05, "loss": 0.1809, "step": 32440 }, { "epoch": 0.76, "learning_rate": 2.446815376311281e-05, "loss": 0.4507, "step": 32445 }, { "epoch": 0.76, "learning_rate": 2.4467273708666776e-05, "loss": 0.416, "step": 32450 }, { "epoch": 0.76, "learning_rate": 2.446639365422074e-05, "loss": 0.2063, "step": 32455 }, { "epoch": 0.76, "learning_rate": 2.4465513599774706e-05, "loss": 0.0316, "step": 32460 }, { "epoch": 0.76, "learning_rate": 2.4464633545328675e-05, "loss": 0.1571, "step": 32465 }, { "epoch": 0.76, "learning_rate": 2.4463753490882636e-05, "loss": 0.2126, "step": 32470 }, { "epoch": 0.76, "learning_rate": 2.44628734364366e-05, "loss": 0.0974, "step": 32475 }, { "epoch": 0.76, "learning_rate": 2.4461993381990566e-05, "loss": 0.2488, "step": 32480 }, { "epoch": 0.76, "learning_rate": 2.446111332754453e-05, "loss": 0.268, "step": 32485 }, { "epoch": 0.76, "learning_rate": 2.4460233273098496e-05, "loss": 0.2542, "step": 32490 }, { "epoch": 0.76, "learning_rate": 2.445935321865246e-05, "loss": 0.3502, "step": 32495 }, { "epoch": 0.76, "learning_rate": 2.4458473164206426e-05, "loss": 0.7912, "step": 32500 }, { "epoch": 0.76, "learning_rate": 2.445759310976039e-05, "loss": 0.0997, "step": 32505 }, { "epoch": 0.76, "learning_rate": 2.4456713055314356e-05, "loss": 0.073, "step": 32510 }, { "epoch": 0.76, "learning_rate": 2.445583300086832e-05, "loss": 0.2109, "step": 32515 }, { "epoch": 0.76, "learning_rate": 2.4454952946422283e-05, "loss": 0.0551, "step": 32520 }, { "epoch": 0.76, "learning_rate": 2.445407289197625e-05, "loss": 0.1568, "step": 32525 }, { "epoch": 0.76, "learning_rate": 2.4453192837530216e-05, "loss": 0.1863, "step": 32530 }, { "epoch": 0.76, "learning_rate": 2.445231278308418e-05, "loss": 0.1844, "step": 32535 }, { "epoch": 0.76, "learning_rate": 2.4451432728638143e-05, "loss": 0.2685, "step": 32540 }, { "epoch": 0.76, "learning_rate": 2.445055267419211e-05, "loss": 0.2048, "step": 32545 }, { "epoch": 0.76, "learning_rate": 2.4449672619746077e-05, "loss": 0.4251, "step": 32550 }, { "epoch": 0.76, "learning_rate": 2.444879256530004e-05, "loss": 0.2672, "step": 32555 }, { "epoch": 0.76, "learning_rate": 2.4447912510854003e-05, "loss": 0.1356, "step": 32560 }, { "epoch": 0.76, "learning_rate": 2.444703245640797e-05, "loss": 0.1628, "step": 32565 }, { "epoch": 0.76, "learning_rate": 2.4446152401961937e-05, "loss": 0.1584, "step": 32570 }, { "epoch": 0.76, "learning_rate": 2.4445272347515898e-05, "loss": 0.1278, "step": 32575 }, { "epoch": 0.76, "learning_rate": 2.4444392293069863e-05, "loss": 0.2621, "step": 32580 }, { "epoch": 0.76, "learning_rate": 2.444351223862383e-05, "loss": 0.1892, "step": 32585 }, { "epoch": 0.76, "learning_rate": 2.4442632184177797e-05, "loss": 0.2015, "step": 32590 }, { "epoch": 0.76, "learning_rate": 2.444175212973176e-05, "loss": 0.2706, "step": 32595 }, { "epoch": 0.76, "learning_rate": 2.4440872075285727e-05, "loss": 0.5273, "step": 32600 }, { "epoch": 0.76, "learning_rate": 2.4439992020839692e-05, "loss": 0.1569, "step": 32605 }, { "epoch": 0.76, "learning_rate": 2.4439111966393653e-05, "loss": 0.0815, "step": 32610 }, { "epoch": 0.76, "learning_rate": 2.443823191194762e-05, "loss": 0.1136, "step": 32615 }, { "epoch": 0.76, "learning_rate": 2.4437351857501587e-05, "loss": 0.0952, "step": 32620 }, { "epoch": 0.76, "learning_rate": 2.4436471803055552e-05, "loss": 0.2277, "step": 32625 }, { "epoch": 0.76, "learning_rate": 2.4435591748609513e-05, "loss": 0.2324, "step": 32630 }, { "epoch": 0.76, "learning_rate": 2.443471169416348e-05, "loss": 0.2721, "step": 32635 }, { "epoch": 0.76, "learning_rate": 2.4433831639717447e-05, "loss": 0.2311, "step": 32640 }, { "epoch": 0.76, "learning_rate": 2.443295158527141e-05, "loss": 0.2452, "step": 32645 }, { "epoch": 0.76, "learning_rate": 2.4432071530825374e-05, "loss": 0.4933, "step": 32650 }, { "epoch": 0.76, "learning_rate": 2.443119147637934e-05, "loss": 0.2667, "step": 32655 }, { "epoch": 0.76, "learning_rate": 2.4430311421933307e-05, "loss": 0.0768, "step": 32660 }, { "epoch": 0.76, "learning_rate": 2.442943136748727e-05, "loss": 0.111, "step": 32665 }, { "epoch": 0.76, "learning_rate": 2.4428551313041234e-05, "loss": 0.1007, "step": 32670 }, { "epoch": 0.76, "learning_rate": 2.44276712585952e-05, "loss": 0.1343, "step": 32675 }, { "epoch": 0.76, "learning_rate": 2.4426791204149164e-05, "loss": 0.1494, "step": 32680 }, { "epoch": 0.76, "learning_rate": 2.442591114970313e-05, "loss": 0.1141, "step": 32685 }, { "epoch": 0.76, "learning_rate": 2.4425031095257094e-05, "loss": 0.232, "step": 32690 }, { "epoch": 0.76, "learning_rate": 2.442415104081106e-05, "loss": 0.452, "step": 32695 }, { "epoch": 0.76, "learning_rate": 2.4423270986365024e-05, "loss": 0.5093, "step": 32700 }, { "epoch": 0.76, "learning_rate": 2.442239093191899e-05, "loss": 0.1942, "step": 32705 }, { "epoch": 0.76, "learning_rate": 2.4421510877472954e-05, "loss": 0.2563, "step": 32710 }, { "epoch": 0.76, "learning_rate": 2.4420630823026915e-05, "loss": 0.1267, "step": 32715 }, { "epoch": 0.76, "learning_rate": 2.4419750768580884e-05, "loss": 0.078, "step": 32720 }, { "epoch": 0.76, "learning_rate": 2.441887071413485e-05, "loss": 0.1111, "step": 32725 }, { "epoch": 0.76, "learning_rate": 2.4417990659688814e-05, "loss": 0.1347, "step": 32730 }, { "epoch": 0.76, "learning_rate": 2.441711060524278e-05, "loss": 0.2832, "step": 32735 }, { "epoch": 0.76, "learning_rate": 2.4416230550796744e-05, "loss": 0.3786, "step": 32740 }, { "epoch": 0.76, "learning_rate": 2.441535049635071e-05, "loss": 0.3216, "step": 32745 }, { "epoch": 0.76, "learning_rate": 2.441447044190467e-05, "loss": 0.514, "step": 32750 }, { "epoch": 0.76, "learning_rate": 2.441359038745864e-05, "loss": 0.2151, "step": 32755 }, { "epoch": 0.76, "learning_rate": 2.4412710333012604e-05, "loss": 0.1113, "step": 32760 }, { "epoch": 0.76, "learning_rate": 2.441183027856657e-05, "loss": 0.1318, "step": 32765 }, { "epoch": 0.76, "learning_rate": 2.441095022412053e-05, "loss": 0.1145, "step": 32770 }, { "epoch": 0.76, "learning_rate": 2.44100701696745e-05, "loss": 0.3261, "step": 32775 }, { "epoch": 0.76, "learning_rate": 2.4409190115228464e-05, "loss": 0.1783, "step": 32780 }, { "epoch": 0.76, "learning_rate": 2.4408310060782426e-05, "loss": 0.2448, "step": 32785 }, { "epoch": 0.77, "learning_rate": 2.440743000633639e-05, "loss": 0.1115, "step": 32790 }, { "epoch": 0.77, "learning_rate": 2.440654995189036e-05, "loss": 0.237, "step": 32795 }, { "epoch": 0.77, "learning_rate": 2.4405669897444324e-05, "loss": 0.9008, "step": 32800 }, { "epoch": 0.77, "learning_rate": 2.4404789842998286e-05, "loss": 0.2975, "step": 32805 }, { "epoch": 0.77, "learning_rate": 2.440390978855225e-05, "loss": 0.0278, "step": 32810 }, { "epoch": 0.77, "learning_rate": 2.440302973410622e-05, "loss": 0.1054, "step": 32815 }, { "epoch": 0.77, "learning_rate": 2.440214967966018e-05, "loss": 0.0536, "step": 32820 }, { "epoch": 0.77, "learning_rate": 2.4401269625214146e-05, "loss": 0.0912, "step": 32825 }, { "epoch": 0.77, "learning_rate": 2.440038957076811e-05, "loss": 0.0695, "step": 32830 }, { "epoch": 0.77, "learning_rate": 2.439950951632208e-05, "loss": 0.2181, "step": 32835 }, { "epoch": 0.77, "learning_rate": 2.439862946187604e-05, "loss": 0.2071, "step": 32840 }, { "epoch": 0.77, "learning_rate": 2.4397749407430006e-05, "loss": 0.1786, "step": 32845 }, { "epoch": 0.77, "learning_rate": 2.439686935298397e-05, "loss": 0.5465, "step": 32850 }, { "epoch": 0.77, "learning_rate": 2.439598929853794e-05, "loss": 0.3088, "step": 32855 }, { "epoch": 0.77, "learning_rate": 2.43951092440919e-05, "loss": 0.1177, "step": 32860 }, { "epoch": 0.77, "learning_rate": 2.4394229189645866e-05, "loss": 0.0892, "step": 32865 }, { "epoch": 0.77, "learning_rate": 2.4393349135199834e-05, "loss": 0.0788, "step": 32870 }, { "epoch": 0.77, "learning_rate": 2.4392469080753796e-05, "loss": 0.2053, "step": 32875 }, { "epoch": 0.77, "learning_rate": 2.439158902630776e-05, "loss": 0.2411, "step": 32880 }, { "epoch": 0.77, "learning_rate": 2.4390708971861726e-05, "loss": 0.1792, "step": 32885 }, { "epoch": 0.77, "learning_rate": 2.4389828917415695e-05, "loss": 0.2403, "step": 32890 }, { "epoch": 0.77, "learning_rate": 2.4388948862969656e-05, "loss": 0.3403, "step": 32895 }, { "epoch": 0.77, "learning_rate": 2.438806880852362e-05, "loss": 0.6544, "step": 32900 }, { "epoch": 0.77, "learning_rate": 2.4387188754077586e-05, "loss": 0.2292, "step": 32905 }, { "epoch": 0.77, "learning_rate": 2.438630869963155e-05, "loss": 0.0939, "step": 32910 }, { "epoch": 0.77, "learning_rate": 2.4385428645185516e-05, "loss": 0.0768, "step": 32915 }, { "epoch": 0.77, "learning_rate": 2.438454859073948e-05, "loss": 0.1976, "step": 32920 }, { "epoch": 0.77, "learning_rate": 2.4383668536293446e-05, "loss": 0.2099, "step": 32925 }, { "epoch": 0.77, "learning_rate": 2.438278848184741e-05, "loss": 0.2137, "step": 32930 }, { "epoch": 0.77, "learning_rate": 2.4381908427401376e-05, "loss": 0.2454, "step": 32935 }, { "epoch": 0.77, "learning_rate": 2.438102837295534e-05, "loss": 0.3766, "step": 32940 }, { "epoch": 0.77, "learning_rate": 2.4380148318509303e-05, "loss": 0.1821, "step": 32945 }, { "epoch": 0.77, "learning_rate": 2.437926826406327e-05, "loss": 0.2997, "step": 32950 }, { "epoch": 0.77, "learning_rate": 2.4378388209617236e-05, "loss": 0.2166, "step": 32955 }, { "epoch": 0.77, "learning_rate": 2.43775081551712e-05, "loss": 0.1121, "step": 32960 }, { "epoch": 0.77, "learning_rate": 2.4376628100725163e-05, "loss": 0.0563, "step": 32965 }, { "epoch": 0.77, "learning_rate": 2.437574804627913e-05, "loss": 0.1434, "step": 32970 }, { "epoch": 0.77, "learning_rate": 2.4374867991833097e-05, "loss": 0.1197, "step": 32975 }, { "epoch": 0.77, "learning_rate": 2.4373987937387058e-05, "loss": 0.126, "step": 32980 }, { "epoch": 0.77, "learning_rate": 2.4373107882941023e-05, "loss": 0.1719, "step": 32985 }, { "epoch": 0.77, "learning_rate": 2.437222782849499e-05, "loss": 0.1605, "step": 32990 }, { "epoch": 0.77, "learning_rate": 2.4371347774048957e-05, "loss": 0.2424, "step": 32995 }, { "epoch": 0.77, "learning_rate": 2.4370467719602918e-05, "loss": 0.4644, "step": 33000 }, { "epoch": 0.77, "learning_rate": 2.4369587665156887e-05, "loss": 0.288, "step": 33005 }, { "epoch": 0.77, "learning_rate": 2.436870761071085e-05, "loss": 0.1246, "step": 33010 }, { "epoch": 0.77, "learning_rate": 2.4367827556264813e-05, "loss": 0.1295, "step": 33015 }, { "epoch": 0.77, "learning_rate": 2.436694750181878e-05, "loss": 0.2096, "step": 33020 }, { "epoch": 0.77, "learning_rate": 2.4366067447372747e-05, "loss": 0.1627, "step": 33025 }, { "epoch": 0.77, "learning_rate": 2.4365187392926712e-05, "loss": 0.2501, "step": 33030 }, { "epoch": 0.77, "learning_rate": 2.4364307338480673e-05, "loss": 0.2006, "step": 33035 }, { "epoch": 0.77, "learning_rate": 2.436342728403464e-05, "loss": 0.1698, "step": 33040 }, { "epoch": 0.77, "learning_rate": 2.4362547229588607e-05, "loss": 0.3167, "step": 33045 }, { "epoch": 0.77, "learning_rate": 2.436166717514257e-05, "loss": 0.7335, "step": 33050 }, { "epoch": 0.77, "learning_rate": 2.4360787120696533e-05, "loss": 0.2272, "step": 33055 }, { "epoch": 0.77, "learning_rate": 2.43599070662505e-05, "loss": 0.0794, "step": 33060 }, { "epoch": 0.77, "learning_rate": 2.4359027011804467e-05, "loss": 0.0462, "step": 33065 }, { "epoch": 0.77, "learning_rate": 2.435814695735843e-05, "loss": 0.0507, "step": 33070 }, { "epoch": 0.77, "learning_rate": 2.4357266902912394e-05, "loss": 0.1738, "step": 33075 }, { "epoch": 0.77, "learning_rate": 2.435638684846636e-05, "loss": 0.269, "step": 33080 }, { "epoch": 0.77, "learning_rate": 2.4355506794020324e-05, "loss": 0.197, "step": 33085 }, { "epoch": 0.77, "learning_rate": 2.435462673957429e-05, "loss": 0.3147, "step": 33090 }, { "epoch": 0.77, "learning_rate": 2.4353746685128254e-05, "loss": 0.4182, "step": 33095 }, { "epoch": 0.77, "learning_rate": 2.435286663068222e-05, "loss": 0.3033, "step": 33100 }, { "epoch": 0.77, "learning_rate": 2.4351986576236184e-05, "loss": 0.3562, "step": 33105 }, { "epoch": 0.77, "learning_rate": 2.435110652179015e-05, "loss": 0.1644, "step": 33110 }, { "epoch": 0.77, "learning_rate": 2.4350226467344114e-05, "loss": 0.2234, "step": 33115 }, { "epoch": 0.77, "learning_rate": 2.4349346412898075e-05, "loss": 0.1017, "step": 33120 }, { "epoch": 0.77, "learning_rate": 2.4348466358452044e-05, "loss": 0.1588, "step": 33125 }, { "epoch": 0.77, "learning_rate": 2.434758630400601e-05, "loss": 0.089, "step": 33130 }, { "epoch": 0.77, "learning_rate": 2.4346706249559974e-05, "loss": 0.1848, "step": 33135 }, { "epoch": 0.77, "learning_rate": 2.434582619511394e-05, "loss": 0.2766, "step": 33140 }, { "epoch": 0.77, "learning_rate": 2.4344946140667904e-05, "loss": 0.3654, "step": 33145 }, { "epoch": 0.77, "learning_rate": 2.434406608622187e-05, "loss": 0.6687, "step": 33150 }, { "epoch": 0.77, "learning_rate": 2.4343186031775834e-05, "loss": 0.2658, "step": 33155 }, { "epoch": 0.77, "learning_rate": 2.43423059773298e-05, "loss": 0.0209, "step": 33160 }, { "epoch": 0.77, "learning_rate": 2.4341425922883764e-05, "loss": 0.1912, "step": 33165 }, { "epoch": 0.77, "learning_rate": 2.434054586843773e-05, "loss": 0.1611, "step": 33170 }, { "epoch": 0.77, "learning_rate": 2.433966581399169e-05, "loss": 0.1772, "step": 33175 }, { "epoch": 0.77, "learning_rate": 2.433878575954566e-05, "loss": 0.1672, "step": 33180 }, { "epoch": 0.77, "learning_rate": 2.4337905705099624e-05, "loss": 0.177, "step": 33185 }, { "epoch": 0.77, "learning_rate": 2.433702565065359e-05, "loss": 0.1182, "step": 33190 }, { "epoch": 0.77, "learning_rate": 2.433614559620755e-05, "loss": 0.2869, "step": 33195 }, { "epoch": 0.77, "learning_rate": 2.433526554176152e-05, "loss": 0.3657, "step": 33200 }, { "epoch": 0.77, "learning_rate": 2.4334385487315484e-05, "loss": 0.3322, "step": 33205 }, { "epoch": 0.77, "learning_rate": 2.4333505432869446e-05, "loss": 0.0711, "step": 33210 }, { "epoch": 0.77, "learning_rate": 2.433262537842341e-05, "loss": 0.0776, "step": 33215 }, { "epoch": 0.78, "learning_rate": 2.433174532397738e-05, "loss": 0.1081, "step": 33220 }, { "epoch": 0.78, "learning_rate": 2.4330865269531344e-05, "loss": 0.1222, "step": 33225 }, { "epoch": 0.78, "learning_rate": 2.4329985215085306e-05, "loss": 0.1854, "step": 33230 }, { "epoch": 0.78, "learning_rate": 2.432910516063927e-05, "loss": 0.1446, "step": 33235 }, { "epoch": 0.78, "learning_rate": 2.432822510619324e-05, "loss": 0.188, "step": 33240 }, { "epoch": 0.78, "learning_rate": 2.43273450517472e-05, "loss": 0.2765, "step": 33245 }, { "epoch": 0.78, "learning_rate": 2.4326464997301166e-05, "loss": 0.5607, "step": 33250 }, { "epoch": 0.78, "learning_rate": 2.432558494285513e-05, "loss": 0.1521, "step": 33255 }, { "epoch": 0.78, "learning_rate": 2.43247048884091e-05, "loss": 0.0844, "step": 33260 }, { "epoch": 0.78, "learning_rate": 2.432382483396306e-05, "loss": 0.0881, "step": 33265 }, { "epoch": 0.78, "learning_rate": 2.4322944779517026e-05, "loss": 0.1234, "step": 33270 }, { "epoch": 0.78, "learning_rate": 2.432206472507099e-05, "loss": 0.1328, "step": 33275 }, { "epoch": 0.78, "learning_rate": 2.4321184670624956e-05, "loss": 0.0584, "step": 33280 }, { "epoch": 0.78, "learning_rate": 2.432030461617892e-05, "loss": 0.2365, "step": 33285 }, { "epoch": 0.78, "learning_rate": 2.4319424561732886e-05, "loss": 0.1573, "step": 33290 }, { "epoch": 0.78, "learning_rate": 2.4318544507286854e-05, "loss": 0.1836, "step": 33295 }, { "epoch": 0.78, "learning_rate": 2.4317664452840816e-05, "loss": 0.934, "step": 33300 }, { "epoch": 0.78, "learning_rate": 2.431678439839478e-05, "loss": 0.2363, "step": 33305 }, { "epoch": 0.78, "learning_rate": 2.4315904343948746e-05, "loss": 0.0627, "step": 33310 }, { "epoch": 0.78, "learning_rate": 2.431502428950271e-05, "loss": 0.0948, "step": 33315 }, { "epoch": 0.78, "learning_rate": 2.4314144235056676e-05, "loss": 0.144, "step": 33320 }, { "epoch": 0.78, "learning_rate": 2.431326418061064e-05, "loss": 0.232, "step": 33325 }, { "epoch": 0.78, "learning_rate": 2.4312384126164606e-05, "loss": 0.1286, "step": 33330 }, { "epoch": 0.78, "learning_rate": 2.431150407171857e-05, "loss": 0.198, "step": 33335 }, { "epoch": 0.78, "learning_rate": 2.4310624017272536e-05, "loss": 0.2542, "step": 33340 }, { "epoch": 0.78, "learning_rate": 2.43097439628265e-05, "loss": 0.1912, "step": 33345 }, { "epoch": 0.78, "learning_rate": 2.4308863908380463e-05, "loss": 0.5112, "step": 33350 }, { "epoch": 0.78, "learning_rate": 2.430798385393443e-05, "loss": 0.2312, "step": 33355 }, { "epoch": 0.78, "learning_rate": 2.4307103799488396e-05, "loss": 0.0739, "step": 33360 }, { "epoch": 0.78, "learning_rate": 2.430622374504236e-05, "loss": 0.1199, "step": 33365 }, { "epoch": 0.78, "learning_rate": 2.4305343690596323e-05, "loss": 0.0523, "step": 33370 }, { "epoch": 0.78, "learning_rate": 2.430446363615029e-05, "loss": 0.2019, "step": 33375 }, { "epoch": 0.78, "learning_rate": 2.4303583581704256e-05, "loss": 0.2309, "step": 33380 }, { "epoch": 0.78, "learning_rate": 2.4302703527258218e-05, "loss": 0.178, "step": 33385 }, { "epoch": 0.78, "learning_rate": 2.4301823472812183e-05, "loss": 0.2024, "step": 33390 }, { "epoch": 0.78, "learning_rate": 2.430094341836615e-05, "loss": 0.2811, "step": 33395 }, { "epoch": 0.78, "learning_rate": 2.4300063363920116e-05, "loss": 0.5188, "step": 33400 }, { "epoch": 0.78, "learning_rate": 2.4299183309474078e-05, "loss": 0.2606, "step": 33405 }, { "epoch": 0.78, "learning_rate": 2.4298303255028047e-05, "loss": 0.0795, "step": 33410 }, { "epoch": 0.78, "learning_rate": 2.429742320058201e-05, "loss": 0.0753, "step": 33415 }, { "epoch": 0.78, "learning_rate": 2.4296543146135973e-05, "loss": 0.0817, "step": 33420 }, { "epoch": 0.78, "learning_rate": 2.4295663091689938e-05, "loss": 0.1952, "step": 33425 }, { "epoch": 0.78, "learning_rate": 2.4294783037243907e-05, "loss": 0.3374, "step": 33430 }, { "epoch": 0.78, "learning_rate": 2.429390298279787e-05, "loss": 0.2435, "step": 33435 }, { "epoch": 0.78, "learning_rate": 2.4293022928351833e-05, "loss": 0.2711, "step": 33440 }, { "epoch": 0.78, "learning_rate": 2.4292142873905798e-05, "loss": 0.2161, "step": 33445 }, { "epoch": 0.78, "learning_rate": 2.4291262819459767e-05, "loss": 0.4422, "step": 33450 }, { "epoch": 0.78, "learning_rate": 2.4290382765013732e-05, "loss": 0.2137, "step": 33455 }, { "epoch": 0.78, "learning_rate": 2.4289502710567693e-05, "loss": 0.2552, "step": 33460 }, { "epoch": 0.78, "learning_rate": 2.428862265612166e-05, "loss": 0.135, "step": 33465 }, { "epoch": 0.78, "learning_rate": 2.4287742601675627e-05, "loss": 0.0777, "step": 33470 }, { "epoch": 0.78, "learning_rate": 2.428686254722959e-05, "loss": 0.116, "step": 33475 }, { "epoch": 0.78, "learning_rate": 2.4285982492783553e-05, "loss": 0.2157, "step": 33480 }, { "epoch": 0.78, "learning_rate": 2.428510243833752e-05, "loss": 0.2343, "step": 33485 }, { "epoch": 0.78, "learning_rate": 2.4284222383891487e-05, "loss": 0.1969, "step": 33490 }, { "epoch": 0.78, "learning_rate": 2.428334232944545e-05, "loss": 0.4306, "step": 33495 }, { "epoch": 0.78, "learning_rate": 2.4282462274999414e-05, "loss": 0.5244, "step": 33500 }, { "epoch": 0.78, "learning_rate": 2.428158222055338e-05, "loss": 0.2676, "step": 33505 }, { "epoch": 0.78, "learning_rate": 2.4280702166107344e-05, "loss": 0.1037, "step": 33510 }, { "epoch": 0.78, "learning_rate": 2.427982211166131e-05, "loss": 0.1183, "step": 33515 }, { "epoch": 0.78, "learning_rate": 2.4278942057215274e-05, "loss": 0.172, "step": 33520 }, { "epoch": 0.78, "learning_rate": 2.427806200276924e-05, "loss": 0.069, "step": 33525 }, { "epoch": 0.78, "learning_rate": 2.4277181948323204e-05, "loss": 0.326, "step": 33530 }, { "epoch": 0.78, "learning_rate": 2.427630189387717e-05, "loss": 0.2006, "step": 33535 }, { "epoch": 0.78, "learning_rate": 2.4275421839431134e-05, "loss": 0.3009, "step": 33540 }, { "epoch": 0.78, "learning_rate": 2.42745417849851e-05, "loss": 0.2216, "step": 33545 }, { "epoch": 0.78, "learning_rate": 2.4273661730539064e-05, "loss": 0.4696, "step": 33550 }, { "epoch": 0.78, "learning_rate": 2.427278167609303e-05, "loss": 0.157, "step": 33555 }, { "epoch": 0.78, "learning_rate": 2.4271901621646994e-05, "loss": 0.0303, "step": 33560 }, { "epoch": 0.78, "learning_rate": 2.427102156720096e-05, "loss": 0.0939, "step": 33565 }, { "epoch": 0.78, "learning_rate": 2.4270141512754924e-05, "loss": 0.0413, "step": 33570 }, { "epoch": 0.78, "learning_rate": 2.426926145830889e-05, "loss": 0.0772, "step": 33575 }, { "epoch": 0.78, "learning_rate": 2.426838140386285e-05, "loss": 0.1668, "step": 33580 }, { "epoch": 0.78, "learning_rate": 2.426750134941682e-05, "loss": 0.2652, "step": 33585 }, { "epoch": 0.78, "learning_rate": 2.4266621294970784e-05, "loss": 0.3847, "step": 33590 }, { "epoch": 0.78, "learning_rate": 2.426574124052475e-05, "loss": 0.2043, "step": 33595 }, { "epoch": 0.78, "learning_rate": 2.426486118607871e-05, "loss": 0.4801, "step": 33600 }, { "epoch": 0.78, "learning_rate": 2.426398113163268e-05, "loss": 0.1364, "step": 33605 }, { "epoch": 0.78, "learning_rate": 2.4263101077186644e-05, "loss": 0.1986, "step": 33610 }, { "epoch": 0.78, "learning_rate": 2.4262221022740606e-05, "loss": 0.0724, "step": 33615 }, { "epoch": 0.78, "learning_rate": 2.426134096829457e-05, "loss": 0.1285, "step": 33620 }, { "epoch": 0.78, "learning_rate": 2.426046091384854e-05, "loss": 0.0914, "step": 33625 }, { "epoch": 0.78, "learning_rate": 2.4259580859402504e-05, "loss": 0.1407, "step": 33630 }, { "epoch": 0.78, "learning_rate": 2.4258700804956466e-05, "loss": 0.1143, "step": 33635 }, { "epoch": 0.78, "learning_rate": 2.425782075051043e-05, "loss": 0.3442, "step": 33640 }, { "epoch": 0.78, "learning_rate": 2.42569406960644e-05, "loss": 0.2532, "step": 33645 }, { "epoch": 0.79, "learning_rate": 2.425606064161836e-05, "loss": 0.6686, "step": 33650 }, { "epoch": 0.79, "learning_rate": 2.4255180587172326e-05, "loss": 0.1753, "step": 33655 }, { "epoch": 0.79, "learning_rate": 2.425430053272629e-05, "loss": 0.1211, "step": 33660 }, { "epoch": 0.79, "learning_rate": 2.425342047828026e-05, "loss": 0.0349, "step": 33665 }, { "epoch": 0.79, "learning_rate": 2.425254042383422e-05, "loss": 0.2256, "step": 33670 }, { "epoch": 0.79, "learning_rate": 2.4251660369388186e-05, "loss": 0.1411, "step": 33675 }, { "epoch": 0.79, "learning_rate": 2.425078031494215e-05, "loss": 0.1934, "step": 33680 }, { "epoch": 0.79, "learning_rate": 2.4249900260496116e-05, "loss": 0.1571, "step": 33685 }, { "epoch": 0.79, "learning_rate": 2.424902020605008e-05, "loss": 0.3169, "step": 33690 }, { "epoch": 0.79, "learning_rate": 2.4248140151604046e-05, "loss": 0.3925, "step": 33695 }, { "epoch": 0.79, "learning_rate": 2.4247260097158014e-05, "loss": 0.4243, "step": 33700 }, { "epoch": 0.79, "learning_rate": 2.4246380042711976e-05, "loss": 0.2778, "step": 33705 }, { "epoch": 0.79, "learning_rate": 2.424549998826594e-05, "loss": 0.1241, "step": 33710 }, { "epoch": 0.79, "learning_rate": 2.4244619933819906e-05, "loss": 0.0469, "step": 33715 }, { "epoch": 0.79, "learning_rate": 2.424373987937387e-05, "loss": 0.076, "step": 33720 }, { "epoch": 0.79, "learning_rate": 2.4242859824927836e-05, "loss": 0.2108, "step": 33725 }, { "epoch": 0.79, "learning_rate": 2.42419797704818e-05, "loss": 0.1842, "step": 33730 }, { "epoch": 0.79, "learning_rate": 2.4241099716035766e-05, "loss": 0.2363, "step": 33735 }, { "epoch": 0.79, "learning_rate": 2.424021966158973e-05, "loss": 0.1962, "step": 33740 }, { "epoch": 0.79, "learning_rate": 2.4239339607143696e-05, "loss": 0.2267, "step": 33745 }, { "epoch": 0.79, "learning_rate": 2.423845955269766e-05, "loss": 0.4275, "step": 33750 }, { "epoch": 0.79, "learning_rate": 2.4237579498251626e-05, "loss": 0.1762, "step": 33755 }, { "epoch": 0.79, "learning_rate": 2.423669944380559e-05, "loss": 0.0839, "step": 33760 }, { "epoch": 0.79, "learning_rate": 2.4235819389359556e-05, "loss": 0.1696, "step": 33765 }, { "epoch": 0.79, "learning_rate": 2.423493933491352e-05, "loss": 0.0352, "step": 33770 }, { "epoch": 0.79, "learning_rate": 2.4234059280467483e-05, "loss": 0.1676, "step": 33775 }, { "epoch": 0.79, "learning_rate": 2.423317922602145e-05, "loss": 0.1504, "step": 33780 }, { "epoch": 0.79, "learning_rate": 2.4232299171575416e-05, "loss": 0.1801, "step": 33785 }, { "epoch": 0.79, "learning_rate": 2.423141911712938e-05, "loss": 0.214, "step": 33790 }, { "epoch": 0.79, "learning_rate": 2.4230539062683343e-05, "loss": 0.3654, "step": 33795 }, { "epoch": 0.79, "learning_rate": 2.422965900823731e-05, "loss": 0.3967, "step": 33800 }, { "epoch": 0.79, "learning_rate": 2.4228778953791276e-05, "loss": 0.178, "step": 33805 }, { "epoch": 0.79, "learning_rate": 2.4227898899345238e-05, "loss": 0.2621, "step": 33810 }, { "epoch": 0.79, "learning_rate": 2.4227018844899203e-05, "loss": 0.0833, "step": 33815 }, { "epoch": 0.79, "learning_rate": 2.422613879045317e-05, "loss": 0.1042, "step": 33820 }, { "epoch": 0.79, "learning_rate": 2.4225258736007136e-05, "loss": 0.1723, "step": 33825 }, { "epoch": 0.79, "learning_rate": 2.4224378681561098e-05, "loss": 0.2189, "step": 33830 }, { "epoch": 0.79, "learning_rate": 2.4223498627115066e-05, "loss": 0.2412, "step": 33835 }, { "epoch": 0.79, "learning_rate": 2.422261857266903e-05, "loss": 0.3153, "step": 33840 }, { "epoch": 0.79, "learning_rate": 2.4221738518222993e-05, "loss": 0.2736, "step": 33845 }, { "epoch": 0.79, "learning_rate": 2.4220858463776958e-05, "loss": 0.6544, "step": 33850 }, { "epoch": 0.79, "learning_rate": 2.4219978409330927e-05, "loss": 0.1776, "step": 33855 }, { "epoch": 0.79, "learning_rate": 2.421909835488489e-05, "loss": 0.0709, "step": 33860 }, { "epoch": 0.79, "learning_rate": 2.4218218300438853e-05, "loss": 0.2923, "step": 33865 }, { "epoch": 0.79, "learning_rate": 2.4217338245992818e-05, "loss": 0.1577, "step": 33870 }, { "epoch": 0.79, "learning_rate": 2.4216458191546787e-05, "loss": 0.0834, "step": 33875 }, { "epoch": 0.79, "learning_rate": 2.4215578137100748e-05, "loss": 0.1908, "step": 33880 }, { "epoch": 0.79, "learning_rate": 2.4214698082654713e-05, "loss": 0.2076, "step": 33885 }, { "epoch": 0.79, "learning_rate": 2.421381802820868e-05, "loss": 0.166, "step": 33890 }, { "epoch": 0.79, "learning_rate": 2.4212937973762647e-05, "loss": 0.4458, "step": 33895 }, { "epoch": 0.79, "learning_rate": 2.421205791931661e-05, "loss": 0.6997, "step": 33900 }, { "epoch": 0.79, "learning_rate": 2.4211177864870573e-05, "loss": 0.1787, "step": 33905 }, { "epoch": 0.79, "learning_rate": 2.421029781042454e-05, "loss": 0.2128, "step": 33910 }, { "epoch": 0.79, "learning_rate": 2.4209417755978503e-05, "loss": 0.1385, "step": 33915 }, { "epoch": 0.79, "learning_rate": 2.420853770153247e-05, "loss": 0.1024, "step": 33920 }, { "epoch": 0.79, "learning_rate": 2.4207657647086433e-05, "loss": 0.3477, "step": 33925 }, { "epoch": 0.79, "learning_rate": 2.42067775926404e-05, "loss": 0.2191, "step": 33930 }, { "epoch": 0.79, "learning_rate": 2.4205897538194364e-05, "loss": 0.2091, "step": 33935 }, { "epoch": 0.79, "learning_rate": 2.420501748374833e-05, "loss": 0.3685, "step": 33940 }, { "epoch": 0.79, "learning_rate": 2.4204137429302294e-05, "loss": 0.283, "step": 33945 }, { "epoch": 0.79, "learning_rate": 2.420325737485626e-05, "loss": 0.286, "step": 33950 }, { "epoch": 0.79, "learning_rate": 2.4202377320410224e-05, "loss": 0.1949, "step": 33955 }, { "epoch": 0.79, "learning_rate": 2.420149726596419e-05, "loss": 0.0493, "step": 33960 }, { "epoch": 0.79, "learning_rate": 2.4200617211518154e-05, "loss": 0.174, "step": 33965 }, { "epoch": 0.79, "learning_rate": 2.419973715707212e-05, "loss": 0.1038, "step": 33970 }, { "epoch": 0.79, "learning_rate": 2.4198857102626084e-05, "loss": 0.1718, "step": 33975 }, { "epoch": 0.79, "learning_rate": 2.419797704818005e-05, "loss": 0.1665, "step": 33980 }, { "epoch": 0.79, "learning_rate": 2.419709699373401e-05, "loss": 0.2792, "step": 33985 }, { "epoch": 0.79, "learning_rate": 2.419621693928798e-05, "loss": 0.2007, "step": 33990 }, { "epoch": 0.79, "learning_rate": 2.4195336884841944e-05, "loss": 0.2362, "step": 33995 }, { "epoch": 0.79, "learning_rate": 2.419445683039591e-05, "loss": 0.4423, "step": 34000 }, { "epoch": 0.79, "learning_rate": 2.419357677594987e-05, "loss": 0.191, "step": 34005 }, { "epoch": 0.79, "learning_rate": 2.419269672150384e-05, "loss": 0.0328, "step": 34010 }, { "epoch": 0.79, "learning_rate": 2.4191816667057804e-05, "loss": 0.0585, "step": 34015 }, { "epoch": 0.79, "learning_rate": 2.4190936612611765e-05, "loss": 0.1082, "step": 34020 }, { "epoch": 0.79, "learning_rate": 2.419005655816573e-05, "loss": 0.0359, "step": 34025 }, { "epoch": 0.79, "learning_rate": 2.41891765037197e-05, "loss": 0.1258, "step": 34030 }, { "epoch": 0.79, "learning_rate": 2.4188296449273664e-05, "loss": 0.236, "step": 34035 }, { "epoch": 0.79, "learning_rate": 2.4187416394827626e-05, "loss": 0.4601, "step": 34040 }, { "epoch": 0.79, "learning_rate": 2.418653634038159e-05, "loss": 0.344, "step": 34045 }, { "epoch": 0.79, "learning_rate": 2.418565628593556e-05, "loss": 0.3857, "step": 34050 }, { "epoch": 0.79, "learning_rate": 2.418477623148952e-05, "loss": 0.2063, "step": 34055 }, { "epoch": 0.79, "learning_rate": 2.4183896177043486e-05, "loss": 0.0684, "step": 34060 }, { "epoch": 0.79, "learning_rate": 2.418301612259745e-05, "loss": 0.1587, "step": 34065 }, { "epoch": 0.79, "learning_rate": 2.418213606815142e-05, "loss": 0.2178, "step": 34070 }, { "epoch": 0.8, "learning_rate": 2.418125601370538e-05, "loss": 0.0625, "step": 34075 }, { "epoch": 0.8, "learning_rate": 2.4180375959259346e-05, "loss": 0.1822, "step": 34080 }, { "epoch": 0.8, "learning_rate": 2.417949590481331e-05, "loss": 0.2733, "step": 34085 }, { "epoch": 0.8, "learning_rate": 2.417861585036728e-05, "loss": 0.3505, "step": 34090 }, { "epoch": 0.8, "learning_rate": 2.417773579592124e-05, "loss": 0.3588, "step": 34095 }, { "epoch": 0.8, "learning_rate": 2.4176855741475206e-05, "loss": 0.329, "step": 34100 }, { "epoch": 0.8, "learning_rate": 2.4175975687029174e-05, "loss": 0.2638, "step": 34105 }, { "epoch": 0.8, "learning_rate": 2.4175095632583136e-05, "loss": 0.1512, "step": 34110 }, { "epoch": 0.8, "learning_rate": 2.41742155781371e-05, "loss": 0.1953, "step": 34115 }, { "epoch": 0.8, "learning_rate": 2.4173335523691066e-05, "loss": 0.1418, "step": 34120 }, { "epoch": 0.8, "learning_rate": 2.4172455469245034e-05, "loss": 0.1214, "step": 34125 }, { "epoch": 0.8, "learning_rate": 2.4171575414798996e-05, "loss": 0.1613, "step": 34130 }, { "epoch": 0.8, "learning_rate": 2.417069536035296e-05, "loss": 0.267, "step": 34135 }, { "epoch": 0.8, "learning_rate": 2.4169815305906926e-05, "loss": 0.232, "step": 34140 }, { "epoch": 0.8, "learning_rate": 2.416893525146089e-05, "loss": 0.3299, "step": 34145 }, { "epoch": 0.8, "learning_rate": 2.4168055197014856e-05, "loss": 0.591, "step": 34150 }, { "epoch": 0.8, "learning_rate": 2.416717514256882e-05, "loss": 0.1931, "step": 34155 }, { "epoch": 0.8, "learning_rate": 2.4166295088122786e-05, "loss": 0.0679, "step": 34160 }, { "epoch": 0.8, "learning_rate": 2.416541503367675e-05, "loss": 0.1373, "step": 34165 }, { "epoch": 0.8, "learning_rate": 2.4164534979230716e-05, "loss": 0.2004, "step": 34170 }, { "epoch": 0.8, "learning_rate": 2.416365492478468e-05, "loss": 0.1853, "step": 34175 }, { "epoch": 0.8, "learning_rate": 2.4162774870338643e-05, "loss": 0.2151, "step": 34180 }, { "epoch": 0.8, "learning_rate": 2.416189481589261e-05, "loss": 0.1243, "step": 34185 }, { "epoch": 0.8, "learning_rate": 2.4161014761446576e-05, "loss": 0.3215, "step": 34190 }, { "epoch": 0.8, "learning_rate": 2.416013470700054e-05, "loss": 0.3128, "step": 34195 }, { "epoch": 0.8, "learning_rate": 2.4159254652554503e-05, "loss": 0.371, "step": 34200 }, { "epoch": 0.8, "learning_rate": 2.415837459810847e-05, "loss": 0.1857, "step": 34205 }, { "epoch": 0.8, "learning_rate": 2.4157494543662436e-05, "loss": 0.1085, "step": 34210 }, { "epoch": 0.8, "learning_rate": 2.4156614489216398e-05, "loss": 0.0469, "step": 34215 }, { "epoch": 0.8, "learning_rate": 2.4155734434770363e-05, "loss": 0.1607, "step": 34220 }, { "epoch": 0.8, "learning_rate": 2.415485438032433e-05, "loss": 0.1188, "step": 34225 }, { "epoch": 0.8, "learning_rate": 2.4153974325878296e-05, "loss": 0.2188, "step": 34230 }, { "epoch": 0.8, "learning_rate": 2.4153094271432258e-05, "loss": 0.2661, "step": 34235 }, { "epoch": 0.8, "learning_rate": 2.4152214216986226e-05, "loss": 0.2939, "step": 34240 }, { "epoch": 0.8, "learning_rate": 2.415133416254019e-05, "loss": 0.196, "step": 34245 }, { "epoch": 0.8, "learning_rate": 2.4150454108094153e-05, "loss": 0.5793, "step": 34250 }, { "epoch": 0.8, "learning_rate": 2.4149574053648118e-05, "loss": 0.2748, "step": 34255 }, { "epoch": 0.8, "learning_rate": 2.4148693999202086e-05, "loss": 0.0627, "step": 34260 }, { "epoch": 0.8, "learning_rate": 2.414781394475605e-05, "loss": 0.0909, "step": 34265 }, { "epoch": 0.8, "learning_rate": 2.4146933890310013e-05, "loss": 0.1444, "step": 34270 }, { "epoch": 0.8, "learning_rate": 2.4146053835863978e-05, "loss": 0.1889, "step": 34275 }, { "epoch": 0.8, "learning_rate": 2.4145173781417947e-05, "loss": 0.315, "step": 34280 }, { "epoch": 0.8, "learning_rate": 2.4144293726971908e-05, "loss": 0.1202, "step": 34285 }, { "epoch": 0.8, "learning_rate": 2.4143413672525873e-05, "loss": 0.3126, "step": 34290 }, { "epoch": 0.8, "learning_rate": 2.4142533618079838e-05, "loss": 0.2446, "step": 34295 }, { "epoch": 0.8, "learning_rate": 2.4141653563633807e-05, "loss": 0.3517, "step": 34300 }, { "epoch": 0.8, "learning_rate": 2.4140773509187768e-05, "loss": 0.2266, "step": 34305 }, { "epoch": 0.8, "learning_rate": 2.4139893454741733e-05, "loss": 0.0312, "step": 34310 }, { "epoch": 0.8, "learning_rate": 2.4139013400295698e-05, "loss": 0.1109, "step": 34315 }, { "epoch": 0.8, "learning_rate": 2.4138133345849663e-05, "loss": 0.1606, "step": 34320 }, { "epoch": 0.8, "learning_rate": 2.413725329140363e-05, "loss": 0.1837, "step": 34325 }, { "epoch": 0.8, "learning_rate": 2.4136373236957593e-05, "loss": 0.286, "step": 34330 }, { "epoch": 0.8, "learning_rate": 2.413549318251156e-05, "loss": 0.1711, "step": 34335 }, { "epoch": 0.8, "learning_rate": 2.4134613128065523e-05, "loss": 0.229, "step": 34340 }, { "epoch": 0.8, "learning_rate": 2.413373307361949e-05, "loss": 0.1887, "step": 34345 }, { "epoch": 0.8, "learning_rate": 2.4132853019173453e-05, "loss": 0.5458, "step": 34350 }, { "epoch": 0.8, "learning_rate": 2.4131972964727415e-05, "loss": 0.1992, "step": 34355 }, { "epoch": 0.8, "learning_rate": 2.4131092910281383e-05, "loss": 0.1288, "step": 34360 }, { "epoch": 0.8, "learning_rate": 2.413021285583535e-05, "loss": 0.1594, "step": 34365 }, { "epoch": 0.8, "learning_rate": 2.4129332801389314e-05, "loss": 0.0404, "step": 34370 }, { "epoch": 0.8, "learning_rate": 2.412845274694328e-05, "loss": 0.1549, "step": 34375 }, { "epoch": 0.8, "learning_rate": 2.4127572692497244e-05, "loss": 0.1576, "step": 34380 }, { "epoch": 0.8, "learning_rate": 2.412669263805121e-05, "loss": 0.3095, "step": 34385 }, { "epoch": 0.8, "learning_rate": 2.4125812583605174e-05, "loss": 0.1802, "step": 34390 }, { "epoch": 0.8, "learning_rate": 2.412493252915914e-05, "loss": 0.2475, "step": 34395 }, { "epoch": 0.8, "learning_rate": 2.4124052474713104e-05, "loss": 0.5365, "step": 34400 }, { "epoch": 0.8, "learning_rate": 2.412317242026707e-05, "loss": 0.2143, "step": 34405 }, { "epoch": 0.8, "learning_rate": 2.412229236582103e-05, "loss": 0.0722, "step": 34410 }, { "epoch": 0.8, "learning_rate": 2.4121412311375e-05, "loss": 0.0964, "step": 34415 }, { "epoch": 0.8, "learning_rate": 2.4120532256928964e-05, "loss": 0.163, "step": 34420 }, { "epoch": 0.8, "learning_rate": 2.411965220248293e-05, "loss": 0.0968, "step": 34425 }, { "epoch": 0.8, "learning_rate": 2.411877214803689e-05, "loss": 0.1146, "step": 34430 }, { "epoch": 0.8, "learning_rate": 2.411789209359086e-05, "loss": 0.2253, "step": 34435 }, { "epoch": 0.8, "learning_rate": 2.4117012039144824e-05, "loss": 0.2337, "step": 34440 }, { "epoch": 0.8, "learning_rate": 2.4116131984698785e-05, "loss": 0.2632, "step": 34445 }, { "epoch": 0.8, "learning_rate": 2.411525193025275e-05, "loss": 0.5607, "step": 34450 }, { "epoch": 0.8, "learning_rate": 2.411437187580672e-05, "loss": 0.1747, "step": 34455 }, { "epoch": 0.8, "learning_rate": 2.4113491821360684e-05, "loss": 0.13, "step": 34460 }, { "epoch": 0.8, "learning_rate": 2.4112611766914646e-05, "loss": 0.0561, "step": 34465 }, { "epoch": 0.8, "learning_rate": 2.411173171246861e-05, "loss": 0.1881, "step": 34470 }, { "epoch": 0.8, "learning_rate": 2.411085165802258e-05, "loss": 0.0839, "step": 34475 }, { "epoch": 0.8, "learning_rate": 2.410997160357654e-05, "loss": 0.1924, "step": 34480 }, { "epoch": 0.8, "learning_rate": 2.4109091549130506e-05, "loss": 0.3084, "step": 34485 }, { "epoch": 0.8, "learning_rate": 2.410821149468447e-05, "loss": 0.3148, "step": 34490 }, { "epoch": 0.8, "learning_rate": 2.410733144023844e-05, "loss": 0.2821, "step": 34495 }, { "epoch": 0.8, "learning_rate": 2.41064513857924e-05, "loss": 0.524, "step": 34500 }, { "epoch": 0.81, "learning_rate": 2.4105571331346366e-05, "loss": 0.2305, "step": 34505 }, { "epoch": 0.81, "learning_rate": 2.4104691276900334e-05, "loss": 0.0661, "step": 34510 }, { "epoch": 0.81, "learning_rate": 2.4103811222454296e-05, "loss": 0.133, "step": 34515 }, { "epoch": 0.81, "learning_rate": 2.410293116800826e-05, "loss": 0.1168, "step": 34520 }, { "epoch": 0.81, "learning_rate": 2.4102051113562226e-05, "loss": 0.1928, "step": 34525 }, { "epoch": 0.81, "learning_rate": 2.4101171059116194e-05, "loss": 0.3364, "step": 34530 }, { "epoch": 0.81, "learning_rate": 2.4100291004670156e-05, "loss": 0.328, "step": 34535 }, { "epoch": 0.81, "learning_rate": 2.409941095022412e-05, "loss": 0.2242, "step": 34540 }, { "epoch": 0.81, "learning_rate": 2.4098530895778086e-05, "loss": 0.3893, "step": 34545 }, { "epoch": 0.81, "learning_rate": 2.409765084133205e-05, "loss": 0.2268, "step": 34550 }, { "epoch": 0.81, "learning_rate": 2.4096770786886016e-05, "loss": 0.1753, "step": 34555 }, { "epoch": 0.81, "learning_rate": 2.409589073243998e-05, "loss": 0.0775, "step": 34560 }, { "epoch": 0.81, "learning_rate": 2.4095010677993946e-05, "loss": 0.1685, "step": 34565 }, { "epoch": 0.81, "learning_rate": 2.409413062354791e-05, "loss": 0.1042, "step": 34570 }, { "epoch": 0.81, "learning_rate": 2.4093250569101876e-05, "loss": 0.1513, "step": 34575 }, { "epoch": 0.81, "learning_rate": 2.409237051465584e-05, "loss": 0.1584, "step": 34580 }, { "epoch": 0.81, "learning_rate": 2.4091490460209803e-05, "loss": 0.2247, "step": 34585 }, { "epoch": 0.81, "learning_rate": 2.409061040576377e-05, "loss": 0.2284, "step": 34590 }, { "epoch": 0.81, "learning_rate": 2.4089730351317736e-05, "loss": 0.2018, "step": 34595 }, { "epoch": 0.81, "learning_rate": 2.40888502968717e-05, "loss": 0.3758, "step": 34600 }, { "epoch": 0.81, "learning_rate": 2.4087970242425663e-05, "loss": 0.2143, "step": 34605 }, { "epoch": 0.81, "learning_rate": 2.408709018797963e-05, "loss": 0.1194, "step": 34610 }, { "epoch": 0.81, "learning_rate": 2.4086210133533596e-05, "loss": 0.1665, "step": 34615 }, { "epoch": 0.81, "learning_rate": 2.4085330079087558e-05, "loss": 0.0842, "step": 34620 }, { "epoch": 0.81, "learning_rate": 2.4084450024641523e-05, "loss": 0.178, "step": 34625 }, { "epoch": 0.81, "learning_rate": 2.408356997019549e-05, "loss": 0.1275, "step": 34630 }, { "epoch": 0.81, "learning_rate": 2.4082689915749456e-05, "loss": 0.3056, "step": 34635 }, { "epoch": 0.81, "learning_rate": 2.4081809861303418e-05, "loss": 0.2785, "step": 34640 }, { "epoch": 0.81, "learning_rate": 2.4080929806857386e-05, "loss": 0.2697, "step": 34645 }, { "epoch": 0.81, "learning_rate": 2.408004975241135e-05, "loss": 0.3213, "step": 34650 }, { "epoch": 0.81, "learning_rate": 2.4079169697965313e-05, "loss": 0.2291, "step": 34655 }, { "epoch": 0.81, "learning_rate": 2.4078289643519278e-05, "loss": 0.0554, "step": 34660 }, { "epoch": 0.81, "learning_rate": 2.4077409589073246e-05, "loss": 0.0777, "step": 34665 }, { "epoch": 0.81, "learning_rate": 2.407652953462721e-05, "loss": 0.1262, "step": 34670 }, { "epoch": 0.81, "learning_rate": 2.4075649480181173e-05, "loss": 0.1182, "step": 34675 }, { "epoch": 0.81, "learning_rate": 2.4074769425735138e-05, "loss": 0.1289, "step": 34680 }, { "epoch": 0.81, "learning_rate": 2.4073889371289106e-05, "loss": 0.1602, "step": 34685 }, { "epoch": 0.81, "learning_rate": 2.407300931684307e-05, "loss": 0.2302, "step": 34690 }, { "epoch": 0.81, "learning_rate": 2.4072129262397033e-05, "loss": 0.4614, "step": 34695 }, { "epoch": 0.81, "learning_rate": 2.4071249207950998e-05, "loss": 0.8337, "step": 34700 }, { "epoch": 0.81, "learning_rate": 2.4070369153504967e-05, "loss": 0.2109, "step": 34705 }, { "epoch": 0.81, "learning_rate": 2.4069489099058928e-05, "loss": 0.1541, "step": 34710 }, { "epoch": 0.81, "learning_rate": 2.4068609044612893e-05, "loss": 0.043, "step": 34715 }, { "epoch": 0.81, "learning_rate": 2.4067728990166858e-05, "loss": 0.1134, "step": 34720 }, { "epoch": 0.81, "learning_rate": 2.4066848935720827e-05, "loss": 0.1822, "step": 34725 }, { "epoch": 0.81, "learning_rate": 2.4065968881274788e-05, "loss": 0.1714, "step": 34730 }, { "epoch": 0.81, "learning_rate": 2.4065088826828753e-05, "loss": 0.2363, "step": 34735 }, { "epoch": 0.81, "learning_rate": 2.4064208772382718e-05, "loss": 0.1782, "step": 34740 }, { "epoch": 0.81, "learning_rate": 2.4063328717936683e-05, "loss": 0.5202, "step": 34745 }, { "epoch": 0.81, "learning_rate": 2.406244866349065e-05, "loss": 0.3129, "step": 34750 }, { "epoch": 0.81, "learning_rate": 2.4061568609044613e-05, "loss": 0.2613, "step": 34755 }, { "epoch": 0.81, "learning_rate": 2.406068855459858e-05, "loss": 0.0717, "step": 34760 }, { "epoch": 0.81, "learning_rate": 2.4059808500152543e-05, "loss": 0.1019, "step": 34765 }, { "epoch": 0.81, "learning_rate": 2.405892844570651e-05, "loss": 0.0642, "step": 34770 }, { "epoch": 0.81, "learning_rate": 2.4058048391260473e-05, "loss": 0.0986, "step": 34775 }, { "epoch": 0.81, "learning_rate": 2.405716833681444e-05, "loss": 0.1488, "step": 34780 }, { "epoch": 0.81, "learning_rate": 2.4056288282368403e-05, "loss": 0.2835, "step": 34785 }, { "epoch": 0.81, "learning_rate": 2.405540822792237e-05, "loss": 0.3043, "step": 34790 }, { "epoch": 0.81, "learning_rate": 2.4054528173476333e-05, "loss": 0.2557, "step": 34795 }, { "epoch": 0.81, "learning_rate": 2.40536481190303e-05, "loss": 0.302, "step": 34800 }, { "epoch": 0.81, "learning_rate": 2.4052768064584264e-05, "loss": 0.2662, "step": 34805 }, { "epoch": 0.81, "learning_rate": 2.405188801013823e-05, "loss": 0.0472, "step": 34810 }, { "epoch": 0.81, "learning_rate": 2.405100795569219e-05, "loss": 0.1854, "step": 34815 }, { "epoch": 0.81, "learning_rate": 2.405012790124616e-05, "loss": 0.1352, "step": 34820 }, { "epoch": 0.81, "learning_rate": 2.4049247846800124e-05, "loss": 0.1809, "step": 34825 }, { "epoch": 0.81, "learning_rate": 2.404836779235409e-05, "loss": 0.3621, "step": 34830 }, { "epoch": 0.81, "learning_rate": 2.404748773790805e-05, "loss": 0.1874, "step": 34835 }, { "epoch": 0.81, "learning_rate": 2.404660768346202e-05, "loss": 0.2071, "step": 34840 }, { "epoch": 0.81, "learning_rate": 2.4045727629015984e-05, "loss": 0.4357, "step": 34845 }, { "epoch": 0.81, "learning_rate": 2.4044847574569945e-05, "loss": 0.4973, "step": 34850 }, { "epoch": 0.81, "learning_rate": 2.404396752012391e-05, "loss": 0.2153, "step": 34855 }, { "epoch": 0.81, "learning_rate": 2.404308746567788e-05, "loss": 0.0506, "step": 34860 }, { "epoch": 0.81, "learning_rate": 2.4042207411231844e-05, "loss": 0.0958, "step": 34865 }, { "epoch": 0.81, "learning_rate": 2.4041327356785805e-05, "loss": 0.1362, "step": 34870 }, { "epoch": 0.81, "learning_rate": 2.404044730233977e-05, "loss": 0.0875, "step": 34875 }, { "epoch": 0.81, "learning_rate": 2.403956724789374e-05, "loss": 0.122, "step": 34880 }, { "epoch": 0.81, "learning_rate": 2.40386871934477e-05, "loss": 0.2793, "step": 34885 }, { "epoch": 0.81, "learning_rate": 2.4037807139001665e-05, "loss": 0.2661, "step": 34890 }, { "epoch": 0.81, "learning_rate": 2.403692708455563e-05, "loss": 0.3149, "step": 34895 }, { "epoch": 0.81, "learning_rate": 2.40360470301096e-05, "loss": 0.8861, "step": 34900 }, { "epoch": 0.81, "learning_rate": 2.403516697566356e-05, "loss": 0.1939, "step": 34905 }, { "epoch": 0.81, "learning_rate": 2.4034286921217526e-05, "loss": 0.1048, "step": 34910 }, { "epoch": 0.81, "learning_rate": 2.4033406866771494e-05, "loss": 0.1007, "step": 34915 }, { "epoch": 0.81, "learning_rate": 2.4032526812325456e-05, "loss": 0.1426, "step": 34920 }, { "epoch": 0.81, "learning_rate": 2.403164675787942e-05, "loss": 0.1278, "step": 34925 }, { "epoch": 0.81, "learning_rate": 2.4030766703433386e-05, "loss": 0.1617, "step": 34930 }, { "epoch": 0.82, "learning_rate": 2.4029886648987354e-05, "loss": 0.2104, "step": 34935 }, { "epoch": 0.82, "learning_rate": 2.4029006594541316e-05, "loss": 0.1199, "step": 34940 }, { "epoch": 0.82, "learning_rate": 2.402812654009528e-05, "loss": 0.2962, "step": 34945 }, { "epoch": 0.82, "learning_rate": 2.4027246485649246e-05, "loss": 0.4396, "step": 34950 }, { "epoch": 0.82, "learning_rate": 2.402636643120321e-05, "loss": 0.3141, "step": 34955 }, { "epoch": 0.82, "learning_rate": 2.4025486376757176e-05, "loss": 0.0303, "step": 34960 }, { "epoch": 0.82, "learning_rate": 2.402460632231114e-05, "loss": 0.051, "step": 34965 }, { "epoch": 0.82, "learning_rate": 2.4023726267865106e-05, "loss": 0.0753, "step": 34970 }, { "epoch": 0.82, "learning_rate": 2.402284621341907e-05, "loss": 0.2448, "step": 34975 }, { "epoch": 0.82, "learning_rate": 2.4021966158973036e-05, "loss": 0.0931, "step": 34980 }, { "epoch": 0.82, "learning_rate": 2.4021086104527e-05, "loss": 0.136, "step": 34985 }, { "epoch": 0.82, "learning_rate": 2.4020206050080966e-05, "loss": 0.3717, "step": 34990 }, { "epoch": 0.82, "learning_rate": 2.401932599563493e-05, "loss": 0.3484, "step": 34995 }, { "epoch": 0.82, "learning_rate": 2.4018445941188896e-05, "loss": 0.4953, "step": 35000 }, { "epoch": 0.82, "learning_rate": 2.401756588674286e-05, "loss": 0.2646, "step": 35005 }, { "epoch": 0.82, "learning_rate": 2.4016685832296823e-05, "loss": 0.0882, "step": 35010 }, { "epoch": 0.82, "learning_rate": 2.401580577785079e-05, "loss": 0.1824, "step": 35015 }, { "epoch": 0.82, "learning_rate": 2.4014925723404756e-05, "loss": 0.0384, "step": 35020 }, { "epoch": 0.82, "learning_rate": 2.401404566895872e-05, "loss": 0.1612, "step": 35025 }, { "epoch": 0.82, "learning_rate": 2.4013165614512683e-05, "loss": 0.0931, "step": 35030 }, { "epoch": 0.82, "learning_rate": 2.401228556006665e-05, "loss": 0.2123, "step": 35035 }, { "epoch": 0.82, "learning_rate": 2.4011405505620616e-05, "loss": 0.17, "step": 35040 }, { "epoch": 0.82, "learning_rate": 2.4010525451174578e-05, "loss": 0.1893, "step": 35045 }, { "epoch": 0.82, "learning_rate": 2.4009645396728546e-05, "loss": 0.3688, "step": 35050 }, { "epoch": 0.82, "learning_rate": 2.400876534228251e-05, "loss": 0.2564, "step": 35055 }, { "epoch": 0.82, "learning_rate": 2.4007885287836476e-05, "loss": 0.097, "step": 35060 }, { "epoch": 0.82, "learning_rate": 2.4007005233390438e-05, "loss": 0.0478, "step": 35065 }, { "epoch": 0.82, "learning_rate": 2.4006125178944406e-05, "loss": 0.0381, "step": 35070 }, { "epoch": 0.82, "learning_rate": 2.400524512449837e-05, "loss": 0.2643, "step": 35075 }, { "epoch": 0.82, "learning_rate": 2.4004365070052333e-05, "loss": 0.1207, "step": 35080 }, { "epoch": 0.82, "learning_rate": 2.4003485015606298e-05, "loss": 0.1931, "step": 35085 }, { "epoch": 0.82, "learning_rate": 2.4002604961160266e-05, "loss": 0.3796, "step": 35090 }, { "epoch": 0.82, "learning_rate": 2.400172490671423e-05, "loss": 0.2255, "step": 35095 }, { "epoch": 0.82, "learning_rate": 2.4000844852268193e-05, "loss": 0.4872, "step": 35100 }, { "epoch": 0.82, "learning_rate": 2.3999964797822158e-05, "loss": 0.2968, "step": 35105 }, { "epoch": 0.82, "learning_rate": 2.3999084743376126e-05, "loss": 0.1039, "step": 35110 }, { "epoch": 0.82, "learning_rate": 2.3998204688930088e-05, "loss": 0.0773, "step": 35115 }, { "epoch": 0.82, "learning_rate": 2.3997324634484053e-05, "loss": 0.1179, "step": 35120 }, { "epoch": 0.82, "learning_rate": 2.3996444580038018e-05, "loss": 0.1246, "step": 35125 }, { "epoch": 0.82, "learning_rate": 2.3995564525591986e-05, "loss": 0.2294, "step": 35130 }, { "epoch": 0.82, "learning_rate": 2.3994684471145948e-05, "loss": 0.2225, "step": 35135 }, { "epoch": 0.82, "learning_rate": 2.3993804416699913e-05, "loss": 0.2831, "step": 35140 }, { "epoch": 0.82, "learning_rate": 2.3992924362253878e-05, "loss": 0.3325, "step": 35145 }, { "epoch": 0.82, "learning_rate": 2.3992044307807843e-05, "loss": 0.7097, "step": 35150 }, { "epoch": 0.82, "learning_rate": 2.3991164253361808e-05, "loss": 0.2241, "step": 35155 }, { "epoch": 0.82, "learning_rate": 2.3990284198915773e-05, "loss": 0.0302, "step": 35160 }, { "epoch": 0.82, "learning_rate": 2.3989404144469738e-05, "loss": 0.064, "step": 35165 }, { "epoch": 0.82, "learning_rate": 2.3988524090023703e-05, "loss": 0.1041, "step": 35170 }, { "epoch": 0.82, "learning_rate": 2.3987644035577668e-05, "loss": 0.3001, "step": 35175 }, { "epoch": 0.82, "learning_rate": 2.3986763981131633e-05, "loss": 0.1129, "step": 35180 }, { "epoch": 0.82, "learning_rate": 2.39858839266856e-05, "loss": 0.2117, "step": 35185 }, { "epoch": 0.82, "learning_rate": 2.3985003872239563e-05, "loss": 0.2633, "step": 35190 }, { "epoch": 0.82, "learning_rate": 2.3984299828682735e-05, "loss": 0.2922, "step": 35195 }, { "epoch": 0.82, "learning_rate": 2.39834197742367e-05, "loss": 0.3393, "step": 35200 }, { "epoch": 0.82, "learning_rate": 2.3982539719790665e-05, "loss": 0.1868, "step": 35205 }, { "epoch": 0.82, "learning_rate": 2.398165966534463e-05, "loss": 0.0593, "step": 35210 }, { "epoch": 0.82, "learning_rate": 2.3980779610898595e-05, "loss": 0.1146, "step": 35215 }, { "epoch": 0.82, "learning_rate": 2.397989955645256e-05, "loss": 0.1318, "step": 35220 }, { "epoch": 0.82, "learning_rate": 2.3979019502006525e-05, "loss": 0.1128, "step": 35225 }, { "epoch": 0.82, "learning_rate": 2.397813944756049e-05, "loss": 0.1009, "step": 35230 }, { "epoch": 0.82, "learning_rate": 2.3977259393114455e-05, "loss": 0.0622, "step": 35235 }, { "epoch": 0.82, "learning_rate": 2.397637933866842e-05, "loss": 0.3224, "step": 35240 }, { "epoch": 0.82, "learning_rate": 2.3975499284222385e-05, "loss": 0.2701, "step": 35245 }, { "epoch": 0.82, "learning_rate": 2.397461922977635e-05, "loss": 0.2689, "step": 35250 }, { "epoch": 0.82, "learning_rate": 2.3973739175330315e-05, "loss": 0.2228, "step": 35255 }, { "epoch": 0.82, "learning_rate": 2.397285912088428e-05, "loss": 0.1571, "step": 35260 }, { "epoch": 0.82, "learning_rate": 2.3971979066438245e-05, "loss": 0.1557, "step": 35265 }, { "epoch": 0.82, "learning_rate": 2.397109901199221e-05, "loss": 0.0974, "step": 35270 }, { "epoch": 0.82, "learning_rate": 2.397021895754617e-05, "loss": 0.2589, "step": 35275 }, { "epoch": 0.82, "learning_rate": 2.396933890310014e-05, "loss": 0.1971, "step": 35280 }, { "epoch": 0.82, "learning_rate": 2.3968458848654105e-05, "loss": 0.2543, "step": 35285 }, { "epoch": 0.82, "learning_rate": 2.396757879420807e-05, "loss": 0.405, "step": 35290 }, { "epoch": 0.82, "learning_rate": 2.3966698739762032e-05, "loss": 0.4839, "step": 35295 }, { "epoch": 0.82, "learning_rate": 2.3965818685316e-05, "loss": 0.3374, "step": 35300 }, { "epoch": 0.82, "learning_rate": 2.3964938630869965e-05, "loss": 0.2548, "step": 35305 }, { "epoch": 0.82, "learning_rate": 2.3964058576423927e-05, "loss": 0.0669, "step": 35310 }, { "epoch": 0.82, "learning_rate": 2.3963178521977892e-05, "loss": 0.0544, "step": 35315 }, { "epoch": 0.82, "learning_rate": 2.396229846753186e-05, "loss": 0.1972, "step": 35320 }, { "epoch": 0.82, "learning_rate": 2.3961418413085825e-05, "loss": 0.2693, "step": 35325 }, { "epoch": 0.82, "learning_rate": 2.3960538358639787e-05, "loss": 0.1883, "step": 35330 }, { "epoch": 0.82, "learning_rate": 2.3959658304193752e-05, "loss": 0.2142, "step": 35335 }, { "epoch": 0.82, "learning_rate": 2.395877824974772e-05, "loss": 0.2743, "step": 35340 }, { "epoch": 0.82, "learning_rate": 2.3957898195301682e-05, "loss": 0.3151, "step": 35345 }, { "epoch": 0.82, "learning_rate": 2.3957018140855647e-05, "loss": 0.5882, "step": 35350 }, { "epoch": 0.82, "learning_rate": 2.3956138086409612e-05, "loss": 0.2849, "step": 35355 }, { "epoch": 0.82, "learning_rate": 2.395525803196358e-05, "loss": 0.0475, "step": 35360 }, { "epoch": 0.83, "learning_rate": 2.3954377977517542e-05, "loss": 0.1457, "step": 35365 }, { "epoch": 0.83, "learning_rate": 2.3953497923071507e-05, "loss": 0.1155, "step": 35370 }, { "epoch": 0.83, "learning_rate": 2.3952617868625475e-05, "loss": 0.1111, "step": 35375 }, { "epoch": 0.83, "learning_rate": 2.3951737814179437e-05, "loss": 0.1843, "step": 35380 }, { "epoch": 0.83, "learning_rate": 2.3950857759733402e-05, "loss": 0.2527, "step": 35385 }, { "epoch": 0.83, "learning_rate": 2.3949977705287367e-05, "loss": 0.2743, "step": 35390 }, { "epoch": 0.83, "learning_rate": 2.3949097650841335e-05, "loss": 0.3379, "step": 35395 }, { "epoch": 0.83, "learning_rate": 2.3948217596395297e-05, "loss": 0.5551, "step": 35400 }, { "epoch": 0.83, "learning_rate": 2.3947337541949262e-05, "loss": 0.1871, "step": 35405 }, { "epoch": 0.83, "learning_rate": 2.3946457487503227e-05, "loss": 0.0718, "step": 35410 }, { "epoch": 0.83, "learning_rate": 2.3945577433057192e-05, "loss": 0.0798, "step": 35415 }, { "epoch": 0.83, "learning_rate": 2.3944697378611157e-05, "loss": 0.0618, "step": 35420 }, { "epoch": 0.83, "learning_rate": 2.3943817324165122e-05, "loss": 0.0458, "step": 35425 }, { "epoch": 0.83, "learning_rate": 2.3942937269719087e-05, "loss": 0.1307, "step": 35430 }, { "epoch": 0.83, "learning_rate": 2.3942057215273052e-05, "loss": 0.211, "step": 35435 }, { "epoch": 0.83, "learning_rate": 2.3941177160827017e-05, "loss": 0.2751, "step": 35440 }, { "epoch": 0.83, "learning_rate": 2.3940297106380982e-05, "loss": 0.2407, "step": 35445 }, { "epoch": 0.83, "learning_rate": 2.3939417051934944e-05, "loss": 0.3578, "step": 35450 }, { "epoch": 0.83, "learning_rate": 2.3938536997488912e-05, "loss": 0.243, "step": 35455 }, { "epoch": 0.83, "learning_rate": 2.3937656943042877e-05, "loss": 0.1349, "step": 35460 }, { "epoch": 0.83, "learning_rate": 2.3936776888596842e-05, "loss": 0.0599, "step": 35465 }, { "epoch": 0.83, "learning_rate": 2.3935896834150804e-05, "loss": 0.0304, "step": 35470 }, { "epoch": 0.83, "learning_rate": 2.3935016779704772e-05, "loss": 0.196, "step": 35475 }, { "epoch": 0.83, "learning_rate": 2.3934136725258737e-05, "loss": 0.259, "step": 35480 }, { "epoch": 0.83, "learning_rate": 2.3933256670812702e-05, "loss": 0.2377, "step": 35485 }, { "epoch": 0.83, "learning_rate": 2.3932376616366664e-05, "loss": 0.1475, "step": 35490 }, { "epoch": 0.83, "learning_rate": 2.3931496561920633e-05, "loss": 0.2468, "step": 35495 }, { "epoch": 0.83, "learning_rate": 2.3930616507474598e-05, "loss": 0.3746, "step": 35500 }, { "epoch": 0.83, "learning_rate": 2.392973645302856e-05, "loss": 0.2255, "step": 35505 }, { "epoch": 0.83, "learning_rate": 2.3928856398582528e-05, "loss": 0.1047, "step": 35510 }, { "epoch": 0.83, "learning_rate": 2.3927976344136493e-05, "loss": 0.2, "step": 35515 }, { "epoch": 0.83, "learning_rate": 2.3927096289690458e-05, "loss": 0.1417, "step": 35520 }, { "epoch": 0.83, "learning_rate": 2.392621623524442e-05, "loss": 0.2103, "step": 35525 }, { "epoch": 0.83, "learning_rate": 2.3925336180798388e-05, "loss": 0.1072, "step": 35530 }, { "epoch": 0.83, "learning_rate": 2.3924456126352353e-05, "loss": 0.1302, "step": 35535 }, { "epoch": 0.83, "learning_rate": 2.3923576071906314e-05, "loss": 0.2827, "step": 35540 }, { "epoch": 0.83, "learning_rate": 2.392269601746028e-05, "loss": 0.3075, "step": 35545 }, { "epoch": 0.83, "learning_rate": 2.3921815963014248e-05, "loss": 0.3726, "step": 35550 }, { "epoch": 0.83, "learning_rate": 2.3920935908568213e-05, "loss": 0.1515, "step": 35555 }, { "epoch": 0.83, "learning_rate": 2.3920055854122174e-05, "loss": 0.1071, "step": 35560 }, { "epoch": 0.83, "learning_rate": 2.391917579967614e-05, "loss": 0.1082, "step": 35565 }, { "epoch": 0.83, "learning_rate": 2.3918295745230108e-05, "loss": 0.076, "step": 35570 }, { "epoch": 0.83, "learning_rate": 2.391741569078407e-05, "loss": 0.1472, "step": 35575 }, { "epoch": 0.83, "learning_rate": 2.3916535636338034e-05, "loss": 0.2921, "step": 35580 }, { "epoch": 0.83, "learning_rate": 2.3915655581892e-05, "loss": 0.1582, "step": 35585 }, { "epoch": 0.83, "learning_rate": 2.3914775527445968e-05, "loss": 0.1574, "step": 35590 }, { "epoch": 0.83, "learning_rate": 2.391389547299993e-05, "loss": 0.3587, "step": 35595 }, { "epoch": 0.83, "learning_rate": 2.3913015418553895e-05, "loss": 0.6552, "step": 35600 }, { "epoch": 0.83, "learning_rate": 2.391213536410786e-05, "loss": 0.2722, "step": 35605 }, { "epoch": 0.83, "learning_rate": 2.3911255309661825e-05, "loss": 0.0751, "step": 35610 }, { "epoch": 0.83, "learning_rate": 2.391037525521579e-05, "loss": 0.0867, "step": 35615 }, { "epoch": 0.83, "learning_rate": 2.3909495200769755e-05, "loss": 0.1259, "step": 35620 }, { "epoch": 0.83, "learning_rate": 2.390861514632372e-05, "loss": 0.1847, "step": 35625 }, { "epoch": 0.83, "learning_rate": 2.3907735091877685e-05, "loss": 0.0789, "step": 35630 }, { "epoch": 0.83, "learning_rate": 2.390685503743165e-05, "loss": 0.1895, "step": 35635 }, { "epoch": 0.83, "learning_rate": 2.3905974982985615e-05, "loss": 0.2608, "step": 35640 }, { "epoch": 0.83, "learning_rate": 2.390509492853958e-05, "loss": 0.276, "step": 35645 }, { "epoch": 0.83, "learning_rate": 2.3904214874093545e-05, "loss": 0.4698, "step": 35650 }, { "epoch": 0.83, "learning_rate": 2.390333481964751e-05, "loss": 0.2869, "step": 35655 }, { "epoch": 0.83, "learning_rate": 2.3902454765201475e-05, "loss": 0.0366, "step": 35660 }, { "epoch": 0.83, "learning_rate": 2.390157471075544e-05, "loss": 0.1607, "step": 35665 }, { "epoch": 0.83, "learning_rate": 2.3900694656309405e-05, "loss": 0.0499, "step": 35670 }, { "epoch": 0.83, "learning_rate": 2.389981460186337e-05, "loss": 0.1431, "step": 35675 }, { "epoch": 0.83, "learning_rate": 2.389893454741733e-05, "loss": 0.204, "step": 35680 }, { "epoch": 0.83, "learning_rate": 2.38980544929713e-05, "loss": 0.1762, "step": 35685 }, { "epoch": 0.83, "learning_rate": 2.3897174438525265e-05, "loss": 0.1759, "step": 35690 }, { "epoch": 0.83, "learning_rate": 2.389629438407923e-05, "loss": 0.1441, "step": 35695 }, { "epoch": 0.83, "learning_rate": 2.389541432963319e-05, "loss": 0.3841, "step": 35700 }, { "epoch": 0.83, "learning_rate": 2.389453427518716e-05, "loss": 0.2739, "step": 35705 }, { "epoch": 0.83, "learning_rate": 2.3893654220741125e-05, "loss": 0.118, "step": 35710 }, { "epoch": 0.83, "learning_rate": 2.3892774166295087e-05, "loss": 0.1301, "step": 35715 }, { "epoch": 0.83, "learning_rate": 2.389189411184905e-05, "loss": 0.0762, "step": 35720 }, { "epoch": 0.83, "learning_rate": 2.389101405740302e-05, "loss": 0.1486, "step": 35725 }, { "epoch": 0.83, "learning_rate": 2.3890134002956985e-05, "loss": 0.2, "step": 35730 }, { "epoch": 0.83, "learning_rate": 2.3889253948510947e-05, "loss": 0.2549, "step": 35735 }, { "epoch": 0.83, "learning_rate": 2.3888373894064912e-05, "loss": 0.198, "step": 35740 }, { "epoch": 0.83, "learning_rate": 2.388749383961888e-05, "loss": 0.342, "step": 35745 }, { "epoch": 0.83, "learning_rate": 2.3886613785172842e-05, "loss": 0.5675, "step": 35750 }, { "epoch": 0.83, "learning_rate": 2.3885733730726807e-05, "loss": 0.2041, "step": 35755 }, { "epoch": 0.83, "learning_rate": 2.3884853676280772e-05, "loss": 0.0717, "step": 35760 }, { "epoch": 0.83, "learning_rate": 2.388397362183474e-05, "loss": 0.1093, "step": 35765 }, { "epoch": 0.83, "learning_rate": 2.3883093567388702e-05, "loss": 0.0671, "step": 35770 }, { "epoch": 0.83, "learning_rate": 2.3882213512942667e-05, "loss": 0.1097, "step": 35775 }, { "epoch": 0.83, "learning_rate": 2.3881333458496635e-05, "loss": 0.3905, "step": 35780 }, { "epoch": 0.83, "learning_rate": 2.38804534040506e-05, "loss": 0.1025, "step": 35785 }, { "epoch": 0.84, "learning_rate": 2.3879573349604562e-05, "loss": 0.1942, "step": 35790 }, { "epoch": 0.84, "learning_rate": 2.3878693295158527e-05, "loss": 0.3148, "step": 35795 }, { "epoch": 0.84, "learning_rate": 2.3877813240712495e-05, "loss": 0.5505, "step": 35800 }, { "epoch": 0.84, "learning_rate": 2.3876933186266457e-05, "loss": 0.1306, "step": 35805 }, { "epoch": 0.84, "learning_rate": 2.3876053131820422e-05, "loss": 0.0777, "step": 35810 }, { "epoch": 0.84, "learning_rate": 2.3875173077374387e-05, "loss": 0.1358, "step": 35815 }, { "epoch": 0.84, "learning_rate": 2.3874293022928355e-05, "loss": 0.147, "step": 35820 }, { "epoch": 0.84, "learning_rate": 2.3873412968482317e-05, "loss": 0.2027, "step": 35825 }, { "epoch": 0.84, "learning_rate": 2.3872532914036282e-05, "loss": 0.4283, "step": 35830 }, { "epoch": 0.84, "learning_rate": 2.3871652859590247e-05, "loss": 0.2371, "step": 35835 }, { "epoch": 0.84, "learning_rate": 2.3870772805144212e-05, "loss": 0.1382, "step": 35840 }, { "epoch": 0.84, "learning_rate": 2.3869892750698177e-05, "loss": 0.363, "step": 35845 }, { "epoch": 0.84, "learning_rate": 2.3869012696252142e-05, "loss": 0.5029, "step": 35850 }, { "epoch": 0.84, "learning_rate": 2.3868132641806107e-05, "loss": 0.1467, "step": 35855 }, { "epoch": 0.84, "learning_rate": 2.3867252587360072e-05, "loss": 0.1427, "step": 35860 }, { "epoch": 0.84, "learning_rate": 2.3866372532914037e-05, "loss": 0.0531, "step": 35865 }, { "epoch": 0.84, "learning_rate": 2.386566848935721e-05, "loss": 0.1652, "step": 35870 }, { "epoch": 0.84, "learning_rate": 2.3864788434911174e-05, "loss": 0.1164, "step": 35875 }, { "epoch": 0.84, "learning_rate": 2.386390838046514e-05, "loss": 0.1665, "step": 35880 }, { "epoch": 0.84, "learning_rate": 2.3863028326019104e-05, "loss": 0.1606, "step": 35885 }, { "epoch": 0.84, "learning_rate": 2.386214827157307e-05, "loss": 0.2591, "step": 35890 }, { "epoch": 0.84, "learning_rate": 2.3861268217127034e-05, "loss": 0.563, "step": 35895 }, { "epoch": 0.84, "learning_rate": 2.3860388162681e-05, "loss": 0.3835, "step": 35900 }, { "epoch": 0.84, "learning_rate": 2.3859508108234964e-05, "loss": 0.2425, "step": 35905 }, { "epoch": 0.84, "learning_rate": 2.3858628053788925e-05, "loss": 0.0724, "step": 35910 }, { "epoch": 0.84, "learning_rate": 2.3857747999342894e-05, "loss": 0.042, "step": 35915 }, { "epoch": 0.84, "learning_rate": 2.385686794489686e-05, "loss": 0.1606, "step": 35920 }, { "epoch": 0.84, "learning_rate": 2.3855987890450824e-05, "loss": 0.2419, "step": 35925 }, { "epoch": 0.84, "learning_rate": 2.3855107836004785e-05, "loss": 0.2066, "step": 35930 }, { "epoch": 0.84, "learning_rate": 2.3854227781558754e-05, "loss": 0.2167, "step": 35935 }, { "epoch": 0.84, "learning_rate": 2.385334772711272e-05, "loss": 0.1989, "step": 35940 }, { "epoch": 0.84, "learning_rate": 2.385246767266668e-05, "loss": 0.2319, "step": 35945 }, { "epoch": 0.84, "learning_rate": 2.3851587618220646e-05, "loss": 0.8107, "step": 35950 }, { "epoch": 0.84, "learning_rate": 2.3850707563774614e-05, "loss": 0.1903, "step": 35955 }, { "epoch": 0.84, "learning_rate": 2.384982750932858e-05, "loss": 0.0877, "step": 35960 }, { "epoch": 0.84, "learning_rate": 2.384894745488254e-05, "loss": 0.0525, "step": 35965 }, { "epoch": 0.84, "learning_rate": 2.384806740043651e-05, "loss": 0.0755, "step": 35970 }, { "epoch": 0.84, "learning_rate": 2.3847187345990474e-05, "loss": 0.1337, "step": 35975 }, { "epoch": 0.84, "learning_rate": 2.384630729154444e-05, "loss": 0.1829, "step": 35980 }, { "epoch": 0.84, "learning_rate": 2.38454272370984e-05, "loss": 0.1963, "step": 35985 }, { "epoch": 0.84, "learning_rate": 2.384454718265237e-05, "loss": 0.4189, "step": 35990 }, { "epoch": 0.84, "learning_rate": 2.3843667128206334e-05, "loss": 0.2495, "step": 35995 }, { "epoch": 0.84, "learning_rate": 2.3842787073760296e-05, "loss": 0.5736, "step": 36000 }, { "epoch": 0.84, "learning_rate": 2.384190701931426e-05, "loss": 0.3872, "step": 36005 }, { "epoch": 0.84, "learning_rate": 2.384102696486823e-05, "loss": 0.087, "step": 36010 }, { "epoch": 0.84, "learning_rate": 2.3840146910422194e-05, "loss": 0.072, "step": 36015 }, { "epoch": 0.84, "learning_rate": 2.3839266855976156e-05, "loss": 0.1233, "step": 36020 }, { "epoch": 0.84, "learning_rate": 2.383838680153012e-05, "loss": 0.1683, "step": 36025 }, { "epoch": 0.84, "learning_rate": 2.383750674708409e-05, "loss": 0.0813, "step": 36030 }, { "epoch": 0.84, "learning_rate": 2.383662669263805e-05, "loss": 0.0919, "step": 36035 }, { "epoch": 0.84, "learning_rate": 2.3835746638192016e-05, "loss": 0.1907, "step": 36040 }, { "epoch": 0.84, "learning_rate": 2.383486658374598e-05, "loss": 0.4232, "step": 36045 }, { "epoch": 0.84, "learning_rate": 2.383398652929995e-05, "loss": 0.7401, "step": 36050 }, { "epoch": 0.84, "learning_rate": 2.383310647485391e-05, "loss": 0.1573, "step": 36055 }, { "epoch": 0.84, "learning_rate": 2.3832226420407876e-05, "loss": 0.0857, "step": 36060 }, { "epoch": 0.84, "learning_rate": 2.383134636596184e-05, "loss": 0.0377, "step": 36065 }, { "epoch": 0.84, "learning_rate": 2.3830466311515806e-05, "loss": 0.2062, "step": 36070 }, { "epoch": 0.84, "learning_rate": 2.382958625706977e-05, "loss": 0.1674, "step": 36075 }, { "epoch": 0.84, "learning_rate": 2.3828706202623736e-05, "loss": 0.2295, "step": 36080 }, { "epoch": 0.84, "learning_rate": 2.38278261481777e-05, "loss": 0.1369, "step": 36085 }, { "epoch": 0.84, "learning_rate": 2.3826946093731666e-05, "loss": 0.223, "step": 36090 }, { "epoch": 0.84, "learning_rate": 2.382606603928563e-05, "loss": 0.2136, "step": 36095 }, { "epoch": 0.84, "learning_rate": 2.3825185984839596e-05, "loss": 0.8158, "step": 36100 }, { "epoch": 0.84, "learning_rate": 2.382430593039356e-05, "loss": 0.2472, "step": 36105 }, { "epoch": 0.84, "learning_rate": 2.3823425875947526e-05, "loss": 0.0317, "step": 36110 }, { "epoch": 0.84, "learning_rate": 2.382254582150149e-05, "loss": 0.0935, "step": 36115 }, { "epoch": 0.84, "learning_rate": 2.3821665767055456e-05, "loss": 0.0935, "step": 36120 }, { "epoch": 0.84, "learning_rate": 2.382078571260942e-05, "loss": 0.1603, "step": 36125 }, { "epoch": 0.84, "learning_rate": 2.3819905658163386e-05, "loss": 0.2144, "step": 36130 }, { "epoch": 0.84, "learning_rate": 2.381902560371735e-05, "loss": 0.1283, "step": 36135 }, { "epoch": 0.84, "learning_rate": 2.3818145549271313e-05, "loss": 0.1606, "step": 36140 }, { "epoch": 0.84, "learning_rate": 2.381726549482528e-05, "loss": 0.3779, "step": 36145 }, { "epoch": 0.84, "learning_rate": 2.3816385440379246e-05, "loss": 0.7577, "step": 36150 }, { "epoch": 0.84, "learning_rate": 2.381550538593321e-05, "loss": 0.2103, "step": 36155 }, { "epoch": 0.84, "learning_rate": 2.3814625331487173e-05, "loss": 0.0752, "step": 36160 }, { "epoch": 0.84, "learning_rate": 2.381374527704114e-05, "loss": 0.105, "step": 36165 }, { "epoch": 0.84, "learning_rate": 2.3812865222595106e-05, "loss": 0.1354, "step": 36170 }, { "epoch": 0.84, "learning_rate": 2.3811985168149068e-05, "loss": 0.109, "step": 36175 }, { "epoch": 0.84, "learning_rate": 2.3811105113703033e-05, "loss": 0.1218, "step": 36180 }, { "epoch": 0.84, "learning_rate": 2.3810225059257e-05, "loss": 0.2602, "step": 36185 }, { "epoch": 0.84, "learning_rate": 2.3809345004810967e-05, "loss": 0.22, "step": 36190 }, { "epoch": 0.84, "learning_rate": 2.3808464950364928e-05, "loss": 0.3326, "step": 36195 }, { "epoch": 0.84, "learning_rate": 2.3807584895918893e-05, "loss": 0.2413, "step": 36200 }, { "epoch": 0.84, "learning_rate": 2.380670484147286e-05, "loss": 0.2449, "step": 36205 }, { "epoch": 0.84, "learning_rate": 2.3805824787026823e-05, "loss": 0.0992, "step": 36210 }, { "epoch": 0.84, "learning_rate": 2.3804944732580788e-05, "loss": 0.0548, "step": 36215 }, { "epoch": 0.85, "learning_rate": 2.3804064678134753e-05, "loss": 0.0726, "step": 36220 }, { "epoch": 0.85, "learning_rate": 2.380318462368872e-05, "loss": 0.3189, "step": 36225 }, { "epoch": 0.85, "learning_rate": 2.3802304569242683e-05, "loss": 0.2011, "step": 36230 }, { "epoch": 0.85, "learning_rate": 2.380142451479665e-05, "loss": 0.0983, "step": 36235 }, { "epoch": 0.85, "learning_rate": 2.3800544460350617e-05, "loss": 0.199, "step": 36240 }, { "epoch": 0.85, "learning_rate": 2.379966440590458e-05, "loss": 0.2958, "step": 36245 }, { "epoch": 0.85, "learning_rate": 2.3798784351458543e-05, "loss": 0.4644, "step": 36250 }, { "epoch": 0.85, "learning_rate": 2.379790429701251e-05, "loss": 0.098, "step": 36255 }, { "epoch": 0.85, "learning_rate": 2.3797024242566477e-05, "loss": 0.1583, "step": 36260 }, { "epoch": 0.85, "learning_rate": 2.379614418812044e-05, "loss": 0.0553, "step": 36265 }, { "epoch": 0.85, "learning_rate": 2.3795264133674403e-05, "loss": 0.0553, "step": 36270 }, { "epoch": 0.85, "learning_rate": 2.379438407922837e-05, "loss": 0.1463, "step": 36275 }, { "epoch": 0.85, "learning_rate": 2.3793504024782337e-05, "loss": 0.1561, "step": 36280 }, { "epoch": 0.85, "learning_rate": 2.37926239703363e-05, "loss": 0.149, "step": 36285 }, { "epoch": 0.85, "learning_rate": 2.3791743915890264e-05, "loss": 0.3009, "step": 36290 }, { "epoch": 0.85, "learning_rate": 2.379086386144423e-05, "loss": 0.4891, "step": 36295 }, { "epoch": 0.85, "learning_rate": 2.3789983806998194e-05, "loss": 0.5965, "step": 36300 }, { "epoch": 0.85, "learning_rate": 2.378910375255216e-05, "loss": 0.2805, "step": 36305 }, { "epoch": 0.85, "learning_rate": 2.3788223698106124e-05, "loss": 0.066, "step": 36310 }, { "epoch": 0.85, "learning_rate": 2.378734364366009e-05, "loss": 0.0627, "step": 36315 }, { "epoch": 0.85, "learning_rate": 2.3786463589214054e-05, "loss": 0.2068, "step": 36320 }, { "epoch": 0.85, "learning_rate": 2.378558353476802e-05, "loss": 0.075, "step": 36325 }, { "epoch": 0.85, "learning_rate": 2.3784703480321984e-05, "loss": 0.1805, "step": 36330 }, { "epoch": 0.85, "learning_rate": 2.3783823425875945e-05, "loss": 0.3462, "step": 36335 }, { "epoch": 0.85, "learning_rate": 2.3782943371429914e-05, "loss": 0.1748, "step": 36340 }, { "epoch": 0.85, "learning_rate": 2.378206331698388e-05, "loss": 0.1637, "step": 36345 }, { "epoch": 0.85, "learning_rate": 2.3781183262537844e-05, "loss": 0.6001, "step": 36350 }, { "epoch": 0.85, "learning_rate": 2.3780303208091805e-05, "loss": 0.2375, "step": 36355 }, { "epoch": 0.85, "learning_rate": 2.3779423153645774e-05, "loss": 0.0417, "step": 36360 }, { "epoch": 0.85, "learning_rate": 2.377854309919974e-05, "loss": 0.2119, "step": 36365 }, { "epoch": 0.85, "learning_rate": 2.37776630447537e-05, "loss": 0.1252, "step": 36370 }, { "epoch": 0.85, "learning_rate": 2.377678299030767e-05, "loss": 0.1103, "step": 36375 }, { "epoch": 0.85, "learning_rate": 2.3775902935861634e-05, "loss": 0.0894, "step": 36380 }, { "epoch": 0.85, "learning_rate": 2.37750228814156e-05, "loss": 0.249, "step": 36385 }, { "epoch": 0.85, "learning_rate": 2.377414282696956e-05, "loss": 0.3578, "step": 36390 }, { "epoch": 0.85, "learning_rate": 2.377326277252353e-05, "loss": 0.2586, "step": 36395 }, { "epoch": 0.85, "learning_rate": 2.3772382718077494e-05, "loss": 0.5129, "step": 36400 }, { "epoch": 0.85, "learning_rate": 2.3771502663631456e-05, "loss": 0.2075, "step": 36405 }, { "epoch": 0.85, "learning_rate": 2.377062260918542e-05, "loss": 0.011, "step": 36410 }, { "epoch": 0.85, "learning_rate": 2.376974255473939e-05, "loss": 0.0887, "step": 36415 }, { "epoch": 0.85, "learning_rate": 2.3768862500293354e-05, "loss": 0.2007, "step": 36420 }, { "epoch": 0.85, "learning_rate": 2.3767982445847316e-05, "loss": 0.0474, "step": 36425 }, { "epoch": 0.85, "learning_rate": 2.376710239140128e-05, "loss": 0.2246, "step": 36430 }, { "epoch": 0.85, "learning_rate": 2.376622233695525e-05, "loss": 0.157, "step": 36435 }, { "epoch": 0.85, "learning_rate": 2.376534228250921e-05, "loss": 0.1194, "step": 36440 }, { "epoch": 0.85, "learning_rate": 2.3764462228063176e-05, "loss": 0.4687, "step": 36445 }, { "epoch": 0.85, "learning_rate": 2.376358217361714e-05, "loss": 0.419, "step": 36450 }, { "epoch": 0.85, "learning_rate": 2.376270211917111e-05, "loss": 0.1977, "step": 36455 }, { "epoch": 0.85, "learning_rate": 2.376182206472507e-05, "loss": 0.1798, "step": 36460 }, { "epoch": 0.85, "learning_rate": 2.3760942010279036e-05, "loss": 0.0506, "step": 36465 }, { "epoch": 0.85, "learning_rate": 2.3760061955833e-05, "loss": 0.1889, "step": 36470 }, { "epoch": 0.85, "learning_rate": 2.3759181901386966e-05, "loss": 0.1066, "step": 36475 }, { "epoch": 0.85, "learning_rate": 2.375830184694093e-05, "loss": 0.2502, "step": 36480 }, { "epoch": 0.85, "learning_rate": 2.3757421792494896e-05, "loss": 0.183, "step": 36485 }, { "epoch": 0.85, "learning_rate": 2.375654173804886e-05, "loss": 0.2681, "step": 36490 }, { "epoch": 0.85, "learning_rate": 2.3755661683602826e-05, "loss": 0.1302, "step": 36495 }, { "epoch": 0.85, "learning_rate": 2.375478162915679e-05, "loss": 0.5472, "step": 36500 }, { "epoch": 0.85, "learning_rate": 2.3753901574710756e-05, "loss": 0.0856, "step": 36505 }, { "epoch": 0.85, "learning_rate": 2.375302152026472e-05, "loss": 0.054, "step": 36510 }, { "epoch": 0.85, "learning_rate": 2.3752141465818686e-05, "loss": 0.1971, "step": 36515 }, { "epoch": 0.85, "learning_rate": 2.375126141137265e-05, "loss": 0.1844, "step": 36520 }, { "epoch": 0.85, "learning_rate": 2.3750381356926616e-05, "loss": 0.1996, "step": 36525 }, { "epoch": 0.85, "learning_rate": 2.374950130248058e-05, "loss": 0.2409, "step": 36530 }, { "epoch": 0.85, "learning_rate": 2.3748621248034546e-05, "loss": 0.3797, "step": 36535 }, { "epoch": 0.85, "learning_rate": 2.374774119358851e-05, "loss": 0.3281, "step": 36540 }, { "epoch": 0.85, "learning_rate": 2.3746861139142473e-05, "loss": 0.3018, "step": 36545 }, { "epoch": 0.85, "learning_rate": 2.374598108469644e-05, "loss": 0.2318, "step": 36550 }, { "epoch": 0.85, "learning_rate": 2.3745101030250406e-05, "loss": 0.1669, "step": 36555 }, { "epoch": 0.85, "learning_rate": 2.374422097580437e-05, "loss": 0.081, "step": 36560 }, { "epoch": 0.85, "learning_rate": 2.3743340921358333e-05, "loss": 0.1968, "step": 36565 }, { "epoch": 0.85, "learning_rate": 2.37424608669123e-05, "loss": 0.07, "step": 36570 }, { "epoch": 0.85, "learning_rate": 2.3741580812466266e-05, "loss": 0.2444, "step": 36575 }, { "epoch": 0.85, "learning_rate": 2.374070075802023e-05, "loss": 0.2159, "step": 36580 }, { "epoch": 0.85, "learning_rate": 2.3739820703574193e-05, "loss": 0.3489, "step": 36585 }, { "epoch": 0.85, "learning_rate": 2.373894064912816e-05, "loss": 0.3164, "step": 36590 }, { "epoch": 0.85, "learning_rate": 2.3738060594682126e-05, "loss": 0.3027, "step": 36595 }, { "epoch": 0.85, "learning_rate": 2.3737180540236088e-05, "loss": 0.5394, "step": 36600 }, { "epoch": 0.85, "learning_rate": 2.3736300485790053e-05, "loss": 0.2559, "step": 36605 }, { "epoch": 0.85, "learning_rate": 2.373542043134402e-05, "loss": 0.1238, "step": 36610 }, { "epoch": 0.85, "learning_rate": 2.3734540376897986e-05, "loss": 0.0236, "step": 36615 }, { "epoch": 0.85, "learning_rate": 2.3733660322451948e-05, "loss": 0.0904, "step": 36620 }, { "epoch": 0.85, "learning_rate": 2.3732780268005913e-05, "loss": 0.133, "step": 36625 }, { "epoch": 0.85, "learning_rate": 2.373190021355988e-05, "loss": 0.2456, "step": 36630 }, { "epoch": 0.85, "learning_rate": 2.3731020159113843e-05, "loss": 0.2872, "step": 36635 }, { "epoch": 0.85, "learning_rate": 2.3730140104667808e-05, "loss": 0.1376, "step": 36640 }, { "epoch": 0.85, "learning_rate": 2.3729260050221773e-05, "loss": 0.5384, "step": 36645 }, { "epoch": 0.86, "learning_rate": 2.372837999577574e-05, "loss": 0.3843, "step": 36650 }, { "epoch": 0.86, "learning_rate": 2.3727499941329703e-05, "loss": 0.3229, "step": 36655 }, { "epoch": 0.86, "learning_rate": 2.3726619886883668e-05, "loss": 0.0356, "step": 36660 }, { "epoch": 0.86, "learning_rate": 2.3725739832437637e-05, "loss": 0.0825, "step": 36665 }, { "epoch": 0.86, "learning_rate": 2.37248597779916e-05, "loss": 0.1075, "step": 36670 }, { "epoch": 0.86, "learning_rate": 2.3723979723545563e-05, "loss": 0.0862, "step": 36675 }, { "epoch": 0.86, "learning_rate": 2.372309966909953e-05, "loss": 0.2547, "step": 36680 }, { "epoch": 0.86, "learning_rate": 2.3722219614653497e-05, "loss": 0.1319, "step": 36685 }, { "epoch": 0.86, "learning_rate": 2.372133956020746e-05, "loss": 0.2559, "step": 36690 }, { "epoch": 0.86, "learning_rate": 2.3720459505761423e-05, "loss": 0.3041, "step": 36695 }, { "epoch": 0.86, "learning_rate": 2.371957945131539e-05, "loss": 0.6036, "step": 36700 }, { "epoch": 0.86, "learning_rate": 2.3718699396869353e-05, "loss": 0.2215, "step": 36705 }, { "epoch": 0.86, "learning_rate": 2.371781934242332e-05, "loss": 0.0397, "step": 36710 }, { "epoch": 0.86, "learning_rate": 2.3716939287977284e-05, "loss": 0.0982, "step": 36715 }, { "epoch": 0.86, "learning_rate": 2.371605923353125e-05, "loss": 0.1166, "step": 36720 }, { "epoch": 0.86, "learning_rate": 2.3715179179085214e-05, "loss": 0.1832, "step": 36725 }, { "epoch": 0.86, "learning_rate": 2.371429912463918e-05, "loss": 0.1499, "step": 36730 }, { "epoch": 0.86, "learning_rate": 2.3713419070193144e-05, "loss": 0.2016, "step": 36735 }, { "epoch": 0.86, "learning_rate": 2.3712539015747105e-05, "loss": 0.3331, "step": 36740 }, { "epoch": 0.86, "learning_rate": 2.3711658961301074e-05, "loss": 0.266, "step": 36745 }, { "epoch": 0.86, "learning_rate": 2.371077890685504e-05, "loss": 0.4019, "step": 36750 }, { "epoch": 0.86, "learning_rate": 2.3709898852409004e-05, "loss": 0.2163, "step": 36755 }, { "epoch": 0.86, "learning_rate": 2.3709018797962965e-05, "loss": 0.1808, "step": 36760 }, { "epoch": 0.86, "learning_rate": 2.3708138743516934e-05, "loss": 0.0605, "step": 36765 }, { "epoch": 0.86, "learning_rate": 2.37072586890709e-05, "loss": 0.2214, "step": 36770 }, { "epoch": 0.86, "learning_rate": 2.370637863462486e-05, "loss": 0.0877, "step": 36775 }, { "epoch": 0.86, "learning_rate": 2.370549858017883e-05, "loss": 0.1568, "step": 36780 }, { "epoch": 0.86, "learning_rate": 2.3704618525732794e-05, "loss": 0.1503, "step": 36785 }, { "epoch": 0.86, "learning_rate": 2.370373847128676e-05, "loss": 0.3535, "step": 36790 }, { "epoch": 0.86, "learning_rate": 2.370285841684072e-05, "loss": 0.2402, "step": 36795 }, { "epoch": 0.86, "learning_rate": 2.370197836239469e-05, "loss": 0.5505, "step": 36800 }, { "epoch": 0.86, "learning_rate": 2.3701098307948654e-05, "loss": 0.1749, "step": 36805 }, { "epoch": 0.86, "learning_rate": 2.3700218253502616e-05, "loss": 0.1195, "step": 36810 }, { "epoch": 0.86, "learning_rate": 2.369933819905658e-05, "loss": 0.0501, "step": 36815 }, { "epoch": 0.86, "learning_rate": 2.369845814461055e-05, "loss": 0.1575, "step": 36820 }, { "epoch": 0.86, "learning_rate": 2.3697578090164514e-05, "loss": 0.0413, "step": 36825 }, { "epoch": 0.86, "learning_rate": 2.3696698035718476e-05, "loss": 0.2511, "step": 36830 }, { "epoch": 0.86, "learning_rate": 2.369581798127244e-05, "loss": 0.2315, "step": 36835 }, { "epoch": 0.86, "learning_rate": 2.369493792682641e-05, "loss": 0.3368, "step": 36840 }, { "epoch": 0.86, "learning_rate": 2.369405787238037e-05, "loss": 0.3307, "step": 36845 }, { "epoch": 0.86, "learning_rate": 2.3693177817934336e-05, "loss": 0.3221, "step": 36850 }, { "epoch": 0.86, "learning_rate": 2.36922977634883e-05, "loss": 0.2051, "step": 36855 }, { "epoch": 0.86, "learning_rate": 2.369141770904227e-05, "loss": 0.1168, "step": 36860 }, { "epoch": 0.86, "learning_rate": 2.369053765459623e-05, "loss": 0.0819, "step": 36865 }, { "epoch": 0.86, "learning_rate": 2.3689657600150196e-05, "loss": 0.1069, "step": 36870 }, { "epoch": 0.86, "learning_rate": 2.368877754570416e-05, "loss": 0.1917, "step": 36875 }, { "epoch": 0.86, "learning_rate": 2.368789749125813e-05, "loss": 0.2276, "step": 36880 }, { "epoch": 0.86, "learning_rate": 2.368701743681209e-05, "loss": 0.1405, "step": 36885 }, { "epoch": 0.86, "learning_rate": 2.3686137382366056e-05, "loss": 0.3954, "step": 36890 }, { "epoch": 0.86, "learning_rate": 2.368525732792002e-05, "loss": 0.308, "step": 36895 }, { "epoch": 0.86, "learning_rate": 2.3684377273473986e-05, "loss": 0.3357, "step": 36900 }, { "epoch": 0.86, "learning_rate": 2.368349721902795e-05, "loss": 0.3128, "step": 36905 }, { "epoch": 0.86, "learning_rate": 2.3682617164581916e-05, "loss": 0.0618, "step": 36910 }, { "epoch": 0.86, "learning_rate": 2.368173711013588e-05, "loss": 0.0873, "step": 36915 }, { "epoch": 0.86, "learning_rate": 2.3680857055689846e-05, "loss": 0.1081, "step": 36920 }, { "epoch": 0.86, "learning_rate": 2.367997700124381e-05, "loss": 0.1349, "step": 36925 }, { "epoch": 0.86, "learning_rate": 2.3679096946797776e-05, "loss": 0.1009, "step": 36930 }, { "epoch": 0.86, "learning_rate": 2.367821689235174e-05, "loss": 0.0901, "step": 36935 }, { "epoch": 0.86, "learning_rate": 2.3677336837905706e-05, "loss": 0.1406, "step": 36940 }, { "epoch": 0.86, "learning_rate": 2.367645678345967e-05, "loss": 0.3743, "step": 36945 }, { "epoch": 0.86, "learning_rate": 2.3675576729013636e-05, "loss": 0.5441, "step": 36950 }, { "epoch": 0.86, "learning_rate": 2.36746966745676e-05, "loss": 0.2728, "step": 36955 }, { "epoch": 0.86, "learning_rate": 2.3673816620121566e-05, "loss": 0.0294, "step": 36960 }, { "epoch": 0.86, "learning_rate": 2.367293656567553e-05, "loss": 0.117, "step": 36965 }, { "epoch": 0.86, "learning_rate": 2.3672056511229493e-05, "loss": 0.0825, "step": 36970 }, { "epoch": 0.86, "learning_rate": 2.367117645678346e-05, "loss": 0.1663, "step": 36975 }, { "epoch": 0.86, "learning_rate": 2.3670296402337426e-05, "loss": 0.1303, "step": 36980 }, { "epoch": 0.86, "learning_rate": 2.366941634789139e-05, "loss": 0.131, "step": 36985 }, { "epoch": 0.86, "learning_rate": 2.3668536293445353e-05, "loss": 0.2215, "step": 36990 }, { "epoch": 0.86, "learning_rate": 2.366765623899932e-05, "loss": 0.2328, "step": 36995 }, { "epoch": 0.86, "learning_rate": 2.3666776184553286e-05, "loss": 0.5803, "step": 37000 }, { "epoch": 0.86, "learning_rate": 2.3665896130107248e-05, "loss": 0.2645, "step": 37005 }, { "epoch": 0.86, "learning_rate": 2.3665016075661213e-05, "loss": 0.0356, "step": 37010 }, { "epoch": 0.86, "learning_rate": 2.366413602121518e-05, "loss": 0.0286, "step": 37015 }, { "epoch": 0.86, "learning_rate": 2.3663255966769146e-05, "loss": 0.0561, "step": 37020 }, { "epoch": 0.86, "learning_rate": 2.3662375912323108e-05, "loss": 0.1748, "step": 37025 }, { "epoch": 0.86, "learning_rate": 2.3661495857877073e-05, "loss": 0.1601, "step": 37030 }, { "epoch": 0.86, "learning_rate": 2.366061580343104e-05, "loss": 0.3417, "step": 37035 }, { "epoch": 0.86, "learning_rate": 2.3659735748985003e-05, "loss": 0.1982, "step": 37040 }, { "epoch": 0.86, "learning_rate": 2.3658855694538968e-05, "loss": 0.422, "step": 37045 }, { "epoch": 0.86, "learning_rate": 2.3657975640092933e-05, "loss": 0.4237, "step": 37050 }, { "epoch": 0.86, "learning_rate": 2.36570955856469e-05, "loss": 0.299, "step": 37055 }, { "epoch": 0.86, "learning_rate": 2.3656215531200863e-05, "loss": 0.0999, "step": 37060 }, { "epoch": 0.86, "learning_rate": 2.3655335476754828e-05, "loss": 0.0958, "step": 37065 }, { "epoch": 0.86, "learning_rate": 2.3654455422308797e-05, "loss": 0.1558, "step": 37070 }, { "epoch": 0.86, "learning_rate": 2.3653575367862758e-05, "loss": 0.1361, "step": 37075 }, { "epoch": 0.87, "learning_rate": 2.3652695313416723e-05, "loss": 0.1608, "step": 37080 }, { "epoch": 0.87, "learning_rate": 2.3651815258970688e-05, "loss": 0.266, "step": 37085 }, { "epoch": 0.87, "learning_rate": 2.3650935204524657e-05, "loss": 0.2237, "step": 37090 }, { "epoch": 0.87, "learning_rate": 2.3650055150078618e-05, "loss": 0.1851, "step": 37095 }, { "epoch": 0.87, "learning_rate": 2.3649175095632583e-05, "loss": 0.5233, "step": 37100 }, { "epoch": 0.87, "learning_rate": 2.364829504118655e-05, "loss": 0.1996, "step": 37105 }, { "epoch": 0.87, "learning_rate": 2.3647414986740513e-05, "loss": 0.088, "step": 37110 }, { "epoch": 0.87, "learning_rate": 2.364653493229448e-05, "loss": 0.0999, "step": 37115 }, { "epoch": 0.87, "learning_rate": 2.3645654877848443e-05, "loss": 0.1583, "step": 37120 }, { "epoch": 0.87, "learning_rate": 2.364477482340241e-05, "loss": 0.204, "step": 37125 }, { "epoch": 0.87, "learning_rate": 2.3643894768956373e-05, "loss": 0.1062, "step": 37130 }, { "epoch": 0.87, "learning_rate": 2.364301471451034e-05, "loss": 0.1799, "step": 37135 }, { "epoch": 0.87, "learning_rate": 2.3642134660064303e-05, "loss": 0.3189, "step": 37140 }, { "epoch": 0.87, "learning_rate": 2.3641254605618265e-05, "loss": 0.2876, "step": 37145 }, { "epoch": 0.87, "learning_rate": 2.3640374551172234e-05, "loss": 0.3619, "step": 37150 }, { "epoch": 0.87, "learning_rate": 2.36394944967262e-05, "loss": 0.2806, "step": 37155 }, { "epoch": 0.87, "learning_rate": 2.3638614442280164e-05, "loss": 0.036, "step": 37160 }, { "epoch": 0.87, "learning_rate": 2.3637734387834125e-05, "loss": 0.1472, "step": 37165 }, { "epoch": 0.87, "learning_rate": 2.3636854333388094e-05, "loss": 0.0664, "step": 37170 }, { "epoch": 0.87, "learning_rate": 2.363597427894206e-05, "loss": 0.1785, "step": 37175 }, { "epoch": 0.87, "learning_rate": 2.3635094224496024e-05, "loss": 0.1889, "step": 37180 }, { "epoch": 0.87, "learning_rate": 2.3634214170049985e-05, "loss": 0.1364, "step": 37185 }, { "epoch": 0.87, "learning_rate": 2.3633334115603954e-05, "loss": 0.2688, "step": 37190 }, { "epoch": 0.87, "learning_rate": 2.363245406115792e-05, "loss": 0.3051, "step": 37195 }, { "epoch": 0.87, "learning_rate": 2.363157400671188e-05, "loss": 0.3386, "step": 37200 }, { "epoch": 0.87, "learning_rate": 2.363069395226585e-05, "loss": 0.1904, "step": 37205 }, { "epoch": 0.87, "learning_rate": 2.3629813897819814e-05, "loss": 0.0515, "step": 37210 }, { "epoch": 0.87, "learning_rate": 2.362893384337378e-05, "loss": 0.0416, "step": 37215 }, { "epoch": 0.87, "learning_rate": 2.362805378892774e-05, "loss": 0.1611, "step": 37220 }, { "epoch": 0.87, "learning_rate": 2.362717373448171e-05, "loss": 0.1272, "step": 37225 }, { "epoch": 0.87, "learning_rate": 2.3626293680035674e-05, "loss": 0.1019, "step": 37230 }, { "epoch": 0.87, "learning_rate": 2.3625413625589635e-05, "loss": 0.1278, "step": 37235 }, { "epoch": 0.87, "learning_rate": 2.36245335711436e-05, "loss": 0.2334, "step": 37240 }, { "epoch": 0.87, "learning_rate": 2.362365351669757e-05, "loss": 0.4271, "step": 37245 }, { "epoch": 0.87, "learning_rate": 2.3622773462251534e-05, "loss": 0.426, "step": 37250 }, { "epoch": 0.87, "learning_rate": 2.3621893407805496e-05, "loss": 0.1574, "step": 37255 }, { "epoch": 0.87, "learning_rate": 2.362101335335946e-05, "loss": 0.0407, "step": 37260 }, { "epoch": 0.87, "learning_rate": 2.362013329891343e-05, "loss": 0.0805, "step": 37265 }, { "epoch": 0.87, "learning_rate": 2.361925324446739e-05, "loss": 0.182, "step": 37270 }, { "epoch": 0.87, "learning_rate": 2.3618373190021356e-05, "loss": 0.2001, "step": 37275 }, { "epoch": 0.87, "learning_rate": 2.361749313557532e-05, "loss": 0.2317, "step": 37280 }, { "epoch": 0.87, "learning_rate": 2.361661308112929e-05, "loss": 0.1633, "step": 37285 }, { "epoch": 0.87, "learning_rate": 2.361573302668325e-05, "loss": 0.1591, "step": 37290 }, { "epoch": 0.87, "learning_rate": 2.3614852972237216e-05, "loss": 0.2276, "step": 37295 }, { "epoch": 0.87, "learning_rate": 2.361397291779118e-05, "loss": 0.3602, "step": 37300 }, { "epoch": 0.87, "learning_rate": 2.3613092863345146e-05, "loss": 0.1739, "step": 37305 }, { "epoch": 0.87, "learning_rate": 2.361221280889911e-05, "loss": 0.0883, "step": 37310 }, { "epoch": 0.87, "learning_rate": 2.3611332754453076e-05, "loss": 0.0459, "step": 37315 }, { "epoch": 0.87, "learning_rate": 2.361045270000704e-05, "loss": 0.0469, "step": 37320 }, { "epoch": 0.87, "learning_rate": 2.3609572645561006e-05, "loss": 0.0851, "step": 37325 }, { "epoch": 0.87, "learning_rate": 2.360869259111497e-05, "loss": 0.1625, "step": 37330 }, { "epoch": 0.87, "learning_rate": 2.3607812536668936e-05, "loss": 0.4533, "step": 37335 }, { "epoch": 0.87, "learning_rate": 2.36069324822229e-05, "loss": 0.1285, "step": 37340 }, { "epoch": 0.87, "learning_rate": 2.3606052427776866e-05, "loss": 0.258, "step": 37345 }, { "epoch": 0.87, "learning_rate": 2.360517237333083e-05, "loss": 0.4917, "step": 37350 }, { "epoch": 0.87, "learning_rate": 2.3604292318884796e-05, "loss": 0.1898, "step": 37355 }, { "epoch": 0.87, "learning_rate": 2.360341226443876e-05, "loss": 0.0696, "step": 37360 }, { "epoch": 0.87, "learning_rate": 2.3602532209992726e-05, "loss": 0.1519, "step": 37365 }, { "epoch": 0.87, "learning_rate": 2.360165215554669e-05, "loss": 0.2041, "step": 37370 }, { "epoch": 0.87, "learning_rate": 2.3600772101100653e-05, "loss": 0.1754, "step": 37375 }, { "epoch": 0.87, "learning_rate": 2.359989204665462e-05, "loss": 0.2117, "step": 37380 }, { "epoch": 0.87, "learning_rate": 2.3599011992208586e-05, "loss": 0.2713, "step": 37385 }, { "epoch": 0.87, "learning_rate": 2.359813193776255e-05, "loss": 0.246, "step": 37390 }, { "epoch": 0.87, "learning_rate": 2.3597251883316513e-05, "loss": 0.2644, "step": 37395 }, { "epoch": 0.87, "learning_rate": 2.359637182887048e-05, "loss": 0.627, "step": 37400 }, { "epoch": 0.87, "learning_rate": 2.3595491774424446e-05, "loss": 0.1587, "step": 37405 }, { "epoch": 0.87, "learning_rate": 2.3594611719978408e-05, "loss": 0.0439, "step": 37410 }, { "epoch": 0.87, "learning_rate": 2.3593731665532373e-05, "loss": 0.1383, "step": 37415 }, { "epoch": 0.87, "learning_rate": 2.359285161108634e-05, "loss": 0.0455, "step": 37420 }, { "epoch": 0.87, "learning_rate": 2.3591971556640306e-05, "loss": 0.1221, "step": 37425 }, { "epoch": 0.87, "learning_rate": 2.3591091502194268e-05, "loss": 0.307, "step": 37430 }, { "epoch": 0.87, "learning_rate": 2.3590211447748233e-05, "loss": 0.0607, "step": 37435 }, { "epoch": 0.87, "learning_rate": 2.35893313933022e-05, "loss": 0.1548, "step": 37440 }, { "epoch": 0.87, "learning_rate": 2.3588451338856163e-05, "loss": 0.2655, "step": 37445 }, { "epoch": 0.87, "learning_rate": 2.3587571284410128e-05, "loss": 0.6642, "step": 37450 }, { "epoch": 0.87, "learning_rate": 2.3586691229964093e-05, "loss": 0.2001, "step": 37455 }, { "epoch": 0.87, "learning_rate": 2.358581117551806e-05, "loss": 0.1123, "step": 37460 }, { "epoch": 0.87, "learning_rate": 2.3584931121072023e-05, "loss": 0.0404, "step": 37465 }, { "epoch": 0.87, "learning_rate": 2.3584051066625988e-05, "loss": 0.0702, "step": 37470 }, { "epoch": 0.87, "learning_rate": 2.3583171012179956e-05, "loss": 0.0743, "step": 37475 }, { "epoch": 0.87, "learning_rate": 2.3582290957733918e-05, "loss": 0.1453, "step": 37480 }, { "epoch": 0.87, "learning_rate": 2.3581410903287883e-05, "loss": 0.1511, "step": 37485 }, { "epoch": 0.87, "learning_rate": 2.3580530848841848e-05, "loss": 0.2628, "step": 37490 }, { "epoch": 0.87, "learning_rate": 2.3579650794395817e-05, "loss": 0.1551, "step": 37495 }, { "epoch": 0.87, "learning_rate": 2.3578770739949778e-05, "loss": 1.073, "step": 37500 }, { "epoch": 0.88, "learning_rate": 2.3577890685503743e-05, "loss": 0.2794, "step": 37505 }, { "epoch": 0.88, "learning_rate": 2.3577010631057708e-05, "loss": 0.0415, "step": 37510 }, { "epoch": 0.88, "learning_rate": 2.3576130576611677e-05, "loss": 0.1528, "step": 37515 }, { "epoch": 0.88, "learning_rate": 2.3575250522165638e-05, "loss": 0.2031, "step": 37520 }, { "epoch": 0.88, "learning_rate": 2.3574370467719603e-05, "loss": 0.203, "step": 37525 }, { "epoch": 0.88, "learning_rate": 2.357349041327357e-05, "loss": 0.1809, "step": 37530 }, { "epoch": 0.88, "learning_rate": 2.3572610358827533e-05, "loss": 0.2438, "step": 37535 }, { "epoch": 0.88, "learning_rate": 2.35717303043815e-05, "loss": 0.1663, "step": 37540 }, { "epoch": 0.88, "learning_rate": 2.3570850249935463e-05, "loss": 0.2269, "step": 37545 }, { "epoch": 0.88, "learning_rate": 2.356997019548943e-05, "loss": 0.5601, "step": 37550 }, { "epoch": 0.88, "learning_rate": 2.3569090141043393e-05, "loss": 0.2053, "step": 37555 }, { "epoch": 0.88, "learning_rate": 2.356821008659736e-05, "loss": 0.1593, "step": 37560 }, { "epoch": 0.88, "learning_rate": 2.3567330032151323e-05, "loss": 0.1261, "step": 37565 }, { "epoch": 0.88, "learning_rate": 2.3566449977705285e-05, "loss": 0.0924, "step": 37570 }, { "epoch": 0.88, "learning_rate": 2.3565569923259253e-05, "loss": 0.0636, "step": 37575 }, { "epoch": 0.88, "learning_rate": 2.356468986881322e-05, "loss": 0.1896, "step": 37580 }, { "epoch": 0.88, "learning_rate": 2.3563809814367184e-05, "loss": 0.319, "step": 37585 }, { "epoch": 0.88, "learning_rate": 2.3562929759921145e-05, "loss": 0.0834, "step": 37590 }, { "epoch": 0.88, "learning_rate": 2.3562049705475114e-05, "loss": 0.1817, "step": 37595 }, { "epoch": 0.88, "learning_rate": 2.356116965102908e-05, "loss": 0.5434, "step": 37600 }, { "epoch": 0.88, "learning_rate": 2.356028959658304e-05, "loss": 0.2873, "step": 37605 }, { "epoch": 0.88, "learning_rate": 2.355940954213701e-05, "loss": 0.0695, "step": 37610 }, { "epoch": 0.88, "learning_rate": 2.3558529487690974e-05, "loss": 0.0811, "step": 37615 }, { "epoch": 0.88, "learning_rate": 2.355764943324494e-05, "loss": 0.0772, "step": 37620 }, { "epoch": 0.88, "learning_rate": 2.35567693787989e-05, "loss": 0.1502, "step": 37625 }, { "epoch": 0.88, "learning_rate": 2.355588932435287e-05, "loss": 0.1114, "step": 37630 }, { "epoch": 0.88, "learning_rate": 2.3555009269906834e-05, "loss": 0.2441, "step": 37635 }, { "epoch": 0.88, "learning_rate": 2.3554129215460795e-05, "loss": 0.2256, "step": 37640 }, { "epoch": 0.88, "learning_rate": 2.355324916101476e-05, "loss": 0.4683, "step": 37645 }, { "epoch": 0.88, "learning_rate": 2.355236910656873e-05, "loss": 0.5178, "step": 37650 }, { "epoch": 0.88, "learning_rate": 2.3551489052122694e-05, "loss": 0.2363, "step": 37655 }, { "epoch": 0.88, "learning_rate": 2.3550608997676655e-05, "loss": 0.0574, "step": 37660 }, { "epoch": 0.88, "learning_rate": 2.354972894323062e-05, "loss": 0.1084, "step": 37665 }, { "epoch": 0.88, "learning_rate": 2.354884888878459e-05, "loss": 0.1895, "step": 37670 }, { "epoch": 0.88, "learning_rate": 2.354796883433855e-05, "loss": 0.1338, "step": 37675 }, { "epoch": 0.88, "learning_rate": 2.3547088779892516e-05, "loss": 0.1623, "step": 37680 }, { "epoch": 0.88, "learning_rate": 2.354620872544648e-05, "loss": 0.3872, "step": 37685 }, { "epoch": 0.88, "learning_rate": 2.354532867100045e-05, "loss": 0.2334, "step": 37690 }, { "epoch": 0.88, "learning_rate": 2.354444861655441e-05, "loss": 0.1258, "step": 37695 }, { "epoch": 0.88, "learning_rate": 2.3543568562108376e-05, "loss": 0.2133, "step": 37700 }, { "epoch": 0.88, "learning_rate": 2.354268850766234e-05, "loss": 0.2981, "step": 37705 }, { "epoch": 0.88, "learning_rate": 2.3541808453216306e-05, "loss": 0.0787, "step": 37710 }, { "epoch": 0.88, "learning_rate": 2.354092839877027e-05, "loss": 0.0225, "step": 37715 }, { "epoch": 0.88, "learning_rate": 2.3540048344324236e-05, "loss": 0.1103, "step": 37720 }, { "epoch": 0.88, "learning_rate": 2.35391682898782e-05, "loss": 0.0862, "step": 37725 }, { "epoch": 0.88, "learning_rate": 2.3538288235432166e-05, "loss": 0.1681, "step": 37730 }, { "epoch": 0.88, "learning_rate": 2.353740818098613e-05, "loss": 0.1597, "step": 37735 }, { "epoch": 0.88, "learning_rate": 2.3536528126540096e-05, "loss": 0.184, "step": 37740 }, { "epoch": 0.88, "learning_rate": 2.353564807209406e-05, "loss": 0.2273, "step": 37745 }, { "epoch": 0.88, "learning_rate": 2.3534768017648026e-05, "loss": 0.5633, "step": 37750 }, { "epoch": 0.88, "learning_rate": 2.353388796320199e-05, "loss": 0.2008, "step": 37755 }, { "epoch": 0.88, "learning_rate": 2.3533007908755956e-05, "loss": 0.0267, "step": 37760 }, { "epoch": 0.88, "learning_rate": 2.353212785430992e-05, "loss": 0.0996, "step": 37765 }, { "epoch": 0.88, "learning_rate": 2.3531247799863886e-05, "loss": 0.1537, "step": 37770 }, { "epoch": 0.88, "learning_rate": 2.353036774541785e-05, "loss": 0.2034, "step": 37775 }, { "epoch": 0.88, "learning_rate": 2.3529487690971813e-05, "loss": 0.0591, "step": 37780 }, { "epoch": 0.88, "learning_rate": 2.352860763652578e-05, "loss": 0.2685, "step": 37785 }, { "epoch": 0.88, "learning_rate": 2.3527727582079746e-05, "loss": 0.1619, "step": 37790 }, { "epoch": 0.88, "learning_rate": 2.352684752763371e-05, "loss": 0.2001, "step": 37795 }, { "epoch": 0.88, "learning_rate": 2.3525967473187673e-05, "loss": 0.4047, "step": 37800 }, { "epoch": 0.88, "learning_rate": 2.352508741874164e-05, "loss": 0.191, "step": 37805 }, { "epoch": 0.88, "learning_rate": 2.3524207364295606e-05, "loss": 0.1048, "step": 37810 }, { "epoch": 0.88, "learning_rate": 2.352332730984957e-05, "loss": 0.0934, "step": 37815 }, { "epoch": 0.88, "learning_rate": 2.3522447255403533e-05, "loss": 0.1638, "step": 37820 }, { "epoch": 0.88, "learning_rate": 2.35215672009575e-05, "loss": 0.1778, "step": 37825 }, { "epoch": 0.88, "learning_rate": 2.3520687146511466e-05, "loss": 0.1658, "step": 37830 }, { "epoch": 0.88, "learning_rate": 2.3519807092065428e-05, "loss": 0.0522, "step": 37835 }, { "epoch": 0.88, "learning_rate": 2.3518927037619393e-05, "loss": 0.2499, "step": 37840 }, { "epoch": 0.88, "learning_rate": 2.351804698317336e-05, "loss": 0.2168, "step": 37845 }, { "epoch": 0.88, "learning_rate": 2.3517166928727326e-05, "loss": 0.7879, "step": 37850 }, { "epoch": 0.88, "learning_rate": 2.3516286874281288e-05, "loss": 0.2014, "step": 37855 }, { "epoch": 0.88, "learning_rate": 2.3515406819835253e-05, "loss": 0.0816, "step": 37860 }, { "epoch": 0.88, "learning_rate": 2.351452676538922e-05, "loss": 0.1429, "step": 37865 }, { "epoch": 0.88, "learning_rate": 2.3513646710943183e-05, "loss": 0.0238, "step": 37870 }, { "epoch": 0.88, "learning_rate": 2.3512766656497148e-05, "loss": 0.1955, "step": 37875 }, { "epoch": 0.88, "learning_rate": 2.3511886602051116e-05, "loss": 0.2006, "step": 37880 }, { "epoch": 0.88, "learning_rate": 2.351100654760508e-05, "loss": 0.1569, "step": 37885 }, { "epoch": 0.88, "learning_rate": 2.3510126493159043e-05, "loss": 0.1793, "step": 37890 }, { "epoch": 0.88, "learning_rate": 2.3509246438713008e-05, "loss": 0.1672, "step": 37895 }, { "epoch": 0.88, "learning_rate": 2.3508366384266976e-05, "loss": 0.3215, "step": 37900 }, { "epoch": 0.88, "learning_rate": 2.3507486329820938e-05, "loss": 0.2341, "step": 37905 }, { "epoch": 0.88, "learning_rate": 2.3506606275374903e-05, "loss": 0.094, "step": 37910 }, { "epoch": 0.88, "learning_rate": 2.3505726220928868e-05, "loss": 0.0703, "step": 37915 }, { "epoch": 0.88, "learning_rate": 2.3504846166482837e-05, "loss": 0.1298, "step": 37920 }, { "epoch": 0.88, "learning_rate": 2.3503966112036798e-05, "loss": 0.1158, "step": 37925 }, { "epoch": 0.88, "learning_rate": 2.3503086057590763e-05, "loss": 0.1414, "step": 37930 }, { "epoch": 0.89, "learning_rate": 2.3502206003144728e-05, "loss": 0.2048, "step": 37935 }, { "epoch": 0.89, "learning_rate": 2.3501325948698693e-05, "loss": 0.1444, "step": 37940 }, { "epoch": 0.89, "learning_rate": 2.3500445894252658e-05, "loss": 0.1447, "step": 37945 }, { "epoch": 0.89, "learning_rate": 2.3499565839806623e-05, "loss": 0.5356, "step": 37950 }, { "epoch": 0.89, "learning_rate": 2.3498685785360588e-05, "loss": 0.2073, "step": 37955 }, { "epoch": 0.89, "learning_rate": 2.3497805730914553e-05, "loss": 0.0257, "step": 37960 }, { "epoch": 0.89, "learning_rate": 2.349692567646852e-05, "loss": 0.0863, "step": 37965 }, { "epoch": 0.89, "learning_rate": 2.3496045622022483e-05, "loss": 0.0972, "step": 37970 }, { "epoch": 0.89, "learning_rate": 2.3495165567576445e-05, "loss": 0.0874, "step": 37975 }, { "epoch": 0.89, "learning_rate": 2.3494285513130413e-05, "loss": 0.167, "step": 37980 }, { "epoch": 0.89, "learning_rate": 2.349340545868438e-05, "loss": 0.2064, "step": 37985 }, { "epoch": 0.89, "learning_rate": 2.3492525404238343e-05, "loss": 0.3512, "step": 37990 }, { "epoch": 0.89, "learning_rate": 2.3491645349792305e-05, "loss": 0.539, "step": 37995 }, { "epoch": 0.89, "learning_rate": 2.3490765295346273e-05, "loss": 0.5514, "step": 38000 }, { "epoch": 0.89, "learning_rate": 2.348988524090024e-05, "loss": 0.2371, "step": 38005 }, { "epoch": 0.89, "learning_rate": 2.34890051864542e-05, "loss": 0.0579, "step": 38010 }, { "epoch": 0.89, "learning_rate": 2.348812513200817e-05, "loss": 0.0166, "step": 38015 }, { "epoch": 0.89, "learning_rate": 2.3487245077562134e-05, "loss": 0.1415, "step": 38020 }, { "epoch": 0.89, "learning_rate": 2.34863650231161e-05, "loss": 0.188, "step": 38025 }, { "epoch": 0.89, "learning_rate": 2.348548496867006e-05, "loss": 0.1868, "step": 38030 }, { "epoch": 0.89, "learning_rate": 2.348460491422403e-05, "loss": 0.1984, "step": 38035 }, { "epoch": 0.89, "learning_rate": 2.3483724859777994e-05, "loss": 0.144, "step": 38040 }, { "epoch": 0.89, "learning_rate": 2.3482844805331955e-05, "loss": 0.2773, "step": 38045 }, { "epoch": 0.89, "learning_rate": 2.348196475088592e-05, "loss": 0.5115, "step": 38050 }, { "epoch": 0.89, "learning_rate": 2.348108469643989e-05, "loss": 0.1244, "step": 38055 }, { "epoch": 0.89, "learning_rate": 2.3480204641993854e-05, "loss": 0.1076, "step": 38060 }, { "epoch": 0.89, "learning_rate": 2.3479324587547815e-05, "loss": 0.1926, "step": 38065 }, { "epoch": 0.89, "learning_rate": 2.347844453310178e-05, "loss": 0.1063, "step": 38070 }, { "epoch": 0.89, "learning_rate": 2.3477740489544955e-05, "loss": 0.1441, "step": 38075 }, { "epoch": 0.89, "learning_rate": 2.347686043509892e-05, "loss": 0.1889, "step": 38080 }, { "epoch": 0.89, "learning_rate": 2.3475980380652882e-05, "loss": 0.1479, "step": 38085 }, { "epoch": 0.89, "learning_rate": 2.347510032620685e-05, "loss": 0.139, "step": 38090 }, { "epoch": 0.89, "learning_rate": 2.3474220271760815e-05, "loss": 0.2082, "step": 38095 }, { "epoch": 0.89, "learning_rate": 2.3473340217314777e-05, "loss": 0.6372, "step": 38100 }, { "epoch": 0.89, "learning_rate": 2.3472460162868742e-05, "loss": 0.1494, "step": 38105 }, { "epoch": 0.89, "learning_rate": 2.347158010842271e-05, "loss": 0.0391, "step": 38110 }, { "epoch": 0.89, "learning_rate": 2.3470700053976675e-05, "loss": 0.0833, "step": 38115 }, { "epoch": 0.89, "learning_rate": 2.3469819999530637e-05, "loss": 0.1161, "step": 38120 }, { "epoch": 0.89, "learning_rate": 2.3468939945084602e-05, "loss": 0.194, "step": 38125 }, { "epoch": 0.89, "learning_rate": 2.346805989063857e-05, "loss": 0.1595, "step": 38130 }, { "epoch": 0.89, "learning_rate": 2.3467179836192532e-05, "loss": 0.2167, "step": 38135 }, { "epoch": 0.89, "learning_rate": 2.3466299781746497e-05, "loss": 0.2143, "step": 38140 }, { "epoch": 0.89, "learning_rate": 2.3465419727300462e-05, "loss": 0.3146, "step": 38145 }, { "epoch": 0.89, "learning_rate": 2.346453967285443e-05, "loss": 0.5711, "step": 38150 }, { "epoch": 0.89, "learning_rate": 2.3463659618408392e-05, "loss": 0.2195, "step": 38155 }, { "epoch": 0.89, "learning_rate": 2.3462779563962357e-05, "loss": 0.061, "step": 38160 }, { "epoch": 0.89, "learning_rate": 2.3461899509516322e-05, "loss": 0.0998, "step": 38165 }, { "epoch": 0.89, "learning_rate": 2.3461019455070287e-05, "loss": 0.0558, "step": 38170 }, { "epoch": 0.89, "learning_rate": 2.3460139400624252e-05, "loss": 0.1189, "step": 38175 }, { "epoch": 0.89, "learning_rate": 2.3459259346178217e-05, "loss": 0.1908, "step": 38180 }, { "epoch": 0.89, "learning_rate": 2.3458379291732182e-05, "loss": 0.2046, "step": 38185 }, { "epoch": 0.89, "learning_rate": 2.3457499237286147e-05, "loss": 0.2224, "step": 38190 }, { "epoch": 0.89, "learning_rate": 2.3456619182840112e-05, "loss": 0.1654, "step": 38195 }, { "epoch": 0.89, "learning_rate": 2.3455739128394077e-05, "loss": 0.3239, "step": 38200 }, { "epoch": 0.89, "learning_rate": 2.3454859073948042e-05, "loss": 0.2274, "step": 38205 }, { "epoch": 0.89, "learning_rate": 2.3453979019502007e-05, "loss": 0.1582, "step": 38210 }, { "epoch": 0.89, "learning_rate": 2.3453098965055972e-05, "loss": 0.1022, "step": 38215 }, { "epoch": 0.89, "learning_rate": 2.3452218910609937e-05, "loss": 0.0678, "step": 38220 }, { "epoch": 0.89, "learning_rate": 2.3451338856163902e-05, "loss": 0.113, "step": 38225 }, { "epoch": 0.89, "learning_rate": 2.3450458801717867e-05, "loss": 0.2022, "step": 38230 }, { "epoch": 0.89, "learning_rate": 2.3449578747271832e-05, "loss": 0.1501, "step": 38235 }, { "epoch": 0.89, "learning_rate": 2.3448698692825794e-05, "loss": 0.268, "step": 38240 }, { "epoch": 0.89, "learning_rate": 2.3447818638379762e-05, "loss": 0.2155, "step": 38245 }, { "epoch": 0.89, "learning_rate": 2.3446938583933727e-05, "loss": 0.3574, "step": 38250 }, { "epoch": 0.89, "learning_rate": 2.3446058529487692e-05, "loss": 0.2182, "step": 38255 }, { "epoch": 0.89, "learning_rate": 2.3445178475041654e-05, "loss": 0.0662, "step": 38260 }, { "epoch": 0.89, "learning_rate": 2.3444298420595622e-05, "loss": 0.1032, "step": 38265 }, { "epoch": 0.89, "learning_rate": 2.3443418366149588e-05, "loss": 0.0832, "step": 38270 }, { "epoch": 0.89, "learning_rate": 2.344253831170355e-05, "loss": 0.1579, "step": 38275 }, { "epoch": 0.89, "learning_rate": 2.3441658257257514e-05, "loss": 0.0995, "step": 38280 }, { "epoch": 0.89, "learning_rate": 2.3440778202811483e-05, "loss": 0.1719, "step": 38285 }, { "epoch": 0.89, "learning_rate": 2.3439898148365448e-05, "loss": 0.2563, "step": 38290 }, { "epoch": 0.89, "learning_rate": 2.343901809391941e-05, "loss": 0.37, "step": 38295 }, { "epoch": 0.89, "learning_rate": 2.3438138039473374e-05, "loss": 0.4708, "step": 38300 }, { "epoch": 0.89, "learning_rate": 2.3437257985027343e-05, "loss": 0.3159, "step": 38305 }, { "epoch": 0.89, "learning_rate": 2.3436377930581308e-05, "loss": 0.0779, "step": 38310 }, { "epoch": 0.89, "learning_rate": 2.343549787613527e-05, "loss": 0.1522, "step": 38315 }, { "epoch": 0.89, "learning_rate": 2.3434617821689234e-05, "loss": 0.066, "step": 38320 }, { "epoch": 0.89, "learning_rate": 2.3433737767243203e-05, "loss": 0.1335, "step": 38325 }, { "epoch": 0.89, "learning_rate": 2.3432857712797164e-05, "loss": 0.3237, "step": 38330 }, { "epoch": 0.89, "learning_rate": 2.343197765835113e-05, "loss": 0.3654, "step": 38335 }, { "epoch": 0.89, "learning_rate": 2.3431097603905098e-05, "loss": 0.2115, "step": 38340 }, { "epoch": 0.89, "learning_rate": 2.3430217549459063e-05, "loss": 0.3739, "step": 38345 }, { "epoch": 0.89, "learning_rate": 2.3429337495013024e-05, "loss": 0.3036, "step": 38350 }, { "epoch": 0.89, "learning_rate": 2.342845744056699e-05, "loss": 0.2227, "step": 38355 }, { "epoch": 0.89, "learning_rate": 2.3427577386120958e-05, "loss": 0.0913, "step": 38360 }, { "epoch": 0.9, "learning_rate": 2.342669733167492e-05, "loss": 0.1494, "step": 38365 }, { "epoch": 0.9, "learning_rate": 2.3425817277228885e-05, "loss": 0.1227, "step": 38370 }, { "epoch": 0.9, "learning_rate": 2.342493722278285e-05, "loss": 0.0877, "step": 38375 }, { "epoch": 0.9, "learning_rate": 2.3424057168336818e-05, "loss": 0.1554, "step": 38380 }, { "epoch": 0.9, "learning_rate": 2.342317711389078e-05, "loss": 0.2149, "step": 38385 }, { "epoch": 0.9, "learning_rate": 2.3422297059444745e-05, "loss": 0.2637, "step": 38390 }, { "epoch": 0.9, "learning_rate": 2.342141700499871e-05, "loss": 0.2592, "step": 38395 }, { "epoch": 0.9, "learning_rate": 2.3420536950552675e-05, "loss": 0.5779, "step": 38400 }, { "epoch": 0.9, "learning_rate": 2.341965689610664e-05, "loss": 0.2097, "step": 38405 }, { "epoch": 0.9, "learning_rate": 2.3418776841660605e-05, "loss": 0.1216, "step": 38410 }, { "epoch": 0.9, "learning_rate": 2.341789678721457e-05, "loss": 0.0893, "step": 38415 }, { "epoch": 0.9, "learning_rate": 2.3417016732768535e-05, "loss": 0.0419, "step": 38420 }, { "epoch": 0.9, "learning_rate": 2.34161366783225e-05, "loss": 0.0294, "step": 38425 }, { "epoch": 0.9, "learning_rate": 2.3415256623876465e-05, "loss": 0.1701, "step": 38430 }, { "epoch": 0.9, "learning_rate": 2.3414376569430426e-05, "loss": 0.2378, "step": 38435 }, { "epoch": 0.9, "learning_rate": 2.3413496514984395e-05, "loss": 0.2634, "step": 38440 }, { "epoch": 0.9, "learning_rate": 2.341261646053836e-05, "loss": 0.3311, "step": 38445 }, { "epoch": 0.9, "learning_rate": 2.3411736406092325e-05, "loss": 0.5056, "step": 38450 }, { "epoch": 0.9, "learning_rate": 2.3410856351646286e-05, "loss": 0.193, "step": 38455 }, { "epoch": 0.9, "learning_rate": 2.3409976297200255e-05, "loss": 0.054, "step": 38460 }, { "epoch": 0.9, "learning_rate": 2.340909624275422e-05, "loss": 0.0576, "step": 38465 }, { "epoch": 0.9, "learning_rate": 2.340821618830818e-05, "loss": 0.0591, "step": 38470 }, { "epoch": 0.9, "learning_rate": 2.340733613386215e-05, "loss": 0.1264, "step": 38475 }, { "epoch": 0.9, "learning_rate": 2.3406456079416115e-05, "loss": 0.1825, "step": 38480 }, { "epoch": 0.9, "learning_rate": 2.340557602497008e-05, "loss": 0.2674, "step": 38485 }, { "epoch": 0.9, "learning_rate": 2.340469597052404e-05, "loss": 0.1797, "step": 38490 }, { "epoch": 0.9, "learning_rate": 2.340381591607801e-05, "loss": 0.1904, "step": 38495 }, { "epoch": 0.9, "learning_rate": 2.3402935861631975e-05, "loss": 0.5302, "step": 38500 }, { "epoch": 0.9, "learning_rate": 2.3402055807185937e-05, "loss": 0.1672, "step": 38505 }, { "epoch": 0.9, "learning_rate": 2.3401175752739902e-05, "loss": 0.0193, "step": 38510 }, { "epoch": 0.9, "learning_rate": 2.340029569829387e-05, "loss": 0.1253, "step": 38515 }, { "epoch": 0.9, "learning_rate": 2.3399415643847835e-05, "loss": 0.2106, "step": 38520 }, { "epoch": 0.9, "learning_rate": 2.3398535589401797e-05, "loss": 0.1287, "step": 38525 }, { "epoch": 0.9, "learning_rate": 2.3397655534955762e-05, "loss": 0.0434, "step": 38530 }, { "epoch": 0.9, "learning_rate": 2.339677548050973e-05, "loss": 0.1524, "step": 38535 }, { "epoch": 0.9, "learning_rate": 2.3395895426063692e-05, "loss": 0.216, "step": 38540 }, { "epoch": 0.9, "learning_rate": 2.3395015371617657e-05, "loss": 0.2287, "step": 38545 }, { "epoch": 0.9, "learning_rate": 2.3394135317171622e-05, "loss": 0.2916, "step": 38550 }, { "epoch": 0.9, "learning_rate": 2.339325526272559e-05, "loss": 0.1965, "step": 38555 }, { "epoch": 0.9, "learning_rate": 2.3392375208279552e-05, "loss": 0.1042, "step": 38560 }, { "epoch": 0.9, "learning_rate": 2.3391495153833517e-05, "loss": 0.0399, "step": 38565 }, { "epoch": 0.9, "learning_rate": 2.3390615099387482e-05, "loss": 0.0434, "step": 38570 }, { "epoch": 0.9, "learning_rate": 2.3389735044941447e-05, "loss": 0.1209, "step": 38575 }, { "epoch": 0.9, "learning_rate": 2.3388854990495412e-05, "loss": 0.1987, "step": 38580 }, { "epoch": 0.9, "learning_rate": 2.3387974936049377e-05, "loss": 0.2371, "step": 38585 }, { "epoch": 0.9, "learning_rate": 2.3387094881603342e-05, "loss": 0.2013, "step": 38590 }, { "epoch": 0.9, "learning_rate": 2.3386214827157307e-05, "loss": 0.3529, "step": 38595 }, { "epoch": 0.9, "learning_rate": 2.3385334772711272e-05, "loss": 0.4564, "step": 38600 }, { "epoch": 0.9, "learning_rate": 2.3384454718265237e-05, "loss": 0.2421, "step": 38605 }, { "epoch": 0.9, "learning_rate": 2.3383574663819206e-05, "loss": 0.0297, "step": 38610 }, { "epoch": 0.9, "learning_rate": 2.3382694609373167e-05, "loss": 0.0894, "step": 38615 }, { "epoch": 0.9, "learning_rate": 2.3381814554927132e-05, "loss": 0.2109, "step": 38620 }, { "epoch": 0.9, "learning_rate": 2.3380934500481097e-05, "loss": 0.1442, "step": 38625 }, { "epoch": 0.9, "learning_rate": 2.3380054446035062e-05, "loss": 0.1762, "step": 38630 }, { "epoch": 0.9, "learning_rate": 2.3379174391589027e-05, "loss": 0.2488, "step": 38635 }, { "epoch": 0.9, "learning_rate": 2.3378294337142992e-05, "loss": 0.265, "step": 38640 }, { "epoch": 0.9, "learning_rate": 2.3377414282696957e-05, "loss": 0.4236, "step": 38645 }, { "epoch": 0.9, "learning_rate": 2.3376534228250922e-05, "loss": 0.7029, "step": 38650 }, { "epoch": 0.9, "learning_rate": 2.3375654173804887e-05, "loss": 0.1732, "step": 38655 }, { "epoch": 0.9, "learning_rate": 2.337495013024806e-05, "loss": 0.0177, "step": 38660 }, { "epoch": 0.9, "learning_rate": 2.3374070075802024e-05, "loss": 0.0965, "step": 38665 }, { "epoch": 0.9, "learning_rate": 2.337319002135599e-05, "loss": 0.0758, "step": 38670 }, { "epoch": 0.9, "learning_rate": 2.3372309966909954e-05, "loss": 0.155, "step": 38675 }, { "epoch": 0.9, "learning_rate": 2.337142991246392e-05, "loss": 0.1421, "step": 38680 }, { "epoch": 0.9, "learning_rate": 2.3370549858017884e-05, "loss": 0.0549, "step": 38685 }, { "epoch": 0.9, "learning_rate": 2.336966980357185e-05, "loss": 0.2011, "step": 38690 }, { "epoch": 0.9, "learning_rate": 2.3368789749125814e-05, "loss": 0.3957, "step": 38695 }, { "epoch": 0.9, "learning_rate": 2.3367909694679775e-05, "loss": 0.5787, "step": 38700 }, { "epoch": 0.9, "learning_rate": 2.3367029640233744e-05, "loss": 0.1521, "step": 38705 }, { "epoch": 0.9, "learning_rate": 2.336614958578771e-05, "loss": 0.0646, "step": 38710 }, { "epoch": 0.9, "learning_rate": 2.3365269531341674e-05, "loss": 0.0525, "step": 38715 }, { "epoch": 0.9, "learning_rate": 2.3364389476895636e-05, "loss": 0.1018, "step": 38720 }, { "epoch": 0.9, "learning_rate": 2.3363509422449604e-05, "loss": 0.2167, "step": 38725 }, { "epoch": 0.9, "learning_rate": 2.336262936800357e-05, "loss": 0.1754, "step": 38730 }, { "epoch": 0.9, "learning_rate": 2.336174931355753e-05, "loss": 0.0865, "step": 38735 }, { "epoch": 0.9, "learning_rate": 2.3360869259111496e-05, "loss": 0.2637, "step": 38740 }, { "epoch": 0.9, "learning_rate": 2.3359989204665464e-05, "loss": 0.1453, "step": 38745 }, { "epoch": 0.9, "learning_rate": 2.335910915021943e-05, "loss": 0.6337, "step": 38750 }, { "epoch": 0.9, "learning_rate": 2.335822909577339e-05, "loss": 0.1513, "step": 38755 }, { "epoch": 0.9, "learning_rate": 2.3357349041327356e-05, "loss": 0.1034, "step": 38760 }, { "epoch": 0.9, "learning_rate": 2.3356468986881324e-05, "loss": 0.048, "step": 38765 }, { "epoch": 0.9, "learning_rate": 2.3355588932435286e-05, "loss": 0.1483, "step": 38770 }, { "epoch": 0.9, "learning_rate": 2.335470887798925e-05, "loss": 0.1544, "step": 38775 }, { "epoch": 0.9, "learning_rate": 2.3353828823543216e-05, "loss": 0.204, "step": 38780 }, { "epoch": 0.9, "learning_rate": 2.3352948769097184e-05, "loss": 0.0976, "step": 38785 }, { "epoch": 0.91, "learning_rate": 2.3352068714651146e-05, "loss": 0.1714, "step": 38790 }, { "epoch": 0.91, "learning_rate": 2.335118866020511e-05, "loss": 0.3236, "step": 38795 }, { "epoch": 0.91, "learning_rate": 2.335030860575908e-05, "loss": 0.5066, "step": 38800 }, { "epoch": 0.91, "learning_rate": 2.3349428551313044e-05, "loss": 0.1923, "step": 38805 }, { "epoch": 0.91, "learning_rate": 2.3348548496867006e-05, "loss": 0.1136, "step": 38810 }, { "epoch": 0.91, "learning_rate": 2.334766844242097e-05, "loss": 0.1379, "step": 38815 }, { "epoch": 0.91, "learning_rate": 2.334678838797494e-05, "loss": 0.0842, "step": 38820 }, { "epoch": 0.91, "learning_rate": 2.33459083335289e-05, "loss": 0.1517, "step": 38825 }, { "epoch": 0.91, "learning_rate": 2.3345028279082866e-05, "loss": 0.2071, "step": 38830 }, { "epoch": 0.91, "learning_rate": 2.334414822463683e-05, "loss": 0.2088, "step": 38835 }, { "epoch": 0.91, "learning_rate": 2.33432681701908e-05, "loss": 0.1565, "step": 38840 }, { "epoch": 0.91, "learning_rate": 2.334238811574476e-05, "loss": 0.1674, "step": 38845 }, { "epoch": 0.91, "learning_rate": 2.3341508061298726e-05, "loss": 0.4465, "step": 38850 }, { "epoch": 0.91, "learning_rate": 2.334062800685269e-05, "loss": 0.1571, "step": 38855 }, { "epoch": 0.91, "learning_rate": 2.3339747952406656e-05, "loss": 0.0894, "step": 38860 }, { "epoch": 0.91, "learning_rate": 2.333886789796062e-05, "loss": 0.0453, "step": 38865 }, { "epoch": 0.91, "learning_rate": 2.3337987843514586e-05, "loss": 0.0516, "step": 38870 }, { "epoch": 0.91, "learning_rate": 2.333710778906855e-05, "loss": 0.1235, "step": 38875 }, { "epoch": 0.91, "learning_rate": 2.3336227734622516e-05, "loss": 0.1582, "step": 38880 }, { "epoch": 0.91, "learning_rate": 2.333534768017648e-05, "loss": 0.2932, "step": 38885 }, { "epoch": 0.91, "learning_rate": 2.3334467625730446e-05, "loss": 0.2145, "step": 38890 }, { "epoch": 0.91, "learning_rate": 2.3333587571284408e-05, "loss": 0.3717, "step": 38895 }, { "epoch": 0.91, "learning_rate": 2.3332707516838376e-05, "loss": 0.23, "step": 38900 }, { "epoch": 0.91, "learning_rate": 2.333182746239234e-05, "loss": 0.1991, "step": 38905 }, { "epoch": 0.91, "learning_rate": 2.3330947407946306e-05, "loss": 0.0453, "step": 38910 }, { "epoch": 0.91, "learning_rate": 2.3330067353500268e-05, "loss": 0.2314, "step": 38915 }, { "epoch": 0.91, "learning_rate": 2.3329187299054236e-05, "loss": 0.1553, "step": 38920 }, { "epoch": 0.91, "learning_rate": 2.33283072446082e-05, "loss": 0.1861, "step": 38925 }, { "epoch": 0.91, "learning_rate": 2.3327427190162163e-05, "loss": 0.1369, "step": 38930 }, { "epoch": 0.91, "learning_rate": 2.332654713571613e-05, "loss": 0.16, "step": 38935 }, { "epoch": 0.91, "learning_rate": 2.3325667081270096e-05, "loss": 0.2437, "step": 38940 }, { "epoch": 0.91, "learning_rate": 2.332478702682406e-05, "loss": 0.6751, "step": 38945 }, { "epoch": 0.91, "learning_rate": 2.3323906972378023e-05, "loss": 0.4941, "step": 38950 }, { "epoch": 0.91, "learning_rate": 2.332302691793199e-05, "loss": 0.2346, "step": 38955 }, { "epoch": 0.91, "learning_rate": 2.3322146863485956e-05, "loss": 0.0739, "step": 38960 }, { "epoch": 0.91, "learning_rate": 2.3321266809039918e-05, "loss": 0.0898, "step": 38965 }, { "epoch": 0.91, "learning_rate": 2.3320386754593883e-05, "loss": 0.0905, "step": 38970 }, { "epoch": 0.91, "learning_rate": 2.331950670014785e-05, "loss": 0.1979, "step": 38975 }, { "epoch": 0.91, "learning_rate": 2.3318626645701817e-05, "loss": 0.14, "step": 38980 }, { "epoch": 0.91, "learning_rate": 2.3317746591255778e-05, "loss": 0.2325, "step": 38985 }, { "epoch": 0.91, "learning_rate": 2.3316866536809743e-05, "loss": 0.1814, "step": 38990 }, { "epoch": 0.91, "learning_rate": 2.331598648236371e-05, "loss": 0.2618, "step": 38995 }, { "epoch": 0.91, "learning_rate": 2.3315106427917673e-05, "loss": 0.267, "step": 39000 }, { "epoch": 0.91, "learning_rate": 2.3314226373471638e-05, "loss": 0.232, "step": 39005 }, { "epoch": 0.91, "learning_rate": 2.3313346319025603e-05, "loss": 0.1598, "step": 39010 }, { "epoch": 0.91, "learning_rate": 2.3312466264579572e-05, "loss": 0.1615, "step": 39015 }, { "epoch": 0.91, "learning_rate": 2.3311586210133533e-05, "loss": 0.1252, "step": 39020 }, { "epoch": 0.91, "learning_rate": 2.33107061556875e-05, "loss": 0.03, "step": 39025 }, { "epoch": 0.91, "learning_rate": 2.3309826101241463e-05, "loss": 0.4033, "step": 39030 }, { "epoch": 0.91, "learning_rate": 2.330894604679543e-05, "loss": 0.0945, "step": 39035 }, { "epoch": 0.91, "learning_rate": 2.3308065992349393e-05, "loss": 0.3881, "step": 39040 }, { "epoch": 0.91, "learning_rate": 2.330718593790336e-05, "loss": 0.2495, "step": 39045 }, { "epoch": 0.91, "learning_rate": 2.3306305883457323e-05, "loss": 0.5278, "step": 39050 }, { "epoch": 0.91, "learning_rate": 2.330542582901129e-05, "loss": 0.2994, "step": 39055 }, { "epoch": 0.91, "learning_rate": 2.3304545774565254e-05, "loss": 0.0637, "step": 39060 }, { "epoch": 0.91, "learning_rate": 2.330366572011922e-05, "loss": 0.1782, "step": 39065 }, { "epoch": 0.91, "learning_rate": 2.3302785665673184e-05, "loss": 0.1765, "step": 39070 }, { "epoch": 0.91, "learning_rate": 2.330190561122715e-05, "loss": 0.1016, "step": 39075 }, { "epoch": 0.91, "learning_rate": 2.3301025556781114e-05, "loss": 0.2395, "step": 39080 }, { "epoch": 0.91, "learning_rate": 2.330014550233508e-05, "loss": 0.2596, "step": 39085 }, { "epoch": 0.91, "learning_rate": 2.3299265447889044e-05, "loss": 0.1279, "step": 39090 }, { "epoch": 0.91, "learning_rate": 2.329838539344301e-05, "loss": 0.1191, "step": 39095 }, { "epoch": 0.91, "learning_rate": 2.3297505338996974e-05, "loss": 0.4824, "step": 39100 }, { "epoch": 0.91, "learning_rate": 2.329662528455094e-05, "loss": 0.3133, "step": 39105 }, { "epoch": 0.91, "learning_rate": 2.3295745230104904e-05, "loss": 0.0556, "step": 39110 }, { "epoch": 0.91, "learning_rate": 2.329486517565887e-05, "loss": 0.1578, "step": 39115 }, { "epoch": 0.91, "learning_rate": 2.3293985121212834e-05, "loss": 0.1082, "step": 39120 }, { "epoch": 0.91, "learning_rate": 2.3293105066766795e-05, "loss": 0.069, "step": 39125 }, { "epoch": 0.91, "learning_rate": 2.3292225012320764e-05, "loss": 0.1996, "step": 39130 }, { "epoch": 0.91, "learning_rate": 2.329134495787473e-05, "loss": 0.0036, "step": 39135 }, { "epoch": 0.91, "learning_rate": 2.3290464903428694e-05, "loss": 0.248, "step": 39140 }, { "epoch": 0.91, "learning_rate": 2.3289584848982655e-05, "loss": 0.4248, "step": 39145 }, { "epoch": 0.91, "learning_rate": 2.3288704794536624e-05, "loss": 0.3317, "step": 39150 }, { "epoch": 0.91, "learning_rate": 2.328782474009059e-05, "loss": 0.2061, "step": 39155 }, { "epoch": 0.91, "learning_rate": 2.328694468564455e-05, "loss": 0.0674, "step": 39160 }, { "epoch": 0.91, "learning_rate": 2.3286064631198516e-05, "loss": 0.0711, "step": 39165 }, { "epoch": 0.91, "learning_rate": 2.3285184576752484e-05, "loss": 0.1069, "step": 39170 }, { "epoch": 0.91, "learning_rate": 2.328430452230645e-05, "loss": 0.1764, "step": 39175 }, { "epoch": 0.91, "learning_rate": 2.328342446786041e-05, "loss": 0.1454, "step": 39180 }, { "epoch": 0.91, "learning_rate": 2.3282544413414376e-05, "loss": 0.1902, "step": 39185 }, { "epoch": 0.91, "learning_rate": 2.3281664358968344e-05, "loss": 0.3428, "step": 39190 }, { "epoch": 0.91, "learning_rate": 2.3280784304522306e-05, "loss": 0.3807, "step": 39195 }, { "epoch": 0.91, "learning_rate": 2.327990425007627e-05, "loss": 0.5761, "step": 39200 }, { "epoch": 0.91, "learning_rate": 2.327902419563024e-05, "loss": 0.2129, "step": 39205 }, { "epoch": 0.91, "learning_rate": 2.3278144141184204e-05, "loss": 0.0492, "step": 39210 }, { "epoch": 0.91, "learning_rate": 2.3277264086738166e-05, "loss": 0.1413, "step": 39215 }, { "epoch": 0.92, "learning_rate": 2.327638403229213e-05, "loss": 0.211, "step": 39220 }, { "epoch": 0.92, "learning_rate": 2.32755039778461e-05, "loss": 0.2163, "step": 39225 }, { "epoch": 0.92, "learning_rate": 2.327462392340006e-05, "loss": 0.128, "step": 39230 }, { "epoch": 0.92, "learning_rate": 2.3273743868954026e-05, "loss": 0.2477, "step": 39235 }, { "epoch": 0.92, "learning_rate": 2.327286381450799e-05, "loss": 0.3264, "step": 39240 }, { "epoch": 0.92, "learning_rate": 2.327198376006196e-05, "loss": 0.3545, "step": 39245 }, { "epoch": 0.92, "learning_rate": 2.327110370561592e-05, "loss": 0.532, "step": 39250 }, { "epoch": 0.92, "learning_rate": 2.3270223651169886e-05, "loss": 0.3303, "step": 39255 }, { "epoch": 0.92, "learning_rate": 2.326934359672385e-05, "loss": 0.0908, "step": 39260 }, { "epoch": 0.92, "learning_rate": 2.3268463542277816e-05, "loss": 0.1565, "step": 39265 }, { "epoch": 0.92, "learning_rate": 2.326758348783178e-05, "loss": 0.1971, "step": 39270 }, { "epoch": 0.92, "learning_rate": 2.3266703433385746e-05, "loss": 0.0619, "step": 39275 }, { "epoch": 0.92, "learning_rate": 2.326582337893971e-05, "loss": 0.1522, "step": 39280 }, { "epoch": 0.92, "learning_rate": 2.3264943324493676e-05, "loss": 0.4507, "step": 39285 }, { "epoch": 0.92, "learning_rate": 2.326406327004764e-05, "loss": 0.2676, "step": 39290 }, { "epoch": 0.92, "learning_rate": 2.3263183215601606e-05, "loss": 0.1427, "step": 39295 }, { "epoch": 0.92, "learning_rate": 2.3262303161155568e-05, "loss": 0.4385, "step": 39300 }, { "epoch": 0.92, "learning_rate": 2.3261423106709536e-05, "loss": 0.328, "step": 39305 }, { "epoch": 0.92, "learning_rate": 2.32605430522635e-05, "loss": 0.0357, "step": 39310 }, { "epoch": 0.92, "learning_rate": 2.3259662997817466e-05, "loss": 0.1139, "step": 39315 }, { "epoch": 0.92, "learning_rate": 2.3258782943371428e-05, "loss": 0.0856, "step": 39320 }, { "epoch": 0.92, "learning_rate": 2.3257902888925396e-05, "loss": 0.133, "step": 39325 }, { "epoch": 0.92, "learning_rate": 2.325702283447936e-05, "loss": 0.1925, "step": 39330 }, { "epoch": 0.92, "learning_rate": 2.3256142780033323e-05, "loss": 0.2914, "step": 39335 }, { "epoch": 0.92, "learning_rate": 2.325526272558729e-05, "loss": 0.218, "step": 39340 }, { "epoch": 0.92, "learning_rate": 2.3254382671141256e-05, "loss": 0.2502, "step": 39345 }, { "epoch": 0.92, "learning_rate": 2.325350261669522e-05, "loss": 0.543, "step": 39350 }, { "epoch": 0.92, "learning_rate": 2.3252622562249183e-05, "loss": 0.1048, "step": 39355 }, { "epoch": 0.92, "learning_rate": 2.325174250780315e-05, "loss": 0.053, "step": 39360 }, { "epoch": 0.92, "learning_rate": 2.3250862453357116e-05, "loss": 0.0763, "step": 39365 }, { "epoch": 0.92, "learning_rate": 2.3249982398911078e-05, "loss": 0.124, "step": 39370 }, { "epoch": 0.92, "learning_rate": 2.3249102344465043e-05, "loss": 0.0919, "step": 39375 }, { "epoch": 0.92, "learning_rate": 2.324822229001901e-05, "loss": 0.0891, "step": 39380 }, { "epoch": 0.92, "learning_rate": 2.3247342235572976e-05, "loss": 0.2055, "step": 39385 }, { "epoch": 0.92, "learning_rate": 2.3246462181126938e-05, "loss": 0.3382, "step": 39390 }, { "epoch": 0.92, "learning_rate": 2.3245582126680903e-05, "loss": 0.3974, "step": 39395 }, { "epoch": 0.92, "learning_rate": 2.324470207223487e-05, "loss": 0.5113, "step": 39400 }, { "epoch": 0.92, "learning_rate": 2.3243822017788837e-05, "loss": 0.1963, "step": 39405 }, { "epoch": 0.92, "learning_rate": 2.3242941963342798e-05, "loss": 0.0551, "step": 39410 }, { "epoch": 0.92, "learning_rate": 2.3242061908896763e-05, "loss": 0.0744, "step": 39415 }, { "epoch": 0.92, "learning_rate": 2.324118185445073e-05, "loss": 0.1877, "step": 39420 }, { "epoch": 0.92, "learning_rate": 2.3240301800004693e-05, "loss": 0.2781, "step": 39425 }, { "epoch": 0.92, "learning_rate": 2.3239421745558658e-05, "loss": 0.1576, "step": 39430 }, { "epoch": 0.92, "learning_rate": 2.3238541691112623e-05, "loss": 0.201, "step": 39435 }, { "epoch": 0.92, "learning_rate": 2.323766163666659e-05, "loss": 0.2216, "step": 39440 }, { "epoch": 0.92, "learning_rate": 2.3236781582220553e-05, "loss": 0.2903, "step": 39445 }, { "epoch": 0.92, "learning_rate": 2.323590152777452e-05, "loss": 0.4529, "step": 39450 }, { "epoch": 0.92, "learning_rate": 2.3235021473328483e-05, "loss": 0.2585, "step": 39455 }, { "epoch": 0.92, "learning_rate": 2.323414141888245e-05, "loss": 0.0605, "step": 39460 }, { "epoch": 0.92, "learning_rate": 2.3233261364436413e-05, "loss": 0.0697, "step": 39465 }, { "epoch": 0.92, "learning_rate": 2.323238130999038e-05, "loss": 0.1245, "step": 39470 }, { "epoch": 0.92, "learning_rate": 2.3231501255544343e-05, "loss": 0.1082, "step": 39475 }, { "epoch": 0.92, "learning_rate": 2.323062120109831e-05, "loss": 0.1442, "step": 39480 }, { "epoch": 0.92, "learning_rate": 2.3229741146652273e-05, "loss": 0.0973, "step": 39485 }, { "epoch": 0.92, "learning_rate": 2.322886109220624e-05, "loss": 0.1378, "step": 39490 }, { "epoch": 0.92, "learning_rate": 2.3227981037760204e-05, "loss": 0.195, "step": 39495 }, { "epoch": 0.92, "learning_rate": 2.322710098331417e-05, "loss": 0.2272, "step": 39500 }, { "epoch": 0.92, "learning_rate": 2.3226220928868134e-05, "loss": 0.1795, "step": 39505 }, { "epoch": 0.92, "learning_rate": 2.32253408744221e-05, "loss": 0.0722, "step": 39510 }, { "epoch": 0.92, "learning_rate": 2.3224460819976064e-05, "loss": 0.0642, "step": 39515 }, { "epoch": 0.92, "learning_rate": 2.322358076553003e-05, "loss": 0.1397, "step": 39520 }, { "epoch": 0.92, "learning_rate": 2.3222700711083994e-05, "loss": 0.0921, "step": 39525 }, { "epoch": 0.92, "learning_rate": 2.3221820656637955e-05, "loss": 0.192, "step": 39530 }, { "epoch": 0.92, "learning_rate": 2.3220940602191924e-05, "loss": 0.1255, "step": 39535 }, { "epoch": 0.92, "learning_rate": 2.322006054774589e-05, "loss": 0.2761, "step": 39540 }, { "epoch": 0.92, "learning_rate": 2.3219180493299854e-05, "loss": 0.4143, "step": 39545 }, { "epoch": 0.92, "learning_rate": 2.3218300438853815e-05, "loss": 0.31, "step": 39550 }, { "epoch": 0.92, "learning_rate": 2.3217420384407784e-05, "loss": 0.1555, "step": 39555 }, { "epoch": 0.92, "learning_rate": 2.321654032996175e-05, "loss": 0.0646, "step": 39560 }, { "epoch": 0.92, "learning_rate": 2.321566027551571e-05, "loss": 0.098, "step": 39565 }, { "epoch": 0.92, "learning_rate": 2.3214780221069675e-05, "loss": 0.1961, "step": 39570 }, { "epoch": 0.92, "learning_rate": 2.3213900166623644e-05, "loss": 0.0827, "step": 39575 }, { "epoch": 0.92, "learning_rate": 2.321302011217761e-05, "loss": 0.1499, "step": 39580 }, { "epoch": 0.92, "learning_rate": 2.321214005773157e-05, "loss": 0.1151, "step": 39585 }, { "epoch": 0.92, "learning_rate": 2.3211260003285536e-05, "loss": 0.2922, "step": 39590 }, { "epoch": 0.92, "learning_rate": 2.3210379948839504e-05, "loss": 0.2196, "step": 39595 }, { "epoch": 0.92, "learning_rate": 2.3209499894393466e-05, "loss": 0.3872, "step": 39600 }, { "epoch": 0.92, "learning_rate": 2.320861983994743e-05, "loss": 0.1919, "step": 39605 }, { "epoch": 0.92, "learning_rate": 2.3207739785501396e-05, "loss": 0.1611, "step": 39610 }, { "epoch": 0.92, "learning_rate": 2.3206859731055364e-05, "loss": 0.0214, "step": 39615 }, { "epoch": 0.92, "learning_rate": 2.3205979676609326e-05, "loss": 0.2137, "step": 39620 }, { "epoch": 0.92, "learning_rate": 2.320509962216329e-05, "loss": 0.163, "step": 39625 }, { "epoch": 0.92, "learning_rate": 2.320421956771726e-05, "loss": 0.121, "step": 39630 }, { "epoch": 0.92, "learning_rate": 2.320333951327122e-05, "loss": 0.4035, "step": 39635 }, { "epoch": 0.92, "learning_rate": 2.3202459458825186e-05, "loss": 0.254, "step": 39640 }, { "epoch": 0.92, "learning_rate": 2.320157940437915e-05, "loss": 0.295, "step": 39645 }, { "epoch": 0.93, "learning_rate": 2.320069934993312e-05, "loss": 0.2964, "step": 39650 }, { "epoch": 0.93, "learning_rate": 2.319981929548708e-05, "loss": 0.2797, "step": 39655 }, { "epoch": 0.93, "learning_rate": 2.3198939241041046e-05, "loss": 0.1411, "step": 39660 }, { "epoch": 0.93, "learning_rate": 2.319805918659501e-05, "loss": 0.1065, "step": 39665 }, { "epoch": 0.93, "learning_rate": 2.3197179132148976e-05, "loss": 0.064, "step": 39670 }, { "epoch": 0.93, "learning_rate": 2.319629907770294e-05, "loss": 0.2198, "step": 39675 }, { "epoch": 0.93, "learning_rate": 2.3195419023256906e-05, "loss": 0.1872, "step": 39680 }, { "epoch": 0.93, "learning_rate": 2.319453896881087e-05, "loss": 0.1119, "step": 39685 }, { "epoch": 0.93, "learning_rate": 2.3193658914364836e-05, "loss": 0.2078, "step": 39690 }, { "epoch": 0.93, "learning_rate": 2.31927788599188e-05, "loss": 0.3993, "step": 39695 }, { "epoch": 0.93, "learning_rate": 2.3191898805472766e-05, "loss": 0.6262, "step": 39700 }, { "epoch": 0.93, "learning_rate": 2.319101875102673e-05, "loss": 0.2627, "step": 39705 }, { "epoch": 0.93, "learning_rate": 2.3190138696580696e-05, "loss": 0.0273, "step": 39710 }, { "epoch": 0.93, "learning_rate": 2.318925864213466e-05, "loss": 0.1491, "step": 39715 }, { "epoch": 0.93, "learning_rate": 2.3188378587688626e-05, "loss": 0.096, "step": 39720 }, { "epoch": 0.93, "learning_rate": 2.3187498533242588e-05, "loss": 0.1841, "step": 39725 }, { "epoch": 0.93, "learning_rate": 2.3186618478796556e-05, "loss": 0.1029, "step": 39730 }, { "epoch": 0.93, "learning_rate": 2.318573842435052e-05, "loss": 0.3498, "step": 39735 }, { "epoch": 0.93, "learning_rate": 2.3184858369904486e-05, "loss": 0.3565, "step": 39740 }, { "epoch": 0.93, "learning_rate": 2.318397831545845e-05, "loss": 0.2274, "step": 39745 }, { "epoch": 0.93, "learning_rate": 2.3183098261012416e-05, "loss": 0.4331, "step": 39750 }, { "epoch": 0.93, "learning_rate": 2.318221820656638e-05, "loss": 0.1747, "step": 39755 }, { "epoch": 0.93, "learning_rate": 2.3181338152120343e-05, "loss": 0.0177, "step": 39760 }, { "epoch": 0.93, "learning_rate": 2.318045809767431e-05, "loss": 0.046, "step": 39765 }, { "epoch": 0.93, "learning_rate": 2.3179578043228276e-05, "loss": 0.1843, "step": 39770 }, { "epoch": 0.93, "learning_rate": 2.317869798878224e-05, "loss": 0.0766, "step": 39775 }, { "epoch": 0.93, "learning_rate": 2.3177817934336203e-05, "loss": 0.0815, "step": 39780 }, { "epoch": 0.93, "learning_rate": 2.317693787989017e-05, "loss": 0.12, "step": 39785 }, { "epoch": 0.93, "learning_rate": 2.3176057825444136e-05, "loss": 0.1683, "step": 39790 }, { "epoch": 0.93, "learning_rate": 2.3175177770998098e-05, "loss": 0.3256, "step": 39795 }, { "epoch": 0.93, "learning_rate": 2.3174297716552063e-05, "loss": 0.3793, "step": 39800 }, { "epoch": 0.93, "learning_rate": 2.317341766210603e-05, "loss": 0.2779, "step": 39805 }, { "epoch": 0.93, "learning_rate": 2.3172537607659996e-05, "loss": 0.0995, "step": 39810 }, { "epoch": 0.93, "learning_rate": 2.3171657553213958e-05, "loss": 0.0745, "step": 39815 }, { "epoch": 0.93, "learning_rate": 2.3170777498767923e-05, "loss": 0.1484, "step": 39820 }, { "epoch": 0.93, "learning_rate": 2.316989744432189e-05, "loss": 0.0986, "step": 39825 }, { "epoch": 0.93, "learning_rate": 2.3169017389875853e-05, "loss": 0.1167, "step": 39830 }, { "epoch": 0.93, "learning_rate": 2.3168137335429818e-05, "loss": 0.3235, "step": 39835 }, { "epoch": 0.93, "learning_rate": 2.3167257280983783e-05, "loss": 0.37, "step": 39840 }, { "epoch": 0.93, "learning_rate": 2.316637722653775e-05, "loss": 0.2835, "step": 39845 }, { "epoch": 0.93, "learning_rate": 2.3165497172091713e-05, "loss": 0.5457, "step": 39850 }, { "epoch": 0.93, "learning_rate": 2.3164617117645678e-05, "loss": 0.2499, "step": 39855 }, { "epoch": 0.93, "learning_rate": 2.3163737063199643e-05, "loss": 0.077, "step": 39860 }, { "epoch": 0.93, "learning_rate": 2.3162857008753608e-05, "loss": 0.0878, "step": 39865 }, { "epoch": 0.93, "learning_rate": 2.3161976954307573e-05, "loss": 0.0973, "step": 39870 }, { "epoch": 0.93, "learning_rate": 2.316109689986154e-05, "loss": 0.2596, "step": 39875 }, { "epoch": 0.93, "learning_rate": 2.3160216845415503e-05, "loss": 0.2374, "step": 39880 }, { "epoch": 0.93, "learning_rate": 2.315933679096947e-05, "loss": 0.1191, "step": 39885 }, { "epoch": 0.93, "learning_rate": 2.3158456736523433e-05, "loss": 0.1761, "step": 39890 }, { "epoch": 0.93, "learning_rate": 2.31575766820774e-05, "loss": 0.2598, "step": 39895 }, { "epoch": 0.93, "learning_rate": 2.3156696627631363e-05, "loss": 0.4568, "step": 39900 }, { "epoch": 0.93, "learning_rate": 2.315581657318533e-05, "loss": 0.2991, "step": 39905 }, { "epoch": 0.93, "learning_rate": 2.3154936518739293e-05, "loss": 0.0795, "step": 39910 }, { "epoch": 0.93, "learning_rate": 2.315405646429326e-05, "loss": 0.0405, "step": 39915 }, { "epoch": 0.93, "learning_rate": 2.3153176409847223e-05, "loss": 0.1058, "step": 39920 }, { "epoch": 0.93, "learning_rate": 2.315229635540119e-05, "loss": 0.2024, "step": 39925 }, { "epoch": 0.93, "learning_rate": 2.3151416300955154e-05, "loss": 0.1386, "step": 39930 }, { "epoch": 0.93, "learning_rate": 2.3150536246509115e-05, "loss": 0.1212, "step": 39935 }, { "epoch": 0.93, "learning_rate": 2.3149656192063084e-05, "loss": 0.2468, "step": 39940 }, { "epoch": 0.93, "learning_rate": 2.314877613761705e-05, "loss": 0.2901, "step": 39945 }, { "epoch": 0.93, "learning_rate": 2.3147896083171014e-05, "loss": 0.4343, "step": 39950 }, { "epoch": 0.93, "learning_rate": 2.3147016028724975e-05, "loss": 0.1955, "step": 39955 }, { "epoch": 0.93, "learning_rate": 2.3146135974278944e-05, "loss": 0.0844, "step": 39960 }, { "epoch": 0.93, "learning_rate": 2.314525591983291e-05, "loss": 0.0899, "step": 39965 }, { "epoch": 0.93, "learning_rate": 2.314437586538687e-05, "loss": 0.1081, "step": 39970 }, { "epoch": 0.93, "learning_rate": 2.3143495810940835e-05, "loss": 0.1268, "step": 39975 }, { "epoch": 0.93, "learning_rate": 2.3142615756494804e-05, "loss": 0.0636, "step": 39980 }, { "epoch": 0.93, "learning_rate": 2.314173570204877e-05, "loss": 0.1254, "step": 39985 }, { "epoch": 0.93, "learning_rate": 2.314085564760273e-05, "loss": 0.3419, "step": 39990 }, { "epoch": 0.93, "learning_rate": 2.3139975593156695e-05, "loss": 0.3104, "step": 39995 }, { "epoch": 0.93, "learning_rate": 2.3139095538710664e-05, "loss": 0.2969, "step": 40000 }, { "epoch": 0.93, "learning_rate": 2.313821548426463e-05, "loss": 0.1167, "step": 40005 }, { "epoch": 0.93, "learning_rate": 2.313733542981859e-05, "loss": 0.0328, "step": 40010 }, { "epoch": 0.93, "learning_rate": 2.3136455375372555e-05, "loss": 0.1214, "step": 40015 }, { "epoch": 0.93, "learning_rate": 2.3135575320926524e-05, "loss": 0.0793, "step": 40020 }, { "epoch": 0.93, "learning_rate": 2.3134695266480486e-05, "loss": 0.1307, "step": 40025 }, { "epoch": 0.93, "learning_rate": 2.313381521203445e-05, "loss": 0.0974, "step": 40030 }, { "epoch": 0.93, "learning_rate": 2.313293515758842e-05, "loss": 0.2112, "step": 40035 }, { "epoch": 0.93, "learning_rate": 2.3132055103142384e-05, "loss": 0.1797, "step": 40040 }, { "epoch": 0.93, "learning_rate": 2.3131175048696346e-05, "loss": 0.3348, "step": 40045 }, { "epoch": 0.93, "learning_rate": 2.313029499425031e-05, "loss": 0.3724, "step": 40050 }, { "epoch": 0.93, "learning_rate": 2.312941493980428e-05, "loss": 0.1549, "step": 40055 }, { "epoch": 0.93, "learning_rate": 2.312853488535824e-05, "loss": 0.0404, "step": 40060 }, { "epoch": 0.93, "learning_rate": 2.3127654830912206e-05, "loss": 0.0266, "step": 40065 }, { "epoch": 0.93, "learning_rate": 2.312677477646617e-05, "loss": 0.1536, "step": 40070 }, { "epoch": 0.93, "learning_rate": 2.312589472202014e-05, "loss": 0.1674, "step": 40075 }, { "epoch": 0.94, "learning_rate": 2.31250146675741e-05, "loss": 0.1378, "step": 40080 }, { "epoch": 0.94, "learning_rate": 2.3124134613128066e-05, "loss": 0.231, "step": 40085 }, { "epoch": 0.94, "learning_rate": 2.312325455868203e-05, "loss": 0.1188, "step": 40090 }, { "epoch": 0.94, "learning_rate": 2.3122374504235996e-05, "loss": 0.2727, "step": 40095 }, { "epoch": 0.94, "learning_rate": 2.312149444978996e-05, "loss": 0.4323, "step": 40100 }, { "epoch": 0.94, "learning_rate": 2.3120614395343926e-05, "loss": 0.1972, "step": 40105 }, { "epoch": 0.94, "learning_rate": 2.311973434089789e-05, "loss": 0.0846, "step": 40110 }, { "epoch": 0.94, "learning_rate": 2.3118854286451856e-05, "loss": 0.0663, "step": 40115 }, { "epoch": 0.94, "learning_rate": 2.311797423200582e-05, "loss": 0.0352, "step": 40120 }, { "epoch": 0.94, "learning_rate": 2.3117094177559786e-05, "loss": 0.0904, "step": 40125 }, { "epoch": 0.94, "learning_rate": 2.3116214123113748e-05, "loss": 0.2203, "step": 40130 }, { "epoch": 0.94, "learning_rate": 2.3115334068667716e-05, "loss": 0.2654, "step": 40135 }, { "epoch": 0.94, "learning_rate": 2.311445401422168e-05, "loss": 0.1804, "step": 40140 }, { "epoch": 0.94, "learning_rate": 2.3113573959775646e-05, "loss": 0.4679, "step": 40145 }, { "epoch": 0.94, "learning_rate": 2.311269390532961e-05, "loss": 0.4914, "step": 40150 }, { "epoch": 0.94, "learning_rate": 2.3111813850883576e-05, "loss": 0.1626, "step": 40155 }, { "epoch": 0.94, "learning_rate": 2.311093379643754e-05, "loss": 0.0274, "step": 40160 }, { "epoch": 0.94, "learning_rate": 2.3110053741991503e-05, "loss": 0.0236, "step": 40165 }, { "epoch": 0.94, "learning_rate": 2.310917368754547e-05, "loss": 0.1277, "step": 40170 }, { "epoch": 0.94, "learning_rate": 2.3108293633099436e-05, "loss": 0.0795, "step": 40175 }, { "epoch": 0.94, "learning_rate": 2.31074135786534e-05, "loss": 0.2415, "step": 40180 }, { "epoch": 0.94, "learning_rate": 2.3106533524207363e-05, "loss": 0.1525, "step": 40185 }, { "epoch": 0.94, "learning_rate": 2.310565346976133e-05, "loss": 0.2131, "step": 40190 }, { "epoch": 0.94, "learning_rate": 2.3104773415315296e-05, "loss": 0.3518, "step": 40195 }, { "epoch": 0.94, "learning_rate": 2.3103893360869258e-05, "loss": 0.3944, "step": 40200 }, { "epoch": 0.94, "learning_rate": 2.3103013306423223e-05, "loss": 0.2728, "step": 40205 }, { "epoch": 0.94, "learning_rate": 2.310213325197719e-05, "loss": 0.1069, "step": 40210 }, { "epoch": 0.94, "learning_rate": 2.3101253197531156e-05, "loss": 0.1844, "step": 40215 }, { "epoch": 0.94, "learning_rate": 2.3100373143085118e-05, "loss": 0.2349, "step": 40220 }, { "epoch": 0.94, "learning_rate": 2.3099493088639083e-05, "loss": 0.0993, "step": 40225 }, { "epoch": 0.94, "learning_rate": 2.309861303419305e-05, "loss": 0.084, "step": 40230 }, { "epoch": 0.94, "learning_rate": 2.3097732979747013e-05, "loss": 0.1449, "step": 40235 }, { "epoch": 0.94, "learning_rate": 2.3096852925300978e-05, "loss": 0.186, "step": 40240 }, { "epoch": 0.94, "learning_rate": 2.3095972870854943e-05, "loss": 0.1159, "step": 40245 }, { "epoch": 0.94, "learning_rate": 2.309509281640891e-05, "loss": 0.5705, "step": 40250 }, { "epoch": 0.94, "learning_rate": 2.3094212761962873e-05, "loss": 0.1658, "step": 40255 }, { "epoch": 0.94, "learning_rate": 2.3093332707516838e-05, "loss": 0.1035, "step": 40260 }, { "epoch": 0.94, "learning_rate": 2.3092452653070803e-05, "loss": 0.1353, "step": 40265 }, { "epoch": 0.94, "learning_rate": 2.3091572598624768e-05, "loss": 0.1232, "step": 40270 }, { "epoch": 0.94, "learning_rate": 2.3090692544178733e-05, "loss": 0.1697, "step": 40275 }, { "epoch": 0.94, "learning_rate": 2.3089812489732698e-05, "loss": 0.1726, "step": 40280 }, { "epoch": 0.94, "learning_rate": 2.3088932435286663e-05, "loss": 0.1428, "step": 40285 }, { "epoch": 0.94, "learning_rate": 2.3088052380840628e-05, "loss": 0.2175, "step": 40290 }, { "epoch": 0.94, "learning_rate": 2.3087172326394593e-05, "loss": 0.372, "step": 40295 }, { "epoch": 0.94, "learning_rate": 2.3086292271948558e-05, "loss": 0.2987, "step": 40300 }, { "epoch": 0.94, "learning_rate": 2.3085412217502527e-05, "loss": 0.2611, "step": 40305 }, { "epoch": 0.94, "learning_rate": 2.308453216305649e-05, "loss": 0.1076, "step": 40310 }, { "epoch": 0.94, "learning_rate": 2.3083652108610453e-05, "loss": 0.172, "step": 40315 }, { "epoch": 0.94, "learning_rate": 2.308277205416442e-05, "loss": 0.1515, "step": 40320 }, { "epoch": 0.94, "learning_rate": 2.3081891999718383e-05, "loss": 0.0962, "step": 40325 }, { "epoch": 0.94, "learning_rate": 2.308101194527235e-05, "loss": 0.1235, "step": 40330 }, { "epoch": 0.94, "learning_rate": 2.3080131890826313e-05, "loss": 0.1124, "step": 40335 }, { "epoch": 0.94, "learning_rate": 2.307925183638028e-05, "loss": 0.2664, "step": 40340 }, { "epoch": 0.94, "learning_rate": 2.3078371781934243e-05, "loss": 0.4751, "step": 40345 }, { "epoch": 0.94, "learning_rate": 2.307749172748821e-05, "loss": 0.3549, "step": 40350 }, { "epoch": 0.94, "learning_rate": 2.3076611673042174e-05, "loss": 0.1764, "step": 40355 }, { "epoch": 0.94, "learning_rate": 2.3075731618596135e-05, "loss": 0.0311, "step": 40360 }, { "epoch": 0.94, "learning_rate": 2.3074851564150104e-05, "loss": 0.1266, "step": 40365 }, { "epoch": 0.94, "learning_rate": 2.307397150970407e-05, "loss": 0.0714, "step": 40370 }, { "epoch": 0.94, "learning_rate": 2.3073091455258034e-05, "loss": 0.1498, "step": 40375 }, { "epoch": 0.94, "learning_rate": 2.3072211400811995e-05, "loss": 0.2164, "step": 40380 }, { "epoch": 0.94, "learning_rate": 2.3071331346365964e-05, "loss": 0.1715, "step": 40385 }, { "epoch": 0.94, "learning_rate": 2.307045129191993e-05, "loss": 0.2907, "step": 40390 }, { "epoch": 0.94, "learning_rate": 2.306957123747389e-05, "loss": 0.384, "step": 40395 }, { "epoch": 0.94, "learning_rate": 2.3068691183027855e-05, "loss": 0.5047, "step": 40400 }, { "epoch": 0.94, "learning_rate": 2.3067811128581824e-05, "loss": 0.1645, "step": 40405 }, { "epoch": 0.94, "learning_rate": 2.306693107413579e-05, "loss": 0.0461, "step": 40410 }, { "epoch": 0.94, "learning_rate": 2.306605101968975e-05, "loss": 0.1118, "step": 40415 }, { "epoch": 0.94, "learning_rate": 2.3065170965243715e-05, "loss": 0.1843, "step": 40420 }, { "epoch": 0.94, "learning_rate": 2.3064290910797684e-05, "loss": 0.2712, "step": 40425 }, { "epoch": 0.94, "learning_rate": 2.3063410856351645e-05, "loss": 0.0522, "step": 40430 }, { "epoch": 0.94, "learning_rate": 2.306253080190561e-05, "loss": 0.2251, "step": 40435 }, { "epoch": 0.94, "learning_rate": 2.306165074745958e-05, "loss": 0.0467, "step": 40440 }, { "epoch": 0.94, "learning_rate": 2.3060770693013544e-05, "loss": 0.3861, "step": 40445 }, { "epoch": 0.94, "learning_rate": 2.3059890638567506e-05, "loss": 0.5126, "step": 40450 }, { "epoch": 0.94, "learning_rate": 2.305901058412147e-05, "loss": 0.2586, "step": 40455 }, { "epoch": 0.94, "learning_rate": 2.305813052967544e-05, "loss": 0.093, "step": 40460 }, { "epoch": 0.94, "learning_rate": 2.30572504752294e-05, "loss": 0.0713, "step": 40465 }, { "epoch": 0.94, "learning_rate": 2.3056370420783366e-05, "loss": 0.1364, "step": 40470 }, { "epoch": 0.94, "learning_rate": 2.305549036633733e-05, "loss": 0.1365, "step": 40475 }, { "epoch": 0.94, "learning_rate": 2.30546103118913e-05, "loss": 0.1333, "step": 40480 }, { "epoch": 0.94, "learning_rate": 2.305373025744526e-05, "loss": 0.1112, "step": 40485 }, { "epoch": 0.94, "learning_rate": 2.3052850202999226e-05, "loss": 0.0985, "step": 40490 }, { "epoch": 0.94, "learning_rate": 2.305197014855319e-05, "loss": 0.2275, "step": 40495 }, { "epoch": 0.94, "learning_rate": 2.3051090094107156e-05, "loss": 0.701, "step": 40500 }, { "epoch": 0.95, "learning_rate": 2.305021003966112e-05, "loss": 0.1451, "step": 40505 }, { "epoch": 0.95, "learning_rate": 2.3049329985215086e-05, "loss": 0.0301, "step": 40510 }, { "epoch": 0.95, "learning_rate": 2.304844993076905e-05, "loss": 0.0781, "step": 40515 }, { "epoch": 0.95, "learning_rate": 2.3047569876323016e-05, "loss": 0.0964, "step": 40520 }, { "epoch": 0.95, "learning_rate": 2.304668982187698e-05, "loss": 0.0553, "step": 40525 }, { "epoch": 0.95, "learning_rate": 2.3045809767430946e-05, "loss": 0.128, "step": 40530 }, { "epoch": 0.95, "learning_rate": 2.3044929712984907e-05, "loss": 0.0979, "step": 40535 }, { "epoch": 0.95, "learning_rate": 2.3044049658538876e-05, "loss": 0.2023, "step": 40540 }, { "epoch": 0.95, "learning_rate": 2.304316960409284e-05, "loss": 0.2759, "step": 40545 }, { "epoch": 0.95, "learning_rate": 2.3042289549646806e-05, "loss": 0.73, "step": 40550 }, { "epoch": 0.95, "learning_rate": 2.3041409495200768e-05, "loss": 0.2415, "step": 40555 }, { "epoch": 0.95, "learning_rate": 2.3040529440754736e-05, "loss": 0.119, "step": 40560 }, { "epoch": 0.95, "learning_rate": 2.30396493863087e-05, "loss": 0.0694, "step": 40565 }, { "epoch": 0.95, "learning_rate": 2.3038769331862663e-05, "loss": 0.0847, "step": 40570 }, { "epoch": 0.95, "learning_rate": 2.303788927741663e-05, "loss": 0.0892, "step": 40575 }, { "epoch": 0.95, "learning_rate": 2.3037009222970596e-05, "loss": 0.1931, "step": 40580 }, { "epoch": 0.95, "learning_rate": 2.303612916852456e-05, "loss": 0.1451, "step": 40585 }, { "epoch": 0.95, "learning_rate": 2.3035249114078523e-05, "loss": 0.1509, "step": 40590 }, { "epoch": 0.95, "learning_rate": 2.303436905963249e-05, "loss": 0.2372, "step": 40595 }, { "epoch": 0.95, "learning_rate": 2.3033489005186456e-05, "loss": 0.2959, "step": 40600 }, { "epoch": 0.95, "learning_rate": 2.303260895074042e-05, "loss": 0.1157, "step": 40605 }, { "epoch": 0.95, "learning_rate": 2.3031728896294383e-05, "loss": 0.1886, "step": 40610 }, { "epoch": 0.95, "learning_rate": 2.303084884184835e-05, "loss": 0.1194, "step": 40615 }, { "epoch": 0.95, "learning_rate": 2.3029968787402316e-05, "loss": 0.1291, "step": 40620 }, { "epoch": 0.95, "learning_rate": 2.3029088732956278e-05, "loss": 0.1319, "step": 40625 }, { "epoch": 0.95, "learning_rate": 2.3028208678510243e-05, "loss": 0.196, "step": 40630 }, { "epoch": 0.95, "learning_rate": 2.302732862406421e-05, "loss": 0.2379, "step": 40635 }, { "epoch": 0.95, "learning_rate": 2.3026448569618176e-05, "loss": 0.3169, "step": 40640 }, { "epoch": 0.95, "learning_rate": 2.3025568515172138e-05, "loss": 0.1696, "step": 40645 }, { "epoch": 0.95, "learning_rate": 2.3024688460726103e-05, "loss": 0.4145, "step": 40650 }, { "epoch": 0.95, "learning_rate": 2.302380840628007e-05, "loss": 0.1936, "step": 40655 }, { "epoch": 0.95, "learning_rate": 2.3022928351834033e-05, "loss": 0.0721, "step": 40660 }, { "epoch": 0.95, "learning_rate": 2.3022048297387998e-05, "loss": 0.0557, "step": 40665 }, { "epoch": 0.95, "learning_rate": 2.3021168242941963e-05, "loss": 0.0688, "step": 40670 }, { "epoch": 0.95, "learning_rate": 2.302028818849593e-05, "loss": 0.1703, "step": 40675 }, { "epoch": 0.95, "learning_rate": 2.3019408134049893e-05, "loss": 0.133, "step": 40680 }, { "epoch": 0.95, "learning_rate": 2.3018528079603858e-05, "loss": 0.1861, "step": 40685 }, { "epoch": 0.95, "learning_rate": 2.3017648025157823e-05, "loss": 0.3069, "step": 40690 }, { "epoch": 0.95, "learning_rate": 2.3016767970711788e-05, "loss": 0.1, "step": 40695 }, { "epoch": 0.95, "learning_rate": 2.3015887916265753e-05, "loss": 0.8214, "step": 40700 }, { "epoch": 0.95, "learning_rate": 2.3015007861819718e-05, "loss": 0.2138, "step": 40705 }, { "epoch": 0.95, "learning_rate": 2.3014127807373687e-05, "loss": 0.0487, "step": 40710 }, { "epoch": 0.95, "learning_rate": 2.3013247752927648e-05, "loss": 0.0939, "step": 40715 }, { "epoch": 0.95, "learning_rate": 2.3012367698481613e-05, "loss": 0.0584, "step": 40720 }, { "epoch": 0.95, "learning_rate": 2.3011487644035578e-05, "loss": 0.1464, "step": 40725 }, { "epoch": 0.95, "learning_rate": 2.3010607589589543e-05, "loss": 0.0883, "step": 40730 }, { "epoch": 0.95, "learning_rate": 2.3009727535143508e-05, "loss": 0.1658, "step": 40735 }, { "epoch": 0.95, "learning_rate": 2.3008847480697473e-05, "loss": 0.2553, "step": 40740 }, { "epoch": 0.95, "learning_rate": 2.300796742625144e-05, "loss": 0.1432, "step": 40745 }, { "epoch": 0.95, "learning_rate": 2.3007087371805403e-05, "loss": 0.4951, "step": 40750 }, { "epoch": 0.95, "learning_rate": 2.300620731735937e-05, "loss": 0.1627, "step": 40755 }, { "epoch": 0.95, "learning_rate": 2.3005327262913333e-05, "loss": 0.0621, "step": 40760 }, { "epoch": 0.95, "learning_rate": 2.3004447208467295e-05, "loss": 0.0789, "step": 40765 }, { "epoch": 0.95, "learning_rate": 2.3003567154021263e-05, "loss": 0.1342, "step": 40770 }, { "epoch": 0.95, "learning_rate": 2.300268709957523e-05, "loss": 0.2025, "step": 40775 }, { "epoch": 0.95, "learning_rate": 2.3001807045129193e-05, "loss": 0.1505, "step": 40780 }, { "epoch": 0.95, "learning_rate": 2.3000926990683155e-05, "loss": 0.2379, "step": 40785 }, { "epoch": 0.95, "learning_rate": 2.3000046936237124e-05, "loss": 0.2839, "step": 40790 }, { "epoch": 0.95, "learning_rate": 2.299916688179109e-05, "loss": 0.2426, "step": 40795 }, { "epoch": 0.95, "learning_rate": 2.299828682734505e-05, "loss": 0.6808, "step": 40800 }, { "epoch": 0.95, "learning_rate": 2.2997406772899015e-05, "loss": 0.2815, "step": 40805 }, { "epoch": 0.95, "learning_rate": 2.2996526718452984e-05, "loss": 0.0624, "step": 40810 }, { "epoch": 0.95, "learning_rate": 2.299564666400695e-05, "loss": 0.0764, "step": 40815 }, { "epoch": 0.95, "learning_rate": 2.299476660956091e-05, "loss": 0.0676, "step": 40820 }, { "epoch": 0.95, "learning_rate": 2.2993886555114875e-05, "loss": 0.0537, "step": 40825 }, { "epoch": 0.95, "learning_rate": 2.2993006500668844e-05, "loss": 0.0507, "step": 40830 }, { "epoch": 0.95, "learning_rate": 2.2992126446222805e-05, "loss": 0.0678, "step": 40835 }, { "epoch": 0.95, "learning_rate": 2.299124639177677e-05, "loss": 0.3164, "step": 40840 }, { "epoch": 0.95, "learning_rate": 2.299036633733074e-05, "loss": 0.3974, "step": 40845 }, { "epoch": 0.95, "learning_rate": 2.2989486282884704e-05, "loss": 0.4675, "step": 40850 }, { "epoch": 0.95, "learning_rate": 2.2988606228438665e-05, "loss": 0.1771, "step": 40855 }, { "epoch": 0.95, "learning_rate": 2.298772617399263e-05, "loss": 0.1284, "step": 40860 }, { "epoch": 0.95, "learning_rate": 2.29868461195466e-05, "loss": 0.1004, "step": 40865 }, { "epoch": 0.95, "learning_rate": 2.298596606510056e-05, "loss": 0.0726, "step": 40870 }, { "epoch": 0.95, "learning_rate": 2.2985086010654525e-05, "loss": 0.1704, "step": 40875 }, { "epoch": 0.95, "learning_rate": 2.298420595620849e-05, "loss": 0.1177, "step": 40880 }, { "epoch": 0.95, "learning_rate": 2.298332590176246e-05, "loss": 0.1778, "step": 40885 }, { "epoch": 0.95, "learning_rate": 2.298244584731642e-05, "loss": 0.1745, "step": 40890 }, { "epoch": 0.95, "learning_rate": 2.2981565792870386e-05, "loss": 0.2976, "step": 40895 }, { "epoch": 0.95, "learning_rate": 2.298068573842435e-05, "loss": 0.2557, "step": 40900 }, { "epoch": 0.95, "learning_rate": 2.2979805683978316e-05, "loss": 0.1721, "step": 40905 }, { "epoch": 0.95, "learning_rate": 2.297892562953228e-05, "loss": 0.1759, "step": 40910 }, { "epoch": 0.95, "learning_rate": 2.2978045575086246e-05, "loss": 0.1878, "step": 40915 }, { "epoch": 0.95, "learning_rate": 2.297716552064021e-05, "loss": 0.1806, "step": 40920 }, { "epoch": 0.95, "learning_rate": 2.2976285466194176e-05, "loss": 0.0929, "step": 40925 }, { "epoch": 0.95, "learning_rate": 2.297540541174814e-05, "loss": 0.2208, "step": 40930 }, { "epoch": 0.96, "learning_rate": 2.2974525357302106e-05, "loss": 0.2522, "step": 40935 }, { "epoch": 0.96, "learning_rate": 2.297364530285607e-05, "loss": 0.111, "step": 40940 }, { "epoch": 0.96, "learning_rate": 2.2972765248410036e-05, "loss": 0.2135, "step": 40945 }, { "epoch": 0.96, "learning_rate": 2.2971885193964e-05, "loss": 0.4884, "step": 40950 }, { "epoch": 0.96, "learning_rate": 2.2971005139517966e-05, "loss": 0.2125, "step": 40955 }, { "epoch": 0.96, "learning_rate": 2.2970125085071927e-05, "loss": 0.0386, "step": 40960 }, { "epoch": 0.96, "learning_rate": 2.2969245030625896e-05, "loss": 0.0592, "step": 40965 }, { "epoch": 0.96, "learning_rate": 2.296836497617986e-05, "loss": 0.1067, "step": 40970 }, { "epoch": 0.96, "learning_rate": 2.2967484921733826e-05, "loss": 0.1098, "step": 40975 }, { "epoch": 0.96, "learning_rate": 2.296660486728779e-05, "loss": 0.1373, "step": 40980 }, { "epoch": 0.96, "learning_rate": 2.2965724812841756e-05, "loss": 0.179, "step": 40985 }, { "epoch": 0.96, "learning_rate": 2.296484475839572e-05, "loss": 0.3522, "step": 40990 }, { "epoch": 0.96, "learning_rate": 2.2963964703949683e-05, "loss": 0.4414, "step": 40995 }, { "epoch": 0.96, "learning_rate": 2.296308464950365e-05, "loss": 0.4159, "step": 41000 }, { "epoch": 0.96, "learning_rate": 2.2962204595057616e-05, "loss": 0.2406, "step": 41005 }, { "epoch": 0.96, "learning_rate": 2.296132454061158e-05, "loss": 0.0204, "step": 41010 }, { "epoch": 0.96, "learning_rate": 2.2960444486165543e-05, "loss": 0.0849, "step": 41015 }, { "epoch": 0.96, "learning_rate": 2.295956443171951e-05, "loss": 0.0654, "step": 41020 }, { "epoch": 0.96, "learning_rate": 2.2958684377273476e-05, "loss": 0.0715, "step": 41025 }, { "epoch": 0.96, "learning_rate": 2.2957804322827438e-05, "loss": 0.1534, "step": 41030 }, { "epoch": 0.96, "learning_rate": 2.2956924268381403e-05, "loss": 0.2823, "step": 41035 }, { "epoch": 0.96, "learning_rate": 2.295604421393537e-05, "loss": 0.2045, "step": 41040 }, { "epoch": 0.96, "learning_rate": 2.2955164159489336e-05, "loss": 0.2369, "step": 41045 }, { "epoch": 0.96, "learning_rate": 2.2954284105043298e-05, "loss": 0.4408, "step": 41050 }, { "epoch": 0.96, "learning_rate": 2.2953404050597263e-05, "loss": 0.2871, "step": 41055 }, { "epoch": 0.96, "learning_rate": 2.295252399615123e-05, "loss": 0.0356, "step": 41060 }, { "epoch": 0.96, "learning_rate": 2.2951643941705193e-05, "loss": 0.0749, "step": 41065 }, { "epoch": 0.96, "learning_rate": 2.2950763887259158e-05, "loss": 0.1097, "step": 41070 }, { "epoch": 0.96, "learning_rate": 2.2949883832813123e-05, "loss": 0.1526, "step": 41075 }, { "epoch": 0.96, "learning_rate": 2.294900377836709e-05, "loss": 0.2101, "step": 41080 }, { "epoch": 0.96, "learning_rate": 2.2948123723921053e-05, "loss": 0.1518, "step": 41085 }, { "epoch": 0.96, "learning_rate": 2.2947243669475018e-05, "loss": 0.2261, "step": 41090 }, { "epoch": 0.96, "learning_rate": 2.2946363615028983e-05, "loss": 0.1952, "step": 41095 }, { "epoch": 0.96, "learning_rate": 2.2945483560582948e-05, "loss": 0.3802, "step": 41100 }, { "epoch": 0.96, "learning_rate": 2.2944603506136913e-05, "loss": 0.193, "step": 41105 }, { "epoch": 0.96, "learning_rate": 2.2943723451690878e-05, "loss": 0.0846, "step": 41110 }, { "epoch": 0.96, "learning_rate": 2.2942843397244846e-05, "loss": 0.0687, "step": 41115 }, { "epoch": 0.96, "learning_rate": 2.2941963342798808e-05, "loss": 0.1857, "step": 41120 }, { "epoch": 0.96, "learning_rate": 2.2941083288352773e-05, "loss": 0.1647, "step": 41125 }, { "epoch": 0.96, "learning_rate": 2.2940203233906738e-05, "loss": 0.2132, "step": 41130 }, { "epoch": 0.96, "learning_rate": 2.2939323179460703e-05, "loss": 0.1771, "step": 41135 }, { "epoch": 0.96, "learning_rate": 2.2938443125014668e-05, "loss": 0.1534, "step": 41140 }, { "epoch": 0.96, "learning_rate": 2.2937563070568633e-05, "loss": 0.266, "step": 41145 }, { "epoch": 0.96, "learning_rate": 2.2936683016122598e-05, "loss": 0.3651, "step": 41150 }, { "epoch": 0.96, "learning_rate": 2.2935802961676563e-05, "loss": 0.1874, "step": 41155 }, { "epoch": 0.96, "learning_rate": 2.2934922907230528e-05, "loss": 0.0634, "step": 41160 }, { "epoch": 0.96, "learning_rate": 2.2934042852784493e-05, "loss": 0.0811, "step": 41165 }, { "epoch": 0.96, "learning_rate": 2.2933162798338455e-05, "loss": 0.1704, "step": 41170 }, { "epoch": 0.96, "learning_rate": 2.2932282743892423e-05, "loss": 0.0789, "step": 41175 }, { "epoch": 0.96, "learning_rate": 2.293140268944639e-05, "loss": 0.148, "step": 41180 }, { "epoch": 0.96, "learning_rate": 2.2930522635000353e-05, "loss": 0.1911, "step": 41185 }, { "epoch": 0.96, "learning_rate": 2.2929642580554315e-05, "loss": 0.1533, "step": 41190 }, { "epoch": 0.96, "learning_rate": 2.2928762526108283e-05, "loss": 0.2555, "step": 41195 }, { "epoch": 0.96, "learning_rate": 2.292788247166225e-05, "loss": 0.3689, "step": 41200 }, { "epoch": 0.96, "learning_rate": 2.292700241721621e-05, "loss": 0.1444, "step": 41205 }, { "epoch": 0.96, "learning_rate": 2.2926122362770175e-05, "loss": 0.0242, "step": 41210 }, { "epoch": 0.96, "learning_rate": 2.2925242308324143e-05, "loss": 0.1099, "step": 41215 }, { "epoch": 0.96, "learning_rate": 2.292436225387811e-05, "loss": 0.0555, "step": 41220 }, { "epoch": 0.96, "learning_rate": 2.292348219943207e-05, "loss": 0.11, "step": 41225 }, { "epoch": 0.96, "learning_rate": 2.2922602144986035e-05, "loss": 0.2327, "step": 41230 }, { "epoch": 0.96, "learning_rate": 2.2921722090540004e-05, "loss": 0.1206, "step": 41235 }, { "epoch": 0.96, "learning_rate": 2.292084203609397e-05, "loss": 0.2974, "step": 41240 }, { "epoch": 0.96, "learning_rate": 2.291996198164793e-05, "loss": 0.3251, "step": 41245 }, { "epoch": 0.96, "learning_rate": 2.29190819272019e-05, "loss": 0.3675, "step": 41250 }, { "epoch": 0.96, "learning_rate": 2.2918201872755864e-05, "loss": 0.2655, "step": 41255 }, { "epoch": 0.96, "learning_rate": 2.2917321818309825e-05, "loss": 0.0329, "step": 41260 }, { "epoch": 0.96, "learning_rate": 2.291644176386379e-05, "loss": 0.1273, "step": 41265 }, { "epoch": 0.96, "learning_rate": 2.291556170941776e-05, "loss": 0.1465, "step": 41270 }, { "epoch": 0.96, "learning_rate": 2.2914681654971724e-05, "loss": 0.2346, "step": 41275 }, { "epoch": 0.96, "learning_rate": 2.2913801600525685e-05, "loss": 0.0541, "step": 41280 }, { "epoch": 0.96, "learning_rate": 2.291292154607965e-05, "loss": 0.2283, "step": 41285 }, { "epoch": 0.96, "learning_rate": 2.291204149163362e-05, "loss": 0.1401, "step": 41290 }, { "epoch": 0.96, "learning_rate": 2.291116143718758e-05, "loss": 0.2349, "step": 41295 }, { "epoch": 0.96, "learning_rate": 2.2910281382741545e-05, "loss": 0.3634, "step": 41300 }, { "epoch": 0.96, "learning_rate": 2.290940132829551e-05, "loss": 0.1659, "step": 41305 }, { "epoch": 0.96, "learning_rate": 2.290852127384948e-05, "loss": 0.0879, "step": 41310 }, { "epoch": 0.96, "learning_rate": 2.290764121940344e-05, "loss": 0.0374, "step": 41315 }, { "epoch": 0.96, "learning_rate": 2.2906761164957406e-05, "loss": 0.1147, "step": 41320 }, { "epoch": 0.96, "learning_rate": 2.290588111051137e-05, "loss": 0.0736, "step": 41325 }, { "epoch": 0.96, "learning_rate": 2.2905001056065336e-05, "loss": 0.1792, "step": 41330 }, { "epoch": 0.96, "learning_rate": 2.29041210016193e-05, "loss": 0.1691, "step": 41335 }, { "epoch": 0.96, "learning_rate": 2.2903240947173266e-05, "loss": 0.1882, "step": 41340 }, { "epoch": 0.96, "learning_rate": 2.290236089272723e-05, "loss": 0.4799, "step": 41345 }, { "epoch": 0.96, "learning_rate": 2.2901480838281196e-05, "loss": 0.4995, "step": 41350 }, { "epoch": 0.96, "learning_rate": 2.290060078383516e-05, "loss": 0.1869, "step": 41355 }, { "epoch": 0.96, "learning_rate": 2.2899720729389126e-05, "loss": 0.0219, "step": 41360 }, { "epoch": 0.97, "learning_rate": 2.2898840674943087e-05, "loss": 0.227, "step": 41365 }, { "epoch": 0.97, "learning_rate": 2.2897960620497056e-05, "loss": 0.1751, "step": 41370 }, { "epoch": 0.97, "learning_rate": 2.289708056605102e-05, "loss": 0.2486, "step": 41375 }, { "epoch": 0.97, "learning_rate": 2.2896200511604986e-05, "loss": 0.0684, "step": 41380 }, { "epoch": 0.97, "learning_rate": 2.289532045715895e-05, "loss": 0.2124, "step": 41385 }, { "epoch": 0.97, "learning_rate": 2.2894440402712916e-05, "loss": 0.2885, "step": 41390 }, { "epoch": 0.97, "learning_rate": 2.289356034826688e-05, "loss": 0.1928, "step": 41395 }, { "epoch": 0.97, "learning_rate": 2.2892680293820842e-05, "loss": 0.5001, "step": 41400 }, { "epoch": 0.97, "learning_rate": 2.289180023937481e-05, "loss": 0.12, "step": 41405 }, { "epoch": 0.97, "learning_rate": 2.2890920184928776e-05, "loss": 0.1579, "step": 41410 }, { "epoch": 0.97, "learning_rate": 2.289004013048274e-05, "loss": 0.146, "step": 41415 }, { "epoch": 0.97, "learning_rate": 2.2889160076036703e-05, "loss": 0.0826, "step": 41420 }, { "epoch": 0.97, "learning_rate": 2.288828002159067e-05, "loss": 0.1789, "step": 41425 }, { "epoch": 0.97, "learning_rate": 2.2887399967144636e-05, "loss": 0.0774, "step": 41430 }, { "epoch": 0.97, "learning_rate": 2.2886519912698598e-05, "loss": 0.135, "step": 41435 }, { "epoch": 0.97, "learning_rate": 2.2885639858252563e-05, "loss": 0.2117, "step": 41440 }, { "epoch": 0.97, "learning_rate": 2.288475980380653e-05, "loss": 0.4228, "step": 41445 }, { "epoch": 0.97, "learning_rate": 2.2883879749360496e-05, "loss": 0.282, "step": 41450 }, { "epoch": 0.97, "learning_rate": 2.2882999694914458e-05, "loss": 0.1152, "step": 41455 }, { "epoch": 0.97, "learning_rate": 2.2882119640468423e-05, "loss": 0.1397, "step": 41460 }, { "epoch": 0.97, "learning_rate": 2.288123958602239e-05, "loss": 0.2521, "step": 41465 }, { "epoch": 0.97, "learning_rate": 2.2880359531576353e-05, "loss": 0.0619, "step": 41470 }, { "epoch": 0.97, "learning_rate": 2.2879479477130318e-05, "loss": 0.079, "step": 41475 }, { "epoch": 0.97, "learning_rate": 2.2878599422684283e-05, "loss": 0.2615, "step": 41480 }, { "epoch": 0.97, "learning_rate": 2.287771936823825e-05, "loss": 0.2039, "step": 41485 }, { "epoch": 0.97, "learning_rate": 2.2876839313792213e-05, "loss": 0.2199, "step": 41490 }, { "epoch": 0.97, "learning_rate": 2.2875959259346178e-05, "loss": 0.416, "step": 41495 }, { "epoch": 0.97, "learning_rate": 2.2875079204900143e-05, "loss": 0.5019, "step": 41500 }, { "epoch": 0.97, "learning_rate": 2.2874199150454108e-05, "loss": 0.3083, "step": 41505 }, { "epoch": 0.97, "learning_rate": 2.2873319096008073e-05, "loss": 0.0945, "step": 41510 }, { "epoch": 0.97, "learning_rate": 2.2872439041562038e-05, "loss": 0.0569, "step": 41515 }, { "epoch": 0.97, "learning_rate": 2.2871558987116006e-05, "loss": 0.1108, "step": 41520 }, { "epoch": 0.97, "learning_rate": 2.2870678932669968e-05, "loss": 0.1432, "step": 41525 }, { "epoch": 0.97, "learning_rate": 2.2869798878223933e-05, "loss": 0.2179, "step": 41530 }, { "epoch": 0.97, "learning_rate": 2.2868918823777898e-05, "loss": 0.0662, "step": 41535 }, { "epoch": 0.97, "learning_rate": 2.2868038769331866e-05, "loss": 0.1478, "step": 41540 }, { "epoch": 0.97, "learning_rate": 2.2867158714885828e-05, "loss": 0.2958, "step": 41545 }, { "epoch": 0.97, "learning_rate": 2.2866278660439793e-05, "loss": 0.2505, "step": 41550 }, { "epoch": 0.97, "learning_rate": 2.2865398605993758e-05, "loss": 0.1234, "step": 41555 }, { "epoch": 0.97, "learning_rate": 2.2864518551547723e-05, "loss": 0.0747, "step": 41560 }, { "epoch": 0.97, "learning_rate": 2.2863638497101688e-05, "loss": 0.1399, "step": 41565 }, { "epoch": 0.97, "learning_rate": 2.2862758442655653e-05, "loss": 0.1438, "step": 41570 }, { "epoch": 0.97, "learning_rate": 2.2861878388209618e-05, "loss": 0.1894, "step": 41575 }, { "epoch": 0.97, "learning_rate": 2.2860998333763583e-05, "loss": 0.1281, "step": 41580 }, { "epoch": 0.97, "learning_rate": 2.2860118279317548e-05, "loss": 0.3124, "step": 41585 }, { "epoch": 0.97, "learning_rate": 2.2859238224871513e-05, "loss": 0.3054, "step": 41590 }, { "epoch": 0.97, "learning_rate": 2.2858358170425475e-05, "loss": 0.3869, "step": 41595 }, { "epoch": 0.97, "learning_rate": 2.2857478115979443e-05, "loss": 0.3431, "step": 41600 }, { "epoch": 0.97, "learning_rate": 2.2856598061533408e-05, "loss": 0.1913, "step": 41605 }, { "epoch": 0.97, "learning_rate": 2.2855718007087373e-05, "loss": 0.0812, "step": 41610 }, { "epoch": 0.97, "learning_rate": 2.2854837952641335e-05, "loss": 0.109, "step": 41615 }, { "epoch": 0.97, "learning_rate": 2.2853957898195303e-05, "loss": 0.0389, "step": 41620 }, { "epoch": 0.97, "learning_rate": 2.285307784374927e-05, "loss": 0.2407, "step": 41625 }, { "epoch": 0.97, "learning_rate": 2.285219778930323e-05, "loss": 0.2419, "step": 41630 }, { "epoch": 0.97, "learning_rate": 2.2851317734857195e-05, "loss": 0.207, "step": 41635 }, { "epoch": 0.97, "learning_rate": 2.2850437680411163e-05, "loss": 0.3345, "step": 41640 }, { "epoch": 0.97, "learning_rate": 2.284955762596513e-05, "loss": 0.2601, "step": 41645 }, { "epoch": 0.97, "learning_rate": 2.284867757151909e-05, "loss": 0.4368, "step": 41650 }, { "epoch": 0.97, "learning_rate": 2.284779751707306e-05, "loss": 0.1836, "step": 41655 }, { "epoch": 0.97, "learning_rate": 2.2846917462627024e-05, "loss": 0.0555, "step": 41660 }, { "epoch": 0.97, "learning_rate": 2.2846037408180985e-05, "loss": 0.0618, "step": 41665 }, { "epoch": 0.97, "learning_rate": 2.284515735373495e-05, "loss": 0.0946, "step": 41670 }, { "epoch": 0.97, "learning_rate": 2.284427729928892e-05, "loss": 0.1289, "step": 41675 }, { "epoch": 0.97, "learning_rate": 2.2843397244842884e-05, "loss": 0.18, "step": 41680 }, { "epoch": 0.97, "learning_rate": 2.2842517190396845e-05, "loss": 0.1341, "step": 41685 }, { "epoch": 0.97, "learning_rate": 2.284163713595081e-05, "loss": 0.1191, "step": 41690 }, { "epoch": 0.97, "learning_rate": 2.284075708150478e-05, "loss": 0.181, "step": 41695 }, { "epoch": 0.97, "learning_rate": 2.283987702705874e-05, "loss": 0.5187, "step": 41700 }, { "epoch": 0.97, "learning_rate": 2.2838996972612705e-05, "loss": 0.149, "step": 41705 }, { "epoch": 0.97, "learning_rate": 2.283811691816667e-05, "loss": 0.124, "step": 41710 }, { "epoch": 0.97, "learning_rate": 2.283723686372064e-05, "loss": 0.109, "step": 41715 }, { "epoch": 0.97, "learning_rate": 2.28363568092746e-05, "loss": 0.1608, "step": 41720 }, { "epoch": 0.97, "learning_rate": 2.2835476754828565e-05, "loss": 0.179, "step": 41725 }, { "epoch": 0.97, "learning_rate": 2.283459670038253e-05, "loss": 0.1152, "step": 41730 }, { "epoch": 0.97, "learning_rate": 2.2833716645936495e-05, "loss": 0.1299, "step": 41735 }, { "epoch": 0.97, "learning_rate": 2.283283659149046e-05, "loss": 0.2732, "step": 41740 }, { "epoch": 0.97, "learning_rate": 2.2831956537044425e-05, "loss": 0.3138, "step": 41745 }, { "epoch": 0.97, "learning_rate": 2.283107648259839e-05, "loss": 0.5521, "step": 41750 }, { "epoch": 0.97, "learning_rate": 2.2830196428152356e-05, "loss": 0.1607, "step": 41755 }, { "epoch": 0.97, "learning_rate": 2.282931637370632e-05, "loss": 0.0673, "step": 41760 }, { "epoch": 0.97, "learning_rate": 2.2828436319260286e-05, "loss": 0.0546, "step": 41765 }, { "epoch": 0.97, "learning_rate": 2.2827556264814247e-05, "loss": 0.116, "step": 41770 }, { "epoch": 0.97, "learning_rate": 2.2826676210368216e-05, "loss": 0.1359, "step": 41775 }, { "epoch": 0.97, "learning_rate": 2.282579615592218e-05, "loss": 0.1839, "step": 41780 }, { "epoch": 0.97, "learning_rate": 2.2824916101476146e-05, "loss": 0.0991, "step": 41785 }, { "epoch": 0.98, "learning_rate": 2.282403604703011e-05, "loss": 0.1518, "step": 41790 }, { "epoch": 0.98, "learning_rate": 2.2823155992584076e-05, "loss": 0.4286, "step": 41795 }, { "epoch": 0.98, "learning_rate": 2.282227593813804e-05, "loss": 0.518, "step": 41800 }, { "epoch": 0.98, "learning_rate": 2.2821395883692002e-05, "loss": 0.2321, "step": 41805 }, { "epoch": 0.98, "learning_rate": 2.282051582924597e-05, "loss": 0.0654, "step": 41810 }, { "epoch": 0.98, "learning_rate": 2.2819635774799936e-05, "loss": 0.0842, "step": 41815 }, { "epoch": 0.98, "learning_rate": 2.28187557203539e-05, "loss": 0.0832, "step": 41820 }, { "epoch": 0.98, "learning_rate": 2.2817875665907862e-05, "loss": 0.1866, "step": 41825 }, { "epoch": 0.98, "learning_rate": 2.281699561146183e-05, "loss": 0.1209, "step": 41830 }, { "epoch": 0.98, "learning_rate": 2.2816115557015796e-05, "loss": 0.2317, "step": 41835 }, { "epoch": 0.98, "learning_rate": 2.281523550256976e-05, "loss": 0.2132, "step": 41840 }, { "epoch": 0.98, "learning_rate": 2.2814355448123723e-05, "loss": 0.332, "step": 41845 }, { "epoch": 0.98, "learning_rate": 2.281347539367769e-05, "loss": 0.3761, "step": 41850 }, { "epoch": 0.98, "learning_rate": 2.2812595339231656e-05, "loss": 0.227, "step": 41855 }, { "epoch": 0.98, "learning_rate": 2.2811715284785618e-05, "loss": 0.0299, "step": 41860 }, { "epoch": 0.98, "learning_rate": 2.2810835230339583e-05, "loss": 0.0935, "step": 41865 }, { "epoch": 0.98, "learning_rate": 2.280995517589355e-05, "loss": 0.0939, "step": 41870 }, { "epoch": 0.98, "learning_rate": 2.2809075121447516e-05, "loss": 0.1364, "step": 41875 }, { "epoch": 0.98, "learning_rate": 2.2808195067001478e-05, "loss": 0.1525, "step": 41880 }, { "epoch": 0.98, "learning_rate": 2.2807315012555443e-05, "loss": 0.2073, "step": 41885 }, { "epoch": 0.98, "learning_rate": 2.280643495810941e-05, "loss": 0.1712, "step": 41890 }, { "epoch": 0.98, "learning_rate": 2.2805554903663373e-05, "loss": 0.2458, "step": 41895 }, { "epoch": 0.98, "learning_rate": 2.2804674849217338e-05, "loss": 0.3638, "step": 41900 }, { "epoch": 0.98, "learning_rate": 2.2803794794771303e-05, "loss": 0.1273, "step": 41905 }, { "epoch": 0.98, "learning_rate": 2.280291474032527e-05, "loss": 0.075, "step": 41910 }, { "epoch": 0.98, "learning_rate": 2.2802034685879233e-05, "loss": 0.0599, "step": 41915 }, { "epoch": 0.98, "learning_rate": 2.2801154631433198e-05, "loss": 0.1123, "step": 41920 }, { "epoch": 0.98, "learning_rate": 2.2800274576987166e-05, "loss": 0.1702, "step": 41925 }, { "epoch": 0.98, "learning_rate": 2.2799394522541128e-05, "loss": 0.1963, "step": 41930 }, { "epoch": 0.98, "learning_rate": 2.2798514468095093e-05, "loss": 0.1697, "step": 41935 }, { "epoch": 0.98, "learning_rate": 2.2797634413649058e-05, "loss": 0.2768, "step": 41940 }, { "epoch": 0.98, "learning_rate": 2.2796754359203026e-05, "loss": 0.4087, "step": 41945 }, { "epoch": 0.98, "learning_rate": 2.2795874304756988e-05, "loss": 0.4487, "step": 41950 }, { "epoch": 0.98, "learning_rate": 2.2794994250310953e-05, "loss": 0.2339, "step": 41955 }, { "epoch": 0.98, "learning_rate": 2.2794114195864918e-05, "loss": 0.0784, "step": 41960 }, { "epoch": 0.98, "learning_rate": 2.2793234141418883e-05, "loss": 0.0351, "step": 41965 }, { "epoch": 0.98, "learning_rate": 2.2792354086972848e-05, "loss": 0.0563, "step": 41970 }, { "epoch": 0.98, "learning_rate": 2.2791474032526813e-05, "loss": 0.1965, "step": 41975 }, { "epoch": 0.98, "learning_rate": 2.2790593978080778e-05, "loss": 0.1432, "step": 41980 }, { "epoch": 0.98, "learning_rate": 2.2789713923634743e-05, "loss": 0.0872, "step": 41985 }, { "epoch": 0.98, "learning_rate": 2.2788833869188708e-05, "loss": 0.2234, "step": 41990 }, { "epoch": 0.98, "learning_rate": 2.2787953814742673e-05, "loss": 0.3484, "step": 41995 }, { "epoch": 0.98, "learning_rate": 2.2787073760296635e-05, "loss": 0.2838, "step": 42000 }, { "epoch": 0.98, "learning_rate": 2.2786193705850603e-05, "loss": 0.2451, "step": 42005 }, { "epoch": 0.98, "learning_rate": 2.2785313651404568e-05, "loss": 0.088, "step": 42010 }, { "epoch": 0.98, "learning_rate": 2.2784433596958533e-05, "loss": 0.096, "step": 42015 }, { "epoch": 0.98, "learning_rate": 2.2783553542512495e-05, "loss": 0.1329, "step": 42020 }, { "epoch": 0.98, "learning_rate": 2.2782673488066463e-05, "loss": 0.0625, "step": 42025 }, { "epoch": 0.98, "learning_rate": 2.2781793433620428e-05, "loss": 0.0867, "step": 42030 }, { "epoch": 0.98, "learning_rate": 2.278091337917439e-05, "loss": 0.0868, "step": 42035 }, { "epoch": 0.98, "learning_rate": 2.2780033324728355e-05, "loss": 0.2774, "step": 42040 }, { "epoch": 0.98, "learning_rate": 2.2779153270282323e-05, "loss": 0.4186, "step": 42045 }, { "epoch": 0.98, "learning_rate": 2.277827321583629e-05, "loss": 0.4362, "step": 42050 }, { "epoch": 0.98, "learning_rate": 2.277739316139025e-05, "loss": 0.1533, "step": 42055 }, { "epoch": 0.98, "learning_rate": 2.277651310694422e-05, "loss": 0.089, "step": 42060 }, { "epoch": 0.98, "learning_rate": 2.2775633052498183e-05, "loss": 0.0874, "step": 42065 }, { "epoch": 0.98, "learning_rate": 2.2774752998052145e-05, "loss": 0.104, "step": 42070 }, { "epoch": 0.98, "learning_rate": 2.277387294360611e-05, "loss": 0.1496, "step": 42075 }, { "epoch": 0.98, "learning_rate": 2.277299288916008e-05, "loss": 0.1547, "step": 42080 }, { "epoch": 0.98, "learning_rate": 2.2772112834714043e-05, "loss": 0.1809, "step": 42085 }, { "epoch": 0.98, "learning_rate": 2.2771232780268005e-05, "loss": 0.1874, "step": 42090 }, { "epoch": 0.98, "learning_rate": 2.277035272582197e-05, "loss": 0.1481, "step": 42095 }, { "epoch": 0.98, "learning_rate": 2.276947267137594e-05, "loss": 0.7153, "step": 42100 }, { "epoch": 0.98, "learning_rate": 2.27685926169299e-05, "loss": 0.1954, "step": 42105 }, { "epoch": 0.98, "learning_rate": 2.2767712562483865e-05, "loss": 0.0147, "step": 42110 }, { "epoch": 0.98, "learning_rate": 2.276683250803783e-05, "loss": 0.0804, "step": 42115 }, { "epoch": 0.98, "learning_rate": 2.27659524535918e-05, "loss": 0.0368, "step": 42120 }, { "epoch": 0.98, "learning_rate": 2.276507239914576e-05, "loss": 0.1169, "step": 42125 }, { "epoch": 0.98, "learning_rate": 2.2764192344699725e-05, "loss": 0.151, "step": 42130 }, { "epoch": 0.98, "learning_rate": 2.276331229025369e-05, "loss": 0.1685, "step": 42135 }, { "epoch": 0.98, "learning_rate": 2.276243223580766e-05, "loss": 0.2997, "step": 42140 }, { "epoch": 0.98, "learning_rate": 2.276155218136162e-05, "loss": 0.3948, "step": 42145 }, { "epoch": 0.98, "learning_rate": 2.2760672126915585e-05, "loss": 0.2703, "step": 42150 }, { "epoch": 0.98, "learning_rate": 2.275979207246955e-05, "loss": 0.1694, "step": 42155 }, { "epoch": 0.98, "learning_rate": 2.2758912018023515e-05, "loss": 0.0355, "step": 42160 }, { "epoch": 0.98, "learning_rate": 2.275803196357748e-05, "loss": 0.0571, "step": 42165 }, { "epoch": 0.98, "learning_rate": 2.2757151909131445e-05, "loss": 0.1097, "step": 42170 }, { "epoch": 0.98, "learning_rate": 2.275627185468541e-05, "loss": 0.1129, "step": 42175 }, { "epoch": 0.98, "learning_rate": 2.2755391800239375e-05, "loss": 0.1491, "step": 42180 }, { "epoch": 0.98, "learning_rate": 2.275451174579334e-05, "loss": 0.3109, "step": 42185 }, { "epoch": 0.98, "learning_rate": 2.2753631691347306e-05, "loss": 0.1408, "step": 42190 }, { "epoch": 0.98, "learning_rate": 2.275275163690127e-05, "loss": 0.3294, "step": 42195 }, { "epoch": 0.98, "learning_rate": 2.2751871582455236e-05, "loss": 0.5891, "step": 42200 }, { "epoch": 0.98, "learning_rate": 2.27509915280092e-05, "loss": 0.1735, "step": 42205 }, { "epoch": 0.98, "learning_rate": 2.2750111473563166e-05, "loss": 0.0499, "step": 42210 }, { "epoch": 0.98, "learning_rate": 2.274923141911713e-05, "loss": 0.0553, "step": 42215 }, { "epoch": 0.99, "learning_rate": 2.2748351364671096e-05, "loss": 0.0434, "step": 42220 }, { "epoch": 0.99, "learning_rate": 2.274747131022506e-05, "loss": 0.1784, "step": 42225 }, { "epoch": 0.99, "learning_rate": 2.2746591255779022e-05, "loss": 0.1153, "step": 42230 }, { "epoch": 0.99, "learning_rate": 2.274571120133299e-05, "loss": 0.1596, "step": 42235 }, { "epoch": 0.99, "learning_rate": 2.2744831146886956e-05, "loss": 0.4432, "step": 42240 }, { "epoch": 0.99, "learning_rate": 2.274395109244092e-05, "loss": 0.2833, "step": 42245 }, { "epoch": 0.99, "learning_rate": 2.2743071037994882e-05, "loss": 0.4861, "step": 42250 }, { "epoch": 0.99, "learning_rate": 2.274219098354885e-05, "loss": 0.2087, "step": 42255 }, { "epoch": 0.99, "learning_rate": 2.2741310929102816e-05, "loss": 0.0687, "step": 42260 }, { "epoch": 0.99, "learning_rate": 2.2740430874656777e-05, "loss": 0.15, "step": 42265 }, { "epoch": 0.99, "learning_rate": 2.2739550820210742e-05, "loss": 0.1757, "step": 42270 }, { "epoch": 0.99, "learning_rate": 2.273867076576471e-05, "loss": 0.121, "step": 42275 }, { "epoch": 0.99, "learning_rate": 2.2737790711318676e-05, "loss": 0.1856, "step": 42280 }, { "epoch": 0.99, "learning_rate": 2.2736910656872638e-05, "loss": 0.2021, "step": 42285 }, { "epoch": 0.99, "learning_rate": 2.2736030602426603e-05, "loss": 0.1383, "step": 42290 }, { "epoch": 0.99, "learning_rate": 2.273515054798057e-05, "loss": 0.1229, "step": 42295 }, { "epoch": 0.99, "learning_rate": 2.2734270493534533e-05, "loss": 0.3256, "step": 42300 }, { "epoch": 0.99, "learning_rate": 2.2733390439088498e-05, "loss": 0.2384, "step": 42305 }, { "epoch": 0.99, "learning_rate": 2.2732510384642463e-05, "loss": 0.0428, "step": 42310 }, { "epoch": 0.99, "learning_rate": 2.273163033019643e-05, "loss": 0.0173, "step": 42315 }, { "epoch": 0.99, "learning_rate": 2.2730750275750393e-05, "loss": 0.0192, "step": 42320 }, { "epoch": 0.99, "learning_rate": 2.2729870221304358e-05, "loss": 0.0775, "step": 42325 }, { "epoch": 0.99, "learning_rate": 2.2728990166858326e-05, "loss": 0.1312, "step": 42330 }, { "epoch": 0.99, "learning_rate": 2.2728110112412288e-05, "loss": 0.2924, "step": 42335 }, { "epoch": 0.99, "learning_rate": 2.2727230057966253e-05, "loss": 0.1417, "step": 42340 }, { "epoch": 0.99, "learning_rate": 2.2726350003520218e-05, "loss": 0.2224, "step": 42345 }, { "epoch": 0.99, "learning_rate": 2.2725469949074186e-05, "loss": 0.609, "step": 42350 }, { "epoch": 0.99, "learning_rate": 2.2724589894628148e-05, "loss": 0.2561, "step": 42355 }, { "epoch": 0.99, "learning_rate": 2.2723709840182113e-05, "loss": 0.0916, "step": 42360 }, { "epoch": 0.99, "learning_rate": 2.2722829785736078e-05, "loss": 0.0243, "step": 42365 }, { "epoch": 0.99, "learning_rate": 2.2721949731290043e-05, "loss": 0.1373, "step": 42370 }, { "epoch": 0.99, "learning_rate": 2.2721069676844008e-05, "loss": 0.0845, "step": 42375 }, { "epoch": 0.99, "learning_rate": 2.2720189622397973e-05, "loss": 0.2089, "step": 42380 }, { "epoch": 0.99, "learning_rate": 2.2719309567951938e-05, "loss": 0.2528, "step": 42385 }, { "epoch": 0.99, "learning_rate": 2.2718429513505903e-05, "loss": 0.1589, "step": 42390 }, { "epoch": 0.99, "learning_rate": 2.2717549459059868e-05, "loss": 0.2931, "step": 42395 }, { "epoch": 0.99, "learning_rate": 2.2716669404613833e-05, "loss": 0.3948, "step": 42400 }, { "epoch": 0.99, "learning_rate": 2.2715789350167795e-05, "loss": 0.1225, "step": 42405 }, { "epoch": 0.99, "learning_rate": 2.2714909295721763e-05, "loss": 0.0264, "step": 42410 }, { "epoch": 0.99, "learning_rate": 2.2714029241275728e-05, "loss": 0.0415, "step": 42415 }, { "epoch": 0.99, "learning_rate": 2.2713149186829693e-05, "loss": 0.1386, "step": 42420 }, { "epoch": 0.99, "learning_rate": 2.2712269132383655e-05, "loss": 0.3408, "step": 42425 }, { "epoch": 0.99, "learning_rate": 2.2711389077937623e-05, "loss": 0.0969, "step": 42430 }, { "epoch": 0.99, "learning_rate": 2.2710509023491588e-05, "loss": 0.047, "step": 42435 }, { "epoch": 0.99, "learning_rate": 2.2709628969045553e-05, "loss": 0.1286, "step": 42440 }, { "epoch": 0.99, "learning_rate": 2.2708748914599515e-05, "loss": 0.2204, "step": 42445 }, { "epoch": 0.99, "learning_rate": 2.2707868860153483e-05, "loss": 0.624, "step": 42450 }, { "epoch": 0.99, "learning_rate": 2.2706988805707448e-05, "loss": 0.2681, "step": 42455 }, { "epoch": 0.99, "learning_rate": 2.270610875126141e-05, "loss": 0.1226, "step": 42460 }, { "epoch": 0.99, "learning_rate": 2.2705228696815378e-05, "loss": 0.0724, "step": 42465 }, { "epoch": 0.99, "learning_rate": 2.2704348642369343e-05, "loss": 0.0821, "step": 42470 }, { "epoch": 0.99, "learning_rate": 2.270346858792331e-05, "loss": 0.1943, "step": 42475 }, { "epoch": 0.99, "learning_rate": 2.270258853347727e-05, "loss": 0.1203, "step": 42480 }, { "epoch": 0.99, "learning_rate": 2.270170847903124e-05, "loss": 0.1831, "step": 42485 }, { "epoch": 0.99, "learning_rate": 2.2700828424585203e-05, "loss": 0.3127, "step": 42490 }, { "epoch": 0.99, "learning_rate": 2.2699948370139165e-05, "loss": 0.2596, "step": 42495 }, { "epoch": 0.99, "learning_rate": 2.269906831569313e-05, "loss": 0.7046, "step": 42500 }, { "epoch": 0.99, "learning_rate": 2.26981882612471e-05, "loss": 0.156, "step": 42505 }, { "epoch": 0.99, "learning_rate": 2.2697308206801063e-05, "loss": 0.2111, "step": 42510 }, { "epoch": 0.99, "learning_rate": 2.2696428152355025e-05, "loss": 0.1445, "step": 42515 }, { "epoch": 0.99, "learning_rate": 2.269554809790899e-05, "loss": 0.1409, "step": 42520 }, { "epoch": 0.99, "learning_rate": 2.269466804346296e-05, "loss": 0.1773, "step": 42525 }, { "epoch": 0.99, "learning_rate": 2.269378798901692e-05, "loss": 0.1407, "step": 42530 }, { "epoch": 0.99, "learning_rate": 2.2692907934570885e-05, "loss": 0.0741, "step": 42535 }, { "epoch": 0.99, "learning_rate": 2.269202788012485e-05, "loss": 0.1436, "step": 42540 }, { "epoch": 0.99, "learning_rate": 2.269114782567882e-05, "loss": 0.2247, "step": 42545 }, { "epoch": 0.99, "learning_rate": 2.269026777123278e-05, "loss": 0.3672, "step": 42550 }, { "epoch": 0.99, "learning_rate": 2.2689387716786745e-05, "loss": 0.2669, "step": 42555 }, { "epoch": 0.99, "learning_rate": 2.268850766234071e-05, "loss": 0.0273, "step": 42560 }, { "epoch": 0.99, "learning_rate": 2.2687627607894675e-05, "loss": 0.1085, "step": 42565 }, { "epoch": 0.99, "learning_rate": 2.268674755344864e-05, "loss": 0.0576, "step": 42570 }, { "epoch": 0.99, "learning_rate": 2.2685867499002605e-05, "loss": 0.0628, "step": 42575 }, { "epoch": 0.99, "learning_rate": 2.268498744455657e-05, "loss": 0.12, "step": 42580 }, { "epoch": 0.99, "learning_rate": 2.2684107390110535e-05, "loss": 0.2592, "step": 42585 }, { "epoch": 0.99, "learning_rate": 2.26832273356645e-05, "loss": 0.3009, "step": 42590 }, { "epoch": 0.99, "learning_rate": 2.2682347281218465e-05, "loss": 0.2958, "step": 42595 }, { "epoch": 0.99, "learning_rate": 2.268146722677243e-05, "loss": 0.3972, "step": 42600 }, { "epoch": 0.99, "learning_rate": 2.2680587172326395e-05, "loss": 0.1502, "step": 42605 }, { "epoch": 0.99, "learning_rate": 2.267970711788036e-05, "loss": 0.1087, "step": 42610 }, { "epoch": 0.99, "learning_rate": 2.2678827063434326e-05, "loss": 0.0654, "step": 42615 }, { "epoch": 0.99, "learning_rate": 2.267794700898829e-05, "loss": 0.1656, "step": 42620 }, { "epoch": 0.99, "learning_rate": 2.2677066954542256e-05, "loss": 0.0972, "step": 42625 }, { "epoch": 0.99, "learning_rate": 2.267618690009622e-05, "loss": 0.2258, "step": 42630 }, { "epoch": 0.99, "learning_rate": 2.2675306845650182e-05, "loss": 0.2555, "step": 42635 }, { "epoch": 0.99, "learning_rate": 2.267442679120415e-05, "loss": 0.0893, "step": 42640 }, { "epoch": 0.99, "learning_rate": 2.2673546736758116e-05, "loss": 0.4389, "step": 42645 }, { "epoch": 1.0, "learning_rate": 2.267266668231208e-05, "loss": 0.4497, "step": 42650 }, { "epoch": 1.0, "learning_rate": 2.2671786627866042e-05, "loss": 0.1544, "step": 42655 }, { "epoch": 1.0, "learning_rate": 2.267090657342001e-05, "loss": 0.0411, "step": 42660 }, { "epoch": 1.0, "learning_rate": 2.2670026518973976e-05, "loss": 0.0413, "step": 42665 }, { "epoch": 1.0, "learning_rate": 2.2669146464527937e-05, "loss": 0.0965, "step": 42670 }, { "epoch": 1.0, "learning_rate": 2.2668266410081902e-05, "loss": 0.0627, "step": 42675 }, { "epoch": 1.0, "learning_rate": 2.266738635563587e-05, "loss": 0.1185, "step": 42680 }, { "epoch": 1.0, "learning_rate": 2.2666506301189836e-05, "loss": 0.2614, "step": 42685 }, { "epoch": 1.0, "learning_rate": 2.2665626246743797e-05, "loss": 0.1448, "step": 42690 }, { "epoch": 1.0, "learning_rate": 2.2664746192297762e-05, "loss": 0.5882, "step": 42695 }, { "epoch": 1.0, "learning_rate": 2.266386613785173e-05, "loss": 0.6964, "step": 42700 }, { "epoch": 1.0, "learning_rate": 2.2662986083405692e-05, "loss": 0.1039, "step": 42705 }, { "epoch": 1.0, "learning_rate": 2.2662106028959658e-05, "loss": 0.0636, "step": 42710 }, { "epoch": 1.0, "learning_rate": 2.2661225974513623e-05, "loss": 0.1184, "step": 42715 }, { "epoch": 1.0, "learning_rate": 2.266034592006759e-05, "loss": 0.1542, "step": 42720 }, { "epoch": 1.0, "learning_rate": 2.2659465865621553e-05, "loss": 0.2038, "step": 42725 }, { "epoch": 1.0, "learning_rate": 2.2658585811175518e-05, "loss": 0.1888, "step": 42730 }, { "epoch": 1.0, "learning_rate": 2.2657705756729483e-05, "loss": 0.2468, "step": 42735 }, { "epoch": 1.0, "learning_rate": 2.2656825702283448e-05, "loss": 0.1633, "step": 42740 }, { "epoch": 1.0, "learning_rate": 2.2655945647837413e-05, "loss": 0.3318, "step": 42745 }, { "epoch": 1.0, "learning_rate": 2.2655065593391378e-05, "loss": 0.642, "step": 42750 }, { "epoch": 1.0, "learning_rate": 2.2654185538945346e-05, "loss": 0.2333, "step": 42755 }, { "epoch": 1.0, "learning_rate": 2.2653305484499308e-05, "loss": 0.109, "step": 42760 }, { "epoch": 1.0, "learning_rate": 2.2652425430053273e-05, "loss": 0.0341, "step": 42765 }, { "epoch": 1.0, "learning_rate": 2.2651545375607238e-05, "loss": 0.0941, "step": 42770 }, { "epoch": 1.0, "learning_rate": 2.2650665321161206e-05, "loss": 0.1742, "step": 42775 }, { "epoch": 1.0, "learning_rate": 2.2649785266715168e-05, "loss": 0.0966, "step": 42780 }, { "epoch": 1.0, "learning_rate": 2.2648905212269133e-05, "loss": 0.2224, "step": 42785 }, { "epoch": 1.0, "learning_rate": 2.2648025157823098e-05, "loss": 0.1477, "step": 42790 }, { "epoch": 1.0, "learning_rate": 2.2647145103377063e-05, "loss": 0.333, "step": 42795 }, { "epoch": 1.0, "learning_rate": 2.2646265048931028e-05, "loss": 0.5462, "step": 42800 }, { "epoch": 1.0, "learning_rate": 2.2645384994484993e-05, "loss": 0.1879, "step": 42805 }, { "epoch": 1.0, "learning_rate": 2.2644504940038958e-05, "loss": 0.1119, "step": 42810 }, { "epoch": 1.0, "learning_rate": 2.2643624885592923e-05, "loss": 0.0781, "step": 42815 }, { "epoch": 1.0, "learning_rate": 2.2642744831146888e-05, "loss": 0.1429, "step": 42820 }, { "epoch": 1.0, "learning_rate": 2.2641864776700853e-05, "loss": 0.0614, "step": 42825 }, { "epoch": 1.0, "learning_rate": 2.2640984722254815e-05, "loss": 0.1426, "step": 42830 }, { "epoch": 1.0, "learning_rate": 2.2640104667808783e-05, "loss": 0.174, "step": 42835 }, { "epoch": 1.0, "learning_rate": 2.2639224613362748e-05, "loss": 0.2794, "step": 42840 }, { "epoch": 1.0, "learning_rate": 2.2638344558916713e-05, "loss": 0.3107, "step": 42845 }, { "epoch": 1.0, "learning_rate": 2.2637464504470675e-05, "loss": 0.431, "step": 42850 }, { "epoch": 1.0, "learning_rate": 2.2636584450024643e-05, "loss": 0.1203, "step": 42855 }, { "epoch": 1.0, "learning_rate": 2.2635880406467814e-05, "loss": 0.1393, "step": 42860 }, { "epoch": 1.0, "eval_cer": 0.01952390277670248, "eval_loss": 0.04072042554616928, "eval_runtime": 711.8193, "eval_samples_per_second": 26.762, "eval_steps_per_second": 6.691, "eval_wer": 0.16809373591707977, "step": 42861 }, { "epoch": 1.0, "learning_rate": 2.2635000352021776e-05, "loss": 0.2734, "step": 42865 }, { "epoch": 1.0, "learning_rate": 2.2634120297575744e-05, "loss": 0.1571, "step": 42870 }, { "epoch": 1.0, "learning_rate": 2.263324024312971e-05, "loss": 0.034, "step": 42875 }, { "epoch": 1.0, "learning_rate": 2.2632360188683675e-05, "loss": 0.1484, "step": 42880 }, { "epoch": 1.0, "learning_rate": 2.2631480134237636e-05, "loss": 0.0894, "step": 42885 }, { "epoch": 1.0, "learning_rate": 2.2630600079791605e-05, "loss": 0.1307, "step": 42890 }, { "epoch": 1.0, "learning_rate": 2.262972002534557e-05, "loss": 0.3143, "step": 42895 }, { "epoch": 1.0, "learning_rate": 2.262883997089953e-05, "loss": 0.1825, "step": 42900 }, { "epoch": 1.0, "learning_rate": 2.2627959916453496e-05, "loss": 0.2353, "step": 42905 }, { "epoch": 1.0, "learning_rate": 2.2627079862007465e-05, "loss": 0.401, "step": 42910 }, { "epoch": 1.0, "learning_rate": 2.262619980756143e-05, "loss": 0.4199, "step": 42915 }, { "epoch": 1.0, "learning_rate": 2.262531975311539e-05, "loss": 0.0642, "step": 42920 }, { "epoch": 1.0, "learning_rate": 2.262443969866936e-05, "loss": 0.0636, "step": 42925 }, { "epoch": 1.0, "learning_rate": 2.2623559644223325e-05, "loss": 0.0402, "step": 42930 }, { "epoch": 1.0, "learning_rate": 2.262267958977729e-05, "loss": 0.154, "step": 42935 }, { "epoch": 1.0, "learning_rate": 2.262179953533125e-05, "loss": 0.1463, "step": 42940 }, { "epoch": 1.0, "learning_rate": 2.262091948088522e-05, "loss": 0.1827, "step": 42945 }, { "epoch": 1.0, "learning_rate": 2.2620039426439185e-05, "loss": 0.1779, "step": 42950 }, { "epoch": 1.0, "learning_rate": 2.2619159371993146e-05, "loss": 0.1902, "step": 42955 }, { "epoch": 1.0, "learning_rate": 2.261827931754711e-05, "loss": 0.297, "step": 42960 }, { "epoch": 1.0, "learning_rate": 2.261739926310108e-05, "loss": 0.2729, "step": 42965 }, { "epoch": 1.0, "learning_rate": 2.2616519208655045e-05, "loss": 0.0649, "step": 42970 }, { "epoch": 1.0, "learning_rate": 2.2615639154209007e-05, "loss": 0.2292, "step": 42975 }, { "epoch": 1.0, "learning_rate": 2.261475909976297e-05, "loss": 0.2433, "step": 42980 }, { "epoch": 1.0, "learning_rate": 2.261387904531694e-05, "loss": 0.2034, "step": 42985 }, { "epoch": 1.0, "learning_rate": 2.26129989908709e-05, "loss": 0.1298, "step": 42990 }, { "epoch": 1.0, "learning_rate": 2.2612118936424867e-05, "loss": 0.2022, "step": 42995 }, { "epoch": 1.0, "learning_rate": 2.261123888197883e-05, "loss": 0.1225, "step": 43000 }, { "epoch": 1.0, "learning_rate": 2.26103588275328e-05, "loss": 0.2355, "step": 43005 }, { "epoch": 1.0, "learning_rate": 2.260947877308676e-05, "loss": 0.2608, "step": 43010 }, { "epoch": 1.0, "learning_rate": 2.2608598718640727e-05, "loss": 0.2148, "step": 43015 }, { "epoch": 1.0, "learning_rate": 2.2607718664194692e-05, "loss": 0.0555, "step": 43020 }, { "epoch": 1.0, "learning_rate": 2.2606838609748657e-05, "loss": 0.1463, "step": 43025 }, { "epoch": 1.0, "learning_rate": 2.2605958555302622e-05, "loss": 0.1213, "step": 43030 }, { "epoch": 1.0, "learning_rate": 2.2605078500856587e-05, "loss": 0.1466, "step": 43035 }, { "epoch": 1.0, "learning_rate": 2.2604198446410552e-05, "loss": 0.2185, "step": 43040 }, { "epoch": 1.0, "learning_rate": 2.2603318391964517e-05, "loss": 0.2258, "step": 43045 }, { "epoch": 1.0, "learning_rate": 2.2602438337518482e-05, "loss": 0.1866, "step": 43050 }, { "epoch": 1.0, "learning_rate": 2.2601558283072447e-05, "loss": 0.1486, "step": 43055 }, { "epoch": 1.0, "learning_rate": 2.2600678228626412e-05, "loss": 0.3067, "step": 43060 }, { "epoch": 1.0, "learning_rate": 2.2599798174180377e-05, "loss": 0.3738, "step": 43065 }, { "epoch": 1.0, "learning_rate": 2.2598918119734342e-05, "loss": 0.0633, "step": 43070 }, { "epoch": 1.0, "learning_rate": 2.2598038065288307e-05, "loss": 0.0434, "step": 43075 }, { "epoch": 1.01, "learning_rate": 2.2597158010842272e-05, "loss": 0.1929, "step": 43080 }, { "epoch": 1.01, "learning_rate": 2.2596277956396237e-05, "loss": 0.128, "step": 43085 }, { "epoch": 1.01, "learning_rate": 2.2595397901950202e-05, "loss": 0.1496, "step": 43090 }, { "epoch": 1.01, "learning_rate": 2.2594517847504164e-05, "loss": 0.1671, "step": 43095 }, { "epoch": 1.01, "learning_rate": 2.2593637793058132e-05, "loss": 0.2719, "step": 43100 }, { "epoch": 1.01, "learning_rate": 2.2592757738612097e-05, "loss": 0.273, "step": 43105 }, { "epoch": 1.01, "learning_rate": 2.2591877684166062e-05, "loss": 0.5096, "step": 43110 }, { "epoch": 1.01, "learning_rate": 2.2590997629720024e-05, "loss": 0.8002, "step": 43115 }, { "epoch": 1.01, "learning_rate": 2.2590117575273992e-05, "loss": 0.0924, "step": 43120 }, { "epoch": 1.01, "learning_rate": 2.2589237520827957e-05, "loss": 0.033, "step": 43125 }, { "epoch": 1.01, "learning_rate": 2.258835746638192e-05, "loss": 0.2384, "step": 43130 }, { "epoch": 1.01, "learning_rate": 2.2587477411935884e-05, "loss": 0.2094, "step": 43135 }, { "epoch": 1.01, "learning_rate": 2.2586597357489852e-05, "loss": 0.1238, "step": 43140 }, { "epoch": 1.01, "learning_rate": 2.2585717303043817e-05, "loss": 0.2061, "step": 43145 }, { "epoch": 1.01, "learning_rate": 2.258483724859778e-05, "loss": 0.1037, "step": 43150 }, { "epoch": 1.01, "learning_rate": 2.2583957194151744e-05, "loss": 0.1603, "step": 43155 }, { "epoch": 1.01, "learning_rate": 2.2583077139705712e-05, "loss": 0.3428, "step": 43160 }, { "epoch": 1.01, "learning_rate": 2.2582197085259674e-05, "loss": 0.3429, "step": 43165 }, { "epoch": 1.01, "learning_rate": 2.258131703081364e-05, "loss": 0.1705, "step": 43170 }, { "epoch": 1.01, "learning_rate": 2.2580436976367604e-05, "loss": 0.1113, "step": 43175 }, { "epoch": 1.01, "learning_rate": 2.2579556921921572e-05, "loss": 0.1047, "step": 43180 }, { "epoch": 1.01, "learning_rate": 2.2578676867475534e-05, "loss": 0.133, "step": 43185 }, { "epoch": 1.01, "learning_rate": 2.25777968130295e-05, "loss": 0.1239, "step": 43190 }, { "epoch": 1.01, "learning_rate": 2.2576916758583464e-05, "loss": 0.2649, "step": 43195 }, { "epoch": 1.01, "learning_rate": 2.257603670413743e-05, "loss": 0.2973, "step": 43200 }, { "epoch": 1.01, "learning_rate": 2.2575156649691394e-05, "loss": 0.3007, "step": 43205 }, { "epoch": 1.01, "learning_rate": 2.257427659524536e-05, "loss": 0.2316, "step": 43210 }, { "epoch": 1.01, "learning_rate": 2.2573396540799328e-05, "loss": 0.2773, "step": 43215 }, { "epoch": 1.01, "learning_rate": 2.257251648635329e-05, "loss": 0.0774, "step": 43220 }, { "epoch": 1.01, "learning_rate": 2.2571636431907254e-05, "loss": 0.0565, "step": 43225 }, { "epoch": 1.01, "learning_rate": 2.257075637746122e-05, "loss": 0.0691, "step": 43230 }, { "epoch": 1.01, "learning_rate": 2.2569876323015188e-05, "loss": 0.244, "step": 43235 }, { "epoch": 1.01, "learning_rate": 2.256899626856915e-05, "loss": 0.1385, "step": 43240 }, { "epoch": 1.01, "learning_rate": 2.2568116214123114e-05, "loss": 0.1391, "step": 43245 }, { "epoch": 1.01, "learning_rate": 2.256723615967708e-05, "loss": 0.1426, "step": 43250 }, { "epoch": 1.01, "learning_rate": 2.2566356105231044e-05, "loss": 0.2848, "step": 43255 }, { "epoch": 1.01, "learning_rate": 2.256547605078501e-05, "loss": 0.341, "step": 43260 }, { "epoch": 1.01, "learning_rate": 2.2564595996338974e-05, "loss": 0.3183, "step": 43265 }, { "epoch": 1.01, "learning_rate": 2.256371594189294e-05, "loss": 0.065, "step": 43270 }, { "epoch": 1.01, "learning_rate": 2.2562835887446904e-05, "loss": 0.0521, "step": 43275 }, { "epoch": 1.01, "learning_rate": 2.256195583300087e-05, "loss": 0.1402, "step": 43280 }, { "epoch": 1.01, "learning_rate": 2.2561075778554834e-05, "loss": 0.085, "step": 43285 }, { "epoch": 1.01, "learning_rate": 2.2560195724108796e-05, "loss": 0.2593, "step": 43290 }, { "epoch": 1.01, "learning_rate": 2.2559315669662764e-05, "loss": 0.1295, "step": 43295 }, { "epoch": 1.01, "learning_rate": 2.255843561521673e-05, "loss": 0.1103, "step": 43300 }, { "epoch": 1.01, "learning_rate": 2.2557555560770694e-05, "loss": 0.337, "step": 43305 }, { "epoch": 1.01, "learning_rate": 2.2556675506324656e-05, "loss": 0.4822, "step": 43310 }, { "epoch": 1.01, "learning_rate": 2.2555795451878625e-05, "loss": 0.5319, "step": 43315 }, { "epoch": 1.01, "learning_rate": 2.255491539743259e-05, "loss": 0.0616, "step": 43320 }, { "epoch": 1.01, "learning_rate": 2.255403534298655e-05, "loss": 0.1395, "step": 43325 }, { "epoch": 1.01, "learning_rate": 2.2553155288540516e-05, "loss": 0.1211, "step": 43330 }, { "epoch": 1.01, "learning_rate": 2.2552275234094485e-05, "loss": 0.0906, "step": 43335 }, { "epoch": 1.01, "learning_rate": 2.255139517964845e-05, "loss": 0.098, "step": 43340 }, { "epoch": 1.01, "learning_rate": 2.255051512520241e-05, "loss": 0.3416, "step": 43345 }, { "epoch": 1.01, "learning_rate": 2.254963507075638e-05, "loss": 0.0982, "step": 43350 }, { "epoch": 1.01, "learning_rate": 2.2548755016310345e-05, "loss": 0.3911, "step": 43355 }, { "epoch": 1.01, "learning_rate": 2.2547874961864306e-05, "loss": 0.5227, "step": 43360 }, { "epoch": 1.01, "learning_rate": 2.254699490741827e-05, "loss": 0.2869, "step": 43365 }, { "epoch": 1.01, "learning_rate": 2.254611485297224e-05, "loss": 0.0334, "step": 43370 }, { "epoch": 1.01, "learning_rate": 2.2545234798526205e-05, "loss": 0.0305, "step": 43375 }, { "epoch": 1.01, "learning_rate": 2.2544354744080166e-05, "loss": 0.1241, "step": 43380 }, { "epoch": 1.01, "learning_rate": 2.254347468963413e-05, "loss": 0.2046, "step": 43385 }, { "epoch": 1.01, "learning_rate": 2.25425946351881e-05, "loss": 0.1558, "step": 43390 }, { "epoch": 1.01, "learning_rate": 2.254171458074206e-05, "loss": 0.0993, "step": 43395 }, { "epoch": 1.01, "learning_rate": 2.2540834526296026e-05, "loss": 0.3601, "step": 43400 }, { "epoch": 1.01, "learning_rate": 2.253995447184999e-05, "loss": 0.3337, "step": 43405 }, { "epoch": 1.01, "learning_rate": 2.253907441740396e-05, "loss": 0.3882, "step": 43410 }, { "epoch": 1.01, "learning_rate": 2.253819436295792e-05, "loss": 0.2617, "step": 43415 }, { "epoch": 1.01, "learning_rate": 2.2537314308511887e-05, "loss": 0.0163, "step": 43420 }, { "epoch": 1.01, "learning_rate": 2.253643425406585e-05, "loss": 0.0378, "step": 43425 }, { "epoch": 1.01, "learning_rate": 2.2535554199619817e-05, "loss": 0.0537, "step": 43430 }, { "epoch": 1.01, "learning_rate": 2.253467414517378e-05, "loss": 0.1571, "step": 43435 }, { "epoch": 1.01, "learning_rate": 2.2533794090727747e-05, "loss": 0.1566, "step": 43440 }, { "epoch": 1.01, "learning_rate": 2.253291403628171e-05, "loss": 0.1898, "step": 43445 }, { "epoch": 1.01, "learning_rate": 2.2532033981835677e-05, "loss": 0.199, "step": 43450 }, { "epoch": 1.01, "learning_rate": 2.2531153927389642e-05, "loss": 0.3163, "step": 43455 }, { "epoch": 1.01, "learning_rate": 2.2530273872943607e-05, "loss": 0.1374, "step": 43460 }, { "epoch": 1.01, "learning_rate": 2.2529393818497572e-05, "loss": 0.2531, "step": 43465 }, { "epoch": 1.01, "learning_rate": 2.2528513764051537e-05, "loss": 0.0486, "step": 43470 }, { "epoch": 1.01, "learning_rate": 2.2527633709605502e-05, "loss": 0.0768, "step": 43475 }, { "epoch": 1.01, "learning_rate": 2.2526753655159467e-05, "loss": 0.0279, "step": 43480 }, { "epoch": 1.01, "learning_rate": 2.2525873600713432e-05, "loss": 0.1536, "step": 43485 }, { "epoch": 1.01, "learning_rate": 2.2524993546267397e-05, "loss": 0.2104, "step": 43490 }, { "epoch": 1.01, "learning_rate": 2.2524113491821362e-05, "loss": 0.1343, "step": 43495 }, { "epoch": 1.01, "learning_rate": 2.2523233437375324e-05, "loss": 0.2068, "step": 43500 }, { "epoch": 1.02, "learning_rate": 2.2522353382929292e-05, "loss": 0.2678, "step": 43505 }, { "epoch": 1.02, "learning_rate": 2.2521473328483257e-05, "loss": 0.3492, "step": 43510 }, { "epoch": 1.02, "learning_rate": 2.2520593274037222e-05, "loss": 0.4179, "step": 43515 }, { "epoch": 1.02, "learning_rate": 2.2519713219591184e-05, "loss": 0.0795, "step": 43520 }, { "epoch": 1.02, "learning_rate": 2.2518833165145152e-05, "loss": 0.1535, "step": 43525 }, { "epoch": 1.02, "learning_rate": 2.2517953110699117e-05, "loss": 0.0562, "step": 43530 }, { "epoch": 1.02, "learning_rate": 2.251707305625308e-05, "loss": 0.2143, "step": 43535 }, { "epoch": 1.02, "learning_rate": 2.2516193001807044e-05, "loss": 0.1129, "step": 43540 }, { "epoch": 1.02, "learning_rate": 2.2515312947361012e-05, "loss": 0.2903, "step": 43545 }, { "epoch": 1.02, "learning_rate": 2.2514432892914977e-05, "loss": 0.1557, "step": 43550 }, { "epoch": 1.02, "learning_rate": 2.251355283846894e-05, "loss": 0.3776, "step": 43555 }, { "epoch": 1.02, "learning_rate": 2.2512672784022904e-05, "loss": 0.4365, "step": 43560 }, { "epoch": 1.02, "learning_rate": 2.2511792729576872e-05, "loss": 0.3578, "step": 43565 }, { "epoch": 1.02, "learning_rate": 2.2510912675130837e-05, "loss": 0.055, "step": 43570 }, { "epoch": 1.02, "learning_rate": 2.25100326206848e-05, "loss": 0.1087, "step": 43575 }, { "epoch": 1.02, "learning_rate": 2.2509152566238764e-05, "loss": 0.1237, "step": 43580 }, { "epoch": 1.02, "learning_rate": 2.2508272511792732e-05, "loss": 0.0523, "step": 43585 }, { "epoch": 1.02, "learning_rate": 2.2507392457346694e-05, "loss": 0.1177, "step": 43590 }, { "epoch": 1.02, "learning_rate": 2.250651240290066e-05, "loss": 0.2119, "step": 43595 }, { "epoch": 1.02, "learning_rate": 2.2505632348454624e-05, "loss": 0.1847, "step": 43600 }, { "epoch": 1.02, "learning_rate": 2.2504752294008592e-05, "loss": 0.316, "step": 43605 }, { "epoch": 1.02, "learning_rate": 2.2503872239562554e-05, "loss": 0.3794, "step": 43610 }, { "epoch": 1.02, "learning_rate": 2.250299218511652e-05, "loss": 0.3772, "step": 43615 }, { "epoch": 1.02, "learning_rate": 2.2502112130670487e-05, "loss": 0.0218, "step": 43620 }, { "epoch": 1.02, "learning_rate": 2.250123207622445e-05, "loss": 0.061, "step": 43625 }, { "epoch": 1.02, "learning_rate": 2.2500352021778414e-05, "loss": 0.1085, "step": 43630 }, { "epoch": 1.02, "learning_rate": 2.249947196733238e-05, "loss": 0.1189, "step": 43635 }, { "epoch": 1.02, "learning_rate": 2.2498591912886347e-05, "loss": 0.0944, "step": 43640 }, { "epoch": 1.02, "learning_rate": 2.249771185844031e-05, "loss": 0.147, "step": 43645 }, { "epoch": 1.02, "learning_rate": 2.2496831803994274e-05, "loss": 0.1252, "step": 43650 }, { "epoch": 1.02, "learning_rate": 2.249595174954824e-05, "loss": 0.3502, "step": 43655 }, { "epoch": 1.02, "learning_rate": 2.2495071695102204e-05, "loss": 0.2496, "step": 43660 }, { "epoch": 1.02, "learning_rate": 2.249419164065617e-05, "loss": 0.8042, "step": 43665 }, { "epoch": 1.02, "learning_rate": 2.2493311586210134e-05, "loss": 0.0317, "step": 43670 }, { "epoch": 1.02, "learning_rate": 2.24924315317641e-05, "loss": 0.1238, "step": 43675 }, { "epoch": 1.02, "learning_rate": 2.2491551477318064e-05, "loss": 0.1772, "step": 43680 }, { "epoch": 1.02, "learning_rate": 2.249067142287203e-05, "loss": 0.2532, "step": 43685 }, { "epoch": 1.02, "learning_rate": 2.2489791368425994e-05, "loss": 0.148, "step": 43690 }, { "epoch": 1.02, "learning_rate": 2.2488911313979956e-05, "loss": 0.0902, "step": 43695 }, { "epoch": 1.02, "learning_rate": 2.2488031259533924e-05, "loss": 0.2458, "step": 43700 }, { "epoch": 1.02, "learning_rate": 2.248715120508789e-05, "loss": 0.1354, "step": 43705 }, { "epoch": 1.02, "learning_rate": 2.2486271150641854e-05, "loss": 0.4948, "step": 43710 }, { "epoch": 1.02, "learning_rate": 2.2485391096195816e-05, "loss": 0.2098, "step": 43715 }, { "epoch": 1.02, "learning_rate": 2.2484511041749784e-05, "loss": 0.0172, "step": 43720 }, { "epoch": 1.02, "learning_rate": 2.248363098730375e-05, "loss": 0.0539, "step": 43725 }, { "epoch": 1.02, "learning_rate": 2.248275093285771e-05, "loss": 0.1038, "step": 43730 }, { "epoch": 1.02, "learning_rate": 2.2481870878411676e-05, "loss": 0.1029, "step": 43735 }, { "epoch": 1.02, "learning_rate": 2.2480990823965645e-05, "loss": 0.1594, "step": 43740 }, { "epoch": 1.02, "learning_rate": 2.248011076951961e-05, "loss": 0.2163, "step": 43745 }, { "epoch": 1.02, "learning_rate": 2.247923071507357e-05, "loss": 0.2656, "step": 43750 }, { "epoch": 1.02, "learning_rate": 2.247835066062754e-05, "loss": 0.3217, "step": 43755 }, { "epoch": 1.02, "learning_rate": 2.2477470606181505e-05, "loss": 0.5124, "step": 43760 }, { "epoch": 1.02, "learning_rate": 2.2476590551735466e-05, "loss": 0.2328, "step": 43765 }, { "epoch": 1.02, "learning_rate": 2.247571049728943e-05, "loss": 0.0739, "step": 43770 }, { "epoch": 1.02, "learning_rate": 2.24748304428434e-05, "loss": 0.0366, "step": 43775 }, { "epoch": 1.02, "learning_rate": 2.2473950388397365e-05, "loss": 0.0626, "step": 43780 }, { "epoch": 1.02, "learning_rate": 2.2473070333951326e-05, "loss": 0.0535, "step": 43785 }, { "epoch": 1.02, "learning_rate": 2.247219027950529e-05, "loss": 0.1369, "step": 43790 }, { "epoch": 1.02, "learning_rate": 2.247131022505926e-05, "loss": 0.1724, "step": 43795 }, { "epoch": 1.02, "learning_rate": 2.247043017061322e-05, "loss": 0.2089, "step": 43800 }, { "epoch": 1.02, "learning_rate": 2.2469550116167186e-05, "loss": 0.2722, "step": 43805 }, { "epoch": 1.02, "learning_rate": 2.246867006172115e-05, "loss": 0.4463, "step": 43810 }, { "epoch": 1.02, "learning_rate": 2.246779000727512e-05, "loss": 0.3203, "step": 43815 }, { "epoch": 1.02, "learning_rate": 2.246690995282908e-05, "loss": 0.0727, "step": 43820 }, { "epoch": 1.02, "learning_rate": 2.2466029898383046e-05, "loss": 0.0732, "step": 43825 }, { "epoch": 1.02, "learning_rate": 2.246514984393701e-05, "loss": 0.1024, "step": 43830 }, { "epoch": 1.02, "learning_rate": 2.2464269789490977e-05, "loss": 0.1092, "step": 43835 }, { "epoch": 1.02, "learning_rate": 2.246338973504494e-05, "loss": 0.1749, "step": 43840 }, { "epoch": 1.02, "learning_rate": 2.2462509680598907e-05, "loss": 0.118, "step": 43845 }, { "epoch": 1.02, "learning_rate": 2.246162962615287e-05, "loss": 0.2331, "step": 43850 }, { "epoch": 1.02, "learning_rate": 2.2460749571706837e-05, "loss": 0.2595, "step": 43855 }, { "epoch": 1.02, "learning_rate": 2.24598695172608e-05, "loss": 0.3434, "step": 43860 }, { "epoch": 1.02, "learning_rate": 2.2458989462814767e-05, "loss": 0.3343, "step": 43865 }, { "epoch": 1.02, "learning_rate": 2.245810940836873e-05, "loss": 0.1079, "step": 43870 }, { "epoch": 1.02, "learning_rate": 2.2457229353922697e-05, "loss": 0.0788, "step": 43875 }, { "epoch": 1.02, "learning_rate": 2.245634929947666e-05, "loss": 0.1496, "step": 43880 }, { "epoch": 1.02, "learning_rate": 2.2455469245030627e-05, "loss": 0.2346, "step": 43885 }, { "epoch": 1.02, "learning_rate": 2.2454589190584592e-05, "loss": 0.1013, "step": 43890 }, { "epoch": 1.02, "learning_rate": 2.2453709136138557e-05, "loss": 0.1514, "step": 43895 }, { "epoch": 1.02, "learning_rate": 2.2452829081692522e-05, "loss": 0.1929, "step": 43900 }, { "epoch": 1.02, "learning_rate": 2.2451949027246487e-05, "loss": 0.1801, "step": 43905 }, { "epoch": 1.02, "learning_rate": 2.2451068972800452e-05, "loss": 0.4715, "step": 43910 }, { "epoch": 1.02, "learning_rate": 2.2450188918354417e-05, "loss": 0.34, "step": 43915 }, { "epoch": 1.02, "learning_rate": 2.2449308863908382e-05, "loss": 0.0619, "step": 43920 }, { "epoch": 1.02, "learning_rate": 2.2448428809462343e-05, "loss": 0.131, "step": 43925 }, { "epoch": 1.02, "learning_rate": 2.2447548755016312e-05, "loss": 0.0583, "step": 43930 }, { "epoch": 1.03, "learning_rate": 2.2446668700570277e-05, "loss": 0.062, "step": 43935 }, { "epoch": 1.03, "learning_rate": 2.2445788646124242e-05, "loss": 0.3348, "step": 43940 }, { "epoch": 1.03, "learning_rate": 2.2444908591678204e-05, "loss": 0.25, "step": 43945 }, { "epoch": 1.03, "learning_rate": 2.2444028537232172e-05, "loss": 0.4003, "step": 43950 }, { "epoch": 1.03, "learning_rate": 2.2443148482786137e-05, "loss": 0.266, "step": 43955 }, { "epoch": 1.03, "learning_rate": 2.24422684283401e-05, "loss": 0.4571, "step": 43960 }, { "epoch": 1.03, "learning_rate": 2.2441388373894064e-05, "loss": 0.3646, "step": 43965 }, { "epoch": 1.03, "learning_rate": 2.2440508319448032e-05, "loss": 0.0338, "step": 43970 }, { "epoch": 1.03, "learning_rate": 2.2439628265001997e-05, "loss": 0.1237, "step": 43975 }, { "epoch": 1.03, "learning_rate": 2.243874821055596e-05, "loss": 0.138, "step": 43980 }, { "epoch": 1.03, "learning_rate": 2.2437868156109924e-05, "loss": 0.1047, "step": 43985 }, { "epoch": 1.03, "learning_rate": 2.2436988101663892e-05, "loss": 0.2721, "step": 43990 }, { "epoch": 1.03, "learning_rate": 2.2436108047217854e-05, "loss": 0.0337, "step": 43995 }, { "epoch": 1.03, "learning_rate": 2.243522799277182e-05, "loss": 0.1796, "step": 44000 }, { "epoch": 1.03, "learning_rate": 2.2434347938325784e-05, "loss": 0.2843, "step": 44005 }, { "epoch": 1.03, "learning_rate": 2.2433467883879752e-05, "loss": 0.2685, "step": 44010 }, { "epoch": 1.03, "learning_rate": 2.2432587829433714e-05, "loss": 0.2984, "step": 44015 }, { "epoch": 1.03, "learning_rate": 2.243170777498768e-05, "loss": 0.0818, "step": 44020 }, { "epoch": 1.03, "learning_rate": 2.2430827720541647e-05, "loss": 0.1574, "step": 44025 }, { "epoch": 1.03, "learning_rate": 2.242994766609561e-05, "loss": 0.1282, "step": 44030 }, { "epoch": 1.03, "learning_rate": 2.2429067611649574e-05, "loss": 0.0994, "step": 44035 }, { "epoch": 1.03, "learning_rate": 2.242818755720354e-05, "loss": 0.152, "step": 44040 }, { "epoch": 1.03, "learning_rate": 2.2427307502757507e-05, "loss": 0.224, "step": 44045 }, { "epoch": 1.03, "learning_rate": 2.242642744831147e-05, "loss": 0.117, "step": 44050 }, { "epoch": 1.03, "learning_rate": 2.2425547393865434e-05, "loss": 0.3099, "step": 44055 }, { "epoch": 1.03, "learning_rate": 2.24246673394194e-05, "loss": 0.488, "step": 44060 }, { "epoch": 1.03, "learning_rate": 2.2423787284973364e-05, "loss": 0.3914, "step": 44065 }, { "epoch": 1.03, "learning_rate": 2.242290723052733e-05, "loss": 0.034, "step": 44070 }, { "epoch": 1.03, "learning_rate": 2.2422027176081294e-05, "loss": 0.1554, "step": 44075 }, { "epoch": 1.03, "learning_rate": 2.242114712163526e-05, "loss": 0.0813, "step": 44080 }, { "epoch": 1.03, "learning_rate": 2.2420267067189224e-05, "loss": 0.1786, "step": 44085 }, { "epoch": 1.03, "learning_rate": 2.241938701274319e-05, "loss": 0.2443, "step": 44090 }, { "epoch": 1.03, "learning_rate": 2.2418506958297154e-05, "loss": 0.1523, "step": 44095 }, { "epoch": 1.03, "learning_rate": 2.2417626903851116e-05, "loss": 0.2254, "step": 44100 }, { "epoch": 1.03, "learning_rate": 2.2416746849405084e-05, "loss": 0.2658, "step": 44105 }, { "epoch": 1.03, "learning_rate": 2.241586679495905e-05, "loss": 0.3462, "step": 44110 }, { "epoch": 1.03, "learning_rate": 2.2414986740513014e-05, "loss": 0.1673, "step": 44115 }, { "epoch": 1.03, "learning_rate": 2.2414106686066976e-05, "loss": 0.1947, "step": 44120 }, { "epoch": 1.03, "learning_rate": 2.2413226631620944e-05, "loss": 0.0295, "step": 44125 }, { "epoch": 1.03, "learning_rate": 2.241234657717491e-05, "loss": 0.0899, "step": 44130 }, { "epoch": 1.03, "learning_rate": 2.241146652272887e-05, "loss": 0.1114, "step": 44135 }, { "epoch": 1.03, "learning_rate": 2.2410586468282836e-05, "loss": 0.1656, "step": 44140 }, { "epoch": 1.03, "learning_rate": 2.2409706413836804e-05, "loss": 0.107, "step": 44145 }, { "epoch": 1.03, "learning_rate": 2.240882635939077e-05, "loss": 0.1947, "step": 44150 }, { "epoch": 1.03, "learning_rate": 2.240794630494473e-05, "loss": 0.3051, "step": 44155 }, { "epoch": 1.03, "learning_rate": 2.24070662504987e-05, "loss": 0.512, "step": 44160 }, { "epoch": 1.03, "learning_rate": 2.2406186196052664e-05, "loss": 0.1892, "step": 44165 }, { "epoch": 1.03, "learning_rate": 2.240530614160663e-05, "loss": 0.1175, "step": 44170 }, { "epoch": 1.03, "learning_rate": 2.240442608716059e-05, "loss": 0.0347, "step": 44175 }, { "epoch": 1.03, "learning_rate": 2.240354603271456e-05, "loss": 0.1382, "step": 44180 }, { "epoch": 1.03, "learning_rate": 2.2402665978268525e-05, "loss": 0.052, "step": 44185 }, { "epoch": 1.03, "learning_rate": 2.2401785923822486e-05, "loss": 0.0879, "step": 44190 }, { "epoch": 1.03, "learning_rate": 2.240090586937645e-05, "loss": 0.1302, "step": 44195 }, { "epoch": 1.03, "learning_rate": 2.240002581493042e-05, "loss": 0.156, "step": 44200 }, { "epoch": 1.03, "learning_rate": 2.2399145760484385e-05, "loss": 0.2207, "step": 44205 }, { "epoch": 1.03, "learning_rate": 2.2398265706038346e-05, "loss": 0.5951, "step": 44210 }, { "epoch": 1.03, "learning_rate": 2.239738565159231e-05, "loss": 0.3981, "step": 44215 }, { "epoch": 1.03, "learning_rate": 2.239650559714628e-05, "loss": 0.1048, "step": 44220 }, { "epoch": 1.03, "learning_rate": 2.239562554270024e-05, "loss": 0.0814, "step": 44225 }, { "epoch": 1.03, "learning_rate": 2.2394745488254206e-05, "loss": 0.1135, "step": 44230 }, { "epoch": 1.03, "learning_rate": 2.239386543380817e-05, "loss": 0.0828, "step": 44235 }, { "epoch": 1.03, "learning_rate": 2.239298537936214e-05, "loss": 0.1747, "step": 44240 }, { "epoch": 1.03, "learning_rate": 2.23921053249161e-05, "loss": 0.2971, "step": 44245 }, { "epoch": 1.03, "learning_rate": 2.2391225270470066e-05, "loss": 0.2079, "step": 44250 }, { "epoch": 1.03, "learning_rate": 2.239034521602403e-05, "loss": 0.4792, "step": 44255 }, { "epoch": 1.03, "learning_rate": 2.2389465161577996e-05, "loss": 0.4058, "step": 44260 }, { "epoch": 1.03, "learning_rate": 2.238858510713196e-05, "loss": 0.2739, "step": 44265 }, { "epoch": 1.03, "learning_rate": 2.2387705052685927e-05, "loss": 0.0942, "step": 44270 }, { "epoch": 1.03, "learning_rate": 2.238682499823989e-05, "loss": 0.0849, "step": 44275 }, { "epoch": 1.03, "learning_rate": 2.2385944943793857e-05, "loss": 0.0547, "step": 44280 }, { "epoch": 1.03, "learning_rate": 2.238506488934782e-05, "loss": 0.1217, "step": 44285 }, { "epoch": 1.03, "learning_rate": 2.2384184834901787e-05, "loss": 0.2145, "step": 44290 }, { "epoch": 1.03, "learning_rate": 2.238330478045575e-05, "loss": 0.1138, "step": 44295 }, { "epoch": 1.03, "learning_rate": 2.2382424726009717e-05, "loss": 0.2986, "step": 44300 }, { "epoch": 1.03, "learning_rate": 2.238154467156368e-05, "loss": 0.2058, "step": 44305 }, { "epoch": 1.03, "learning_rate": 2.2380664617117647e-05, "loss": 0.3159, "step": 44310 }, { "epoch": 1.03, "learning_rate": 2.2379784562671612e-05, "loss": 0.2189, "step": 44315 }, { "epoch": 1.03, "learning_rate": 2.2378904508225577e-05, "loss": 0.0718, "step": 44320 }, { "epoch": 1.03, "learning_rate": 2.2378024453779542e-05, "loss": 0.0444, "step": 44325 }, { "epoch": 1.03, "learning_rate": 2.2377144399333503e-05, "loss": 0.1459, "step": 44330 }, { "epoch": 1.03, "learning_rate": 2.2376264344887472e-05, "loss": 0.1203, "step": 44335 }, { "epoch": 1.03, "learning_rate": 2.2375384290441437e-05, "loss": 0.1432, "step": 44340 }, { "epoch": 1.03, "learning_rate": 2.2374504235995402e-05, "loss": 0.1222, "step": 44345 }, { "epoch": 1.03, "learning_rate": 2.2373624181549363e-05, "loss": 0.1947, "step": 44350 }, { "epoch": 1.03, "learning_rate": 2.2372744127103332e-05, "loss": 0.3254, "step": 44355 }, { "epoch": 1.03, "learning_rate": 2.2371864072657297e-05, "loss": 0.3463, "step": 44360 }, { "epoch": 1.04, "learning_rate": 2.237098401821126e-05, "loss": 0.3155, "step": 44365 }, { "epoch": 1.04, "learning_rate": 2.2370103963765224e-05, "loss": 0.0835, "step": 44370 }, { "epoch": 1.04, "learning_rate": 2.2369223909319192e-05, "loss": 0.0057, "step": 44375 }, { "epoch": 1.04, "learning_rate": 2.2368343854873157e-05, "loss": 0.1071, "step": 44380 }, { "epoch": 1.04, "learning_rate": 2.236746380042712e-05, "loss": 0.134, "step": 44385 }, { "epoch": 1.04, "learning_rate": 2.2366583745981084e-05, "loss": 0.2023, "step": 44390 }, { "epoch": 1.04, "learning_rate": 2.2365703691535052e-05, "loss": 0.1527, "step": 44395 }, { "epoch": 1.04, "learning_rate": 2.2364823637089014e-05, "loss": 0.0929, "step": 44400 }, { "epoch": 1.04, "learning_rate": 2.236394358264298e-05, "loss": 0.2676, "step": 44405 }, { "epoch": 1.04, "learning_rate": 2.2363063528196944e-05, "loss": 0.4376, "step": 44410 }, { "epoch": 1.04, "learning_rate": 2.2362183473750912e-05, "loss": 0.4119, "step": 44415 }, { "epoch": 1.04, "learning_rate": 2.2361303419304874e-05, "loss": 0.0541, "step": 44420 }, { "epoch": 1.04, "learning_rate": 2.236042336485884e-05, "loss": 0.0858, "step": 44425 }, { "epoch": 1.04, "learning_rate": 2.2359543310412807e-05, "loss": 0.0703, "step": 44430 }, { "epoch": 1.04, "learning_rate": 2.235866325596677e-05, "loss": 0.0616, "step": 44435 }, { "epoch": 1.04, "learning_rate": 2.2357783201520734e-05, "loss": 0.1304, "step": 44440 }, { "epoch": 1.04, "learning_rate": 2.23569031470747e-05, "loss": 0.2388, "step": 44445 }, { "epoch": 1.04, "learning_rate": 2.2356023092628667e-05, "loss": 0.4174, "step": 44450 }, { "epoch": 1.04, "learning_rate": 2.235514303818263e-05, "loss": 0.1866, "step": 44455 }, { "epoch": 1.04, "learning_rate": 2.2354262983736594e-05, "loss": 0.402, "step": 44460 }, { "epoch": 1.04, "learning_rate": 2.235338292929056e-05, "loss": 0.3785, "step": 44465 }, { "epoch": 1.04, "learning_rate": 2.2352502874844527e-05, "loss": 0.0208, "step": 44470 }, { "epoch": 1.04, "learning_rate": 2.235162282039849e-05, "loss": 0.0773, "step": 44475 }, { "epoch": 1.04, "learning_rate": 2.2350742765952454e-05, "loss": 0.0124, "step": 44480 }, { "epoch": 1.04, "learning_rate": 2.234986271150642e-05, "loss": 0.1149, "step": 44485 }, { "epoch": 1.04, "learning_rate": 2.2348982657060384e-05, "loss": 0.099, "step": 44490 }, { "epoch": 1.04, "learning_rate": 2.234810260261435e-05, "loss": 0.1323, "step": 44495 }, { "epoch": 1.04, "learning_rate": 2.2347222548168314e-05, "loss": 0.1457, "step": 44500 }, { "epoch": 1.04, "learning_rate": 2.234634249372228e-05, "loss": 0.277, "step": 44505 }, { "epoch": 1.04, "learning_rate": 2.2345462439276244e-05, "loss": 0.3939, "step": 44510 }, { "epoch": 1.04, "learning_rate": 2.234458238483021e-05, "loss": 0.4402, "step": 44515 }, { "epoch": 1.04, "learning_rate": 2.2343702330384174e-05, "loss": 0.0553, "step": 44520 }, { "epoch": 1.04, "learning_rate": 2.2342822275938136e-05, "loss": 0.0539, "step": 44525 }, { "epoch": 1.04, "learning_rate": 2.2341942221492104e-05, "loss": 0.0276, "step": 44530 }, { "epoch": 1.04, "learning_rate": 2.234106216704607e-05, "loss": 0.1094, "step": 44535 }, { "epoch": 1.04, "learning_rate": 2.2340182112600034e-05, "loss": 0.1581, "step": 44540 }, { "epoch": 1.04, "learning_rate": 2.2339302058153996e-05, "loss": 0.1258, "step": 44545 }, { "epoch": 1.04, "learning_rate": 2.2338422003707964e-05, "loss": 0.4227, "step": 44550 }, { "epoch": 1.04, "learning_rate": 2.233754194926193e-05, "loss": 0.21, "step": 44555 }, { "epoch": 1.04, "learning_rate": 2.233666189481589e-05, "loss": 0.3884, "step": 44560 }, { "epoch": 1.04, "learning_rate": 2.233578184036986e-05, "loss": 0.3746, "step": 44565 }, { "epoch": 1.04, "learning_rate": 2.2334901785923824e-05, "loss": 0.1254, "step": 44570 }, { "epoch": 1.04, "learning_rate": 2.233402173147779e-05, "loss": 0.062, "step": 44575 }, { "epoch": 1.04, "learning_rate": 2.233314167703175e-05, "loss": 0.0658, "step": 44580 }, { "epoch": 1.04, "learning_rate": 2.233226162258572e-05, "loss": 0.1104, "step": 44585 }, { "epoch": 1.04, "learning_rate": 2.2331381568139684e-05, "loss": 0.0913, "step": 44590 }, { "epoch": 1.04, "learning_rate": 2.2330501513693646e-05, "loss": 0.1655, "step": 44595 }, { "epoch": 1.04, "learning_rate": 2.232962145924761e-05, "loss": 0.2571, "step": 44600 }, { "epoch": 1.04, "learning_rate": 2.232874140480158e-05, "loss": 0.1918, "step": 44605 }, { "epoch": 1.04, "learning_rate": 2.2327861350355545e-05, "loss": 0.2234, "step": 44610 }, { "epoch": 1.04, "learning_rate": 2.2326981295909506e-05, "loss": 0.1924, "step": 44615 }, { "epoch": 1.04, "learning_rate": 2.232610124146347e-05, "loss": 0.1059, "step": 44620 }, { "epoch": 1.04, "learning_rate": 2.232522118701744e-05, "loss": 0.0398, "step": 44625 }, { "epoch": 1.04, "learning_rate": 2.23243411325714e-05, "loss": 0.139, "step": 44630 }, { "epoch": 1.04, "learning_rate": 2.2323461078125366e-05, "loss": 0.1975, "step": 44635 }, { "epoch": 1.04, "learning_rate": 2.232258102367933e-05, "loss": 0.11, "step": 44640 }, { "epoch": 1.04, "learning_rate": 2.23217009692333e-05, "loss": 0.2299, "step": 44645 }, { "epoch": 1.04, "learning_rate": 2.232082091478726e-05, "loss": 0.1162, "step": 44650 }, { "epoch": 1.04, "learning_rate": 2.2319940860341226e-05, "loss": 0.2034, "step": 44655 }, { "epoch": 1.04, "learning_rate": 2.231906080589519e-05, "loss": 0.4301, "step": 44660 }, { "epoch": 1.04, "learning_rate": 2.2318180751449156e-05, "loss": 0.2214, "step": 44665 }, { "epoch": 1.04, "learning_rate": 2.231730069700312e-05, "loss": 0.0777, "step": 44670 }, { "epoch": 1.04, "learning_rate": 2.2316420642557086e-05, "loss": 0.0622, "step": 44675 }, { "epoch": 1.04, "learning_rate": 2.231554058811105e-05, "loss": 0.1995, "step": 44680 }, { "epoch": 1.04, "learning_rate": 2.2314660533665016e-05, "loss": 0.0793, "step": 44685 }, { "epoch": 1.04, "learning_rate": 2.231378047921898e-05, "loss": 0.1187, "step": 44690 }, { "epoch": 1.04, "learning_rate": 2.2312900424772946e-05, "loss": 0.1792, "step": 44695 }, { "epoch": 1.04, "learning_rate": 2.231202037032691e-05, "loss": 0.2246, "step": 44700 }, { "epoch": 1.04, "learning_rate": 2.2311140315880877e-05, "loss": 0.2445, "step": 44705 }, { "epoch": 1.04, "learning_rate": 2.231026026143484e-05, "loss": 0.2909, "step": 44710 }, { "epoch": 1.04, "learning_rate": 2.2309380206988807e-05, "loss": 0.4889, "step": 44715 }, { "epoch": 1.04, "learning_rate": 2.230850015254277e-05, "loss": 0.0358, "step": 44720 }, { "epoch": 1.04, "learning_rate": 2.2307620098096737e-05, "loss": 0.0835, "step": 44725 }, { "epoch": 1.04, "learning_rate": 2.23067400436507e-05, "loss": 0.1288, "step": 44730 }, { "epoch": 1.04, "learning_rate": 2.2305859989204663e-05, "loss": 0.1208, "step": 44735 }, { "epoch": 1.04, "learning_rate": 2.230497993475863e-05, "loss": 0.1065, "step": 44740 }, { "epoch": 1.04, "learning_rate": 2.2304099880312597e-05, "loss": 0.105, "step": 44745 }, { "epoch": 1.04, "learning_rate": 2.2303219825866562e-05, "loss": 0.172, "step": 44750 }, { "epoch": 1.04, "learning_rate": 2.2302339771420523e-05, "loss": 0.3719, "step": 44755 }, { "epoch": 1.04, "learning_rate": 2.2301459716974492e-05, "loss": 0.3059, "step": 44760 }, { "epoch": 1.04, "learning_rate": 2.2300579662528457e-05, "loss": 0.3256, "step": 44765 }, { "epoch": 1.04, "learning_rate": 2.2299699608082422e-05, "loss": 0.1043, "step": 44770 }, { "epoch": 1.04, "learning_rate": 2.2298819553636383e-05, "loss": 0.1517, "step": 44775 }, { "epoch": 1.04, "learning_rate": 2.2297939499190352e-05, "loss": 0.1766, "step": 44780 }, { "epoch": 1.04, "learning_rate": 2.2297059444744317e-05, "loss": 0.204, "step": 44785 }, { "epoch": 1.05, "learning_rate": 2.229617939029828e-05, "loss": 0.1445, "step": 44790 }, { "epoch": 1.05, "learning_rate": 2.2295299335852244e-05, "loss": 0.1351, "step": 44795 }, { "epoch": 1.05, "learning_rate": 2.2294419281406212e-05, "loss": 0.1961, "step": 44800 }, { "epoch": 1.05, "learning_rate": 2.2293539226960177e-05, "loss": 0.2256, "step": 44805 }, { "epoch": 1.05, "learning_rate": 2.229265917251414e-05, "loss": 0.4279, "step": 44810 }, { "epoch": 1.05, "learning_rate": 2.2291779118068104e-05, "loss": 0.3961, "step": 44815 }, { "epoch": 1.05, "learning_rate": 2.2290899063622072e-05, "loss": 0.0973, "step": 44820 }, { "epoch": 1.05, "learning_rate": 2.2290019009176034e-05, "loss": 0.1222, "step": 44825 }, { "epoch": 1.05, "learning_rate": 2.228913895473e-05, "loss": 0.2311, "step": 44830 }, { "epoch": 1.05, "learning_rate": 2.2288258900283967e-05, "loss": 0.138, "step": 44835 }, { "epoch": 1.05, "learning_rate": 2.2287378845837932e-05, "loss": 0.1274, "step": 44840 }, { "epoch": 1.05, "learning_rate": 2.2286498791391894e-05, "loss": 0.2085, "step": 44845 }, { "epoch": 1.05, "learning_rate": 2.228561873694586e-05, "loss": 0.2406, "step": 44850 }, { "epoch": 1.05, "learning_rate": 2.2284738682499827e-05, "loss": 0.1395, "step": 44855 }, { "epoch": 1.05, "learning_rate": 2.228385862805379e-05, "loss": 0.2208, "step": 44860 }, { "epoch": 1.05, "learning_rate": 2.2282978573607754e-05, "loss": 0.319, "step": 44865 }, { "epoch": 1.05, "learning_rate": 2.228209851916172e-05, "loss": 0.0349, "step": 44870 }, { "epoch": 1.05, "learning_rate": 2.2281218464715687e-05, "loss": 0.0266, "step": 44875 }, { "epoch": 1.05, "learning_rate": 2.228033841026965e-05, "loss": 0.0732, "step": 44880 }, { "epoch": 1.05, "learning_rate": 2.2279458355823614e-05, "loss": 0.1557, "step": 44885 }, { "epoch": 1.05, "learning_rate": 2.227857830137758e-05, "loss": 0.0694, "step": 44890 }, { "epoch": 1.05, "learning_rate": 2.2277698246931544e-05, "loss": 0.2471, "step": 44895 }, { "epoch": 1.05, "learning_rate": 2.227681819248551e-05, "loss": 0.2874, "step": 44900 }, { "epoch": 1.05, "learning_rate": 2.2275938138039474e-05, "loss": 0.3417, "step": 44905 }, { "epoch": 1.05, "learning_rate": 2.227505808359344e-05, "loss": 0.3313, "step": 44910 }, { "epoch": 1.05, "learning_rate": 2.2274178029147404e-05, "loss": 0.4734, "step": 44915 }, { "epoch": 1.05, "learning_rate": 2.227329797470137e-05, "loss": 0.1039, "step": 44920 }, { "epoch": 1.05, "learning_rate": 2.2272417920255334e-05, "loss": 0.1481, "step": 44925 }, { "epoch": 1.05, "learning_rate": 2.2271537865809296e-05, "loss": 0.0939, "step": 44930 }, { "epoch": 1.05, "learning_rate": 2.2270657811363264e-05, "loss": 0.097, "step": 44935 }, { "epoch": 1.05, "learning_rate": 2.226977775691723e-05, "loss": 0.1068, "step": 44940 }, { "epoch": 1.05, "learning_rate": 2.2268897702471194e-05, "loss": 0.4717, "step": 44945 }, { "epoch": 1.05, "learning_rate": 2.2268017648025156e-05, "loss": 0.2424, "step": 44950 }, { "epoch": 1.05, "learning_rate": 2.2267137593579124e-05, "loss": 0.3328, "step": 44955 }, { "epoch": 1.05, "learning_rate": 2.226625753913309e-05, "loss": 0.5237, "step": 44960 }, { "epoch": 1.05, "learning_rate": 2.226537748468705e-05, "loss": 0.3439, "step": 44965 }, { "epoch": 1.05, "learning_rate": 2.226449743024102e-05, "loss": 0.0908, "step": 44970 }, { "epoch": 1.05, "learning_rate": 2.2263617375794984e-05, "loss": 0.0348, "step": 44975 }, { "epoch": 1.05, "learning_rate": 2.226273732134895e-05, "loss": 0.069, "step": 44980 }, { "epoch": 1.05, "learning_rate": 2.226185726690291e-05, "loss": 0.1371, "step": 44985 }, { "epoch": 1.05, "learning_rate": 2.226097721245688e-05, "loss": 0.0925, "step": 44990 }, { "epoch": 1.05, "learning_rate": 2.2260097158010844e-05, "loss": 0.18, "step": 44995 }, { "epoch": 1.05, "learning_rate": 2.2259217103564806e-05, "loss": 0.2806, "step": 45000 }, { "epoch": 1.05, "learning_rate": 2.225833704911877e-05, "loss": 0.3119, "step": 45005 }, { "epoch": 1.05, "learning_rate": 2.225745699467274e-05, "loss": 0.3983, "step": 45010 }, { "epoch": 1.05, "learning_rate": 2.2256576940226704e-05, "loss": 0.3056, "step": 45015 }, { "epoch": 1.05, "learning_rate": 2.2255696885780666e-05, "loss": 0.0829, "step": 45020 }, { "epoch": 1.05, "learning_rate": 2.225481683133463e-05, "loss": 0.1337, "step": 45025 }, { "epoch": 1.05, "learning_rate": 2.22539367768886e-05, "loss": 0.12, "step": 45030 }, { "epoch": 1.05, "learning_rate": 2.225305672244256e-05, "loss": 0.0423, "step": 45035 }, { "epoch": 1.05, "learning_rate": 2.2252176667996526e-05, "loss": 0.1691, "step": 45040 }, { "epoch": 1.05, "learning_rate": 2.225129661355049e-05, "loss": 0.1752, "step": 45045 }, { "epoch": 1.05, "learning_rate": 2.225041655910446e-05, "loss": 0.1896, "step": 45050 }, { "epoch": 1.05, "learning_rate": 2.224953650465842e-05, "loss": 0.113, "step": 45055 }, { "epoch": 1.05, "learning_rate": 2.2248656450212386e-05, "loss": 0.2955, "step": 45060 }, { "epoch": 1.05, "learning_rate": 2.224777639576635e-05, "loss": 0.3357, "step": 45065 }, { "epoch": 1.05, "learning_rate": 2.224689634132032e-05, "loss": 0.0516, "step": 45070 }, { "epoch": 1.05, "learning_rate": 2.224601628687428e-05, "loss": 0.0938, "step": 45075 }, { "epoch": 1.05, "learning_rate": 2.2245136232428246e-05, "loss": 0.0681, "step": 45080 }, { "epoch": 1.05, "learning_rate": 2.224425617798221e-05, "loss": 0.0828, "step": 45085 }, { "epoch": 1.05, "learning_rate": 2.2243376123536176e-05, "loss": 0.1651, "step": 45090 }, { "epoch": 1.05, "learning_rate": 2.224249606909014e-05, "loss": 0.1407, "step": 45095 }, { "epoch": 1.05, "learning_rate": 2.2241616014644106e-05, "loss": 0.1768, "step": 45100 }, { "epoch": 1.05, "learning_rate": 2.2240735960198075e-05, "loss": 0.2629, "step": 45105 }, { "epoch": 1.05, "learning_rate": 2.2239855905752036e-05, "loss": 0.5191, "step": 45110 }, { "epoch": 1.05, "learning_rate": 2.2238975851306e-05, "loss": 0.4052, "step": 45115 }, { "epoch": 1.05, "learning_rate": 2.2238095796859966e-05, "loss": 0.026, "step": 45120 }, { "epoch": 1.05, "learning_rate": 2.223721574241393e-05, "loss": 0.1047, "step": 45125 }, { "epoch": 1.05, "learning_rate": 2.2236335687967896e-05, "loss": 0.1297, "step": 45130 }, { "epoch": 1.05, "learning_rate": 2.223545563352186e-05, "loss": 0.2077, "step": 45135 }, { "epoch": 1.05, "learning_rate": 2.2234575579075827e-05, "loss": 0.154, "step": 45140 }, { "epoch": 1.05, "learning_rate": 2.223369552462979e-05, "loss": 0.0563, "step": 45145 }, { "epoch": 1.05, "learning_rate": 2.2232815470183757e-05, "loss": 0.2092, "step": 45150 }, { "epoch": 1.05, "learning_rate": 2.223193541573772e-05, "loss": 0.264, "step": 45155 }, { "epoch": 1.05, "learning_rate": 2.2231055361291683e-05, "loss": 0.2511, "step": 45160 }, { "epoch": 1.05, "learning_rate": 2.223017530684565e-05, "loss": 0.2139, "step": 45165 }, { "epoch": 1.05, "learning_rate": 2.2229295252399617e-05, "loss": 0.0557, "step": 45170 }, { "epoch": 1.05, "learning_rate": 2.222841519795358e-05, "loss": 0.0601, "step": 45175 }, { "epoch": 1.05, "learning_rate": 2.2227535143507543e-05, "loss": 0.1317, "step": 45180 }, { "epoch": 1.05, "learning_rate": 2.2226655089061512e-05, "loss": 0.0646, "step": 45185 }, { "epoch": 1.05, "learning_rate": 2.2225775034615477e-05, "loss": 0.1462, "step": 45190 }, { "epoch": 1.05, "learning_rate": 2.222489498016944e-05, "loss": 0.1792, "step": 45195 }, { "epoch": 1.05, "learning_rate": 2.2224014925723403e-05, "loss": 0.1245, "step": 45200 }, { "epoch": 1.05, "learning_rate": 2.2223134871277372e-05, "loss": 0.1969, "step": 45205 }, { "epoch": 1.05, "learning_rate": 2.2222254816831337e-05, "loss": 0.3727, "step": 45210 }, { "epoch": 1.05, "learning_rate": 2.22213747623853e-05, "loss": 0.2831, "step": 45215 }, { "epoch": 1.06, "learning_rate": 2.2220494707939263e-05, "loss": 0.0606, "step": 45220 }, { "epoch": 1.06, "learning_rate": 2.2219614653493232e-05, "loss": 0.1671, "step": 45225 }, { "epoch": 1.06, "learning_rate": 2.2218734599047194e-05, "loss": 0.0653, "step": 45230 }, { "epoch": 1.06, "learning_rate": 2.221785454460116e-05, "loss": 0.1047, "step": 45235 }, { "epoch": 1.06, "learning_rate": 2.2216974490155127e-05, "loss": 0.0919, "step": 45240 }, { "epoch": 1.06, "learning_rate": 2.2216094435709092e-05, "loss": 0.0383, "step": 45245 }, { "epoch": 1.06, "learning_rate": 2.2215214381263054e-05, "loss": 0.2483, "step": 45250 }, { "epoch": 1.06, "learning_rate": 2.221433432681702e-05, "loss": 0.1435, "step": 45255 }, { "epoch": 1.06, "learning_rate": 2.2213454272370987e-05, "loss": 0.6353, "step": 45260 }, { "epoch": 1.06, "learning_rate": 2.221257421792495e-05, "loss": 0.3578, "step": 45265 }, { "epoch": 1.06, "learning_rate": 2.2211694163478914e-05, "loss": 0.0709, "step": 45270 }, { "epoch": 1.06, "learning_rate": 2.221081410903288e-05, "loss": 0.0968, "step": 45275 }, { "epoch": 1.06, "learning_rate": 2.2209934054586847e-05, "loss": 0.0222, "step": 45280 }, { "epoch": 1.06, "learning_rate": 2.220905400014081e-05, "loss": 0.1246, "step": 45285 }, { "epoch": 1.06, "learning_rate": 2.2208173945694774e-05, "loss": 0.1263, "step": 45290 }, { "epoch": 1.06, "learning_rate": 2.220729389124874e-05, "loss": 0.0868, "step": 45295 }, { "epoch": 1.06, "learning_rate": 2.2206413836802704e-05, "loss": 0.1709, "step": 45300 }, { "epoch": 1.06, "learning_rate": 2.220553378235667e-05, "loss": 0.3962, "step": 45305 }, { "epoch": 1.06, "learning_rate": 2.2204653727910634e-05, "loss": 0.3554, "step": 45310 }, { "epoch": 1.06, "learning_rate": 2.22037736734646e-05, "loss": 0.3283, "step": 45315 }, { "epoch": 1.06, "learning_rate": 2.2202893619018564e-05, "loss": 0.0982, "step": 45320 }, { "epoch": 1.06, "learning_rate": 2.220201356457253e-05, "loss": 0.113, "step": 45325 }, { "epoch": 1.06, "learning_rate": 2.2201133510126494e-05, "loss": 0.1135, "step": 45330 }, { "epoch": 1.06, "learning_rate": 2.2200253455680456e-05, "loss": 0.2185, "step": 45335 }, { "epoch": 1.06, "learning_rate": 2.2199373401234424e-05, "loss": 0.1949, "step": 45340 }, { "epoch": 1.06, "learning_rate": 2.219849334678839e-05, "loss": 0.2282, "step": 45345 }, { "epoch": 1.06, "learning_rate": 2.2197613292342354e-05, "loss": 0.161, "step": 45350 }, { "epoch": 1.06, "learning_rate": 2.2196733237896316e-05, "loss": 0.2622, "step": 45355 }, { "epoch": 1.06, "learning_rate": 2.2195853183450284e-05, "loss": 0.329, "step": 45360 }, { "epoch": 1.06, "learning_rate": 2.219497312900425e-05, "loss": 0.3185, "step": 45365 }, { "epoch": 1.06, "learning_rate": 2.219409307455821e-05, "loss": 0.0438, "step": 45370 }, { "epoch": 1.06, "learning_rate": 2.219321302011218e-05, "loss": 0.0986, "step": 45375 }, { "epoch": 1.06, "learning_rate": 2.2192332965666144e-05, "loss": 0.2086, "step": 45380 }, { "epoch": 1.06, "learning_rate": 2.219145291122011e-05, "loss": 0.1519, "step": 45385 }, { "epoch": 1.06, "learning_rate": 2.219057285677407e-05, "loss": 0.1467, "step": 45390 }, { "epoch": 1.06, "learning_rate": 2.218969280232804e-05, "loss": 0.1724, "step": 45395 }, { "epoch": 1.06, "learning_rate": 2.2188812747882004e-05, "loss": 0.2482, "step": 45400 }, { "epoch": 1.06, "learning_rate": 2.218793269343597e-05, "loss": 0.3682, "step": 45405 }, { "epoch": 1.06, "learning_rate": 2.218705263898993e-05, "loss": 0.2643, "step": 45410 }, { "epoch": 1.06, "learning_rate": 2.21861725845439e-05, "loss": 0.238, "step": 45415 }, { "epoch": 1.06, "learning_rate": 2.2185292530097864e-05, "loss": 0.0746, "step": 45420 }, { "epoch": 1.06, "learning_rate": 2.2184412475651826e-05, "loss": 0.0573, "step": 45425 }, { "epoch": 1.06, "learning_rate": 2.218353242120579e-05, "loss": 0.0322, "step": 45430 }, { "epoch": 1.06, "learning_rate": 2.218265236675976e-05, "loss": 0.0731, "step": 45435 }, { "epoch": 1.06, "learning_rate": 2.2181772312313724e-05, "loss": 0.0882, "step": 45440 }, { "epoch": 1.06, "learning_rate": 2.2180892257867686e-05, "loss": 0.1929, "step": 45445 }, { "epoch": 1.06, "learning_rate": 2.218001220342165e-05, "loss": 0.1499, "step": 45450 }, { "epoch": 1.06, "learning_rate": 2.217913214897562e-05, "loss": 0.2163, "step": 45455 }, { "epoch": 1.06, "learning_rate": 2.217825209452958e-05, "loss": 0.3809, "step": 45460 }, { "epoch": 1.06, "learning_rate": 2.2177372040083546e-05, "loss": 0.4142, "step": 45465 }, { "epoch": 1.06, "learning_rate": 2.217649198563751e-05, "loss": 0.1121, "step": 45470 }, { "epoch": 1.06, "learning_rate": 2.217561193119148e-05, "loss": 0.0893, "step": 45475 }, { "epoch": 1.06, "learning_rate": 2.217473187674544e-05, "loss": 0.1591, "step": 45480 }, { "epoch": 1.06, "learning_rate": 2.2173851822299406e-05, "loss": 0.2317, "step": 45485 }, { "epoch": 1.06, "learning_rate": 2.217297176785337e-05, "loss": 0.1821, "step": 45490 }, { "epoch": 1.06, "learning_rate": 2.2172091713407336e-05, "loss": 0.1369, "step": 45495 }, { "epoch": 1.06, "learning_rate": 2.21712116589613e-05, "loss": 0.1551, "step": 45500 }, { "epoch": 1.06, "learning_rate": 2.2170331604515266e-05, "loss": 0.2347, "step": 45505 }, { "epoch": 1.06, "learning_rate": 2.216945155006923e-05, "loss": 0.4807, "step": 45510 }, { "epoch": 1.06, "learning_rate": 2.2168571495623196e-05, "loss": 0.3339, "step": 45515 }, { "epoch": 1.06, "learning_rate": 2.216769144117716e-05, "loss": 0.1079, "step": 45520 }, { "epoch": 1.06, "learning_rate": 2.2166811386731126e-05, "loss": 0.0385, "step": 45525 }, { "epoch": 1.06, "learning_rate": 2.216593133228509e-05, "loss": 0.0362, "step": 45530 }, { "epoch": 1.06, "learning_rate": 2.2165051277839056e-05, "loss": 0.1524, "step": 45535 }, { "epoch": 1.06, "learning_rate": 2.216417122339302e-05, "loss": 0.117, "step": 45540 }, { "epoch": 1.06, "learning_rate": 2.2163291168946986e-05, "loss": 0.1392, "step": 45545 }, { "epoch": 1.06, "learning_rate": 2.216241111450095e-05, "loss": 0.0421, "step": 45550 }, { "epoch": 1.06, "learning_rate": 2.2161531060054916e-05, "loss": 0.2813, "step": 45555 }, { "epoch": 1.06, "learning_rate": 2.216065100560888e-05, "loss": 0.2641, "step": 45560 }, { "epoch": 1.06, "learning_rate": 2.2159770951162843e-05, "loss": 0.2767, "step": 45565 }, { "epoch": 1.06, "learning_rate": 2.215889089671681e-05, "loss": 0.0478, "step": 45570 }, { "epoch": 1.06, "learning_rate": 2.2158010842270777e-05, "loss": 0.1452, "step": 45575 }, { "epoch": 1.06, "learning_rate": 2.215713078782474e-05, "loss": 0.0962, "step": 45580 }, { "epoch": 1.06, "learning_rate": 2.2156250733378703e-05, "loss": 0.1115, "step": 45585 }, { "epoch": 1.06, "learning_rate": 2.215537067893267e-05, "loss": 0.2238, "step": 45590 }, { "epoch": 1.06, "learning_rate": 2.2154490624486637e-05, "loss": 0.114, "step": 45595 }, { "epoch": 1.06, "learning_rate": 2.2153610570040598e-05, "loss": 0.249, "step": 45600 }, { "epoch": 1.06, "learning_rate": 2.2152730515594563e-05, "loss": 0.1461, "step": 45605 }, { "epoch": 1.06, "learning_rate": 2.215185046114853e-05, "loss": 0.2726, "step": 45610 }, { "epoch": 1.06, "learning_rate": 2.2150970406702497e-05, "loss": 0.3647, "step": 45615 }, { "epoch": 1.06, "learning_rate": 2.215009035225646e-05, "loss": 0.0544, "step": 45620 }, { "epoch": 1.06, "learning_rate": 2.2149210297810423e-05, "loss": 0.153, "step": 45625 }, { "epoch": 1.06, "learning_rate": 2.2148330243364392e-05, "loss": 0.0717, "step": 45630 }, { "epoch": 1.06, "learning_rate": 2.2147450188918353e-05, "loss": 0.1886, "step": 45635 }, { "epoch": 1.06, "learning_rate": 2.214657013447232e-05, "loss": 0.0877, "step": 45640 }, { "epoch": 1.06, "learning_rate": 2.2145690080026287e-05, "loss": 0.1728, "step": 45645 }, { "epoch": 1.07, "learning_rate": 2.2144810025580252e-05, "loss": 0.211, "step": 45650 }, { "epoch": 1.07, "learning_rate": 2.2143929971134213e-05, "loss": 0.2589, "step": 45655 }, { "epoch": 1.07, "learning_rate": 2.214304991668818e-05, "loss": 0.4747, "step": 45660 }, { "epoch": 1.07, "learning_rate": 2.2142169862242147e-05, "loss": 0.2652, "step": 45665 }, { "epoch": 1.07, "learning_rate": 2.214128980779611e-05, "loss": 0.0531, "step": 45670 }, { "epoch": 1.07, "learning_rate": 2.2140409753350074e-05, "loss": 0.1143, "step": 45675 }, { "epoch": 1.07, "learning_rate": 2.213952969890404e-05, "loss": 0.1101, "step": 45680 }, { "epoch": 1.07, "learning_rate": 2.2138649644458007e-05, "loss": 0.1069, "step": 45685 }, { "epoch": 1.07, "learning_rate": 2.213776959001197e-05, "loss": 0.1799, "step": 45690 }, { "epoch": 1.07, "learning_rate": 2.2136889535565934e-05, "loss": 0.2394, "step": 45695 }, { "epoch": 1.07, "learning_rate": 2.21360094811199e-05, "loss": 0.1595, "step": 45700 }, { "epoch": 1.07, "learning_rate": 2.2135129426673867e-05, "loss": 0.3619, "step": 45705 }, { "epoch": 1.07, "learning_rate": 2.213424937222783e-05, "loss": 0.4076, "step": 45710 }, { "epoch": 1.07, "learning_rate": 2.2133369317781794e-05, "loss": 0.3502, "step": 45715 }, { "epoch": 1.07, "learning_rate": 2.213248926333576e-05, "loss": 0.0505, "step": 45720 }, { "epoch": 1.07, "learning_rate": 2.2131609208889724e-05, "loss": 0.1007, "step": 45725 }, { "epoch": 1.07, "learning_rate": 2.213072915444369e-05, "loss": 0.0115, "step": 45730 }, { "epoch": 1.07, "learning_rate": 2.2129849099997654e-05, "loss": 0.0796, "step": 45735 }, { "epoch": 1.07, "learning_rate": 2.212896904555162e-05, "loss": 0.1066, "step": 45740 }, { "epoch": 1.07, "learning_rate": 2.2128088991105584e-05, "loss": 0.1636, "step": 45745 }, { "epoch": 1.07, "learning_rate": 2.212720893665955e-05, "loss": 0.4032, "step": 45750 }, { "epoch": 1.07, "learning_rate": 2.2126328882213514e-05, "loss": 0.3939, "step": 45755 }, { "epoch": 1.07, "learning_rate": 2.2125448827767476e-05, "loss": 0.273, "step": 45760 }, { "epoch": 1.07, "learning_rate": 2.2124568773321444e-05, "loss": 0.2841, "step": 45765 }, { "epoch": 1.07, "learning_rate": 2.212368871887541e-05, "loss": 0.0991, "step": 45770 }, { "epoch": 1.07, "learning_rate": 2.2122808664429374e-05, "loss": 0.1695, "step": 45775 }, { "epoch": 1.07, "learning_rate": 2.212192860998334e-05, "loss": 0.0894, "step": 45780 }, { "epoch": 1.07, "learning_rate": 2.2121048555537304e-05, "loss": 0.0733, "step": 45785 }, { "epoch": 1.07, "learning_rate": 2.212016850109127e-05, "loss": 0.104, "step": 45790 }, { "epoch": 1.07, "learning_rate": 2.211928844664523e-05, "loss": 0.1082, "step": 45795 }, { "epoch": 1.07, "learning_rate": 2.21184083921992e-05, "loss": 0.1763, "step": 45800 }, { "epoch": 1.07, "learning_rate": 2.2117528337753164e-05, "loss": 0.1545, "step": 45805 }, { "epoch": 1.07, "learning_rate": 2.211664828330713e-05, "loss": 0.3008, "step": 45810 }, { "epoch": 1.07, "learning_rate": 2.211576822886109e-05, "loss": 0.31, "step": 45815 }, { "epoch": 1.07, "learning_rate": 2.211488817441506e-05, "loss": 0.0799, "step": 45820 }, { "epoch": 1.07, "learning_rate": 2.2114008119969024e-05, "loss": 0.0239, "step": 45825 }, { "epoch": 1.07, "learning_rate": 2.2113128065522986e-05, "loss": 0.1532, "step": 45830 }, { "epoch": 1.07, "learning_rate": 2.211224801107695e-05, "loss": 0.2261, "step": 45835 }, { "epoch": 1.07, "learning_rate": 2.211136795663092e-05, "loss": 0.1757, "step": 45840 }, { "epoch": 1.07, "learning_rate": 2.2110487902184884e-05, "loss": 0.1424, "step": 45845 }, { "epoch": 1.07, "learning_rate": 2.2109607847738846e-05, "loss": 0.1585, "step": 45850 }, { "epoch": 1.07, "learning_rate": 2.210872779329281e-05, "loss": 0.1142, "step": 45855 }, { "epoch": 1.07, "learning_rate": 2.210784773884678e-05, "loss": 0.2535, "step": 45860 }, { "epoch": 1.07, "learning_rate": 2.210696768440074e-05, "loss": 0.2092, "step": 45865 }, { "epoch": 1.07, "learning_rate": 2.2106087629954706e-05, "loss": 0.1469, "step": 45870 }, { "epoch": 1.07, "learning_rate": 2.210520757550867e-05, "loss": 0.1124, "step": 45875 }, { "epoch": 1.07, "learning_rate": 2.210432752106264e-05, "loss": 0.2634, "step": 45880 }, { "epoch": 1.07, "learning_rate": 2.21034474666166e-05, "loss": 0.1461, "step": 45885 }, { "epoch": 1.07, "learning_rate": 2.2102567412170566e-05, "loss": 0.1011, "step": 45890 }, { "epoch": 1.07, "learning_rate": 2.210168735772453e-05, "loss": 0.1413, "step": 45895 }, { "epoch": 1.07, "learning_rate": 2.2100807303278496e-05, "loss": 0.1106, "step": 45900 }, { "epoch": 1.07, "learning_rate": 2.209992724883246e-05, "loss": 0.3505, "step": 45905 }, { "epoch": 1.07, "learning_rate": 2.2099047194386426e-05, "loss": 0.4372, "step": 45910 }, { "epoch": 1.07, "learning_rate": 2.209816713994039e-05, "loss": 0.3458, "step": 45915 }, { "epoch": 1.07, "learning_rate": 2.2097287085494356e-05, "loss": 0.1452, "step": 45920 }, { "epoch": 1.07, "learning_rate": 2.209640703104832e-05, "loss": 0.1241, "step": 45925 }, { "epoch": 1.07, "learning_rate": 2.2095526976602286e-05, "loss": 0.1239, "step": 45930 }, { "epoch": 1.07, "learning_rate": 2.209464692215625e-05, "loss": 0.1226, "step": 45935 }, { "epoch": 1.07, "learning_rate": 2.2093766867710216e-05, "loss": 0.1989, "step": 45940 }, { "epoch": 1.07, "learning_rate": 2.209288681326418e-05, "loss": 0.3191, "step": 45945 }, { "epoch": 1.07, "learning_rate": 2.2092006758818146e-05, "loss": 0.2219, "step": 45950 }, { "epoch": 1.07, "learning_rate": 2.209112670437211e-05, "loss": 0.1269, "step": 45955 }, { "epoch": 1.07, "learning_rate": 2.2090246649926076e-05, "loss": 0.4214, "step": 45960 }, { "epoch": 1.07, "learning_rate": 2.208936659548004e-05, "loss": 0.226, "step": 45965 }, { "epoch": 1.07, "learning_rate": 2.2088486541034003e-05, "loss": 0.0642, "step": 45970 }, { "epoch": 1.07, "learning_rate": 2.208760648658797e-05, "loss": 0.0502, "step": 45975 }, { "epoch": 1.07, "learning_rate": 2.2086726432141936e-05, "loss": 0.1606, "step": 45980 }, { "epoch": 1.07, "learning_rate": 2.20858463776959e-05, "loss": 0.0448, "step": 45985 }, { "epoch": 1.07, "learning_rate": 2.2084966323249863e-05, "loss": 0.1385, "step": 45990 }, { "epoch": 1.07, "learning_rate": 2.208408626880383e-05, "loss": 0.1043, "step": 45995 }, { "epoch": 1.07, "learning_rate": 2.2083206214357797e-05, "loss": 0.1546, "step": 46000 }, { "epoch": 1.07, "learning_rate": 2.208232615991176e-05, "loss": 0.3363, "step": 46005 }, { "epoch": 1.07, "learning_rate": 2.2081446105465723e-05, "loss": 0.2345, "step": 46010 }, { "epoch": 1.07, "learning_rate": 2.208056605101969e-05, "loss": 0.2084, "step": 46015 }, { "epoch": 1.07, "learning_rate": 2.2079685996573657e-05, "loss": 0.1484, "step": 46020 }, { "epoch": 1.07, "learning_rate": 2.2078805942127618e-05, "loss": 0.0207, "step": 46025 }, { "epoch": 1.07, "learning_rate": 2.2077925887681583e-05, "loss": 0.1044, "step": 46030 }, { "epoch": 1.07, "learning_rate": 2.207704583323555e-05, "loss": 0.0976, "step": 46035 }, { "epoch": 1.07, "learning_rate": 2.2076165778789517e-05, "loss": 0.2881, "step": 46040 }, { "epoch": 1.07, "learning_rate": 2.207528572434348e-05, "loss": 0.1153, "step": 46045 }, { "epoch": 1.07, "learning_rate": 2.2074405669897443e-05, "loss": 0.1591, "step": 46050 }, { "epoch": 1.07, "learning_rate": 2.2073525615451412e-05, "loss": 0.3113, "step": 46055 }, { "epoch": 1.07, "learning_rate": 2.2072645561005373e-05, "loss": 0.4064, "step": 46060 }, { "epoch": 1.07, "learning_rate": 2.207176550655934e-05, "loss": 0.2421, "step": 46065 }, { "epoch": 1.07, "learning_rate": 2.2070885452113307e-05, "loss": 0.1242, "step": 46070 }, { "epoch": 1.07, "learning_rate": 2.2070005397667272e-05, "loss": 0.1285, "step": 46075 }, { "epoch": 1.08, "learning_rate": 2.2069125343221233e-05, "loss": 0.0889, "step": 46080 }, { "epoch": 1.08, "learning_rate": 2.20682452887752e-05, "loss": 0.1061, "step": 46085 }, { "epoch": 1.08, "learning_rate": 2.2067365234329167e-05, "loss": 0.1891, "step": 46090 }, { "epoch": 1.08, "learning_rate": 2.206648517988313e-05, "loss": 0.4173, "step": 46095 }, { "epoch": 1.08, "learning_rate": 2.2065605125437094e-05, "loss": 0.1109, "step": 46100 }, { "epoch": 1.08, "learning_rate": 2.206472507099106e-05, "loss": 0.0867, "step": 46105 }, { "epoch": 1.08, "learning_rate": 2.2063845016545027e-05, "loss": 0.4524, "step": 46110 }, { "epoch": 1.08, "learning_rate": 2.206296496209899e-05, "loss": 0.3996, "step": 46115 }, { "epoch": 1.08, "learning_rate": 2.2062084907652954e-05, "loss": 0.0511, "step": 46120 }, { "epoch": 1.08, "learning_rate": 2.206120485320692e-05, "loss": 0.0423, "step": 46125 }, { "epoch": 1.08, "learning_rate": 2.2060324798760884e-05, "loss": 0.0465, "step": 46130 }, { "epoch": 1.08, "learning_rate": 2.205944474431485e-05, "loss": 0.1206, "step": 46135 }, { "epoch": 1.08, "learning_rate": 2.2058564689868814e-05, "loss": 0.0426, "step": 46140 }, { "epoch": 1.08, "learning_rate": 2.205768463542278e-05, "loss": 0.1539, "step": 46145 }, { "epoch": 1.08, "learning_rate": 2.2056804580976744e-05, "loss": 0.1829, "step": 46150 }, { "epoch": 1.08, "learning_rate": 2.205592452653071e-05, "loss": 0.2829, "step": 46155 }, { "epoch": 1.08, "learning_rate": 2.2055044472084674e-05, "loss": 0.3746, "step": 46160 }, { "epoch": 1.08, "learning_rate": 2.2054164417638635e-05, "loss": 0.2493, "step": 46165 }, { "epoch": 1.08, "learning_rate": 2.2053284363192604e-05, "loss": 0.0407, "step": 46170 }, { "epoch": 1.08, "learning_rate": 2.205240430874657e-05, "loss": 0.0246, "step": 46175 }, { "epoch": 1.08, "learning_rate": 2.2051524254300534e-05, "loss": 0.2124, "step": 46180 }, { "epoch": 1.08, "learning_rate": 2.20506441998545e-05, "loss": 0.1228, "step": 46185 }, { "epoch": 1.08, "learning_rate": 2.2049764145408464e-05, "loss": 0.0982, "step": 46190 }, { "epoch": 1.08, "learning_rate": 2.204888409096243e-05, "loss": 0.1295, "step": 46195 }, { "epoch": 1.08, "learning_rate": 2.204800403651639e-05, "loss": 0.2389, "step": 46200 }, { "epoch": 1.08, "learning_rate": 2.204712398207036e-05, "loss": 0.1945, "step": 46205 }, { "epoch": 1.08, "learning_rate": 2.2046243927624324e-05, "loss": 0.3562, "step": 46210 }, { "epoch": 1.08, "learning_rate": 2.204536387317829e-05, "loss": 0.7475, "step": 46215 }, { "epoch": 1.08, "learning_rate": 2.204448381873225e-05, "loss": 0.0847, "step": 46220 }, { "epoch": 1.08, "learning_rate": 2.204360376428622e-05, "loss": 0.1075, "step": 46225 }, { "epoch": 1.08, "learning_rate": 2.2042723709840184e-05, "loss": 0.1212, "step": 46230 }, { "epoch": 1.08, "learning_rate": 2.2041843655394146e-05, "loss": 0.0974, "step": 46235 }, { "epoch": 1.08, "learning_rate": 2.204096360094811e-05, "loss": 0.0951, "step": 46240 }, { "epoch": 1.08, "learning_rate": 2.204008354650208e-05, "loss": 0.2395, "step": 46245 }, { "epoch": 1.08, "learning_rate": 2.2039203492056044e-05, "loss": 0.336, "step": 46250 }, { "epoch": 1.08, "learning_rate": 2.2038323437610006e-05, "loss": 0.1329, "step": 46255 }, { "epoch": 1.08, "learning_rate": 2.203744338316397e-05, "loss": 0.3262, "step": 46260 }, { "epoch": 1.08, "learning_rate": 2.203656332871794e-05, "loss": 0.4907, "step": 46265 }, { "epoch": 1.08, "learning_rate": 2.20356832742719e-05, "loss": 0.0376, "step": 46270 }, { "epoch": 1.08, "learning_rate": 2.2034803219825866e-05, "loss": 0.1484, "step": 46275 }, { "epoch": 1.08, "learning_rate": 2.203392316537983e-05, "loss": 0.1065, "step": 46280 }, { "epoch": 1.08, "learning_rate": 2.20330431109338e-05, "loss": 0.1258, "step": 46285 }, { "epoch": 1.08, "learning_rate": 2.203216305648776e-05, "loss": 0.1876, "step": 46290 }, { "epoch": 1.08, "learning_rate": 2.2031283002041726e-05, "loss": 0.1632, "step": 46295 }, { "epoch": 1.08, "learning_rate": 2.203040294759569e-05, "loss": 0.1009, "step": 46300 }, { "epoch": 1.08, "learning_rate": 2.202952289314966e-05, "loss": 0.1733, "step": 46305 }, { "epoch": 1.08, "learning_rate": 2.202864283870362e-05, "loss": 0.3982, "step": 46310 }, { "epoch": 1.08, "learning_rate": 2.2027762784257586e-05, "loss": 0.3691, "step": 46315 }, { "epoch": 1.08, "learning_rate": 2.202688272981155e-05, "loss": 0.1132, "step": 46320 }, { "epoch": 1.08, "learning_rate": 2.2026002675365516e-05, "loss": 0.0369, "step": 46325 }, { "epoch": 1.08, "learning_rate": 2.202512262091948e-05, "loss": 0.2002, "step": 46330 }, { "epoch": 1.08, "learning_rate": 2.2024242566473446e-05, "loss": 0.0891, "step": 46335 }, { "epoch": 1.08, "learning_rate": 2.2023362512027415e-05, "loss": 0.0755, "step": 46340 }, { "epoch": 1.08, "learning_rate": 2.2022658468470582e-05, "loss": 0.1828, "step": 46345 }, { "epoch": 1.08, "learning_rate": 2.2021778414024547e-05, "loss": 0.2836, "step": 46350 }, { "epoch": 1.08, "learning_rate": 2.2020898359578513e-05, "loss": 0.3066, "step": 46355 }, { "epoch": 1.08, "learning_rate": 2.2020018305132478e-05, "loss": 0.608, "step": 46360 }, { "epoch": 1.08, "learning_rate": 2.2019138250686443e-05, "loss": 0.3372, "step": 46365 }, { "epoch": 1.08, "learning_rate": 2.2018258196240408e-05, "loss": 0.1115, "step": 46370 }, { "epoch": 1.08, "learning_rate": 2.2017378141794373e-05, "loss": 0.0595, "step": 46375 }, { "epoch": 1.08, "learning_rate": 2.2016498087348338e-05, "loss": 0.0821, "step": 46380 }, { "epoch": 1.08, "learning_rate": 2.2015618032902303e-05, "loss": 0.2475, "step": 46385 }, { "epoch": 1.08, "learning_rate": 2.2014737978456268e-05, "loss": 0.1711, "step": 46390 }, { "epoch": 1.08, "learning_rate": 2.2013857924010233e-05, "loss": 0.1471, "step": 46395 }, { "epoch": 1.08, "learning_rate": 2.2012977869564198e-05, "loss": 0.1887, "step": 46400 }, { "epoch": 1.08, "learning_rate": 2.2012097815118163e-05, "loss": 0.2195, "step": 46405 }, { "epoch": 1.08, "learning_rate": 2.2011217760672128e-05, "loss": 0.3736, "step": 46410 }, { "epoch": 1.08, "learning_rate": 2.2010337706226093e-05, "loss": 0.3607, "step": 46415 }, { "epoch": 1.08, "learning_rate": 2.2009457651780058e-05, "loss": 0.0755, "step": 46420 }, { "epoch": 1.08, "learning_rate": 2.2008577597334023e-05, "loss": 0.113, "step": 46425 }, { "epoch": 1.08, "learning_rate": 2.2007697542887984e-05, "loss": 0.0493, "step": 46430 }, { "epoch": 1.08, "learning_rate": 2.2006817488441953e-05, "loss": 0.0497, "step": 46435 }, { "epoch": 1.08, "learning_rate": 2.2005937433995918e-05, "loss": 0.1243, "step": 46440 }, { "epoch": 1.08, "learning_rate": 2.2005057379549883e-05, "loss": 0.1348, "step": 46445 }, { "epoch": 1.08, "learning_rate": 2.2004177325103845e-05, "loss": 0.283, "step": 46450 }, { "epoch": 1.08, "learning_rate": 2.2003297270657813e-05, "loss": 0.2015, "step": 46455 }, { "epoch": 1.08, "learning_rate": 2.2002417216211778e-05, "loss": 0.4243, "step": 46460 }, { "epoch": 1.08, "learning_rate": 2.200153716176574e-05, "loss": 0.2796, "step": 46465 }, { "epoch": 1.08, "learning_rate": 2.2000657107319705e-05, "loss": 0.0299, "step": 46470 }, { "epoch": 1.08, "learning_rate": 2.1999777052873673e-05, "loss": 0.0709, "step": 46475 }, { "epoch": 1.08, "learning_rate": 2.1998896998427638e-05, "loss": 0.0645, "step": 46480 }, { "epoch": 1.08, "learning_rate": 2.19980169439816e-05, "loss": 0.1356, "step": 46485 }, { "epoch": 1.08, "learning_rate": 2.1997136889535565e-05, "loss": 0.1351, "step": 46490 }, { "epoch": 1.08, "learning_rate": 2.1996256835089533e-05, "loss": 0.1166, "step": 46495 }, { "epoch": 1.08, "learning_rate": 2.1995376780643498e-05, "loss": 0.1018, "step": 46500 }, { "epoch": 1.09, "learning_rate": 2.199449672619746e-05, "loss": 0.3061, "step": 46505 }, { "epoch": 1.09, "learning_rate": 2.1993616671751425e-05, "loss": 0.2509, "step": 46510 }, { "epoch": 1.09, "learning_rate": 2.1992736617305393e-05, "loss": 0.4207, "step": 46515 }, { "epoch": 1.09, "learning_rate": 2.1991856562859355e-05, "loss": 0.0344, "step": 46520 }, { "epoch": 1.09, "learning_rate": 2.199097650841332e-05, "loss": 0.0749, "step": 46525 }, { "epoch": 1.09, "learning_rate": 2.1990096453967288e-05, "loss": 0.1136, "step": 46530 }, { "epoch": 1.09, "learning_rate": 2.1989216399521253e-05, "loss": 0.0801, "step": 46535 }, { "epoch": 1.09, "learning_rate": 2.1988336345075215e-05, "loss": 0.0594, "step": 46540 }, { "epoch": 1.09, "learning_rate": 2.198745629062918e-05, "loss": 0.1504, "step": 46545 }, { "epoch": 1.09, "learning_rate": 2.198657623618315e-05, "loss": 0.3287, "step": 46550 }, { "epoch": 1.09, "learning_rate": 2.198569618173711e-05, "loss": 0.2378, "step": 46555 }, { "epoch": 1.09, "learning_rate": 2.1984816127291075e-05, "loss": 0.3553, "step": 46560 }, { "epoch": 1.09, "learning_rate": 2.198393607284504e-05, "loss": 0.2679, "step": 46565 }, { "epoch": 1.09, "learning_rate": 2.198305601839901e-05, "loss": 0.0286, "step": 46570 }, { "epoch": 1.09, "learning_rate": 2.198217596395297e-05, "loss": 0.1109, "step": 46575 }, { "epoch": 1.09, "learning_rate": 2.1981295909506935e-05, "loss": 0.0535, "step": 46580 }, { "epoch": 1.09, "learning_rate": 2.19804158550609e-05, "loss": 0.2374, "step": 46585 }, { "epoch": 1.09, "learning_rate": 2.1979535800614865e-05, "loss": 0.1107, "step": 46590 }, { "epoch": 1.09, "learning_rate": 2.197865574616883e-05, "loss": 0.1842, "step": 46595 }, { "epoch": 1.09, "learning_rate": 2.1977775691722795e-05, "loss": 0.215, "step": 46600 }, { "epoch": 1.09, "learning_rate": 2.197689563727676e-05, "loss": 0.3627, "step": 46605 }, { "epoch": 1.09, "learning_rate": 2.1976015582830725e-05, "loss": 0.6031, "step": 46610 }, { "epoch": 1.09, "learning_rate": 2.197513552838469e-05, "loss": 0.403, "step": 46615 }, { "epoch": 1.09, "learning_rate": 2.1974255473938655e-05, "loss": 0.0089, "step": 46620 }, { "epoch": 1.09, "learning_rate": 2.1973375419492617e-05, "loss": 0.018, "step": 46625 }, { "epoch": 1.09, "learning_rate": 2.1972495365046585e-05, "loss": 0.0818, "step": 46630 }, { "epoch": 1.09, "learning_rate": 2.197161531060055e-05, "loss": 0.1457, "step": 46635 }, { "epoch": 1.09, "learning_rate": 2.1970735256154515e-05, "loss": 0.0696, "step": 46640 }, { "epoch": 1.09, "learning_rate": 2.1969855201708477e-05, "loss": 0.2321, "step": 46645 }, { "epoch": 1.09, "learning_rate": 2.1968975147262445e-05, "loss": 0.1592, "step": 46650 }, { "epoch": 1.09, "learning_rate": 2.196809509281641e-05, "loss": 0.2708, "step": 46655 }, { "epoch": 1.09, "learning_rate": 2.1967215038370372e-05, "loss": 0.2033, "step": 46660 }, { "epoch": 1.09, "learning_rate": 2.196633498392434e-05, "loss": 0.2954, "step": 46665 }, { "epoch": 1.09, "learning_rate": 2.1965454929478305e-05, "loss": 0.1335, "step": 46670 }, { "epoch": 1.09, "learning_rate": 2.196457487503227e-05, "loss": 0.1011, "step": 46675 }, { "epoch": 1.09, "learning_rate": 2.1963694820586232e-05, "loss": 0.0863, "step": 46680 }, { "epoch": 1.09, "learning_rate": 2.19628147661402e-05, "loss": 0.0951, "step": 46685 }, { "epoch": 1.09, "learning_rate": 2.1961934711694166e-05, "loss": 0.1726, "step": 46690 }, { "epoch": 1.09, "learning_rate": 2.1961054657248127e-05, "loss": 0.0519, "step": 46695 }, { "epoch": 1.09, "learning_rate": 2.1960174602802092e-05, "loss": 0.1741, "step": 46700 }, { "epoch": 1.09, "learning_rate": 2.195929454835606e-05, "loss": 0.3341, "step": 46705 }, { "epoch": 1.09, "learning_rate": 2.1958414493910026e-05, "loss": 0.5908, "step": 46710 }, { "epoch": 1.09, "learning_rate": 2.1957534439463987e-05, "loss": 0.2672, "step": 46715 }, { "epoch": 1.09, "learning_rate": 2.1956654385017952e-05, "loss": 0.0726, "step": 46720 }, { "epoch": 1.09, "learning_rate": 2.195577433057192e-05, "loss": 0.0341, "step": 46725 }, { "epoch": 1.09, "learning_rate": 2.1954894276125882e-05, "loss": 0.1002, "step": 46730 }, { "epoch": 1.09, "learning_rate": 2.1954014221679847e-05, "loss": 0.0532, "step": 46735 }, { "epoch": 1.09, "learning_rate": 2.1953134167233812e-05, "loss": 0.144, "step": 46740 }, { "epoch": 1.09, "learning_rate": 2.195225411278778e-05, "loss": 0.1256, "step": 46745 }, { "epoch": 1.09, "learning_rate": 2.1951374058341742e-05, "loss": 0.1771, "step": 46750 }, { "epoch": 1.09, "learning_rate": 2.1950494003895707e-05, "loss": 0.3606, "step": 46755 }, { "epoch": 1.09, "learning_rate": 2.1949613949449672e-05, "loss": 0.5266, "step": 46760 }, { "epoch": 1.09, "learning_rate": 2.1948733895003637e-05, "loss": 0.3176, "step": 46765 }, { "epoch": 1.09, "learning_rate": 2.1947853840557602e-05, "loss": 0.0617, "step": 46770 }, { "epoch": 1.09, "learning_rate": 2.1946973786111567e-05, "loss": 0.1226, "step": 46775 }, { "epoch": 1.09, "learning_rate": 2.1946093731665532e-05, "loss": 0.0747, "step": 46780 }, { "epoch": 1.09, "learning_rate": 2.1945213677219498e-05, "loss": 0.1614, "step": 46785 }, { "epoch": 1.09, "learning_rate": 2.1944333622773463e-05, "loss": 0.1738, "step": 46790 }, { "epoch": 1.09, "learning_rate": 2.1943453568327428e-05, "loss": 0.1484, "step": 46795 }, { "epoch": 1.09, "learning_rate": 2.1942573513881396e-05, "loss": 0.2158, "step": 46800 }, { "epoch": 1.09, "learning_rate": 2.1941693459435358e-05, "loss": 0.2064, "step": 46805 }, { "epoch": 1.09, "learning_rate": 2.1940813404989323e-05, "loss": 0.2668, "step": 46810 }, { "epoch": 1.09, "learning_rate": 2.1939933350543288e-05, "loss": 0.3096, "step": 46815 }, { "epoch": 1.09, "learning_rate": 2.1939053296097253e-05, "loss": 0.0437, "step": 46820 }, { "epoch": 1.09, "learning_rate": 2.1938173241651218e-05, "loss": 0.2002, "step": 46825 }, { "epoch": 1.09, "learning_rate": 2.1937293187205183e-05, "loss": 0.0817, "step": 46830 }, { "epoch": 1.09, "learning_rate": 2.1936413132759148e-05, "loss": 0.1759, "step": 46835 }, { "epoch": 1.09, "learning_rate": 2.1935533078313113e-05, "loss": 0.1387, "step": 46840 }, { "epoch": 1.09, "learning_rate": 2.1934653023867078e-05, "loss": 0.1865, "step": 46845 }, { "epoch": 1.09, "learning_rate": 2.1933772969421043e-05, "loss": 0.2625, "step": 46850 }, { "epoch": 1.09, "learning_rate": 2.1932892914975004e-05, "loss": 0.333, "step": 46855 }, { "epoch": 1.09, "learning_rate": 2.1932012860528973e-05, "loss": 0.6413, "step": 46860 }, { "epoch": 1.09, "learning_rate": 2.1931132806082938e-05, "loss": 0.2374, "step": 46865 }, { "epoch": 1.09, "learning_rate": 2.1930252751636903e-05, "loss": 0.0489, "step": 46870 }, { "epoch": 1.09, "learning_rate": 2.1929372697190864e-05, "loss": 0.0668, "step": 46875 }, { "epoch": 1.09, "learning_rate": 2.1928492642744833e-05, "loss": 0.1594, "step": 46880 }, { "epoch": 1.09, "learning_rate": 2.1927612588298798e-05, "loss": 0.0702, "step": 46885 }, { "epoch": 1.09, "learning_rate": 2.192673253385276e-05, "loss": 0.1707, "step": 46890 }, { "epoch": 1.09, "learning_rate": 2.1925852479406725e-05, "loss": 0.0642, "step": 46895 }, { "epoch": 1.09, "learning_rate": 2.1924972424960693e-05, "loss": 0.2331, "step": 46900 }, { "epoch": 1.09, "learning_rate": 2.1924092370514658e-05, "loss": 0.2552, "step": 46905 }, { "epoch": 1.09, "learning_rate": 2.192321231606862e-05, "loss": 0.2521, "step": 46910 }, { "epoch": 1.09, "learning_rate": 2.1922332261622585e-05, "loss": 0.3018, "step": 46915 }, { "epoch": 1.09, "learning_rate": 2.1921452207176553e-05, "loss": 0.0575, "step": 46920 }, { "epoch": 1.09, "learning_rate": 2.1920572152730515e-05, "loss": 0.0665, "step": 46925 }, { "epoch": 1.09, "learning_rate": 2.191969209828448e-05, "loss": 0.1025, "step": 46930 }, { "epoch": 1.1, "learning_rate": 2.1918812043838448e-05, "loss": 0.0972, "step": 46935 }, { "epoch": 1.1, "learning_rate": 2.1917931989392413e-05, "loss": 0.1044, "step": 46940 }, { "epoch": 1.1, "learning_rate": 2.1917051934946375e-05, "loss": 0.0601, "step": 46945 }, { "epoch": 1.1, "learning_rate": 2.191617188050034e-05, "loss": 0.2901, "step": 46950 }, { "epoch": 1.1, "learning_rate": 2.1915291826054308e-05, "loss": 0.1907, "step": 46955 }, { "epoch": 1.1, "learning_rate": 2.191441177160827e-05, "loss": 0.3146, "step": 46960 }, { "epoch": 1.1, "learning_rate": 2.1913531717162235e-05, "loss": 0.3872, "step": 46965 }, { "epoch": 1.1, "learning_rate": 2.19126516627162e-05, "loss": 0.04, "step": 46970 }, { "epoch": 1.1, "learning_rate": 2.1911771608270168e-05, "loss": 0.0605, "step": 46975 }, { "epoch": 1.1, "learning_rate": 2.191089155382413e-05, "loss": 0.1764, "step": 46980 }, { "epoch": 1.1, "learning_rate": 2.1910011499378095e-05, "loss": 0.0818, "step": 46985 }, { "epoch": 1.1, "learning_rate": 2.190913144493206e-05, "loss": 0.1214, "step": 46990 }, { "epoch": 1.1, "learning_rate": 2.1908251390486025e-05, "loss": 0.2402, "step": 46995 }, { "epoch": 1.1, "learning_rate": 2.190737133603999e-05, "loss": 0.2471, "step": 47000 }, { "epoch": 1.1, "learning_rate": 2.1906491281593955e-05, "loss": 0.2569, "step": 47005 }, { "epoch": 1.1, "learning_rate": 2.190561122714792e-05, "loss": 0.2916, "step": 47010 }, { "epoch": 1.1, "learning_rate": 2.1904731172701885e-05, "loss": 0.2806, "step": 47015 }, { "epoch": 1.1, "learning_rate": 2.190385111825585e-05, "loss": 0.0175, "step": 47020 }, { "epoch": 1.1, "learning_rate": 2.1902971063809815e-05, "loss": 0.1348, "step": 47025 }, { "epoch": 1.1, "learning_rate": 2.1902091009363777e-05, "loss": 0.0439, "step": 47030 }, { "epoch": 1.1, "learning_rate": 2.1901210954917745e-05, "loss": 0.2221, "step": 47035 }, { "epoch": 1.1, "learning_rate": 2.190033090047171e-05, "loss": 0.1094, "step": 47040 }, { "epoch": 1.1, "learning_rate": 2.1899450846025675e-05, "loss": 0.1162, "step": 47045 }, { "epoch": 1.1, "learning_rate": 2.1898570791579637e-05, "loss": 0.2581, "step": 47050 }, { "epoch": 1.1, "learning_rate": 2.1897690737133605e-05, "loss": 0.2954, "step": 47055 }, { "epoch": 1.1, "learning_rate": 2.189681068268757e-05, "loss": 0.4086, "step": 47060 }, { "epoch": 1.1, "learning_rate": 2.1895930628241532e-05, "loss": 0.3225, "step": 47065 }, { "epoch": 1.1, "learning_rate": 2.18950505737955e-05, "loss": 0.1057, "step": 47070 }, { "epoch": 1.1, "learning_rate": 2.1894170519349465e-05, "loss": 0.1173, "step": 47075 }, { "epoch": 1.1, "learning_rate": 2.189329046490343e-05, "loss": 0.1486, "step": 47080 }, { "epoch": 1.1, "learning_rate": 2.1892410410457392e-05, "loss": 0.1016, "step": 47085 }, { "epoch": 1.1, "learning_rate": 2.189153035601136e-05, "loss": 0.1174, "step": 47090 }, { "epoch": 1.1, "learning_rate": 2.1890650301565325e-05, "loss": 0.151, "step": 47095 }, { "epoch": 1.1, "learning_rate": 2.188977024711929e-05, "loss": 0.1962, "step": 47100 }, { "epoch": 1.1, "learning_rate": 2.1888890192673252e-05, "loss": 0.1911, "step": 47105 }, { "epoch": 1.1, "learning_rate": 2.188801013822722e-05, "loss": 0.4062, "step": 47110 }, { "epoch": 1.1, "learning_rate": 2.1887130083781185e-05, "loss": 0.48, "step": 47115 }, { "epoch": 1.1, "learning_rate": 2.1886250029335147e-05, "loss": 0.0153, "step": 47120 }, { "epoch": 1.1, "learning_rate": 2.1885369974889112e-05, "loss": 0.0921, "step": 47125 }, { "epoch": 1.1, "learning_rate": 2.188448992044308e-05, "loss": 0.1325, "step": 47130 }, { "epoch": 1.1, "learning_rate": 2.1883609865997046e-05, "loss": 0.1866, "step": 47135 }, { "epoch": 1.1, "learning_rate": 2.1882729811551007e-05, "loss": 0.0671, "step": 47140 }, { "epoch": 1.1, "learning_rate": 2.1881849757104972e-05, "loss": 0.3302, "step": 47145 }, { "epoch": 1.1, "learning_rate": 2.188096970265894e-05, "loss": 0.2212, "step": 47150 }, { "epoch": 1.1, "learning_rate": 2.1880089648212902e-05, "loss": 0.1955, "step": 47155 }, { "epoch": 1.1, "learning_rate": 2.1879209593766867e-05, "loss": 0.3671, "step": 47160 }, { "epoch": 1.1, "learning_rate": 2.1878329539320832e-05, "loss": 0.2724, "step": 47165 }, { "epoch": 1.1, "learning_rate": 2.18774494848748e-05, "loss": 0.0585, "step": 47170 }, { "epoch": 1.1, "learning_rate": 2.1876569430428762e-05, "loss": 0.1895, "step": 47175 }, { "epoch": 1.1, "learning_rate": 2.1875689375982727e-05, "loss": 0.131, "step": 47180 }, { "epoch": 1.1, "learning_rate": 2.1874809321536692e-05, "loss": 0.0606, "step": 47185 }, { "epoch": 1.1, "learning_rate": 2.1873929267090657e-05, "loss": 0.1287, "step": 47190 }, { "epoch": 1.1, "learning_rate": 2.1873049212644622e-05, "loss": 0.0836, "step": 47195 }, { "epoch": 1.1, "learning_rate": 2.1872169158198587e-05, "loss": 0.2639, "step": 47200 }, { "epoch": 1.1, "learning_rate": 2.1871289103752556e-05, "loss": 0.197, "step": 47205 }, { "epoch": 1.1, "learning_rate": 2.1870409049306517e-05, "loss": 0.4298, "step": 47210 }, { "epoch": 1.1, "learning_rate": 2.1869528994860482e-05, "loss": 0.3642, "step": 47215 }, { "epoch": 1.1, "learning_rate": 2.1868648940414448e-05, "loss": 0.0759, "step": 47220 }, { "epoch": 1.1, "learning_rate": 2.1867768885968413e-05, "loss": 0.0339, "step": 47225 }, { "epoch": 1.1, "learning_rate": 2.1866888831522378e-05, "loss": 0.1154, "step": 47230 }, { "epoch": 1.1, "learning_rate": 2.1866008777076343e-05, "loss": 0.1389, "step": 47235 }, { "epoch": 1.1, "learning_rate": 2.1865128722630308e-05, "loss": 0.089, "step": 47240 }, { "epoch": 1.1, "learning_rate": 2.1864248668184273e-05, "loss": 0.3079, "step": 47245 }, { "epoch": 1.1, "learning_rate": 2.1863368613738238e-05, "loss": 0.264, "step": 47250 }, { "epoch": 1.1, "learning_rate": 2.1862488559292203e-05, "loss": 0.2191, "step": 47255 }, { "epoch": 1.1, "learning_rate": 2.1861608504846164e-05, "loss": 1.0166, "step": 47260 }, { "epoch": 1.1, "learning_rate": 2.1860728450400133e-05, "loss": 0.2808, "step": 47265 }, { "epoch": 1.1, "learning_rate": 2.1859848395954098e-05, "loss": 0.0577, "step": 47270 }, { "epoch": 1.1, "learning_rate": 2.1858968341508063e-05, "loss": 0.0547, "step": 47275 }, { "epoch": 1.1, "learning_rate": 2.1858088287062024e-05, "loss": 0.04, "step": 47280 }, { "epoch": 1.1, "learning_rate": 2.1857208232615993e-05, "loss": 0.0742, "step": 47285 }, { "epoch": 1.1, "learning_rate": 2.1856328178169958e-05, "loss": 0.1699, "step": 47290 }, { "epoch": 1.1, "learning_rate": 2.185544812372392e-05, "loss": 0.1468, "step": 47295 }, { "epoch": 1.1, "learning_rate": 2.1854568069277884e-05, "loss": 0.2001, "step": 47300 }, { "epoch": 1.1, "learning_rate": 2.1853688014831853e-05, "loss": 0.3125, "step": 47305 }, { "epoch": 1.1, "learning_rate": 2.1852807960385818e-05, "loss": 0.5852, "step": 47310 }, { "epoch": 1.1, "learning_rate": 2.185192790593978e-05, "loss": 0.2359, "step": 47315 }, { "epoch": 1.1, "learning_rate": 2.1851047851493745e-05, "loss": 0.0857, "step": 47320 }, { "epoch": 1.1, "learning_rate": 2.1850167797047713e-05, "loss": 0.038, "step": 47325 }, { "epoch": 1.1, "learning_rate": 2.1849287742601675e-05, "loss": 0.0774, "step": 47330 }, { "epoch": 1.1, "learning_rate": 2.184840768815564e-05, "loss": 0.1483, "step": 47335 }, { "epoch": 1.1, "learning_rate": 2.1847527633709608e-05, "loss": 0.1563, "step": 47340 }, { "epoch": 1.1, "learning_rate": 2.1846647579263573e-05, "loss": 0.1131, "step": 47345 }, { "epoch": 1.1, "learning_rate": 2.1845767524817535e-05, "loss": 0.0583, "step": 47350 }, { "epoch": 1.1, "learning_rate": 2.18448874703715e-05, "loss": 0.1525, "step": 47355 }, { "epoch": 1.1, "learning_rate": 2.1844007415925468e-05, "loss": 0.5227, "step": 47360 }, { "epoch": 1.11, "learning_rate": 2.184312736147943e-05, "loss": 0.2774, "step": 47365 }, { "epoch": 1.11, "learning_rate": 2.1842247307033395e-05, "loss": 0.0268, "step": 47370 }, { "epoch": 1.11, "learning_rate": 2.184136725258736e-05, "loss": 0.0594, "step": 47375 }, { "epoch": 1.11, "learning_rate": 2.1840487198141328e-05, "loss": 0.1343, "step": 47380 }, { "epoch": 1.11, "learning_rate": 2.183960714369529e-05, "loss": 0.1436, "step": 47385 }, { "epoch": 1.11, "learning_rate": 2.1838727089249255e-05, "loss": 0.1877, "step": 47390 }, { "epoch": 1.11, "learning_rate": 2.183784703480322e-05, "loss": 0.2488, "step": 47395 }, { "epoch": 1.11, "learning_rate": 2.1836966980357188e-05, "loss": 0.1823, "step": 47400 }, { "epoch": 1.11, "learning_rate": 2.183608692591115e-05, "loss": 0.3582, "step": 47405 }, { "epoch": 1.11, "learning_rate": 2.1835206871465115e-05, "loss": 0.4465, "step": 47410 }, { "epoch": 1.11, "learning_rate": 2.183432681701908e-05, "loss": 0.3053, "step": 47415 }, { "epoch": 1.11, "learning_rate": 2.1833446762573045e-05, "loss": 0.023, "step": 47420 }, { "epoch": 1.11, "learning_rate": 2.183256670812701e-05, "loss": 0.011, "step": 47425 }, { "epoch": 1.11, "learning_rate": 2.1831686653680975e-05, "loss": 0.1188, "step": 47430 }, { "epoch": 1.11, "learning_rate": 2.183080659923494e-05, "loss": 0.1074, "step": 47435 }, { "epoch": 1.11, "learning_rate": 2.1829926544788905e-05, "loss": 0.1964, "step": 47440 }, { "epoch": 1.11, "learning_rate": 2.182904649034287e-05, "loss": 0.3304, "step": 47445 }, { "epoch": 1.11, "learning_rate": 2.1828166435896835e-05, "loss": 0.1533, "step": 47450 }, { "epoch": 1.11, "learning_rate": 2.1827286381450797e-05, "loss": 0.3908, "step": 47455 }, { "epoch": 1.11, "learning_rate": 2.1826406327004765e-05, "loss": 0.3003, "step": 47460 }, { "epoch": 1.11, "learning_rate": 2.182552627255873e-05, "loss": 0.2656, "step": 47465 }, { "epoch": 1.11, "learning_rate": 2.1824646218112695e-05, "loss": 0.0876, "step": 47470 }, { "epoch": 1.11, "learning_rate": 2.182376616366666e-05, "loss": 0.1069, "step": 47475 }, { "epoch": 1.11, "learning_rate": 2.1822886109220625e-05, "loss": 0.1075, "step": 47480 }, { "epoch": 1.11, "learning_rate": 2.182200605477459e-05, "loss": 0.0944, "step": 47485 }, { "epoch": 1.11, "learning_rate": 2.1821126000328552e-05, "loss": 0.1571, "step": 47490 }, { "epoch": 1.11, "learning_rate": 2.182024594588252e-05, "loss": 0.2874, "step": 47495 }, { "epoch": 1.11, "learning_rate": 2.1819365891436485e-05, "loss": 0.076, "step": 47500 }, { "epoch": 1.11, "learning_rate": 2.181848583699045e-05, "loss": 0.2574, "step": 47505 }, { "epoch": 1.11, "learning_rate": 2.1817605782544412e-05, "loss": 0.2324, "step": 47510 }, { "epoch": 1.11, "learning_rate": 2.181672572809838e-05, "loss": 0.3309, "step": 47515 }, { "epoch": 1.11, "learning_rate": 2.1815845673652345e-05, "loss": 0.0486, "step": 47520 }, { "epoch": 1.11, "learning_rate": 2.1814965619206307e-05, "loss": 0.0653, "step": 47525 }, { "epoch": 1.11, "learning_rate": 2.1814085564760272e-05, "loss": 0.1957, "step": 47530 }, { "epoch": 1.11, "learning_rate": 2.181320551031424e-05, "loss": 0.1295, "step": 47535 }, { "epoch": 1.11, "learning_rate": 2.1812325455868205e-05, "loss": 0.1182, "step": 47540 }, { "epoch": 1.11, "learning_rate": 2.1811445401422167e-05, "loss": 0.1725, "step": 47545 }, { "epoch": 1.11, "learning_rate": 2.1810565346976132e-05, "loss": 0.1848, "step": 47550 }, { "epoch": 1.11, "learning_rate": 2.18096852925301e-05, "loss": 0.3242, "step": 47555 }, { "epoch": 1.11, "learning_rate": 2.1808805238084062e-05, "loss": 0.4088, "step": 47560 }, { "epoch": 1.11, "learning_rate": 2.1807925183638027e-05, "loss": 0.1984, "step": 47565 }, { "epoch": 1.11, "learning_rate": 2.1807045129191992e-05, "loss": 0.1118, "step": 47570 }, { "epoch": 1.11, "learning_rate": 2.180616507474596e-05, "loss": 0.0338, "step": 47575 }, { "epoch": 1.11, "learning_rate": 2.1805285020299922e-05, "loss": 0.0923, "step": 47580 }, { "epoch": 1.11, "learning_rate": 2.1804404965853887e-05, "loss": 0.2057, "step": 47585 }, { "epoch": 1.11, "learning_rate": 2.1803524911407852e-05, "loss": 0.0713, "step": 47590 }, { "epoch": 1.11, "learning_rate": 2.1802644856961817e-05, "loss": 0.3555, "step": 47595 }, { "epoch": 1.11, "learning_rate": 2.1801764802515782e-05, "loss": 0.2109, "step": 47600 }, { "epoch": 1.11, "learning_rate": 2.1800884748069747e-05, "loss": 0.3416, "step": 47605 }, { "epoch": 1.11, "learning_rate": 2.1800004693623716e-05, "loss": 0.3277, "step": 47610 }, { "epoch": 1.11, "learning_rate": 2.1799124639177677e-05, "loss": 0.3168, "step": 47615 }, { "epoch": 1.11, "learning_rate": 2.1798244584731642e-05, "loss": 0.0807, "step": 47620 }, { "epoch": 1.11, "learning_rate": 2.1797364530285607e-05, "loss": 0.0608, "step": 47625 }, { "epoch": 1.11, "learning_rate": 2.1796484475839572e-05, "loss": 0.1136, "step": 47630 }, { "epoch": 1.11, "learning_rate": 2.1795604421393537e-05, "loss": 0.1066, "step": 47635 }, { "epoch": 1.11, "learning_rate": 2.1794724366947502e-05, "loss": 0.1853, "step": 47640 }, { "epoch": 1.11, "learning_rate": 2.1793844312501467e-05, "loss": 0.044, "step": 47645 }, { "epoch": 1.11, "learning_rate": 2.1792964258055433e-05, "loss": 0.1715, "step": 47650 }, { "epoch": 1.11, "learning_rate": 2.1792084203609398e-05, "loss": 0.3211, "step": 47655 }, { "epoch": 1.11, "learning_rate": 2.1791204149163363e-05, "loss": 0.338, "step": 47660 }, { "epoch": 1.11, "learning_rate": 2.1790324094717324e-05, "loss": 0.2869, "step": 47665 }, { "epoch": 1.11, "learning_rate": 2.1789444040271293e-05, "loss": 0.0907, "step": 47670 }, { "epoch": 1.11, "learning_rate": 2.1788563985825258e-05, "loss": 0.0794, "step": 47675 }, { "epoch": 1.11, "learning_rate": 2.1787683931379223e-05, "loss": 0.0705, "step": 47680 }, { "epoch": 1.11, "learning_rate": 2.1786979887822394e-05, "loss": 0.1549, "step": 47685 }, { "epoch": 1.11, "learning_rate": 2.178609983337636e-05, "loss": 0.1545, "step": 47690 }, { "epoch": 1.11, "learning_rate": 2.1785219778930324e-05, "loss": 0.1482, "step": 47695 }, { "epoch": 1.11, "learning_rate": 2.178433972448429e-05, "loss": 0.234, "step": 47700 }, { "epoch": 1.11, "learning_rate": 2.1783459670038254e-05, "loss": 0.3162, "step": 47705 }, { "epoch": 1.11, "learning_rate": 2.178257961559222e-05, "loss": 0.4187, "step": 47710 }, { "epoch": 1.11, "learning_rate": 2.1781699561146184e-05, "loss": 0.257, "step": 47715 }, { "epoch": 1.11, "learning_rate": 2.1780819506700146e-05, "loss": 0.0559, "step": 47720 }, { "epoch": 1.11, "learning_rate": 2.1779939452254114e-05, "loss": 0.0784, "step": 47725 }, { "epoch": 1.11, "learning_rate": 2.177905939780808e-05, "loss": 0.1226, "step": 47730 }, { "epoch": 1.11, "learning_rate": 2.1778179343362044e-05, "loss": 0.1102, "step": 47735 }, { "epoch": 1.11, "learning_rate": 2.1777299288916006e-05, "loss": 0.1334, "step": 47740 }, { "epoch": 1.11, "learning_rate": 2.1776419234469974e-05, "loss": 0.1915, "step": 47745 }, { "epoch": 1.11, "learning_rate": 2.177553918002394e-05, "loss": 0.1538, "step": 47750 }, { "epoch": 1.11, "learning_rate": 2.17746591255779e-05, "loss": 0.2564, "step": 47755 }, { "epoch": 1.11, "learning_rate": 2.1773779071131866e-05, "loss": 0.912, "step": 47760 }, { "epoch": 1.11, "learning_rate": 2.1772899016685834e-05, "loss": 0.3594, "step": 47765 }, { "epoch": 1.11, "learning_rate": 2.17720189622398e-05, "loss": 0.0298, "step": 47770 }, { "epoch": 1.11, "learning_rate": 2.177113890779376e-05, "loss": 0.0678, "step": 47775 }, { "epoch": 1.11, "learning_rate": 2.1770258853347726e-05, "loss": 0.1821, "step": 47780 }, { "epoch": 1.11, "learning_rate": 2.1769378798901694e-05, "loss": 0.0533, "step": 47785 }, { "epoch": 1.11, "learning_rate": 2.1768498744455656e-05, "loss": 0.1427, "step": 47790 }, { "epoch": 1.12, "learning_rate": 2.176761869000962e-05, "loss": 0.1624, "step": 47795 }, { "epoch": 1.12, "learning_rate": 2.176673863556359e-05, "loss": 0.2129, "step": 47800 }, { "epoch": 1.12, "learning_rate": 2.1765858581117554e-05, "loss": 0.1751, "step": 47805 }, { "epoch": 1.12, "learning_rate": 2.1764978526671516e-05, "loss": 0.4622, "step": 47810 }, { "epoch": 1.12, "learning_rate": 2.176409847222548e-05, "loss": 0.2802, "step": 47815 }, { "epoch": 1.12, "learning_rate": 2.176321841777945e-05, "loss": 0.0988, "step": 47820 }, { "epoch": 1.12, "learning_rate": 2.176233836333341e-05, "loss": 0.0565, "step": 47825 }, { "epoch": 1.12, "learning_rate": 2.1761458308887376e-05, "loss": 0.0363, "step": 47830 }, { "epoch": 1.12, "learning_rate": 2.176057825444134e-05, "loss": 0.0837, "step": 47835 }, { "epoch": 1.12, "learning_rate": 2.175969819999531e-05, "loss": 0.1538, "step": 47840 }, { "epoch": 1.12, "learning_rate": 2.175881814554927e-05, "loss": 0.174, "step": 47845 }, { "epoch": 1.12, "learning_rate": 2.1757938091103236e-05, "loss": 0.2757, "step": 47850 }, { "epoch": 1.12, "learning_rate": 2.17570580366572e-05, "loss": 0.1839, "step": 47855 }, { "epoch": 1.12, "learning_rate": 2.1756177982211166e-05, "loss": 0.2634, "step": 47860 }, { "epoch": 1.12, "learning_rate": 2.175529792776513e-05, "loss": 0.4725, "step": 47865 }, { "epoch": 1.12, "learning_rate": 2.1754417873319096e-05, "loss": 0.0332, "step": 47870 }, { "epoch": 1.12, "learning_rate": 2.175353781887306e-05, "loss": 0.025, "step": 47875 }, { "epoch": 1.12, "learning_rate": 2.1752657764427026e-05, "loss": 0.1289, "step": 47880 }, { "epoch": 1.12, "learning_rate": 2.175177770998099e-05, "loss": 0.1278, "step": 47885 }, { "epoch": 1.12, "learning_rate": 2.1750897655534956e-05, "loss": 0.0913, "step": 47890 }, { "epoch": 1.12, "learning_rate": 2.175001760108892e-05, "loss": 0.0301, "step": 47895 }, { "epoch": 1.12, "learning_rate": 2.1749137546642886e-05, "loss": 0.2042, "step": 47900 }, { "epoch": 1.12, "learning_rate": 2.174825749219685e-05, "loss": 0.2462, "step": 47905 }, { "epoch": 1.12, "learning_rate": 2.1747377437750816e-05, "loss": 0.3436, "step": 47910 }, { "epoch": 1.12, "learning_rate": 2.1746497383304778e-05, "loss": 0.3072, "step": 47915 }, { "epoch": 1.12, "learning_rate": 2.1745617328858747e-05, "loss": 0.1153, "step": 47920 }, { "epoch": 1.12, "learning_rate": 2.174473727441271e-05, "loss": 0.0523, "step": 47925 }, { "epoch": 1.12, "learning_rate": 2.1743857219966677e-05, "loss": 0.0379, "step": 47930 }, { "epoch": 1.12, "learning_rate": 2.174297716552064e-05, "loss": 0.2454, "step": 47935 }, { "epoch": 1.12, "learning_rate": 2.1742097111074607e-05, "loss": 0.0563, "step": 47940 }, { "epoch": 1.12, "learning_rate": 2.174121705662857e-05, "loss": 0.1569, "step": 47945 }, { "epoch": 1.12, "learning_rate": 2.1740337002182533e-05, "loss": 0.1258, "step": 47950 }, { "epoch": 1.12, "learning_rate": 2.17394569477365e-05, "loss": 0.1223, "step": 47955 }, { "epoch": 1.12, "learning_rate": 2.1738576893290467e-05, "loss": 0.5146, "step": 47960 }, { "epoch": 1.12, "learning_rate": 2.1737696838844432e-05, "loss": 0.1581, "step": 47965 }, { "epoch": 1.12, "learning_rate": 2.1736816784398393e-05, "loss": 0.0969, "step": 47970 }, { "epoch": 1.12, "learning_rate": 2.1735936729952362e-05, "loss": 0.1224, "step": 47975 }, { "epoch": 1.12, "learning_rate": 2.1735056675506327e-05, "loss": 0.0484, "step": 47980 }, { "epoch": 1.12, "learning_rate": 2.173417662106029e-05, "loss": 0.1088, "step": 47985 }, { "epoch": 1.12, "learning_rate": 2.1733296566614253e-05, "loss": 0.1687, "step": 47990 }, { "epoch": 1.12, "learning_rate": 2.1732416512168222e-05, "loss": 0.242, "step": 47995 }, { "epoch": 1.12, "learning_rate": 2.1731536457722187e-05, "loss": 0.1748, "step": 48000 }, { "epoch": 1.12, "learning_rate": 2.173065640327615e-05, "loss": 0.3084, "step": 48005 }, { "epoch": 1.12, "learning_rate": 2.1729776348830114e-05, "loss": 0.482, "step": 48010 }, { "epoch": 1.12, "learning_rate": 2.1728896294384082e-05, "loss": 0.2644, "step": 48015 }, { "epoch": 1.12, "learning_rate": 2.1728016239938044e-05, "loss": 0.0753, "step": 48020 }, { "epoch": 1.12, "learning_rate": 2.172713618549201e-05, "loss": 0.0252, "step": 48025 }, { "epoch": 1.12, "learning_rate": 2.1726256131045974e-05, "loss": 0.1284, "step": 48030 }, { "epoch": 1.12, "learning_rate": 2.1725376076599942e-05, "loss": 0.0454, "step": 48035 }, { "epoch": 1.12, "learning_rate": 2.1724496022153904e-05, "loss": 0.095, "step": 48040 }, { "epoch": 1.12, "learning_rate": 2.172361596770787e-05, "loss": 0.0939, "step": 48045 }, { "epoch": 1.12, "learning_rate": 2.1722735913261834e-05, "loss": 0.2369, "step": 48050 }, { "epoch": 1.12, "learning_rate": 2.17218558588158e-05, "loss": 0.3424, "step": 48055 }, { "epoch": 1.12, "learning_rate": 2.1720975804369764e-05, "loss": 0.455, "step": 48060 }, { "epoch": 1.12, "learning_rate": 2.172009574992373e-05, "loss": 0.485, "step": 48065 }, { "epoch": 1.12, "learning_rate": 2.1719215695477697e-05, "loss": 0.0379, "step": 48070 }, { "epoch": 1.12, "learning_rate": 2.171833564103166e-05, "loss": 0.0317, "step": 48075 }, { "epoch": 1.12, "learning_rate": 2.1717455586585624e-05, "loss": 0.0876, "step": 48080 }, { "epoch": 1.12, "learning_rate": 2.171657553213959e-05, "loss": 0.2704, "step": 48085 }, { "epoch": 1.12, "learning_rate": 2.1715695477693554e-05, "loss": 0.0852, "step": 48090 }, { "epoch": 1.12, "learning_rate": 2.171481542324752e-05, "loss": 0.0732, "step": 48095 }, { "epoch": 1.12, "learning_rate": 2.1713935368801484e-05, "loss": 0.4057, "step": 48100 }, { "epoch": 1.12, "learning_rate": 2.171305531435545e-05, "loss": 0.2853, "step": 48105 }, { "epoch": 1.12, "learning_rate": 2.1712175259909414e-05, "loss": 0.3318, "step": 48110 }, { "epoch": 1.12, "learning_rate": 2.171129520546338e-05, "loss": 0.211, "step": 48115 }, { "epoch": 1.12, "learning_rate": 2.1710415151017344e-05, "loss": 0.0101, "step": 48120 }, { "epoch": 1.12, "learning_rate": 2.1709535096571306e-05, "loss": 0.102, "step": 48125 }, { "epoch": 1.12, "learning_rate": 2.1708655042125274e-05, "loss": 0.1765, "step": 48130 }, { "epoch": 1.12, "learning_rate": 2.170777498767924e-05, "loss": 0.0406, "step": 48135 }, { "epoch": 1.12, "learning_rate": 2.1706894933233204e-05, "loss": 0.1047, "step": 48140 }, { "epoch": 1.12, "learning_rate": 2.1706014878787166e-05, "loss": 0.0584, "step": 48145 }, { "epoch": 1.12, "learning_rate": 2.1705134824341134e-05, "loss": 0.1366, "step": 48150 }, { "epoch": 1.12, "learning_rate": 2.17042547698951e-05, "loss": 0.1125, "step": 48155 }, { "epoch": 1.12, "learning_rate": 2.170337471544906e-05, "loss": 0.6109, "step": 48160 }, { "epoch": 1.12, "learning_rate": 2.1702494661003026e-05, "loss": 0.3919, "step": 48165 }, { "epoch": 1.12, "learning_rate": 2.1701614606556994e-05, "loss": 0.0786, "step": 48170 }, { "epoch": 1.12, "learning_rate": 2.170073455211096e-05, "loss": 0.0545, "step": 48175 }, { "epoch": 1.12, "learning_rate": 2.169985449766492e-05, "loss": 0.165, "step": 48180 }, { "epoch": 1.12, "learning_rate": 2.1698974443218886e-05, "loss": 0.0793, "step": 48185 }, { "epoch": 1.12, "learning_rate": 2.1698094388772854e-05, "loss": 0.0769, "step": 48190 }, { "epoch": 1.12, "learning_rate": 2.169721433432682e-05, "loss": 0.1791, "step": 48195 }, { "epoch": 1.12, "learning_rate": 2.169633427988078e-05, "loss": 0.18, "step": 48200 }, { "epoch": 1.12, "learning_rate": 2.169545422543475e-05, "loss": 0.1964, "step": 48205 }, { "epoch": 1.12, "learning_rate": 2.1694574170988714e-05, "loss": 0.3537, "step": 48210 }, { "epoch": 1.12, "learning_rate": 2.1693694116542676e-05, "loss": 0.3144, "step": 48215 }, { "epoch": 1.13, "learning_rate": 2.169281406209664e-05, "loss": 0.0278, "step": 48220 }, { "epoch": 1.13, "learning_rate": 2.169193400765061e-05, "loss": 0.1255, "step": 48225 }, { "epoch": 1.13, "learning_rate": 2.1691053953204574e-05, "loss": 0.1186, "step": 48230 }, { "epoch": 1.13, "learning_rate": 2.1690173898758536e-05, "loss": 0.0747, "step": 48235 }, { "epoch": 1.13, "learning_rate": 2.16892938443125e-05, "loss": 0.1528, "step": 48240 }, { "epoch": 1.13, "learning_rate": 2.168841378986647e-05, "loss": 0.0736, "step": 48245 }, { "epoch": 1.13, "learning_rate": 2.168753373542043e-05, "loss": 0.1929, "step": 48250 }, { "epoch": 1.13, "learning_rate": 2.1686653680974396e-05, "loss": 0.1708, "step": 48255 }, { "epoch": 1.13, "learning_rate": 2.168577362652836e-05, "loss": 0.5523, "step": 48260 }, { "epoch": 1.13, "learning_rate": 2.168489357208233e-05, "loss": 0.2391, "step": 48265 }, { "epoch": 1.13, "learning_rate": 2.168401351763629e-05, "loss": 0.0904, "step": 48270 }, { "epoch": 1.13, "learning_rate": 2.1683133463190256e-05, "loss": 0.1066, "step": 48275 }, { "epoch": 1.13, "learning_rate": 2.168225340874422e-05, "loss": 0.0854, "step": 48280 }, { "epoch": 1.13, "learning_rate": 2.1681373354298186e-05, "loss": 0.0853, "step": 48285 }, { "epoch": 1.13, "learning_rate": 2.168049329985215e-05, "loss": 0.185, "step": 48290 }, { "epoch": 1.13, "learning_rate": 2.1679613245406116e-05, "loss": 0.224, "step": 48295 }, { "epoch": 1.13, "learning_rate": 2.167873319096008e-05, "loss": 0.1487, "step": 48300 }, { "epoch": 1.13, "learning_rate": 2.1677853136514046e-05, "loss": 0.1923, "step": 48305 }, { "epoch": 1.13, "learning_rate": 2.167697308206801e-05, "loss": 0.2428, "step": 48310 }, { "epoch": 1.13, "learning_rate": 2.1676093027621976e-05, "loss": 0.3492, "step": 48315 }, { "epoch": 1.13, "learning_rate": 2.1675212973175938e-05, "loss": 0.1006, "step": 48320 }, { "epoch": 1.13, "learning_rate": 2.1674332918729906e-05, "loss": 0.066, "step": 48325 }, { "epoch": 1.13, "learning_rate": 2.167345286428387e-05, "loss": 0.0668, "step": 48330 }, { "epoch": 1.13, "learning_rate": 2.1672572809837836e-05, "loss": 0.1606, "step": 48335 }, { "epoch": 1.13, "learning_rate": 2.16716927553918e-05, "loss": 0.1013, "step": 48340 }, { "epoch": 1.13, "learning_rate": 2.1670812700945767e-05, "loss": 0.1699, "step": 48345 }, { "epoch": 1.13, "learning_rate": 2.166993264649973e-05, "loss": 0.3038, "step": 48350 }, { "epoch": 1.13, "learning_rate": 2.1669052592053693e-05, "loss": 0.2569, "step": 48355 }, { "epoch": 1.13, "learning_rate": 2.166817253760766e-05, "loss": 0.5147, "step": 48360 }, { "epoch": 1.13, "learning_rate": 2.1667292483161627e-05, "loss": 0.2173, "step": 48365 }, { "epoch": 1.13, "learning_rate": 2.166641242871559e-05, "loss": 0.0971, "step": 48370 }, { "epoch": 1.13, "learning_rate": 2.1665532374269553e-05, "loss": 0.0279, "step": 48375 }, { "epoch": 1.13, "learning_rate": 2.166465231982352e-05, "loss": 0.0846, "step": 48380 }, { "epoch": 1.13, "learning_rate": 2.1663772265377487e-05, "loss": 0.1554, "step": 48385 }, { "epoch": 1.13, "learning_rate": 2.1662892210931448e-05, "loss": 0.0887, "step": 48390 }, { "epoch": 1.13, "learning_rate": 2.1662012156485413e-05, "loss": 0.1419, "step": 48395 }, { "epoch": 1.13, "learning_rate": 2.1661132102039382e-05, "loss": 0.1503, "step": 48400 }, { "epoch": 1.13, "learning_rate": 2.1660252047593347e-05, "loss": 0.2451, "step": 48405 }, { "epoch": 1.13, "learning_rate": 2.165937199314731e-05, "loss": 0.4509, "step": 48410 }, { "epoch": 1.13, "learning_rate": 2.1658491938701273e-05, "loss": 0.2279, "step": 48415 }, { "epoch": 1.13, "learning_rate": 2.1657611884255242e-05, "loss": 0.0285, "step": 48420 }, { "epoch": 1.13, "learning_rate": 2.1656731829809203e-05, "loss": 0.0667, "step": 48425 }, { "epoch": 1.13, "learning_rate": 2.165585177536317e-05, "loss": 0.0642, "step": 48430 }, { "epoch": 1.13, "learning_rate": 2.1654971720917133e-05, "loss": 0.088, "step": 48435 }, { "epoch": 1.13, "learning_rate": 2.1654091666471102e-05, "loss": 0.0255, "step": 48440 }, { "epoch": 1.13, "learning_rate": 2.1653211612025064e-05, "loss": 0.0771, "step": 48445 }, { "epoch": 1.13, "learning_rate": 2.165233155757903e-05, "loss": 0.1138, "step": 48450 }, { "epoch": 1.13, "learning_rate": 2.1651451503132994e-05, "loss": 0.1551, "step": 48455 }, { "epoch": 1.13, "learning_rate": 2.165057144868696e-05, "loss": 0.4292, "step": 48460 }, { "epoch": 1.13, "learning_rate": 2.1649691394240924e-05, "loss": 0.4222, "step": 48465 }, { "epoch": 1.13, "learning_rate": 2.164881133979489e-05, "loss": 0.0294, "step": 48470 }, { "epoch": 1.13, "learning_rate": 2.1647931285348857e-05, "loss": 0.1314, "step": 48475 }, { "epoch": 1.13, "learning_rate": 2.164705123090282e-05, "loss": 0.1734, "step": 48480 }, { "epoch": 1.13, "learning_rate": 2.1646171176456784e-05, "loss": 0.1281, "step": 48485 }, { "epoch": 1.13, "learning_rate": 2.164529112201075e-05, "loss": 0.1919, "step": 48490 }, { "epoch": 1.13, "learning_rate": 2.1644411067564717e-05, "loss": 0.2523, "step": 48495 }, { "epoch": 1.13, "learning_rate": 2.164353101311868e-05, "loss": 0.2346, "step": 48500 }, { "epoch": 1.13, "learning_rate": 2.1642650958672644e-05, "loss": 0.2694, "step": 48505 }, { "epoch": 1.13, "learning_rate": 2.164177090422661e-05, "loss": 0.2778, "step": 48510 }, { "epoch": 1.13, "learning_rate": 2.1640890849780574e-05, "loss": 0.2383, "step": 48515 }, { "epoch": 1.13, "learning_rate": 2.164001079533454e-05, "loss": 0.1186, "step": 48520 }, { "epoch": 1.13, "learning_rate": 2.1639130740888504e-05, "loss": 0.1241, "step": 48525 }, { "epoch": 1.13, "learning_rate": 2.163825068644247e-05, "loss": 0.0733, "step": 48530 }, { "epoch": 1.13, "learning_rate": 2.1637370631996434e-05, "loss": 0.1945, "step": 48535 }, { "epoch": 1.13, "learning_rate": 2.16364905775504e-05, "loss": 0.0935, "step": 48540 }, { "epoch": 1.13, "learning_rate": 2.1635610523104364e-05, "loss": 0.281, "step": 48545 }, { "epoch": 1.13, "learning_rate": 2.1634730468658326e-05, "loss": 0.1017, "step": 48550 }, { "epoch": 1.13, "learning_rate": 2.1633850414212294e-05, "loss": 0.3259, "step": 48555 }, { "epoch": 1.13, "learning_rate": 2.163297035976626e-05, "loss": 0.3227, "step": 48560 }, { "epoch": 1.13, "learning_rate": 2.1632090305320224e-05, "loss": 0.5006, "step": 48565 }, { "epoch": 1.13, "learning_rate": 2.1631210250874186e-05, "loss": 0.1234, "step": 48570 }, { "epoch": 1.13, "learning_rate": 2.1630330196428154e-05, "loss": 0.0816, "step": 48575 }, { "epoch": 1.13, "learning_rate": 2.162945014198212e-05, "loss": 0.0507, "step": 48580 }, { "epoch": 1.13, "learning_rate": 2.162857008753608e-05, "loss": 0.1069, "step": 48585 }, { "epoch": 1.13, "learning_rate": 2.1627690033090046e-05, "loss": 0.2327, "step": 48590 }, { "epoch": 1.13, "learning_rate": 2.1626809978644014e-05, "loss": 0.1171, "step": 48595 }, { "epoch": 1.13, "learning_rate": 2.162592992419798e-05, "loss": 0.2503, "step": 48600 }, { "epoch": 1.13, "learning_rate": 2.162504986975194e-05, "loss": 0.2917, "step": 48605 }, { "epoch": 1.13, "learning_rate": 2.162416981530591e-05, "loss": 0.3697, "step": 48610 }, { "epoch": 1.13, "learning_rate": 2.1623289760859874e-05, "loss": 0.2579, "step": 48615 }, { "epoch": 1.13, "learning_rate": 2.1622409706413836e-05, "loss": 0.1433, "step": 48620 }, { "epoch": 1.13, "learning_rate": 2.16215296519678e-05, "loss": 0.0937, "step": 48625 }, { "epoch": 1.13, "learning_rate": 2.162064959752177e-05, "loss": 0.035, "step": 48630 }, { "epoch": 1.13, "learning_rate": 2.1619769543075734e-05, "loss": 0.0239, "step": 48635 }, { "epoch": 1.13, "learning_rate": 2.1618889488629696e-05, "loss": 0.1541, "step": 48640 }, { "epoch": 1.13, "learning_rate": 2.161800943418366e-05, "loss": 0.2584, "step": 48645 }, { "epoch": 1.14, "learning_rate": 2.161712937973763e-05, "loss": 0.1129, "step": 48650 }, { "epoch": 1.14, "learning_rate": 2.161624932529159e-05, "loss": 0.1405, "step": 48655 }, { "epoch": 1.14, "learning_rate": 2.1615369270845556e-05, "loss": 0.4774, "step": 48660 }, { "epoch": 1.14, "learning_rate": 2.161448921639952e-05, "loss": 0.241, "step": 48665 }, { "epoch": 1.14, "learning_rate": 2.161360916195349e-05, "loss": 0.0231, "step": 48670 }, { "epoch": 1.14, "learning_rate": 2.161272910750745e-05, "loss": 0.068, "step": 48675 }, { "epoch": 1.14, "learning_rate": 2.1611849053061416e-05, "loss": 0.09, "step": 48680 }, { "epoch": 1.14, "learning_rate": 2.161096899861538e-05, "loss": 0.1032, "step": 48685 }, { "epoch": 1.14, "learning_rate": 2.1610088944169346e-05, "loss": 0.2361, "step": 48690 }, { "epoch": 1.14, "learning_rate": 2.160920888972331e-05, "loss": 0.2218, "step": 48695 }, { "epoch": 1.14, "learning_rate": 2.1608328835277276e-05, "loss": 0.252, "step": 48700 }, { "epoch": 1.14, "learning_rate": 2.160744878083124e-05, "loss": 0.1688, "step": 48705 }, { "epoch": 1.14, "learning_rate": 2.1606568726385206e-05, "loss": 0.2418, "step": 48710 }, { "epoch": 1.14, "learning_rate": 2.160568867193917e-05, "loss": 0.1581, "step": 48715 }, { "epoch": 1.14, "learning_rate": 2.1604808617493136e-05, "loss": 0.1509, "step": 48720 }, { "epoch": 1.14, "learning_rate": 2.1603928563047098e-05, "loss": 0.0377, "step": 48725 }, { "epoch": 1.14, "learning_rate": 2.1603048508601066e-05, "loss": 0.0289, "step": 48730 }, { "epoch": 1.14, "learning_rate": 2.160216845415503e-05, "loss": 0.0991, "step": 48735 }, { "epoch": 1.14, "learning_rate": 2.1601288399708996e-05, "loss": 0.1994, "step": 48740 }, { "epoch": 1.14, "learning_rate": 2.160040834526296e-05, "loss": 0.2063, "step": 48745 }, { "epoch": 1.14, "learning_rate": 2.1599528290816926e-05, "loss": 0.2013, "step": 48750 }, { "epoch": 1.14, "learning_rate": 2.159864823637089e-05, "loss": 0.068, "step": 48755 }, { "epoch": 1.14, "learning_rate": 2.1597768181924853e-05, "loss": 0.3281, "step": 48760 }, { "epoch": 1.14, "learning_rate": 2.159688812747882e-05, "loss": 0.4133, "step": 48765 }, { "epoch": 1.14, "learning_rate": 2.1596008073032786e-05, "loss": 0.0573, "step": 48770 }, { "epoch": 1.14, "learning_rate": 2.159512801858675e-05, "loss": 0.0429, "step": 48775 }, { "epoch": 1.14, "learning_rate": 2.1594247964140713e-05, "loss": 0.1258, "step": 48780 }, { "epoch": 1.14, "learning_rate": 2.159336790969468e-05, "loss": 0.1268, "step": 48785 }, { "epoch": 1.14, "learning_rate": 2.1592487855248647e-05, "loss": 0.2096, "step": 48790 }, { "epoch": 1.14, "learning_rate": 2.1591607800802608e-05, "loss": 0.0748, "step": 48795 }, { "epoch": 1.14, "learning_rate": 2.1590727746356573e-05, "loss": 0.1322, "step": 48800 }, { "epoch": 1.14, "learning_rate": 2.158984769191054e-05, "loss": 0.1854, "step": 48805 }, { "epoch": 1.14, "learning_rate": 2.1588967637464507e-05, "loss": 0.1937, "step": 48810 }, { "epoch": 1.14, "learning_rate": 2.1588087583018468e-05, "loss": 0.3701, "step": 48815 }, { "epoch": 1.14, "learning_rate": 2.1587207528572433e-05, "loss": 0.0813, "step": 48820 }, { "epoch": 1.14, "learning_rate": 2.1586327474126402e-05, "loss": 0.0694, "step": 48825 }, { "epoch": 1.14, "learning_rate": 2.1585447419680367e-05, "loss": 0.228, "step": 48830 }, { "epoch": 1.14, "learning_rate": 2.158456736523433e-05, "loss": 0.1353, "step": 48835 }, { "epoch": 1.14, "learning_rate": 2.1583687310788293e-05, "loss": 0.2455, "step": 48840 }, { "epoch": 1.14, "learning_rate": 2.1582807256342262e-05, "loss": 0.2712, "step": 48845 }, { "epoch": 1.14, "learning_rate": 2.1581927201896223e-05, "loss": 0.2389, "step": 48850 }, { "epoch": 1.14, "learning_rate": 2.158104714745019e-05, "loss": 0.4523, "step": 48855 }, { "epoch": 1.14, "learning_rate": 2.1580167093004153e-05, "loss": 0.395, "step": 48860 }, { "epoch": 1.14, "learning_rate": 2.1579287038558122e-05, "loss": 0.403, "step": 48865 }, { "epoch": 1.14, "learning_rate": 2.1578406984112083e-05, "loss": 0.1217, "step": 48870 }, { "epoch": 1.14, "learning_rate": 2.157752692966605e-05, "loss": 0.0874, "step": 48875 }, { "epoch": 1.14, "learning_rate": 2.1576646875220014e-05, "loss": 0.0997, "step": 48880 }, { "epoch": 1.14, "learning_rate": 2.157576682077398e-05, "loss": 0.2097, "step": 48885 }, { "epoch": 1.14, "learning_rate": 2.1574886766327944e-05, "loss": 0.1399, "step": 48890 }, { "epoch": 1.14, "learning_rate": 2.157400671188191e-05, "loss": 0.108, "step": 48895 }, { "epoch": 1.14, "learning_rate": 2.1573126657435877e-05, "loss": 0.1691, "step": 48900 }, { "epoch": 1.14, "learning_rate": 2.157224660298984e-05, "loss": 0.1626, "step": 48905 }, { "epoch": 1.14, "learning_rate": 2.1571366548543804e-05, "loss": 0.3249, "step": 48910 }, { "epoch": 1.14, "learning_rate": 2.157048649409777e-05, "loss": 0.3391, "step": 48915 }, { "epoch": 1.14, "learning_rate": 2.1569606439651734e-05, "loss": 0.0408, "step": 48920 }, { "epoch": 1.14, "learning_rate": 2.15687263852057e-05, "loss": 0.0336, "step": 48925 }, { "epoch": 1.14, "learning_rate": 2.1567846330759664e-05, "loss": 0.1111, "step": 48930 }, { "epoch": 1.14, "learning_rate": 2.156696627631363e-05, "loss": 0.0989, "step": 48935 }, { "epoch": 1.14, "learning_rate": 2.1566086221867594e-05, "loss": 0.1332, "step": 48940 }, { "epoch": 1.14, "learning_rate": 2.156520616742156e-05, "loss": 0.1246, "step": 48945 }, { "epoch": 1.14, "learning_rate": 2.1564326112975524e-05, "loss": 0.1518, "step": 48950 }, { "epoch": 1.14, "learning_rate": 2.1563446058529485e-05, "loss": 0.1298, "step": 48955 }, { "epoch": 1.14, "learning_rate": 2.1562566004083454e-05, "loss": 0.4183, "step": 48960 }, { "epoch": 1.14, "learning_rate": 2.156168594963742e-05, "loss": 0.3238, "step": 48965 }, { "epoch": 1.14, "learning_rate": 2.1560805895191384e-05, "loss": 0.2164, "step": 48970 }, { "epoch": 1.14, "learning_rate": 2.1559925840745346e-05, "loss": 0.0625, "step": 48975 }, { "epoch": 1.14, "learning_rate": 2.1559045786299314e-05, "loss": 0.0644, "step": 48980 }, { "epoch": 1.14, "learning_rate": 2.155816573185328e-05, "loss": 0.0835, "step": 48985 }, { "epoch": 1.14, "learning_rate": 2.155728567740724e-05, "loss": 0.0982, "step": 48990 }, { "epoch": 1.14, "learning_rate": 2.1556405622961206e-05, "loss": 0.0766, "step": 48995 }, { "epoch": 1.14, "learning_rate": 2.1555525568515174e-05, "loss": 0.1566, "step": 49000 }, { "epoch": 1.14, "learning_rate": 2.155464551406914e-05, "loss": 0.2503, "step": 49005 }, { "epoch": 1.14, "learning_rate": 2.15537654596231e-05, "loss": 0.5075, "step": 49010 }, { "epoch": 1.14, "learning_rate": 2.155288540517707e-05, "loss": 0.4049, "step": 49015 }, { "epoch": 1.14, "learning_rate": 2.1552005350731034e-05, "loss": 0.0231, "step": 49020 }, { "epoch": 1.14, "learning_rate": 2.1551125296284996e-05, "loss": 0.0489, "step": 49025 }, { "epoch": 1.14, "learning_rate": 2.155024524183896e-05, "loss": 0.0847, "step": 49030 }, { "epoch": 1.14, "learning_rate": 2.154936518739293e-05, "loss": 0.0252, "step": 49035 }, { "epoch": 1.14, "learning_rate": 2.1548485132946894e-05, "loss": 0.1755, "step": 49040 }, { "epoch": 1.14, "learning_rate": 2.1547605078500856e-05, "loss": 0.2166, "step": 49045 }, { "epoch": 1.14, "learning_rate": 2.154672502405482e-05, "loss": 0.2277, "step": 49050 }, { "epoch": 1.14, "learning_rate": 2.154584496960879e-05, "loss": 0.2829, "step": 49055 }, { "epoch": 1.14, "learning_rate": 2.154496491516275e-05, "loss": 0.1953, "step": 49060 }, { "epoch": 1.14, "learning_rate": 2.1544084860716716e-05, "loss": 0.4055, "step": 49065 }, { "epoch": 1.14, "learning_rate": 2.154320480627068e-05, "loss": 0.1367, "step": 49070 }, { "epoch": 1.14, "learning_rate": 2.154232475182465e-05, "loss": 0.0511, "step": 49075 }, { "epoch": 1.15, "learning_rate": 2.154144469737861e-05, "loss": 0.1431, "step": 49080 }, { "epoch": 1.15, "learning_rate": 2.1540564642932576e-05, "loss": 0.0687, "step": 49085 }, { "epoch": 1.15, "learning_rate": 2.153968458848654e-05, "loss": 0.1189, "step": 49090 }, { "epoch": 1.15, "learning_rate": 2.1538804534040506e-05, "loss": 0.1883, "step": 49095 }, { "epoch": 1.15, "learning_rate": 2.153792447959447e-05, "loss": 0.1231, "step": 49100 }, { "epoch": 1.15, "learning_rate": 2.1537044425148436e-05, "loss": 0.1536, "step": 49105 }, { "epoch": 1.15, "learning_rate": 2.15361643707024e-05, "loss": 0.2998, "step": 49110 }, { "epoch": 1.15, "learning_rate": 2.1535284316256366e-05, "loss": 0.416, "step": 49115 }, { "epoch": 1.15, "learning_rate": 2.153440426181033e-05, "loss": 0.061, "step": 49120 }, { "epoch": 1.15, "learning_rate": 2.1533524207364296e-05, "loss": 0.0485, "step": 49125 }, { "epoch": 1.15, "learning_rate": 2.153264415291826e-05, "loss": 0.0507, "step": 49130 }, { "epoch": 1.15, "learning_rate": 2.1531764098472226e-05, "loss": 0.1619, "step": 49135 }, { "epoch": 1.15, "learning_rate": 2.153088404402619e-05, "loss": 0.353, "step": 49140 }, { "epoch": 1.15, "learning_rate": 2.1530003989580156e-05, "loss": 0.0758, "step": 49145 }, { "epoch": 1.15, "learning_rate": 2.152912393513412e-05, "loss": 0.2088, "step": 49150 }, { "epoch": 1.15, "learning_rate": 2.1528243880688086e-05, "loss": 0.2128, "step": 49155 }, { "epoch": 1.15, "learning_rate": 2.152736382624205e-05, "loss": 0.3463, "step": 49160 }, { "epoch": 1.15, "learning_rate": 2.1526483771796016e-05, "loss": 0.2167, "step": 49165 }, { "epoch": 1.15, "learning_rate": 2.152560371734998e-05, "loss": 0.0552, "step": 49170 }, { "epoch": 1.15, "learning_rate": 2.1524723662903946e-05, "loss": 0.1456, "step": 49175 }, { "epoch": 1.15, "learning_rate": 2.152384360845791e-05, "loss": 0.0342, "step": 49180 }, { "epoch": 1.15, "learning_rate": 2.1522963554011873e-05, "loss": 0.0721, "step": 49185 }, { "epoch": 1.15, "learning_rate": 2.152208349956584e-05, "loss": 0.1829, "step": 49190 }, { "epoch": 1.15, "learning_rate": 2.1521203445119806e-05, "loss": 0.2142, "step": 49195 }, { "epoch": 1.15, "learning_rate": 2.152032339067377e-05, "loss": 0.2636, "step": 49200 }, { "epoch": 1.15, "learning_rate": 2.1519443336227733e-05, "loss": 0.1658, "step": 49205 }, { "epoch": 1.15, "learning_rate": 2.15185632817817e-05, "loss": 0.195, "step": 49210 }, { "epoch": 1.15, "learning_rate": 2.1517683227335667e-05, "loss": 0.4009, "step": 49215 }, { "epoch": 1.15, "learning_rate": 2.1516803172889628e-05, "loss": 0.0738, "step": 49220 }, { "epoch": 1.15, "learning_rate": 2.1515923118443593e-05, "loss": 0.0871, "step": 49225 }, { "epoch": 1.15, "learning_rate": 2.151504306399756e-05, "loss": 0.0119, "step": 49230 }, { "epoch": 1.15, "learning_rate": 2.1514163009551527e-05, "loss": 0.049, "step": 49235 }, { "epoch": 1.15, "learning_rate": 2.1513282955105488e-05, "loss": 0.0972, "step": 49240 }, { "epoch": 1.15, "learning_rate": 2.1512402900659453e-05, "loss": 0.1638, "step": 49245 }, { "epoch": 1.15, "learning_rate": 2.151152284621342e-05, "loss": 0.1893, "step": 49250 }, { "epoch": 1.15, "learning_rate": 2.1510642791767383e-05, "loss": 0.2397, "step": 49255 }, { "epoch": 1.15, "learning_rate": 2.150976273732135e-05, "loss": 0.2623, "step": 49260 }, { "epoch": 1.15, "learning_rate": 2.1508882682875313e-05, "loss": 0.4272, "step": 49265 }, { "epoch": 1.15, "learning_rate": 2.1508002628429282e-05, "loss": 0.1277, "step": 49270 }, { "epoch": 1.15, "learning_rate": 2.1507122573983243e-05, "loss": 0.0796, "step": 49275 }, { "epoch": 1.15, "learning_rate": 2.150624251953721e-05, "loss": 0.1627, "step": 49280 }, { "epoch": 1.15, "learning_rate": 2.1505362465091173e-05, "loss": 0.1747, "step": 49285 }, { "epoch": 1.15, "learning_rate": 2.150448241064514e-05, "loss": 0.1551, "step": 49290 }, { "epoch": 1.15, "learning_rate": 2.1503602356199103e-05, "loss": 0.2183, "step": 49295 }, { "epoch": 1.15, "learning_rate": 2.150272230175307e-05, "loss": 0.0577, "step": 49300 }, { "epoch": 1.15, "learning_rate": 2.1501842247307037e-05, "loss": 0.2719, "step": 49305 }, { "epoch": 1.15, "learning_rate": 2.1500962192861e-05, "loss": 0.2602, "step": 49310 }, { "epoch": 1.15, "learning_rate": 2.1500082138414964e-05, "loss": 0.1286, "step": 49315 }, { "epoch": 1.15, "learning_rate": 2.149920208396893e-05, "loss": 0.0623, "step": 49320 }, { "epoch": 1.15, "learning_rate": 2.1498322029522894e-05, "loss": 0.0687, "step": 49325 }, { "epoch": 1.15, "learning_rate": 2.149744197507686e-05, "loss": 0.0483, "step": 49330 }, { "epoch": 1.15, "learning_rate": 2.1496561920630824e-05, "loss": 0.1462, "step": 49335 }, { "epoch": 1.15, "learning_rate": 2.149568186618479e-05, "loss": 0.069, "step": 49340 }, { "epoch": 1.15, "learning_rate": 2.1494801811738754e-05, "loss": 0.167, "step": 49345 }, { "epoch": 1.15, "learning_rate": 2.149392175729272e-05, "loss": 0.0815, "step": 49350 }, { "epoch": 1.15, "learning_rate": 2.1493041702846684e-05, "loss": 0.2188, "step": 49355 }, { "epoch": 1.15, "learning_rate": 2.1492161648400645e-05, "loss": 0.5502, "step": 49360 }, { "epoch": 1.15, "learning_rate": 2.1491281593954614e-05, "loss": 0.2622, "step": 49365 }, { "epoch": 1.15, "learning_rate": 2.149040153950858e-05, "loss": 0.0582, "step": 49370 }, { "epoch": 1.15, "learning_rate": 2.1489521485062544e-05, "loss": 0.1031, "step": 49375 }, { "epoch": 1.15, "learning_rate": 2.1488641430616505e-05, "loss": 0.0478, "step": 49380 }, { "epoch": 1.15, "learning_rate": 2.1487761376170474e-05, "loss": 0.1112, "step": 49385 }, { "epoch": 1.15, "learning_rate": 2.148688132172444e-05, "loss": 0.2243, "step": 49390 }, { "epoch": 1.15, "learning_rate": 2.14860012672784e-05, "loss": 0.1234, "step": 49395 }, { "epoch": 1.15, "learning_rate": 2.1485121212832366e-05, "loss": 0.1275, "step": 49400 }, { "epoch": 1.15, "learning_rate": 2.1484241158386334e-05, "loss": 0.3078, "step": 49405 }, { "epoch": 1.15, "learning_rate": 2.14833611039403e-05, "loss": 0.3207, "step": 49410 }, { "epoch": 1.15, "learning_rate": 2.148248104949426e-05, "loss": 0.3891, "step": 49415 }, { "epoch": 1.15, "learning_rate": 2.1481600995048226e-05, "loss": 0.1734, "step": 49420 }, { "epoch": 1.15, "learning_rate": 2.1480720940602194e-05, "loss": 0.0899, "step": 49425 }, { "epoch": 1.15, "learning_rate": 2.147984088615616e-05, "loss": 0.1283, "step": 49430 }, { "epoch": 1.15, "learning_rate": 2.147896083171012e-05, "loss": 0.0826, "step": 49435 }, { "epoch": 1.15, "learning_rate": 2.147808077726409e-05, "loss": 0.1373, "step": 49440 }, { "epoch": 1.15, "learning_rate": 2.1477200722818054e-05, "loss": 0.3221, "step": 49445 }, { "epoch": 1.15, "learning_rate": 2.1476320668372016e-05, "loss": 0.1399, "step": 49450 }, { "epoch": 1.15, "learning_rate": 2.147544061392598e-05, "loss": 0.4558, "step": 49455 }, { "epoch": 1.15, "learning_rate": 2.147456055947995e-05, "loss": 0.2959, "step": 49460 }, { "epoch": 1.15, "learning_rate": 2.1473680505033914e-05, "loss": 0.4288, "step": 49465 }, { "epoch": 1.15, "learning_rate": 2.1472800450587876e-05, "loss": 0.0681, "step": 49470 }, { "epoch": 1.15, "learning_rate": 2.147192039614184e-05, "loss": 0.088, "step": 49475 }, { "epoch": 1.15, "learning_rate": 2.147104034169581e-05, "loss": 0.227, "step": 49480 }, { "epoch": 1.15, "learning_rate": 2.147016028724977e-05, "loss": 0.0814, "step": 49485 }, { "epoch": 1.15, "learning_rate": 2.1469280232803736e-05, "loss": 0.1964, "step": 49490 }, { "epoch": 1.15, "learning_rate": 2.14684001783577e-05, "loss": 0.089, "step": 49495 }, { "epoch": 1.15, "learning_rate": 2.146752012391167e-05, "loss": 0.145, "step": 49500 }, { "epoch": 1.16, "learning_rate": 2.146664006946563e-05, "loss": 0.2406, "step": 49505 }, { "epoch": 1.16, "learning_rate": 2.1465760015019596e-05, "loss": 0.1467, "step": 49510 }, { "epoch": 1.16, "learning_rate": 2.146487996057356e-05, "loss": 0.2855, "step": 49515 }, { "epoch": 1.16, "learning_rate": 2.1463999906127526e-05, "loss": 0.0704, "step": 49520 }, { "epoch": 1.16, "learning_rate": 2.146311985168149e-05, "loss": 0.0286, "step": 49525 }, { "epoch": 1.16, "learning_rate": 2.1462239797235456e-05, "loss": 0.0452, "step": 49530 }, { "epoch": 1.16, "learning_rate": 2.146135974278942e-05, "loss": 0.128, "step": 49535 }, { "epoch": 1.16, "learning_rate": 2.1460479688343386e-05, "loss": 0.1991, "step": 49540 }, { "epoch": 1.16, "learning_rate": 2.145959963389735e-05, "loss": 0.3344, "step": 49545 }, { "epoch": 1.16, "learning_rate": 2.1458719579451316e-05, "loss": 0.1734, "step": 49550 }, { "epoch": 1.16, "learning_rate": 2.145783952500528e-05, "loss": 0.3176, "step": 49555 }, { "epoch": 1.16, "learning_rate": 2.1456959470559246e-05, "loss": 0.4908, "step": 49560 }, { "epoch": 1.16, "learning_rate": 2.145607941611321e-05, "loss": 0.228, "step": 49565 }, { "epoch": 1.16, "learning_rate": 2.1455199361667176e-05, "loss": 0.1379, "step": 49570 }, { "epoch": 1.16, "learning_rate": 2.145431930722114e-05, "loss": 0.0619, "step": 49575 }, { "epoch": 1.16, "learning_rate": 2.1453439252775106e-05, "loss": 0.1331, "step": 49580 }, { "epoch": 1.16, "learning_rate": 2.145255919832907e-05, "loss": 0.0917, "step": 49585 }, { "epoch": 1.16, "learning_rate": 2.1451679143883033e-05, "loss": 0.2006, "step": 49590 }, { "epoch": 1.16, "learning_rate": 2.1450799089437e-05, "loss": 0.3882, "step": 49595 }, { "epoch": 1.16, "learning_rate": 2.1449919034990966e-05, "loss": 0.1913, "step": 49600 }, { "epoch": 1.16, "learning_rate": 2.144903898054493e-05, "loss": 0.1696, "step": 49605 }, { "epoch": 1.16, "learning_rate": 2.1448158926098893e-05, "loss": 0.5641, "step": 49610 }, { "epoch": 1.16, "learning_rate": 2.144727887165286e-05, "loss": 0.4672, "step": 49615 }, { "epoch": 1.16, "learning_rate": 2.1446398817206826e-05, "loss": 0.0072, "step": 49620 }, { "epoch": 1.16, "learning_rate": 2.1445518762760788e-05, "loss": 0.0877, "step": 49625 }, { "epoch": 1.16, "learning_rate": 2.1444638708314753e-05, "loss": 0.0354, "step": 49630 }, { "epoch": 1.16, "learning_rate": 2.144375865386872e-05, "loss": 0.0626, "step": 49635 }, { "epoch": 1.16, "learning_rate": 2.1442878599422686e-05, "loss": 0.052, "step": 49640 }, { "epoch": 1.16, "learning_rate": 2.1441998544976648e-05, "loss": 0.0785, "step": 49645 }, { "epoch": 1.16, "learning_rate": 2.1441118490530613e-05, "loss": 0.1947, "step": 49650 }, { "epoch": 1.16, "learning_rate": 2.144023843608458e-05, "loss": 0.3521, "step": 49655 }, { "epoch": 1.16, "learning_rate": 2.1439358381638543e-05, "loss": 0.2005, "step": 49660 }, { "epoch": 1.16, "learning_rate": 2.1438478327192508e-05, "loss": 0.2587, "step": 49665 }, { "epoch": 1.16, "learning_rate": 2.1437598272746473e-05, "loss": 0.0267, "step": 49670 }, { "epoch": 1.16, "learning_rate": 2.143671821830044e-05, "loss": 0.0344, "step": 49675 }, { "epoch": 1.16, "learning_rate": 2.1435838163854403e-05, "loss": 0.0858, "step": 49680 }, { "epoch": 1.16, "learning_rate": 2.1434958109408368e-05, "loss": 0.217, "step": 49685 }, { "epoch": 1.16, "learning_rate": 2.1434078054962333e-05, "loss": 0.1365, "step": 49690 }, { "epoch": 1.16, "learning_rate": 2.14331980005163e-05, "loss": 0.1496, "step": 49695 }, { "epoch": 1.16, "learning_rate": 2.1432317946070263e-05, "loss": 0.1714, "step": 49700 }, { "epoch": 1.16, "learning_rate": 2.143143789162423e-05, "loss": 0.2051, "step": 49705 }, { "epoch": 1.16, "learning_rate": 2.1430557837178197e-05, "loss": 0.3189, "step": 49710 }, { "epoch": 1.16, "learning_rate": 2.142967778273216e-05, "loss": 0.504, "step": 49715 }, { "epoch": 1.16, "learning_rate": 2.1428797728286123e-05, "loss": 0.2149, "step": 49720 }, { "epoch": 1.16, "learning_rate": 2.142791767384009e-05, "loss": 0.0423, "step": 49725 }, { "epoch": 1.16, "learning_rate": 2.1427037619394057e-05, "loss": 0.0571, "step": 49730 }, { "epoch": 1.16, "learning_rate": 2.142615756494802e-05, "loss": 0.0386, "step": 49735 }, { "epoch": 1.16, "learning_rate": 2.1425277510501984e-05, "loss": 0.108, "step": 49740 }, { "epoch": 1.16, "learning_rate": 2.142439745605595e-05, "loss": 0.0561, "step": 49745 }, { "epoch": 1.16, "learning_rate": 2.1423517401609914e-05, "loss": 0.1596, "step": 49750 }, { "epoch": 1.16, "learning_rate": 2.142263734716388e-05, "loss": 0.1933, "step": 49755 }, { "epoch": 1.16, "learning_rate": 2.1421757292717844e-05, "loss": 0.3219, "step": 49760 }, { "epoch": 1.16, "learning_rate": 2.142087723827181e-05, "loss": 0.6309, "step": 49765 }, { "epoch": 1.16, "learning_rate": 2.1419997183825774e-05, "loss": 0.121, "step": 49770 }, { "epoch": 1.16, "learning_rate": 2.141911712937974e-05, "loss": 0.0165, "step": 49775 }, { "epoch": 1.16, "learning_rate": 2.1418237074933704e-05, "loss": 0.1478, "step": 49780 }, { "epoch": 1.16, "learning_rate": 2.1417357020487665e-05, "loss": 0.0886, "step": 49785 }, { "epoch": 1.16, "learning_rate": 2.1416476966041634e-05, "loss": 0.2034, "step": 49790 }, { "epoch": 1.16, "learning_rate": 2.14155969115956e-05, "loss": 0.2068, "step": 49795 }, { "epoch": 1.16, "learning_rate": 2.1414716857149564e-05, "loss": 0.1746, "step": 49800 }, { "epoch": 1.16, "learning_rate": 2.1413836802703525e-05, "loss": 0.295, "step": 49805 }, { "epoch": 1.16, "learning_rate": 2.1412956748257494e-05, "loss": 0.3874, "step": 49810 }, { "epoch": 1.16, "learning_rate": 2.141207669381146e-05, "loss": 0.2644, "step": 49815 }, { "epoch": 1.16, "learning_rate": 2.141119663936542e-05, "loss": 0.0314, "step": 49820 }, { "epoch": 1.16, "learning_rate": 2.1410316584919385e-05, "loss": 0.076, "step": 49825 }, { "epoch": 1.16, "learning_rate": 2.1409436530473354e-05, "loss": 0.0963, "step": 49830 }, { "epoch": 1.16, "learning_rate": 2.140855647602732e-05, "loss": 0.1511, "step": 49835 }, { "epoch": 1.16, "learning_rate": 2.140767642158128e-05, "loss": 0.115, "step": 49840 }, { "epoch": 1.16, "learning_rate": 2.140679636713525e-05, "loss": 0.1338, "step": 49845 }, { "epoch": 1.16, "learning_rate": 2.1405916312689214e-05, "loss": 0.108, "step": 49850 }, { "epoch": 1.16, "learning_rate": 2.1405036258243176e-05, "loss": 0.1719, "step": 49855 }, { "epoch": 1.16, "learning_rate": 2.140415620379714e-05, "loss": 0.1807, "step": 49860 }, { "epoch": 1.16, "learning_rate": 2.140327614935111e-05, "loss": 0.2633, "step": 49865 }, { "epoch": 1.16, "learning_rate": 2.1402396094905074e-05, "loss": 0.0936, "step": 49870 }, { "epoch": 1.16, "learning_rate": 2.1401516040459036e-05, "loss": 0.0711, "step": 49875 }, { "epoch": 1.16, "learning_rate": 2.1400635986013e-05, "loss": 0.1122, "step": 49880 }, { "epoch": 1.16, "learning_rate": 2.139975593156697e-05, "loss": 0.0424, "step": 49885 }, { "epoch": 1.16, "learning_rate": 2.139887587712093e-05, "loss": 0.1393, "step": 49890 }, { "epoch": 1.16, "learning_rate": 2.1397995822674896e-05, "loss": 0.1548, "step": 49895 }, { "epoch": 1.16, "learning_rate": 2.139711576822886e-05, "loss": 0.1191, "step": 49900 }, { "epoch": 1.16, "learning_rate": 2.139623571378283e-05, "loss": 0.3373, "step": 49905 }, { "epoch": 1.16, "learning_rate": 2.139535565933679e-05, "loss": 0.4507, "step": 49910 }, { "epoch": 1.16, "learning_rate": 2.1394475604890756e-05, "loss": 0.2842, "step": 49915 }, { "epoch": 1.16, "learning_rate": 2.139359555044472e-05, "loss": 0.0953, "step": 49920 }, { "epoch": 1.16, "learning_rate": 2.1392715495998686e-05, "loss": 0.0865, "step": 49925 }, { "epoch": 1.16, "learning_rate": 2.139183544155265e-05, "loss": 0.0229, "step": 49930 }, { "epoch": 1.17, "learning_rate": 2.1390955387106616e-05, "loss": 0.1454, "step": 49935 }, { "epoch": 1.17, "learning_rate": 2.139007533266058e-05, "loss": 0.2326, "step": 49940 }, { "epoch": 1.17, "learning_rate": 2.1389195278214546e-05, "loss": 0.2468, "step": 49945 }, { "epoch": 1.17, "learning_rate": 2.138831522376851e-05, "loss": 0.1919, "step": 49950 }, { "epoch": 1.17, "learning_rate": 2.1387435169322476e-05, "loss": 0.3394, "step": 49955 }, { "epoch": 1.17, "learning_rate": 2.1386555114876438e-05, "loss": 0.4525, "step": 49960 }, { "epoch": 1.17, "learning_rate": 2.1385675060430406e-05, "loss": 0.3973, "step": 49965 }, { "epoch": 1.17, "learning_rate": 2.138479500598437e-05, "loss": 0.0608, "step": 49970 }, { "epoch": 1.17, "learning_rate": 2.1383914951538336e-05, "loss": 0.1416, "step": 49975 }, { "epoch": 1.17, "learning_rate": 2.13830348970923e-05, "loss": 0.0562, "step": 49980 }, { "epoch": 1.17, "learning_rate": 2.1382154842646266e-05, "loss": 0.073, "step": 49985 }, { "epoch": 1.17, "learning_rate": 2.138127478820023e-05, "loss": 0.1703, "step": 49990 }, { "epoch": 1.17, "learning_rate": 2.1380394733754193e-05, "loss": 0.0585, "step": 49995 }, { "epoch": 1.17, "learning_rate": 2.137951467930816e-05, "loss": 0.3716, "step": 50000 }, { "epoch": 1.17, "learning_rate": 2.1378634624862126e-05, "loss": 0.1634, "step": 50005 }, { "epoch": 1.17, "learning_rate": 2.137775457041609e-05, "loss": 0.5058, "step": 50010 }, { "epoch": 1.17, "learning_rate": 2.1376874515970053e-05, "loss": 0.2151, "step": 50015 }, { "epoch": 1.17, "learning_rate": 2.137599446152402e-05, "loss": 0.0762, "step": 50020 }, { "epoch": 1.17, "learning_rate": 2.1375114407077986e-05, "loss": 0.0557, "step": 50025 }, { "epoch": 1.17, "learning_rate": 2.137423435263195e-05, "loss": 0.137, "step": 50030 }, { "epoch": 1.17, "learning_rate": 2.1373354298185913e-05, "loss": 0.1076, "step": 50035 }, { "epoch": 1.17, "learning_rate": 2.137247424373988e-05, "loss": 0.1751, "step": 50040 }, { "epoch": 1.17, "learning_rate": 2.1371594189293846e-05, "loss": 0.1445, "step": 50045 }, { "epoch": 1.17, "learning_rate": 2.1370714134847808e-05, "loss": 0.2025, "step": 50050 }, { "epoch": 1.17, "learning_rate": 2.1369834080401773e-05, "loss": 0.2696, "step": 50055 }, { "epoch": 1.17, "learning_rate": 2.136895402595574e-05, "loss": 0.4908, "step": 50060 }, { "epoch": 1.17, "learning_rate": 2.1368073971509706e-05, "loss": 0.2417, "step": 50065 }, { "epoch": 1.17, "learning_rate": 2.1367193917063668e-05, "loss": 0.0576, "step": 50070 }, { "epoch": 1.17, "learning_rate": 2.1366313862617633e-05, "loss": 0.0334, "step": 50075 }, { "epoch": 1.17, "learning_rate": 2.13654338081716e-05, "loss": 0.0872, "step": 50080 }, { "epoch": 1.17, "learning_rate": 2.1364553753725563e-05, "loss": 0.0404, "step": 50085 }, { "epoch": 1.17, "learning_rate": 2.1363673699279528e-05, "loss": 0.1708, "step": 50090 }, { "epoch": 1.17, "learning_rate": 2.1362793644833493e-05, "loss": 0.2011, "step": 50095 }, { "epoch": 1.17, "learning_rate": 2.136191359038746e-05, "loss": 0.1242, "step": 50100 }, { "epoch": 1.17, "learning_rate": 2.1361033535941423e-05, "loss": 0.2033, "step": 50105 }, { "epoch": 1.17, "learning_rate": 2.1360153481495388e-05, "loss": 0.5232, "step": 50110 }, { "epoch": 1.17, "learning_rate": 2.1359273427049357e-05, "loss": 0.3415, "step": 50115 }, { "epoch": 1.17, "learning_rate": 2.1358393372603318e-05, "loss": 0.047, "step": 50120 }, { "epoch": 1.17, "learning_rate": 2.1357513318157283e-05, "loss": 0.1327, "step": 50125 }, { "epoch": 1.17, "learning_rate": 2.135663326371125e-05, "loss": 0.0638, "step": 50130 }, { "epoch": 1.17, "learning_rate": 2.1355753209265217e-05, "loss": 0.0769, "step": 50135 }, { "epoch": 1.17, "learning_rate": 2.135487315481918e-05, "loss": 0.1802, "step": 50140 }, { "epoch": 1.17, "learning_rate": 2.1353993100373143e-05, "loss": 0.1025, "step": 50145 }, { "epoch": 1.17, "learning_rate": 2.135311304592711e-05, "loss": 0.1087, "step": 50150 }, { "epoch": 1.17, "learning_rate": 2.1352232991481073e-05, "loss": 0.1559, "step": 50155 }, { "epoch": 1.17, "learning_rate": 2.135135293703504e-05, "loss": 0.4583, "step": 50160 }, { "epoch": 1.17, "learning_rate": 2.1350472882589003e-05, "loss": 0.3948, "step": 50165 }, { "epoch": 1.17, "learning_rate": 2.134959282814297e-05, "loss": 0.07, "step": 50170 }, { "epoch": 1.17, "learning_rate": 2.1348712773696934e-05, "loss": 0.0841, "step": 50175 }, { "epoch": 1.17, "learning_rate": 2.13478327192509e-05, "loss": 0.088, "step": 50180 }, { "epoch": 1.17, "learning_rate": 2.1346952664804864e-05, "loss": 0.1074, "step": 50185 }, { "epoch": 1.17, "learning_rate": 2.1346072610358825e-05, "loss": 0.2776, "step": 50190 }, { "epoch": 1.17, "learning_rate": 2.1345192555912794e-05, "loss": 0.2191, "step": 50195 }, { "epoch": 1.17, "learning_rate": 2.134431250146676e-05, "loss": 0.2384, "step": 50200 }, { "epoch": 1.17, "learning_rate": 2.1343432447020724e-05, "loss": 0.2697, "step": 50205 }, { "epoch": 1.17, "learning_rate": 2.1342552392574685e-05, "loss": 0.2206, "step": 50210 }, { "epoch": 1.17, "learning_rate": 2.1341672338128654e-05, "loss": 0.2266, "step": 50215 }, { "epoch": 1.17, "learning_rate": 2.134079228368262e-05, "loss": 0.0799, "step": 50220 }, { "epoch": 1.17, "learning_rate": 2.133991222923658e-05, "loss": 0.0557, "step": 50225 }, { "epoch": 1.17, "learning_rate": 2.1339032174790545e-05, "loss": 0.0832, "step": 50230 }, { "epoch": 1.17, "learning_rate": 2.1338152120344514e-05, "loss": 0.0639, "step": 50235 }, { "epoch": 1.17, "learning_rate": 2.133727206589848e-05, "loss": 0.248, "step": 50240 }, { "epoch": 1.17, "learning_rate": 2.133639201145244e-05, "loss": 0.1561, "step": 50245 }, { "epoch": 1.17, "learning_rate": 2.133551195700641e-05, "loss": 0.1739, "step": 50250 }, { "epoch": 1.17, "learning_rate": 2.1334631902560374e-05, "loss": 0.2453, "step": 50255 }, { "epoch": 1.17, "learning_rate": 2.1333751848114335e-05, "loss": 0.201, "step": 50260 }, { "epoch": 1.17, "learning_rate": 2.13328717936683e-05, "loss": 0.2624, "step": 50265 }, { "epoch": 1.17, "learning_rate": 2.133199173922227e-05, "loss": 0.07, "step": 50270 }, { "epoch": 1.17, "learning_rate": 2.1331111684776234e-05, "loss": 0.0773, "step": 50275 }, { "epoch": 1.17, "learning_rate": 2.1330231630330196e-05, "loss": 0.0814, "step": 50280 }, { "epoch": 1.17, "learning_rate": 2.132935157588416e-05, "loss": 0.1069, "step": 50285 }, { "epoch": 1.17, "learning_rate": 2.132847152143813e-05, "loss": 0.1061, "step": 50290 }, { "epoch": 1.17, "learning_rate": 2.132759146699209e-05, "loss": 0.0855, "step": 50295 }, { "epoch": 1.17, "learning_rate": 2.1326711412546056e-05, "loss": 0.3407, "step": 50300 }, { "epoch": 1.17, "learning_rate": 2.132583135810002e-05, "loss": 0.3674, "step": 50305 }, { "epoch": 1.17, "learning_rate": 2.132495130365399e-05, "loss": 0.3681, "step": 50310 }, { "epoch": 1.17, "learning_rate": 2.132407124920795e-05, "loss": 0.3421, "step": 50315 }, { "epoch": 1.17, "learning_rate": 2.1323191194761916e-05, "loss": 0.0782, "step": 50320 }, { "epoch": 1.17, "learning_rate": 2.132231114031588e-05, "loss": 0.0049, "step": 50325 }, { "epoch": 1.17, "learning_rate": 2.132143108586985e-05, "loss": 0.1138, "step": 50330 }, { "epoch": 1.17, "learning_rate": 2.132055103142381e-05, "loss": 0.0586, "step": 50335 }, { "epoch": 1.17, "learning_rate": 2.1319670976977776e-05, "loss": 0.1575, "step": 50340 }, { "epoch": 1.17, "learning_rate": 2.131879092253174e-05, "loss": 0.1122, "step": 50345 }, { "epoch": 1.17, "learning_rate": 2.1317910868085706e-05, "loss": 0.1712, "step": 50350 }, { "epoch": 1.17, "learning_rate": 2.131703081363967e-05, "loss": 0.1867, "step": 50355 }, { "epoch": 1.17, "learning_rate": 2.1316150759193636e-05, "loss": 0.2844, "step": 50360 }, { "epoch": 1.18, "learning_rate": 2.13152707047476e-05, "loss": 0.2629, "step": 50365 }, { "epoch": 1.18, "learning_rate": 2.1314390650301566e-05, "loss": 0.042, "step": 50370 }, { "epoch": 1.18, "learning_rate": 2.131351059585553e-05, "loss": 0.1087, "step": 50375 }, { "epoch": 1.18, "learning_rate": 2.1312630541409496e-05, "loss": 0.0247, "step": 50380 }, { "epoch": 1.18, "learning_rate": 2.131175048696346e-05, "loss": 0.043, "step": 50385 }, { "epoch": 1.18, "learning_rate": 2.1310870432517426e-05, "loss": 0.0584, "step": 50390 }, { "epoch": 1.18, "learning_rate": 2.130999037807139e-05, "loss": 0.1406, "step": 50395 }, { "epoch": 1.18, "learning_rate": 2.1309110323625356e-05, "loss": 0.2196, "step": 50400 }, { "epoch": 1.18, "learning_rate": 2.130823026917932e-05, "loss": 0.1805, "step": 50405 }, { "epoch": 1.18, "learning_rate": 2.1307350214733286e-05, "loss": 0.5588, "step": 50410 }, { "epoch": 1.18, "learning_rate": 2.130647016028725e-05, "loss": 0.3924, "step": 50415 }, { "epoch": 1.18, "learning_rate": 2.1305590105841213e-05, "loss": 0.12, "step": 50420 }, { "epoch": 1.18, "learning_rate": 2.130471005139518e-05, "loss": 0.0437, "step": 50425 }, { "epoch": 1.18, "learning_rate": 2.1303829996949146e-05, "loss": 0.031, "step": 50430 }, { "epoch": 1.18, "learning_rate": 2.130294994250311e-05, "loss": 0.091, "step": 50435 }, { "epoch": 1.18, "learning_rate": 2.1302069888057073e-05, "loss": 0.0991, "step": 50440 }, { "epoch": 1.18, "learning_rate": 2.130118983361104e-05, "loss": 0.2512, "step": 50445 }, { "epoch": 1.18, "learning_rate": 2.1300309779165006e-05, "loss": 0.1097, "step": 50450 }, { "epoch": 1.18, "learning_rate": 2.1299429724718968e-05, "loss": 0.1861, "step": 50455 }, { "epoch": 1.18, "learning_rate": 2.1298549670272933e-05, "loss": 0.3349, "step": 50460 }, { "epoch": 1.18, "learning_rate": 2.12976696158269e-05, "loss": 0.2811, "step": 50465 }, { "epoch": 1.18, "learning_rate": 2.1296789561380866e-05, "loss": 0.0886, "step": 50470 }, { "epoch": 1.18, "learning_rate": 2.1295909506934828e-05, "loss": 0.0426, "step": 50475 }, { "epoch": 1.18, "learning_rate": 2.1295029452488793e-05, "loss": 0.0585, "step": 50480 }, { "epoch": 1.18, "learning_rate": 2.129414939804276e-05, "loss": 0.1228, "step": 50485 }, { "epoch": 1.18, "learning_rate": 2.1293269343596723e-05, "loss": 0.2028, "step": 50490 }, { "epoch": 1.18, "learning_rate": 2.1292389289150688e-05, "loss": 0.1425, "step": 50495 }, { "epoch": 1.18, "learning_rate": 2.1291509234704653e-05, "loss": 0.2379, "step": 50500 }, { "epoch": 1.18, "learning_rate": 2.129062918025862e-05, "loss": 0.2492, "step": 50505 }, { "epoch": 1.18, "learning_rate": 2.1289749125812583e-05, "loss": 0.2491, "step": 50510 }, { "epoch": 1.18, "learning_rate": 2.1288869071366548e-05, "loss": 0.2748, "step": 50515 }, { "epoch": 1.18, "learning_rate": 2.1287989016920517e-05, "loss": 0.0541, "step": 50520 }, { "epoch": 1.18, "learning_rate": 2.1287108962474478e-05, "loss": 0.0495, "step": 50525 }, { "epoch": 1.18, "learning_rate": 2.1286228908028443e-05, "loss": 0.0898, "step": 50530 }, { "epoch": 1.18, "learning_rate": 2.1285348853582408e-05, "loss": 0.1133, "step": 50535 }, { "epoch": 1.18, "learning_rate": 2.1284468799136377e-05, "loss": 0.0908, "step": 50540 }, { "epoch": 1.18, "learning_rate": 2.1283588744690338e-05, "loss": 0.1966, "step": 50545 }, { "epoch": 1.18, "learning_rate": 2.1282708690244303e-05, "loss": 0.0924, "step": 50550 }, { "epoch": 1.18, "learning_rate": 2.128182863579827e-05, "loss": 0.1452, "step": 50555 }, { "epoch": 1.18, "learning_rate": 2.1280948581352233e-05, "loss": 0.2643, "step": 50560 }, { "epoch": 1.18, "learning_rate": 2.12800685269062e-05, "loss": 0.2677, "step": 50565 }, { "epoch": 1.18, "learning_rate": 2.1279188472460163e-05, "loss": 0.0436, "step": 50570 }, { "epoch": 1.18, "learning_rate": 2.127830841801413e-05, "loss": 0.0436, "step": 50575 }, { "epoch": 1.18, "learning_rate": 2.1277428363568093e-05, "loss": 0.1036, "step": 50580 }, { "epoch": 1.18, "learning_rate": 2.127654830912206e-05, "loss": 0.1625, "step": 50585 }, { "epoch": 1.18, "learning_rate": 2.1275668254676023e-05, "loss": 0.1315, "step": 50590 }, { "epoch": 1.18, "learning_rate": 2.1274788200229985e-05, "loss": 0.1705, "step": 50595 }, { "epoch": 1.18, "learning_rate": 2.1273908145783953e-05, "loss": 0.157, "step": 50600 }, { "epoch": 1.18, "learning_rate": 2.127302809133792e-05, "loss": 0.1837, "step": 50605 }, { "epoch": 1.18, "learning_rate": 2.1272148036891884e-05, "loss": 0.2294, "step": 50610 }, { "epoch": 1.18, "learning_rate": 2.1271267982445845e-05, "loss": 0.4663, "step": 50615 }, { "epoch": 1.18, "learning_rate": 2.1270387927999814e-05, "loss": 0.0882, "step": 50620 }, { "epoch": 1.18, "learning_rate": 2.126950787355378e-05, "loss": 0.0518, "step": 50625 }, { "epoch": 1.18, "learning_rate": 2.126862781910774e-05, "loss": 0.1926, "step": 50630 }, { "epoch": 1.18, "learning_rate": 2.1267747764661705e-05, "loss": 0.0752, "step": 50635 }, { "epoch": 1.18, "learning_rate": 2.1266867710215674e-05, "loss": 0.2009, "step": 50640 }, { "epoch": 1.18, "learning_rate": 2.126598765576964e-05, "loss": 0.1007, "step": 50645 }, { "epoch": 1.18, "learning_rate": 2.12651076013236e-05, "loss": 0.3236, "step": 50650 }, { "epoch": 1.18, "learning_rate": 2.126422754687757e-05, "loss": 0.246, "step": 50655 }, { "epoch": 1.18, "learning_rate": 2.1263347492431534e-05, "loss": 0.1874, "step": 50660 }, { "epoch": 1.18, "learning_rate": 2.12624674379855e-05, "loss": 0.3779, "step": 50665 }, { "epoch": 1.18, "learning_rate": 2.126158738353946e-05, "loss": 0.0748, "step": 50670 }, { "epoch": 1.18, "learning_rate": 2.126070732909343e-05, "loss": 0.0185, "step": 50675 }, { "epoch": 1.18, "learning_rate": 2.1259827274647394e-05, "loss": 0.0584, "step": 50680 }, { "epoch": 1.18, "learning_rate": 2.1258947220201355e-05, "loss": 0.1057, "step": 50685 }, { "epoch": 1.18, "learning_rate": 2.125806716575532e-05, "loss": 0.1734, "step": 50690 }, { "epoch": 1.18, "learning_rate": 2.125718711130929e-05, "loss": 0.1625, "step": 50695 }, { "epoch": 1.18, "learning_rate": 2.1256307056863254e-05, "loss": 0.1123, "step": 50700 }, { "epoch": 1.18, "learning_rate": 2.1255427002417216e-05, "loss": 0.3005, "step": 50705 }, { "epoch": 1.18, "learning_rate": 2.125454694797118e-05, "loss": 0.3885, "step": 50710 }, { "epoch": 1.18, "learning_rate": 2.125366689352515e-05, "loss": 0.2966, "step": 50715 }, { "epoch": 1.18, "learning_rate": 2.125278683907911e-05, "loss": 0.0224, "step": 50720 }, { "epoch": 1.18, "learning_rate": 2.1251906784633076e-05, "loss": 0.0277, "step": 50725 }, { "epoch": 1.18, "learning_rate": 2.125102673018704e-05, "loss": 0.0896, "step": 50730 }, { "epoch": 1.18, "learning_rate": 2.125014667574101e-05, "loss": 0.165, "step": 50735 }, { "epoch": 1.18, "learning_rate": 2.124926662129497e-05, "loss": 0.1023, "step": 50740 }, { "epoch": 1.18, "learning_rate": 2.1248386566848936e-05, "loss": 0.4227, "step": 50745 }, { "epoch": 1.18, "learning_rate": 2.12475065124029e-05, "loss": 0.1305, "step": 50750 }, { "epoch": 1.18, "learning_rate": 2.1246626457956866e-05, "loss": 0.3656, "step": 50755 }, { "epoch": 1.18, "learning_rate": 2.124574640351083e-05, "loss": 0.5028, "step": 50760 }, { "epoch": 1.18, "learning_rate": 2.1244866349064796e-05, "loss": 0.1595, "step": 50765 }, { "epoch": 1.18, "learning_rate": 2.124398629461876e-05, "loss": 0.0728, "step": 50770 }, { "epoch": 1.18, "learning_rate": 2.1243106240172726e-05, "loss": 0.092, "step": 50775 }, { "epoch": 1.18, "learning_rate": 2.124222618572669e-05, "loss": 0.199, "step": 50780 }, { "epoch": 1.18, "learning_rate": 2.1241346131280656e-05, "loss": 0.2087, "step": 50785 }, { "epoch": 1.18, "learning_rate": 2.124046607683462e-05, "loss": 0.09, "step": 50790 }, { "epoch": 1.19, "learning_rate": 2.1239586022388586e-05, "loss": 0.0816, "step": 50795 }, { "epoch": 1.19, "learning_rate": 2.123870596794255e-05, "loss": 0.1972, "step": 50800 }, { "epoch": 1.19, "learning_rate": 2.1237825913496516e-05, "loss": 0.2535, "step": 50805 }, { "epoch": 1.19, "learning_rate": 2.123694585905048e-05, "loss": 0.5419, "step": 50810 }, { "epoch": 1.19, "learning_rate": 2.1236065804604446e-05, "loss": 0.363, "step": 50815 }, { "epoch": 1.19, "learning_rate": 2.123518575015841e-05, "loss": 0.0483, "step": 50820 }, { "epoch": 1.19, "learning_rate": 2.1234305695712373e-05, "loss": 0.0307, "step": 50825 }, { "epoch": 1.19, "learning_rate": 2.123342564126634e-05, "loss": 0.1059, "step": 50830 }, { "epoch": 1.19, "learning_rate": 2.1232545586820306e-05, "loss": 0.1877, "step": 50835 }, { "epoch": 1.19, "learning_rate": 2.123166553237427e-05, "loss": 0.0803, "step": 50840 }, { "epoch": 1.19, "learning_rate": 2.1230785477928233e-05, "loss": 0.1993, "step": 50845 }, { "epoch": 1.19, "learning_rate": 2.12299054234822e-05, "loss": 0.2899, "step": 50850 }, { "epoch": 1.19, "learning_rate": 2.1229025369036166e-05, "loss": 0.1773, "step": 50855 }, { "epoch": 1.19, "learning_rate": 2.1228145314590128e-05, "loss": 0.322, "step": 50860 }, { "epoch": 1.19, "learning_rate": 2.1227265260144093e-05, "loss": 0.261, "step": 50865 }, { "epoch": 1.19, "learning_rate": 2.122638520569806e-05, "loss": 0.2101, "step": 50870 }, { "epoch": 1.19, "learning_rate": 2.1225505151252026e-05, "loss": 0.0204, "step": 50875 }, { "epoch": 1.19, "learning_rate": 2.1224625096805988e-05, "loss": 0.0584, "step": 50880 }, { "epoch": 1.19, "learning_rate": 2.1223745042359953e-05, "loss": 0.1926, "step": 50885 }, { "epoch": 1.19, "learning_rate": 2.122286498791392e-05, "loss": 0.1354, "step": 50890 }, { "epoch": 1.19, "learning_rate": 2.1221984933467883e-05, "loss": 0.2808, "step": 50895 }, { "epoch": 1.19, "learning_rate": 2.1221104879021848e-05, "loss": 0.1966, "step": 50900 }, { "epoch": 1.19, "learning_rate": 2.1220224824575813e-05, "loss": 0.2354, "step": 50905 }, { "epoch": 1.19, "learning_rate": 2.121934477012978e-05, "loss": 0.2848, "step": 50910 }, { "epoch": 1.19, "learning_rate": 2.1218464715683743e-05, "loss": 0.2849, "step": 50915 }, { "epoch": 1.19, "learning_rate": 2.1217584661237708e-05, "loss": 0.029, "step": 50920 }, { "epoch": 1.19, "learning_rate": 2.1216704606791676e-05, "loss": 0.0596, "step": 50925 }, { "epoch": 1.19, "learning_rate": 2.1215824552345638e-05, "loss": 0.0358, "step": 50930 }, { "epoch": 1.19, "learning_rate": 2.1214944497899603e-05, "loss": 0.1066, "step": 50935 }, { "epoch": 1.19, "learning_rate": 2.1214064443453568e-05, "loss": 0.1685, "step": 50940 }, { "epoch": 1.19, "learning_rate": 2.1213184389007537e-05, "loss": 0.0688, "step": 50945 }, { "epoch": 1.19, "learning_rate": 2.1212304334561498e-05, "loss": 0.2054, "step": 50950 }, { "epoch": 1.19, "learning_rate": 2.1211424280115463e-05, "loss": 0.243, "step": 50955 }, { "epoch": 1.19, "learning_rate": 2.1210544225669428e-05, "loss": 0.2089, "step": 50960 }, { "epoch": 1.19, "learning_rate": 2.1209664171223397e-05, "loss": 0.264, "step": 50965 }, { "epoch": 1.19, "learning_rate": 2.1208784116777358e-05, "loss": 0.0515, "step": 50970 }, { "epoch": 1.19, "learning_rate": 2.1207904062331323e-05, "loss": 0.0558, "step": 50975 }, { "epoch": 1.19, "learning_rate": 2.1207024007885288e-05, "loss": 0.1306, "step": 50980 }, { "epoch": 1.19, "learning_rate": 2.1206143953439253e-05, "loss": 0.0914, "step": 50985 }, { "epoch": 1.19, "learning_rate": 2.120526389899322e-05, "loss": 0.2707, "step": 50990 }, { "epoch": 1.19, "learning_rate": 2.1204383844547183e-05, "loss": 0.1584, "step": 50995 }, { "epoch": 1.19, "learning_rate": 2.120350379010115e-05, "loss": 0.2002, "step": 51000 }, { "epoch": 1.19, "learning_rate": 2.1202623735655113e-05, "loss": 0.3144, "step": 51005 }, { "epoch": 1.19, "learning_rate": 2.120174368120908e-05, "loss": 0.2998, "step": 51010 }, { "epoch": 1.19, "learning_rate": 2.1200863626763043e-05, "loss": 0.1436, "step": 51015 }, { "epoch": 1.19, "learning_rate": 2.1199983572317005e-05, "loss": 0.051, "step": 51020 }, { "epoch": 1.19, "learning_rate": 2.1199103517870973e-05, "loss": 0.105, "step": 51025 }, { "epoch": 1.19, "learning_rate": 2.119822346342494e-05, "loss": 0.0827, "step": 51030 }, { "epoch": 1.19, "learning_rate": 2.1197343408978904e-05, "loss": 0.1155, "step": 51035 }, { "epoch": 1.19, "learning_rate": 2.1196463354532865e-05, "loss": 0.0758, "step": 51040 }, { "epoch": 1.19, "learning_rate": 2.1195583300086834e-05, "loss": 0.1117, "step": 51045 }, { "epoch": 1.19, "learning_rate": 2.11947032456408e-05, "loss": 0.2771, "step": 51050 }, { "epoch": 1.19, "learning_rate": 2.119382319119476e-05, "loss": 0.2145, "step": 51055 }, { "epoch": 1.19, "learning_rate": 2.119294313674873e-05, "loss": 0.3193, "step": 51060 }, { "epoch": 1.19, "learning_rate": 2.1192063082302694e-05, "loss": 0.3933, "step": 51065 }, { "epoch": 1.19, "learning_rate": 2.119118302785666e-05, "loss": 0.0147, "step": 51070 }, { "epoch": 1.19, "learning_rate": 2.119030297341062e-05, "loss": 0.0248, "step": 51075 }, { "epoch": 1.19, "learning_rate": 2.118942291896459e-05, "loss": 0.0647, "step": 51080 }, { "epoch": 1.19, "learning_rate": 2.1188542864518554e-05, "loss": 0.1085, "step": 51085 }, { "epoch": 1.19, "learning_rate": 2.1187662810072515e-05, "loss": 0.1655, "step": 51090 }, { "epoch": 1.19, "learning_rate": 2.118678275562648e-05, "loss": 0.1255, "step": 51095 }, { "epoch": 1.19, "learning_rate": 2.118590270118045e-05, "loss": 0.1436, "step": 51100 }, { "epoch": 1.19, "learning_rate": 2.1185022646734414e-05, "loss": 0.1514, "step": 51105 }, { "epoch": 1.19, "learning_rate": 2.1184142592288375e-05, "loss": 0.3071, "step": 51110 }, { "epoch": 1.19, "learning_rate": 2.118326253784234e-05, "loss": 0.2337, "step": 51115 }, { "epoch": 1.19, "learning_rate": 2.118238248339631e-05, "loss": 0.0443, "step": 51120 }, { "epoch": 1.19, "learning_rate": 2.118150242895027e-05, "loss": 0.0909, "step": 51125 }, { "epoch": 1.19, "learning_rate": 2.1180622374504236e-05, "loss": 0.053, "step": 51130 }, { "epoch": 1.19, "learning_rate": 2.11797423200582e-05, "loss": 0.1704, "step": 51135 }, { "epoch": 1.19, "learning_rate": 2.117886226561217e-05, "loss": 0.1706, "step": 51140 }, { "epoch": 1.19, "learning_rate": 2.117798221116613e-05, "loss": 0.0506, "step": 51145 }, { "epoch": 1.19, "learning_rate": 2.1177102156720096e-05, "loss": 0.1378, "step": 51150 }, { "epoch": 1.19, "learning_rate": 2.117622210227406e-05, "loss": 0.1588, "step": 51155 }, { "epoch": 1.19, "learning_rate": 2.1175342047828026e-05, "loss": 0.2761, "step": 51160 }, { "epoch": 1.19, "learning_rate": 2.117446199338199e-05, "loss": 0.2627, "step": 51165 }, { "epoch": 1.19, "learning_rate": 2.1173581938935956e-05, "loss": 0.0574, "step": 51170 }, { "epoch": 1.19, "learning_rate": 2.117270188448992e-05, "loss": 0.0914, "step": 51175 }, { "epoch": 1.19, "learning_rate": 2.1171821830043886e-05, "loss": 0.0723, "step": 51180 }, { "epoch": 1.19, "learning_rate": 2.117094177559785e-05, "loss": 0.0561, "step": 51185 }, { "epoch": 1.19, "learning_rate": 2.1170061721151816e-05, "loss": 0.1206, "step": 51190 }, { "epoch": 1.19, "learning_rate": 2.116918166670578e-05, "loss": 0.1921, "step": 51195 }, { "epoch": 1.19, "learning_rate": 2.1168301612259746e-05, "loss": 0.2063, "step": 51200 }, { "epoch": 1.19, "learning_rate": 2.116742155781371e-05, "loss": 0.1628, "step": 51205 }, { "epoch": 1.19, "learning_rate": 2.1166541503367676e-05, "loss": 0.2518, "step": 51210 }, { "epoch": 1.19, "learning_rate": 2.116566144892164e-05, "loss": 0.2597, "step": 51215 }, { "epoch": 1.2, "learning_rate": 2.1164781394475606e-05, "loss": 0.0542, "step": 51220 }, { "epoch": 1.2, "learning_rate": 2.116390134002957e-05, "loss": 0.0346, "step": 51225 }, { "epoch": 1.2, "learning_rate": 2.1163021285583533e-05, "loss": 0.1787, "step": 51230 }, { "epoch": 1.2, "learning_rate": 2.11621412311375e-05, "loss": 0.1511, "step": 51235 }, { "epoch": 1.2, "learning_rate": 2.1161261176691466e-05, "loss": 0.204, "step": 51240 }, { "epoch": 1.2, "learning_rate": 2.116038112224543e-05, "loss": 0.1707, "step": 51245 }, { "epoch": 1.2, "learning_rate": 2.1159501067799393e-05, "loss": 0.2251, "step": 51250 }, { "epoch": 1.2, "learning_rate": 2.115862101335336e-05, "loss": 0.2686, "step": 51255 }, { "epoch": 1.2, "learning_rate": 2.1157740958907326e-05, "loss": 0.2285, "step": 51260 }, { "epoch": 1.2, "learning_rate": 2.115686090446129e-05, "loss": 0.4444, "step": 51265 }, { "epoch": 1.2, "learning_rate": 2.1155980850015253e-05, "loss": 0.0602, "step": 51270 }, { "epoch": 1.2, "learning_rate": 2.115510079556922e-05, "loss": 0.0239, "step": 51275 }, { "epoch": 1.2, "learning_rate": 2.1154220741123186e-05, "loss": 0.0538, "step": 51280 }, { "epoch": 1.2, "learning_rate": 2.1153340686677148e-05, "loss": 0.1115, "step": 51285 }, { "epoch": 1.2, "learning_rate": 2.1152460632231113e-05, "loss": 0.0579, "step": 51290 }, { "epoch": 1.2, "learning_rate": 2.115158057778508e-05, "loss": 0.128, "step": 51295 }, { "epoch": 1.2, "learning_rate": 2.1150700523339046e-05, "loss": 0.1939, "step": 51300 }, { "epoch": 1.2, "learning_rate": 2.1149820468893008e-05, "loss": 0.2598, "step": 51305 }, { "epoch": 1.2, "learning_rate": 2.1148940414446973e-05, "loss": 0.2623, "step": 51310 }, { "epoch": 1.2, "learning_rate": 2.114806036000094e-05, "loss": 0.2274, "step": 51315 }, { "epoch": 1.2, "learning_rate": 2.1147180305554903e-05, "loss": 0.0451, "step": 51320 }, { "epoch": 1.2, "learning_rate": 2.1146300251108868e-05, "loss": 0.0402, "step": 51325 }, { "epoch": 1.2, "learning_rate": 2.1145420196662836e-05, "loss": 0.1245, "step": 51330 }, { "epoch": 1.2, "learning_rate": 2.11445401422168e-05, "loss": 0.0971, "step": 51335 }, { "epoch": 1.2, "learning_rate": 2.1143660087770763e-05, "loss": 0.1079, "step": 51340 }, { "epoch": 1.2, "learning_rate": 2.1142780033324728e-05, "loss": 0.0841, "step": 51345 }, { "epoch": 1.2, "learning_rate": 2.1141899978878696e-05, "loss": 0.2383, "step": 51350 }, { "epoch": 1.2, "learning_rate": 2.1141019924432658e-05, "loss": 0.1248, "step": 51355 }, { "epoch": 1.2, "learning_rate": 2.1140139869986623e-05, "loss": 0.3192, "step": 51360 }, { "epoch": 1.2, "learning_rate": 2.1139259815540588e-05, "loss": 0.2611, "step": 51365 }, { "epoch": 1.2, "learning_rate": 2.1138379761094556e-05, "loss": 0.1049, "step": 51370 }, { "epoch": 1.2, "learning_rate": 2.1137499706648518e-05, "loss": 0.0055, "step": 51375 }, { "epoch": 1.2, "learning_rate": 2.1136619652202483e-05, "loss": 0.1162, "step": 51380 }, { "epoch": 1.2, "learning_rate": 2.1135739597756448e-05, "loss": 0.1325, "step": 51385 }, { "epoch": 1.2, "learning_rate": 2.1134859543310413e-05, "loss": 0.2191, "step": 51390 }, { "epoch": 1.2, "learning_rate": 2.1133979488864378e-05, "loss": 0.0965, "step": 51395 }, { "epoch": 1.2, "learning_rate": 2.1133099434418343e-05, "loss": 0.1522, "step": 51400 }, { "epoch": 1.2, "learning_rate": 2.1132219379972308e-05, "loss": 0.2306, "step": 51405 }, { "epoch": 1.2, "learning_rate": 2.1131339325526273e-05, "loss": 0.2294, "step": 51410 }, { "epoch": 1.2, "learning_rate": 2.1130459271080238e-05, "loss": 0.3885, "step": 51415 }, { "epoch": 1.2, "learning_rate": 2.1129579216634203e-05, "loss": 0.1062, "step": 51420 }, { "epoch": 1.2, "learning_rate": 2.1128699162188165e-05, "loss": 0.2034, "step": 51425 }, { "epoch": 1.2, "learning_rate": 2.1127819107742133e-05, "loss": 0.0799, "step": 51430 }, { "epoch": 1.2, "learning_rate": 2.11269390532961e-05, "loss": 0.067, "step": 51435 }, { "epoch": 1.2, "learning_rate": 2.1126058998850063e-05, "loss": 0.0936, "step": 51440 }, { "epoch": 1.2, "learning_rate": 2.1125178944404025e-05, "loss": 0.2204, "step": 51445 }, { "epoch": 1.2, "learning_rate": 2.1124298889957993e-05, "loss": 0.1948, "step": 51450 }, { "epoch": 1.2, "learning_rate": 2.112341883551196e-05, "loss": 0.1992, "step": 51455 }, { "epoch": 1.2, "learning_rate": 2.112253878106592e-05, "loss": 0.2784, "step": 51460 }, { "epoch": 1.2, "learning_rate": 2.112165872661989e-05, "loss": 0.3158, "step": 51465 }, { "epoch": 1.2, "learning_rate": 2.1120778672173854e-05, "loss": 0.0645, "step": 51470 }, { "epoch": 1.2, "learning_rate": 2.111989861772782e-05, "loss": 0.1217, "step": 51475 }, { "epoch": 1.2, "learning_rate": 2.111901856328178e-05, "loss": 0.1378, "step": 51480 }, { "epoch": 1.2, "learning_rate": 2.111813850883575e-05, "loss": 0.1063, "step": 51485 }, { "epoch": 1.2, "learning_rate": 2.1117258454389714e-05, "loss": 0.102, "step": 51490 }, { "epoch": 1.2, "learning_rate": 2.1116378399943675e-05, "loss": 0.2371, "step": 51495 }, { "epoch": 1.2, "learning_rate": 2.111549834549764e-05, "loss": 0.2841, "step": 51500 }, { "epoch": 1.2, "learning_rate": 2.111461829105161e-05, "loss": 0.277, "step": 51505 }, { "epoch": 1.2, "learning_rate": 2.1113738236605574e-05, "loss": 0.4889, "step": 51510 }, { "epoch": 1.2, "learning_rate": 2.1112858182159535e-05, "loss": 0.2905, "step": 51515 }, { "epoch": 1.2, "learning_rate": 2.11119781277135e-05, "loss": 0.0295, "step": 51520 }, { "epoch": 1.2, "learning_rate": 2.111109807326747e-05, "loss": 0.1267, "step": 51525 }, { "epoch": 1.2, "learning_rate": 2.111021801882143e-05, "loss": 0.3999, "step": 51530 }, { "epoch": 1.2, "learning_rate": 2.1109337964375395e-05, "loss": 0.067, "step": 51535 }, { "epoch": 1.2, "learning_rate": 2.110845790992936e-05, "loss": 0.0361, "step": 51540 }, { "epoch": 1.2, "learning_rate": 2.110757785548333e-05, "loss": 0.1539, "step": 51545 }, { "epoch": 1.2, "learning_rate": 2.110669780103729e-05, "loss": 0.2258, "step": 51550 }, { "epoch": 1.2, "learning_rate": 2.1105817746591255e-05, "loss": 0.1769, "step": 51555 }, { "epoch": 1.2, "learning_rate": 2.110493769214522e-05, "loss": 0.2799, "step": 51560 }, { "epoch": 1.2, "learning_rate": 2.110405763769919e-05, "loss": 0.1978, "step": 51565 }, { "epoch": 1.2, "learning_rate": 2.110317758325315e-05, "loss": 0.0971, "step": 51570 }, { "epoch": 1.2, "learning_rate": 2.1102297528807116e-05, "loss": 0.0901, "step": 51575 }, { "epoch": 1.2, "learning_rate": 2.110141747436108e-05, "loss": 0.2187, "step": 51580 }, { "epoch": 1.2, "learning_rate": 2.1100537419915046e-05, "loss": 0.2169, "step": 51585 }, { "epoch": 1.2, "learning_rate": 2.109965736546901e-05, "loss": 0.072, "step": 51590 }, { "epoch": 1.2, "learning_rate": 2.1098777311022976e-05, "loss": 0.1202, "step": 51595 }, { "epoch": 1.2, "learning_rate": 2.109789725657694e-05, "loss": 0.1529, "step": 51600 }, { "epoch": 1.2, "learning_rate": 2.1097017202130906e-05, "loss": 0.2084, "step": 51605 }, { "epoch": 1.2, "learning_rate": 2.109613714768487e-05, "loss": 0.4112, "step": 51610 }, { "epoch": 1.2, "learning_rate": 2.1095257093238836e-05, "loss": 0.2973, "step": 51615 }, { "epoch": 1.2, "learning_rate": 2.10943770387928e-05, "loss": 0.0655, "step": 51620 }, { "epoch": 1.2, "learning_rate": 2.1093496984346766e-05, "loss": 0.0571, "step": 51625 }, { "epoch": 1.2, "learning_rate": 2.109261692990073e-05, "loss": 0.0456, "step": 51630 }, { "epoch": 1.2, "learning_rate": 2.1091736875454696e-05, "loss": 0.0971, "step": 51635 }, { "epoch": 1.2, "learning_rate": 2.109085682100866e-05, "loss": 0.1276, "step": 51640 }, { "epoch": 1.2, "learning_rate": 2.1089976766562626e-05, "loss": 0.1218, "step": 51645 }, { "epoch": 1.21, "learning_rate": 2.108909671211659e-05, "loss": 0.1512, "step": 51650 }, { "epoch": 1.21, "learning_rate": 2.1088216657670552e-05, "loss": 0.2872, "step": 51655 }, { "epoch": 1.21, "learning_rate": 2.108733660322452e-05, "loss": 0.4229, "step": 51660 }, { "epoch": 1.21, "learning_rate": 2.1086456548778486e-05, "loss": 0.1982, "step": 51665 }, { "epoch": 1.21, "learning_rate": 2.108557649433245e-05, "loss": 0.0906, "step": 51670 }, { "epoch": 1.21, "learning_rate": 2.1084696439886413e-05, "loss": 0.0384, "step": 51675 }, { "epoch": 1.21, "learning_rate": 2.108381638544038e-05, "loss": 0.0386, "step": 51680 }, { "epoch": 1.21, "learning_rate": 2.1082936330994346e-05, "loss": 0.0262, "step": 51685 }, { "epoch": 1.21, "learning_rate": 2.1082056276548308e-05, "loss": 0.1934, "step": 51690 }, { "epoch": 1.21, "learning_rate": 2.1081176222102273e-05, "loss": 0.0572, "step": 51695 }, { "epoch": 1.21, "learning_rate": 2.108029616765624e-05, "loss": 0.1101, "step": 51700 }, { "epoch": 1.21, "learning_rate": 2.1079416113210206e-05, "loss": 0.1763, "step": 51705 }, { "epoch": 1.21, "learning_rate": 2.1078536058764168e-05, "loss": 0.3924, "step": 51710 }, { "epoch": 1.21, "learning_rate": 2.1077656004318133e-05, "loss": 0.3576, "step": 51715 }, { "epoch": 1.21, "learning_rate": 2.10767759498721e-05, "loss": 0.1295, "step": 51720 }, { "epoch": 1.21, "learning_rate": 2.1075895895426063e-05, "loss": 0.0407, "step": 51725 }, { "epoch": 1.21, "learning_rate": 2.1075015840980028e-05, "loss": 0.0446, "step": 51730 }, { "epoch": 1.21, "learning_rate": 2.1074135786533996e-05, "loss": 0.1132, "step": 51735 }, { "epoch": 1.21, "learning_rate": 2.107325573208796e-05, "loss": 0.1098, "step": 51740 }, { "epoch": 1.21, "learning_rate": 2.1072375677641923e-05, "loss": 0.097, "step": 51745 }, { "epoch": 1.21, "learning_rate": 2.1071495623195888e-05, "loss": 0.1721, "step": 51750 }, { "epoch": 1.21, "learning_rate": 2.1070615568749856e-05, "loss": 0.2225, "step": 51755 }, { "epoch": 1.21, "learning_rate": 2.1069735514303818e-05, "loss": 0.6202, "step": 51760 }, { "epoch": 1.21, "learning_rate": 2.1068855459857783e-05, "loss": 0.2669, "step": 51765 }, { "epoch": 1.21, "learning_rate": 2.1067975405411748e-05, "loss": 0.0424, "step": 51770 }, { "epoch": 1.21, "learning_rate": 2.1067095350965716e-05, "loss": 0.0227, "step": 51775 }, { "epoch": 1.21, "learning_rate": 2.1066215296519678e-05, "loss": 0.0756, "step": 51780 }, { "epoch": 1.21, "learning_rate": 2.1065335242073643e-05, "loss": 0.124, "step": 51785 }, { "epoch": 1.21, "learning_rate": 2.1064455187627608e-05, "loss": 0.1553, "step": 51790 }, { "epoch": 1.21, "learning_rate": 2.1063575133181573e-05, "loss": 0.0589, "step": 51795 }, { "epoch": 1.21, "learning_rate": 2.1062695078735538e-05, "loss": 0.118, "step": 51800 }, { "epoch": 1.21, "learning_rate": 2.1061815024289503e-05, "loss": 0.0832, "step": 51805 }, { "epoch": 1.21, "learning_rate": 2.1060934969843468e-05, "loss": 0.4237, "step": 51810 }, { "epoch": 1.21, "learning_rate": 2.1060054915397433e-05, "loss": 0.4074, "step": 51815 }, { "epoch": 1.21, "learning_rate": 2.1059174860951398e-05, "loss": 0.1497, "step": 51820 }, { "epoch": 1.21, "learning_rate": 2.1058294806505363e-05, "loss": 0.0514, "step": 51825 }, { "epoch": 1.21, "learning_rate": 2.1057414752059325e-05, "loss": 0.0598, "step": 51830 }, { "epoch": 1.21, "learning_rate": 2.1056534697613293e-05, "loss": 0.1013, "step": 51835 }, { "epoch": 1.21, "learning_rate": 2.1055654643167258e-05, "loss": 0.0186, "step": 51840 }, { "epoch": 1.21, "learning_rate": 2.1054774588721223e-05, "loss": 0.1173, "step": 51845 }, { "epoch": 1.21, "learning_rate": 2.1053894534275185e-05, "loss": 0.2609, "step": 51850 }, { "epoch": 1.21, "learning_rate": 2.1053014479829153e-05, "loss": 0.0788, "step": 51855 }, { "epoch": 1.21, "learning_rate": 2.105213442538312e-05, "loss": 0.4084, "step": 51860 }, { "epoch": 1.21, "learning_rate": 2.1051254370937083e-05, "loss": 0.3537, "step": 51865 }, { "epoch": 1.21, "learning_rate": 2.105037431649105e-05, "loss": 0.032, "step": 51870 }, { "epoch": 1.21, "learning_rate": 2.1049494262045013e-05, "loss": 0.1503, "step": 51875 }, { "epoch": 1.21, "learning_rate": 2.104861420759898e-05, "loss": 0.0853, "step": 51880 }, { "epoch": 1.21, "learning_rate": 2.104773415315294e-05, "loss": 0.1704, "step": 51885 }, { "epoch": 1.21, "learning_rate": 2.104685409870691e-05, "loss": 0.2963, "step": 51890 }, { "epoch": 1.21, "learning_rate": 2.1045974044260873e-05, "loss": 0.1168, "step": 51895 }, { "epoch": 1.21, "learning_rate": 2.104509398981484e-05, "loss": 0.1004, "step": 51900 }, { "epoch": 1.21, "learning_rate": 2.10442139353688e-05, "loss": 0.2084, "step": 51905 }, { "epoch": 1.21, "learning_rate": 2.104333388092277e-05, "loss": 0.4996, "step": 51910 }, { "epoch": 1.21, "learning_rate": 2.1042453826476734e-05, "loss": 0.1445, "step": 51915 }, { "epoch": 1.21, "learning_rate": 2.1041573772030695e-05, "loss": 0.0255, "step": 51920 }, { "epoch": 1.21, "learning_rate": 2.104069371758466e-05, "loss": 0.0857, "step": 51925 }, { "epoch": 1.21, "learning_rate": 2.103981366313863e-05, "loss": 0.128, "step": 51930 }, { "epoch": 1.21, "learning_rate": 2.1038933608692594e-05, "loss": 0.081, "step": 51935 }, { "epoch": 1.21, "learning_rate": 2.1038053554246555e-05, "loss": 0.198, "step": 51940 }, { "epoch": 1.21, "learning_rate": 2.103717349980052e-05, "loss": 0.163, "step": 51945 }, { "epoch": 1.21, "learning_rate": 2.103629344535449e-05, "loss": 0.1438, "step": 51950 }, { "epoch": 1.21, "learning_rate": 2.103541339090845e-05, "loss": 0.2464, "step": 51955 }, { "epoch": 1.21, "learning_rate": 2.1034533336462415e-05, "loss": 0.2828, "step": 51960 }, { "epoch": 1.21, "learning_rate": 2.103365328201638e-05, "loss": 0.2655, "step": 51965 }, { "epoch": 1.21, "learning_rate": 2.103277322757035e-05, "loss": 0.029, "step": 51970 }, { "epoch": 1.21, "learning_rate": 2.103189317312431e-05, "loss": 0.0688, "step": 51975 }, { "epoch": 1.21, "learning_rate": 2.1031013118678275e-05, "loss": 0.1104, "step": 51980 }, { "epoch": 1.21, "learning_rate": 2.103013306423224e-05, "loss": 0.1224, "step": 51985 }, { "epoch": 1.21, "learning_rate": 2.1029253009786205e-05, "loss": 0.1407, "step": 51990 }, { "epoch": 1.21, "learning_rate": 2.102837295534017e-05, "loss": 0.186, "step": 51995 }, { "epoch": 1.21, "learning_rate": 2.1027492900894136e-05, "loss": 0.2194, "step": 52000 }, { "epoch": 1.21, "learning_rate": 2.10266128464481e-05, "loss": 0.3647, "step": 52005 }, { "epoch": 1.21, "learning_rate": 2.1025732792002066e-05, "loss": 0.3582, "step": 52010 }, { "epoch": 1.21, "learning_rate": 2.102485273755603e-05, "loss": 0.2247, "step": 52015 }, { "epoch": 1.21, "learning_rate": 2.1023972683109996e-05, "loss": 0.1142, "step": 52020 }, { "epoch": 1.21, "learning_rate": 2.102309262866396e-05, "loss": 0.0193, "step": 52025 }, { "epoch": 1.21, "learning_rate": 2.1022212574217926e-05, "loss": 0.0787, "step": 52030 }, { "epoch": 1.21, "learning_rate": 2.102133251977189e-05, "loss": 0.0819, "step": 52035 }, { "epoch": 1.21, "learning_rate": 2.1020452465325856e-05, "loss": 0.0963, "step": 52040 }, { "epoch": 1.21, "learning_rate": 2.101957241087982e-05, "loss": 0.154, "step": 52045 }, { "epoch": 1.21, "learning_rate": 2.1018692356433786e-05, "loss": 0.1873, "step": 52050 }, { "epoch": 1.21, "learning_rate": 2.101781230198775e-05, "loss": 0.3561, "step": 52055 }, { "epoch": 1.21, "learning_rate": 2.1016932247541712e-05, "loss": 0.2069, "step": 52060 }, { "epoch": 1.21, "learning_rate": 2.101605219309568e-05, "loss": 0.2411, "step": 52065 }, { "epoch": 1.21, "learning_rate": 2.1015172138649646e-05, "loss": 0.0561, "step": 52070 }, { "epoch": 1.21, "learning_rate": 2.101429208420361e-05, "loss": 0.1148, "step": 52075 }, { "epoch": 1.22, "learning_rate": 2.1013412029757572e-05, "loss": 0.108, "step": 52080 }, { "epoch": 1.22, "learning_rate": 2.101253197531154e-05, "loss": 0.0472, "step": 52085 }, { "epoch": 1.22, "learning_rate": 2.1011651920865506e-05, "loss": 0.0441, "step": 52090 }, { "epoch": 1.22, "learning_rate": 2.1010771866419468e-05, "loss": 0.1236, "step": 52095 }, { "epoch": 1.22, "learning_rate": 2.1009891811973433e-05, "loss": 0.1065, "step": 52100 }, { "epoch": 1.22, "learning_rate": 2.10090117575274e-05, "loss": 0.2705, "step": 52105 }, { "epoch": 1.22, "learning_rate": 2.1008131703081366e-05, "loss": 0.1946, "step": 52110 }, { "epoch": 1.22, "learning_rate": 2.1007251648635328e-05, "loss": 0.2437, "step": 52115 }, { "epoch": 1.22, "learning_rate": 2.1006371594189293e-05, "loss": 0.0734, "step": 52120 }, { "epoch": 1.22, "learning_rate": 2.100549153974326e-05, "loss": 0.0142, "step": 52125 }, { "epoch": 1.22, "learning_rate": 2.1004611485297223e-05, "loss": 0.0366, "step": 52130 }, { "epoch": 1.22, "learning_rate": 2.1003731430851188e-05, "loss": 0.1213, "step": 52135 }, { "epoch": 1.22, "learning_rate": 2.1002851376405153e-05, "loss": 0.04, "step": 52140 }, { "epoch": 1.22, "learning_rate": 2.100197132195912e-05, "loss": 0.1171, "step": 52145 }, { "epoch": 1.22, "learning_rate": 2.1001091267513083e-05, "loss": 0.1338, "step": 52150 }, { "epoch": 1.22, "learning_rate": 2.1000211213067048e-05, "loss": 0.2816, "step": 52155 }, { "epoch": 1.22, "learning_rate": 2.0999331158621016e-05, "loss": 0.3759, "step": 52160 }, { "epoch": 1.22, "learning_rate": 2.099845110417498e-05, "loss": 0.4314, "step": 52165 }, { "epoch": 1.22, "learning_rate": 2.0997571049728943e-05, "loss": 0.074, "step": 52170 }, { "epoch": 1.22, "learning_rate": 2.0996690995282908e-05, "loss": 0.0718, "step": 52175 }, { "epoch": 1.22, "learning_rate": 2.0995810940836876e-05, "loss": 0.1334, "step": 52180 }, { "epoch": 1.22, "learning_rate": 2.0994930886390838e-05, "loss": 0.1562, "step": 52185 }, { "epoch": 1.22, "learning_rate": 2.0994050831944803e-05, "loss": 0.0733, "step": 52190 }, { "epoch": 1.22, "learning_rate": 2.0993170777498768e-05, "loss": 0.1786, "step": 52195 }, { "epoch": 1.22, "learning_rate": 2.0992290723052736e-05, "loss": 0.2291, "step": 52200 }, { "epoch": 1.22, "learning_rate": 2.0991410668606698e-05, "loss": 0.1646, "step": 52205 }, { "epoch": 1.22, "learning_rate": 2.0990530614160663e-05, "loss": 0.4264, "step": 52210 }, { "epoch": 1.22, "learning_rate": 2.0989650559714628e-05, "loss": 0.31, "step": 52215 }, { "epoch": 1.22, "learning_rate": 2.0988770505268593e-05, "loss": 0.1028, "step": 52220 }, { "epoch": 1.22, "learning_rate": 2.0987890450822558e-05, "loss": 0.0306, "step": 52225 }, { "epoch": 1.22, "learning_rate": 2.0987010396376523e-05, "loss": 0.1871, "step": 52230 }, { "epoch": 1.22, "learning_rate": 2.0986130341930488e-05, "loss": 0.0425, "step": 52235 }, { "epoch": 1.22, "learning_rate": 2.0985250287484453e-05, "loss": 0.2852, "step": 52240 }, { "epoch": 1.22, "learning_rate": 2.0984370233038418e-05, "loss": 0.228, "step": 52245 }, { "epoch": 1.22, "learning_rate": 2.0983490178592383e-05, "loss": 0.1112, "step": 52250 }, { "epoch": 1.22, "learning_rate": 2.0982610124146345e-05, "loss": 0.3508, "step": 52255 }, { "epoch": 1.22, "learning_rate": 2.0981730069700313e-05, "loss": 0.2525, "step": 52260 }, { "epoch": 1.22, "learning_rate": 2.0980850015254278e-05, "loss": 0.4259, "step": 52265 }, { "epoch": 1.22, "learning_rate": 2.0979969960808243e-05, "loss": 0.0095, "step": 52270 }, { "epoch": 1.22, "learning_rate": 2.0979089906362208e-05, "loss": 0.0407, "step": 52275 }, { "epoch": 1.22, "learning_rate": 2.0978209851916173e-05, "loss": 0.1072, "step": 52280 }, { "epoch": 1.22, "learning_rate": 2.097732979747014e-05, "loss": 0.1934, "step": 52285 }, { "epoch": 1.22, "learning_rate": 2.09764497430241e-05, "loss": 0.13, "step": 52290 }, { "epoch": 1.22, "learning_rate": 2.097556968857807e-05, "loss": 0.1357, "step": 52295 }, { "epoch": 1.22, "learning_rate": 2.0974689634132033e-05, "loss": 0.2374, "step": 52300 }, { "epoch": 1.22, "learning_rate": 2.0973809579686e-05, "loss": 0.1433, "step": 52305 }, { "epoch": 1.22, "learning_rate": 2.097292952523996e-05, "loss": 0.4016, "step": 52310 }, { "epoch": 1.22, "learning_rate": 2.097204947079393e-05, "loss": 0.1309, "step": 52315 }, { "epoch": 1.22, "learning_rate": 2.0971169416347893e-05, "loss": 0.0634, "step": 52320 }, { "epoch": 1.22, "learning_rate": 2.0970289361901855e-05, "loss": 0.0485, "step": 52325 }, { "epoch": 1.22, "learning_rate": 2.096940930745582e-05, "loss": 0.137, "step": 52330 }, { "epoch": 1.22, "learning_rate": 2.096852925300979e-05, "loss": 0.1402, "step": 52335 }, { "epoch": 1.22, "learning_rate": 2.0967649198563754e-05, "loss": 0.0471, "step": 52340 }, { "epoch": 1.22, "learning_rate": 2.0966769144117715e-05, "loss": 0.1019, "step": 52345 }, { "epoch": 1.22, "learning_rate": 2.096588908967168e-05, "loss": 0.2644, "step": 52350 }, { "epoch": 1.22, "learning_rate": 2.096500903522565e-05, "loss": 0.2608, "step": 52355 }, { "epoch": 1.22, "learning_rate": 2.096412898077961e-05, "loss": 0.5171, "step": 52360 }, { "epoch": 1.22, "learning_rate": 2.0963248926333575e-05, "loss": 0.2218, "step": 52365 }, { "epoch": 1.22, "learning_rate": 2.096236887188754e-05, "loss": 0.0684, "step": 52370 }, { "epoch": 1.22, "learning_rate": 2.096148881744151e-05, "loss": 0.1876, "step": 52375 }, { "epoch": 1.22, "learning_rate": 2.096060876299547e-05, "loss": 0.0405, "step": 52380 }, { "epoch": 1.22, "learning_rate": 2.0959728708549435e-05, "loss": 0.1614, "step": 52385 }, { "epoch": 1.22, "learning_rate": 2.09588486541034e-05, "loss": 0.0844, "step": 52390 }, { "epoch": 1.22, "learning_rate": 2.0957968599657365e-05, "loss": 0.1243, "step": 52395 }, { "epoch": 1.22, "learning_rate": 2.095708854521133e-05, "loss": 0.3114, "step": 52400 }, { "epoch": 1.22, "learning_rate": 2.0956208490765295e-05, "loss": 0.3258, "step": 52405 }, { "epoch": 1.22, "learning_rate": 2.095532843631926e-05, "loss": 0.2501, "step": 52410 }, { "epoch": 1.22, "learning_rate": 2.0954448381873225e-05, "loss": 0.21, "step": 52415 }, { "epoch": 1.22, "learning_rate": 2.095356832742719e-05, "loss": 0.0292, "step": 52420 }, { "epoch": 1.22, "learning_rate": 2.0952688272981155e-05, "loss": 0.0875, "step": 52425 }, { "epoch": 1.22, "learning_rate": 2.095180821853512e-05, "loss": 0.0723, "step": 52430 }, { "epoch": 1.22, "learning_rate": 2.0950928164089086e-05, "loss": 0.094, "step": 52435 }, { "epoch": 1.22, "learning_rate": 2.095004810964305e-05, "loss": 0.1823, "step": 52440 }, { "epoch": 1.22, "learning_rate": 2.0949168055197016e-05, "loss": 0.1131, "step": 52445 }, { "epoch": 1.22, "learning_rate": 2.094828800075098e-05, "loss": 0.2422, "step": 52450 }, { "epoch": 1.22, "learning_rate": 2.0947407946304946e-05, "loss": 0.3117, "step": 52455 }, { "epoch": 1.22, "learning_rate": 2.094652789185891e-05, "loss": 0.2027, "step": 52460 }, { "epoch": 1.22, "learning_rate": 2.0945647837412872e-05, "loss": 0.3586, "step": 52465 }, { "epoch": 1.22, "learning_rate": 2.094476778296684e-05, "loss": 0.1131, "step": 52470 }, { "epoch": 1.22, "learning_rate": 2.0943887728520806e-05, "loss": 0.1385, "step": 52475 }, { "epoch": 1.22, "learning_rate": 2.094300767407477e-05, "loss": 0.045, "step": 52480 }, { "epoch": 1.22, "learning_rate": 2.0942127619628732e-05, "loss": 0.0973, "step": 52485 }, { "epoch": 1.22, "learning_rate": 2.09412475651827e-05, "loss": 0.1451, "step": 52490 }, { "epoch": 1.22, "learning_rate": 2.0940367510736666e-05, "loss": 0.1029, "step": 52495 }, { "epoch": 1.22, "learning_rate": 2.093948745629063e-05, "loss": 0.1621, "step": 52500 }, { "epoch": 1.23, "learning_rate": 2.0938607401844592e-05, "loss": 0.2143, "step": 52505 }, { "epoch": 1.23, "learning_rate": 2.093772734739856e-05, "loss": 0.55, "step": 52510 }, { "epoch": 1.23, "learning_rate": 2.0936847292952526e-05, "loss": 0.2699, "step": 52515 }, { "epoch": 1.23, "learning_rate": 2.0935967238506487e-05, "loss": 0.0708, "step": 52520 }, { "epoch": 1.23, "learning_rate": 2.0935087184060453e-05, "loss": 0.0407, "step": 52525 }, { "epoch": 1.23, "learning_rate": 2.093420712961442e-05, "loss": 0.0262, "step": 52530 }, { "epoch": 1.23, "learning_rate": 2.0933327075168386e-05, "loss": 0.0305, "step": 52535 }, { "epoch": 1.23, "learning_rate": 2.0932447020722348e-05, "loss": 0.1293, "step": 52540 }, { "epoch": 1.23, "learning_rate": 2.0931566966276313e-05, "loss": 0.1627, "step": 52545 }, { "epoch": 1.23, "learning_rate": 2.093068691183028e-05, "loss": 0.3234, "step": 52550 }, { "epoch": 1.23, "learning_rate": 2.0929806857384243e-05, "loss": 0.3355, "step": 52555 }, { "epoch": 1.23, "learning_rate": 2.0928926802938208e-05, "loss": 0.4025, "step": 52560 }, { "epoch": 1.23, "learning_rate": 2.0928046748492176e-05, "loss": 0.4861, "step": 52565 }, { "epoch": 1.23, "learning_rate": 2.092716669404614e-05, "loss": 0.0599, "step": 52570 }, { "epoch": 1.23, "learning_rate": 2.0926286639600103e-05, "loss": 0.0554, "step": 52575 }, { "epoch": 1.23, "learning_rate": 2.0925406585154068e-05, "loss": 0.0978, "step": 52580 }, { "epoch": 1.23, "learning_rate": 2.0924526530708036e-05, "loss": 0.0916, "step": 52585 }, { "epoch": 1.23, "learning_rate": 2.0923646476261998e-05, "loss": 0.2194, "step": 52590 }, { "epoch": 1.23, "learning_rate": 2.0922766421815963e-05, "loss": 0.0778, "step": 52595 }, { "epoch": 1.23, "learning_rate": 2.0921886367369928e-05, "loss": 0.1724, "step": 52600 }, { "epoch": 1.23, "learning_rate": 2.0921006312923896e-05, "loss": 0.2951, "step": 52605 }, { "epoch": 1.23, "learning_rate": 2.0920126258477858e-05, "loss": 0.255, "step": 52610 }, { "epoch": 1.23, "learning_rate": 2.0919246204031823e-05, "loss": 0.3057, "step": 52615 }, { "epoch": 1.23, "learning_rate": 2.0918366149585788e-05, "loss": 0.0406, "step": 52620 }, { "epoch": 1.23, "learning_rate": 2.0917486095139753e-05, "loss": 0.0907, "step": 52625 }, { "epoch": 1.23, "learning_rate": 2.0916606040693718e-05, "loss": 0.1577, "step": 52630 }, { "epoch": 1.23, "learning_rate": 2.0915725986247683e-05, "loss": 0.1501, "step": 52635 }, { "epoch": 1.23, "learning_rate": 2.0914845931801648e-05, "loss": 0.1519, "step": 52640 }, { "epoch": 1.23, "learning_rate": 2.0913965877355613e-05, "loss": 0.2199, "step": 52645 }, { "epoch": 1.23, "learning_rate": 2.0913085822909578e-05, "loss": 0.1272, "step": 52650 }, { "epoch": 1.23, "learning_rate": 2.0912205768463543e-05, "loss": 0.1917, "step": 52655 }, { "epoch": 1.23, "learning_rate": 2.0911325714017505e-05, "loss": 0.5279, "step": 52660 }, { "epoch": 1.23, "learning_rate": 2.0910445659571473e-05, "loss": 0.2868, "step": 52665 }, { "epoch": 1.23, "learning_rate": 2.0909565605125438e-05, "loss": 0.0667, "step": 52670 }, { "epoch": 1.23, "learning_rate": 2.0908685550679403e-05, "loss": 0.0589, "step": 52675 }, { "epoch": 1.23, "learning_rate": 2.0907805496233365e-05, "loss": 0.0455, "step": 52680 }, { "epoch": 1.23, "learning_rate": 2.0906925441787333e-05, "loss": 0.1469, "step": 52685 }, { "epoch": 1.23, "learning_rate": 2.0906045387341298e-05, "loss": 0.1138, "step": 52690 }, { "epoch": 1.23, "learning_rate": 2.090516533289526e-05, "loss": 0.266, "step": 52695 }, { "epoch": 1.23, "learning_rate": 2.0904285278449228e-05, "loss": 0.1183, "step": 52700 }, { "epoch": 1.23, "learning_rate": 2.0903405224003193e-05, "loss": 0.2349, "step": 52705 }, { "epoch": 1.23, "learning_rate": 2.0902525169557158e-05, "loss": 0.2811, "step": 52710 }, { "epoch": 1.23, "learning_rate": 2.090164511511112e-05, "loss": 0.2264, "step": 52715 }, { "epoch": 1.23, "learning_rate": 2.090076506066509e-05, "loss": 0.0238, "step": 52720 }, { "epoch": 1.23, "learning_rate": 2.0899885006219053e-05, "loss": 0.1324, "step": 52725 }, { "epoch": 1.23, "learning_rate": 2.0899004951773015e-05, "loss": 0.0818, "step": 52730 }, { "epoch": 1.23, "learning_rate": 2.089812489732698e-05, "loss": 0.1012, "step": 52735 }, { "epoch": 1.23, "learning_rate": 2.089724484288095e-05, "loss": 0.2028, "step": 52740 }, { "epoch": 1.23, "learning_rate": 2.0896364788434913e-05, "loss": 0.0795, "step": 52745 }, { "epoch": 1.23, "learning_rate": 2.0895484733988875e-05, "loss": 0.1406, "step": 52750 }, { "epoch": 1.23, "learning_rate": 2.089460467954284e-05, "loss": 0.2062, "step": 52755 }, { "epoch": 1.23, "learning_rate": 2.089372462509681e-05, "loss": 0.3726, "step": 52760 }, { "epoch": 1.23, "learning_rate": 2.089284457065077e-05, "loss": 0.2768, "step": 52765 }, { "epoch": 1.23, "learning_rate": 2.0891964516204735e-05, "loss": 0.1015, "step": 52770 }, { "epoch": 1.23, "learning_rate": 2.08910844617587e-05, "loss": 0.099, "step": 52775 }, { "epoch": 1.23, "learning_rate": 2.089020440731267e-05, "loss": 0.0547, "step": 52780 }, { "epoch": 1.23, "learning_rate": 2.088932435286663e-05, "loss": 0.107, "step": 52785 }, { "epoch": 1.23, "learning_rate": 2.0888444298420595e-05, "loss": 0.0405, "step": 52790 }, { "epoch": 1.23, "learning_rate": 2.088756424397456e-05, "loss": 0.1451, "step": 52795 }, { "epoch": 1.23, "learning_rate": 2.088668418952853e-05, "loss": 0.3501, "step": 52800 }, { "epoch": 1.23, "learning_rate": 2.088580413508249e-05, "loss": 0.5692, "step": 52805 }, { "epoch": 1.23, "learning_rate": 2.0884924080636455e-05, "loss": 0.4438, "step": 52810 }, { "epoch": 1.23, "learning_rate": 2.088404402619042e-05, "loss": 0.2683, "step": 52815 }, { "epoch": 1.23, "learning_rate": 2.0883163971744385e-05, "loss": 0.0345, "step": 52820 }, { "epoch": 1.23, "learning_rate": 2.088228391729835e-05, "loss": 0.0579, "step": 52825 }, { "epoch": 1.23, "learning_rate": 2.0881403862852315e-05, "loss": 0.1093, "step": 52830 }, { "epoch": 1.23, "learning_rate": 2.0880523808406284e-05, "loss": 0.0625, "step": 52835 }, { "epoch": 1.23, "learning_rate": 2.0879643753960245e-05, "loss": 0.0825, "step": 52840 }, { "epoch": 1.23, "learning_rate": 2.087876369951421e-05, "loss": 0.1065, "step": 52845 }, { "epoch": 1.23, "learning_rate": 2.0877883645068175e-05, "loss": 0.2493, "step": 52850 }, { "epoch": 1.23, "learning_rate": 2.087700359062214e-05, "loss": 0.2388, "step": 52855 }, { "epoch": 1.23, "learning_rate": 2.0876123536176106e-05, "loss": 0.4664, "step": 52860 }, { "epoch": 1.23, "learning_rate": 2.087524348173007e-05, "loss": 0.5758, "step": 52865 }, { "epoch": 1.23, "learning_rate": 2.0874363427284036e-05, "loss": 0.0807, "step": 52870 }, { "epoch": 1.23, "learning_rate": 2.0873483372838e-05, "loss": 0.0549, "step": 52875 }, { "epoch": 1.23, "learning_rate": 2.0872603318391966e-05, "loss": 0.0546, "step": 52880 }, { "epoch": 1.23, "learning_rate": 2.087172326394593e-05, "loss": 0.0816, "step": 52885 }, { "epoch": 1.23, "learning_rate": 2.0870843209499892e-05, "loss": 0.0483, "step": 52890 }, { "epoch": 1.23, "learning_rate": 2.086996315505386e-05, "loss": 0.1818, "step": 52895 }, { "epoch": 1.23, "learning_rate": 2.0869083100607826e-05, "loss": 0.1417, "step": 52900 }, { "epoch": 1.23, "learning_rate": 2.086820304616179e-05, "loss": 0.2329, "step": 52905 }, { "epoch": 1.23, "learning_rate": 2.0867322991715752e-05, "loss": 0.4853, "step": 52910 }, { "epoch": 1.23, "learning_rate": 2.086644293726972e-05, "loss": 0.5712, "step": 52915 }, { "epoch": 1.23, "learning_rate": 2.0865562882823686e-05, "loss": 0.0618, "step": 52920 }, { "epoch": 1.23, "learning_rate": 2.0864682828377647e-05, "loss": 0.0306, "step": 52925 }, { "epoch": 1.23, "learning_rate": 2.0863802773931612e-05, "loss": 0.1365, "step": 52930 }, { "epoch": 1.24, "learning_rate": 2.086292271948558e-05, "loss": 0.2041, "step": 52935 }, { "epoch": 1.24, "learning_rate": 2.0862042665039546e-05, "loss": 0.1064, "step": 52940 }, { "epoch": 1.24, "learning_rate": 2.0861162610593507e-05, "loss": 0.2006, "step": 52945 }, { "epoch": 1.24, "learning_rate": 2.0860282556147472e-05, "loss": 0.147, "step": 52950 }, { "epoch": 1.24, "learning_rate": 2.085940250170144e-05, "loss": 0.1633, "step": 52955 }, { "epoch": 1.24, "learning_rate": 2.0858522447255403e-05, "loss": 0.2303, "step": 52960 }, { "epoch": 1.24, "learning_rate": 2.0857642392809368e-05, "loss": 0.2365, "step": 52965 }, { "epoch": 1.24, "learning_rate": 2.0856762338363336e-05, "loss": 0.0531, "step": 52970 }, { "epoch": 1.24, "learning_rate": 2.08558822839173e-05, "loss": 0.1066, "step": 52975 }, { "epoch": 1.24, "learning_rate": 2.0855002229471263e-05, "loss": 0.147, "step": 52980 }, { "epoch": 1.24, "learning_rate": 2.0854122175025228e-05, "loss": 0.1803, "step": 52985 }, { "epoch": 1.24, "learning_rate": 2.0853242120579196e-05, "loss": 0.1849, "step": 52990 }, { "epoch": 1.24, "learning_rate": 2.0852362066133158e-05, "loss": 0.0497, "step": 52995 }, { "epoch": 1.24, "learning_rate": 2.0851482011687123e-05, "loss": 0.2212, "step": 53000 }, { "epoch": 1.24, "learning_rate": 2.0850601957241088e-05, "loss": 0.1312, "step": 53005 }, { "epoch": 1.24, "learning_rate": 2.0849721902795056e-05, "loss": 0.474, "step": 53010 }, { "epoch": 1.24, "learning_rate": 2.0848841848349018e-05, "loss": 0.3599, "step": 53015 }, { "epoch": 1.24, "learning_rate": 2.0847961793902983e-05, "loss": 0.1546, "step": 53020 }, { "epoch": 1.24, "learning_rate": 2.0847081739456948e-05, "loss": 0.0635, "step": 53025 }, { "epoch": 1.24, "learning_rate": 2.0846201685010913e-05, "loss": 0.0721, "step": 53030 }, { "epoch": 1.24, "learning_rate": 2.0845321630564878e-05, "loss": 0.1162, "step": 53035 }, { "epoch": 1.24, "learning_rate": 2.0844441576118843e-05, "loss": 0.0986, "step": 53040 }, { "epoch": 1.24, "learning_rate": 2.0843561521672808e-05, "loss": 0.1164, "step": 53045 }, { "epoch": 1.24, "learning_rate": 2.0842681467226773e-05, "loss": 0.1725, "step": 53050 }, { "epoch": 1.24, "learning_rate": 2.0841801412780738e-05, "loss": 0.1637, "step": 53055 }, { "epoch": 1.24, "learning_rate": 2.0840921358334703e-05, "loss": 0.3785, "step": 53060 }, { "epoch": 1.24, "learning_rate": 2.0840041303888665e-05, "loss": 0.2234, "step": 53065 }, { "epoch": 1.24, "learning_rate": 2.0839161249442633e-05, "loss": 0.0495, "step": 53070 }, { "epoch": 1.24, "learning_rate": 2.0838281194996598e-05, "loss": 0.0463, "step": 53075 }, { "epoch": 1.24, "learning_rate": 2.0837401140550563e-05, "loss": 0.0878, "step": 53080 }, { "epoch": 1.24, "learning_rate": 2.0836521086104525e-05, "loss": 0.2172, "step": 53085 }, { "epoch": 1.24, "learning_rate": 2.0835641031658493e-05, "loss": 0.2068, "step": 53090 }, { "epoch": 1.24, "learning_rate": 2.0834760977212458e-05, "loss": 0.1324, "step": 53095 }, { "epoch": 1.24, "learning_rate": 2.0833880922766423e-05, "loss": 0.1506, "step": 53100 }, { "epoch": 1.24, "learning_rate": 2.0833000868320388e-05, "loss": 0.2323, "step": 53105 }, { "epoch": 1.24, "learning_rate": 2.0832120813874353e-05, "loss": 0.8237, "step": 53110 }, { "epoch": 1.24, "learning_rate": 2.0831240759428318e-05, "loss": 0.3555, "step": 53115 }, { "epoch": 1.24, "learning_rate": 2.083036070498228e-05, "loss": 0.0682, "step": 53120 }, { "epoch": 1.24, "learning_rate": 2.0829480650536248e-05, "loss": 0.0948, "step": 53125 }, { "epoch": 1.24, "learning_rate": 2.0828600596090213e-05, "loss": 0.0789, "step": 53130 }, { "epoch": 1.24, "learning_rate": 2.0827720541644178e-05, "loss": 0.0799, "step": 53135 }, { "epoch": 1.24, "learning_rate": 2.082684048719814e-05, "loss": 0.117, "step": 53140 }, { "epoch": 1.24, "learning_rate": 2.0825960432752108e-05, "loss": 0.0975, "step": 53145 }, { "epoch": 1.24, "learning_rate": 2.0825080378306073e-05, "loss": 0.1074, "step": 53150 }, { "epoch": 1.24, "learning_rate": 2.0824200323860035e-05, "loss": 0.1221, "step": 53155 }, { "epoch": 1.24, "learning_rate": 2.0823320269414e-05, "loss": 0.1988, "step": 53160 }, { "epoch": 1.24, "learning_rate": 2.082244021496797e-05, "loss": 0.4021, "step": 53165 }, { "epoch": 1.24, "learning_rate": 2.0821560160521933e-05, "loss": 0.0497, "step": 53170 }, { "epoch": 1.24, "learning_rate": 2.0820680106075895e-05, "loss": 0.104, "step": 53175 }, { "epoch": 1.24, "learning_rate": 2.081980005162986e-05, "loss": 0.134, "step": 53180 }, { "epoch": 1.24, "learning_rate": 2.081891999718383e-05, "loss": 0.1353, "step": 53185 }, { "epoch": 1.24, "learning_rate": 2.081803994273779e-05, "loss": 0.1463, "step": 53190 }, { "epoch": 1.24, "learning_rate": 2.0817159888291755e-05, "loss": 0.1699, "step": 53195 }, { "epoch": 1.24, "learning_rate": 2.081627983384572e-05, "loss": 0.1714, "step": 53200 }, { "epoch": 1.24, "learning_rate": 2.081539977939969e-05, "loss": 0.1698, "step": 53205 }, { "epoch": 1.24, "learning_rate": 2.081451972495365e-05, "loss": 0.4742, "step": 53210 }, { "epoch": 1.24, "learning_rate": 2.0813639670507615e-05, "loss": 0.2623, "step": 53215 }, { "epoch": 1.24, "learning_rate": 2.081275961606158e-05, "loss": 0.1304, "step": 53220 }, { "epoch": 1.24, "learning_rate": 2.0811879561615545e-05, "loss": 0.0222, "step": 53225 }, { "epoch": 1.24, "learning_rate": 2.081099950716951e-05, "loss": 0.1157, "step": 53230 }, { "epoch": 1.24, "learning_rate": 2.0810119452723475e-05, "loss": 0.1299, "step": 53235 }, { "epoch": 1.24, "learning_rate": 2.0809239398277444e-05, "loss": 0.0918, "step": 53240 }, { "epoch": 1.24, "learning_rate": 2.0808359343831405e-05, "loss": 0.0431, "step": 53245 }, { "epoch": 1.24, "learning_rate": 2.080747928938537e-05, "loss": 0.0952, "step": 53250 }, { "epoch": 1.24, "learning_rate": 2.0806599234939335e-05, "loss": 0.2508, "step": 53255 }, { "epoch": 1.24, "learning_rate": 2.08057191804933e-05, "loss": 0.3154, "step": 53260 }, { "epoch": 1.24, "learning_rate": 2.0804839126047265e-05, "loss": 0.1979, "step": 53265 }, { "epoch": 1.24, "learning_rate": 2.080395907160123e-05, "loss": 0.074, "step": 53270 }, { "epoch": 1.24, "learning_rate": 2.0803079017155195e-05, "loss": 0.1035, "step": 53275 }, { "epoch": 1.24, "learning_rate": 2.080219896270916e-05, "loss": 0.1375, "step": 53280 }, { "epoch": 1.24, "learning_rate": 2.0801318908263125e-05, "loss": 0.0457, "step": 53285 }, { "epoch": 1.24, "learning_rate": 2.080043885381709e-05, "loss": 0.2443, "step": 53290 }, { "epoch": 1.24, "learning_rate": 2.0799558799371052e-05, "loss": 0.1169, "step": 53295 }, { "epoch": 1.24, "learning_rate": 2.079867874492502e-05, "loss": 0.1322, "step": 53300 }, { "epoch": 1.24, "learning_rate": 2.0797798690478986e-05, "loss": 0.3009, "step": 53305 }, { "epoch": 1.24, "learning_rate": 2.079691863603295e-05, "loss": 0.5731, "step": 53310 }, { "epoch": 1.24, "learning_rate": 2.0796038581586912e-05, "loss": 0.2488, "step": 53315 }, { "epoch": 1.24, "learning_rate": 2.079515852714088e-05, "loss": 0.1007, "step": 53320 }, { "epoch": 1.24, "learning_rate": 2.0794278472694846e-05, "loss": 0.1249, "step": 53325 }, { "epoch": 1.24, "learning_rate": 2.0793398418248807e-05, "loss": 0.0979, "step": 53330 }, { "epoch": 1.24, "learning_rate": 2.0792518363802772e-05, "loss": 0.173, "step": 53335 }, { "epoch": 1.24, "learning_rate": 2.079163830935674e-05, "loss": 0.157, "step": 53340 }, { "epoch": 1.24, "learning_rate": 2.0790758254910706e-05, "loss": 0.1192, "step": 53345 }, { "epoch": 1.24, "learning_rate": 2.0789878200464667e-05, "loss": 0.1571, "step": 53350 }, { "epoch": 1.24, "learning_rate": 2.0788998146018632e-05, "loss": 0.2039, "step": 53355 }, { "epoch": 1.24, "learning_rate": 2.07881180915726e-05, "loss": 0.3709, "step": 53360 }, { "epoch": 1.25, "learning_rate": 2.0787238037126562e-05, "loss": 0.2058, "step": 53365 }, { "epoch": 1.25, "learning_rate": 2.0786357982680527e-05, "loss": 0.0301, "step": 53370 }, { "epoch": 1.25, "learning_rate": 2.0785477928234496e-05, "loss": 0.1379, "step": 53375 }, { "epoch": 1.25, "learning_rate": 2.078459787378846e-05, "loss": 0.0617, "step": 53380 }, { "epoch": 1.25, "learning_rate": 2.0783717819342422e-05, "loss": 0.0327, "step": 53385 }, { "epoch": 1.25, "learning_rate": 2.0782837764896388e-05, "loss": 0.1888, "step": 53390 }, { "epoch": 1.25, "learning_rate": 2.0781957710450356e-05, "loss": 0.3213, "step": 53395 }, { "epoch": 1.25, "learning_rate": 2.078107765600432e-05, "loss": 0.5033, "step": 53400 }, { "epoch": 1.25, "learning_rate": 2.0780197601558283e-05, "loss": 0.3145, "step": 53405 }, { "epoch": 1.25, "learning_rate": 2.0779317547112248e-05, "loss": 0.3866, "step": 53410 }, { "epoch": 1.25, "learning_rate": 2.0778437492666216e-05, "loss": 0.427, "step": 53415 }, { "epoch": 1.25, "learning_rate": 2.0777557438220178e-05, "loss": 0.0917, "step": 53420 }, { "epoch": 1.25, "learning_rate": 2.0776677383774143e-05, "loss": 0.0518, "step": 53425 }, { "epoch": 1.25, "learning_rate": 2.0775797329328108e-05, "loss": 0.0356, "step": 53430 }, { "epoch": 1.25, "learning_rate": 2.0774917274882076e-05, "loss": 0.0776, "step": 53435 }, { "epoch": 1.25, "learning_rate": 2.0774037220436038e-05, "loss": 0.0684, "step": 53440 }, { "epoch": 1.25, "learning_rate": 2.0773157165990003e-05, "loss": 0.1752, "step": 53445 }, { "epoch": 1.25, "learning_rate": 2.0772277111543968e-05, "loss": 0.1345, "step": 53450 }, { "epoch": 1.25, "learning_rate": 2.0771397057097933e-05, "loss": 0.253, "step": 53455 }, { "epoch": 1.25, "learning_rate": 2.0770517002651898e-05, "loss": 0.4194, "step": 53460 }, { "epoch": 1.25, "learning_rate": 2.0769636948205863e-05, "loss": 0.3432, "step": 53465 }, { "epoch": 1.25, "learning_rate": 2.0768756893759828e-05, "loss": 0.0656, "step": 53470 }, { "epoch": 1.25, "learning_rate": 2.0768052850203e-05, "loss": 0.1708, "step": 53475 }, { "epoch": 1.25, "learning_rate": 2.0767172795756964e-05, "loss": 0.1492, "step": 53480 }, { "epoch": 1.25, "learning_rate": 2.076629274131093e-05, "loss": 0.144, "step": 53485 }, { "epoch": 1.25, "learning_rate": 2.0765412686864894e-05, "loss": 0.1875, "step": 53490 }, { "epoch": 1.25, "learning_rate": 2.076453263241886e-05, "loss": 0.1053, "step": 53495 }, { "epoch": 1.25, "learning_rate": 2.0763652577972824e-05, "loss": 0.2283, "step": 53500 }, { "epoch": 1.25, "learning_rate": 2.076277252352679e-05, "loss": 0.4542, "step": 53505 }, { "epoch": 1.25, "learning_rate": 2.0761892469080754e-05, "loss": 0.462, "step": 53510 }, { "epoch": 1.25, "learning_rate": 2.076101241463472e-05, "loss": 0.5235, "step": 53515 }, { "epoch": 1.25, "learning_rate": 2.0760132360188684e-05, "loss": 0.0269, "step": 53520 }, { "epoch": 1.25, "learning_rate": 2.0759252305742646e-05, "loss": 0.1365, "step": 53525 }, { "epoch": 1.25, "learning_rate": 2.0758372251296614e-05, "loss": 0.0611, "step": 53530 }, { "epoch": 1.25, "learning_rate": 2.075749219685058e-05, "loss": 0.0801, "step": 53535 }, { "epoch": 1.25, "learning_rate": 2.0756612142404544e-05, "loss": 0.0814, "step": 53540 }, { "epoch": 1.25, "learning_rate": 2.0755732087958506e-05, "loss": 0.1841, "step": 53545 }, { "epoch": 1.25, "learning_rate": 2.0754852033512474e-05, "loss": 0.14, "step": 53550 }, { "epoch": 1.25, "learning_rate": 2.075397197906644e-05, "loss": 0.2234, "step": 53555 }, { "epoch": 1.25, "learning_rate": 2.07530919246204e-05, "loss": 0.2765, "step": 53560 }, { "epoch": 1.25, "learning_rate": 2.075221187017437e-05, "loss": 0.2435, "step": 53565 }, { "epoch": 1.25, "learning_rate": 2.0751331815728335e-05, "loss": 0.119, "step": 53570 }, { "epoch": 1.25, "learning_rate": 2.07504517612823e-05, "loss": 0.0238, "step": 53575 }, { "epoch": 1.25, "learning_rate": 2.074957170683626e-05, "loss": 0.1336, "step": 53580 }, { "epoch": 1.25, "learning_rate": 2.074869165239023e-05, "loss": 0.0946, "step": 53585 }, { "epoch": 1.25, "learning_rate": 2.0747811597944195e-05, "loss": 0.0477, "step": 53590 }, { "epoch": 1.25, "learning_rate": 2.074693154349816e-05, "loss": 0.1221, "step": 53595 }, { "epoch": 1.25, "learning_rate": 2.074605148905212e-05, "loss": 0.1332, "step": 53600 }, { "epoch": 1.25, "learning_rate": 2.074517143460609e-05, "loss": 0.135, "step": 53605 }, { "epoch": 1.25, "learning_rate": 2.0744291380160055e-05, "loss": 0.602, "step": 53610 }, { "epoch": 1.25, "learning_rate": 2.0743411325714016e-05, "loss": 0.255, "step": 53615 }, { "epoch": 1.25, "learning_rate": 2.074253127126798e-05, "loss": 0.1518, "step": 53620 }, { "epoch": 1.25, "learning_rate": 2.074165121682195e-05, "loss": 0.0742, "step": 53625 }, { "epoch": 1.25, "learning_rate": 2.0740771162375915e-05, "loss": 0.0314, "step": 53630 }, { "epoch": 1.25, "learning_rate": 2.0739891107929876e-05, "loss": 0.1484, "step": 53635 }, { "epoch": 1.25, "learning_rate": 2.073901105348384e-05, "loss": 0.1903, "step": 53640 }, { "epoch": 1.25, "learning_rate": 2.073813099903781e-05, "loss": 0.1383, "step": 53645 }, { "epoch": 1.25, "learning_rate": 2.073725094459177e-05, "loss": 0.1543, "step": 53650 }, { "epoch": 1.25, "learning_rate": 2.0736370890145737e-05, "loss": 0.223, "step": 53655 }, { "epoch": 1.25, "learning_rate": 2.07354908356997e-05, "loss": 0.4795, "step": 53660 }, { "epoch": 1.25, "learning_rate": 2.073461078125367e-05, "loss": 0.1857, "step": 53665 }, { "epoch": 1.25, "learning_rate": 2.073373072680763e-05, "loss": 0.0517, "step": 53670 }, { "epoch": 1.25, "learning_rate": 2.0732850672361597e-05, "loss": 0.0668, "step": 53675 }, { "epoch": 1.25, "learning_rate": 2.073197061791556e-05, "loss": 0.102, "step": 53680 }, { "epoch": 1.25, "learning_rate": 2.0731090563469527e-05, "loss": 0.1419, "step": 53685 }, { "epoch": 1.25, "learning_rate": 2.073021050902349e-05, "loss": 0.2151, "step": 53690 }, { "epoch": 1.25, "learning_rate": 2.0729330454577457e-05, "loss": 0.0829, "step": 53695 }, { "epoch": 1.25, "learning_rate": 2.0728450400131425e-05, "loss": 0.0522, "step": 53700 }, { "epoch": 1.25, "learning_rate": 2.0727570345685387e-05, "loss": 0.0463, "step": 53705 }, { "epoch": 1.25, "learning_rate": 2.0726690291239352e-05, "loss": 0.3531, "step": 53710 }, { "epoch": 1.25, "learning_rate": 2.0725810236793317e-05, "loss": 0.3693, "step": 53715 }, { "epoch": 1.25, "learning_rate": 2.0724930182347282e-05, "loss": 0.1611, "step": 53720 }, { "epoch": 1.25, "learning_rate": 2.0724050127901247e-05, "loss": 0.0371, "step": 53725 }, { "epoch": 1.25, "learning_rate": 2.0723170073455212e-05, "loss": 0.0733, "step": 53730 }, { "epoch": 1.25, "learning_rate": 2.0722290019009177e-05, "loss": 0.0732, "step": 53735 }, { "epoch": 1.25, "learning_rate": 2.0721409964563142e-05, "loss": 0.1317, "step": 53740 }, { "epoch": 1.25, "learning_rate": 2.0720529910117107e-05, "loss": 0.1559, "step": 53745 }, { "epoch": 1.25, "learning_rate": 2.0719649855671072e-05, "loss": 0.1947, "step": 53750 }, { "epoch": 1.25, "learning_rate": 2.0718769801225034e-05, "loss": 0.0986, "step": 53755 }, { "epoch": 1.25, "learning_rate": 2.0717889746779002e-05, "loss": 0.5937, "step": 53760 }, { "epoch": 1.25, "learning_rate": 2.0717009692332967e-05, "loss": 0.1968, "step": 53765 }, { "epoch": 1.25, "learning_rate": 2.0716129637886932e-05, "loss": 0.1027, "step": 53770 }, { "epoch": 1.25, "learning_rate": 2.0715249583440894e-05, "loss": 0.0199, "step": 53775 }, { "epoch": 1.25, "learning_rate": 2.0714369528994862e-05, "loss": 0.2082, "step": 53780 }, { "epoch": 1.25, "learning_rate": 2.0713489474548827e-05, "loss": 0.018, "step": 53785 }, { "epoch": 1.25, "learning_rate": 2.071260942010279e-05, "loss": 0.1283, "step": 53790 }, { "epoch": 1.26, "learning_rate": 2.0711729365656754e-05, "loss": 0.0922, "step": 53795 }, { "epoch": 1.26, "learning_rate": 2.0710849311210722e-05, "loss": 0.0894, "step": 53800 }, { "epoch": 1.26, "learning_rate": 2.0709969256764687e-05, "loss": 0.2827, "step": 53805 }, { "epoch": 1.26, "learning_rate": 2.070908920231865e-05, "loss": 0.3261, "step": 53810 }, { "epoch": 1.26, "learning_rate": 2.0708209147872614e-05, "loss": 0.1899, "step": 53815 }, { "epoch": 1.26, "learning_rate": 2.0707329093426582e-05, "loss": 0.0952, "step": 53820 }, { "epoch": 1.26, "learning_rate": 2.0706449038980544e-05, "loss": 0.1297, "step": 53825 }, { "epoch": 1.26, "learning_rate": 2.070556898453451e-05, "loss": 0.0588, "step": 53830 }, { "epoch": 1.26, "learning_rate": 2.0704688930088477e-05, "loss": 0.1989, "step": 53835 }, { "epoch": 1.26, "learning_rate": 2.0703808875642442e-05, "loss": 0.1, "step": 53840 }, { "epoch": 1.26, "learning_rate": 2.0702928821196404e-05, "loss": 0.0315, "step": 53845 }, { "epoch": 1.26, "learning_rate": 2.070204876675037e-05, "loss": 0.3397, "step": 53850 }, { "epoch": 1.26, "learning_rate": 2.0701168712304337e-05, "loss": 0.1756, "step": 53855 }, { "epoch": 1.26, "learning_rate": 2.07002886578583e-05, "loss": 0.473, "step": 53860 }, { "epoch": 1.26, "learning_rate": 2.0699408603412264e-05, "loss": 0.2482, "step": 53865 }, { "epoch": 1.26, "learning_rate": 2.069852854896623e-05, "loss": 0.1327, "step": 53870 }, { "epoch": 1.26, "learning_rate": 2.0697648494520197e-05, "loss": 0.0322, "step": 53875 }, { "epoch": 1.26, "learning_rate": 2.069676844007416e-05, "loss": 0.0634, "step": 53880 }, { "epoch": 1.26, "learning_rate": 2.0695888385628124e-05, "loss": 0.1984, "step": 53885 }, { "epoch": 1.26, "learning_rate": 2.069500833118209e-05, "loss": 0.1654, "step": 53890 }, { "epoch": 1.26, "learning_rate": 2.0694128276736058e-05, "loss": 0.1541, "step": 53895 }, { "epoch": 1.26, "learning_rate": 2.069324822229002e-05, "loss": 0.1654, "step": 53900 }, { "epoch": 1.26, "learning_rate": 2.0692368167843984e-05, "loss": 0.4056, "step": 53905 }, { "epoch": 1.26, "learning_rate": 2.069148811339795e-05, "loss": 0.1486, "step": 53910 }, { "epoch": 1.26, "learning_rate": 2.0690608058951914e-05, "loss": 0.2234, "step": 53915 }, { "epoch": 1.26, "learning_rate": 2.068972800450588e-05, "loss": 0.0976, "step": 53920 }, { "epoch": 1.26, "learning_rate": 2.0688847950059844e-05, "loss": 0.0655, "step": 53925 }, { "epoch": 1.26, "learning_rate": 2.068796789561381e-05, "loss": 0.0971, "step": 53930 }, { "epoch": 1.26, "learning_rate": 2.0687087841167774e-05, "loss": 0.1083, "step": 53935 }, { "epoch": 1.26, "learning_rate": 2.068620778672174e-05, "loss": 0.0799, "step": 53940 }, { "epoch": 1.26, "learning_rate": 2.0685327732275704e-05, "loss": 0.1071, "step": 53945 }, { "epoch": 1.26, "learning_rate": 2.0684447677829666e-05, "loss": 0.2208, "step": 53950 }, { "epoch": 1.26, "learning_rate": 2.0683567623383634e-05, "loss": 0.1329, "step": 53955 }, { "epoch": 1.26, "learning_rate": 2.06826875689376e-05, "loss": 0.2656, "step": 53960 }, { "epoch": 1.26, "learning_rate": 2.0681807514491564e-05, "loss": 0.3165, "step": 53965 }, { "epoch": 1.26, "learning_rate": 2.068092746004553e-05, "loss": 0.1049, "step": 53970 }, { "epoch": 1.26, "learning_rate": 2.0680047405599494e-05, "loss": 0.0392, "step": 53975 }, { "epoch": 1.26, "learning_rate": 2.067916735115346e-05, "loss": 0.0609, "step": 53980 }, { "epoch": 1.26, "learning_rate": 2.067828729670742e-05, "loss": 0.0503, "step": 53985 }, { "epoch": 1.26, "learning_rate": 2.067740724226139e-05, "loss": 0.1356, "step": 53990 }, { "epoch": 1.26, "learning_rate": 2.0676527187815355e-05, "loss": 0.1305, "step": 53995 }, { "epoch": 1.26, "learning_rate": 2.067564713336932e-05, "loss": 0.1571, "step": 54000 }, { "epoch": 1.26, "learning_rate": 2.067476707892328e-05, "loss": 0.1326, "step": 54005 }, { "epoch": 1.26, "learning_rate": 2.067388702447725e-05, "loss": 0.586, "step": 54010 }, { "epoch": 1.26, "learning_rate": 2.0673006970031215e-05, "loss": 0.2418, "step": 54015 }, { "epoch": 1.26, "learning_rate": 2.0672126915585176e-05, "loss": 0.0529, "step": 54020 }, { "epoch": 1.26, "learning_rate": 2.067124686113914e-05, "loss": 0.0093, "step": 54025 }, { "epoch": 1.26, "learning_rate": 2.067036680669311e-05, "loss": 0.0865, "step": 54030 }, { "epoch": 1.26, "learning_rate": 2.0669486752247075e-05, "loss": 0.071, "step": 54035 }, { "epoch": 1.26, "learning_rate": 2.0668606697801036e-05, "loss": 0.154, "step": 54040 }, { "epoch": 1.26, "learning_rate": 2.0667726643355e-05, "loss": 0.1477, "step": 54045 }, { "epoch": 1.26, "learning_rate": 2.066684658890897e-05, "loss": 0.1561, "step": 54050 }, { "epoch": 1.26, "learning_rate": 2.066596653446293e-05, "loss": 0.3198, "step": 54055 }, { "epoch": 1.26, "learning_rate": 2.0665086480016896e-05, "loss": 0.5042, "step": 54060 }, { "epoch": 1.26, "learning_rate": 2.066420642557086e-05, "loss": 0.2623, "step": 54065 }, { "epoch": 1.26, "learning_rate": 2.066332637112483e-05, "loss": 0.0973, "step": 54070 }, { "epoch": 1.26, "learning_rate": 2.066244631667879e-05, "loss": 0.0367, "step": 54075 }, { "epoch": 1.26, "learning_rate": 2.0661566262232757e-05, "loss": 0.1108, "step": 54080 }, { "epoch": 1.26, "learning_rate": 2.066068620778672e-05, "loss": 0.091, "step": 54085 }, { "epoch": 1.26, "learning_rate": 2.0659806153340687e-05, "loss": 0.1079, "step": 54090 }, { "epoch": 1.26, "learning_rate": 2.065892609889465e-05, "loss": 0.153, "step": 54095 }, { "epoch": 1.26, "learning_rate": 2.0658046044448617e-05, "loss": 0.1438, "step": 54100 }, { "epoch": 1.26, "learning_rate": 2.0657165990002585e-05, "loss": 0.1443, "step": 54105 }, { "epoch": 1.26, "learning_rate": 2.0656285935556547e-05, "loss": 0.3082, "step": 54110 }, { "epoch": 1.26, "learning_rate": 2.065540588111051e-05, "loss": 0.2836, "step": 54115 }, { "epoch": 1.26, "learning_rate": 2.0654525826664477e-05, "loss": 0.0626, "step": 54120 }, { "epoch": 1.26, "learning_rate": 2.065364577221844e-05, "loss": 0.1043, "step": 54125 }, { "epoch": 1.26, "learning_rate": 2.0652765717772407e-05, "loss": 0.1145, "step": 54130 }, { "epoch": 1.26, "learning_rate": 2.0651885663326372e-05, "loss": 0.0802, "step": 54135 }, { "epoch": 1.26, "learning_rate": 2.0651005608880337e-05, "loss": 0.1055, "step": 54140 }, { "epoch": 1.26, "learning_rate": 2.0650125554434302e-05, "loss": 0.1373, "step": 54145 }, { "epoch": 1.26, "learning_rate": 2.0649245499988267e-05, "loss": 0.2597, "step": 54150 }, { "epoch": 1.26, "learning_rate": 2.0648365445542232e-05, "loss": 0.0915, "step": 54155 }, { "epoch": 1.26, "learning_rate": 2.0647485391096193e-05, "loss": 0.3196, "step": 54160 }, { "epoch": 1.26, "learning_rate": 2.0646605336650162e-05, "loss": 0.2044, "step": 54165 }, { "epoch": 1.26, "learning_rate": 2.0645725282204127e-05, "loss": 0.1311, "step": 54170 }, { "epoch": 1.26, "learning_rate": 2.0644845227758092e-05, "loss": 0.0796, "step": 54175 }, { "epoch": 1.26, "learning_rate": 2.0643965173312054e-05, "loss": 0.0206, "step": 54180 }, { "epoch": 1.26, "learning_rate": 2.0643085118866022e-05, "loss": 0.1123, "step": 54185 }, { "epoch": 1.26, "learning_rate": 2.0642205064419987e-05, "loss": 0.1323, "step": 54190 }, { "epoch": 1.26, "learning_rate": 2.0641325009973952e-05, "loss": 0.0908, "step": 54195 }, { "epoch": 1.26, "learning_rate": 2.0640444955527914e-05, "loss": 0.1065, "step": 54200 }, { "epoch": 1.26, "learning_rate": 2.0639564901081882e-05, "loss": 0.2653, "step": 54205 }, { "epoch": 1.26, "learning_rate": 2.0638684846635847e-05, "loss": 0.7536, "step": 54210 }, { "epoch": 1.26, "learning_rate": 2.063780479218981e-05, "loss": 0.2759, "step": 54215 }, { "epoch": 1.27, "learning_rate": 2.0636924737743774e-05, "loss": 0.031, "step": 54220 }, { "epoch": 1.27, "learning_rate": 2.0636044683297742e-05, "loss": 0.0018, "step": 54225 }, { "epoch": 1.27, "learning_rate": 2.0635164628851707e-05, "loss": 0.1496, "step": 54230 }, { "epoch": 1.27, "learning_rate": 2.063428457440567e-05, "loss": 0.0461, "step": 54235 }, { "epoch": 1.27, "learning_rate": 2.0633404519959637e-05, "loss": 0.2096, "step": 54240 }, { "epoch": 1.27, "learning_rate": 2.0632524465513602e-05, "loss": 0.0922, "step": 54245 }, { "epoch": 1.27, "learning_rate": 2.0631644411067564e-05, "loss": 0.1649, "step": 54250 }, { "epoch": 1.27, "learning_rate": 2.063076435662153e-05, "loss": 0.121, "step": 54255 }, { "epoch": 1.27, "learning_rate": 2.0629884302175497e-05, "loss": 0.2196, "step": 54260 }, { "epoch": 1.27, "learning_rate": 2.0629004247729462e-05, "loss": 0.2864, "step": 54265 }, { "epoch": 1.27, "learning_rate": 2.0628124193283424e-05, "loss": 0.0903, "step": 54270 }, { "epoch": 1.27, "learning_rate": 2.062724413883739e-05, "loss": 0.0698, "step": 54275 }, { "epoch": 1.27, "learning_rate": 2.0626364084391357e-05, "loss": 0.0695, "step": 54280 }, { "epoch": 1.27, "learning_rate": 2.062548402994532e-05, "loss": 0.099, "step": 54285 }, { "epoch": 1.27, "learning_rate": 2.0624603975499284e-05, "loss": 0.1046, "step": 54290 }, { "epoch": 1.27, "learning_rate": 2.062372392105325e-05, "loss": 0.1902, "step": 54295 }, { "epoch": 1.27, "learning_rate": 2.0622843866607217e-05, "loss": 0.27, "step": 54300 }, { "epoch": 1.27, "learning_rate": 2.062196381216118e-05, "loss": 0.3552, "step": 54305 }, { "epoch": 1.27, "learning_rate": 2.0621083757715144e-05, "loss": 0.2901, "step": 54310 }, { "epoch": 1.27, "learning_rate": 2.062020370326911e-05, "loss": 0.28, "step": 54315 }, { "epoch": 1.27, "learning_rate": 2.0619323648823074e-05, "loss": 0.0549, "step": 54320 }, { "epoch": 1.27, "learning_rate": 2.061844359437704e-05, "loss": 0.1036, "step": 54325 }, { "epoch": 1.27, "learning_rate": 2.0617563539931004e-05, "loss": 0.0841, "step": 54330 }, { "epoch": 1.27, "learning_rate": 2.061668348548497e-05, "loss": 0.0382, "step": 54335 }, { "epoch": 1.27, "learning_rate": 2.0615803431038934e-05, "loss": 0.1326, "step": 54340 }, { "epoch": 1.27, "learning_rate": 2.06149233765929e-05, "loss": 0.2581, "step": 54345 }, { "epoch": 1.27, "learning_rate": 2.0614043322146864e-05, "loss": 0.1186, "step": 54350 }, { "epoch": 1.27, "learning_rate": 2.0613163267700826e-05, "loss": 0.2499, "step": 54355 }, { "epoch": 1.27, "learning_rate": 2.0612283213254794e-05, "loss": 0.2867, "step": 54360 }, { "epoch": 1.27, "learning_rate": 2.061140315880876e-05, "loss": 0.2358, "step": 54365 }, { "epoch": 1.27, "learning_rate": 2.0610523104362724e-05, "loss": 0.0428, "step": 54370 }, { "epoch": 1.27, "learning_rate": 2.060964304991669e-05, "loss": 0.044, "step": 54375 }, { "epoch": 1.27, "learning_rate": 2.0608762995470654e-05, "loss": 0.1303, "step": 54380 }, { "epoch": 1.27, "learning_rate": 2.060788294102462e-05, "loss": 0.0704, "step": 54385 }, { "epoch": 1.27, "learning_rate": 2.060700288657858e-05, "loss": 0.1042, "step": 54390 }, { "epoch": 1.27, "learning_rate": 2.060612283213255e-05, "loss": 0.1192, "step": 54395 }, { "epoch": 1.27, "learning_rate": 2.0605242777686514e-05, "loss": 0.3046, "step": 54400 }, { "epoch": 1.27, "learning_rate": 2.060436272324048e-05, "loss": 0.3047, "step": 54405 }, { "epoch": 1.27, "learning_rate": 2.060348266879444e-05, "loss": 0.3547, "step": 54410 }, { "epoch": 1.27, "learning_rate": 2.060260261434841e-05, "loss": 0.2365, "step": 54415 }, { "epoch": 1.27, "learning_rate": 2.0601722559902375e-05, "loss": 0.1283, "step": 54420 }, { "epoch": 1.27, "learning_rate": 2.0600842505456336e-05, "loss": 0.0751, "step": 54425 }, { "epoch": 1.27, "learning_rate": 2.05999624510103e-05, "loss": 0.0631, "step": 54430 }, { "epoch": 1.27, "learning_rate": 2.059908239656427e-05, "loss": 0.1213, "step": 54435 }, { "epoch": 1.27, "learning_rate": 2.0598202342118235e-05, "loss": 0.2563, "step": 54440 }, { "epoch": 1.27, "learning_rate": 2.0597322287672196e-05, "loss": 0.1224, "step": 54445 }, { "epoch": 1.27, "learning_rate": 2.059644223322616e-05, "loss": 0.0634, "step": 54450 }, { "epoch": 1.27, "learning_rate": 2.059556217878013e-05, "loss": 0.2141, "step": 54455 }, { "epoch": 1.27, "learning_rate": 2.059468212433409e-05, "loss": 0.3663, "step": 54460 }, { "epoch": 1.27, "learning_rate": 2.0593802069888056e-05, "loss": 0.2226, "step": 54465 }, { "epoch": 1.27, "learning_rate": 2.059292201544202e-05, "loss": 0.0594, "step": 54470 }, { "epoch": 1.27, "learning_rate": 2.059204196099599e-05, "loss": 0.0183, "step": 54475 }, { "epoch": 1.27, "learning_rate": 2.059116190654995e-05, "loss": 0.1279, "step": 54480 }, { "epoch": 1.27, "learning_rate": 2.0590281852103916e-05, "loss": 0.0619, "step": 54485 }, { "epoch": 1.27, "learning_rate": 2.058940179765788e-05, "loss": 0.259, "step": 54490 }, { "epoch": 1.27, "learning_rate": 2.058852174321185e-05, "loss": 0.0731, "step": 54495 }, { "epoch": 1.27, "learning_rate": 2.058764168876581e-05, "loss": 0.1691, "step": 54500 }, { "epoch": 1.27, "learning_rate": 2.0586761634319776e-05, "loss": 0.2113, "step": 54505 }, { "epoch": 1.27, "learning_rate": 2.0585881579873745e-05, "loss": 0.3882, "step": 54510 }, { "epoch": 1.27, "learning_rate": 2.0585001525427707e-05, "loss": 0.2628, "step": 54515 }, { "epoch": 1.27, "learning_rate": 2.058412147098167e-05, "loss": 0.0747, "step": 54520 }, { "epoch": 1.27, "learning_rate": 2.0583241416535637e-05, "loss": 0.0737, "step": 54525 }, { "epoch": 1.27, "learning_rate": 2.0582361362089605e-05, "loss": 0.0615, "step": 54530 }, { "epoch": 1.27, "learning_rate": 2.0581481307643567e-05, "loss": 0.1388, "step": 54535 }, { "epoch": 1.27, "learning_rate": 2.058060125319753e-05, "loss": 0.2165, "step": 54540 }, { "epoch": 1.27, "learning_rate": 2.0579721198751497e-05, "loss": 0.0842, "step": 54545 }, { "epoch": 1.27, "learning_rate": 2.057884114430546e-05, "loss": 0.1471, "step": 54550 }, { "epoch": 1.27, "learning_rate": 2.0577961089859427e-05, "loss": 0.2408, "step": 54555 }, { "epoch": 1.27, "learning_rate": 2.0577081035413392e-05, "loss": 0.3147, "step": 54560 }, { "epoch": 1.27, "learning_rate": 2.0576200980967357e-05, "loss": 0.2997, "step": 54565 }, { "epoch": 1.27, "learning_rate": 2.0575320926521322e-05, "loss": 0.0184, "step": 54570 }, { "epoch": 1.27, "learning_rate": 2.0574440872075287e-05, "loss": 0.0727, "step": 54575 }, { "epoch": 1.27, "learning_rate": 2.0573560817629252e-05, "loss": 0.1438, "step": 54580 }, { "epoch": 1.27, "learning_rate": 2.0572680763183213e-05, "loss": 0.1127, "step": 54585 }, { "epoch": 1.27, "learning_rate": 2.0571800708737182e-05, "loss": 0.1889, "step": 54590 }, { "epoch": 1.27, "learning_rate": 2.0570920654291147e-05, "loss": 0.1586, "step": 54595 }, { "epoch": 1.27, "learning_rate": 2.0570040599845112e-05, "loss": 0.1059, "step": 54600 }, { "epoch": 1.27, "learning_rate": 2.0569160545399073e-05, "loss": 0.2822, "step": 54605 }, { "epoch": 1.27, "learning_rate": 2.0568280490953042e-05, "loss": 0.1569, "step": 54610 }, { "epoch": 1.27, "learning_rate": 2.0567400436507007e-05, "loss": 0.2925, "step": 54615 }, { "epoch": 1.27, "learning_rate": 2.056652038206097e-05, "loss": 0.0362, "step": 54620 }, { "epoch": 1.27, "learning_rate": 2.0565640327614934e-05, "loss": 0.1054, "step": 54625 }, { "epoch": 1.27, "learning_rate": 2.0564760273168902e-05, "loss": 0.0848, "step": 54630 }, { "epoch": 1.27, "learning_rate": 2.0563880218722867e-05, "loss": 0.0918, "step": 54635 }, { "epoch": 1.27, "learning_rate": 2.056300016427683e-05, "loss": 0.196, "step": 54640 }, { "epoch": 1.27, "learning_rate": 2.0562120109830797e-05, "loss": 0.0727, "step": 54645 }, { "epoch": 1.28, "learning_rate": 2.0561240055384762e-05, "loss": 0.1532, "step": 54650 }, { "epoch": 1.28, "learning_rate": 2.0560360000938724e-05, "loss": 0.3437, "step": 54655 }, { "epoch": 1.28, "learning_rate": 2.055947994649269e-05, "loss": 0.3787, "step": 54660 }, { "epoch": 1.28, "learning_rate": 2.0558599892046657e-05, "loss": 0.3197, "step": 54665 }, { "epoch": 1.28, "learning_rate": 2.0557719837600622e-05, "loss": 0.0215, "step": 54670 }, { "epoch": 1.28, "learning_rate": 2.0556839783154584e-05, "loss": 0.0676, "step": 54675 }, { "epoch": 1.28, "learning_rate": 2.055595972870855e-05, "loss": 0.0363, "step": 54680 }, { "epoch": 1.28, "learning_rate": 2.0555079674262517e-05, "loss": 0.1061, "step": 54685 }, { "epoch": 1.28, "learning_rate": 2.055419961981648e-05, "loss": 0.1292, "step": 54690 }, { "epoch": 1.28, "learning_rate": 2.0553319565370444e-05, "loss": 0.2558, "step": 54695 }, { "epoch": 1.28, "learning_rate": 2.055243951092441e-05, "loss": 0.1499, "step": 54700 }, { "epoch": 1.28, "learning_rate": 2.0551559456478377e-05, "loss": 0.2142, "step": 54705 }, { "epoch": 1.28, "learning_rate": 2.055067940203234e-05, "loss": 0.559, "step": 54710 }, { "epoch": 1.28, "learning_rate": 2.0549799347586304e-05, "loss": 0.2642, "step": 54715 }, { "epoch": 1.28, "learning_rate": 2.054891929314027e-05, "loss": 0.0832, "step": 54720 }, { "epoch": 1.28, "learning_rate": 2.0548039238694234e-05, "loss": 0.0574, "step": 54725 }, { "epoch": 1.28, "learning_rate": 2.05471591842482e-05, "loss": 0.0939, "step": 54730 }, { "epoch": 1.28, "learning_rate": 2.0546279129802164e-05, "loss": 0.036, "step": 54735 }, { "epoch": 1.28, "learning_rate": 2.054539907535613e-05, "loss": 0.1219, "step": 54740 }, { "epoch": 1.28, "learning_rate": 2.0544519020910094e-05, "loss": 0.0791, "step": 54745 }, { "epoch": 1.28, "learning_rate": 2.054363896646406e-05, "loss": 0.3338, "step": 54750 }, { "epoch": 1.28, "learning_rate": 2.0542758912018024e-05, "loss": 0.196, "step": 54755 }, { "epoch": 1.28, "learning_rate": 2.0541878857571986e-05, "loss": 0.3946, "step": 54760 }, { "epoch": 1.28, "learning_rate": 2.0540998803125954e-05, "loss": 0.1501, "step": 54765 }, { "epoch": 1.28, "learning_rate": 2.054011874867992e-05, "loss": 0.0447, "step": 54770 }, { "epoch": 1.28, "learning_rate": 2.0539238694233884e-05, "loss": 0.0847, "step": 54775 }, { "epoch": 1.28, "learning_rate": 2.053835863978785e-05, "loss": 0.0935, "step": 54780 }, { "epoch": 1.28, "learning_rate": 2.0537478585341814e-05, "loss": 0.1685, "step": 54785 }, { "epoch": 1.28, "learning_rate": 2.053659853089578e-05, "loss": 0.1209, "step": 54790 }, { "epoch": 1.28, "learning_rate": 2.0535718476449744e-05, "loss": 0.1284, "step": 54795 }, { "epoch": 1.28, "learning_rate": 2.053483842200371e-05, "loss": 0.1895, "step": 54800 }, { "epoch": 1.28, "learning_rate": 2.0533958367557674e-05, "loss": 0.1944, "step": 54805 }, { "epoch": 1.28, "learning_rate": 2.053307831311164e-05, "loss": 0.3854, "step": 54810 }, { "epoch": 1.28, "learning_rate": 2.05321982586656e-05, "loss": 0.2499, "step": 54815 }, { "epoch": 1.28, "learning_rate": 2.053131820421957e-05, "loss": 0.0444, "step": 54820 }, { "epoch": 1.28, "learning_rate": 2.0530438149773534e-05, "loss": 0.039, "step": 54825 }, { "epoch": 1.28, "learning_rate": 2.05295580953275e-05, "loss": 0.0347, "step": 54830 }, { "epoch": 1.28, "learning_rate": 2.052867804088146e-05, "loss": 0.078, "step": 54835 }, { "epoch": 1.28, "learning_rate": 2.052779798643543e-05, "loss": 0.1114, "step": 54840 }, { "epoch": 1.28, "learning_rate": 2.0526917931989394e-05, "loss": 0.1456, "step": 54845 }, { "epoch": 1.28, "learning_rate": 2.0526037877543356e-05, "loss": 0.1613, "step": 54850 }, { "epoch": 1.28, "learning_rate": 2.052515782309732e-05, "loss": 0.2842, "step": 54855 }, { "epoch": 1.28, "learning_rate": 2.052427776865129e-05, "loss": 0.6552, "step": 54860 }, { "epoch": 1.28, "learning_rate": 2.0523397714205255e-05, "loss": 0.2095, "step": 54865 }, { "epoch": 1.28, "learning_rate": 2.0522517659759216e-05, "loss": 0.0422, "step": 54870 }, { "epoch": 1.28, "learning_rate": 2.052163760531318e-05, "loss": 0.1079, "step": 54875 }, { "epoch": 1.28, "learning_rate": 2.052075755086715e-05, "loss": 0.0622, "step": 54880 }, { "epoch": 1.28, "learning_rate": 2.051987749642111e-05, "loss": 0.0998, "step": 54885 }, { "epoch": 1.28, "learning_rate": 2.0518997441975076e-05, "loss": 0.1291, "step": 54890 }, { "epoch": 1.28, "learning_rate": 2.051811738752904e-05, "loss": 0.1459, "step": 54895 }, { "epoch": 1.28, "learning_rate": 2.051723733308301e-05, "loss": 0.2262, "step": 54900 }, { "epoch": 1.28, "learning_rate": 2.051635727863697e-05, "loss": 0.2417, "step": 54905 }, { "epoch": 1.28, "learning_rate": 2.0515477224190936e-05, "loss": 0.351, "step": 54910 }, { "epoch": 1.28, "learning_rate": 2.05145971697449e-05, "loss": 0.2998, "step": 54915 }, { "epoch": 1.28, "learning_rate": 2.0513717115298866e-05, "loss": 0.0677, "step": 54920 }, { "epoch": 1.28, "learning_rate": 2.051283706085283e-05, "loss": 0.0844, "step": 54925 }, { "epoch": 1.28, "learning_rate": 2.0511957006406796e-05, "loss": 0.0924, "step": 54930 }, { "epoch": 1.28, "learning_rate": 2.0511076951960765e-05, "loss": 0.0725, "step": 54935 }, { "epoch": 1.28, "learning_rate": 2.0510196897514726e-05, "loss": 0.1217, "step": 54940 }, { "epoch": 1.28, "learning_rate": 2.050931684306869e-05, "loss": 0.1702, "step": 54945 }, { "epoch": 1.28, "learning_rate": 2.0508436788622657e-05, "loss": 0.2184, "step": 54950 }, { "epoch": 1.28, "learning_rate": 2.050755673417662e-05, "loss": 0.2545, "step": 54955 }, { "epoch": 1.28, "learning_rate": 2.0506676679730587e-05, "loss": 0.511, "step": 54960 }, { "epoch": 1.28, "learning_rate": 2.050579662528455e-05, "loss": 0.2834, "step": 54965 }, { "epoch": 1.28, "learning_rate": 2.0504916570838517e-05, "loss": 0.0239, "step": 54970 }, { "epoch": 1.28, "learning_rate": 2.050403651639248e-05, "loss": 0.0596, "step": 54975 }, { "epoch": 1.28, "learning_rate": 2.0503156461946447e-05, "loss": 0.1281, "step": 54980 }, { "epoch": 1.28, "learning_rate": 2.050227640750041e-05, "loss": 0.1372, "step": 54985 }, { "epoch": 1.28, "learning_rate": 2.0501396353054373e-05, "loss": 0.052, "step": 54990 }, { "epoch": 1.28, "learning_rate": 2.0500516298608342e-05, "loss": 0.1011, "step": 54995 }, { "epoch": 1.28, "learning_rate": 2.0499636244162307e-05, "loss": 0.1401, "step": 55000 }, { "epoch": 1.28, "learning_rate": 2.0498756189716272e-05, "loss": 0.1192, "step": 55005 }, { "epoch": 1.28, "learning_rate": 2.0497876135270233e-05, "loss": 0.227, "step": 55010 }, { "epoch": 1.28, "learning_rate": 2.0496996080824202e-05, "loss": 0.427, "step": 55015 }, { "epoch": 1.28, "learning_rate": 2.0496116026378167e-05, "loss": 0.0932, "step": 55020 }, { "epoch": 1.28, "learning_rate": 2.049523597193213e-05, "loss": 0.0977, "step": 55025 }, { "epoch": 1.28, "learning_rate": 2.0494355917486093e-05, "loss": 0.0461, "step": 55030 }, { "epoch": 1.28, "learning_rate": 2.0493475863040062e-05, "loss": 0.0884, "step": 55035 }, { "epoch": 1.28, "learning_rate": 2.0492595808594027e-05, "loss": 0.0581, "step": 55040 }, { "epoch": 1.28, "learning_rate": 2.049171575414799e-05, "loss": 0.3297, "step": 55045 }, { "epoch": 1.28, "learning_rate": 2.0490835699701957e-05, "loss": 0.2278, "step": 55050 }, { "epoch": 1.28, "learning_rate": 2.0489955645255922e-05, "loss": 0.1209, "step": 55055 }, { "epoch": 1.28, "learning_rate": 2.0489075590809884e-05, "loss": 0.4456, "step": 55060 }, { "epoch": 1.28, "learning_rate": 2.048819553636385e-05, "loss": 0.4128, "step": 55065 }, { "epoch": 1.28, "learning_rate": 2.0487315481917817e-05, "loss": 0.0706, "step": 55070 }, { "epoch": 1.28, "learning_rate": 2.0486435427471782e-05, "loss": 0.0575, "step": 55075 }, { "epoch": 1.29, "learning_rate": 2.0485555373025744e-05, "loss": 0.0691, "step": 55080 }, { "epoch": 1.29, "learning_rate": 2.048467531857971e-05, "loss": 0.1294, "step": 55085 }, { "epoch": 1.29, "learning_rate": 2.0483795264133677e-05, "loss": 0.0807, "step": 55090 }, { "epoch": 1.29, "learning_rate": 2.048291520968764e-05, "loss": 0.1813, "step": 55095 }, { "epoch": 1.29, "learning_rate": 2.0482035155241604e-05, "loss": 0.2416, "step": 55100 }, { "epoch": 1.29, "learning_rate": 2.048115510079557e-05, "loss": 0.3814, "step": 55105 }, { "epoch": 1.29, "learning_rate": 2.0480275046349537e-05, "loss": 0.2313, "step": 55110 }, { "epoch": 1.29, "learning_rate": 2.04793949919035e-05, "loss": 0.2035, "step": 55115 }, { "epoch": 1.29, "learning_rate": 2.0478514937457464e-05, "loss": 0.0316, "step": 55120 }, { "epoch": 1.29, "learning_rate": 2.047763488301143e-05, "loss": 0.0242, "step": 55125 }, { "epoch": 1.29, "learning_rate": 2.0476754828565397e-05, "loss": 0.1839, "step": 55130 }, { "epoch": 1.29, "learning_rate": 2.047587477411936e-05, "loss": 0.1148, "step": 55135 }, { "epoch": 1.29, "learning_rate": 2.0474994719673324e-05, "loss": 0.1437, "step": 55140 }, { "epoch": 1.29, "learning_rate": 2.047411466522729e-05, "loss": 0.1563, "step": 55145 }, { "epoch": 1.29, "learning_rate": 2.0473234610781254e-05, "loss": 0.2008, "step": 55150 }, { "epoch": 1.29, "learning_rate": 2.047235455633522e-05, "loss": 0.2862, "step": 55155 }, { "epoch": 1.29, "learning_rate": 2.0471474501889184e-05, "loss": 0.3232, "step": 55160 }, { "epoch": 1.29, "learning_rate": 2.047059444744315e-05, "loss": 0.2432, "step": 55165 }, { "epoch": 1.29, "learning_rate": 2.0469714392997114e-05, "loss": 0.1018, "step": 55170 }, { "epoch": 1.29, "learning_rate": 2.046883433855108e-05, "loss": 0.0521, "step": 55175 }, { "epoch": 1.29, "learning_rate": 2.0467954284105044e-05, "loss": 0.0434, "step": 55180 }, { "epoch": 1.29, "learning_rate": 2.046707422965901e-05, "loss": 0.0757, "step": 55185 }, { "epoch": 1.29, "learning_rate": 2.0466194175212974e-05, "loss": 0.0711, "step": 55190 }, { "epoch": 1.29, "learning_rate": 2.046531412076694e-05, "loss": 0.1181, "step": 55195 }, { "epoch": 1.29, "learning_rate": 2.0464434066320904e-05, "loss": 0.1785, "step": 55200 }, { "epoch": 1.29, "learning_rate": 2.046355401187487e-05, "loss": 0.1823, "step": 55205 }, { "epoch": 1.29, "learning_rate": 2.0462673957428834e-05, "loss": 0.2864, "step": 55210 }, { "epoch": 1.29, "learning_rate": 2.04617939029828e-05, "loss": 0.1339, "step": 55215 }, { "epoch": 1.29, "learning_rate": 2.046091384853676e-05, "loss": 0.0386, "step": 55220 }, { "epoch": 1.29, "learning_rate": 2.046003379409073e-05, "loss": 0.0306, "step": 55225 }, { "epoch": 1.29, "learning_rate": 2.0459153739644694e-05, "loss": 0.0599, "step": 55230 }, { "epoch": 1.29, "learning_rate": 2.045827368519866e-05, "loss": 0.0813, "step": 55235 }, { "epoch": 1.29, "learning_rate": 2.045739363075262e-05, "loss": 0.1113, "step": 55240 }, { "epoch": 1.29, "learning_rate": 2.045651357630659e-05, "loss": 0.1148, "step": 55245 }, { "epoch": 1.29, "learning_rate": 2.0455633521860554e-05, "loss": 0.1895, "step": 55250 }, { "epoch": 1.29, "learning_rate": 2.0454753467414516e-05, "loss": 0.1494, "step": 55255 }, { "epoch": 1.29, "learning_rate": 2.045387341296848e-05, "loss": 0.1876, "step": 55260 }, { "epoch": 1.29, "learning_rate": 2.045299335852245e-05, "loss": 0.2813, "step": 55265 }, { "epoch": 1.29, "learning_rate": 2.0452113304076414e-05, "loss": 0.0334, "step": 55270 }, { "epoch": 1.29, "learning_rate": 2.0451233249630376e-05, "loss": 0.0179, "step": 55275 }, { "epoch": 1.29, "learning_rate": 2.045035319518434e-05, "loss": 0.064, "step": 55280 }, { "epoch": 1.29, "learning_rate": 2.044947314073831e-05, "loss": 0.0691, "step": 55285 }, { "epoch": 1.29, "learning_rate": 2.044859308629227e-05, "loss": 0.1185, "step": 55290 }, { "epoch": 1.29, "learning_rate": 2.0447713031846236e-05, "loss": 0.1116, "step": 55295 }, { "epoch": 1.29, "learning_rate": 2.04468329774002e-05, "loss": 0.2061, "step": 55300 }, { "epoch": 1.29, "learning_rate": 2.044595292295417e-05, "loss": 0.269, "step": 55305 }, { "epoch": 1.29, "learning_rate": 2.044507286850813e-05, "loss": 0.2965, "step": 55310 }, { "epoch": 1.29, "learning_rate": 2.0444192814062096e-05, "loss": 0.2952, "step": 55315 }, { "epoch": 1.29, "learning_rate": 2.044331275961606e-05, "loss": 0.1393, "step": 55320 }, { "epoch": 1.29, "learning_rate": 2.0442432705170026e-05, "loss": 0.0549, "step": 55325 }, { "epoch": 1.29, "learning_rate": 2.044155265072399e-05, "loss": 0.0397, "step": 55330 }, { "epoch": 1.29, "learning_rate": 2.0440672596277956e-05, "loss": 0.1343, "step": 55335 }, { "epoch": 1.29, "learning_rate": 2.0439792541831925e-05, "loss": 0.1401, "step": 55340 }, { "epoch": 1.29, "learning_rate": 2.0438912487385886e-05, "loss": 0.0697, "step": 55345 }, { "epoch": 1.29, "learning_rate": 2.043803243293985e-05, "loss": 0.3873, "step": 55350 }, { "epoch": 1.29, "learning_rate": 2.0437152378493816e-05, "loss": 0.1038, "step": 55355 }, { "epoch": 1.29, "learning_rate": 2.043627232404778e-05, "loss": 0.2708, "step": 55360 }, { "epoch": 1.29, "learning_rate": 2.0435392269601746e-05, "loss": 0.2189, "step": 55365 }, { "epoch": 1.29, "learning_rate": 2.043451221515571e-05, "loss": 0.0701, "step": 55370 }, { "epoch": 1.29, "learning_rate": 2.0433632160709676e-05, "loss": 0.0121, "step": 55375 }, { "epoch": 1.29, "learning_rate": 2.043275210626364e-05, "loss": 0.1121, "step": 55380 }, { "epoch": 1.29, "learning_rate": 2.0431872051817607e-05, "loss": 0.0416, "step": 55385 }, { "epoch": 1.29, "learning_rate": 2.043099199737157e-05, "loss": 0.2254, "step": 55390 }, { "epoch": 1.29, "learning_rate": 2.0430111942925533e-05, "loss": 0.1172, "step": 55395 }, { "epoch": 1.29, "learning_rate": 2.04292318884795e-05, "loss": 0.177, "step": 55400 }, { "epoch": 1.29, "learning_rate": 2.0428351834033467e-05, "loss": 0.126, "step": 55405 }, { "epoch": 1.29, "learning_rate": 2.042747177958743e-05, "loss": 0.4501, "step": 55410 }, { "epoch": 1.29, "learning_rate": 2.0426591725141393e-05, "loss": 0.36, "step": 55415 }, { "epoch": 1.29, "learning_rate": 2.042571167069536e-05, "loss": 0.0021, "step": 55420 }, { "epoch": 1.29, "learning_rate": 2.0424831616249327e-05, "loss": 0.1543, "step": 55425 }, { "epoch": 1.29, "learning_rate": 2.0423951561803292e-05, "loss": 0.0839, "step": 55430 }, { "epoch": 1.29, "learning_rate": 2.0423071507357253e-05, "loss": 0.2356, "step": 55435 }, { "epoch": 1.29, "learning_rate": 2.0422191452911222e-05, "loss": 0.0968, "step": 55440 }, { "epoch": 1.29, "learning_rate": 2.0421311398465187e-05, "loss": 0.1507, "step": 55445 }, { "epoch": 1.29, "learning_rate": 2.042043134401915e-05, "loss": 0.1471, "step": 55450 }, { "epoch": 1.29, "learning_rate": 2.0419551289573113e-05, "loss": 0.2118, "step": 55455 }, { "epoch": 1.29, "learning_rate": 2.0418671235127082e-05, "loss": 0.4286, "step": 55460 }, { "epoch": 1.29, "learning_rate": 2.0417791180681047e-05, "loss": 0.2472, "step": 55465 }, { "epoch": 1.29, "learning_rate": 2.041691112623501e-05, "loss": 0.061, "step": 55470 }, { "epoch": 1.29, "learning_rate": 2.0416031071788977e-05, "loss": 0.0192, "step": 55475 }, { "epoch": 1.29, "learning_rate": 2.0415151017342942e-05, "loss": 0.0542, "step": 55480 }, { "epoch": 1.29, "learning_rate": 2.0414270962896904e-05, "loss": 0.1258, "step": 55485 }, { "epoch": 1.29, "learning_rate": 2.041339090845087e-05, "loss": 0.1119, "step": 55490 }, { "epoch": 1.29, "learning_rate": 2.0412510854004837e-05, "loss": 0.1317, "step": 55495 }, { "epoch": 1.29, "learning_rate": 2.0411630799558802e-05, "loss": 0.1612, "step": 55500 }, { "epoch": 1.29, "learning_rate": 2.0410750745112764e-05, "loss": 0.1755, "step": 55505 }, { "epoch": 1.3, "learning_rate": 2.040987069066673e-05, "loss": 0.2644, "step": 55510 }, { "epoch": 1.3, "learning_rate": 2.0408990636220697e-05, "loss": 0.1848, "step": 55515 }, { "epoch": 1.3, "learning_rate": 2.040811058177466e-05, "loss": 0.0693, "step": 55520 }, { "epoch": 1.3, "learning_rate": 2.0407230527328624e-05, "loss": 0.0912, "step": 55525 }, { "epoch": 1.3, "learning_rate": 2.040635047288259e-05, "loss": 0.0254, "step": 55530 }, { "epoch": 1.3, "learning_rate": 2.0405470418436557e-05, "loss": 0.2189, "step": 55535 }, { "epoch": 1.3, "learning_rate": 2.040459036399052e-05, "loss": 0.1135, "step": 55540 }, { "epoch": 1.3, "learning_rate": 2.0403710309544484e-05, "loss": 0.1106, "step": 55545 }, { "epoch": 1.3, "learning_rate": 2.040283025509845e-05, "loss": 0.1773, "step": 55550 }, { "epoch": 1.3, "learning_rate": 2.0401950200652414e-05, "loss": 0.1264, "step": 55555 }, { "epoch": 1.3, "learning_rate": 2.040107014620638e-05, "loss": 0.1944, "step": 55560 }, { "epoch": 1.3, "learning_rate": 2.0400190091760344e-05, "loss": 0.2766, "step": 55565 }, { "epoch": 1.3, "learning_rate": 2.039931003731431e-05, "loss": 0.1449, "step": 55570 }, { "epoch": 1.3, "learning_rate": 2.0398429982868274e-05, "loss": 0.0221, "step": 55575 }, { "epoch": 1.3, "learning_rate": 2.039754992842224e-05, "loss": 0.106, "step": 55580 }, { "epoch": 1.3, "learning_rate": 2.0396669873976204e-05, "loss": 0.2708, "step": 55585 }, { "epoch": 1.3, "learning_rate": 2.039578981953017e-05, "loss": 0.0529, "step": 55590 }, { "epoch": 1.3, "learning_rate": 2.0394909765084134e-05, "loss": 0.0806, "step": 55595 }, { "epoch": 1.3, "learning_rate": 2.03940297106381e-05, "loss": 0.0816, "step": 55600 }, { "epoch": 1.3, "learning_rate": 2.0393149656192064e-05, "loss": 0.0977, "step": 55605 }, { "epoch": 1.3, "learning_rate": 2.039226960174603e-05, "loss": 0.4497, "step": 55610 }, { "epoch": 1.3, "learning_rate": 2.0391389547299994e-05, "loss": 0.298, "step": 55615 }, { "epoch": 1.3, "learning_rate": 2.039050949285396e-05, "loss": 0.1218, "step": 55620 }, { "epoch": 1.3, "learning_rate": 2.038962943840792e-05, "loss": 0.1235, "step": 55625 }, { "epoch": 1.3, "learning_rate": 2.038874938396189e-05, "loss": 0.2161, "step": 55630 }, { "epoch": 1.3, "learning_rate": 2.0387869329515854e-05, "loss": 0.0881, "step": 55635 }, { "epoch": 1.3, "learning_rate": 2.038698927506982e-05, "loss": 0.0339, "step": 55640 }, { "epoch": 1.3, "learning_rate": 2.038610922062378e-05, "loss": 0.2186, "step": 55645 }, { "epoch": 1.3, "learning_rate": 2.038522916617775e-05, "loss": 0.1381, "step": 55650 }, { "epoch": 1.3, "learning_rate": 2.0384349111731714e-05, "loss": 0.2231, "step": 55655 }, { "epoch": 1.3, "learning_rate": 2.0383469057285676e-05, "loss": 0.3993, "step": 55660 }, { "epoch": 1.3, "learning_rate": 2.038258900283964e-05, "loss": 0.427, "step": 55665 }, { "epoch": 1.3, "learning_rate": 2.038170894839361e-05, "loss": 0.11, "step": 55670 }, { "epoch": 1.3, "learning_rate": 2.0380828893947574e-05, "loss": 0.0315, "step": 55675 }, { "epoch": 1.3, "learning_rate": 2.0379948839501536e-05, "loss": 0.074, "step": 55680 }, { "epoch": 1.3, "learning_rate": 2.03790687850555e-05, "loss": 0.0925, "step": 55685 }, { "epoch": 1.3, "learning_rate": 2.037818873060947e-05, "loss": 0.1476, "step": 55690 }, { "epoch": 1.3, "learning_rate": 2.037730867616343e-05, "loss": 0.0866, "step": 55695 }, { "epoch": 1.3, "learning_rate": 2.0376428621717396e-05, "loss": 0.2267, "step": 55700 }, { "epoch": 1.3, "learning_rate": 2.037554856727136e-05, "loss": 0.2179, "step": 55705 }, { "epoch": 1.3, "learning_rate": 2.037466851282533e-05, "loss": 0.3841, "step": 55710 }, { "epoch": 1.3, "learning_rate": 2.037378845837929e-05, "loss": 0.4517, "step": 55715 }, { "epoch": 1.3, "learning_rate": 2.0372908403933256e-05, "loss": 0.0647, "step": 55720 }, { "epoch": 1.3, "learning_rate": 2.037202834948722e-05, "loss": 0.1155, "step": 55725 }, { "epoch": 1.3, "learning_rate": 2.037114829504119e-05, "loss": 0.1551, "step": 55730 }, { "epoch": 1.3, "learning_rate": 2.037026824059515e-05, "loss": 0.0766, "step": 55735 }, { "epoch": 1.3, "learning_rate": 2.0369388186149116e-05, "loss": 0.1197, "step": 55740 }, { "epoch": 1.3, "learning_rate": 2.0368508131703085e-05, "loss": 0.1993, "step": 55745 }, { "epoch": 1.3, "learning_rate": 2.0367628077257046e-05, "loss": 0.0923, "step": 55750 }, { "epoch": 1.3, "learning_rate": 2.036674802281101e-05, "loss": 0.2786, "step": 55755 }, { "epoch": 1.3, "learning_rate": 2.0365867968364976e-05, "loss": 0.2086, "step": 55760 }, { "epoch": 1.3, "learning_rate": 2.0364987913918945e-05, "loss": 0.1966, "step": 55765 }, { "epoch": 1.3, "learning_rate": 2.0364107859472906e-05, "loss": 0.0214, "step": 55770 }, { "epoch": 1.3, "learning_rate": 2.036322780502687e-05, "loss": 0.0649, "step": 55775 }, { "epoch": 1.3, "learning_rate": 2.0362347750580836e-05, "loss": 0.0917, "step": 55780 }, { "epoch": 1.3, "learning_rate": 2.03614676961348e-05, "loss": 0.054, "step": 55785 }, { "epoch": 1.3, "learning_rate": 2.0360587641688766e-05, "loss": 0.1254, "step": 55790 }, { "epoch": 1.3, "learning_rate": 2.035970758724273e-05, "loss": 0.1007, "step": 55795 }, { "epoch": 1.3, "learning_rate": 2.0358827532796696e-05, "loss": 0.1148, "step": 55800 }, { "epoch": 1.3, "learning_rate": 2.035794747835066e-05, "loss": 0.2815, "step": 55805 }, { "epoch": 1.3, "learning_rate": 2.0357067423904627e-05, "loss": 0.3731, "step": 55810 }, { "epoch": 1.3, "learning_rate": 2.035618736945859e-05, "loss": 0.1575, "step": 55815 }, { "epoch": 1.3, "learning_rate": 2.0355307315012553e-05, "loss": 0.0663, "step": 55820 }, { "epoch": 1.3, "learning_rate": 2.035442726056652e-05, "loss": 0.1021, "step": 55825 }, { "epoch": 1.3, "learning_rate": 2.0353547206120487e-05, "loss": 0.0848, "step": 55830 }, { "epoch": 1.3, "learning_rate": 2.035266715167445e-05, "loss": 0.1605, "step": 55835 }, { "epoch": 1.3, "learning_rate": 2.0351787097228413e-05, "loss": 0.1834, "step": 55840 }, { "epoch": 1.3, "learning_rate": 2.035090704278238e-05, "loss": 0.0879, "step": 55845 }, { "epoch": 1.3, "learning_rate": 2.0350026988336347e-05, "loss": 0.1619, "step": 55850 }, { "epoch": 1.3, "learning_rate": 2.0349146933890308e-05, "loss": 0.3205, "step": 55855 }, { "epoch": 1.3, "learning_rate": 2.0348266879444273e-05, "loss": 0.4103, "step": 55860 }, { "epoch": 1.3, "learning_rate": 2.0347386824998242e-05, "loss": 0.3175, "step": 55865 }, { "epoch": 1.3, "learning_rate": 2.0346506770552207e-05, "loss": 0.0549, "step": 55870 }, { "epoch": 1.3, "learning_rate": 2.034562671610617e-05, "loss": 0.0521, "step": 55875 }, { "epoch": 1.3, "learning_rate": 2.0344746661660137e-05, "loss": 0.0877, "step": 55880 }, { "epoch": 1.3, "learning_rate": 2.0343866607214102e-05, "loss": 0.0796, "step": 55885 }, { "epoch": 1.3, "learning_rate": 2.0342986552768063e-05, "loss": 0.0962, "step": 55890 }, { "epoch": 1.3, "learning_rate": 2.034210649832203e-05, "loss": 0.1825, "step": 55895 }, { "epoch": 1.3, "learning_rate": 2.0341226443875997e-05, "loss": 0.297, "step": 55900 }, { "epoch": 1.3, "learning_rate": 2.0340346389429962e-05, "loss": 0.1506, "step": 55905 }, { "epoch": 1.3, "learning_rate": 2.0339466334983924e-05, "loss": 0.4306, "step": 55910 }, { "epoch": 1.3, "learning_rate": 2.033858628053789e-05, "loss": 0.1063, "step": 55915 }, { "epoch": 1.3, "learning_rate": 2.0337706226091857e-05, "loss": 0.0243, "step": 55920 }, { "epoch": 1.3, "learning_rate": 2.033682617164582e-05, "loss": 0.1397, "step": 55925 }, { "epoch": 1.3, "learning_rate": 2.0335946117199784e-05, "loss": 0.1126, "step": 55930 }, { "epoch": 1.31, "learning_rate": 2.033506606275375e-05, "loss": 0.0872, "step": 55935 }, { "epoch": 1.31, "learning_rate": 2.0334186008307717e-05, "loss": 0.2159, "step": 55940 }, { "epoch": 1.31, "learning_rate": 2.033330595386168e-05, "loss": 0.1565, "step": 55945 }, { "epoch": 1.31, "learning_rate": 2.0332425899415644e-05, "loss": 0.1248, "step": 55950 }, { "epoch": 1.31, "learning_rate": 2.033154584496961e-05, "loss": 0.1735, "step": 55955 }, { "epoch": 1.31, "learning_rate": 2.0330665790523574e-05, "loss": 0.3428, "step": 55960 }, { "epoch": 1.31, "learning_rate": 2.032978573607754e-05, "loss": 0.3396, "step": 55965 }, { "epoch": 1.31, "learning_rate": 2.0328905681631504e-05, "loss": 0.0323, "step": 55970 }, { "epoch": 1.31, "learning_rate": 2.032802562718547e-05, "loss": 0.0589, "step": 55975 }, { "epoch": 1.31, "learning_rate": 2.0327145572739434e-05, "loss": 0.1485, "step": 55980 }, { "epoch": 1.31, "learning_rate": 2.03262655182934e-05, "loss": 0.0895, "step": 55985 }, { "epoch": 1.31, "learning_rate": 2.0325385463847364e-05, "loss": 0.1614, "step": 55990 }, { "epoch": 1.31, "learning_rate": 2.0324505409401325e-05, "loss": 0.143, "step": 55995 }, { "epoch": 1.31, "learning_rate": 2.0323625354955294e-05, "loss": 0.2859, "step": 56000 }, { "epoch": 1.31, "learning_rate": 2.032274530050926e-05, "loss": 0.225, "step": 56005 }, { "epoch": 1.31, "learning_rate": 2.0321865246063224e-05, "loss": 0.472, "step": 56010 }, { "epoch": 1.31, "learning_rate": 2.032098519161719e-05, "loss": 0.2167, "step": 56015 }, { "epoch": 1.31, "learning_rate": 2.0320105137171154e-05, "loss": 0.0645, "step": 56020 }, { "epoch": 1.31, "learning_rate": 2.031922508272512e-05, "loss": 0.12, "step": 56025 }, { "epoch": 1.31, "learning_rate": 2.0318345028279084e-05, "loss": 0.0909, "step": 56030 }, { "epoch": 1.31, "learning_rate": 2.031746497383305e-05, "loss": 0.0526, "step": 56035 }, { "epoch": 1.31, "learning_rate": 2.0316584919387014e-05, "loss": 0.2084, "step": 56040 }, { "epoch": 1.31, "learning_rate": 2.031570486494098e-05, "loss": 0.1332, "step": 56045 }, { "epoch": 1.31, "learning_rate": 2.031482481049494e-05, "loss": 0.1915, "step": 56050 }, { "epoch": 1.31, "learning_rate": 2.031394475604891e-05, "loss": 0.3316, "step": 56055 }, { "epoch": 1.31, "learning_rate": 2.0313064701602874e-05, "loss": 0.2937, "step": 56060 }, { "epoch": 1.31, "learning_rate": 2.031218464715684e-05, "loss": 0.2529, "step": 56065 }, { "epoch": 1.31, "learning_rate": 2.03113045927108e-05, "loss": 0.0924, "step": 56070 }, { "epoch": 1.31, "learning_rate": 2.031042453826477e-05, "loss": 0.0083, "step": 56075 }, { "epoch": 1.31, "learning_rate": 2.0309544483818734e-05, "loss": 0.1081, "step": 56080 }, { "epoch": 1.31, "learning_rate": 2.0308664429372696e-05, "loss": 0.041, "step": 56085 }, { "epoch": 1.31, "learning_rate": 2.030778437492666e-05, "loss": 0.1386, "step": 56090 }, { "epoch": 1.31, "learning_rate": 2.030690432048063e-05, "loss": 0.1297, "step": 56095 }, { "epoch": 1.31, "learning_rate": 2.03062002769238e-05, "loss": 0.3209, "step": 56100 }, { "epoch": 1.31, "learning_rate": 2.0305320222477762e-05, "loss": 0.193, "step": 56105 }, { "epoch": 1.31, "learning_rate": 2.030444016803173e-05, "loss": 0.5803, "step": 56110 }, { "epoch": 1.31, "learning_rate": 2.0303560113585696e-05, "loss": 0.3002, "step": 56115 }, { "epoch": 1.31, "learning_rate": 2.0302680059139657e-05, "loss": 0.0696, "step": 56120 }, { "epoch": 1.31, "learning_rate": 2.0301800004693622e-05, "loss": 0.0287, "step": 56125 }, { "epoch": 1.31, "learning_rate": 2.030091995024759e-05, "loss": 0.101, "step": 56130 }, { "epoch": 1.31, "learning_rate": 2.0300039895801556e-05, "loss": 0.1196, "step": 56135 }, { "epoch": 1.31, "learning_rate": 2.0299159841355517e-05, "loss": 0.0761, "step": 56140 }, { "epoch": 1.31, "learning_rate": 2.0298279786909482e-05, "loss": 0.2483, "step": 56145 }, { "epoch": 1.31, "learning_rate": 2.029739973246345e-05, "loss": 0.1093, "step": 56150 }, { "epoch": 1.31, "learning_rate": 2.0296519678017412e-05, "loss": 0.1752, "step": 56155 }, { "epoch": 1.31, "learning_rate": 2.0295639623571377e-05, "loss": 0.2365, "step": 56160 }, { "epoch": 1.31, "learning_rate": 2.0294759569125342e-05, "loss": 0.1131, "step": 56165 }, { "epoch": 1.31, "learning_rate": 2.029387951467931e-05, "loss": 0.0576, "step": 56170 }, { "epoch": 1.31, "learning_rate": 2.0292999460233273e-05, "loss": 0.0112, "step": 56175 }, { "epoch": 1.31, "learning_rate": 2.0292119405787238e-05, "loss": 0.0984, "step": 56180 }, { "epoch": 1.31, "learning_rate": 2.0291239351341203e-05, "loss": 0.0632, "step": 56185 }, { "epoch": 1.31, "learning_rate": 2.0290359296895168e-05, "loss": 0.1286, "step": 56190 }, { "epoch": 1.31, "learning_rate": 2.0289479242449133e-05, "loss": 0.2044, "step": 56195 }, { "epoch": 1.31, "learning_rate": 2.0288599188003098e-05, "loss": 0.2687, "step": 56200 }, { "epoch": 1.31, "learning_rate": 2.0287719133557066e-05, "loss": 0.2778, "step": 56205 }, { "epoch": 1.31, "learning_rate": 2.0286839079111028e-05, "loss": 0.1753, "step": 56210 }, { "epoch": 1.31, "learning_rate": 2.0285959024664993e-05, "loss": 0.2703, "step": 56215 }, { "epoch": 1.31, "learning_rate": 2.0285078970218958e-05, "loss": 0.0724, "step": 56220 }, { "epoch": 1.31, "learning_rate": 2.0284198915772926e-05, "loss": 0.0722, "step": 56225 }, { "epoch": 1.31, "learning_rate": 2.0283318861326888e-05, "loss": 0.1181, "step": 56230 }, { "epoch": 1.31, "learning_rate": 2.0282438806880853e-05, "loss": 0.189, "step": 56235 }, { "epoch": 1.31, "learning_rate": 2.0281558752434818e-05, "loss": 0.1388, "step": 56240 }, { "epoch": 1.31, "learning_rate": 2.0280678697988783e-05, "loss": 0.1431, "step": 56245 }, { "epoch": 1.31, "learning_rate": 2.0279798643542748e-05, "loss": 0.1105, "step": 56250 }, { "epoch": 1.31, "learning_rate": 2.0278918589096713e-05, "loss": 0.1819, "step": 56255 }, { "epoch": 1.31, "learning_rate": 2.0278038534650678e-05, "loss": 0.2232, "step": 56260 }, { "epoch": 1.31, "learning_rate": 2.0277158480204643e-05, "loss": 0.3071, "step": 56265 }, { "epoch": 1.31, "learning_rate": 2.0276278425758608e-05, "loss": 0.0693, "step": 56270 }, { "epoch": 1.31, "learning_rate": 2.0275398371312573e-05, "loss": 0.0045, "step": 56275 }, { "epoch": 1.31, "learning_rate": 2.0274518316866535e-05, "loss": 0.087, "step": 56280 }, { "epoch": 1.31, "learning_rate": 2.0273638262420503e-05, "loss": 0.0453, "step": 56285 }, { "epoch": 1.31, "learning_rate": 2.0272758207974468e-05, "loss": 0.1466, "step": 56290 }, { "epoch": 1.31, "learning_rate": 2.0271878153528433e-05, "loss": 0.1797, "step": 56295 }, { "epoch": 1.31, "learning_rate": 2.0270998099082395e-05, "loss": 0.1996, "step": 56300 }, { "epoch": 1.31, "learning_rate": 2.0270118044636363e-05, "loss": 0.357, "step": 56305 }, { "epoch": 1.31, "learning_rate": 2.0269237990190328e-05, "loss": 0.413, "step": 56310 }, { "epoch": 1.31, "learning_rate": 2.026835793574429e-05, "loss": 0.3236, "step": 56315 }, { "epoch": 1.31, "learning_rate": 2.0267477881298255e-05, "loss": 0.0356, "step": 56320 }, { "epoch": 1.31, "learning_rate": 2.0266597826852223e-05, "loss": 0.1015, "step": 56325 }, { "epoch": 1.31, "learning_rate": 2.0265717772406188e-05, "loss": 0.0177, "step": 56330 }, { "epoch": 1.31, "learning_rate": 2.026483771796015e-05, "loss": 0.1522, "step": 56335 }, { "epoch": 1.31, "learning_rate": 2.0263957663514118e-05, "loss": 0.0915, "step": 56340 }, { "epoch": 1.31, "learning_rate": 2.0263077609068083e-05, "loss": 0.095, "step": 56345 }, { "epoch": 1.31, "learning_rate": 2.0262197554622045e-05, "loss": 0.0905, "step": 56350 }, { "epoch": 1.31, "learning_rate": 2.026131750017601e-05, "loss": 0.433, "step": 56355 }, { "epoch": 1.31, "learning_rate": 2.0260437445729978e-05, "loss": 0.4203, "step": 56360 }, { "epoch": 1.32, "learning_rate": 2.0259557391283943e-05, "loss": 0.3773, "step": 56365 }, { "epoch": 1.32, "learning_rate": 2.0258677336837905e-05, "loss": 0.0891, "step": 56370 }, { "epoch": 1.32, "learning_rate": 2.025779728239187e-05, "loss": 0.0295, "step": 56375 }, { "epoch": 1.32, "learning_rate": 2.025691722794584e-05, "loss": 0.0607, "step": 56380 }, { "epoch": 1.32, "learning_rate": 2.02560371734998e-05, "loss": 0.0918, "step": 56385 }, { "epoch": 1.32, "learning_rate": 2.0255157119053765e-05, "loss": 0.1674, "step": 56390 }, { "epoch": 1.32, "learning_rate": 2.025427706460773e-05, "loss": 0.1216, "step": 56395 }, { "epoch": 1.32, "learning_rate": 2.02533970101617e-05, "loss": 0.1963, "step": 56400 }, { "epoch": 1.32, "learning_rate": 2.025251695571566e-05, "loss": 0.3152, "step": 56405 }, { "epoch": 1.32, "learning_rate": 2.0251636901269625e-05, "loss": 0.4565, "step": 56410 }, { "epoch": 1.32, "learning_rate": 2.025075684682359e-05, "loss": 0.2281, "step": 56415 }, { "epoch": 1.32, "learning_rate": 2.0249876792377555e-05, "loss": 0.0536, "step": 56420 }, { "epoch": 1.32, "learning_rate": 2.024899673793152e-05, "loss": 0.0609, "step": 56425 }, { "epoch": 1.32, "learning_rate": 2.0248116683485485e-05, "loss": 0.0361, "step": 56430 }, { "epoch": 1.32, "learning_rate": 2.024723662903945e-05, "loss": 0.0631, "step": 56435 }, { "epoch": 1.32, "learning_rate": 2.0246356574593415e-05, "loss": 0.0955, "step": 56440 }, { "epoch": 1.32, "learning_rate": 2.024547652014738e-05, "loss": 0.0797, "step": 56445 }, { "epoch": 1.32, "learning_rate": 2.0244596465701345e-05, "loss": 0.1571, "step": 56450 }, { "epoch": 1.32, "learning_rate": 2.0243716411255307e-05, "loss": 0.2471, "step": 56455 }, { "epoch": 1.32, "learning_rate": 2.0242836356809275e-05, "loss": 0.4994, "step": 56460 }, { "epoch": 1.32, "learning_rate": 2.024195630236324e-05, "loss": 0.2667, "step": 56465 }, { "epoch": 1.32, "learning_rate": 2.0241076247917205e-05, "loss": 0.0901, "step": 56470 }, { "epoch": 1.32, "learning_rate": 2.024019619347117e-05, "loss": 0.1123, "step": 56475 }, { "epoch": 1.32, "learning_rate": 2.0239316139025135e-05, "loss": 0.1725, "step": 56480 }, { "epoch": 1.32, "learning_rate": 2.02384360845791e-05, "loss": 0.096, "step": 56485 }, { "epoch": 1.32, "learning_rate": 2.0237556030133062e-05, "loss": 0.0822, "step": 56490 }, { "epoch": 1.32, "learning_rate": 2.023667597568703e-05, "loss": 0.1257, "step": 56495 }, { "epoch": 1.32, "learning_rate": 2.0235795921240995e-05, "loss": 0.1999, "step": 56500 }, { "epoch": 1.32, "learning_rate": 2.023491586679496e-05, "loss": 0.1961, "step": 56505 }, { "epoch": 1.32, "learning_rate": 2.0234035812348922e-05, "loss": 0.3081, "step": 56510 }, { "epoch": 1.32, "learning_rate": 2.023315575790289e-05, "loss": 0.2807, "step": 56515 }, { "epoch": 1.32, "learning_rate": 2.0232275703456856e-05, "loss": 0.0487, "step": 56520 }, { "epoch": 1.32, "learning_rate": 2.023139564901082e-05, "loss": 0.1051, "step": 56525 }, { "epoch": 1.32, "learning_rate": 2.0230515594564782e-05, "loss": 0.0378, "step": 56530 }, { "epoch": 1.32, "learning_rate": 2.022963554011875e-05, "loss": 0.0989, "step": 56535 }, { "epoch": 1.32, "learning_rate": 2.0228755485672716e-05, "loss": 0.1105, "step": 56540 }, { "epoch": 1.32, "learning_rate": 2.0227875431226677e-05, "loss": 0.294, "step": 56545 }, { "epoch": 1.32, "learning_rate": 2.0226995376780642e-05, "loss": 0.2234, "step": 56550 }, { "epoch": 1.32, "learning_rate": 2.022611532233461e-05, "loss": 0.1647, "step": 56555 }, { "epoch": 1.32, "learning_rate": 2.0225235267888576e-05, "loss": 0.3975, "step": 56560 }, { "epoch": 1.32, "learning_rate": 2.0224355213442537e-05, "loss": 0.1871, "step": 56565 }, { "epoch": 1.32, "learning_rate": 2.0223475158996502e-05, "loss": 0.0472, "step": 56570 }, { "epoch": 1.32, "learning_rate": 2.022259510455047e-05, "loss": 0.04, "step": 56575 }, { "epoch": 1.32, "learning_rate": 2.0221715050104432e-05, "loss": 0.0549, "step": 56580 }, { "epoch": 1.32, "learning_rate": 2.0220834995658397e-05, "loss": 0.1148, "step": 56585 }, { "epoch": 1.32, "learning_rate": 2.0219954941212362e-05, "loss": 0.1008, "step": 56590 }, { "epoch": 1.32, "learning_rate": 2.021907488676633e-05, "loss": 0.1527, "step": 56595 }, { "epoch": 1.32, "learning_rate": 2.0218194832320293e-05, "loss": 0.1538, "step": 56600 }, { "epoch": 1.32, "learning_rate": 2.0217314777874258e-05, "loss": 0.1838, "step": 56605 }, { "epoch": 1.32, "learning_rate": 2.0216434723428226e-05, "loss": 0.2433, "step": 56610 }, { "epoch": 1.32, "learning_rate": 2.0215554668982188e-05, "loss": 0.5871, "step": 56615 }, { "epoch": 1.32, "learning_rate": 2.0214674614536153e-05, "loss": 0.1461, "step": 56620 }, { "epoch": 1.32, "learning_rate": 2.0213794560090118e-05, "loss": 0.086, "step": 56625 }, { "epoch": 1.32, "learning_rate": 2.0212914505644086e-05, "loss": 0.1281, "step": 56630 }, { "epoch": 1.32, "learning_rate": 2.0212034451198048e-05, "loss": 0.1197, "step": 56635 }, { "epoch": 1.32, "learning_rate": 2.0211154396752013e-05, "loss": 0.1827, "step": 56640 }, { "epoch": 1.32, "learning_rate": 2.0210274342305978e-05, "loss": 0.1114, "step": 56645 }, { "epoch": 1.32, "learning_rate": 2.0209394287859943e-05, "loss": 0.1382, "step": 56650 }, { "epoch": 1.32, "learning_rate": 2.0208514233413908e-05, "loss": 0.1782, "step": 56655 }, { "epoch": 1.32, "learning_rate": 2.0207634178967873e-05, "loss": 0.3243, "step": 56660 }, { "epoch": 1.32, "learning_rate": 2.0206754124521838e-05, "loss": 0.1971, "step": 56665 }, { "epoch": 1.32, "learning_rate": 2.0205874070075803e-05, "loss": 0.0575, "step": 56670 }, { "epoch": 1.32, "learning_rate": 2.0204994015629768e-05, "loss": 0.0689, "step": 56675 }, { "epoch": 1.32, "learning_rate": 2.0204113961183733e-05, "loss": 0.1272, "step": 56680 }, { "epoch": 1.32, "learning_rate": 2.0203233906737694e-05, "loss": 0.2073, "step": 56685 }, { "epoch": 1.32, "learning_rate": 2.0202353852291663e-05, "loss": 0.2071, "step": 56690 }, { "epoch": 1.32, "learning_rate": 2.0201473797845628e-05, "loss": 0.1795, "step": 56695 }, { "epoch": 1.32, "learning_rate": 2.0200593743399593e-05, "loss": 0.1556, "step": 56700 }, { "epoch": 1.32, "learning_rate": 2.0199713688953555e-05, "loss": 0.149, "step": 56705 }, { "epoch": 1.32, "learning_rate": 2.0198833634507523e-05, "loss": 0.444, "step": 56710 }, { "epoch": 1.32, "learning_rate": 2.0197953580061488e-05, "loss": 0.2686, "step": 56715 }, { "epoch": 1.32, "learning_rate": 2.019707352561545e-05, "loss": 0.0273, "step": 56720 }, { "epoch": 1.32, "learning_rate": 2.0196193471169415e-05, "loss": 0.0242, "step": 56725 }, { "epoch": 1.32, "learning_rate": 2.0195313416723383e-05, "loss": 0.0602, "step": 56730 }, { "epoch": 1.32, "learning_rate": 2.0194433362277348e-05, "loss": 0.1477, "step": 56735 }, { "epoch": 1.32, "learning_rate": 2.019355330783131e-05, "loss": 0.1138, "step": 56740 }, { "epoch": 1.32, "learning_rate": 2.0192673253385278e-05, "loss": 0.1402, "step": 56745 }, { "epoch": 1.32, "learning_rate": 2.0191793198939243e-05, "loss": 0.0691, "step": 56750 }, { "epoch": 1.32, "learning_rate": 2.0190913144493205e-05, "loss": 0.1667, "step": 56755 }, { "epoch": 1.32, "learning_rate": 2.019003309004717e-05, "loss": 0.1794, "step": 56760 }, { "epoch": 1.32, "learning_rate": 2.0189153035601138e-05, "loss": 0.1856, "step": 56765 }, { "epoch": 1.32, "learning_rate": 2.0188272981155103e-05, "loss": 0.0261, "step": 56770 }, { "epoch": 1.32, "learning_rate": 2.0187392926709065e-05, "loss": 0.0919, "step": 56775 }, { "epoch": 1.32, "learning_rate": 2.018651287226303e-05, "loss": 0.098, "step": 56780 }, { "epoch": 1.32, "learning_rate": 2.0185632817816998e-05, "loss": 0.0686, "step": 56785 }, { "epoch": 1.32, "learning_rate": 2.018475276337096e-05, "loss": 0.0975, "step": 56790 }, { "epoch": 1.33, "learning_rate": 2.0183872708924925e-05, "loss": 0.1887, "step": 56795 }, { "epoch": 1.33, "learning_rate": 2.018299265447889e-05, "loss": 0.3085, "step": 56800 }, { "epoch": 1.33, "learning_rate": 2.018211260003286e-05, "loss": 0.2292, "step": 56805 }, { "epoch": 1.33, "learning_rate": 2.018123254558682e-05, "loss": 0.4561, "step": 56810 }, { "epoch": 1.33, "learning_rate": 2.0180352491140785e-05, "loss": 0.2334, "step": 56815 }, { "epoch": 1.33, "learning_rate": 2.017947243669475e-05, "loss": 0.038, "step": 56820 }, { "epoch": 1.33, "learning_rate": 2.017859238224872e-05, "loss": 0.0971, "step": 56825 }, { "epoch": 1.33, "learning_rate": 2.017771232780268e-05, "loss": 0.006, "step": 56830 }, { "epoch": 1.33, "learning_rate": 2.0176832273356645e-05, "loss": 0.1618, "step": 56835 }, { "epoch": 1.33, "learning_rate": 2.017595221891061e-05, "loss": 0.1117, "step": 56840 }, { "epoch": 1.33, "learning_rate": 2.0175072164464575e-05, "loss": 0.12, "step": 56845 }, { "epoch": 1.33, "learning_rate": 2.017419211001854e-05, "loss": 0.2115, "step": 56850 }, { "epoch": 1.33, "learning_rate": 2.0173312055572505e-05, "loss": 0.1052, "step": 56855 }, { "epoch": 1.33, "learning_rate": 2.017243200112647e-05, "loss": 0.545, "step": 56860 }, { "epoch": 1.33, "learning_rate": 2.0171551946680435e-05, "loss": 0.1957, "step": 56865 }, { "epoch": 1.33, "learning_rate": 2.01706718922344e-05, "loss": 0.1115, "step": 56870 }, { "epoch": 1.33, "learning_rate": 2.0169791837788365e-05, "loss": 0.0172, "step": 56875 }, { "epoch": 1.33, "learning_rate": 2.016891178334233e-05, "loss": 0.0744, "step": 56880 }, { "epoch": 1.33, "learning_rate": 2.0168031728896295e-05, "loss": 0.0449, "step": 56885 }, { "epoch": 1.33, "learning_rate": 2.016715167445026e-05, "loss": 0.1217, "step": 56890 }, { "epoch": 1.33, "learning_rate": 2.0166271620004225e-05, "loss": 0.1437, "step": 56895 }, { "epoch": 1.33, "learning_rate": 2.016539156555819e-05, "loss": 0.1668, "step": 56900 }, { "epoch": 1.33, "learning_rate": 2.0164511511112155e-05, "loss": 0.1077, "step": 56905 }, { "epoch": 1.33, "learning_rate": 2.016363145666612e-05, "loss": 0.3368, "step": 56910 }, { "epoch": 1.33, "learning_rate": 2.0162751402220082e-05, "loss": 0.2184, "step": 56915 }, { "epoch": 1.33, "learning_rate": 2.016187134777405e-05, "loss": 0.034, "step": 56920 }, { "epoch": 1.33, "learning_rate": 2.0160991293328015e-05, "loss": 0.128, "step": 56925 }, { "epoch": 1.33, "learning_rate": 2.016011123888198e-05, "loss": 0.162, "step": 56930 }, { "epoch": 1.33, "learning_rate": 2.0159231184435942e-05, "loss": 0.0659, "step": 56935 }, { "epoch": 1.33, "learning_rate": 2.015835112998991e-05, "loss": 0.0306, "step": 56940 }, { "epoch": 1.33, "learning_rate": 2.0157471075543876e-05, "loss": 0.1024, "step": 56945 }, { "epoch": 1.33, "learning_rate": 2.0156591021097837e-05, "loss": 0.147, "step": 56950 }, { "epoch": 1.33, "learning_rate": 2.0155710966651802e-05, "loss": 0.3112, "step": 56955 }, { "epoch": 1.33, "learning_rate": 2.015483091220577e-05, "loss": 0.533, "step": 56960 }, { "epoch": 1.33, "learning_rate": 2.0153950857759736e-05, "loss": 0.3554, "step": 56965 }, { "epoch": 1.33, "learning_rate": 2.0153070803313697e-05, "loss": 0.0212, "step": 56970 }, { "epoch": 1.33, "learning_rate": 2.0152190748867662e-05, "loss": 0.0466, "step": 56975 }, { "epoch": 1.33, "learning_rate": 2.015131069442163e-05, "loss": 0.0739, "step": 56980 }, { "epoch": 1.33, "learning_rate": 2.0150430639975592e-05, "loss": 0.1567, "step": 56985 }, { "epoch": 1.33, "learning_rate": 2.0149550585529557e-05, "loss": 0.1042, "step": 56990 }, { "epoch": 1.33, "learning_rate": 2.0148670531083522e-05, "loss": 0.1075, "step": 56995 }, { "epoch": 1.33, "learning_rate": 2.014779047663749e-05, "loss": 0.115, "step": 57000 }, { "epoch": 1.33, "learning_rate": 2.0146910422191452e-05, "loss": 0.1135, "step": 57005 }, { "epoch": 1.33, "learning_rate": 2.0146030367745417e-05, "loss": 0.1391, "step": 57010 }, { "epoch": 1.33, "learning_rate": 2.0145150313299386e-05, "loss": 0.2706, "step": 57015 }, { "epoch": 1.33, "learning_rate": 2.0144270258853347e-05, "loss": 0.0351, "step": 57020 }, { "epoch": 1.33, "learning_rate": 2.0143390204407312e-05, "loss": 0.0457, "step": 57025 }, { "epoch": 1.33, "learning_rate": 2.0142510149961277e-05, "loss": 0.0567, "step": 57030 }, { "epoch": 1.33, "learning_rate": 2.0141630095515246e-05, "loss": 0.1082, "step": 57035 }, { "epoch": 1.33, "learning_rate": 2.0140750041069208e-05, "loss": 0.2255, "step": 57040 }, { "epoch": 1.33, "learning_rate": 2.0139869986623173e-05, "loss": 0.14, "step": 57045 }, { "epoch": 1.33, "learning_rate": 2.0138989932177138e-05, "loss": 0.0241, "step": 57050 }, { "epoch": 1.33, "learning_rate": 2.0138109877731103e-05, "loss": 0.1219, "step": 57055 }, { "epoch": 1.33, "learning_rate": 2.0137229823285068e-05, "loss": 0.4988, "step": 57060 }, { "epoch": 1.33, "learning_rate": 2.0136349768839033e-05, "loss": 0.4479, "step": 57065 }, { "epoch": 1.33, "learning_rate": 2.0135469714392998e-05, "loss": 0.054, "step": 57070 }, { "epoch": 1.33, "learning_rate": 2.0134589659946963e-05, "loss": 0.1164, "step": 57075 }, { "epoch": 1.33, "learning_rate": 2.0133709605500928e-05, "loss": 0.1209, "step": 57080 }, { "epoch": 1.33, "learning_rate": 2.0132829551054893e-05, "loss": 0.0523, "step": 57085 }, { "epoch": 1.33, "learning_rate": 2.0131949496608854e-05, "loss": 0.0775, "step": 57090 }, { "epoch": 1.33, "learning_rate": 2.0131069442162823e-05, "loss": 0.1789, "step": 57095 }, { "epoch": 1.33, "learning_rate": 2.0130189387716788e-05, "loss": 0.2531, "step": 57100 }, { "epoch": 1.33, "learning_rate": 2.0129309333270753e-05, "loss": 0.3053, "step": 57105 }, { "epoch": 1.33, "learning_rate": 2.0128429278824714e-05, "loss": 0.1689, "step": 57110 }, { "epoch": 1.33, "learning_rate": 2.0127549224378683e-05, "loss": 0.3371, "step": 57115 }, { "epoch": 1.33, "learning_rate": 2.0126669169932648e-05, "loss": 0.0363, "step": 57120 }, { "epoch": 1.33, "learning_rate": 2.0125789115486613e-05, "loss": 0.0275, "step": 57125 }, { "epoch": 1.33, "learning_rate": 2.0124909061040575e-05, "loss": 0.049, "step": 57130 }, { "epoch": 1.33, "learning_rate": 2.0124029006594543e-05, "loss": 0.114, "step": 57135 }, { "epoch": 1.33, "learning_rate": 2.0123148952148508e-05, "loss": 0.0967, "step": 57140 }, { "epoch": 1.33, "learning_rate": 2.012226889770247e-05, "loss": 0.155, "step": 57145 }, { "epoch": 1.33, "learning_rate": 2.0121388843256438e-05, "loss": 0.1675, "step": 57150 }, { "epoch": 1.33, "learning_rate": 2.0120508788810403e-05, "loss": 0.2147, "step": 57155 }, { "epoch": 1.33, "learning_rate": 2.0119628734364368e-05, "loss": 0.3631, "step": 57160 }, { "epoch": 1.33, "learning_rate": 2.011874867991833e-05, "loss": 0.2052, "step": 57165 }, { "epoch": 1.33, "learning_rate": 2.0117868625472298e-05, "loss": 0.0614, "step": 57170 }, { "epoch": 1.33, "learning_rate": 2.0116988571026263e-05, "loss": 0.0936, "step": 57175 }, { "epoch": 1.33, "learning_rate": 2.0116108516580225e-05, "loss": 0.0731, "step": 57180 }, { "epoch": 1.33, "learning_rate": 2.011522846213419e-05, "loss": 0.0795, "step": 57185 }, { "epoch": 1.33, "learning_rate": 2.0114348407688158e-05, "loss": 0.1957, "step": 57190 }, { "epoch": 1.33, "learning_rate": 2.0113468353242123e-05, "loss": 0.1449, "step": 57195 }, { "epoch": 1.33, "learning_rate": 2.0112588298796085e-05, "loss": 0.1636, "step": 57200 }, { "epoch": 1.33, "learning_rate": 2.011170824435005e-05, "loss": 0.2267, "step": 57205 }, { "epoch": 1.33, "learning_rate": 2.0110828189904018e-05, "loss": 0.4564, "step": 57210 }, { "epoch": 1.33, "learning_rate": 2.010994813545798e-05, "loss": 0.3155, "step": 57215 }, { "epoch": 1.34, "learning_rate": 2.0109068081011945e-05, "loss": 0.0268, "step": 57220 }, { "epoch": 1.34, "learning_rate": 2.010818802656591e-05, "loss": 0.0185, "step": 57225 }, { "epoch": 1.34, "learning_rate": 2.010730797211988e-05, "loss": 0.0874, "step": 57230 }, { "epoch": 1.34, "learning_rate": 2.010642791767384e-05, "loss": 0.2768, "step": 57235 }, { "epoch": 1.34, "learning_rate": 2.0105547863227805e-05, "loss": 0.2077, "step": 57240 }, { "epoch": 1.34, "learning_rate": 2.010466780878177e-05, "loss": 0.1496, "step": 57245 }, { "epoch": 1.34, "learning_rate": 2.0103787754335735e-05, "loss": 0.256, "step": 57250 }, { "epoch": 1.34, "learning_rate": 2.01029076998897e-05, "loss": 0.2653, "step": 57255 }, { "epoch": 1.34, "learning_rate": 2.0102027645443665e-05, "loss": 0.3526, "step": 57260 }, { "epoch": 1.34, "learning_rate": 2.010114759099763e-05, "loss": 0.2327, "step": 57265 }, { "epoch": 1.34, "learning_rate": 2.0100267536551595e-05, "loss": 0.0746, "step": 57270 }, { "epoch": 1.34, "learning_rate": 2.009938748210556e-05, "loss": 0.0962, "step": 57275 }, { "epoch": 1.34, "learning_rate": 2.0098507427659525e-05, "loss": 0.053, "step": 57280 }, { "epoch": 1.34, "learning_rate": 2.009762737321349e-05, "loss": 0.1239, "step": 57285 }, { "epoch": 1.34, "learning_rate": 2.0096747318767455e-05, "loss": 0.0768, "step": 57290 }, { "epoch": 1.34, "learning_rate": 2.009586726432142e-05, "loss": 0.2514, "step": 57295 }, { "epoch": 1.34, "learning_rate": 2.0094987209875385e-05, "loss": 0.1566, "step": 57300 }, { "epoch": 1.34, "learning_rate": 2.009410715542935e-05, "loss": 0.1258, "step": 57305 }, { "epoch": 1.34, "learning_rate": 2.0093227100983315e-05, "loss": 0.4338, "step": 57310 }, { "epoch": 1.34, "learning_rate": 2.009234704653728e-05, "loss": 0.185, "step": 57315 }, { "epoch": 1.34, "learning_rate": 2.0091466992091242e-05, "loss": 0.012, "step": 57320 }, { "epoch": 1.34, "learning_rate": 2.009058693764521e-05, "loss": 0.0705, "step": 57325 }, { "epoch": 1.34, "learning_rate": 2.0089706883199175e-05, "loss": 0.0359, "step": 57330 }, { "epoch": 1.34, "learning_rate": 2.008882682875314e-05, "loss": 0.1058, "step": 57335 }, { "epoch": 1.34, "learning_rate": 2.0087946774307102e-05, "loss": 0.1693, "step": 57340 }, { "epoch": 1.34, "learning_rate": 2.008706671986107e-05, "loss": 0.1703, "step": 57345 }, { "epoch": 1.34, "learning_rate": 2.0086186665415035e-05, "loss": 0.1032, "step": 57350 }, { "epoch": 1.34, "learning_rate": 2.0085306610968997e-05, "loss": 0.4779, "step": 57355 }, { "epoch": 1.34, "learning_rate": 2.0084426556522962e-05, "loss": 0.3782, "step": 57360 }, { "epoch": 1.34, "learning_rate": 2.008354650207693e-05, "loss": 0.2351, "step": 57365 }, { "epoch": 1.34, "learning_rate": 2.0082666447630896e-05, "loss": 0.0347, "step": 57370 }, { "epoch": 1.34, "learning_rate": 2.0081786393184857e-05, "loss": 0.0037, "step": 57375 }, { "epoch": 1.34, "learning_rate": 2.0080906338738822e-05, "loss": 0.0848, "step": 57380 }, { "epoch": 1.34, "learning_rate": 2.008002628429279e-05, "loss": 0.0959, "step": 57385 }, { "epoch": 1.34, "learning_rate": 2.0079146229846752e-05, "loss": 0.0964, "step": 57390 }, { "epoch": 1.34, "learning_rate": 2.0078266175400717e-05, "loss": 0.1501, "step": 57395 }, { "epoch": 1.34, "learning_rate": 2.0077386120954682e-05, "loss": 0.2577, "step": 57400 }, { "epoch": 1.34, "learning_rate": 2.007650606650865e-05, "loss": 0.2231, "step": 57405 }, { "epoch": 1.34, "learning_rate": 2.0075626012062612e-05, "loss": 0.2286, "step": 57410 }, { "epoch": 1.34, "learning_rate": 2.0074745957616577e-05, "loss": 0.3458, "step": 57415 }, { "epoch": 1.34, "learning_rate": 2.0073865903170546e-05, "loss": 0.0866, "step": 57420 }, { "epoch": 1.34, "learning_rate": 2.007298584872451e-05, "loss": 0.1028, "step": 57425 }, { "epoch": 1.34, "learning_rate": 2.0072105794278472e-05, "loss": 0.0466, "step": 57430 }, { "epoch": 1.34, "learning_rate": 2.0071225739832437e-05, "loss": 0.0209, "step": 57435 }, { "epoch": 1.34, "learning_rate": 2.0070345685386406e-05, "loss": 0.1188, "step": 57440 }, { "epoch": 1.34, "learning_rate": 2.0069465630940367e-05, "loss": 0.1184, "step": 57445 }, { "epoch": 1.34, "learning_rate": 2.0068585576494332e-05, "loss": 0.1822, "step": 57450 }, { "epoch": 1.34, "learning_rate": 2.0067705522048297e-05, "loss": 0.3161, "step": 57455 }, { "epoch": 1.34, "learning_rate": 2.0066825467602266e-05, "loss": 0.4061, "step": 57460 }, { "epoch": 1.34, "learning_rate": 2.0065945413156228e-05, "loss": 0.442, "step": 57465 }, { "epoch": 1.34, "learning_rate": 2.0065065358710193e-05, "loss": 0.0486, "step": 57470 }, { "epoch": 1.34, "learning_rate": 2.0064185304264158e-05, "loss": 0.0028, "step": 57475 }, { "epoch": 1.34, "learning_rate": 2.0063305249818123e-05, "loss": 0.1412, "step": 57480 }, { "epoch": 1.34, "learning_rate": 2.0062425195372088e-05, "loss": 0.0294, "step": 57485 }, { "epoch": 1.34, "learning_rate": 2.0061545140926053e-05, "loss": 0.0508, "step": 57490 }, { "epoch": 1.34, "learning_rate": 2.0060665086480018e-05, "loss": 0.1228, "step": 57495 }, { "epoch": 1.34, "learning_rate": 2.0059785032033983e-05, "loss": 0.1439, "step": 57500 }, { "epoch": 1.34, "learning_rate": 2.0058904977587948e-05, "loss": 0.1019, "step": 57505 }, { "epoch": 1.34, "learning_rate": 2.0058024923141913e-05, "loss": 0.2723, "step": 57510 }, { "epoch": 1.34, "learning_rate": 2.0057144868695874e-05, "loss": 0.3388, "step": 57515 }, { "epoch": 1.34, "learning_rate": 2.0056264814249843e-05, "loss": 0.0597, "step": 57520 }, { "epoch": 1.34, "learning_rate": 2.0055384759803808e-05, "loss": 0.0653, "step": 57525 }, { "epoch": 1.34, "learning_rate": 2.0054504705357773e-05, "loss": 0.1386, "step": 57530 }, { "epoch": 1.34, "learning_rate": 2.0053624650911734e-05, "loss": 0.107, "step": 57535 }, { "epoch": 1.34, "learning_rate": 2.0052744596465703e-05, "loss": 0.173, "step": 57540 }, { "epoch": 1.34, "learning_rate": 2.0051864542019668e-05, "loss": 0.0897, "step": 57545 }, { "epoch": 1.34, "learning_rate": 2.005098448757363e-05, "loss": 0.1279, "step": 57550 }, { "epoch": 1.34, "learning_rate": 2.0050104433127598e-05, "loss": 0.2833, "step": 57555 }, { "epoch": 1.34, "learning_rate": 2.0049224378681563e-05, "loss": 0.3247, "step": 57560 }, { "epoch": 1.34, "learning_rate": 2.0048344324235528e-05, "loss": 0.1833, "step": 57565 }, { "epoch": 1.34, "learning_rate": 2.004746426978949e-05, "loss": 0.1504, "step": 57570 }, { "epoch": 1.34, "learning_rate": 2.0046584215343458e-05, "loss": 0.0439, "step": 57575 }, { "epoch": 1.34, "learning_rate": 2.0045704160897423e-05, "loss": 0.1067, "step": 57580 }, { "epoch": 1.34, "learning_rate": 2.0044824106451385e-05, "loss": 0.1045, "step": 57585 }, { "epoch": 1.34, "learning_rate": 2.004394405200535e-05, "loss": 0.0759, "step": 57590 }, { "epoch": 1.34, "learning_rate": 2.0043063997559318e-05, "loss": 0.1405, "step": 57595 }, { "epoch": 1.34, "learning_rate": 2.0042183943113283e-05, "loss": 0.1778, "step": 57600 }, { "epoch": 1.34, "learning_rate": 2.0041303888667245e-05, "loss": 0.3137, "step": 57605 }, { "epoch": 1.34, "learning_rate": 2.004042383422121e-05, "loss": 0.3362, "step": 57610 }, { "epoch": 1.34, "learning_rate": 2.0039543779775178e-05, "loss": 0.3743, "step": 57615 }, { "epoch": 1.34, "learning_rate": 2.003866372532914e-05, "loss": 0.0263, "step": 57620 }, { "epoch": 1.34, "learning_rate": 2.0037783670883105e-05, "loss": 0.0389, "step": 57625 }, { "epoch": 1.34, "learning_rate": 2.003690361643707e-05, "loss": 0.1034, "step": 57630 }, { "epoch": 1.34, "learning_rate": 2.0036023561991038e-05, "loss": 0.137, "step": 57635 }, { "epoch": 1.34, "learning_rate": 2.0035143507545e-05, "loss": 0.1584, "step": 57640 }, { "epoch": 1.34, "learning_rate": 2.0034263453098965e-05, "loss": 0.2143, "step": 57645 }, { "epoch": 1.35, "learning_rate": 2.003338339865293e-05, "loss": 0.1924, "step": 57650 }, { "epoch": 1.35, "learning_rate": 2.0032503344206895e-05, "loss": 0.2626, "step": 57655 }, { "epoch": 1.35, "learning_rate": 2.003162328976086e-05, "loss": 0.4938, "step": 57660 }, { "epoch": 1.35, "learning_rate": 2.0030743235314825e-05, "loss": 0.2966, "step": 57665 }, { "epoch": 1.35, "learning_rate": 2.002986318086879e-05, "loss": 0.0455, "step": 57670 }, { "epoch": 1.35, "learning_rate": 2.0028983126422755e-05, "loss": 0.0691, "step": 57675 }, { "epoch": 1.35, "learning_rate": 2.002810307197672e-05, "loss": 0.0411, "step": 57680 }, { "epoch": 1.35, "learning_rate": 2.0027223017530685e-05, "loss": 0.1464, "step": 57685 }, { "epoch": 1.35, "learning_rate": 2.002634296308465e-05, "loss": 0.1306, "step": 57690 }, { "epoch": 1.35, "learning_rate": 2.0025462908638615e-05, "loss": 0.2014, "step": 57695 }, { "epoch": 1.35, "learning_rate": 2.002458285419258e-05, "loss": 0.0993, "step": 57700 }, { "epoch": 1.35, "learning_rate": 2.0023702799746545e-05, "loss": 0.4713, "step": 57705 }, { "epoch": 1.35, "learning_rate": 2.002282274530051e-05, "loss": 0.364, "step": 57710 }, { "epoch": 1.35, "learning_rate": 2.0021942690854475e-05, "loss": 0.2862, "step": 57715 }, { "epoch": 1.35, "learning_rate": 2.002106263640844e-05, "loss": 0.0392, "step": 57720 }, { "epoch": 1.35, "learning_rate": 2.0020182581962402e-05, "loss": 0.171, "step": 57725 }, { "epoch": 1.35, "learning_rate": 2.001930252751637e-05, "loss": 0.0298, "step": 57730 }, { "epoch": 1.35, "learning_rate": 2.0018422473070335e-05, "loss": 0.0396, "step": 57735 }, { "epoch": 1.35, "learning_rate": 2.00175424186243e-05, "loss": 0.2035, "step": 57740 }, { "epoch": 1.35, "learning_rate": 2.0016662364178262e-05, "loss": 0.2012, "step": 57745 }, { "epoch": 1.35, "learning_rate": 2.001578230973223e-05, "loss": 0.1846, "step": 57750 }, { "epoch": 1.35, "learning_rate": 2.0014902255286195e-05, "loss": 0.3163, "step": 57755 }, { "epoch": 1.35, "learning_rate": 2.001402220084016e-05, "loss": 0.1782, "step": 57760 }, { "epoch": 1.35, "learning_rate": 2.0013142146394122e-05, "loss": 0.2469, "step": 57765 }, { "epoch": 1.35, "learning_rate": 2.001226209194809e-05, "loss": 0.0608, "step": 57770 }, { "epoch": 1.35, "learning_rate": 2.0011382037502055e-05, "loss": 0.0382, "step": 57775 }, { "epoch": 1.35, "learning_rate": 2.0010501983056017e-05, "loss": 0.1208, "step": 57780 }, { "epoch": 1.35, "learning_rate": 2.0009621928609982e-05, "loss": 0.0868, "step": 57785 }, { "epoch": 1.35, "learning_rate": 2.000874187416395e-05, "loss": 0.1414, "step": 57790 }, { "epoch": 1.35, "learning_rate": 2.0007861819717915e-05, "loss": 0.1331, "step": 57795 }, { "epoch": 1.35, "learning_rate": 2.0006981765271877e-05, "loss": 0.1456, "step": 57800 }, { "epoch": 1.35, "learning_rate": 2.0006101710825842e-05, "loss": 0.1325, "step": 57805 }, { "epoch": 1.35, "learning_rate": 2.000522165637981e-05, "loss": 0.4329, "step": 57810 }, { "epoch": 1.35, "learning_rate": 2.0004341601933772e-05, "loss": 0.2925, "step": 57815 }, { "epoch": 1.35, "learning_rate": 2.0003461547487737e-05, "loss": 0.1479, "step": 57820 }, { "epoch": 1.35, "learning_rate": 2.0002581493041706e-05, "loss": 0.1564, "step": 57825 }, { "epoch": 1.35, "learning_rate": 2.000170143859567e-05, "loss": 0.0938, "step": 57830 }, { "epoch": 1.35, "learning_rate": 2.0000821384149632e-05, "loss": 0.0577, "step": 57835 }, { "epoch": 1.35, "learning_rate": 1.9999941329703597e-05, "loss": 0.357, "step": 57840 }, { "epoch": 1.35, "learning_rate": 1.9999061275257566e-05, "loss": 0.1398, "step": 57845 }, { "epoch": 1.35, "learning_rate": 1.9998181220811527e-05, "loss": 0.2076, "step": 57850 }, { "epoch": 1.35, "learning_rate": 1.9997301166365492e-05, "loss": 0.2388, "step": 57855 }, { "epoch": 1.35, "learning_rate": 1.9996421111919457e-05, "loss": 0.3389, "step": 57860 }, { "epoch": 1.35, "learning_rate": 1.9995541057473426e-05, "loss": 0.3258, "step": 57865 }, { "epoch": 1.35, "learning_rate": 1.9994661003027387e-05, "loss": 0.1242, "step": 57870 }, { "epoch": 1.35, "learning_rate": 1.9993780948581352e-05, "loss": 0.06, "step": 57875 }, { "epoch": 1.35, "learning_rate": 1.9992900894135317e-05, "loss": 0.0876, "step": 57880 }, { "epoch": 1.35, "learning_rate": 1.9992020839689282e-05, "loss": 0.1529, "step": 57885 }, { "epoch": 1.35, "learning_rate": 1.9991140785243247e-05, "loss": 0.1154, "step": 57890 }, { "epoch": 1.35, "learning_rate": 1.9990260730797212e-05, "loss": 0.1814, "step": 57895 }, { "epoch": 1.35, "learning_rate": 1.9989380676351178e-05, "loss": 0.2273, "step": 57900 }, { "epoch": 1.35, "learning_rate": 1.9988500621905143e-05, "loss": 0.2452, "step": 57905 }, { "epoch": 1.35, "learning_rate": 1.9987620567459108e-05, "loss": 0.2826, "step": 57910 }, { "epoch": 1.35, "learning_rate": 1.9986740513013073e-05, "loss": 0.3553, "step": 57915 }, { "epoch": 1.35, "learning_rate": 1.9985860458567034e-05, "loss": 0.0899, "step": 57920 }, { "epoch": 1.35, "learning_rate": 1.9984980404121003e-05, "loss": 0.0708, "step": 57925 }, { "epoch": 1.35, "learning_rate": 1.9984100349674968e-05, "loss": 0.0942, "step": 57930 }, { "epoch": 1.35, "learning_rate": 1.9983220295228933e-05, "loss": 0.0725, "step": 57935 }, { "epoch": 1.35, "learning_rate": 1.9982340240782894e-05, "loss": 0.1815, "step": 57940 }, { "epoch": 1.35, "learning_rate": 1.9981460186336863e-05, "loss": 0.1106, "step": 57945 }, { "epoch": 1.35, "learning_rate": 1.9980580131890828e-05, "loss": 0.2897, "step": 57950 }, { "epoch": 1.35, "learning_rate": 1.997970007744479e-05, "loss": 0.1892, "step": 57955 }, { "epoch": 1.35, "learning_rate": 1.9978820022998758e-05, "loss": 0.2037, "step": 57960 }, { "epoch": 1.35, "learning_rate": 1.9977939968552723e-05, "loss": 0.221, "step": 57965 }, { "epoch": 1.35, "learning_rate": 1.9977059914106688e-05, "loss": 0.0047, "step": 57970 }, { "epoch": 1.35, "learning_rate": 1.997617985966065e-05, "loss": 0.0423, "step": 57975 }, { "epoch": 1.35, "learning_rate": 1.9975299805214618e-05, "loss": 0.1155, "step": 57980 }, { "epoch": 1.35, "learning_rate": 1.9974419750768583e-05, "loss": 0.1655, "step": 57985 }, { "epoch": 1.35, "learning_rate": 1.9973539696322544e-05, "loss": 0.0716, "step": 57990 }, { "epoch": 1.35, "learning_rate": 1.997265964187651e-05, "loss": 0.168, "step": 57995 }, { "epoch": 1.35, "learning_rate": 1.9971779587430478e-05, "loss": 0.1861, "step": 58000 }, { "epoch": 1.35, "learning_rate": 1.9970899532984443e-05, "loss": 0.2037, "step": 58005 }, { "epoch": 1.35, "learning_rate": 1.9970019478538405e-05, "loss": 0.401, "step": 58010 }, { "epoch": 1.35, "learning_rate": 1.996913942409237e-05, "loss": 0.3483, "step": 58015 }, { "epoch": 1.35, "learning_rate": 1.9968259369646338e-05, "loss": 0.0373, "step": 58020 }, { "epoch": 1.35, "learning_rate": 1.99673793152003e-05, "loss": 0.0274, "step": 58025 }, { "epoch": 1.35, "learning_rate": 1.9966499260754265e-05, "loss": 0.1556, "step": 58030 }, { "epoch": 1.35, "learning_rate": 1.996561920630823e-05, "loss": 0.051, "step": 58035 }, { "epoch": 1.35, "learning_rate": 1.9964739151862198e-05, "loss": 0.0997, "step": 58040 }, { "epoch": 1.35, "learning_rate": 1.996385909741616e-05, "loss": 0.1145, "step": 58045 }, { "epoch": 1.35, "learning_rate": 1.9962979042970125e-05, "loss": 0.1099, "step": 58050 }, { "epoch": 1.35, "learning_rate": 1.996209898852409e-05, "loss": 0.2351, "step": 58055 }, { "epoch": 1.35, "learning_rate": 1.9961218934078058e-05, "loss": 0.6958, "step": 58060 }, { "epoch": 1.35, "learning_rate": 1.996033887963202e-05, "loss": 0.2278, "step": 58065 }, { "epoch": 1.35, "learning_rate": 1.9959458825185985e-05, "loss": 0.0769, "step": 58070 }, { "epoch": 1.35, "learning_rate": 1.995857877073995e-05, "loss": 0.0573, "step": 58075 }, { "epoch": 1.36, "learning_rate": 1.9957698716293915e-05, "loss": 0.056, "step": 58080 }, { "epoch": 1.36, "learning_rate": 1.995681866184788e-05, "loss": 0.0086, "step": 58085 }, { "epoch": 1.36, "learning_rate": 1.9955938607401845e-05, "loss": 0.1029, "step": 58090 }, { "epoch": 1.36, "learning_rate": 1.995505855295581e-05, "loss": 0.1049, "step": 58095 }, { "epoch": 1.36, "learning_rate": 1.9954178498509775e-05, "loss": 0.1872, "step": 58100 }, { "epoch": 1.36, "learning_rate": 1.995329844406374e-05, "loss": 0.3442, "step": 58105 }, { "epoch": 1.36, "learning_rate": 1.9952418389617705e-05, "loss": 0.2491, "step": 58110 }, { "epoch": 1.36, "learning_rate": 1.995153833517167e-05, "loss": 0.0986, "step": 58115 }, { "epoch": 1.36, "learning_rate": 1.9950658280725635e-05, "loss": 0.0035, "step": 58120 }, { "epoch": 1.36, "learning_rate": 1.99497782262796e-05, "loss": 0.1074, "step": 58125 }, { "epoch": 1.36, "learning_rate": 1.9948898171833565e-05, "loss": 0.0568, "step": 58130 }, { "epoch": 1.36, "learning_rate": 1.994801811738753e-05, "loss": 0.2845, "step": 58135 }, { "epoch": 1.36, "learning_rate": 1.9947138062941495e-05, "loss": 0.1092, "step": 58140 }, { "epoch": 1.36, "learning_rate": 1.994625800849546e-05, "loss": 0.0904, "step": 58145 }, { "epoch": 1.36, "learning_rate": 1.9945377954049422e-05, "loss": 0.1492, "step": 58150 }, { "epoch": 1.36, "learning_rate": 1.994449789960339e-05, "loss": 0.3234, "step": 58155 }, { "epoch": 1.36, "learning_rate": 1.9943617845157355e-05, "loss": 0.2545, "step": 58160 }, { "epoch": 1.36, "learning_rate": 1.994273779071132e-05, "loss": 0.3458, "step": 58165 }, { "epoch": 1.36, "learning_rate": 1.9941857736265282e-05, "loss": 0.0223, "step": 58170 }, { "epoch": 1.36, "learning_rate": 1.994097768181925e-05, "loss": 0.1386, "step": 58175 }, { "epoch": 1.36, "learning_rate": 1.9940097627373215e-05, "loss": 0.123, "step": 58180 }, { "epoch": 1.36, "learning_rate": 1.9939217572927177e-05, "loss": 0.0823, "step": 58185 }, { "epoch": 1.36, "learning_rate": 1.9938337518481142e-05, "loss": 0.2015, "step": 58190 }, { "epoch": 1.36, "learning_rate": 1.993745746403511e-05, "loss": 0.1771, "step": 58195 }, { "epoch": 1.36, "learning_rate": 1.9936577409589075e-05, "loss": 0.1618, "step": 58200 }, { "epoch": 1.36, "learning_rate": 1.9935697355143037e-05, "loss": 0.1108, "step": 58205 }, { "epoch": 1.36, "learning_rate": 1.9934817300697002e-05, "loss": 0.2949, "step": 58210 }, { "epoch": 1.36, "learning_rate": 1.993393724625097e-05, "loss": 0.3016, "step": 58215 }, { "epoch": 1.36, "learning_rate": 1.9933057191804932e-05, "loss": 0.0427, "step": 58220 }, { "epoch": 1.36, "learning_rate": 1.9932177137358897e-05, "loss": 0.0956, "step": 58225 }, { "epoch": 1.36, "learning_rate": 1.9931297082912862e-05, "loss": 0.0775, "step": 58230 }, { "epoch": 1.36, "learning_rate": 1.993041702846683e-05, "loss": 0.0617, "step": 58235 }, { "epoch": 1.36, "learning_rate": 1.9929536974020792e-05, "loss": 0.0447, "step": 58240 }, { "epoch": 1.36, "learning_rate": 1.9928656919574757e-05, "loss": 0.2659, "step": 58245 }, { "epoch": 1.36, "learning_rate": 1.9927776865128726e-05, "loss": 0.1321, "step": 58250 }, { "epoch": 1.36, "learning_rate": 1.9926896810682687e-05, "loss": 0.1932, "step": 58255 }, { "epoch": 1.36, "learning_rate": 1.9926016756236652e-05, "loss": 0.3252, "step": 58260 }, { "epoch": 1.36, "learning_rate": 1.9925136701790617e-05, "loss": 0.2857, "step": 58265 }, { "epoch": 1.36, "learning_rate": 1.9924256647344586e-05, "loss": 0.0322, "step": 58270 }, { "epoch": 1.36, "learning_rate": 1.9923376592898547e-05, "loss": 0.108, "step": 58275 }, { "epoch": 1.36, "learning_rate": 1.9922496538452512e-05, "loss": 0.1585, "step": 58280 }, { "epoch": 1.36, "learning_rate": 1.9921616484006477e-05, "loss": 0.0555, "step": 58285 }, { "epoch": 1.36, "learning_rate": 1.9920736429560442e-05, "loss": 0.068, "step": 58290 }, { "epoch": 1.36, "learning_rate": 1.9919856375114407e-05, "loss": 0.3356, "step": 58295 }, { "epoch": 1.36, "learning_rate": 1.9918976320668372e-05, "loss": 0.1601, "step": 58300 }, { "epoch": 1.36, "learning_rate": 1.9918272277111547e-05, "loss": 0.2389, "step": 58305 }, { "epoch": 1.36, "learning_rate": 1.991739222266551e-05, "loss": 0.451, "step": 58310 }, { "epoch": 1.36, "learning_rate": 1.9916512168219474e-05, "loss": 0.3161, "step": 58315 }, { "epoch": 1.36, "learning_rate": 1.991563211377344e-05, "loss": 0.0368, "step": 58320 }, { "epoch": 1.36, "learning_rate": 1.9914752059327407e-05, "loss": 0.0499, "step": 58325 }, { "epoch": 1.36, "learning_rate": 1.991387200488137e-05, "loss": 0.0405, "step": 58330 }, { "epoch": 1.36, "learning_rate": 1.9912991950435334e-05, "loss": 0.1847, "step": 58335 }, { "epoch": 1.36, "learning_rate": 1.99121118959893e-05, "loss": 0.0314, "step": 58340 }, { "epoch": 1.36, "learning_rate": 1.9911231841543264e-05, "loss": 0.1033, "step": 58345 }, { "epoch": 1.36, "learning_rate": 1.991035178709723e-05, "loss": 0.2812, "step": 58350 }, { "epoch": 1.36, "learning_rate": 1.9909471732651194e-05, "loss": 0.2465, "step": 58355 }, { "epoch": 1.36, "learning_rate": 1.990859167820516e-05, "loss": 0.2273, "step": 58360 }, { "epoch": 1.36, "learning_rate": 1.9907711623759124e-05, "loss": 0.2144, "step": 58365 }, { "epoch": 1.36, "learning_rate": 1.990683156931309e-05, "loss": 0.0557, "step": 58370 }, { "epoch": 1.36, "learning_rate": 1.9905951514867054e-05, "loss": 0.0567, "step": 58375 }, { "epoch": 1.36, "learning_rate": 1.9905071460421016e-05, "loss": 0.1675, "step": 58380 }, { "epoch": 1.36, "learning_rate": 1.9904191405974984e-05, "loss": 0.0878, "step": 58385 }, { "epoch": 1.36, "learning_rate": 1.990331135152895e-05, "loss": 0.0464, "step": 58390 }, { "epoch": 1.36, "learning_rate": 1.9902431297082914e-05, "loss": 0.1358, "step": 58395 }, { "epoch": 1.36, "learning_rate": 1.9901551242636876e-05, "loss": 0.3426, "step": 58400 }, { "epoch": 1.36, "learning_rate": 1.9900671188190844e-05, "loss": 0.158, "step": 58405 }, { "epoch": 1.36, "learning_rate": 1.989979113374481e-05, "loss": 0.1943, "step": 58410 }, { "epoch": 1.36, "learning_rate": 1.989891107929877e-05, "loss": 0.2939, "step": 58415 }, { "epoch": 1.36, "learning_rate": 1.989803102485274e-05, "loss": 0.0143, "step": 58420 }, { "epoch": 1.36, "learning_rate": 1.9897150970406704e-05, "loss": 0.029, "step": 58425 }, { "epoch": 1.36, "learning_rate": 1.989627091596067e-05, "loss": 0.0435, "step": 58430 }, { "epoch": 1.36, "learning_rate": 1.989539086151463e-05, "loss": 0.0869, "step": 58435 }, { "epoch": 1.36, "learning_rate": 1.98945108070686e-05, "loss": 0.1341, "step": 58440 }, { "epoch": 1.36, "learning_rate": 1.9893630752622564e-05, "loss": 0.1211, "step": 58445 }, { "epoch": 1.36, "learning_rate": 1.9892750698176526e-05, "loss": 0.1752, "step": 58450 }, { "epoch": 1.36, "learning_rate": 1.989187064373049e-05, "loss": 0.043, "step": 58455 }, { "epoch": 1.36, "learning_rate": 1.989099058928446e-05, "loss": 0.3567, "step": 58460 }, { "epoch": 1.36, "learning_rate": 1.9890110534838424e-05, "loss": 0.2681, "step": 58465 }, { "epoch": 1.36, "learning_rate": 1.9889230480392386e-05, "loss": 0.0805, "step": 58470 }, { "epoch": 1.36, "learning_rate": 1.988835042594635e-05, "loss": 0.0534, "step": 58475 }, { "epoch": 1.36, "learning_rate": 1.988747037150032e-05, "loss": 0.0924, "step": 58480 }, { "epoch": 1.36, "learning_rate": 1.988659031705428e-05, "loss": 0.137, "step": 58485 }, { "epoch": 1.36, "learning_rate": 1.9885710262608246e-05, "loss": 0.1567, "step": 58490 }, { "epoch": 1.36, "learning_rate": 1.988483020816221e-05, "loss": 0.1911, "step": 58495 }, { "epoch": 1.36, "learning_rate": 1.988395015371618e-05, "loss": 0.2257, "step": 58500 }, { "epoch": 1.36, "learning_rate": 1.988307009927014e-05, "loss": 0.1954, "step": 58505 }, { "epoch": 1.37, "learning_rate": 1.9882190044824106e-05, "loss": 0.3115, "step": 58510 }, { "epoch": 1.37, "learning_rate": 1.988130999037807e-05, "loss": 0.2381, "step": 58515 }, { "epoch": 1.37, "learning_rate": 1.9880429935932036e-05, "loss": 0.0902, "step": 58520 }, { "epoch": 1.37, "learning_rate": 1.9879549881486e-05, "loss": 0.1028, "step": 58525 }, { "epoch": 1.37, "learning_rate": 1.9878669827039966e-05, "loss": 0.061, "step": 58530 }, { "epoch": 1.37, "learning_rate": 1.987778977259393e-05, "loss": 0.1337, "step": 58535 }, { "epoch": 1.37, "learning_rate": 1.9876909718147896e-05, "loss": 0.1197, "step": 58540 }, { "epoch": 1.37, "learning_rate": 1.987602966370186e-05, "loss": 0.0904, "step": 58545 }, { "epoch": 1.37, "learning_rate": 1.9875149609255826e-05, "loss": 0.0638, "step": 58550 }, { "epoch": 1.37, "learning_rate": 1.987426955480979e-05, "loss": 0.2062, "step": 58555 }, { "epoch": 1.37, "learning_rate": 1.9873389500363756e-05, "loss": 0.3832, "step": 58560 }, { "epoch": 1.37, "learning_rate": 1.987250944591772e-05, "loss": 0.3465, "step": 58565 }, { "epoch": 1.37, "learning_rate": 1.9871629391471686e-05, "loss": 0.056, "step": 58570 }, { "epoch": 1.37, "learning_rate": 1.987074933702565e-05, "loss": 0.0403, "step": 58575 }, { "epoch": 1.37, "learning_rate": 1.9869869282579616e-05, "loss": 0.1745, "step": 58580 }, { "epoch": 1.37, "learning_rate": 1.986898922813358e-05, "loss": 0.0992, "step": 58585 }, { "epoch": 1.37, "learning_rate": 1.9868109173687547e-05, "loss": 0.0648, "step": 58590 }, { "epoch": 1.37, "learning_rate": 1.986722911924151e-05, "loss": 0.1391, "step": 58595 }, { "epoch": 1.37, "learning_rate": 1.9866349064795477e-05, "loss": 0.2619, "step": 58600 }, { "epoch": 1.37, "learning_rate": 1.986546901034944e-05, "loss": 0.1275, "step": 58605 }, { "epoch": 1.37, "learning_rate": 1.9864588955903403e-05, "loss": 0.3173, "step": 58610 }, { "epoch": 1.37, "learning_rate": 1.986370890145737e-05, "loss": 0.1789, "step": 58615 }, { "epoch": 1.37, "learning_rate": 1.9862828847011337e-05, "loss": 0.157, "step": 58620 }, { "epoch": 1.37, "learning_rate": 1.98619487925653e-05, "loss": 0.0505, "step": 58625 }, { "epoch": 1.37, "learning_rate": 1.9861068738119263e-05, "loss": 0.131, "step": 58630 }, { "epoch": 1.37, "learning_rate": 1.986018868367323e-05, "loss": 0.122, "step": 58635 }, { "epoch": 1.37, "learning_rate": 1.9859308629227197e-05, "loss": 0.2928, "step": 58640 }, { "epoch": 1.37, "learning_rate": 1.985842857478116e-05, "loss": 0.2069, "step": 58645 }, { "epoch": 1.37, "learning_rate": 1.9857548520335123e-05, "loss": 0.115, "step": 58650 }, { "epoch": 1.37, "learning_rate": 1.9856668465889092e-05, "loss": 0.1785, "step": 58655 }, { "epoch": 1.37, "learning_rate": 1.9855788411443057e-05, "loss": 0.3761, "step": 58660 }, { "epoch": 1.37, "learning_rate": 1.985490835699702e-05, "loss": 0.2479, "step": 58665 }, { "epoch": 1.37, "learning_rate": 1.9854028302550983e-05, "loss": 0.1831, "step": 58670 }, { "epoch": 1.37, "learning_rate": 1.9853148248104952e-05, "loss": 0.0317, "step": 58675 }, { "epoch": 1.37, "learning_rate": 1.9852268193658913e-05, "loss": 0.1584, "step": 58680 }, { "epoch": 1.37, "learning_rate": 1.985138813921288e-05, "loss": 0.0458, "step": 58685 }, { "epoch": 1.37, "learning_rate": 1.9850508084766844e-05, "loss": 0.0728, "step": 58690 }, { "epoch": 1.37, "learning_rate": 1.9849628030320812e-05, "loss": 0.0988, "step": 58695 }, { "epoch": 1.37, "learning_rate": 1.9848747975874774e-05, "loss": 0.199, "step": 58700 }, { "epoch": 1.37, "learning_rate": 1.984786792142874e-05, "loss": 0.2135, "step": 58705 }, { "epoch": 1.37, "learning_rate": 1.9846987866982707e-05, "loss": 0.283, "step": 58710 }, { "epoch": 1.37, "learning_rate": 1.984610781253667e-05, "loss": 0.2869, "step": 58715 }, { "epoch": 1.37, "learning_rate": 1.9845227758090634e-05, "loss": 0.0618, "step": 58720 }, { "epoch": 1.37, "learning_rate": 1.98443477036446e-05, "loss": 0.0567, "step": 58725 }, { "epoch": 1.37, "learning_rate": 1.9843467649198567e-05, "loss": 0.0577, "step": 58730 }, { "epoch": 1.37, "learning_rate": 1.984258759475253e-05, "loss": 0.0902, "step": 58735 }, { "epoch": 1.37, "learning_rate": 1.9841707540306494e-05, "loss": 0.063, "step": 58740 }, { "epoch": 1.37, "learning_rate": 1.984082748586046e-05, "loss": 0.1536, "step": 58745 }, { "epoch": 1.37, "learning_rate": 1.9839947431414424e-05, "loss": 0.1904, "step": 58750 }, { "epoch": 1.37, "learning_rate": 1.983906737696839e-05, "loss": 0.2304, "step": 58755 }, { "epoch": 1.37, "learning_rate": 1.9838187322522354e-05, "loss": 0.4364, "step": 58760 }, { "epoch": 1.37, "learning_rate": 1.983730726807632e-05, "loss": 0.1778, "step": 58765 }, { "epoch": 1.37, "learning_rate": 1.9836427213630284e-05, "loss": 0.0151, "step": 58770 }, { "epoch": 1.37, "learning_rate": 1.983554715918425e-05, "loss": 0.0431, "step": 58775 }, { "epoch": 1.37, "learning_rate": 1.9834667104738214e-05, "loss": 0.0675, "step": 58780 }, { "epoch": 1.37, "learning_rate": 1.9833787050292176e-05, "loss": 0.0877, "step": 58785 }, { "epoch": 1.37, "learning_rate": 1.9832906995846144e-05, "loss": 0.0901, "step": 58790 }, { "epoch": 1.37, "learning_rate": 1.983202694140011e-05, "loss": 0.2187, "step": 58795 }, { "epoch": 1.37, "learning_rate": 1.9831146886954074e-05, "loss": 0.1073, "step": 58800 }, { "epoch": 1.37, "learning_rate": 1.9830266832508036e-05, "loss": 0.2034, "step": 58805 }, { "epoch": 1.37, "learning_rate": 1.9829386778062004e-05, "loss": 0.3036, "step": 58810 }, { "epoch": 1.37, "learning_rate": 1.982850672361597e-05, "loss": 0.1226, "step": 58815 }, { "epoch": 1.37, "learning_rate": 1.982762666916993e-05, "loss": 0.0619, "step": 58820 }, { "epoch": 1.37, "learning_rate": 1.9826746614723896e-05, "loss": 0.1742, "step": 58825 }, { "epoch": 1.37, "learning_rate": 1.9825866560277864e-05, "loss": 0.0211, "step": 58830 }, { "epoch": 1.37, "learning_rate": 1.982498650583183e-05, "loss": 0.0917, "step": 58835 }, { "epoch": 1.37, "learning_rate": 1.982410645138579e-05, "loss": 0.1474, "step": 58840 }, { "epoch": 1.37, "learning_rate": 1.982322639693976e-05, "loss": 0.1511, "step": 58845 }, { "epoch": 1.37, "learning_rate": 1.9822346342493724e-05, "loss": 0.1339, "step": 58850 }, { "epoch": 1.37, "learning_rate": 1.982146628804769e-05, "loss": 0.1996, "step": 58855 }, { "epoch": 1.37, "learning_rate": 1.982058623360165e-05, "loss": 0.3108, "step": 58860 }, { "epoch": 1.37, "learning_rate": 1.981970617915562e-05, "loss": 0.2374, "step": 58865 }, { "epoch": 1.37, "learning_rate": 1.9818826124709584e-05, "loss": 0.0226, "step": 58870 }, { "epoch": 1.37, "learning_rate": 1.9817946070263546e-05, "loss": 0.0913, "step": 58875 }, { "epoch": 1.37, "learning_rate": 1.981706601581751e-05, "loss": 0.0684, "step": 58880 }, { "epoch": 1.37, "learning_rate": 1.981618596137148e-05, "loss": 0.0891, "step": 58885 }, { "epoch": 1.37, "learning_rate": 1.9815305906925444e-05, "loss": 0.1026, "step": 58890 }, { "epoch": 1.37, "learning_rate": 1.9814425852479406e-05, "loss": 0.0412, "step": 58895 }, { "epoch": 1.37, "learning_rate": 1.981354579803337e-05, "loss": 0.3528, "step": 58900 }, { "epoch": 1.37, "learning_rate": 1.981266574358734e-05, "loss": 0.3097, "step": 58905 }, { "epoch": 1.37, "learning_rate": 1.98117856891413e-05, "loss": 0.3771, "step": 58910 }, { "epoch": 1.37, "learning_rate": 1.9810905634695266e-05, "loss": 0.2157, "step": 58915 }, { "epoch": 1.37, "learning_rate": 1.981002558024923e-05, "loss": 0.0674, "step": 58920 }, { "epoch": 1.37, "learning_rate": 1.98091455258032e-05, "loss": 0.2425, "step": 58925 }, { "epoch": 1.37, "learning_rate": 1.980826547135716e-05, "loss": 0.0703, "step": 58930 }, { "epoch": 1.38, "learning_rate": 1.9807385416911126e-05, "loss": 0.0507, "step": 58935 }, { "epoch": 1.38, "learning_rate": 1.980650536246509e-05, "loss": 0.1594, "step": 58940 }, { "epoch": 1.38, "learning_rate": 1.9805625308019056e-05, "loss": 0.0494, "step": 58945 }, { "epoch": 1.38, "learning_rate": 1.980474525357302e-05, "loss": 0.3073, "step": 58950 }, { "epoch": 1.38, "learning_rate": 1.9803865199126986e-05, "loss": 0.1442, "step": 58955 }, { "epoch": 1.38, "learning_rate": 1.980298514468095e-05, "loss": 0.3403, "step": 58960 }, { "epoch": 1.38, "learning_rate": 1.9802105090234916e-05, "loss": 0.1942, "step": 58965 }, { "epoch": 1.38, "learning_rate": 1.980122503578888e-05, "loss": 0.106, "step": 58970 }, { "epoch": 1.38, "learning_rate": 1.9800344981342846e-05, "loss": 0.0245, "step": 58975 }, { "epoch": 1.38, "learning_rate": 1.979946492689681e-05, "loss": 0.0811, "step": 58980 }, { "epoch": 1.38, "learning_rate": 1.9798584872450776e-05, "loss": 0.0779, "step": 58985 }, { "epoch": 1.38, "learning_rate": 1.979770481800474e-05, "loss": 0.1913, "step": 58990 }, { "epoch": 1.38, "learning_rate": 1.9796824763558706e-05, "loss": 0.1903, "step": 58995 }, { "epoch": 1.38, "learning_rate": 1.979594470911267e-05, "loss": 0.2867, "step": 59000 }, { "epoch": 1.38, "learning_rate": 1.9795064654666636e-05, "loss": 0.1853, "step": 59005 }, { "epoch": 1.38, "learning_rate": 1.97941846002206e-05, "loss": 0.5407, "step": 59010 }, { "epoch": 1.38, "learning_rate": 1.9793304545774563e-05, "loss": 0.2038, "step": 59015 }, { "epoch": 1.38, "learning_rate": 1.979242449132853e-05, "loss": 0.02, "step": 59020 }, { "epoch": 1.38, "learning_rate": 1.9791544436882497e-05, "loss": 0.1505, "step": 59025 }, { "epoch": 1.38, "learning_rate": 1.979066438243646e-05, "loss": 0.1216, "step": 59030 }, { "epoch": 1.38, "learning_rate": 1.9789784327990423e-05, "loss": 0.064, "step": 59035 }, { "epoch": 1.38, "learning_rate": 1.978890427354439e-05, "loss": 0.0743, "step": 59040 }, { "epoch": 1.38, "learning_rate": 1.9788024219098357e-05, "loss": 0.216, "step": 59045 }, { "epoch": 1.38, "learning_rate": 1.9787144164652318e-05, "loss": 0.1639, "step": 59050 }, { "epoch": 1.38, "learning_rate": 1.9786264110206283e-05, "loss": 0.2697, "step": 59055 }, { "epoch": 1.38, "learning_rate": 1.978538405576025e-05, "loss": 0.3423, "step": 59060 }, { "epoch": 1.38, "learning_rate": 1.9784504001314217e-05, "loss": 0.318, "step": 59065 }, { "epoch": 1.38, "learning_rate": 1.978362394686818e-05, "loss": 0.0113, "step": 59070 }, { "epoch": 1.38, "learning_rate": 1.9782743892422143e-05, "loss": 0.0362, "step": 59075 }, { "epoch": 1.38, "learning_rate": 1.9781863837976112e-05, "loss": 0.2099, "step": 59080 }, { "epoch": 1.38, "learning_rate": 1.9780983783530073e-05, "loss": 0.0178, "step": 59085 }, { "epoch": 1.38, "learning_rate": 1.978010372908404e-05, "loss": 0.1009, "step": 59090 }, { "epoch": 1.38, "learning_rate": 1.9779223674638003e-05, "loss": 0.2001, "step": 59095 }, { "epoch": 1.38, "learning_rate": 1.9778343620191972e-05, "loss": 0.2796, "step": 59100 }, { "epoch": 1.38, "learning_rate": 1.9777463565745933e-05, "loss": 0.2602, "step": 59105 }, { "epoch": 1.38, "learning_rate": 1.97765835112999e-05, "loss": 0.254, "step": 59110 }, { "epoch": 1.38, "learning_rate": 1.9775703456853867e-05, "loss": 0.2729, "step": 59115 }, { "epoch": 1.38, "learning_rate": 1.977482340240783e-05, "loss": 0.1194, "step": 59120 }, { "epoch": 1.38, "learning_rate": 1.9773943347961794e-05, "loss": 0.0496, "step": 59125 }, { "epoch": 1.38, "learning_rate": 1.977306329351576e-05, "loss": 0.1509, "step": 59130 }, { "epoch": 1.38, "learning_rate": 1.9772183239069727e-05, "loss": 0.126, "step": 59135 }, { "epoch": 1.38, "learning_rate": 1.977130318462369e-05, "loss": 0.1716, "step": 59140 }, { "epoch": 1.38, "learning_rate": 1.9770423130177654e-05, "loss": 0.1133, "step": 59145 }, { "epoch": 1.38, "learning_rate": 1.976954307573162e-05, "loss": 0.0769, "step": 59150 }, { "epoch": 1.38, "learning_rate": 1.9768663021285587e-05, "loss": 0.1831, "step": 59155 }, { "epoch": 1.38, "learning_rate": 1.976778296683955e-05, "loss": 0.2067, "step": 59160 }, { "epoch": 1.38, "learning_rate": 1.9766902912393514e-05, "loss": 0.369, "step": 59165 }, { "epoch": 1.38, "learning_rate": 1.976602285794748e-05, "loss": 0.0246, "step": 59170 }, { "epoch": 1.38, "learning_rate": 1.9765142803501444e-05, "loss": 0.0518, "step": 59175 }, { "epoch": 1.38, "learning_rate": 1.976426274905541e-05, "loss": 0.0603, "step": 59180 }, { "epoch": 1.38, "learning_rate": 1.9763382694609374e-05, "loss": 0.085, "step": 59185 }, { "epoch": 1.38, "learning_rate": 1.976250264016334e-05, "loss": 0.0366, "step": 59190 }, { "epoch": 1.38, "learning_rate": 1.9761622585717304e-05, "loss": 0.141, "step": 59195 }, { "epoch": 1.38, "learning_rate": 1.976074253127127e-05, "loss": 0.3081, "step": 59200 }, { "epoch": 1.38, "learning_rate": 1.9759862476825234e-05, "loss": 0.3558, "step": 59205 }, { "epoch": 1.38, "learning_rate": 1.9758982422379195e-05, "loss": 0.3445, "step": 59210 }, { "epoch": 1.38, "learning_rate": 1.9758102367933164e-05, "loss": 0.4513, "step": 59215 }, { "epoch": 1.38, "learning_rate": 1.975722231348713e-05, "loss": 0.0622, "step": 59220 }, { "epoch": 1.38, "learning_rate": 1.9756342259041094e-05, "loss": 0.0813, "step": 59225 }, { "epoch": 1.38, "learning_rate": 1.9755462204595056e-05, "loss": 0.072, "step": 59230 }, { "epoch": 1.38, "learning_rate": 1.9754582150149024e-05, "loss": 0.1286, "step": 59235 }, { "epoch": 1.38, "learning_rate": 1.975370209570299e-05, "loss": 0.0502, "step": 59240 }, { "epoch": 1.38, "learning_rate": 1.975282204125695e-05, "loss": 0.1345, "step": 59245 }, { "epoch": 1.38, "learning_rate": 1.975194198681092e-05, "loss": 0.0903, "step": 59250 }, { "epoch": 1.38, "learning_rate": 1.9751061932364884e-05, "loss": 0.2625, "step": 59255 }, { "epoch": 1.38, "learning_rate": 1.975018187791885e-05, "loss": 0.6721, "step": 59260 }, { "epoch": 1.38, "learning_rate": 1.974930182347281e-05, "loss": 0.3712, "step": 59265 }, { "epoch": 1.38, "learning_rate": 1.974842176902678e-05, "loss": 0.0708, "step": 59270 }, { "epoch": 1.38, "learning_rate": 1.9747541714580744e-05, "loss": 0.0222, "step": 59275 }, { "epoch": 1.38, "learning_rate": 1.9746661660134706e-05, "loss": 0.0909, "step": 59280 }, { "epoch": 1.38, "learning_rate": 1.974578160568867e-05, "loss": 0.0711, "step": 59285 }, { "epoch": 1.38, "learning_rate": 1.974490155124264e-05, "loss": 0.0432, "step": 59290 }, { "epoch": 1.38, "learning_rate": 1.9744021496796604e-05, "loss": 0.1986, "step": 59295 }, { "epoch": 1.38, "learning_rate": 1.9743141442350566e-05, "loss": 0.1592, "step": 59300 }, { "epoch": 1.38, "learning_rate": 1.974226138790453e-05, "loss": 0.1678, "step": 59305 }, { "epoch": 1.38, "learning_rate": 1.97413813334585e-05, "loss": 0.2354, "step": 59310 }, { "epoch": 1.38, "learning_rate": 1.974050127901246e-05, "loss": 0.1855, "step": 59315 }, { "epoch": 1.38, "learning_rate": 1.9739621224566426e-05, "loss": 0.0417, "step": 59320 }, { "epoch": 1.38, "learning_rate": 1.973874117012039e-05, "loss": 0.0931, "step": 59325 }, { "epoch": 1.38, "learning_rate": 1.973786111567436e-05, "loss": 0.0487, "step": 59330 }, { "epoch": 1.38, "learning_rate": 1.973698106122832e-05, "loss": 0.1323, "step": 59335 }, { "epoch": 1.38, "learning_rate": 1.9736101006782286e-05, "loss": 0.0329, "step": 59340 }, { "epoch": 1.38, "learning_rate": 1.973522095233625e-05, "loss": 0.2315, "step": 59345 }, { "epoch": 1.38, "learning_rate": 1.9734340897890216e-05, "loss": 0.0772, "step": 59350 }, { "epoch": 1.38, "learning_rate": 1.973346084344418e-05, "loss": 0.2297, "step": 59355 }, { "epoch": 1.38, "learning_rate": 1.9732580788998146e-05, "loss": 0.2715, "step": 59360 }, { "epoch": 1.39, "learning_rate": 1.973170073455211e-05, "loss": 0.1587, "step": 59365 }, { "epoch": 1.39, "learning_rate": 1.9730820680106076e-05, "loss": 0.0187, "step": 59370 }, { "epoch": 1.39, "learning_rate": 1.972994062566004e-05, "loss": 0.0726, "step": 59375 }, { "epoch": 1.39, "learning_rate": 1.9729060571214006e-05, "loss": 0.0558, "step": 59380 }, { "epoch": 1.39, "learning_rate": 1.972818051676797e-05, "loss": 0.0787, "step": 59385 }, { "epoch": 1.39, "learning_rate": 1.9727300462321936e-05, "loss": 0.2466, "step": 59390 }, { "epoch": 1.39, "learning_rate": 1.97264204078759e-05, "loss": 0.2405, "step": 59395 }, { "epoch": 1.39, "learning_rate": 1.9725540353429866e-05, "loss": 0.1454, "step": 59400 }, { "epoch": 1.39, "learning_rate": 1.972466029898383e-05, "loss": 0.1635, "step": 59405 }, { "epoch": 1.39, "learning_rate": 1.9723780244537796e-05, "loss": 0.2432, "step": 59410 }, { "epoch": 1.39, "learning_rate": 1.972290019009176e-05, "loss": 0.2739, "step": 59415 }, { "epoch": 1.39, "learning_rate": 1.9722020135645723e-05, "loss": 0.0681, "step": 59420 }, { "epoch": 1.39, "learning_rate": 1.972114008119969e-05, "loss": 0.0243, "step": 59425 }, { "epoch": 1.39, "learning_rate": 1.9720260026753656e-05, "loss": 0.0175, "step": 59430 }, { "epoch": 1.39, "learning_rate": 1.971937997230762e-05, "loss": 0.2228, "step": 59435 }, { "epoch": 1.39, "learning_rate": 1.9718499917861583e-05, "loss": 0.077, "step": 59440 }, { "epoch": 1.39, "learning_rate": 1.971761986341555e-05, "loss": 0.0542, "step": 59445 }, { "epoch": 1.39, "learning_rate": 1.9716739808969516e-05, "loss": 0.0668, "step": 59450 }, { "epoch": 1.39, "learning_rate": 1.971585975452348e-05, "loss": 0.1336, "step": 59455 }, { "epoch": 1.39, "learning_rate": 1.9714979700077443e-05, "loss": 0.4405, "step": 59460 }, { "epoch": 1.39, "learning_rate": 1.971409964563141e-05, "loss": 0.2881, "step": 59465 }, { "epoch": 1.39, "learning_rate": 1.9713219591185377e-05, "loss": 0.0726, "step": 59470 }, { "epoch": 1.39, "learning_rate": 1.9712339536739338e-05, "loss": 0.1488, "step": 59475 }, { "epoch": 1.39, "learning_rate": 1.9711459482293303e-05, "loss": 0.0196, "step": 59480 }, { "epoch": 1.39, "learning_rate": 1.971057942784727e-05, "loss": 0.0496, "step": 59485 }, { "epoch": 1.39, "learning_rate": 1.9709699373401237e-05, "loss": 0.2566, "step": 59490 }, { "epoch": 1.39, "learning_rate": 1.9708819318955198e-05, "loss": 0.1223, "step": 59495 }, { "epoch": 1.39, "learning_rate": 1.9707939264509163e-05, "loss": 0.2973, "step": 59500 }, { "epoch": 1.39, "learning_rate": 1.9707059210063132e-05, "loss": 0.256, "step": 59505 }, { "epoch": 1.39, "learning_rate": 1.9706179155617093e-05, "loss": 0.3682, "step": 59510 }, { "epoch": 1.39, "learning_rate": 1.970529910117106e-05, "loss": 0.2393, "step": 59515 }, { "epoch": 1.39, "learning_rate": 1.9704419046725027e-05, "loss": 0.1372, "step": 59520 }, { "epoch": 1.39, "learning_rate": 1.9703538992278992e-05, "loss": 0.0431, "step": 59525 }, { "epoch": 1.39, "learning_rate": 1.9702658937832953e-05, "loss": 0.0753, "step": 59530 }, { "epoch": 1.39, "learning_rate": 1.970177888338692e-05, "loss": 0.0384, "step": 59535 }, { "epoch": 1.39, "learning_rate": 1.9700898828940887e-05, "loss": 0.1059, "step": 59540 }, { "epoch": 1.39, "learning_rate": 1.970001877449485e-05, "loss": 0.1126, "step": 59545 }, { "epoch": 1.39, "learning_rate": 1.9699138720048814e-05, "loss": 0.0753, "step": 59550 }, { "epoch": 1.39, "learning_rate": 1.969825866560278e-05, "loss": 0.2528, "step": 59555 }, { "epoch": 1.39, "learning_rate": 1.9697378611156747e-05, "loss": 0.3978, "step": 59560 }, { "epoch": 1.39, "learning_rate": 1.969649855671071e-05, "loss": 0.3945, "step": 59565 }, { "epoch": 1.39, "learning_rate": 1.9695618502264674e-05, "loss": 0.1378, "step": 59570 }, { "epoch": 1.39, "learning_rate": 1.969473844781864e-05, "loss": 0.1206, "step": 59575 }, { "epoch": 1.39, "learning_rate": 1.9693858393372604e-05, "loss": 0.111, "step": 59580 }, { "epoch": 1.39, "learning_rate": 1.969297833892657e-05, "loss": 0.1173, "step": 59585 }, { "epoch": 1.39, "learning_rate": 1.9692098284480534e-05, "loss": 0.1919, "step": 59590 }, { "epoch": 1.39, "learning_rate": 1.96912182300345e-05, "loss": 0.1223, "step": 59595 }, { "epoch": 1.39, "learning_rate": 1.9690338175588464e-05, "loss": 0.2176, "step": 59600 }, { "epoch": 1.39, "learning_rate": 1.968945812114243e-05, "loss": 0.2281, "step": 59605 }, { "epoch": 1.39, "learning_rate": 1.9688578066696394e-05, "loss": 0.282, "step": 59610 }, { "epoch": 1.39, "learning_rate": 1.9687698012250355e-05, "loss": 0.2247, "step": 59615 }, { "epoch": 1.39, "learning_rate": 1.9686817957804324e-05, "loss": 0.0355, "step": 59620 }, { "epoch": 1.39, "learning_rate": 1.968593790335829e-05, "loss": 0.0621, "step": 59625 }, { "epoch": 1.39, "learning_rate": 1.9685057848912254e-05, "loss": 0.058, "step": 59630 }, { "epoch": 1.39, "learning_rate": 1.9684177794466215e-05, "loss": 0.172, "step": 59635 }, { "epoch": 1.39, "learning_rate": 1.9683297740020184e-05, "loss": 0.2535, "step": 59640 }, { "epoch": 1.39, "learning_rate": 1.968241768557415e-05, "loss": 0.1157, "step": 59645 }, { "epoch": 1.39, "learning_rate": 1.968153763112811e-05, "loss": 0.112, "step": 59650 }, { "epoch": 1.39, "learning_rate": 1.968065757668208e-05, "loss": 0.1473, "step": 59655 }, { "epoch": 1.39, "learning_rate": 1.9679777522236044e-05, "loss": 0.3652, "step": 59660 }, { "epoch": 1.39, "learning_rate": 1.967889746779001e-05, "loss": 0.532, "step": 59665 }, { "epoch": 1.39, "learning_rate": 1.967801741334397e-05, "loss": 0.0732, "step": 59670 }, { "epoch": 1.39, "learning_rate": 1.967713735889794e-05, "loss": 0.0552, "step": 59675 }, { "epoch": 1.39, "learning_rate": 1.9676257304451904e-05, "loss": 0.0455, "step": 59680 }, { "epoch": 1.39, "learning_rate": 1.9675377250005866e-05, "loss": 0.1512, "step": 59685 }, { "epoch": 1.39, "learning_rate": 1.967449719555983e-05, "loss": 0.1482, "step": 59690 }, { "epoch": 1.39, "learning_rate": 1.96736171411138e-05, "loss": 0.1142, "step": 59695 }, { "epoch": 1.39, "learning_rate": 1.9672737086667764e-05, "loss": 0.0786, "step": 59700 }, { "epoch": 1.39, "learning_rate": 1.9671857032221726e-05, "loss": 0.281, "step": 59705 }, { "epoch": 1.39, "learning_rate": 1.967097697777569e-05, "loss": 0.2438, "step": 59710 }, { "epoch": 1.39, "learning_rate": 1.967009692332966e-05, "loss": 0.3111, "step": 59715 }, { "epoch": 1.39, "learning_rate": 1.966921686888362e-05, "loss": 0.0815, "step": 59720 }, { "epoch": 1.39, "learning_rate": 1.9668336814437586e-05, "loss": 0.1729, "step": 59725 }, { "epoch": 1.39, "learning_rate": 1.966745675999155e-05, "loss": 0.0614, "step": 59730 }, { "epoch": 1.39, "learning_rate": 1.966657670554552e-05, "loss": 0.1458, "step": 59735 }, { "epoch": 1.39, "learning_rate": 1.966569665109948e-05, "loss": 0.0832, "step": 59740 }, { "epoch": 1.39, "learning_rate": 1.9664816596653446e-05, "loss": 0.0982, "step": 59745 }, { "epoch": 1.39, "learning_rate": 1.966393654220741e-05, "loss": 0.12, "step": 59750 }, { "epoch": 1.39, "learning_rate": 1.966305648776138e-05, "loss": 0.186, "step": 59755 }, { "epoch": 1.39, "learning_rate": 1.966217643331534e-05, "loss": 0.3326, "step": 59760 }, { "epoch": 1.39, "learning_rate": 1.9661296378869306e-05, "loss": 0.3814, "step": 59765 }, { "epoch": 1.39, "learning_rate": 1.966041632442327e-05, "loss": 0.0486, "step": 59770 }, { "epoch": 1.39, "learning_rate": 1.9659536269977236e-05, "loss": 0.0213, "step": 59775 }, { "epoch": 1.39, "learning_rate": 1.96586562155312e-05, "loss": 0.0457, "step": 59780 }, { "epoch": 1.39, "learning_rate": 1.9657776161085166e-05, "loss": 0.1428, "step": 59785 }, { "epoch": 1.39, "learning_rate": 1.9656896106639134e-05, "loss": 0.0761, "step": 59790 }, { "epoch": 1.4, "learning_rate": 1.9656016052193096e-05, "loss": 0.072, "step": 59795 }, { "epoch": 1.4, "learning_rate": 1.965513599774706e-05, "loss": 0.2272, "step": 59800 }, { "epoch": 1.4, "learning_rate": 1.9654255943301026e-05, "loss": 0.2341, "step": 59805 }, { "epoch": 1.4, "learning_rate": 1.965337588885499e-05, "loss": 0.3762, "step": 59810 }, { "epoch": 1.4, "learning_rate": 1.9652495834408956e-05, "loss": 0.2446, "step": 59815 }, { "epoch": 1.4, "learning_rate": 1.965161577996292e-05, "loss": 0.0475, "step": 59820 }, { "epoch": 1.4, "learning_rate": 1.9650735725516886e-05, "loss": 0.0438, "step": 59825 }, { "epoch": 1.4, "learning_rate": 1.964985567107085e-05, "loss": 0.1611, "step": 59830 }, { "epoch": 1.4, "learning_rate": 1.9648975616624816e-05, "loss": 0.1041, "step": 59835 }, { "epoch": 1.4, "learning_rate": 1.964809556217878e-05, "loss": 0.0093, "step": 59840 }, { "epoch": 1.4, "learning_rate": 1.9647215507732743e-05, "loss": 0.0761, "step": 59845 }, { "epoch": 1.4, "learning_rate": 1.964633545328671e-05, "loss": 0.2324, "step": 59850 }, { "epoch": 1.4, "learning_rate": 1.9645455398840676e-05, "loss": 0.3506, "step": 59855 }, { "epoch": 1.4, "learning_rate": 1.964457534439464e-05, "loss": 0.1709, "step": 59860 }, { "epoch": 1.4, "learning_rate": 1.9643695289948603e-05, "loss": 0.2122, "step": 59865 }, { "epoch": 1.4, "learning_rate": 1.964281523550257e-05, "loss": 0.0673, "step": 59870 }, { "epoch": 1.4, "learning_rate": 1.9641935181056536e-05, "loss": 0.1049, "step": 59875 }, { "epoch": 1.4, "learning_rate": 1.9641055126610498e-05, "loss": 0.0697, "step": 59880 }, { "epoch": 1.4, "learning_rate": 1.9640175072164463e-05, "loss": 0.1193, "step": 59885 }, { "epoch": 1.4, "learning_rate": 1.963929501771843e-05, "loss": 0.0615, "step": 59890 }, { "epoch": 1.4, "learning_rate": 1.9638414963272397e-05, "loss": 0.1375, "step": 59895 }, { "epoch": 1.4, "learning_rate": 1.9637534908826358e-05, "loss": 0.1268, "step": 59900 }, { "epoch": 1.4, "learning_rate": 1.9636654854380323e-05, "loss": 0.0749, "step": 59905 }, { "epoch": 1.4, "learning_rate": 1.963577479993429e-05, "loss": 0.268, "step": 59910 }, { "epoch": 1.4, "learning_rate": 1.9634894745488253e-05, "loss": 0.2212, "step": 59915 }, { "epoch": 1.4, "learning_rate": 1.9634014691042218e-05, "loss": 0.0424, "step": 59920 }, { "epoch": 1.4, "learning_rate": 1.9633134636596187e-05, "loss": 0.0358, "step": 59925 }, { "epoch": 1.4, "learning_rate": 1.963225458215015e-05, "loss": 0.0415, "step": 59930 }, { "epoch": 1.4, "learning_rate": 1.9631374527704113e-05, "loss": 0.1139, "step": 59935 }, { "epoch": 1.4, "learning_rate": 1.963049447325808e-05, "loss": 0.0519, "step": 59940 }, { "epoch": 1.4, "learning_rate": 1.9629614418812047e-05, "loss": 0.0568, "step": 59945 }, { "epoch": 1.4, "learning_rate": 1.962873436436601e-05, "loss": 0.277, "step": 59950 }, { "epoch": 1.4, "learning_rate": 1.9627854309919973e-05, "loss": 0.118, "step": 59955 }, { "epoch": 1.4, "learning_rate": 1.962697425547394e-05, "loss": 0.4925, "step": 59960 }, { "epoch": 1.4, "learning_rate": 1.9626094201027907e-05, "loss": 0.3287, "step": 59965 }, { "epoch": 1.4, "learning_rate": 1.962521414658187e-05, "loss": 0.0784, "step": 59970 }, { "epoch": 1.4, "learning_rate": 1.9624334092135833e-05, "loss": 0.0598, "step": 59975 }, { "epoch": 1.4, "learning_rate": 1.96234540376898e-05, "loss": 0.1003, "step": 59980 }, { "epoch": 1.4, "learning_rate": 1.9622573983243764e-05, "loss": 0.0699, "step": 59985 }, { "epoch": 1.4, "learning_rate": 1.962169392879773e-05, "loss": 0.1412, "step": 59990 }, { "epoch": 1.4, "learning_rate": 1.9620813874351694e-05, "loss": 0.125, "step": 59995 }, { "epoch": 1.4, "learning_rate": 1.961993381990566e-05, "loss": 0.1424, "step": 60000 }, { "epoch": 1.4, "learning_rate": 1.9619053765459624e-05, "loss": 0.2168, "step": 60005 }, { "epoch": 1.4, "learning_rate": 1.961817371101359e-05, "loss": 0.3739, "step": 60010 }, { "epoch": 1.4, "learning_rate": 1.9617293656567554e-05, "loss": 0.2545, "step": 60015 }, { "epoch": 1.4, "learning_rate": 1.9616413602121515e-05, "loss": 0.0897, "step": 60020 }, { "epoch": 1.4, "learning_rate": 1.9615533547675484e-05, "loss": 0.0844, "step": 60025 }, { "epoch": 1.4, "learning_rate": 1.961465349322945e-05, "loss": 0.0791, "step": 60030 }, { "epoch": 1.4, "learning_rate": 1.9613773438783414e-05, "loss": 0.0358, "step": 60035 }, { "epoch": 1.4, "learning_rate": 1.9612893384337375e-05, "loss": 0.0559, "step": 60040 }, { "epoch": 1.4, "learning_rate": 1.9612013329891344e-05, "loss": 0.1559, "step": 60045 }, { "epoch": 1.4, "learning_rate": 1.961113327544531e-05, "loss": 0.1303, "step": 60050 }, { "epoch": 1.4, "learning_rate": 1.9610253220999274e-05, "loss": 0.142, "step": 60055 }, { "epoch": 1.4, "learning_rate": 1.960937316655324e-05, "loss": 0.3626, "step": 60060 }, { "epoch": 1.4, "learning_rate": 1.9608493112107204e-05, "loss": 0.2227, "step": 60065 }, { "epoch": 1.4, "learning_rate": 1.960761305766117e-05, "loss": 0.0258, "step": 60070 }, { "epoch": 1.4, "learning_rate": 1.960673300321513e-05, "loss": 0.0759, "step": 60075 }, { "epoch": 1.4, "learning_rate": 1.96058529487691e-05, "loss": 0.076, "step": 60080 }, { "epoch": 1.4, "learning_rate": 1.9604972894323064e-05, "loss": 0.0456, "step": 60085 }, { "epoch": 1.4, "learning_rate": 1.960409283987703e-05, "loss": 0.1255, "step": 60090 }, { "epoch": 1.4, "learning_rate": 1.960321278543099e-05, "loss": 0.1273, "step": 60095 }, { "epoch": 1.4, "learning_rate": 1.960233273098496e-05, "loss": 0.1395, "step": 60100 }, { "epoch": 1.4, "learning_rate": 1.9601452676538924e-05, "loss": 0.2159, "step": 60105 }, { "epoch": 1.4, "learning_rate": 1.9600572622092886e-05, "loss": 0.2459, "step": 60110 }, { "epoch": 1.4, "learning_rate": 1.959969256764685e-05, "loss": 0.2237, "step": 60115 }, { "epoch": 1.4, "learning_rate": 1.959881251320082e-05, "loss": 0.0205, "step": 60120 }, { "epoch": 1.4, "learning_rate": 1.9597932458754784e-05, "loss": 0.0982, "step": 60125 }, { "epoch": 1.4, "learning_rate": 1.9597052404308746e-05, "loss": 0.1104, "step": 60130 }, { "epoch": 1.4, "learning_rate": 1.959617234986271e-05, "loss": 0.0477, "step": 60135 }, { "epoch": 1.4, "learning_rate": 1.959529229541668e-05, "loss": 0.1824, "step": 60140 }, { "epoch": 1.4, "learning_rate": 1.959441224097064e-05, "loss": 0.036, "step": 60145 }, { "epoch": 1.4, "learning_rate": 1.9593532186524606e-05, "loss": 0.3419, "step": 60150 }, { "epoch": 1.4, "learning_rate": 1.959265213207857e-05, "loss": 0.1479, "step": 60155 }, { "epoch": 1.4, "learning_rate": 1.959177207763254e-05, "loss": 0.3857, "step": 60160 }, { "epoch": 1.4, "learning_rate": 1.95908920231865e-05, "loss": 0.0896, "step": 60165 }, { "epoch": 1.4, "learning_rate": 1.9590011968740466e-05, "loss": 0.0845, "step": 60170 }, { "epoch": 1.4, "learning_rate": 1.958913191429443e-05, "loss": 0.0277, "step": 60175 }, { "epoch": 1.4, "learning_rate": 1.9588251859848396e-05, "loss": 0.1316, "step": 60180 }, { "epoch": 1.4, "learning_rate": 1.958737180540236e-05, "loss": 0.0637, "step": 60185 }, { "epoch": 1.4, "learning_rate": 1.9586491750956326e-05, "loss": 0.0453, "step": 60190 }, { "epoch": 1.4, "learning_rate": 1.9585611696510294e-05, "loss": 0.2313, "step": 60195 }, { "epoch": 1.4, "learning_rate": 1.9584731642064256e-05, "loss": 0.1597, "step": 60200 }, { "epoch": 1.4, "learning_rate": 1.958385158761822e-05, "loss": 0.1812, "step": 60205 }, { "epoch": 1.4, "learning_rate": 1.9582971533172186e-05, "loss": 0.3597, "step": 60210 }, { "epoch": 1.4, "learning_rate": 1.958209147872615e-05, "loss": 0.2704, "step": 60215 }, { "epoch": 1.41, "learning_rate": 1.9581211424280116e-05, "loss": 0.1413, "step": 60220 }, { "epoch": 1.41, "learning_rate": 1.958033136983408e-05, "loss": 0.002, "step": 60225 }, { "epoch": 1.41, "learning_rate": 1.9579451315388046e-05, "loss": 0.1378, "step": 60230 }, { "epoch": 1.41, "learning_rate": 1.957857126094201e-05, "loss": 0.0091, "step": 60235 }, { "epoch": 1.41, "learning_rate": 1.9577691206495976e-05, "loss": 0.1072, "step": 60240 }, { "epoch": 1.41, "learning_rate": 1.957681115204994e-05, "loss": 0.1415, "step": 60245 }, { "epoch": 1.41, "learning_rate": 1.9575931097603903e-05, "loss": 0.1296, "step": 60250 }, { "epoch": 1.41, "learning_rate": 1.957505104315787e-05, "loss": 0.2444, "step": 60255 }, { "epoch": 1.41, "learning_rate": 1.9574170988711836e-05, "loss": 0.3965, "step": 60260 }, { "epoch": 1.41, "learning_rate": 1.95732909342658e-05, "loss": 0.3604, "step": 60265 }, { "epoch": 1.41, "learning_rate": 1.9572410879819763e-05, "loss": 0.0883, "step": 60270 }, { "epoch": 1.41, "learning_rate": 1.957153082537373e-05, "loss": 0.0876, "step": 60275 }, { "epoch": 1.41, "learning_rate": 1.9570650770927696e-05, "loss": 0.0146, "step": 60280 }, { "epoch": 1.41, "learning_rate": 1.9569770716481658e-05, "loss": 0.1925, "step": 60285 }, { "epoch": 1.41, "learning_rate": 1.9568890662035623e-05, "loss": 0.1078, "step": 60290 }, { "epoch": 1.41, "learning_rate": 1.956801060758959e-05, "loss": 0.1005, "step": 60295 }, { "epoch": 1.41, "learning_rate": 1.9567130553143556e-05, "loss": 0.289, "step": 60300 }, { "epoch": 1.41, "learning_rate": 1.9566250498697518e-05, "loss": 0.306, "step": 60305 }, { "epoch": 1.41, "learning_rate": 1.9565370444251483e-05, "loss": 0.3671, "step": 60310 }, { "epoch": 1.41, "learning_rate": 1.956449038980545e-05, "loss": 0.3358, "step": 60315 }, { "epoch": 1.41, "learning_rate": 1.9563610335359413e-05, "loss": 0.0582, "step": 60320 }, { "epoch": 1.41, "learning_rate": 1.9562730280913378e-05, "loss": 0.0325, "step": 60325 }, { "epoch": 1.41, "learning_rate": 1.9561850226467347e-05, "loss": 0.0757, "step": 60330 }, { "epoch": 1.41, "learning_rate": 1.956097017202131e-05, "loss": 0.1238, "step": 60335 }, { "epoch": 1.41, "learning_rate": 1.9560090117575273e-05, "loss": 0.1032, "step": 60340 }, { "epoch": 1.41, "learning_rate": 1.9559210063129238e-05, "loss": 0.0776, "step": 60345 }, { "epoch": 1.41, "learning_rate": 1.9558330008683207e-05, "loss": 0.1238, "step": 60350 }, { "epoch": 1.41, "learning_rate": 1.9557449954237168e-05, "loss": 0.2599, "step": 60355 }, { "epoch": 1.41, "learning_rate": 1.9556569899791133e-05, "loss": 0.1411, "step": 60360 }, { "epoch": 1.41, "learning_rate": 1.9555689845345098e-05, "loss": 0.3755, "step": 60365 }, { "epoch": 1.41, "learning_rate": 1.9554809790899067e-05, "loss": 0.0433, "step": 60370 }, { "epoch": 1.41, "learning_rate": 1.955392973645303e-05, "loss": 0.0465, "step": 60375 }, { "epoch": 1.41, "learning_rate": 1.9553049682006993e-05, "loss": 0.1824, "step": 60380 }, { "epoch": 1.41, "learning_rate": 1.955216962756096e-05, "loss": 0.0876, "step": 60385 }, { "epoch": 1.41, "learning_rate": 1.9551289573114927e-05, "loss": 0.0662, "step": 60390 }, { "epoch": 1.41, "learning_rate": 1.955040951866889e-05, "loss": 0.1127, "step": 60395 }, { "epoch": 1.41, "learning_rate": 1.9549529464222853e-05, "loss": 0.1619, "step": 60400 }, { "epoch": 1.41, "learning_rate": 1.954864940977682e-05, "loss": 0.271, "step": 60405 }, { "epoch": 1.41, "learning_rate": 1.9547769355330783e-05, "loss": 0.255, "step": 60410 }, { "epoch": 1.41, "learning_rate": 1.954688930088475e-05, "loss": 0.2254, "step": 60415 }, { "epoch": 1.41, "learning_rate": 1.9546009246438714e-05, "loss": 0.1031, "step": 60420 }, { "epoch": 1.41, "learning_rate": 1.954512919199268e-05, "loss": 0.0372, "step": 60425 }, { "epoch": 1.41, "learning_rate": 1.9544249137546644e-05, "loss": 0.0993, "step": 60430 }, { "epoch": 1.41, "learning_rate": 1.954336908310061e-05, "loss": 0.0777, "step": 60435 }, { "epoch": 1.41, "learning_rate": 1.9542489028654574e-05, "loss": 0.1158, "step": 60440 }, { "epoch": 1.41, "learning_rate": 1.9541608974208535e-05, "loss": 0.1155, "step": 60445 }, { "epoch": 1.41, "learning_rate": 1.9540728919762504e-05, "loss": 0.1882, "step": 60450 }, { "epoch": 1.41, "learning_rate": 1.953984886531647e-05, "loss": 0.4355, "step": 60455 }, { "epoch": 1.41, "learning_rate": 1.9538968810870434e-05, "loss": 0.4282, "step": 60460 }, { "epoch": 1.41, "learning_rate": 1.95380887564244e-05, "loss": 0.2553, "step": 60465 }, { "epoch": 1.41, "learning_rate": 1.9537208701978364e-05, "loss": 0.0458, "step": 60470 }, { "epoch": 1.41, "learning_rate": 1.953632864753233e-05, "loss": 0.0554, "step": 60475 }, { "epoch": 1.41, "learning_rate": 1.953544859308629e-05, "loss": 0.0786, "step": 60480 }, { "epoch": 1.41, "learning_rate": 1.953456853864026e-05, "loss": 0.0628, "step": 60485 }, { "epoch": 1.41, "learning_rate": 1.9533688484194224e-05, "loss": 0.1696, "step": 60490 }, { "epoch": 1.41, "learning_rate": 1.953280842974819e-05, "loss": 0.1511, "step": 60495 }, { "epoch": 1.41, "learning_rate": 1.953192837530215e-05, "loss": 0.2991, "step": 60500 }, { "epoch": 1.41, "learning_rate": 1.953104832085612e-05, "loss": 0.2715, "step": 60505 }, { "epoch": 1.41, "learning_rate": 1.9530168266410084e-05, "loss": 0.3368, "step": 60510 }, { "epoch": 1.41, "learning_rate": 1.9529288211964046e-05, "loss": 0.3149, "step": 60515 }, { "epoch": 1.41, "learning_rate": 1.952840815751801e-05, "loss": 0.0401, "step": 60520 }, { "epoch": 1.41, "learning_rate": 1.952752810307198e-05, "loss": 0.0882, "step": 60525 }, { "epoch": 1.41, "learning_rate": 1.9526648048625944e-05, "loss": 0.1025, "step": 60530 }, { "epoch": 1.41, "learning_rate": 1.9525767994179906e-05, "loss": 0.0661, "step": 60535 }, { "epoch": 1.41, "learning_rate": 1.952488793973387e-05, "loss": 0.0286, "step": 60540 }, { "epoch": 1.41, "learning_rate": 1.952400788528784e-05, "loss": 0.2564, "step": 60545 }, { "epoch": 1.41, "learning_rate": 1.95231278308418e-05, "loss": 0.3641, "step": 60550 }, { "epoch": 1.41, "learning_rate": 1.9522247776395766e-05, "loss": 0.3366, "step": 60555 }, { "epoch": 1.41, "learning_rate": 1.952136772194973e-05, "loss": 0.2725, "step": 60560 }, { "epoch": 1.41, "learning_rate": 1.95204876675037e-05, "loss": 0.3133, "step": 60565 }, { "epoch": 1.41, "learning_rate": 1.951960761305766e-05, "loss": 0.0786, "step": 60570 }, { "epoch": 1.41, "learning_rate": 1.9518727558611626e-05, "loss": 0.0905, "step": 60575 }, { "epoch": 1.41, "learning_rate": 1.951784750416559e-05, "loss": 0.104, "step": 60580 }, { "epoch": 1.41, "learning_rate": 1.9516967449719556e-05, "loss": 0.1136, "step": 60585 }, { "epoch": 1.41, "learning_rate": 1.951608739527352e-05, "loss": 0.0815, "step": 60590 }, { "epoch": 1.41, "learning_rate": 1.9515207340827486e-05, "loss": 0.171, "step": 60595 }, { "epoch": 1.41, "learning_rate": 1.9514327286381454e-05, "loss": 0.0946, "step": 60600 }, { "epoch": 1.41, "learning_rate": 1.9513447231935416e-05, "loss": 0.1658, "step": 60605 }, { "epoch": 1.41, "learning_rate": 1.951256717748938e-05, "loss": 0.3271, "step": 60610 }, { "epoch": 1.41, "learning_rate": 1.9511687123043346e-05, "loss": 0.271, "step": 60615 }, { "epoch": 1.41, "learning_rate": 1.951080706859731e-05, "loss": 0.11, "step": 60620 }, { "epoch": 1.41, "learning_rate": 1.9509927014151276e-05, "loss": 0.083, "step": 60625 }, { "epoch": 1.41, "learning_rate": 1.950904695970524e-05, "loss": 0.1148, "step": 60630 }, { "epoch": 1.41, "learning_rate": 1.9508166905259206e-05, "loss": 0.0548, "step": 60635 }, { "epoch": 1.41, "learning_rate": 1.950728685081317e-05, "loss": 0.0295, "step": 60640 }, { "epoch": 1.41, "learning_rate": 1.9506406796367136e-05, "loss": 0.0759, "step": 60645 }, { "epoch": 1.42, "learning_rate": 1.95055267419211e-05, "loss": 0.287, "step": 60650 }, { "epoch": 1.42, "learning_rate": 1.9504646687475063e-05, "loss": 0.2709, "step": 60655 }, { "epoch": 1.42, "learning_rate": 1.950376663302903e-05, "loss": 0.3944, "step": 60660 }, { "epoch": 1.42, "learning_rate": 1.9502886578582996e-05, "loss": 0.2364, "step": 60665 }, { "epoch": 1.42, "learning_rate": 1.950200652413696e-05, "loss": 0.0132, "step": 60670 }, { "epoch": 1.42, "learning_rate": 1.9501126469690923e-05, "loss": 0.0169, "step": 60675 }, { "epoch": 1.42, "learning_rate": 1.950024641524489e-05, "loss": 0.0907, "step": 60680 }, { "epoch": 1.42, "learning_rate": 1.9499366360798856e-05, "loss": 0.1694, "step": 60685 }, { "epoch": 1.42, "learning_rate": 1.949848630635282e-05, "loss": 0.1802, "step": 60690 }, { "epoch": 1.42, "learning_rate": 1.9497606251906783e-05, "loss": 0.0791, "step": 60695 }, { "epoch": 1.42, "learning_rate": 1.949672619746075e-05, "loss": 0.1311, "step": 60700 }, { "epoch": 1.42, "learning_rate": 1.9495846143014716e-05, "loss": 0.3006, "step": 60705 }, { "epoch": 1.42, "learning_rate": 1.9494966088568678e-05, "loss": 0.2766, "step": 60710 }, { "epoch": 1.42, "learning_rate": 1.9494086034122643e-05, "loss": 0.1773, "step": 60715 }, { "epoch": 1.42, "learning_rate": 1.949320597967661e-05, "loss": 0.0577, "step": 60720 }, { "epoch": 1.42, "learning_rate": 1.9492325925230576e-05, "loss": 0.062, "step": 60725 }, { "epoch": 1.42, "learning_rate": 1.9491445870784538e-05, "loss": 0.0328, "step": 60730 }, { "epoch": 1.42, "learning_rate": 1.9490565816338506e-05, "loss": 0.1561, "step": 60735 }, { "epoch": 1.42, "learning_rate": 1.948968576189247e-05, "loss": 0.0505, "step": 60740 }, { "epoch": 1.42, "learning_rate": 1.9488805707446433e-05, "loss": 0.0692, "step": 60745 }, { "epoch": 1.42, "learning_rate": 1.9487925653000398e-05, "loss": 0.1099, "step": 60750 }, { "epoch": 1.42, "learning_rate": 1.9487045598554367e-05, "loss": 0.2466, "step": 60755 }, { "epoch": 1.42, "learning_rate": 1.948616554410833e-05, "loss": 0.531, "step": 60760 }, { "epoch": 1.42, "learning_rate": 1.9485285489662293e-05, "loss": 0.3125, "step": 60765 }, { "epoch": 1.42, "learning_rate": 1.9484405435216258e-05, "loss": 0.0299, "step": 60770 }, { "epoch": 1.42, "learning_rate": 1.9483525380770227e-05, "loss": 0.0348, "step": 60775 }, { "epoch": 1.42, "learning_rate": 1.9482645326324188e-05, "loss": 0.0413, "step": 60780 }, { "epoch": 1.42, "learning_rate": 1.9481765271878153e-05, "loss": 0.0928, "step": 60785 }, { "epoch": 1.42, "learning_rate": 1.9480885217432118e-05, "loss": 0.1377, "step": 60790 }, { "epoch": 1.42, "learning_rate": 1.9480181173875293e-05, "loss": 0.1122, "step": 60795 }, { "epoch": 1.42, "learning_rate": 1.9479301119429255e-05, "loss": 0.2105, "step": 60800 }, { "epoch": 1.42, "learning_rate": 1.947842106498322e-05, "loss": 0.2111, "step": 60805 }, { "epoch": 1.42, "learning_rate": 1.9477541010537188e-05, "loss": 0.5169, "step": 60810 }, { "epoch": 1.42, "learning_rate": 1.947666095609115e-05, "loss": 0.208, "step": 60815 }, { "epoch": 1.42, "learning_rate": 1.9475780901645115e-05, "loss": 0.1308, "step": 60820 }, { "epoch": 1.42, "learning_rate": 1.947490084719908e-05, "loss": 0.0681, "step": 60825 }, { "epoch": 1.42, "learning_rate": 1.9474020792753048e-05, "loss": 0.0472, "step": 60830 }, { "epoch": 1.42, "learning_rate": 1.947314073830701e-05, "loss": 0.0202, "step": 60835 }, { "epoch": 1.42, "learning_rate": 1.9472260683860975e-05, "loss": 0.134, "step": 60840 }, { "epoch": 1.42, "learning_rate": 1.947138062941494e-05, "loss": 0.1309, "step": 60845 }, { "epoch": 1.42, "learning_rate": 1.9470500574968908e-05, "loss": 0.1057, "step": 60850 }, { "epoch": 1.42, "learning_rate": 1.946962052052287e-05, "loss": 0.1969, "step": 60855 }, { "epoch": 1.42, "learning_rate": 1.9468740466076835e-05, "loss": 0.344, "step": 60860 }, { "epoch": 1.42, "learning_rate": 1.94678604116308e-05, "loss": 0.2322, "step": 60865 }, { "epoch": 1.42, "learning_rate": 1.9466980357184765e-05, "loss": 0.0241, "step": 60870 }, { "epoch": 1.42, "learning_rate": 1.946610030273873e-05, "loss": 0.0365, "step": 60875 }, { "epoch": 1.42, "learning_rate": 1.9465220248292695e-05, "loss": 0.0335, "step": 60880 }, { "epoch": 1.42, "learning_rate": 1.946434019384666e-05, "loss": 0.1289, "step": 60885 }, { "epoch": 1.42, "learning_rate": 1.9463460139400625e-05, "loss": 0.1891, "step": 60890 }, { "epoch": 1.42, "learning_rate": 1.946258008495459e-05, "loss": 0.1793, "step": 60895 }, { "epoch": 1.42, "learning_rate": 1.9461700030508555e-05, "loss": 0.1742, "step": 60900 }, { "epoch": 1.42, "learning_rate": 1.9460819976062517e-05, "loss": 0.1861, "step": 60905 }, { "epoch": 1.42, "learning_rate": 1.9459939921616485e-05, "loss": 0.2819, "step": 60910 }, { "epoch": 1.42, "learning_rate": 1.945905986717045e-05, "loss": 0.3832, "step": 60915 }, { "epoch": 1.42, "learning_rate": 1.9458179812724415e-05, "loss": 0.0566, "step": 60920 }, { "epoch": 1.42, "learning_rate": 1.945729975827838e-05, "loss": 0.0101, "step": 60925 }, { "epoch": 1.42, "learning_rate": 1.9456419703832345e-05, "loss": 0.0314, "step": 60930 }, { "epoch": 1.42, "learning_rate": 1.945553964938631e-05, "loss": 0.0615, "step": 60935 }, { "epoch": 1.42, "learning_rate": 1.9454659594940272e-05, "loss": 0.092, "step": 60940 }, { "epoch": 1.42, "learning_rate": 1.945377954049424e-05, "loss": 0.1982, "step": 60945 }, { "epoch": 1.42, "learning_rate": 1.9452899486048205e-05, "loss": 0.1028, "step": 60950 }, { "epoch": 1.42, "learning_rate": 1.945201943160217e-05, "loss": 0.1601, "step": 60955 }, { "epoch": 1.42, "learning_rate": 1.9451139377156132e-05, "loss": 0.3096, "step": 60960 }, { "epoch": 1.42, "learning_rate": 1.94502593227101e-05, "loss": 0.2476, "step": 60965 }, { "epoch": 1.42, "learning_rate": 1.9449379268264065e-05, "loss": 0.0916, "step": 60970 }, { "epoch": 1.42, "learning_rate": 1.9448499213818027e-05, "loss": 0.0893, "step": 60975 }, { "epoch": 1.42, "learning_rate": 1.9447619159371992e-05, "loss": 0.1776, "step": 60980 }, { "epoch": 1.42, "learning_rate": 1.944673910492596e-05, "loss": 0.0913, "step": 60985 }, { "epoch": 1.42, "learning_rate": 1.9445859050479925e-05, "loss": 0.1281, "step": 60990 }, { "epoch": 1.42, "learning_rate": 1.9444978996033887e-05, "loss": 0.2433, "step": 60995 }, { "epoch": 1.42, "learning_rate": 1.9444098941587852e-05, "loss": 0.1811, "step": 61000 }, { "epoch": 1.42, "learning_rate": 1.944321888714182e-05, "loss": 0.1766, "step": 61005 }, { "epoch": 1.42, "learning_rate": 1.9442338832695782e-05, "loss": 0.4198, "step": 61010 }, { "epoch": 1.42, "learning_rate": 1.9441458778249747e-05, "loss": 0.287, "step": 61015 }, { "epoch": 1.42, "learning_rate": 1.9440578723803712e-05, "loss": 0.0672, "step": 61020 }, { "epoch": 1.42, "learning_rate": 1.943969866935768e-05, "loss": 0.062, "step": 61025 }, { "epoch": 1.42, "learning_rate": 1.9438818614911642e-05, "loss": 0.0468, "step": 61030 }, { "epoch": 1.42, "learning_rate": 1.9437938560465607e-05, "loss": 0.1261, "step": 61035 }, { "epoch": 1.42, "learning_rate": 1.9437058506019572e-05, "loss": 0.1105, "step": 61040 }, { "epoch": 1.42, "learning_rate": 1.9436178451573537e-05, "loss": 0.1196, "step": 61045 }, { "epoch": 1.42, "learning_rate": 1.9435298397127502e-05, "loss": 0.1954, "step": 61050 }, { "epoch": 1.42, "learning_rate": 1.9434418342681467e-05, "loss": 0.2, "step": 61055 }, { "epoch": 1.42, "learning_rate": 1.9433538288235432e-05, "loss": 0.2362, "step": 61060 }, { "epoch": 1.42, "learning_rate": 1.9432658233789397e-05, "loss": 0.2264, "step": 61065 }, { "epoch": 1.42, "learning_rate": 1.9431778179343362e-05, "loss": 0.0108, "step": 61070 }, { "epoch": 1.42, "learning_rate": 1.9430898124897327e-05, "loss": 0.0372, "step": 61075 }, { "epoch": 1.43, "learning_rate": 1.9430018070451292e-05, "loss": 0.0786, "step": 61080 }, { "epoch": 1.43, "learning_rate": 1.9429138016005257e-05, "loss": 0.1756, "step": 61085 }, { "epoch": 1.43, "learning_rate": 1.9428257961559222e-05, "loss": 0.1917, "step": 61090 }, { "epoch": 1.43, "learning_rate": 1.9427377907113187e-05, "loss": 0.1677, "step": 61095 }, { "epoch": 1.43, "learning_rate": 1.9426497852667152e-05, "loss": 0.1293, "step": 61100 }, { "epoch": 1.43, "learning_rate": 1.9425617798221117e-05, "loss": 0.1835, "step": 61105 }, { "epoch": 1.43, "learning_rate": 1.9424737743775083e-05, "loss": 0.2878, "step": 61110 }, { "epoch": 1.43, "learning_rate": 1.9423857689329044e-05, "loss": 0.2312, "step": 61115 }, { "epoch": 1.43, "learning_rate": 1.9422977634883013e-05, "loss": 0.0504, "step": 61120 }, { "epoch": 1.43, "learning_rate": 1.9422097580436978e-05, "loss": 0.088, "step": 61125 }, { "epoch": 1.43, "learning_rate": 1.9421217525990943e-05, "loss": 0.0928, "step": 61130 }, { "epoch": 1.43, "learning_rate": 1.9420337471544904e-05, "loss": 0.1696, "step": 61135 }, { "epoch": 1.43, "learning_rate": 1.9419457417098873e-05, "loss": 0.117, "step": 61140 }, { "epoch": 1.43, "learning_rate": 1.9418577362652838e-05, "loss": 0.1171, "step": 61145 }, { "epoch": 1.43, "learning_rate": 1.94176973082068e-05, "loss": 0.0663, "step": 61150 }, { "epoch": 1.43, "learning_rate": 1.9416817253760764e-05, "loss": 0.1336, "step": 61155 }, { "epoch": 1.43, "learning_rate": 1.9415937199314733e-05, "loss": 0.3631, "step": 61160 }, { "epoch": 1.43, "learning_rate": 1.9415057144868698e-05, "loss": 0.2267, "step": 61165 }, { "epoch": 1.43, "learning_rate": 1.941417709042266e-05, "loss": 0.0761, "step": 61170 }, { "epoch": 1.43, "learning_rate": 1.9413297035976624e-05, "loss": 0.1657, "step": 61175 }, { "epoch": 1.43, "learning_rate": 1.9412416981530593e-05, "loss": 0.1428, "step": 61180 }, { "epoch": 1.43, "learning_rate": 1.9411536927084558e-05, "loss": 0.1235, "step": 61185 }, { "epoch": 1.43, "learning_rate": 1.941065687263852e-05, "loss": 0.1797, "step": 61190 }, { "epoch": 1.43, "learning_rate": 1.9409776818192488e-05, "loss": 0.2492, "step": 61195 }, { "epoch": 1.43, "learning_rate": 1.9408896763746453e-05, "loss": 0.1498, "step": 61200 }, { "epoch": 1.43, "learning_rate": 1.9408016709300415e-05, "loss": 0.0527, "step": 61205 }, { "epoch": 1.43, "learning_rate": 1.940713665485438e-05, "loss": 0.595, "step": 61210 }, { "epoch": 1.43, "learning_rate": 1.9406256600408348e-05, "loss": 0.2982, "step": 61215 }, { "epoch": 1.43, "learning_rate": 1.9405376545962313e-05, "loss": 0.023, "step": 61220 }, { "epoch": 1.43, "learning_rate": 1.9404496491516275e-05, "loss": 0.0538, "step": 61225 }, { "epoch": 1.43, "learning_rate": 1.940361643707024e-05, "loss": 0.0436, "step": 61230 }, { "epoch": 1.43, "learning_rate": 1.9402736382624208e-05, "loss": 0.0634, "step": 61235 }, { "epoch": 1.43, "learning_rate": 1.940185632817817e-05, "loss": 0.0455, "step": 61240 }, { "epoch": 1.43, "learning_rate": 1.9400976273732135e-05, "loss": 0.0522, "step": 61245 }, { "epoch": 1.43, "learning_rate": 1.94000962192861e-05, "loss": 0.15, "step": 61250 }, { "epoch": 1.43, "learning_rate": 1.9399216164840068e-05, "loss": 0.1324, "step": 61255 }, { "epoch": 1.43, "learning_rate": 1.939833611039403e-05, "loss": 0.4567, "step": 61260 }, { "epoch": 1.43, "learning_rate": 1.9397456055947995e-05, "loss": 0.1605, "step": 61265 }, { "epoch": 1.43, "learning_rate": 1.939657600150196e-05, "loss": 0.0511, "step": 61270 }, { "epoch": 1.43, "learning_rate": 1.9395695947055925e-05, "loss": 0.0694, "step": 61275 }, { "epoch": 1.43, "learning_rate": 1.939481589260989e-05, "loss": 0.0475, "step": 61280 }, { "epoch": 1.43, "learning_rate": 1.9393935838163855e-05, "loss": 0.0595, "step": 61285 }, { "epoch": 1.43, "learning_rate": 1.939305578371782e-05, "loss": 0.118, "step": 61290 }, { "epoch": 1.43, "learning_rate": 1.9392175729271785e-05, "loss": 0.2227, "step": 61295 }, { "epoch": 1.43, "learning_rate": 1.939129567482575e-05, "loss": 0.1524, "step": 61300 }, { "epoch": 1.43, "learning_rate": 1.9390415620379715e-05, "loss": 0.3582, "step": 61305 }, { "epoch": 1.43, "learning_rate": 1.9389535565933677e-05, "loss": 0.4828, "step": 61310 }, { "epoch": 1.43, "learning_rate": 1.9388655511487645e-05, "loss": 0.226, "step": 61315 }, { "epoch": 1.43, "learning_rate": 1.938777545704161e-05, "loss": 0.0825, "step": 61320 }, { "epoch": 1.43, "learning_rate": 1.9386895402595575e-05, "loss": 0.0703, "step": 61325 }, { "epoch": 1.43, "learning_rate": 1.938601534814954e-05, "loss": 0.0932, "step": 61330 }, { "epoch": 1.43, "learning_rate": 1.9385135293703505e-05, "loss": 0.1273, "step": 61335 }, { "epoch": 1.43, "learning_rate": 1.938425523925747e-05, "loss": 0.0761, "step": 61340 }, { "epoch": 1.43, "learning_rate": 1.9383375184811432e-05, "loss": 0.1829, "step": 61345 }, { "epoch": 1.43, "learning_rate": 1.93824951303654e-05, "loss": 0.2064, "step": 61350 }, { "epoch": 1.43, "learning_rate": 1.9381615075919365e-05, "loss": 0.1922, "step": 61355 }, { "epoch": 1.43, "learning_rate": 1.938073502147333e-05, "loss": 0.2227, "step": 61360 }, { "epoch": 1.43, "learning_rate": 1.9379854967027292e-05, "loss": 0.2876, "step": 61365 }, { "epoch": 1.43, "learning_rate": 1.937897491258126e-05, "loss": 0.0354, "step": 61370 }, { "epoch": 1.43, "learning_rate": 1.9378094858135225e-05, "loss": 0.0535, "step": 61375 }, { "epoch": 1.43, "learning_rate": 1.9377214803689187e-05, "loss": 0.0585, "step": 61380 }, { "epoch": 1.43, "learning_rate": 1.9376334749243152e-05, "loss": 0.1118, "step": 61385 }, { "epoch": 1.43, "learning_rate": 1.937545469479712e-05, "loss": 0.0548, "step": 61390 }, { "epoch": 1.43, "learning_rate": 1.9374574640351085e-05, "loss": 0.0692, "step": 61395 }, { "epoch": 1.43, "learning_rate": 1.9373694585905047e-05, "loss": 0.1655, "step": 61400 }, { "epoch": 1.43, "learning_rate": 1.9372814531459012e-05, "loss": 0.3361, "step": 61405 }, { "epoch": 1.43, "learning_rate": 1.937193447701298e-05, "loss": 0.422, "step": 61410 }, { "epoch": 1.43, "learning_rate": 1.9371054422566942e-05, "loss": 0.1582, "step": 61415 }, { "epoch": 1.43, "learning_rate": 1.9370174368120907e-05, "loss": 0.0092, "step": 61420 }, { "epoch": 1.43, "learning_rate": 1.9369294313674872e-05, "loss": 0.0582, "step": 61425 }, { "epoch": 1.43, "learning_rate": 1.936841425922884e-05, "loss": 0.1014, "step": 61430 }, { "epoch": 1.43, "learning_rate": 1.9367534204782802e-05, "loss": 0.1063, "step": 61435 }, { "epoch": 1.43, "learning_rate": 1.9366654150336767e-05, "loss": 0.1998, "step": 61440 }, { "epoch": 1.43, "learning_rate": 1.9365774095890732e-05, "loss": 0.1164, "step": 61445 }, { "epoch": 1.43, "learning_rate": 1.9364894041444697e-05, "loss": 0.1257, "step": 61450 }, { "epoch": 1.43, "learning_rate": 1.9364013986998662e-05, "loss": 0.2517, "step": 61455 }, { "epoch": 1.43, "learning_rate": 1.9363133932552627e-05, "loss": 0.258, "step": 61460 }, { "epoch": 1.43, "learning_rate": 1.9362253878106592e-05, "loss": 0.2565, "step": 61465 }, { "epoch": 1.43, "learning_rate": 1.9361373823660557e-05, "loss": 0.0471, "step": 61470 }, { "epoch": 1.43, "learning_rate": 1.9360493769214522e-05, "loss": 0.069, "step": 61475 }, { "epoch": 1.43, "learning_rate": 1.9359613714768487e-05, "loss": 0.0257, "step": 61480 }, { "epoch": 1.43, "learning_rate": 1.9358733660322456e-05, "loss": 0.1471, "step": 61485 }, { "epoch": 1.43, "learning_rate": 1.9357853605876417e-05, "loss": 0.0762, "step": 61490 }, { "epoch": 1.43, "learning_rate": 1.9356973551430382e-05, "loss": 0.2292, "step": 61495 }, { "epoch": 1.43, "learning_rate": 1.9356093496984347e-05, "loss": 0.1778, "step": 61500 }, { "epoch": 1.43, "learning_rate": 1.9355213442538312e-05, "loss": 0.1322, "step": 61505 }, { "epoch": 1.44, "learning_rate": 1.9354333388092277e-05, "loss": 0.1562, "step": 61510 }, { "epoch": 1.44, "learning_rate": 1.9353453333646242e-05, "loss": 0.1436, "step": 61515 }, { "epoch": 1.44, "learning_rate": 1.9352573279200207e-05, "loss": 0.1196, "step": 61520 }, { "epoch": 1.44, "learning_rate": 1.9351693224754172e-05, "loss": 0.0759, "step": 61525 }, { "epoch": 1.44, "learning_rate": 1.9350813170308137e-05, "loss": 0.0855, "step": 61530 }, { "epoch": 1.44, "learning_rate": 1.9349933115862102e-05, "loss": 0.0712, "step": 61535 }, { "epoch": 1.44, "learning_rate": 1.9349053061416064e-05, "loss": 0.1734, "step": 61540 }, { "epoch": 1.44, "learning_rate": 1.9348173006970033e-05, "loss": 0.1238, "step": 61545 }, { "epoch": 1.44, "learning_rate": 1.9347292952523998e-05, "loss": 0.1878, "step": 61550 }, { "epoch": 1.44, "learning_rate": 1.9346412898077963e-05, "loss": 0.4095, "step": 61555 }, { "epoch": 1.44, "learning_rate": 1.9345532843631924e-05, "loss": 0.3759, "step": 61560 }, { "epoch": 1.44, "learning_rate": 1.9344652789185893e-05, "loss": 0.3679, "step": 61565 }, { "epoch": 1.44, "learning_rate": 1.9343772734739858e-05, "loss": 0.0578, "step": 61570 }, { "epoch": 1.44, "learning_rate": 1.934289268029382e-05, "loss": 0.1294, "step": 61575 }, { "epoch": 1.44, "learning_rate": 1.9342012625847784e-05, "loss": 0.2102, "step": 61580 }, { "epoch": 1.44, "learning_rate": 1.9341132571401753e-05, "loss": 0.0268, "step": 61585 }, { "epoch": 1.44, "learning_rate": 1.9340252516955718e-05, "loss": 0.0845, "step": 61590 }, { "epoch": 1.44, "learning_rate": 1.933937246250968e-05, "loss": 0.244, "step": 61595 }, { "epoch": 1.44, "learning_rate": 1.9338492408063644e-05, "loss": 0.2587, "step": 61600 }, { "epoch": 1.44, "learning_rate": 1.9337612353617613e-05, "loss": 0.2246, "step": 61605 }, { "epoch": 1.44, "learning_rate": 1.9336732299171574e-05, "loss": 0.2277, "step": 61610 }, { "epoch": 1.44, "learning_rate": 1.933585224472554e-05, "loss": 0.1919, "step": 61615 }, { "epoch": 1.44, "learning_rate": 1.9334972190279508e-05, "loss": 0.0725, "step": 61620 }, { "epoch": 1.44, "learning_rate": 1.9334092135833473e-05, "loss": 0.0298, "step": 61625 }, { "epoch": 1.44, "learning_rate": 1.9333212081387434e-05, "loss": 0.0253, "step": 61630 }, { "epoch": 1.44, "learning_rate": 1.93323320269414e-05, "loss": 0.0767, "step": 61635 }, { "epoch": 1.44, "learning_rate": 1.9331451972495368e-05, "loss": 0.082, "step": 61640 }, { "epoch": 1.44, "learning_rate": 1.933057191804933e-05, "loss": 0.3995, "step": 61645 }, { "epoch": 1.44, "learning_rate": 1.9329691863603295e-05, "loss": 0.2528, "step": 61650 }, { "epoch": 1.44, "learning_rate": 1.932881180915726e-05, "loss": 0.1682, "step": 61655 }, { "epoch": 1.44, "learning_rate": 1.9327931754711228e-05, "loss": 0.404, "step": 61660 }, { "epoch": 1.44, "learning_rate": 1.932705170026519e-05, "loss": 0.3252, "step": 61665 }, { "epoch": 1.44, "learning_rate": 1.9326171645819155e-05, "loss": 0.0403, "step": 61670 }, { "epoch": 1.44, "learning_rate": 1.932529159137312e-05, "loss": 0.0258, "step": 61675 }, { "epoch": 1.44, "learning_rate": 1.9324411536927085e-05, "loss": 0.1615, "step": 61680 }, { "epoch": 1.44, "learning_rate": 1.932353148248105e-05, "loss": 0.0963, "step": 61685 }, { "epoch": 1.44, "learning_rate": 1.9322651428035015e-05, "loss": 0.0564, "step": 61690 }, { "epoch": 1.44, "learning_rate": 1.932177137358898e-05, "loss": 0.0756, "step": 61695 }, { "epoch": 1.44, "learning_rate": 1.9320891319142945e-05, "loss": 0.0436, "step": 61700 }, { "epoch": 1.44, "learning_rate": 1.932001126469691e-05, "loss": 0.1379, "step": 61705 }, { "epoch": 1.44, "learning_rate": 1.9319131210250875e-05, "loss": 0.2779, "step": 61710 }, { "epoch": 1.44, "learning_rate": 1.9318251155804836e-05, "loss": 0.2591, "step": 61715 }, { "epoch": 1.44, "learning_rate": 1.9317371101358805e-05, "loss": 0.0423, "step": 61720 }, { "epoch": 1.44, "learning_rate": 1.931649104691277e-05, "loss": 0.0642, "step": 61725 }, { "epoch": 1.44, "learning_rate": 1.9315610992466735e-05, "loss": 0.1015, "step": 61730 }, { "epoch": 1.44, "learning_rate": 1.93147309380207e-05, "loss": 0.0836, "step": 61735 }, { "epoch": 1.44, "learning_rate": 1.9313850883574665e-05, "loss": 0.14, "step": 61740 }, { "epoch": 1.44, "learning_rate": 1.931297082912863e-05, "loss": 0.1062, "step": 61745 }, { "epoch": 1.44, "learning_rate": 1.931209077468259e-05, "loss": 0.0357, "step": 61750 }, { "epoch": 1.44, "learning_rate": 1.931121072023656e-05, "loss": 0.2475, "step": 61755 }, { "epoch": 1.44, "learning_rate": 1.9310330665790525e-05, "loss": 0.3362, "step": 61760 }, { "epoch": 1.44, "learning_rate": 1.930945061134449e-05, "loss": 0.3104, "step": 61765 }, { "epoch": 1.44, "learning_rate": 1.930857055689845e-05, "loss": 0.0183, "step": 61770 }, { "epoch": 1.44, "learning_rate": 1.930769050245242e-05, "loss": 0.0485, "step": 61775 }, { "epoch": 1.44, "learning_rate": 1.9306810448006385e-05, "loss": 0.0675, "step": 61780 }, { "epoch": 1.44, "learning_rate": 1.930593039356035e-05, "loss": 0.085, "step": 61785 }, { "epoch": 1.44, "learning_rate": 1.9305050339114312e-05, "loss": 0.0424, "step": 61790 }, { "epoch": 1.44, "learning_rate": 1.930417028466828e-05, "loss": 0.0584, "step": 61795 }, { "epoch": 1.44, "learning_rate": 1.9303290230222245e-05, "loss": 0.1511, "step": 61800 }, { "epoch": 1.44, "learning_rate": 1.9302410175776207e-05, "loss": 0.1289, "step": 61805 }, { "epoch": 1.44, "learning_rate": 1.9301530121330172e-05, "loss": 0.1353, "step": 61810 }, { "epoch": 1.44, "learning_rate": 1.930065006688414e-05, "loss": 0.2643, "step": 61815 }, { "epoch": 1.44, "learning_rate": 1.9299770012438105e-05, "loss": 0.0609, "step": 61820 }, { "epoch": 1.44, "learning_rate": 1.9298889957992067e-05, "loss": 0.0575, "step": 61825 }, { "epoch": 1.44, "learning_rate": 1.9298009903546032e-05, "loss": 0.127, "step": 61830 }, { "epoch": 1.44, "learning_rate": 1.92971298491e-05, "loss": 0.0297, "step": 61835 }, { "epoch": 1.44, "learning_rate": 1.9296249794653962e-05, "loss": 0.0942, "step": 61840 }, { "epoch": 1.44, "learning_rate": 1.9295369740207927e-05, "loss": 0.1859, "step": 61845 }, { "epoch": 1.44, "learning_rate": 1.9294489685761892e-05, "loss": 0.1659, "step": 61850 }, { "epoch": 1.44, "learning_rate": 1.929360963131586e-05, "loss": 0.2262, "step": 61855 }, { "epoch": 1.44, "learning_rate": 1.9292729576869822e-05, "loss": 0.3739, "step": 61860 }, { "epoch": 1.44, "learning_rate": 1.9291849522423787e-05, "loss": 0.2269, "step": 61865 }, { "epoch": 1.44, "learning_rate": 1.9290969467977752e-05, "loss": 0.1098, "step": 61870 }, { "epoch": 1.44, "learning_rate": 1.9290089413531717e-05, "loss": 0.0269, "step": 61875 }, { "epoch": 1.44, "learning_rate": 1.9289209359085682e-05, "loss": 0.0382, "step": 61880 }, { "epoch": 1.44, "learning_rate": 1.9288329304639647e-05, "loss": 0.0973, "step": 61885 }, { "epoch": 1.44, "learning_rate": 1.9287449250193616e-05, "loss": 0.1485, "step": 61890 }, { "epoch": 1.44, "learning_rate": 1.9286569195747577e-05, "loss": 0.1894, "step": 61895 }, { "epoch": 1.44, "learning_rate": 1.9285689141301542e-05, "loss": 0.1201, "step": 61900 }, { "epoch": 1.44, "learning_rate": 1.9284809086855507e-05, "loss": 0.2066, "step": 61905 }, { "epoch": 1.44, "learning_rate": 1.9283929032409472e-05, "loss": 0.1804, "step": 61910 }, { "epoch": 1.44, "learning_rate": 1.9283048977963437e-05, "loss": 0.206, "step": 61915 }, { "epoch": 1.44, "learning_rate": 1.9282168923517402e-05, "loss": 0.0347, "step": 61920 }, { "epoch": 1.44, "learning_rate": 1.9281288869071367e-05, "loss": 0.1224, "step": 61925 }, { "epoch": 1.44, "learning_rate": 1.9280408814625332e-05, "loss": 0.113, "step": 61930 }, { "epoch": 1.45, "learning_rate": 1.9279528760179297e-05, "loss": 0.1376, "step": 61935 }, { "epoch": 1.45, "learning_rate": 1.9278648705733262e-05, "loss": 0.0504, "step": 61940 }, { "epoch": 1.45, "learning_rate": 1.9277768651287224e-05, "loss": 0.0763, "step": 61945 }, { "epoch": 1.45, "learning_rate": 1.9276888596841192e-05, "loss": 0.105, "step": 61950 }, { "epoch": 1.45, "learning_rate": 1.9276008542395157e-05, "loss": 0.1619, "step": 61955 }, { "epoch": 1.45, "learning_rate": 1.9275128487949122e-05, "loss": 0.1342, "step": 61960 }, { "epoch": 1.45, "learning_rate": 1.9274248433503084e-05, "loss": 0.2543, "step": 61965 }, { "epoch": 1.45, "learning_rate": 1.9273368379057052e-05, "loss": 0.0482, "step": 61970 }, { "epoch": 1.45, "learning_rate": 1.9272488324611018e-05, "loss": 0.0435, "step": 61975 }, { "epoch": 1.45, "learning_rate": 1.927160827016498e-05, "loss": 0.0334, "step": 61980 }, { "epoch": 1.45, "learning_rate": 1.9270728215718944e-05, "loss": 0.1695, "step": 61985 }, { "epoch": 1.45, "learning_rate": 1.9269848161272913e-05, "loss": 0.1376, "step": 61990 }, { "epoch": 1.45, "learning_rate": 1.9268968106826878e-05, "loss": 0.2286, "step": 61995 }, { "epoch": 1.45, "learning_rate": 1.926808805238084e-05, "loss": 0.0619, "step": 62000 }, { "epoch": 1.45, "learning_rate": 1.9267207997934804e-05, "loss": 0.1274, "step": 62005 }, { "epoch": 1.45, "learning_rate": 1.9266327943488773e-05, "loss": 0.4424, "step": 62010 }, { "epoch": 1.45, "learning_rate": 1.9265447889042734e-05, "loss": 0.3332, "step": 62015 }, { "epoch": 1.45, "learning_rate": 1.92645678345967e-05, "loss": 0.1149, "step": 62020 }, { "epoch": 1.45, "learning_rate": 1.9263687780150668e-05, "loss": 0.0525, "step": 62025 }, { "epoch": 1.45, "learning_rate": 1.9262807725704633e-05, "loss": 0.1582, "step": 62030 }, { "epoch": 1.45, "learning_rate": 1.9261927671258594e-05, "loss": 0.0857, "step": 62035 }, { "epoch": 1.45, "learning_rate": 1.926104761681256e-05, "loss": 0.0425, "step": 62040 }, { "epoch": 1.45, "learning_rate": 1.9260167562366528e-05, "loss": 0.2187, "step": 62045 }, { "epoch": 1.45, "learning_rate": 1.925928750792049e-05, "loss": 0.1578, "step": 62050 }, { "epoch": 1.45, "learning_rate": 1.9258407453474454e-05, "loss": 0.0487, "step": 62055 }, { "epoch": 1.45, "learning_rate": 1.925752739902842e-05, "loss": 0.3918, "step": 62060 }, { "epoch": 1.45, "learning_rate": 1.9256647344582388e-05, "loss": 0.2035, "step": 62065 }, { "epoch": 1.45, "learning_rate": 1.925576729013635e-05, "loss": 0.0714, "step": 62070 }, { "epoch": 1.45, "learning_rate": 1.9254887235690315e-05, "loss": 0.0224, "step": 62075 }, { "epoch": 1.45, "learning_rate": 1.925400718124428e-05, "loss": 0.0737, "step": 62080 }, { "epoch": 1.45, "learning_rate": 1.9253127126798248e-05, "loss": 0.1021, "step": 62085 }, { "epoch": 1.45, "learning_rate": 1.925224707235221e-05, "loss": 0.1893, "step": 62090 }, { "epoch": 1.45, "learning_rate": 1.9251367017906175e-05, "loss": 0.1582, "step": 62095 }, { "epoch": 1.45, "learning_rate": 1.925048696346014e-05, "loss": 0.2058, "step": 62100 }, { "epoch": 1.45, "learning_rate": 1.9249606909014105e-05, "loss": 0.2645, "step": 62105 }, { "epoch": 1.45, "learning_rate": 1.924872685456807e-05, "loss": 0.4966, "step": 62110 }, { "epoch": 1.45, "learning_rate": 1.9247846800122035e-05, "loss": 0.1228, "step": 62115 }, { "epoch": 1.45, "learning_rate": 1.9246966745676e-05, "loss": 0.1029, "step": 62120 }, { "epoch": 1.45, "learning_rate": 1.9246086691229965e-05, "loss": 0.0701, "step": 62125 }, { "epoch": 1.45, "learning_rate": 1.924520663678393e-05, "loss": 0.1148, "step": 62130 }, { "epoch": 1.45, "learning_rate": 1.9244326582337895e-05, "loss": 0.0486, "step": 62135 }, { "epoch": 1.45, "learning_rate": 1.9243446527891856e-05, "loss": 0.0938, "step": 62140 }, { "epoch": 1.45, "learning_rate": 1.9242566473445825e-05, "loss": 0.1614, "step": 62145 }, { "epoch": 1.45, "learning_rate": 1.924168641899979e-05, "loss": 0.1568, "step": 62150 }, { "epoch": 1.45, "learning_rate": 1.9240806364553755e-05, "loss": 0.1529, "step": 62155 }, { "epoch": 1.45, "learning_rate": 1.923992631010772e-05, "loss": 0.3871, "step": 62160 }, { "epoch": 1.45, "learning_rate": 1.9239046255661685e-05, "loss": 0.296, "step": 62165 }, { "epoch": 1.45, "learning_rate": 1.923816620121565e-05, "loss": 0.1013, "step": 62170 }, { "epoch": 1.45, "learning_rate": 1.923728614676961e-05, "loss": 0.0358, "step": 62175 }, { "epoch": 1.45, "learning_rate": 1.923640609232358e-05, "loss": 0.0025, "step": 62180 }, { "epoch": 1.45, "learning_rate": 1.9235526037877545e-05, "loss": 0.1166, "step": 62185 }, { "epoch": 1.45, "learning_rate": 1.923464598343151e-05, "loss": 0.1429, "step": 62190 }, { "epoch": 1.45, "learning_rate": 1.923376592898547e-05, "loss": 0.2247, "step": 62195 }, { "epoch": 1.45, "learning_rate": 1.923288587453944e-05, "loss": 0.2055, "step": 62200 }, { "epoch": 1.45, "learning_rate": 1.9232005820093405e-05, "loss": 0.2446, "step": 62205 }, { "epoch": 1.45, "learning_rate": 1.9231125765647367e-05, "loss": 0.272, "step": 62210 }, { "epoch": 1.45, "learning_rate": 1.9230245711201332e-05, "loss": 0.2323, "step": 62215 }, { "epoch": 1.45, "learning_rate": 1.92293656567553e-05, "loss": 0.034, "step": 62220 }, { "epoch": 1.45, "learning_rate": 1.9228485602309265e-05, "loss": 0.0769, "step": 62225 }, { "epoch": 1.45, "learning_rate": 1.9227605547863227e-05, "loss": 0.0395, "step": 62230 }, { "epoch": 1.45, "learning_rate": 1.9226725493417192e-05, "loss": 0.2392, "step": 62235 }, { "epoch": 1.45, "learning_rate": 1.922584543897116e-05, "loss": 0.0593, "step": 62240 }, { "epoch": 1.45, "learning_rate": 1.9224965384525122e-05, "loss": 0.0803, "step": 62245 }, { "epoch": 1.45, "learning_rate": 1.9224085330079087e-05, "loss": 0.1642, "step": 62250 }, { "epoch": 1.45, "learning_rate": 1.9223205275633052e-05, "loss": 0.2085, "step": 62255 }, { "epoch": 1.45, "learning_rate": 1.922232522118702e-05, "loss": 0.3577, "step": 62260 }, { "epoch": 1.45, "learning_rate": 1.9221445166740982e-05, "loss": 0.4173, "step": 62265 }, { "epoch": 1.45, "learning_rate": 1.9220565112294947e-05, "loss": 0.0811, "step": 62270 }, { "epoch": 1.45, "learning_rate": 1.9219685057848912e-05, "loss": 0.105, "step": 62275 }, { "epoch": 1.45, "learning_rate": 1.9218805003402877e-05, "loss": 0.0607, "step": 62280 }, { "epoch": 1.45, "learning_rate": 1.9217924948956842e-05, "loss": 0.112, "step": 62285 }, { "epoch": 1.45, "learning_rate": 1.9217044894510807e-05, "loss": 0.1724, "step": 62290 }, { "epoch": 1.45, "learning_rate": 1.9216164840064775e-05, "loss": 0.1319, "step": 62295 }, { "epoch": 1.45, "learning_rate": 1.9215284785618737e-05, "loss": 0.1221, "step": 62300 }, { "epoch": 1.45, "learning_rate": 1.9214404731172702e-05, "loss": 0.1625, "step": 62305 }, { "epoch": 1.45, "learning_rate": 1.9213524676726667e-05, "loss": 0.3279, "step": 62310 }, { "epoch": 1.45, "learning_rate": 1.9212644622280632e-05, "loss": 0.2295, "step": 62315 }, { "epoch": 1.45, "learning_rate": 1.9211764567834597e-05, "loss": 0.0186, "step": 62320 }, { "epoch": 1.45, "learning_rate": 1.9210884513388562e-05, "loss": 0.0344, "step": 62325 }, { "epoch": 1.45, "learning_rate": 1.9210004458942527e-05, "loss": 0.1622, "step": 62330 }, { "epoch": 1.45, "learning_rate": 1.9209124404496492e-05, "loss": 0.1267, "step": 62335 }, { "epoch": 1.45, "learning_rate": 1.9208244350050457e-05, "loss": 0.294, "step": 62340 }, { "epoch": 1.45, "learning_rate": 1.9207364295604422e-05, "loss": 0.0253, "step": 62345 }, { "epoch": 1.45, "learning_rate": 1.9206484241158384e-05, "loss": 0.2169, "step": 62350 }, { "epoch": 1.45, "learning_rate": 1.9205604186712352e-05, "loss": 0.4201, "step": 62355 }, { "epoch": 1.45, "learning_rate": 1.9204724132266317e-05, "loss": 0.3749, "step": 62360 }, { "epoch": 1.46, "learning_rate": 1.9203844077820282e-05, "loss": 0.1996, "step": 62365 }, { "epoch": 1.46, "learning_rate": 1.9202964023374244e-05, "loss": 0.069, "step": 62370 }, { "epoch": 1.46, "learning_rate": 1.9202083968928212e-05, "loss": 0.0729, "step": 62375 }, { "epoch": 1.46, "learning_rate": 1.9201203914482177e-05, "loss": 0.1245, "step": 62380 }, { "epoch": 1.46, "learning_rate": 1.9200323860036142e-05, "loss": 0.0299, "step": 62385 }, { "epoch": 1.46, "learning_rate": 1.9199443805590104e-05, "loss": 0.1215, "step": 62390 }, { "epoch": 1.46, "learning_rate": 1.9198563751144072e-05, "loss": 0.1778, "step": 62395 }, { "epoch": 1.46, "learning_rate": 1.9197683696698037e-05, "loss": 0.0511, "step": 62400 }, { "epoch": 1.46, "learning_rate": 1.9196803642252e-05, "loss": 0.2385, "step": 62405 }, { "epoch": 1.46, "learning_rate": 1.9195923587805964e-05, "loss": 0.4338, "step": 62410 }, { "epoch": 1.46, "learning_rate": 1.9195043533359933e-05, "loss": 0.306, "step": 62415 }, { "epoch": 1.46, "learning_rate": 1.9194163478913898e-05, "loss": 0.037, "step": 62420 }, { "epoch": 1.46, "learning_rate": 1.919328342446786e-05, "loss": 0.1287, "step": 62425 }, { "epoch": 1.46, "learning_rate": 1.9192403370021828e-05, "loss": 0.0126, "step": 62430 }, { "epoch": 1.46, "learning_rate": 1.9191523315575793e-05, "loss": 0.0377, "step": 62435 }, { "epoch": 1.46, "learning_rate": 1.9190643261129754e-05, "loss": 0.0983, "step": 62440 }, { "epoch": 1.46, "learning_rate": 1.918976320668372e-05, "loss": 0.0904, "step": 62445 }, { "epoch": 1.46, "learning_rate": 1.9188883152237688e-05, "loss": 0.0848, "step": 62450 }, { "epoch": 1.46, "learning_rate": 1.9188003097791653e-05, "loss": 0.1328, "step": 62455 }, { "epoch": 1.46, "learning_rate": 1.9187123043345614e-05, "loss": 0.5606, "step": 62460 }, { "epoch": 1.46, "learning_rate": 1.918624298889958e-05, "loss": 0.217, "step": 62465 }, { "epoch": 1.46, "learning_rate": 1.9185362934453548e-05, "loss": 0.0474, "step": 62470 }, { "epoch": 1.46, "learning_rate": 1.918448288000751e-05, "loss": 0.1119, "step": 62475 }, { "epoch": 1.46, "learning_rate": 1.9183602825561474e-05, "loss": 0.0662, "step": 62480 }, { "epoch": 1.46, "learning_rate": 1.918272277111544e-05, "loss": 0.1405, "step": 62485 }, { "epoch": 1.46, "learning_rate": 1.9181842716669408e-05, "loss": 0.1277, "step": 62490 }, { "epoch": 1.46, "learning_rate": 1.918096266222337e-05, "loss": 0.0973, "step": 62495 }, { "epoch": 1.46, "learning_rate": 1.9180082607777334e-05, "loss": 0.1964, "step": 62500 }, { "epoch": 1.46, "learning_rate": 1.91792025533313e-05, "loss": 0.2573, "step": 62505 }, { "epoch": 1.46, "learning_rate": 1.9178322498885265e-05, "loss": 0.3525, "step": 62510 }, { "epoch": 1.46, "learning_rate": 1.917744244443923e-05, "loss": 0.2115, "step": 62515 }, { "epoch": 1.46, "learning_rate": 1.9176562389993195e-05, "loss": 0.0602, "step": 62520 }, { "epoch": 1.46, "learning_rate": 1.917568233554716e-05, "loss": 0.0715, "step": 62525 }, { "epoch": 1.46, "learning_rate": 1.9174802281101125e-05, "loss": 0.0783, "step": 62530 }, { "epoch": 1.46, "learning_rate": 1.917392222665509e-05, "loss": 0.1211, "step": 62535 }, { "epoch": 1.46, "learning_rate": 1.9173042172209055e-05, "loss": 0.1308, "step": 62540 }, { "epoch": 1.46, "learning_rate": 1.9172162117763016e-05, "loss": 0.1411, "step": 62545 }, { "epoch": 1.46, "learning_rate": 1.9171282063316985e-05, "loss": 0.3229, "step": 62550 }, { "epoch": 1.46, "learning_rate": 1.917040200887095e-05, "loss": 0.149, "step": 62555 }, { "epoch": 1.46, "learning_rate": 1.9169521954424915e-05, "loss": 0.2488, "step": 62560 }, { "epoch": 1.46, "learning_rate": 1.916864189997888e-05, "loss": 0.3178, "step": 62565 }, { "epoch": 1.46, "learning_rate": 1.9167761845532845e-05, "loss": 0.0131, "step": 62570 }, { "epoch": 1.46, "learning_rate": 1.916688179108681e-05, "loss": 0.0944, "step": 62575 }, { "epoch": 1.46, "learning_rate": 1.916600173664077e-05, "loss": 0.0762, "step": 62580 }, { "epoch": 1.46, "learning_rate": 1.916512168219474e-05, "loss": 0.0841, "step": 62585 }, { "epoch": 1.46, "learning_rate": 1.9164241627748705e-05, "loss": 0.0623, "step": 62590 }, { "epoch": 1.46, "learning_rate": 1.916336157330267e-05, "loss": 0.1292, "step": 62595 }, { "epoch": 1.46, "learning_rate": 1.916248151885663e-05, "loss": 0.1824, "step": 62600 }, { "epoch": 1.46, "learning_rate": 1.91616014644106e-05, "loss": 0.2815, "step": 62605 }, { "epoch": 1.46, "learning_rate": 1.9160721409964565e-05, "loss": 0.3156, "step": 62610 }, { "epoch": 1.46, "learning_rate": 1.9159841355518527e-05, "loss": 0.3581, "step": 62615 }, { "epoch": 1.46, "learning_rate": 1.915896130107249e-05, "loss": 0.0804, "step": 62620 }, { "epoch": 1.46, "learning_rate": 1.915808124662646e-05, "loss": 0.1696, "step": 62625 }, { "epoch": 1.46, "learning_rate": 1.9157201192180425e-05, "loss": 0.07, "step": 62630 }, { "epoch": 1.46, "learning_rate": 1.9156321137734387e-05, "loss": 0.0792, "step": 62635 }, { "epoch": 1.46, "learning_rate": 1.915544108328835e-05, "loss": 0.1358, "step": 62640 }, { "epoch": 1.46, "learning_rate": 1.915456102884232e-05, "loss": 0.1712, "step": 62645 }, { "epoch": 1.46, "learning_rate": 1.9153680974396282e-05, "loss": 0.2159, "step": 62650 }, { "epoch": 1.46, "learning_rate": 1.9152800919950247e-05, "loss": 0.1964, "step": 62655 }, { "epoch": 1.46, "learning_rate": 1.9151920865504212e-05, "loss": 0.4416, "step": 62660 }, { "epoch": 1.46, "learning_rate": 1.915104081105818e-05, "loss": 0.2327, "step": 62665 }, { "epoch": 1.46, "learning_rate": 1.9150160756612142e-05, "loss": 0.043, "step": 62670 }, { "epoch": 1.46, "learning_rate": 1.9149280702166107e-05, "loss": 0.0365, "step": 62675 }, { "epoch": 1.46, "learning_rate": 1.9148400647720072e-05, "loss": 0.0256, "step": 62680 }, { "epoch": 1.46, "learning_rate": 1.914752059327404e-05, "loss": 0.0642, "step": 62685 }, { "epoch": 1.46, "learning_rate": 1.9146640538828002e-05, "loss": 0.071, "step": 62690 }, { "epoch": 1.46, "learning_rate": 1.9145760484381967e-05, "loss": 0.0521, "step": 62695 }, { "epoch": 1.46, "learning_rate": 1.9144880429935935e-05, "loss": 0.1619, "step": 62700 }, { "epoch": 1.46, "learning_rate": 1.9144000375489897e-05, "loss": 0.3759, "step": 62705 }, { "epoch": 1.46, "learning_rate": 1.9143120321043862e-05, "loss": 0.2911, "step": 62710 }, { "epoch": 1.46, "learning_rate": 1.9142240266597827e-05, "loss": 0.1724, "step": 62715 }, { "epoch": 1.46, "learning_rate": 1.9141360212151795e-05, "loss": 0.0087, "step": 62720 }, { "epoch": 1.46, "learning_rate": 1.9140480157705757e-05, "loss": 0.0781, "step": 62725 }, { "epoch": 1.46, "learning_rate": 1.9139600103259722e-05, "loss": 0.1085, "step": 62730 }, { "epoch": 1.46, "learning_rate": 1.9138720048813687e-05, "loss": 0.1435, "step": 62735 }, { "epoch": 1.46, "learning_rate": 1.9137839994367652e-05, "loss": 0.2927, "step": 62740 }, { "epoch": 1.46, "learning_rate": 1.9136959939921617e-05, "loss": 0.0767, "step": 62745 }, { "epoch": 1.46, "learning_rate": 1.9136079885475582e-05, "loss": 0.2764, "step": 62750 }, { "epoch": 1.46, "learning_rate": 1.9135199831029547e-05, "loss": 0.407, "step": 62755 }, { "epoch": 1.46, "learning_rate": 1.9134319776583512e-05, "loss": 0.4347, "step": 62760 }, { "epoch": 1.46, "learning_rate": 1.9133439722137477e-05, "loss": 0.1816, "step": 62765 }, { "epoch": 1.46, "learning_rate": 1.9132559667691442e-05, "loss": 0.0412, "step": 62770 }, { "epoch": 1.46, "learning_rate": 1.9131679613245404e-05, "loss": 0.0916, "step": 62775 }, { "epoch": 1.46, "learning_rate": 1.9130799558799372e-05, "loss": 0.0206, "step": 62780 }, { "epoch": 1.46, "learning_rate": 1.9129919504353337e-05, "loss": 0.1383, "step": 62785 }, { "epoch": 1.46, "learning_rate": 1.9129039449907302e-05, "loss": 0.0335, "step": 62790 }, { "epoch": 1.47, "learning_rate": 1.9128159395461264e-05, "loss": 0.3135, "step": 62795 }, { "epoch": 1.47, "learning_rate": 1.9127279341015232e-05, "loss": 0.2516, "step": 62800 }, { "epoch": 1.47, "learning_rate": 1.9126399286569197e-05, "loss": 0.1931, "step": 62805 }, { "epoch": 1.47, "learning_rate": 1.912551923212316e-05, "loss": 0.229, "step": 62810 }, { "epoch": 1.47, "learning_rate": 1.9124639177677124e-05, "loss": 0.312, "step": 62815 }, { "epoch": 1.47, "learning_rate": 1.9123759123231092e-05, "loss": 0.0864, "step": 62820 }, { "epoch": 1.47, "learning_rate": 1.9122879068785057e-05, "loss": 0.0402, "step": 62825 }, { "epoch": 1.47, "learning_rate": 1.912199901433902e-05, "loss": 0.1499, "step": 62830 }, { "epoch": 1.47, "learning_rate": 1.9121118959892987e-05, "loss": 0.0629, "step": 62835 }, { "epoch": 1.47, "learning_rate": 1.9120238905446953e-05, "loss": 0.1559, "step": 62840 }, { "epoch": 1.47, "learning_rate": 1.9119358851000914e-05, "loss": 0.0803, "step": 62845 }, { "epoch": 1.47, "learning_rate": 1.911847879655488e-05, "loss": 0.1692, "step": 62850 }, { "epoch": 1.47, "learning_rate": 1.9117598742108848e-05, "loss": 0.4805, "step": 62855 }, { "epoch": 1.47, "learning_rate": 1.9116718687662813e-05, "loss": 0.3969, "step": 62860 }, { "epoch": 1.47, "learning_rate": 1.9115838633216774e-05, "loss": 0.2968, "step": 62865 }, { "epoch": 1.47, "learning_rate": 1.911495857877074e-05, "loss": 0.0297, "step": 62870 }, { "epoch": 1.47, "learning_rate": 1.9114078524324708e-05, "loss": 0.074, "step": 62875 }, { "epoch": 1.47, "learning_rate": 1.911319846987867e-05, "loss": 0.0614, "step": 62880 }, { "epoch": 1.47, "learning_rate": 1.9112318415432634e-05, "loss": 0.0791, "step": 62885 }, { "epoch": 1.47, "learning_rate": 1.91114383609866e-05, "loss": 0.066, "step": 62890 }, { "epoch": 1.47, "learning_rate": 1.9110558306540568e-05, "loss": 0.1729, "step": 62895 }, { "epoch": 1.47, "learning_rate": 1.910967825209453e-05, "loss": 0.121, "step": 62900 }, { "epoch": 1.47, "learning_rate": 1.9108798197648494e-05, "loss": 0.08, "step": 62905 }, { "epoch": 1.47, "learning_rate": 1.910791814320246e-05, "loss": 0.327, "step": 62910 }, { "epoch": 1.47, "learning_rate": 1.9107038088756424e-05, "loss": 0.451, "step": 62915 }, { "epoch": 1.47, "learning_rate": 1.910615803431039e-05, "loss": 0.0588, "step": 62920 }, { "epoch": 1.47, "learning_rate": 1.9105277979864354e-05, "loss": 0.058, "step": 62925 }, { "epoch": 1.47, "learning_rate": 1.910439792541832e-05, "loss": 0.082, "step": 62930 }, { "epoch": 1.47, "learning_rate": 1.9103517870972285e-05, "loss": 0.0871, "step": 62935 }, { "epoch": 1.47, "learning_rate": 1.910263781652625e-05, "loss": 0.1875, "step": 62940 }, { "epoch": 1.47, "learning_rate": 1.9101757762080215e-05, "loss": 0.0839, "step": 62945 }, { "epoch": 1.47, "learning_rate": 1.9100877707634176e-05, "loss": 0.0581, "step": 62950 }, { "epoch": 1.47, "learning_rate": 1.9099997653188145e-05, "loss": 0.2436, "step": 62955 }, { "epoch": 1.47, "learning_rate": 1.909911759874211e-05, "loss": 0.5173, "step": 62960 }, { "epoch": 1.47, "learning_rate": 1.9098237544296075e-05, "loss": 0.1813, "step": 62965 }, { "epoch": 1.47, "learning_rate": 1.909735748985004e-05, "loss": 0.0462, "step": 62970 }, { "epoch": 1.47, "learning_rate": 1.9096477435404005e-05, "loss": 0.0857, "step": 62975 }, { "epoch": 1.47, "learning_rate": 1.909559738095797e-05, "loss": 0.0388, "step": 62980 }, { "epoch": 1.47, "learning_rate": 1.909471732651193e-05, "loss": 0.0683, "step": 62985 }, { "epoch": 1.47, "learning_rate": 1.90938372720659e-05, "loss": 0.0263, "step": 62990 }, { "epoch": 1.47, "learning_rate": 1.9092957217619865e-05, "loss": 0.0854, "step": 62995 }, { "epoch": 1.47, "learning_rate": 1.909207716317383e-05, "loss": 0.3183, "step": 63000 }, { "epoch": 1.47, "learning_rate": 1.909119710872779e-05, "loss": 0.1147, "step": 63005 }, { "epoch": 1.47, "learning_rate": 1.909031705428176e-05, "loss": 0.5296, "step": 63010 }, { "epoch": 1.47, "learning_rate": 1.9089436999835725e-05, "loss": 0.2938, "step": 63015 }, { "epoch": 1.47, "learning_rate": 1.908855694538969e-05, "loss": 0.0931, "step": 63020 }, { "epoch": 1.47, "learning_rate": 1.908767689094365e-05, "loss": 0.1526, "step": 63025 }, { "epoch": 1.47, "learning_rate": 1.908679683649762e-05, "loss": 0.11, "step": 63030 }, { "epoch": 1.47, "learning_rate": 1.9085916782051585e-05, "loss": 0.0633, "step": 63035 }, { "epoch": 1.47, "learning_rate": 1.9085036727605547e-05, "loss": 0.1229, "step": 63040 }, { "epoch": 1.47, "learning_rate": 1.908415667315951e-05, "loss": 0.1634, "step": 63045 }, { "epoch": 1.47, "learning_rate": 1.908327661871348e-05, "loss": 0.1199, "step": 63050 }, { "epoch": 1.47, "learning_rate": 1.9082396564267445e-05, "loss": 0.0966, "step": 63055 }, { "epoch": 1.47, "learning_rate": 1.9081516509821407e-05, "loss": 0.5206, "step": 63060 }, { "epoch": 1.47, "learning_rate": 1.908063645537537e-05, "loss": 0.287, "step": 63065 }, { "epoch": 1.47, "learning_rate": 1.907975640092934e-05, "loss": 0.0137, "step": 63070 }, { "epoch": 1.47, "learning_rate": 1.9078876346483302e-05, "loss": 0.077, "step": 63075 }, { "epoch": 1.47, "learning_rate": 1.9077996292037267e-05, "loss": 0.0619, "step": 63080 }, { "epoch": 1.47, "learning_rate": 1.9077116237591232e-05, "loss": 0.0137, "step": 63085 }, { "epoch": 1.47, "learning_rate": 1.90762361831452e-05, "loss": 0.0606, "step": 63090 }, { "epoch": 1.47, "learning_rate": 1.9075356128699162e-05, "loss": 0.0464, "step": 63095 }, { "epoch": 1.47, "learning_rate": 1.9074476074253127e-05, "loss": 0.1283, "step": 63100 }, { "epoch": 1.47, "learning_rate": 1.9073596019807095e-05, "loss": 0.1551, "step": 63105 }, { "epoch": 1.47, "learning_rate": 1.9072715965361057e-05, "loss": 0.2448, "step": 63110 }, { "epoch": 1.47, "learning_rate": 1.9071835910915022e-05, "loss": 0.2135, "step": 63115 }, { "epoch": 1.47, "learning_rate": 1.9070955856468987e-05, "loss": 0.0297, "step": 63120 }, { "epoch": 1.47, "learning_rate": 1.9070075802022955e-05, "loss": 0.1377, "step": 63125 }, { "epoch": 1.47, "learning_rate": 1.9069195747576917e-05, "loss": 0.0523, "step": 63130 }, { "epoch": 1.47, "learning_rate": 1.9068315693130882e-05, "loss": 0.1095, "step": 63135 }, { "epoch": 1.47, "learning_rate": 1.9067435638684847e-05, "loss": 0.149, "step": 63140 }, { "epoch": 1.47, "learning_rate": 1.9066555584238812e-05, "loss": 0.0994, "step": 63145 }, { "epoch": 1.47, "learning_rate": 1.9065675529792777e-05, "loss": 0.0975, "step": 63150 }, { "epoch": 1.47, "learning_rate": 1.9064795475346742e-05, "loss": 0.1562, "step": 63155 }, { "epoch": 1.47, "learning_rate": 1.9063915420900707e-05, "loss": 0.2325, "step": 63160 }, { "epoch": 1.47, "learning_rate": 1.9063035366454672e-05, "loss": 0.2993, "step": 63165 }, { "epoch": 1.47, "learning_rate": 1.9062155312008637e-05, "loss": 0.038, "step": 63170 }, { "epoch": 1.47, "learning_rate": 1.9061275257562602e-05, "loss": 0.0717, "step": 63175 }, { "epoch": 1.47, "learning_rate": 1.9060395203116564e-05, "loss": 0.0414, "step": 63180 }, { "epoch": 1.47, "learning_rate": 1.9059515148670532e-05, "loss": 0.0936, "step": 63185 }, { "epoch": 1.47, "learning_rate": 1.9058635094224497e-05, "loss": 0.1233, "step": 63190 }, { "epoch": 1.47, "learning_rate": 1.9057755039778462e-05, "loss": 0.0735, "step": 63195 }, { "epoch": 1.47, "learning_rate": 1.9056874985332424e-05, "loss": 0.1291, "step": 63200 }, { "epoch": 1.47, "learning_rate": 1.9055994930886392e-05, "loss": 0.2731, "step": 63205 }, { "epoch": 1.47, "learning_rate": 1.9055114876440357e-05, "loss": 0.2582, "step": 63210 }, { "epoch": 1.47, "learning_rate": 1.905423482199432e-05, "loss": 0.1799, "step": 63215 }, { "epoch": 1.47, "learning_rate": 1.9053354767548284e-05, "loss": 0.0938, "step": 63220 }, { "epoch": 1.48, "learning_rate": 1.9052474713102252e-05, "loss": 0.0689, "step": 63225 }, { "epoch": 1.48, "learning_rate": 1.9051594658656217e-05, "loss": 0.111, "step": 63230 }, { "epoch": 1.48, "learning_rate": 1.905071460421018e-05, "loss": 0.1519, "step": 63235 }, { "epoch": 1.48, "learning_rate": 1.9049834549764147e-05, "loss": 0.2124, "step": 63240 }, { "epoch": 1.48, "learning_rate": 1.9048954495318112e-05, "loss": 0.1036, "step": 63245 }, { "epoch": 1.48, "learning_rate": 1.9048074440872074e-05, "loss": 0.1041, "step": 63250 }, { "epoch": 1.48, "learning_rate": 1.904719438642604e-05, "loss": 0.1522, "step": 63255 }, { "epoch": 1.48, "learning_rate": 1.9046314331980007e-05, "loss": 0.2016, "step": 63260 }, { "epoch": 1.48, "learning_rate": 1.9045434277533972e-05, "loss": 0.3007, "step": 63265 }, { "epoch": 1.48, "learning_rate": 1.9044554223087934e-05, "loss": 0.1293, "step": 63270 }, { "epoch": 1.48, "learning_rate": 1.90436741686419e-05, "loss": 0.0749, "step": 63275 }, { "epoch": 1.48, "learning_rate": 1.9042794114195868e-05, "loss": 0.0788, "step": 63280 }, { "epoch": 1.48, "learning_rate": 1.904191405974983e-05, "loss": 0.0475, "step": 63285 }, { "epoch": 1.48, "learning_rate": 1.9041034005303794e-05, "loss": 0.1087, "step": 63290 }, { "epoch": 1.48, "learning_rate": 1.904015395085776e-05, "loss": 0.1047, "step": 63295 }, { "epoch": 1.48, "learning_rate": 1.9039273896411728e-05, "loss": 0.0673, "step": 63300 }, { "epoch": 1.48, "learning_rate": 1.903839384196569e-05, "loss": 0.3435, "step": 63305 }, { "epoch": 1.48, "learning_rate": 1.9037513787519654e-05, "loss": 0.447, "step": 63310 }, { "epoch": 1.48, "learning_rate": 1.903663373307362e-05, "loss": 0.3737, "step": 63315 }, { "epoch": 1.48, "learning_rate": 1.9035753678627588e-05, "loss": 0.0295, "step": 63320 }, { "epoch": 1.48, "learning_rate": 1.903487362418155e-05, "loss": 0.1098, "step": 63325 }, { "epoch": 1.48, "learning_rate": 1.9033993569735514e-05, "loss": 0.0296, "step": 63330 }, { "epoch": 1.48, "learning_rate": 1.903311351528948e-05, "loss": 0.062, "step": 63335 }, { "epoch": 1.48, "learning_rate": 1.9032233460843444e-05, "loss": 0.1905, "step": 63340 }, { "epoch": 1.48, "learning_rate": 1.903135340639741e-05, "loss": 0.1654, "step": 63345 }, { "epoch": 1.48, "learning_rate": 1.9030473351951374e-05, "loss": 0.2373, "step": 63350 }, { "epoch": 1.48, "learning_rate": 1.902959329750534e-05, "loss": 0.2025, "step": 63355 }, { "epoch": 1.48, "learning_rate": 1.9028713243059304e-05, "loss": 0.2421, "step": 63360 }, { "epoch": 1.48, "learning_rate": 1.902783318861327e-05, "loss": 0.2759, "step": 63365 }, { "epoch": 1.48, "learning_rate": 1.9026953134167235e-05, "loss": 0.0431, "step": 63370 }, { "epoch": 1.48, "learning_rate": 1.90260730797212e-05, "loss": 0.0493, "step": 63375 }, { "epoch": 1.48, "learning_rate": 1.9025193025275165e-05, "loss": 0.0239, "step": 63380 }, { "epoch": 1.48, "learning_rate": 1.902431297082913e-05, "loss": 0.1251, "step": 63385 }, { "epoch": 1.48, "learning_rate": 1.9023432916383095e-05, "loss": 0.1274, "step": 63390 }, { "epoch": 1.48, "learning_rate": 1.902255286193706e-05, "loss": 0.0933, "step": 63395 }, { "epoch": 1.48, "learning_rate": 1.9021672807491025e-05, "loss": 0.0901, "step": 63400 }, { "epoch": 1.48, "learning_rate": 1.902079275304499e-05, "loss": 0.2505, "step": 63405 }, { "epoch": 1.48, "learning_rate": 1.901991269859895e-05, "loss": 0.2317, "step": 63410 }, { "epoch": 1.48, "learning_rate": 1.901903264415292e-05, "loss": 0.2909, "step": 63415 }, { "epoch": 1.48, "learning_rate": 1.9018152589706885e-05, "loss": 0.1286, "step": 63420 }, { "epoch": 1.48, "learning_rate": 1.901727253526085e-05, "loss": 0.0699, "step": 63425 }, { "epoch": 1.48, "learning_rate": 1.901639248081481e-05, "loss": 0.0891, "step": 63430 }, { "epoch": 1.48, "learning_rate": 1.901551242636878e-05, "loss": 0.0966, "step": 63435 }, { "epoch": 1.48, "learning_rate": 1.9014632371922745e-05, "loss": 0.092, "step": 63440 }, { "epoch": 1.48, "learning_rate": 1.9013752317476706e-05, "loss": 0.1145, "step": 63445 }, { "epoch": 1.48, "learning_rate": 1.901287226303067e-05, "loss": 0.1638, "step": 63450 }, { "epoch": 1.48, "learning_rate": 1.901199220858464e-05, "loss": 0.2245, "step": 63455 }, { "epoch": 1.48, "learning_rate": 1.9011112154138605e-05, "loss": 0.359, "step": 63460 }, { "epoch": 1.48, "learning_rate": 1.9010232099692567e-05, "loss": 0.1861, "step": 63465 }, { "epoch": 1.48, "learning_rate": 1.900935204524653e-05, "loss": 0.1047, "step": 63470 }, { "epoch": 1.48, "learning_rate": 1.90084719908005e-05, "loss": 0.0219, "step": 63475 }, { "epoch": 1.48, "learning_rate": 1.900759193635446e-05, "loss": 0.1229, "step": 63480 }, { "epoch": 1.48, "learning_rate": 1.9006711881908427e-05, "loss": 0.0627, "step": 63485 }, { "epoch": 1.48, "learning_rate": 1.900583182746239e-05, "loss": 0.0621, "step": 63490 }, { "epoch": 1.48, "learning_rate": 1.900495177301636e-05, "loss": 0.1171, "step": 63495 }, { "epoch": 1.48, "learning_rate": 1.900407171857032e-05, "loss": 0.1125, "step": 63500 }, { "epoch": 1.48, "learning_rate": 1.9003191664124287e-05, "loss": 0.1567, "step": 63505 }, { "epoch": 1.48, "learning_rate": 1.9002311609678255e-05, "loss": 0.2525, "step": 63510 }, { "epoch": 1.48, "learning_rate": 1.9001431555232217e-05, "loss": 0.2691, "step": 63515 }, { "epoch": 1.48, "learning_rate": 1.9000551500786182e-05, "loss": 0.1183, "step": 63520 }, { "epoch": 1.48, "learning_rate": 1.8999671446340147e-05, "loss": 0.0727, "step": 63525 }, { "epoch": 1.48, "learning_rate": 1.8998791391894115e-05, "loss": 0.0476, "step": 63530 }, { "epoch": 1.48, "learning_rate": 1.8997911337448077e-05, "loss": 0.0178, "step": 63535 }, { "epoch": 1.48, "learning_rate": 1.8997031283002042e-05, "loss": 0.1443, "step": 63540 }, { "epoch": 1.48, "learning_rate": 1.8996151228556007e-05, "loss": 0.078, "step": 63545 }, { "epoch": 1.48, "learning_rate": 1.8995271174109972e-05, "loss": 0.2529, "step": 63550 }, { "epoch": 1.48, "learning_rate": 1.8994391119663937e-05, "loss": 0.2957, "step": 63555 }, { "epoch": 1.48, "learning_rate": 1.8993511065217902e-05, "loss": 0.1786, "step": 63560 }, { "epoch": 1.48, "learning_rate": 1.8992631010771867e-05, "loss": 0.2219, "step": 63565 }, { "epoch": 1.48, "learning_rate": 1.8991750956325832e-05, "loss": 0.0806, "step": 63570 }, { "epoch": 1.48, "learning_rate": 1.8990870901879797e-05, "loss": 0.0778, "step": 63575 }, { "epoch": 1.48, "learning_rate": 1.8989990847433762e-05, "loss": 0.0119, "step": 63580 }, { "epoch": 1.48, "learning_rate": 1.8989110792987724e-05, "loss": 0.0403, "step": 63585 }, { "epoch": 1.48, "learning_rate": 1.8988230738541692e-05, "loss": 0.1405, "step": 63590 }, { "epoch": 1.48, "learning_rate": 1.8987350684095657e-05, "loss": 0.1082, "step": 63595 }, { "epoch": 1.48, "learning_rate": 1.8986470629649622e-05, "loss": 0.1946, "step": 63600 }, { "epoch": 1.48, "learning_rate": 1.8985590575203584e-05, "loss": 0.2025, "step": 63605 }, { "epoch": 1.48, "learning_rate": 1.8984710520757552e-05, "loss": 0.1201, "step": 63610 }, { "epoch": 1.48, "learning_rate": 1.8983830466311517e-05, "loss": 0.2197, "step": 63615 }, { "epoch": 1.48, "learning_rate": 1.8982950411865482e-05, "loss": 0.0632, "step": 63620 }, { "epoch": 1.48, "learning_rate": 1.8982070357419444e-05, "loss": 0.0256, "step": 63625 }, { "epoch": 1.48, "learning_rate": 1.8981190302973412e-05, "loss": 0.0848, "step": 63630 }, { "epoch": 1.48, "learning_rate": 1.8980310248527377e-05, "loss": 0.0935, "step": 63635 }, { "epoch": 1.48, "learning_rate": 1.897943019408134e-05, "loss": 0.1465, "step": 63640 }, { "epoch": 1.48, "learning_rate": 1.8978550139635307e-05, "loss": 0.2149, "step": 63645 }, { "epoch": 1.49, "learning_rate": 1.8977670085189272e-05, "loss": 0.192, "step": 63650 }, { "epoch": 1.49, "learning_rate": 1.8976790030743237e-05, "loss": 0.2012, "step": 63655 }, { "epoch": 1.49, "learning_rate": 1.89759099762972e-05, "loss": 0.3172, "step": 63660 }, { "epoch": 1.49, "learning_rate": 1.8975029921851167e-05, "loss": 0.328, "step": 63665 }, { "epoch": 1.49, "learning_rate": 1.8974149867405132e-05, "loss": 0.0301, "step": 63670 }, { "epoch": 1.49, "learning_rate": 1.8973269812959094e-05, "loss": 0.0686, "step": 63675 }, { "epoch": 1.49, "learning_rate": 1.897238975851306e-05, "loss": 0.0643, "step": 63680 }, { "epoch": 1.49, "learning_rate": 1.8971509704067027e-05, "loss": 0.0828, "step": 63685 }, { "epoch": 1.49, "learning_rate": 1.8970629649620992e-05, "loss": 0.0793, "step": 63690 }, { "epoch": 1.49, "learning_rate": 1.8969749595174954e-05, "loss": 0.1259, "step": 63695 }, { "epoch": 1.49, "learning_rate": 1.896886954072892e-05, "loss": 0.1304, "step": 63700 }, { "epoch": 1.49, "learning_rate": 1.8967989486282888e-05, "loss": 0.2751, "step": 63705 }, { "epoch": 1.49, "learning_rate": 1.896710943183685e-05, "loss": 0.1782, "step": 63710 }, { "epoch": 1.49, "learning_rate": 1.8966229377390814e-05, "loss": 0.2986, "step": 63715 }, { "epoch": 1.49, "learning_rate": 1.896534932294478e-05, "loss": 0.0515, "step": 63720 }, { "epoch": 1.49, "learning_rate": 1.8964469268498748e-05, "loss": 0.0305, "step": 63725 }, { "epoch": 1.49, "learning_rate": 1.896358921405271e-05, "loss": 0.0144, "step": 63730 }, { "epoch": 1.49, "learning_rate": 1.8962709159606674e-05, "loss": 0.1331, "step": 63735 }, { "epoch": 1.49, "learning_rate": 1.896182910516064e-05, "loss": 0.2047, "step": 63740 }, { "epoch": 1.49, "learning_rate": 1.8960949050714604e-05, "loss": 0.0722, "step": 63745 }, { "epoch": 1.49, "learning_rate": 1.896006899626857e-05, "loss": 0.2094, "step": 63750 }, { "epoch": 1.49, "learning_rate": 1.8959188941822534e-05, "loss": 0.0466, "step": 63755 }, { "epoch": 1.49, "learning_rate": 1.89583088873765e-05, "loss": 0.5104, "step": 63760 }, { "epoch": 1.49, "learning_rate": 1.8957428832930464e-05, "loss": 0.3222, "step": 63765 }, { "epoch": 1.49, "learning_rate": 1.895654877848443e-05, "loss": 0.0397, "step": 63770 }, { "epoch": 1.49, "learning_rate": 1.8955668724038394e-05, "loss": 0.1126, "step": 63775 }, { "epoch": 1.49, "learning_rate": 1.895478866959236e-05, "loss": 0.0618, "step": 63780 }, { "epoch": 1.49, "learning_rate": 1.8953908615146324e-05, "loss": 0.1839, "step": 63785 }, { "epoch": 1.49, "learning_rate": 1.895302856070029e-05, "loss": 0.0847, "step": 63790 }, { "epoch": 1.49, "learning_rate": 1.8952148506254254e-05, "loss": 0.1018, "step": 63795 }, { "epoch": 1.49, "learning_rate": 1.895126845180822e-05, "loss": 0.3439, "step": 63800 }, { "epoch": 1.49, "learning_rate": 1.8950388397362185e-05, "loss": 0.1888, "step": 63805 }, { "epoch": 1.49, "learning_rate": 1.894950834291615e-05, "loss": 0.257, "step": 63810 }, { "epoch": 1.49, "learning_rate": 1.894862828847011e-05, "loss": 0.2085, "step": 63815 }, { "epoch": 1.49, "learning_rate": 1.894774823402408e-05, "loss": 0.0242, "step": 63820 }, { "epoch": 1.49, "learning_rate": 1.8946868179578045e-05, "loss": 0.0436, "step": 63825 }, { "epoch": 1.49, "learning_rate": 1.894598812513201e-05, "loss": 0.0494, "step": 63830 }, { "epoch": 1.49, "learning_rate": 1.894510807068597e-05, "loss": 0.0989, "step": 63835 }, { "epoch": 1.49, "learning_rate": 1.894422801623994e-05, "loss": 0.0759, "step": 63840 }, { "epoch": 1.49, "learning_rate": 1.8943347961793905e-05, "loss": 0.0822, "step": 63845 }, { "epoch": 1.49, "learning_rate": 1.8942467907347866e-05, "loss": 0.1528, "step": 63850 }, { "epoch": 1.49, "learning_rate": 1.894158785290183e-05, "loss": 0.2158, "step": 63855 }, { "epoch": 1.49, "learning_rate": 1.89407077984558e-05, "loss": 0.3222, "step": 63860 }, { "epoch": 1.49, "learning_rate": 1.8939827744009765e-05, "loss": 0.2052, "step": 63865 }, { "epoch": 1.49, "learning_rate": 1.8938947689563726e-05, "loss": 0.0866, "step": 63870 }, { "epoch": 1.49, "learning_rate": 1.893806763511769e-05, "loss": 0.0712, "step": 63875 }, { "epoch": 1.49, "learning_rate": 1.893718758067166e-05, "loss": 0.1383, "step": 63880 }, { "epoch": 1.49, "learning_rate": 1.893630752622562e-05, "loss": 0.1176, "step": 63885 }, { "epoch": 1.49, "learning_rate": 1.8935427471779586e-05, "loss": 0.0767, "step": 63890 }, { "epoch": 1.49, "learning_rate": 1.893454741733355e-05, "loss": 0.1221, "step": 63895 }, { "epoch": 1.49, "learning_rate": 1.893366736288752e-05, "loss": 0.1777, "step": 63900 }, { "epoch": 1.49, "learning_rate": 1.893278730844148e-05, "loss": 0.2311, "step": 63905 }, { "epoch": 1.49, "learning_rate": 1.8931907253995447e-05, "loss": 0.2273, "step": 63910 }, { "epoch": 1.49, "learning_rate": 1.8931027199549415e-05, "loss": 0.248, "step": 63915 }, { "epoch": 1.49, "learning_rate": 1.893014714510338e-05, "loss": 0.0333, "step": 63920 }, { "epoch": 1.49, "learning_rate": 1.892926709065734e-05, "loss": 0.0994, "step": 63925 }, { "epoch": 1.49, "learning_rate": 1.8928387036211307e-05, "loss": 0.0917, "step": 63930 }, { "epoch": 1.49, "learning_rate": 1.8927506981765275e-05, "loss": 0.1115, "step": 63935 }, { "epoch": 1.49, "learning_rate": 1.8926626927319237e-05, "loss": 0.1151, "step": 63940 }, { "epoch": 1.49, "learning_rate": 1.8925746872873202e-05, "loss": 0.0816, "step": 63945 }, { "epoch": 1.49, "learning_rate": 1.8924866818427167e-05, "loss": 0.1251, "step": 63950 }, { "epoch": 1.49, "learning_rate": 1.8923986763981135e-05, "loss": 0.0822, "step": 63955 }, { "epoch": 1.49, "learning_rate": 1.8923106709535097e-05, "loss": 0.5089, "step": 63960 }, { "epoch": 1.49, "learning_rate": 1.8922226655089062e-05, "loss": 0.2195, "step": 63965 }, { "epoch": 1.49, "learning_rate": 1.8921346600643027e-05, "loss": 0.1769, "step": 63970 }, { "epoch": 1.49, "learning_rate": 1.8920466546196992e-05, "loss": 0.0396, "step": 63975 }, { "epoch": 1.49, "learning_rate": 1.8919586491750957e-05, "loss": 0.0701, "step": 63980 }, { "epoch": 1.49, "learning_rate": 1.8918706437304922e-05, "loss": 0.119, "step": 63985 }, { "epoch": 1.49, "learning_rate": 1.8917826382858887e-05, "loss": 0.1948, "step": 63990 }, { "epoch": 1.49, "learning_rate": 1.8916946328412852e-05, "loss": 0.2042, "step": 63995 }, { "epoch": 1.49, "learning_rate": 1.8916066273966817e-05, "loss": 0.1971, "step": 64000 }, { "epoch": 1.49, "learning_rate": 1.8915186219520782e-05, "loss": 0.136, "step": 64005 }, { "epoch": 1.49, "learning_rate": 1.8914306165074744e-05, "loss": 0.2862, "step": 64010 }, { "epoch": 1.49, "learning_rate": 1.8913426110628712e-05, "loss": 0.257, "step": 64015 }, { "epoch": 1.49, "learning_rate": 1.8912546056182677e-05, "loss": 0.0267, "step": 64020 }, { "epoch": 1.49, "learning_rate": 1.8911666001736642e-05, "loss": 0.0711, "step": 64025 }, { "epoch": 1.49, "learning_rate": 1.8910785947290604e-05, "loss": 0.0598, "step": 64030 }, { "epoch": 1.49, "learning_rate": 1.8909905892844572e-05, "loss": 0.1283, "step": 64035 }, { "epoch": 1.49, "learning_rate": 1.8909025838398537e-05, "loss": 0.1177, "step": 64040 }, { "epoch": 1.49, "learning_rate": 1.89081457839525e-05, "loss": 0.1552, "step": 64045 }, { "epoch": 1.49, "learning_rate": 1.8907265729506467e-05, "loss": 0.1788, "step": 64050 }, { "epoch": 1.49, "learning_rate": 1.8906385675060432e-05, "loss": 0.1607, "step": 64055 }, { "epoch": 1.49, "learning_rate": 1.8905505620614397e-05, "loss": 0.2521, "step": 64060 }, { "epoch": 1.49, "learning_rate": 1.890462556616836e-05, "loss": 0.3605, "step": 64065 }, { "epoch": 1.49, "learning_rate": 1.8903745511722327e-05, "loss": 0.1179, "step": 64070 }, { "epoch": 1.49, "learning_rate": 1.8902865457276292e-05, "loss": 0.0878, "step": 64075 }, { "epoch": 1.5, "learning_rate": 1.8901985402830254e-05, "loss": 0.0356, "step": 64080 }, { "epoch": 1.5, "learning_rate": 1.890110534838422e-05, "loss": 0.1268, "step": 64085 }, { "epoch": 1.5, "learning_rate": 1.8900225293938187e-05, "loss": 0.1604, "step": 64090 }, { "epoch": 1.5, "learning_rate": 1.8899345239492152e-05, "loss": 0.0837, "step": 64095 }, { "epoch": 1.5, "learning_rate": 1.8898465185046114e-05, "loss": 0.2062, "step": 64100 }, { "epoch": 1.5, "learning_rate": 1.889758513060008e-05, "loss": 0.1399, "step": 64105 }, { "epoch": 1.5, "learning_rate": 1.8896705076154047e-05, "loss": 0.2223, "step": 64110 }, { "epoch": 1.5, "learning_rate": 1.889582502170801e-05, "loss": 0.2136, "step": 64115 }, { "epoch": 1.5, "learning_rate": 1.8894944967261974e-05, "loss": 0.0134, "step": 64120 }, { "epoch": 1.5, "learning_rate": 1.889406491281594e-05, "loss": 0.0307, "step": 64125 }, { "epoch": 1.5, "learning_rate": 1.8893184858369907e-05, "loss": 0.0673, "step": 64130 }, { "epoch": 1.5, "learning_rate": 1.889230480392387e-05, "loss": 0.0829, "step": 64135 }, { "epoch": 1.5, "learning_rate": 1.8891424749477834e-05, "loss": 0.1065, "step": 64140 }, { "epoch": 1.5, "learning_rate": 1.88905446950318e-05, "loss": 0.1057, "step": 64145 }, { "epoch": 1.5, "learning_rate": 1.8889664640585764e-05, "loss": 0.2457, "step": 64150 }, { "epoch": 1.5, "learning_rate": 1.888878458613973e-05, "loss": 0.1361, "step": 64155 }, { "epoch": 1.5, "learning_rate": 1.8887904531693694e-05, "loss": 0.1523, "step": 64160 }, { "epoch": 1.5, "learning_rate": 1.888702447724766e-05, "loss": 0.325, "step": 64165 }, { "epoch": 1.5, "learning_rate": 1.8886144422801624e-05, "loss": 0.0334, "step": 64170 }, { "epoch": 1.5, "learning_rate": 1.888526436835559e-05, "loss": 0.0085, "step": 64175 }, { "epoch": 1.5, "learning_rate": 1.8884384313909554e-05, "loss": 0.0341, "step": 64180 }, { "epoch": 1.5, "learning_rate": 1.888350425946352e-05, "loss": 0.084, "step": 64185 }, { "epoch": 1.5, "learning_rate": 1.8882624205017484e-05, "loss": 0.2861, "step": 64190 }, { "epoch": 1.5, "learning_rate": 1.888174415057145e-05, "loss": 0.3748, "step": 64195 }, { "epoch": 1.5, "learning_rate": 1.8880864096125414e-05, "loss": 0.1162, "step": 64200 }, { "epoch": 1.5, "learning_rate": 1.887998404167938e-05, "loss": 0.0594, "step": 64205 }, { "epoch": 1.5, "learning_rate": 1.8879103987233344e-05, "loss": 0.0807, "step": 64210 }, { "epoch": 1.5, "learning_rate": 1.887822393278731e-05, "loss": 0.3687, "step": 64215 }, { "epoch": 1.5, "learning_rate": 1.8877343878341274e-05, "loss": 0.0879, "step": 64220 }, { "epoch": 1.5, "learning_rate": 1.887646382389524e-05, "loss": 0.082, "step": 64225 }, { "epoch": 1.5, "learning_rate": 1.8875583769449204e-05, "loss": 0.0962, "step": 64230 }, { "epoch": 1.5, "learning_rate": 1.887470371500317e-05, "loss": 0.0969, "step": 64235 }, { "epoch": 1.5, "learning_rate": 1.887382366055713e-05, "loss": 0.0629, "step": 64240 }, { "epoch": 1.5, "learning_rate": 1.88729436061111e-05, "loss": 0.1109, "step": 64245 }, { "epoch": 1.5, "learning_rate": 1.8872063551665065e-05, "loss": 0.0677, "step": 64250 }, { "epoch": 1.5, "learning_rate": 1.887118349721903e-05, "loss": 0.1327, "step": 64255 }, { "epoch": 1.5, "learning_rate": 1.887030344277299e-05, "loss": 0.1092, "step": 64260 }, { "epoch": 1.5, "learning_rate": 1.886942338832696e-05, "loss": 0.3032, "step": 64265 }, { "epoch": 1.5, "learning_rate": 1.8868543333880925e-05, "loss": 0.0103, "step": 64270 }, { "epoch": 1.5, "learning_rate": 1.8867663279434886e-05, "loss": 0.0234, "step": 64275 }, { "epoch": 1.5, "learning_rate": 1.886678322498885e-05, "loss": 0.1636, "step": 64280 }, { "epoch": 1.5, "learning_rate": 1.886590317054282e-05, "loss": 0.0792, "step": 64285 }, { "epoch": 1.5, "learning_rate": 1.8865023116096785e-05, "loss": 0.1167, "step": 64290 }, { "epoch": 1.5, "learning_rate": 1.8864143061650746e-05, "loss": 0.0718, "step": 64295 }, { "epoch": 1.5, "learning_rate": 1.886326300720471e-05, "loss": 0.2496, "step": 64300 }, { "epoch": 1.5, "learning_rate": 1.886238295275868e-05, "loss": 0.2795, "step": 64305 }, { "epoch": 1.5, "learning_rate": 1.886150289831264e-05, "loss": 0.2417, "step": 64310 }, { "epoch": 1.5, "learning_rate": 1.8860622843866606e-05, "loss": 0.1902, "step": 64315 }, { "epoch": 1.5, "learning_rate": 1.8859742789420575e-05, "loss": 0.0384, "step": 64320 }, { "epoch": 1.5, "learning_rate": 1.885886273497454e-05, "loss": 0.0938, "step": 64325 }, { "epoch": 1.5, "learning_rate": 1.88579826805285e-05, "loss": 0.1225, "step": 64330 }, { "epoch": 1.5, "learning_rate": 1.8857102626082467e-05, "loss": 0.0667, "step": 64335 }, { "epoch": 1.5, "learning_rate": 1.8856222571636435e-05, "loss": 0.1838, "step": 64340 }, { "epoch": 1.5, "learning_rate": 1.8855342517190397e-05, "loss": 0.1142, "step": 64345 }, { "epoch": 1.5, "learning_rate": 1.885446246274436e-05, "loss": 0.0762, "step": 64350 }, { "epoch": 1.5, "learning_rate": 1.8853582408298327e-05, "loss": 0.1279, "step": 64355 }, { "epoch": 1.5, "learning_rate": 1.8852702353852295e-05, "loss": 0.3155, "step": 64360 }, { "epoch": 1.5, "learning_rate": 1.8851822299406257e-05, "loss": 0.1515, "step": 64365 }, { "epoch": 1.5, "learning_rate": 1.885094224496022e-05, "loss": 0.0569, "step": 64370 }, { "epoch": 1.5, "learning_rate": 1.8850062190514187e-05, "loss": 0.0376, "step": 64375 }, { "epoch": 1.5, "learning_rate": 1.8849182136068152e-05, "loss": 0.0532, "step": 64380 }, { "epoch": 1.5, "learning_rate": 1.8848302081622117e-05, "loss": 0.1006, "step": 64385 }, { "epoch": 1.5, "learning_rate": 1.8847422027176082e-05, "loss": 0.2045, "step": 64390 }, { "epoch": 1.5, "learning_rate": 1.8846541972730047e-05, "loss": 0.0984, "step": 64395 }, { "epoch": 1.5, "learning_rate": 1.8845661918284012e-05, "loss": 0.1179, "step": 64400 }, { "epoch": 1.5, "learning_rate": 1.8844781863837977e-05, "loss": 0.3547, "step": 64405 }, { "epoch": 1.5, "learning_rate": 1.8843901809391942e-05, "loss": 0.3345, "step": 64410 }, { "epoch": 1.5, "learning_rate": 1.8843021754945903e-05, "loss": 0.3676, "step": 64415 }, { "epoch": 1.5, "learning_rate": 1.8842317711389078e-05, "loss": 0.1586, "step": 64420 }, { "epoch": 1.5, "learning_rate": 1.8841437656943043e-05, "loss": 0.0988, "step": 64425 }, { "epoch": 1.5, "learning_rate": 1.8840557602497008e-05, "loss": 0.1104, "step": 64430 }, { "epoch": 1.5, "learning_rate": 1.8839677548050973e-05, "loss": 0.091, "step": 64435 }, { "epoch": 1.5, "learning_rate": 1.8838797493604938e-05, "loss": 0.236, "step": 64440 }, { "epoch": 1.5, "learning_rate": 1.8837917439158903e-05, "loss": 0.1507, "step": 64445 }, { "epoch": 1.5, "learning_rate": 1.883703738471287e-05, "loss": 0.1387, "step": 64450 }, { "epoch": 1.5, "learning_rate": 1.8836157330266833e-05, "loss": 0.2209, "step": 64455 }, { "epoch": 1.5, "learning_rate": 1.88352772758208e-05, "loss": 0.422, "step": 64460 }, { "epoch": 1.5, "learning_rate": 1.8834397221374763e-05, "loss": 0.2388, "step": 64465 }, { "epoch": 1.5, "learning_rate": 1.8833517166928725e-05, "loss": 0.1069, "step": 64470 }, { "epoch": 1.5, "learning_rate": 1.8832637112482693e-05, "loss": 0.1491, "step": 64475 }, { "epoch": 1.5, "learning_rate": 1.883175705803666e-05, "loss": 0.0543, "step": 64480 }, { "epoch": 1.5, "learning_rate": 1.8830877003590623e-05, "loss": 0.1355, "step": 64485 }, { "epoch": 1.5, "learning_rate": 1.8829996949144585e-05, "loss": 0.0539, "step": 64490 }, { "epoch": 1.5, "learning_rate": 1.8829116894698554e-05, "loss": 0.1807, "step": 64495 }, { "epoch": 1.5, "learning_rate": 1.882823684025252e-05, "loss": 0.3231, "step": 64500 }, { "epoch": 1.5, "learning_rate": 1.882735678580648e-05, "loss": 0.1704, "step": 64505 }, { "epoch": 1.51, "learning_rate": 1.882647673136045e-05, "loss": 0.2379, "step": 64510 }, { "epoch": 1.51, "learning_rate": 1.8825596676914414e-05, "loss": 0.1996, "step": 64515 }, { "epoch": 1.51, "learning_rate": 1.882471662246838e-05, "loss": 0.0867, "step": 64520 }, { "epoch": 1.51, "learning_rate": 1.882383656802234e-05, "loss": 0.0581, "step": 64525 }, { "epoch": 1.51, "learning_rate": 1.882295651357631e-05, "loss": 0.0342, "step": 64530 }, { "epoch": 1.51, "learning_rate": 1.8822076459130274e-05, "loss": 0.069, "step": 64535 }, { "epoch": 1.51, "learning_rate": 1.8821196404684235e-05, "loss": 0.0454, "step": 64540 }, { "epoch": 1.51, "learning_rate": 1.88203163502382e-05, "loss": 0.1455, "step": 64545 }, { "epoch": 1.51, "learning_rate": 1.881943629579217e-05, "loss": 0.2103, "step": 64550 }, { "epoch": 1.51, "learning_rate": 1.8818556241346134e-05, "loss": 0.185, "step": 64555 }, { "epoch": 1.51, "learning_rate": 1.8817676186900095e-05, "loss": 0.4419, "step": 64560 }, { "epoch": 1.51, "learning_rate": 1.881679613245406e-05, "loss": 0.2907, "step": 64565 }, { "epoch": 1.51, "learning_rate": 1.881591607800803e-05, "loss": 0.0669, "step": 64570 }, { "epoch": 1.51, "learning_rate": 1.881503602356199e-05, "loss": 0.0703, "step": 64575 }, { "epoch": 1.51, "learning_rate": 1.8814155969115955e-05, "loss": 0.0843, "step": 64580 }, { "epoch": 1.51, "learning_rate": 1.881327591466992e-05, "loss": 0.1654, "step": 64585 }, { "epoch": 1.51, "learning_rate": 1.881239586022389e-05, "loss": 0.0943, "step": 64590 }, { "epoch": 1.51, "learning_rate": 1.881151580577785e-05, "loss": 0.1036, "step": 64595 }, { "epoch": 1.51, "learning_rate": 1.8810635751331816e-05, "loss": 0.1929, "step": 64600 }, { "epoch": 1.51, "learning_rate": 1.880975569688578e-05, "loss": 0.207, "step": 64605 }, { "epoch": 1.51, "learning_rate": 1.8808875642439746e-05, "loss": 0.2486, "step": 64610 }, { "epoch": 1.51, "learning_rate": 1.880799558799371e-05, "loss": 0.468, "step": 64615 }, { "epoch": 1.51, "learning_rate": 1.8807115533547676e-05, "loss": 0.0646, "step": 64620 }, { "epoch": 1.51, "learning_rate": 1.880623547910164e-05, "loss": 0.0371, "step": 64625 }, { "epoch": 1.51, "learning_rate": 1.8805355424655606e-05, "loss": 0.0506, "step": 64630 }, { "epoch": 1.51, "learning_rate": 1.880447537020957e-05, "loss": 0.1145, "step": 64635 }, { "epoch": 1.51, "learning_rate": 1.8803595315763536e-05, "loss": 0.0476, "step": 64640 }, { "epoch": 1.51, "learning_rate": 1.88027152613175e-05, "loss": 0.2126, "step": 64645 }, { "epoch": 1.51, "learning_rate": 1.8801835206871466e-05, "loss": 0.1216, "step": 64650 }, { "epoch": 1.51, "learning_rate": 1.880095515242543e-05, "loss": 0.2541, "step": 64655 }, { "epoch": 1.51, "learning_rate": 1.8800075097979396e-05, "loss": 0.3041, "step": 64660 }, { "epoch": 1.51, "learning_rate": 1.879919504353336e-05, "loss": 0.1691, "step": 64665 }, { "epoch": 1.51, "learning_rate": 1.8798314989087326e-05, "loss": 0.0388, "step": 64670 }, { "epoch": 1.51, "learning_rate": 1.879743493464129e-05, "loss": 0.0216, "step": 64675 }, { "epoch": 1.51, "learning_rate": 1.8796554880195252e-05, "loss": 0.0416, "step": 64680 }, { "epoch": 1.51, "learning_rate": 1.879567482574922e-05, "loss": 0.0637, "step": 64685 }, { "epoch": 1.51, "learning_rate": 1.8794794771303186e-05, "loss": 0.0602, "step": 64690 }, { "epoch": 1.51, "learning_rate": 1.879391471685715e-05, "loss": 0.1477, "step": 64695 }, { "epoch": 1.51, "learning_rate": 1.8793034662411113e-05, "loss": 0.3277, "step": 64700 }, { "epoch": 1.51, "learning_rate": 1.879215460796508e-05, "loss": 0.1834, "step": 64705 }, { "epoch": 1.51, "learning_rate": 1.8791274553519046e-05, "loss": 0.2867, "step": 64710 }, { "epoch": 1.51, "learning_rate": 1.879039449907301e-05, "loss": 0.211, "step": 64715 }, { "epoch": 1.51, "learning_rate": 1.8789514444626973e-05, "loss": 0.0782, "step": 64720 }, { "epoch": 1.51, "learning_rate": 1.878863439018094e-05, "loss": 0.0507, "step": 64725 }, { "epoch": 1.51, "learning_rate": 1.8787754335734906e-05, "loss": 0.0382, "step": 64730 }, { "epoch": 1.51, "learning_rate": 1.8786874281288868e-05, "loss": 0.0664, "step": 64735 }, { "epoch": 1.51, "learning_rate": 1.8785994226842833e-05, "loss": 0.1289, "step": 64740 }, { "epoch": 1.51, "learning_rate": 1.87851141723968e-05, "loss": 0.1657, "step": 64745 }, { "epoch": 1.51, "learning_rate": 1.8784234117950766e-05, "loss": 0.188, "step": 64750 }, { "epoch": 1.51, "learning_rate": 1.8783354063504728e-05, "loss": 0.1517, "step": 64755 }, { "epoch": 1.51, "learning_rate": 1.8782474009058693e-05, "loss": 0.1892, "step": 64760 }, { "epoch": 1.51, "learning_rate": 1.878159395461266e-05, "loss": 0.2509, "step": 64765 }, { "epoch": 1.51, "learning_rate": 1.8780713900166623e-05, "loss": 0.0151, "step": 64770 }, { "epoch": 1.51, "learning_rate": 1.8779833845720588e-05, "loss": 0.1624, "step": 64775 }, { "epoch": 1.51, "learning_rate": 1.8778953791274553e-05, "loss": 0.0176, "step": 64780 }, { "epoch": 1.51, "learning_rate": 1.877807373682852e-05, "loss": 0.1706, "step": 64785 }, { "epoch": 1.51, "learning_rate": 1.8777193682382483e-05, "loss": 0.1489, "step": 64790 }, { "epoch": 1.51, "learning_rate": 1.8776313627936448e-05, "loss": 0.1589, "step": 64795 }, { "epoch": 1.51, "learning_rate": 1.8775433573490416e-05, "loss": 0.1881, "step": 64800 }, { "epoch": 1.51, "learning_rate": 1.8774553519044378e-05, "loss": 0.1759, "step": 64805 }, { "epoch": 1.51, "learning_rate": 1.8773673464598343e-05, "loss": 0.3404, "step": 64810 }, { "epoch": 1.51, "learning_rate": 1.8772793410152308e-05, "loss": 0.418, "step": 64815 }, { "epoch": 1.51, "learning_rate": 1.8771913355706276e-05, "loss": 0.0081, "step": 64820 }, { "epoch": 1.51, "learning_rate": 1.8771033301260238e-05, "loss": 0.0949, "step": 64825 }, { "epoch": 1.51, "learning_rate": 1.8770153246814203e-05, "loss": 0.0412, "step": 64830 }, { "epoch": 1.51, "learning_rate": 1.8769273192368168e-05, "loss": 0.1347, "step": 64835 }, { "epoch": 1.51, "learning_rate": 1.8768393137922133e-05, "loss": 0.0431, "step": 64840 }, { "epoch": 1.51, "learning_rate": 1.8767513083476098e-05, "loss": 0.2188, "step": 64845 }, { "epoch": 1.51, "learning_rate": 1.8766633029030063e-05, "loss": 0.3027, "step": 64850 }, { "epoch": 1.51, "learning_rate": 1.8765752974584028e-05, "loss": 0.3335, "step": 64855 }, { "epoch": 1.51, "learning_rate": 1.8764872920137993e-05, "loss": 0.4195, "step": 64860 }, { "epoch": 1.51, "learning_rate": 1.8763992865691958e-05, "loss": 0.2696, "step": 64865 }, { "epoch": 1.51, "learning_rate": 1.8763112811245923e-05, "loss": 0.0323, "step": 64870 }, { "epoch": 1.51, "learning_rate": 1.8762232756799885e-05, "loss": 0.0769, "step": 64875 }, { "epoch": 1.51, "learning_rate": 1.8761352702353853e-05, "loss": 0.0859, "step": 64880 }, { "epoch": 1.51, "learning_rate": 1.876047264790782e-05, "loss": 0.0828, "step": 64885 }, { "epoch": 1.51, "learning_rate": 1.8759592593461783e-05, "loss": 0.2452, "step": 64890 }, { "epoch": 1.51, "learning_rate": 1.8758712539015745e-05, "loss": 0.1691, "step": 64895 }, { "epoch": 1.51, "learning_rate": 1.8757832484569713e-05, "loss": 0.1578, "step": 64900 }, { "epoch": 1.51, "learning_rate": 1.875695243012368e-05, "loss": 0.1902, "step": 64905 }, { "epoch": 1.51, "learning_rate": 1.875607237567764e-05, "loss": 0.2931, "step": 64910 }, { "epoch": 1.51, "learning_rate": 1.8755192321231605e-05, "loss": 0.1641, "step": 64915 }, { "epoch": 1.51, "learning_rate": 1.8754312266785573e-05, "loss": 0.0824, "step": 64920 }, { "epoch": 1.51, "learning_rate": 1.875343221233954e-05, "loss": 0.0861, "step": 64925 }, { "epoch": 1.51, "learning_rate": 1.87525521578935e-05, "loss": 0.0919, "step": 64930 }, { "epoch": 1.52, "learning_rate": 1.875167210344747e-05, "loss": 0.0708, "step": 64935 }, { "epoch": 1.52, "learning_rate": 1.8750792049001434e-05, "loss": 0.0201, "step": 64940 }, { "epoch": 1.52, "learning_rate": 1.8749911994555395e-05, "loss": 0.0834, "step": 64945 }, { "epoch": 1.52, "learning_rate": 1.874903194010936e-05, "loss": 0.2541, "step": 64950 }, { "epoch": 1.52, "learning_rate": 1.874815188566333e-05, "loss": 0.2399, "step": 64955 }, { "epoch": 1.52, "learning_rate": 1.8747271831217294e-05, "loss": 0.1799, "step": 64960 }, { "epoch": 1.52, "learning_rate": 1.8746391776771255e-05, "loss": 0.4461, "step": 64965 }, { "epoch": 1.52, "learning_rate": 1.874551172232522e-05, "loss": 0.0142, "step": 64970 }, { "epoch": 1.52, "learning_rate": 1.874463166787919e-05, "loss": 0.0454, "step": 64975 }, { "epoch": 1.52, "learning_rate": 1.874375161343315e-05, "loss": 0.0591, "step": 64980 }, { "epoch": 1.52, "learning_rate": 1.8742871558987115e-05, "loss": 0.0722, "step": 64985 }, { "epoch": 1.52, "learning_rate": 1.874199150454108e-05, "loss": 0.1427, "step": 64990 }, { "epoch": 1.52, "learning_rate": 1.874111145009505e-05, "loss": 0.1222, "step": 64995 }, { "epoch": 1.52, "learning_rate": 1.874023139564901e-05, "loss": 0.1213, "step": 65000 }, { "epoch": 1.52, "learning_rate": 1.8739351341202975e-05, "loss": 0.1222, "step": 65005 }, { "epoch": 1.52, "learning_rate": 1.873847128675694e-05, "loss": 0.4791, "step": 65010 }, { "epoch": 1.52, "learning_rate": 1.873759123231091e-05, "loss": 0.3184, "step": 65015 }, { "epoch": 1.52, "learning_rate": 1.873671117786487e-05, "loss": 0.102, "step": 65020 }, { "epoch": 1.52, "learning_rate": 1.8735831123418836e-05, "loss": 0.0263, "step": 65025 }, { "epoch": 1.52, "learning_rate": 1.87349510689728e-05, "loss": 0.1244, "step": 65030 }, { "epoch": 1.52, "learning_rate": 1.8734071014526766e-05, "loss": 0.1444, "step": 65035 }, { "epoch": 1.52, "learning_rate": 1.873319096008073e-05, "loss": 0.0856, "step": 65040 }, { "epoch": 1.52, "learning_rate": 1.8732310905634696e-05, "loss": 0.2712, "step": 65045 }, { "epoch": 1.52, "learning_rate": 1.873143085118866e-05, "loss": 0.1511, "step": 65050 }, { "epoch": 1.52, "learning_rate": 1.8730550796742626e-05, "loss": 0.1866, "step": 65055 }, { "epoch": 1.52, "learning_rate": 1.872967074229659e-05, "loss": 0.4221, "step": 65060 }, { "epoch": 1.52, "learning_rate": 1.8728790687850556e-05, "loss": 0.3263, "step": 65065 }, { "epoch": 1.52, "learning_rate": 1.872791063340452e-05, "loss": 0.0566, "step": 65070 }, { "epoch": 1.52, "learning_rate": 1.8727030578958486e-05, "loss": 0.0358, "step": 65075 }, { "epoch": 1.52, "learning_rate": 1.872615052451245e-05, "loss": 0.0648, "step": 65080 }, { "epoch": 1.52, "learning_rate": 1.8725270470066416e-05, "loss": 0.0979, "step": 65085 }, { "epoch": 1.52, "learning_rate": 1.872439041562038e-05, "loss": 0.0479, "step": 65090 }, { "epoch": 1.52, "learning_rate": 1.8723510361174346e-05, "loss": 0.2272, "step": 65095 }, { "epoch": 1.52, "learning_rate": 1.872263030672831e-05, "loss": 0.1873, "step": 65100 }, { "epoch": 1.52, "learning_rate": 1.8721750252282272e-05, "loss": 0.1708, "step": 65105 }, { "epoch": 1.52, "learning_rate": 1.872087019783624e-05, "loss": 0.303, "step": 65110 }, { "epoch": 1.52, "learning_rate": 1.8719990143390206e-05, "loss": 0.1756, "step": 65115 }, { "epoch": 1.52, "learning_rate": 1.871911008894417e-05, "loss": 0.0722, "step": 65120 }, { "epoch": 1.52, "learning_rate": 1.8718230034498133e-05, "loss": 0.0311, "step": 65125 }, { "epoch": 1.52, "learning_rate": 1.87173499800521e-05, "loss": 0.0688, "step": 65130 }, { "epoch": 1.52, "learning_rate": 1.8716469925606066e-05, "loss": 0.1594, "step": 65135 }, { "epoch": 1.52, "learning_rate": 1.8715589871160028e-05, "loss": 0.1313, "step": 65140 }, { "epoch": 1.52, "learning_rate": 1.8714709816713993e-05, "loss": 0.1321, "step": 65145 }, { "epoch": 1.52, "learning_rate": 1.871382976226796e-05, "loss": 0.0472, "step": 65150 }, { "epoch": 1.52, "learning_rate": 1.8712949707821926e-05, "loss": 0.1147, "step": 65155 }, { "epoch": 1.52, "learning_rate": 1.8712069653375888e-05, "loss": 0.6303, "step": 65160 }, { "epoch": 1.52, "learning_rate": 1.8711189598929853e-05, "loss": 0.4302, "step": 65165 }, { "epoch": 1.52, "learning_rate": 1.871030954448382e-05, "loss": 0.0373, "step": 65170 }, { "epoch": 1.52, "learning_rate": 1.8709429490037783e-05, "loss": 0.1811, "step": 65175 }, { "epoch": 1.52, "learning_rate": 1.8708549435591748e-05, "loss": 0.0303, "step": 65180 }, { "epoch": 1.52, "learning_rate": 1.8707669381145713e-05, "loss": 0.0515, "step": 65185 }, { "epoch": 1.52, "learning_rate": 1.870678932669968e-05, "loss": 0.1882, "step": 65190 }, { "epoch": 1.52, "learning_rate": 1.8705909272253643e-05, "loss": 0.1569, "step": 65195 }, { "epoch": 1.52, "learning_rate": 1.8705029217807608e-05, "loss": 0.3448, "step": 65200 }, { "epoch": 1.52, "learning_rate": 1.8704149163361576e-05, "loss": 0.2582, "step": 65205 }, { "epoch": 1.52, "learning_rate": 1.8703269108915538e-05, "loss": 0.4319, "step": 65210 }, { "epoch": 1.52, "learning_rate": 1.8702389054469503e-05, "loss": 0.3479, "step": 65215 }, { "epoch": 1.52, "learning_rate": 1.8701509000023468e-05, "loss": 0.0223, "step": 65220 }, { "epoch": 1.52, "learning_rate": 1.8700628945577436e-05, "loss": 0.0502, "step": 65225 }, { "epoch": 1.52, "learning_rate": 1.8699748891131398e-05, "loss": 0.0818, "step": 65230 }, { "epoch": 1.52, "learning_rate": 1.8698868836685363e-05, "loss": 0.139, "step": 65235 }, { "epoch": 1.52, "learning_rate": 1.8697988782239328e-05, "loss": 0.0574, "step": 65240 }, { "epoch": 1.52, "learning_rate": 1.8697108727793293e-05, "loss": 0.1134, "step": 65245 }, { "epoch": 1.52, "learning_rate": 1.8696228673347258e-05, "loss": 0.0708, "step": 65250 }, { "epoch": 1.52, "learning_rate": 1.8695348618901223e-05, "loss": 0.2419, "step": 65255 }, { "epoch": 1.52, "learning_rate": 1.8694468564455188e-05, "loss": 0.4487, "step": 65260 }, { "epoch": 1.52, "learning_rate": 1.8693588510009153e-05, "loss": 0.2561, "step": 65265 }, { "epoch": 1.52, "learning_rate": 1.8692708455563118e-05, "loss": 0.048, "step": 65270 }, { "epoch": 1.52, "learning_rate": 1.8691828401117083e-05, "loss": 0.005, "step": 65275 }, { "epoch": 1.52, "learning_rate": 1.8690948346671045e-05, "loss": 0.0435, "step": 65280 }, { "epoch": 1.52, "learning_rate": 1.8690068292225013e-05, "loss": 0.0957, "step": 65285 }, { "epoch": 1.52, "learning_rate": 1.8689188237778978e-05, "loss": 0.0395, "step": 65290 }, { "epoch": 1.52, "learning_rate": 1.8688308183332943e-05, "loss": 0.0569, "step": 65295 }, { "epoch": 1.52, "learning_rate": 1.8687428128886905e-05, "loss": 0.0624, "step": 65300 }, { "epoch": 1.52, "learning_rate": 1.8686548074440873e-05, "loss": 0.2176, "step": 65305 }, { "epoch": 1.52, "learning_rate": 1.868566801999484e-05, "loss": 0.4527, "step": 65310 }, { "epoch": 1.52, "learning_rate": 1.8684787965548803e-05, "loss": 0.1384, "step": 65315 }, { "epoch": 1.52, "learning_rate": 1.8683907911102765e-05, "loss": 0.0521, "step": 65320 }, { "epoch": 1.52, "learning_rate": 1.8683027856656733e-05, "loss": 0.0264, "step": 65325 }, { "epoch": 1.52, "learning_rate": 1.86821478022107e-05, "loss": 0.1287, "step": 65330 }, { "epoch": 1.52, "learning_rate": 1.868126774776466e-05, "loss": 0.0487, "step": 65335 }, { "epoch": 1.52, "learning_rate": 1.868038769331863e-05, "loss": 0.0924, "step": 65340 }, { "epoch": 1.52, "learning_rate": 1.8679507638872593e-05, "loss": 0.0771, "step": 65345 }, { "epoch": 1.52, "learning_rate": 1.867862758442656e-05, "loss": 0.0933, "step": 65350 }, { "epoch": 1.52, "learning_rate": 1.867774752998052e-05, "loss": 0.1467, "step": 65355 }, { "epoch": 1.52, "learning_rate": 1.867686747553449e-05, "loss": 0.5108, "step": 65360 }, { "epoch": 1.53, "learning_rate": 1.8675987421088454e-05, "loss": 0.3095, "step": 65365 }, { "epoch": 1.53, "learning_rate": 1.8675107366642415e-05, "loss": 0.092, "step": 65370 }, { "epoch": 1.53, "learning_rate": 1.867422731219638e-05, "loss": 0.0879, "step": 65375 }, { "epoch": 1.53, "learning_rate": 1.867334725775035e-05, "loss": 0.0264, "step": 65380 }, { "epoch": 1.53, "learning_rate": 1.8672467203304314e-05, "loss": 0.0403, "step": 65385 }, { "epoch": 1.53, "learning_rate": 1.8671587148858275e-05, "loss": 0.0336, "step": 65390 }, { "epoch": 1.53, "learning_rate": 1.867070709441224e-05, "loss": 0.1834, "step": 65395 }, { "epoch": 1.53, "learning_rate": 1.866982703996621e-05, "loss": 0.0833, "step": 65400 }, { "epoch": 1.53, "learning_rate": 1.866894698552017e-05, "loss": 0.2096, "step": 65405 }, { "epoch": 1.53, "learning_rate": 1.8668066931074135e-05, "loss": 0.2209, "step": 65410 }, { "epoch": 1.53, "learning_rate": 1.86671868766281e-05, "loss": 0.2992, "step": 65415 }, { "epoch": 1.53, "learning_rate": 1.866630682218207e-05, "loss": 0.1029, "step": 65420 }, { "epoch": 1.53, "learning_rate": 1.866542676773603e-05, "loss": 0.121, "step": 65425 }, { "epoch": 1.53, "learning_rate": 1.8664546713289995e-05, "loss": 0.0441, "step": 65430 }, { "epoch": 1.53, "learning_rate": 1.866366665884396e-05, "loss": 0.0532, "step": 65435 }, { "epoch": 1.53, "learning_rate": 1.8662786604397925e-05, "loss": 0.0702, "step": 65440 }, { "epoch": 1.53, "learning_rate": 1.866190654995189e-05, "loss": 0.1586, "step": 65445 }, { "epoch": 1.53, "learning_rate": 1.8661026495505855e-05, "loss": 0.271, "step": 65450 }, { "epoch": 1.53, "learning_rate": 1.866014644105982e-05, "loss": 0.1424, "step": 65455 }, { "epoch": 1.53, "learning_rate": 1.8659266386613786e-05, "loss": 0.3359, "step": 65460 }, { "epoch": 1.53, "learning_rate": 1.865838633216775e-05, "loss": 0.2155, "step": 65465 }, { "epoch": 1.53, "learning_rate": 1.8657506277721716e-05, "loss": 0.0389, "step": 65470 }, { "epoch": 1.53, "learning_rate": 1.865662622327568e-05, "loss": 0.0908, "step": 65475 }, { "epoch": 1.53, "learning_rate": 1.8655746168829646e-05, "loss": 0.08, "step": 65480 }, { "epoch": 1.53, "learning_rate": 1.865486611438361e-05, "loss": 0.0155, "step": 65485 }, { "epoch": 1.53, "learning_rate": 1.8653986059937576e-05, "loss": 0.09, "step": 65490 }, { "epoch": 1.53, "learning_rate": 1.865310600549154e-05, "loss": 0.1107, "step": 65495 }, { "epoch": 1.53, "learning_rate": 1.8652225951045506e-05, "loss": 0.1463, "step": 65500 }, { "epoch": 1.53, "learning_rate": 1.865134589659947e-05, "loss": 0.235, "step": 65505 }, { "epoch": 1.53, "learning_rate": 1.8650465842153432e-05, "loss": 0.1782, "step": 65510 }, { "epoch": 1.53, "learning_rate": 1.86495857877074e-05, "loss": 0.3289, "step": 65515 }, { "epoch": 1.53, "learning_rate": 1.8648705733261366e-05, "loss": 0.0189, "step": 65520 }, { "epoch": 1.53, "learning_rate": 1.864782567881533e-05, "loss": 0.0166, "step": 65525 }, { "epoch": 1.53, "learning_rate": 1.8646945624369292e-05, "loss": 0.179, "step": 65530 }, { "epoch": 1.53, "learning_rate": 1.864606556992326e-05, "loss": 0.0956, "step": 65535 }, { "epoch": 1.53, "learning_rate": 1.8645185515477226e-05, "loss": 0.0939, "step": 65540 }, { "epoch": 1.53, "learning_rate": 1.8644305461031187e-05, "loss": 0.137, "step": 65545 }, { "epoch": 1.53, "learning_rate": 1.8643425406585153e-05, "loss": 0.1372, "step": 65550 }, { "epoch": 1.53, "learning_rate": 1.864254535213912e-05, "loss": 0.2867, "step": 65555 }, { "epoch": 1.53, "learning_rate": 1.8641665297693086e-05, "loss": 0.4139, "step": 65560 }, { "epoch": 1.53, "learning_rate": 1.8640785243247048e-05, "loss": 0.194, "step": 65565 }, { "epoch": 1.53, "learning_rate": 1.8639905188801013e-05, "loss": 0.1521, "step": 65570 }, { "epoch": 1.53, "learning_rate": 1.863902513435498e-05, "loss": 0.0879, "step": 65575 }, { "epoch": 1.53, "learning_rate": 1.8638145079908943e-05, "loss": 0.0051, "step": 65580 }, { "epoch": 1.53, "learning_rate": 1.8637265025462908e-05, "loss": 0.0124, "step": 65585 }, { "epoch": 1.53, "learning_rate": 1.8636384971016873e-05, "loss": 0.1101, "step": 65590 }, { "epoch": 1.53, "learning_rate": 1.863550491657084e-05, "loss": 0.204, "step": 65595 }, { "epoch": 1.53, "learning_rate": 1.8634624862124803e-05, "loss": 0.1045, "step": 65600 }, { "epoch": 1.53, "learning_rate": 1.8633744807678768e-05, "loss": 0.198, "step": 65605 }, { "epoch": 1.53, "learning_rate": 1.8632864753232736e-05, "loss": 0.501, "step": 65610 }, { "epoch": 1.53, "learning_rate": 1.8631984698786698e-05, "loss": 0.2373, "step": 65615 }, { "epoch": 1.53, "learning_rate": 1.8631104644340663e-05, "loss": 0.0346, "step": 65620 }, { "epoch": 1.53, "learning_rate": 1.8630224589894628e-05, "loss": 0.0345, "step": 65625 }, { "epoch": 1.53, "learning_rate": 1.8629344535448596e-05, "loss": 0.097, "step": 65630 }, { "epoch": 1.53, "learning_rate": 1.8628464481002558e-05, "loss": 0.0516, "step": 65635 }, { "epoch": 1.53, "learning_rate": 1.8627584426556523e-05, "loss": 0.1806, "step": 65640 }, { "epoch": 1.53, "learning_rate": 1.8626704372110488e-05, "loss": 0.1004, "step": 65645 }, { "epoch": 1.53, "learning_rate": 1.8625824317664456e-05, "loss": 0.0617, "step": 65650 }, { "epoch": 1.53, "learning_rate": 1.8624944263218418e-05, "loss": 0.239, "step": 65655 }, { "epoch": 1.53, "learning_rate": 1.8624064208772383e-05, "loss": 0.136, "step": 65660 }, { "epoch": 1.53, "learning_rate": 1.8623184154326348e-05, "loss": 0.1841, "step": 65665 }, { "epoch": 1.53, "learning_rate": 1.8622304099880313e-05, "loss": 0.0472, "step": 65670 }, { "epoch": 1.53, "learning_rate": 1.8621424045434278e-05, "loss": 0.0786, "step": 65675 }, { "epoch": 1.53, "learning_rate": 1.8620543990988243e-05, "loss": 0.1213, "step": 65680 }, { "epoch": 1.53, "learning_rate": 1.8619663936542208e-05, "loss": 0.0625, "step": 65685 }, { "epoch": 1.53, "learning_rate": 1.8618783882096173e-05, "loss": 0.0848, "step": 65690 }, { "epoch": 1.53, "learning_rate": 1.8617903827650138e-05, "loss": 0.1429, "step": 65695 }, { "epoch": 1.53, "learning_rate": 1.8617023773204103e-05, "loss": 0.0991, "step": 65700 }, { "epoch": 1.53, "learning_rate": 1.8616143718758065e-05, "loss": 0.1971, "step": 65705 }, { "epoch": 1.53, "learning_rate": 1.8615263664312033e-05, "loss": 0.3848, "step": 65710 }, { "epoch": 1.53, "learning_rate": 1.8614383609865998e-05, "loss": 0.1523, "step": 65715 }, { "epoch": 1.53, "learning_rate": 1.8613503555419963e-05, "loss": 0.0065, "step": 65720 }, { "epoch": 1.53, "learning_rate": 1.8612623500973925e-05, "loss": 0.0397, "step": 65725 }, { "epoch": 1.53, "learning_rate": 1.8611743446527893e-05, "loss": 0.0789, "step": 65730 }, { "epoch": 1.53, "learning_rate": 1.8610863392081858e-05, "loss": 0.1624, "step": 65735 }, { "epoch": 1.53, "learning_rate": 1.860998333763582e-05, "loss": 0.0451, "step": 65740 }, { "epoch": 1.53, "learning_rate": 1.860910328318979e-05, "loss": 0.1167, "step": 65745 }, { "epoch": 1.53, "learning_rate": 1.8608223228743753e-05, "loss": 0.1688, "step": 65750 }, { "epoch": 1.53, "learning_rate": 1.860734317429772e-05, "loss": 0.1242, "step": 65755 }, { "epoch": 1.53, "learning_rate": 1.860646311985168e-05, "loss": 0.3278, "step": 65760 }, { "epoch": 1.53, "learning_rate": 1.860558306540565e-05, "loss": 0.445, "step": 65765 }, { "epoch": 1.53, "learning_rate": 1.8604703010959613e-05, "loss": 0.0448, "step": 65770 }, { "epoch": 1.53, "learning_rate": 1.8603822956513575e-05, "loss": 0.1068, "step": 65775 }, { "epoch": 1.53, "learning_rate": 1.860294290206754e-05, "loss": 0.0263, "step": 65780 }, { "epoch": 1.53, "learning_rate": 1.860206284762151e-05, "loss": 0.0677, "step": 65785 }, { "epoch": 1.53, "learning_rate": 1.8601182793175474e-05, "loss": 0.1247, "step": 65790 }, { "epoch": 1.54, "learning_rate": 1.8600302738729435e-05, "loss": 0.2574, "step": 65795 }, { "epoch": 1.54, "learning_rate": 1.85994226842834e-05, "loss": 0.2138, "step": 65800 }, { "epoch": 1.54, "learning_rate": 1.859854262983737e-05, "loss": 0.2294, "step": 65805 }, { "epoch": 1.54, "learning_rate": 1.859766257539133e-05, "loss": 0.3301, "step": 65810 }, { "epoch": 1.54, "learning_rate": 1.8596782520945295e-05, "loss": 0.2511, "step": 65815 }, { "epoch": 1.54, "learning_rate": 1.859590246649926e-05, "loss": 0.086, "step": 65820 }, { "epoch": 1.54, "learning_rate": 1.859502241205323e-05, "loss": 0.0286, "step": 65825 }, { "epoch": 1.54, "learning_rate": 1.859414235760719e-05, "loss": 0.1011, "step": 65830 }, { "epoch": 1.54, "learning_rate": 1.8593262303161155e-05, "loss": 0.0875, "step": 65835 }, { "epoch": 1.54, "learning_rate": 1.859238224871512e-05, "loss": 0.1135, "step": 65840 }, { "epoch": 1.54, "learning_rate": 1.8591502194269085e-05, "loss": 0.1334, "step": 65845 }, { "epoch": 1.54, "learning_rate": 1.859062213982305e-05, "loss": 0.0669, "step": 65850 }, { "epoch": 1.54, "learning_rate": 1.8589742085377015e-05, "loss": 0.1684, "step": 65855 }, { "epoch": 1.54, "learning_rate": 1.858886203093098e-05, "loss": 0.3173, "step": 65860 }, { "epoch": 1.54, "learning_rate": 1.8587981976484945e-05, "loss": 0.3522, "step": 65865 }, { "epoch": 1.54, "learning_rate": 1.858710192203891e-05, "loss": 0.0046, "step": 65870 }, { "epoch": 1.54, "learning_rate": 1.8586221867592875e-05, "loss": 0.0502, "step": 65875 }, { "epoch": 1.54, "learning_rate": 1.858534181314684e-05, "loss": 0.2108, "step": 65880 }, { "epoch": 1.54, "learning_rate": 1.8584461758700806e-05, "loss": 0.0451, "step": 65885 }, { "epoch": 1.54, "learning_rate": 1.858358170425477e-05, "loss": 0.11, "step": 65890 }, { "epoch": 1.54, "learning_rate": 1.8582701649808736e-05, "loss": 0.1063, "step": 65895 }, { "epoch": 1.54, "learning_rate": 1.85818215953627e-05, "loss": 0.2357, "step": 65900 }, { "epoch": 1.54, "learning_rate": 1.8580941540916666e-05, "loss": 0.2514, "step": 65905 }, { "epoch": 1.54, "learning_rate": 1.858006148647063e-05, "loss": 0.2823, "step": 65910 }, { "epoch": 1.54, "learning_rate": 1.8579181432024592e-05, "loss": 0.1546, "step": 65915 }, { "epoch": 1.54, "learning_rate": 1.857830137757856e-05, "loss": 0.0848, "step": 65920 }, { "epoch": 1.54, "learning_rate": 1.8577421323132526e-05, "loss": 0.0326, "step": 65925 }, { "epoch": 1.54, "learning_rate": 1.857654126868649e-05, "loss": 0.0662, "step": 65930 }, { "epoch": 1.54, "learning_rate": 1.8575661214240452e-05, "loss": 0.1491, "step": 65935 }, { "epoch": 1.54, "learning_rate": 1.857478115979442e-05, "loss": 0.0928, "step": 65940 }, { "epoch": 1.54, "learning_rate": 1.8573901105348386e-05, "loss": 0.0709, "step": 65945 }, { "epoch": 1.54, "learning_rate": 1.857302105090235e-05, "loss": 0.1441, "step": 65950 }, { "epoch": 1.54, "learning_rate": 1.8572140996456312e-05, "loss": 0.2898, "step": 65955 }, { "epoch": 1.54, "learning_rate": 1.857126094201028e-05, "loss": 0.3134, "step": 65960 }, { "epoch": 1.54, "learning_rate": 1.8570380887564246e-05, "loss": 0.294, "step": 65965 }, { "epoch": 1.54, "learning_rate": 1.8569500833118207e-05, "loss": 0.0728, "step": 65970 }, { "epoch": 1.54, "learning_rate": 1.8568620778672172e-05, "loss": 0.0673, "step": 65975 }, { "epoch": 1.54, "learning_rate": 1.856774072422614e-05, "loss": 0.0614, "step": 65980 }, { "epoch": 1.54, "learning_rate": 1.8566860669780106e-05, "loss": 0.069, "step": 65985 }, { "epoch": 1.54, "learning_rate": 1.8565980615334068e-05, "loss": 0.1127, "step": 65990 }, { "epoch": 1.54, "learning_rate": 1.8565100560888033e-05, "loss": 0.0936, "step": 65995 }, { "epoch": 1.54, "learning_rate": 1.8564220506442e-05, "loss": 0.0434, "step": 66000 }, { "epoch": 1.54, "learning_rate": 1.8563340451995963e-05, "loss": 0.1558, "step": 66005 }, { "epoch": 1.54, "learning_rate": 1.8562460397549928e-05, "loss": 0.3724, "step": 66010 }, { "epoch": 1.54, "learning_rate": 1.8561580343103896e-05, "loss": 0.3022, "step": 66015 }, { "epoch": 1.54, "learning_rate": 1.856070028865786e-05, "loss": 0.0307, "step": 66020 }, { "epoch": 1.54, "learning_rate": 1.8559820234211823e-05, "loss": 0.0476, "step": 66025 }, { "epoch": 1.54, "learning_rate": 1.8558940179765788e-05, "loss": 0.0449, "step": 66030 }, { "epoch": 1.54, "learning_rate": 1.8558060125319756e-05, "loss": 0.0526, "step": 66035 }, { "epoch": 1.54, "learning_rate": 1.8557180070873718e-05, "loss": 0.1053, "step": 66040 }, { "epoch": 1.54, "learning_rate": 1.8556300016427683e-05, "loss": 0.1105, "step": 66045 }, { "epoch": 1.54, "learning_rate": 1.8555419961981648e-05, "loss": 0.1059, "step": 66050 }, { "epoch": 1.54, "learning_rate": 1.8554539907535616e-05, "loss": 0.1811, "step": 66055 }, { "epoch": 1.54, "learning_rate": 1.8553659853089578e-05, "loss": 0.1166, "step": 66060 }, { "epoch": 1.54, "learning_rate": 1.8552779798643543e-05, "loss": 0.1848, "step": 66065 }, { "epoch": 1.54, "learning_rate": 1.8551899744197508e-05, "loss": 0.1123, "step": 66070 }, { "epoch": 1.54, "learning_rate": 1.8551019689751473e-05, "loss": 0.0933, "step": 66075 }, { "epoch": 1.54, "learning_rate": 1.8550139635305438e-05, "loss": 0.0313, "step": 66080 }, { "epoch": 1.54, "learning_rate": 1.8549259580859403e-05, "loss": 0.0042, "step": 66085 }, { "epoch": 1.54, "learning_rate": 1.8548379526413368e-05, "loss": 0.1156, "step": 66090 }, { "epoch": 1.54, "learning_rate": 1.8547499471967333e-05, "loss": 0.1316, "step": 66095 }, { "epoch": 1.54, "learning_rate": 1.8546619417521298e-05, "loss": 0.094, "step": 66100 }, { "epoch": 1.54, "learning_rate": 1.8545739363075263e-05, "loss": 0.2413, "step": 66105 }, { "epoch": 1.54, "learning_rate": 1.8544859308629225e-05, "loss": 0.3078, "step": 66110 }, { "epoch": 1.54, "learning_rate": 1.8543979254183193e-05, "loss": 0.2218, "step": 66115 }, { "epoch": 1.54, "learning_rate": 1.8543099199737158e-05, "loss": 0.0006, "step": 66120 }, { "epoch": 1.54, "learning_rate": 1.8542219145291123e-05, "loss": 0.0319, "step": 66125 }, { "epoch": 1.54, "learning_rate": 1.8541339090845085e-05, "loss": 0.0428, "step": 66130 }, { "epoch": 1.54, "learning_rate": 1.8540459036399053e-05, "loss": 0.0255, "step": 66135 }, { "epoch": 1.54, "learning_rate": 1.8539578981953018e-05, "loss": 0.0787, "step": 66140 }, { "epoch": 1.54, "learning_rate": 1.853869892750698e-05, "loss": 0.0836, "step": 66145 }, { "epoch": 1.54, "learning_rate": 1.8537818873060948e-05, "loss": 0.1256, "step": 66150 }, { "epoch": 1.54, "learning_rate": 1.8536938818614913e-05, "loss": 0.3017, "step": 66155 }, { "epoch": 1.54, "learning_rate": 1.8536058764168878e-05, "loss": 0.6006, "step": 66160 }, { "epoch": 1.54, "learning_rate": 1.853517870972284e-05, "loss": 0.3206, "step": 66165 }, { "epoch": 1.54, "learning_rate": 1.8534298655276808e-05, "loss": 0.0795, "step": 66170 }, { "epoch": 1.54, "learning_rate": 1.8533418600830773e-05, "loss": 0.0356, "step": 66175 }, { "epoch": 1.54, "learning_rate": 1.8532538546384735e-05, "loss": 0.0581, "step": 66180 }, { "epoch": 1.54, "learning_rate": 1.85316584919387e-05, "loss": 0.0457, "step": 66185 }, { "epoch": 1.54, "learning_rate": 1.853077843749267e-05, "loss": 0.0603, "step": 66190 }, { "epoch": 1.54, "learning_rate": 1.8529898383046633e-05, "loss": 0.1262, "step": 66195 }, { "epoch": 1.54, "learning_rate": 1.8529018328600595e-05, "loss": 0.1669, "step": 66200 }, { "epoch": 1.54, "learning_rate": 1.852813827415456e-05, "loss": 0.1913, "step": 66205 }, { "epoch": 1.54, "learning_rate": 1.852725821970853e-05, "loss": 0.2526, "step": 66210 }, { "epoch": 1.54, "learning_rate": 1.852637816526249e-05, "loss": 0.2617, "step": 66215 }, { "epoch": 1.54, "learning_rate": 1.8525498110816455e-05, "loss": 0.1102, "step": 66220 }, { "epoch": 1.55, "learning_rate": 1.852461805637042e-05, "loss": 0.0376, "step": 66225 }, { "epoch": 1.55, "learning_rate": 1.852373800192439e-05, "loss": 0.1382, "step": 66230 }, { "epoch": 1.55, "learning_rate": 1.852285794747835e-05, "loss": 0.0273, "step": 66235 }, { "epoch": 1.55, "learning_rate": 1.8521977893032315e-05, "loss": 0.1244, "step": 66240 }, { "epoch": 1.55, "learning_rate": 1.852109783858628e-05, "loss": 0.1063, "step": 66245 }, { "epoch": 1.55, "learning_rate": 1.852021778414025e-05, "loss": 0.0771, "step": 66250 }, { "epoch": 1.55, "learning_rate": 1.851933772969421e-05, "loss": 0.2174, "step": 66255 }, { "epoch": 1.55, "learning_rate": 1.8518457675248175e-05, "loss": 0.2794, "step": 66260 }, { "epoch": 1.55, "learning_rate": 1.851757762080214e-05, "loss": 0.369, "step": 66265 }, { "epoch": 1.55, "learning_rate": 1.8516697566356105e-05, "loss": 0.2073, "step": 66270 }, { "epoch": 1.55, "learning_rate": 1.851581751191007e-05, "loss": 0.0028, "step": 66275 }, { "epoch": 1.55, "learning_rate": 1.8514937457464035e-05, "loss": 0.0772, "step": 66280 }, { "epoch": 1.55, "learning_rate": 1.8514057403018004e-05, "loss": 0.1769, "step": 66285 }, { "epoch": 1.55, "learning_rate": 1.8513177348571965e-05, "loss": 0.0569, "step": 66290 }, { "epoch": 1.55, "learning_rate": 1.851229729412593e-05, "loss": 0.36, "step": 66295 }, { "epoch": 1.55, "learning_rate": 1.8511417239679895e-05, "loss": 0.1358, "step": 66300 }, { "epoch": 1.55, "learning_rate": 1.851053718523386e-05, "loss": 0.2823, "step": 66305 }, { "epoch": 1.55, "learning_rate": 1.8509657130787825e-05, "loss": 0.3498, "step": 66310 }, { "epoch": 1.55, "learning_rate": 1.850877707634179e-05, "loss": 0.2707, "step": 66315 }, { "epoch": 1.55, "learning_rate": 1.8507897021895756e-05, "loss": 0.1298, "step": 66320 }, { "epoch": 1.55, "learning_rate": 1.850701696744972e-05, "loss": 0.1052, "step": 66325 }, { "epoch": 1.55, "learning_rate": 1.8506136913003686e-05, "loss": 0.0551, "step": 66330 }, { "epoch": 1.55, "learning_rate": 1.850525685855765e-05, "loss": 0.1342, "step": 66335 }, { "epoch": 1.55, "learning_rate": 1.8504376804111612e-05, "loss": 0.1949, "step": 66340 }, { "epoch": 1.55, "learning_rate": 1.850349674966558e-05, "loss": 0.1655, "step": 66345 }, { "epoch": 1.55, "learning_rate": 1.8502616695219546e-05, "loss": 0.1466, "step": 66350 }, { "epoch": 1.55, "learning_rate": 1.850173664077351e-05, "loss": 0.2059, "step": 66355 }, { "epoch": 1.55, "learning_rate": 1.8500856586327472e-05, "loss": 0.3642, "step": 66360 }, { "epoch": 1.55, "learning_rate": 1.849997653188144e-05, "loss": 0.2251, "step": 66365 }, { "epoch": 1.55, "learning_rate": 1.8499096477435406e-05, "loss": 0.1119, "step": 66370 }, { "epoch": 1.55, "learning_rate": 1.8498216422989367e-05, "loss": 0.2186, "step": 66375 }, { "epoch": 1.55, "learning_rate": 1.8497336368543332e-05, "loss": 0.0702, "step": 66380 }, { "epoch": 1.55, "learning_rate": 1.84964563140973e-05, "loss": 0.0498, "step": 66385 }, { "epoch": 1.55, "learning_rate": 1.8495576259651266e-05, "loss": 0.1091, "step": 66390 }, { "epoch": 1.55, "learning_rate": 1.8494696205205227e-05, "loss": 0.1287, "step": 66395 }, { "epoch": 1.55, "learning_rate": 1.8493816150759192e-05, "loss": 0.219, "step": 66400 }, { "epoch": 1.55, "learning_rate": 1.849293609631316e-05, "loss": 0.3554, "step": 66405 }, { "epoch": 1.55, "learning_rate": 1.8492056041867122e-05, "loss": 0.3615, "step": 66410 }, { "epoch": 1.55, "learning_rate": 1.8491175987421088e-05, "loss": 0.3242, "step": 66415 }, { "epoch": 1.55, "learning_rate": 1.8490295932975056e-05, "loss": 0.058, "step": 66420 }, { "epoch": 1.55, "learning_rate": 1.848941587852902e-05, "loss": 0.0202, "step": 66425 }, { "epoch": 1.55, "learning_rate": 1.8488535824082983e-05, "loss": 0.0548, "step": 66430 }, { "epoch": 1.55, "learning_rate": 1.8487655769636948e-05, "loss": 0.1322, "step": 66435 }, { "epoch": 1.55, "learning_rate": 1.8486775715190916e-05, "loss": 0.1173, "step": 66440 }, { "epoch": 1.55, "learning_rate": 1.8485895660744878e-05, "loss": 0.1553, "step": 66445 }, { "epoch": 1.55, "learning_rate": 1.8485015606298843e-05, "loss": 0.1504, "step": 66450 }, { "epoch": 1.55, "learning_rate": 1.8484135551852808e-05, "loss": 0.313, "step": 66455 }, { "epoch": 1.55, "learning_rate": 1.8483255497406776e-05, "loss": 0.2203, "step": 66460 }, { "epoch": 1.55, "learning_rate": 1.8482375442960738e-05, "loss": 0.1951, "step": 66465 }, { "epoch": 1.55, "learning_rate": 1.8481495388514703e-05, "loss": 0.0437, "step": 66470 }, { "epoch": 1.55, "learning_rate": 1.8480615334068668e-05, "loss": 0.0336, "step": 66475 }, { "epoch": 1.55, "learning_rate": 1.8479735279622633e-05, "loss": 0.0782, "step": 66480 }, { "epoch": 1.55, "learning_rate": 1.8478855225176598e-05, "loss": 0.2446, "step": 66485 }, { "epoch": 1.55, "learning_rate": 1.8477975170730563e-05, "loss": 0.1908, "step": 66490 }, { "epoch": 1.55, "learning_rate": 1.8477095116284528e-05, "loss": 0.0867, "step": 66495 }, { "epoch": 1.55, "learning_rate": 1.8476215061838493e-05, "loss": 0.0841, "step": 66500 }, { "epoch": 1.55, "learning_rate": 1.8475335007392458e-05, "loss": 0.2863, "step": 66505 }, { "epoch": 1.55, "learning_rate": 1.8474454952946423e-05, "loss": 0.2678, "step": 66510 }, { "epoch": 1.55, "learning_rate": 1.8473574898500385e-05, "loss": 0.2843, "step": 66515 }, { "epoch": 1.55, "learning_rate": 1.8472694844054353e-05, "loss": 0.0249, "step": 66520 }, { "epoch": 1.55, "learning_rate": 1.8471814789608318e-05, "loss": 0.0442, "step": 66525 }, { "epoch": 1.55, "learning_rate": 1.8470934735162283e-05, "loss": 0.0664, "step": 66530 }, { "epoch": 1.55, "learning_rate": 1.8470054680716245e-05, "loss": 0.1185, "step": 66535 }, { "epoch": 1.55, "learning_rate": 1.8469174626270213e-05, "loss": 0.1371, "step": 66540 }, { "epoch": 1.55, "learning_rate": 1.8468294571824178e-05, "loss": 0.0735, "step": 66545 }, { "epoch": 1.55, "learning_rate": 1.8467414517378143e-05, "loss": 0.21, "step": 66550 }, { "epoch": 1.55, "learning_rate": 1.8466534462932108e-05, "loss": 0.2187, "step": 66555 }, { "epoch": 1.55, "learning_rate": 1.8465654408486073e-05, "loss": 0.35, "step": 66560 }, { "epoch": 1.55, "learning_rate": 1.8464774354040038e-05, "loss": 0.2668, "step": 66565 }, { "epoch": 1.55, "learning_rate": 1.8463894299594e-05, "loss": 0.0695, "step": 66570 }, { "epoch": 1.55, "learning_rate": 1.8463014245147968e-05, "loss": 0.0928, "step": 66575 }, { "epoch": 1.55, "learning_rate": 1.8462134190701933e-05, "loss": 0.0785, "step": 66580 }, { "epoch": 1.55, "learning_rate": 1.8461254136255898e-05, "loss": 0.0054, "step": 66585 }, { "epoch": 1.55, "learning_rate": 1.846037408180986e-05, "loss": 0.1172, "step": 66590 }, { "epoch": 1.55, "learning_rate": 1.8459494027363828e-05, "loss": 0.1474, "step": 66595 }, { "epoch": 1.55, "learning_rate": 1.8458613972917793e-05, "loss": 0.2289, "step": 66600 }, { "epoch": 1.55, "learning_rate": 1.8457733918471755e-05, "loss": 0.1353, "step": 66605 }, { "epoch": 1.55, "learning_rate": 1.845685386402572e-05, "loss": 0.5279, "step": 66610 }, { "epoch": 1.55, "learning_rate": 1.845597380957969e-05, "loss": 0.2809, "step": 66615 }, { "epoch": 1.55, "learning_rate": 1.8455093755133653e-05, "loss": 0.0594, "step": 66620 }, { "epoch": 1.55, "learning_rate": 1.8454213700687615e-05, "loss": 0.0168, "step": 66625 }, { "epoch": 1.55, "learning_rate": 1.845333364624158e-05, "loss": 0.0457, "step": 66630 }, { "epoch": 1.55, "learning_rate": 1.845245359179555e-05, "loss": 0.0985, "step": 66635 }, { "epoch": 1.55, "learning_rate": 1.845157353734951e-05, "loss": 0.0949, "step": 66640 }, { "epoch": 1.55, "learning_rate": 1.8450693482903475e-05, "loss": 0.1541, "step": 66645 }, { "epoch": 1.56, "learning_rate": 1.844981342845744e-05, "loss": 0.2176, "step": 66650 }, { "epoch": 1.56, "learning_rate": 1.844893337401141e-05, "loss": 0.1749, "step": 66655 }, { "epoch": 1.56, "learning_rate": 1.844805331956537e-05, "loss": 0.2522, "step": 66660 }, { "epoch": 1.56, "learning_rate": 1.8447173265119335e-05, "loss": 0.2072, "step": 66665 }, { "epoch": 1.56, "learning_rate": 1.84462932106733e-05, "loss": 0.1284, "step": 66670 }, { "epoch": 1.56, "learning_rate": 1.8445413156227265e-05, "loss": 0.0692, "step": 66675 }, { "epoch": 1.56, "learning_rate": 1.844453310178123e-05, "loss": 0.1314, "step": 66680 }, { "epoch": 1.56, "learning_rate": 1.8443653047335195e-05, "loss": 0.0389, "step": 66685 }, { "epoch": 1.56, "learning_rate": 1.8442772992889164e-05, "loss": 0.1591, "step": 66690 }, { "epoch": 1.56, "learning_rate": 1.8441892938443125e-05, "loss": 0.0406, "step": 66695 }, { "epoch": 1.56, "learning_rate": 1.844101288399709e-05, "loss": 0.2599, "step": 66700 }, { "epoch": 1.56, "learning_rate": 1.8440132829551055e-05, "loss": 0.2226, "step": 66705 }, { "epoch": 1.56, "learning_rate": 1.843925277510502e-05, "loss": 0.6187, "step": 66710 }, { "epoch": 1.56, "learning_rate": 1.8438372720658985e-05, "loss": 0.1347, "step": 66715 }, { "epoch": 1.56, "learning_rate": 1.843749266621295e-05, "loss": 0.0377, "step": 66720 }, { "epoch": 1.56, "learning_rate": 1.8436612611766915e-05, "loss": 0.0412, "step": 66725 }, { "epoch": 1.56, "learning_rate": 1.843573255732088e-05, "loss": 0.047, "step": 66730 }, { "epoch": 1.56, "learning_rate": 1.8434852502874845e-05, "loss": 0.0669, "step": 66735 }, { "epoch": 1.56, "learning_rate": 1.843397244842881e-05, "loss": 0.0776, "step": 66740 }, { "epoch": 1.56, "learning_rate": 1.8433092393982772e-05, "loss": 0.1024, "step": 66745 }, { "epoch": 1.56, "learning_rate": 1.843221233953674e-05, "loss": 0.1161, "step": 66750 }, { "epoch": 1.56, "learning_rate": 1.8431332285090706e-05, "loss": 0.1167, "step": 66755 }, { "epoch": 1.56, "learning_rate": 1.843045223064467e-05, "loss": 0.3813, "step": 66760 }, { "epoch": 1.56, "learning_rate": 1.8429572176198632e-05, "loss": 0.4253, "step": 66765 }, { "epoch": 1.56, "learning_rate": 1.84286921217526e-05, "loss": 0.0543, "step": 66770 }, { "epoch": 1.56, "learning_rate": 1.8427812067306566e-05, "loss": 0.0184, "step": 66775 }, { "epoch": 1.56, "learning_rate": 1.8426932012860527e-05, "loss": 0.094, "step": 66780 }, { "epoch": 1.56, "learning_rate": 1.8426051958414492e-05, "loss": 0.0886, "step": 66785 }, { "epoch": 1.56, "learning_rate": 1.842517190396846e-05, "loss": 0.0678, "step": 66790 }, { "epoch": 1.56, "learning_rate": 1.8424291849522426e-05, "loss": 0.0291, "step": 66795 }, { "epoch": 1.56, "learning_rate": 1.8423411795076387e-05, "loss": 0.1628, "step": 66800 }, { "epoch": 1.56, "learning_rate": 1.8422531740630352e-05, "loss": 0.2571, "step": 66805 }, { "epoch": 1.56, "learning_rate": 1.842165168618432e-05, "loss": 0.3179, "step": 66810 }, { "epoch": 1.56, "learning_rate": 1.8420771631738282e-05, "loss": 0.1416, "step": 66815 }, { "epoch": 1.56, "learning_rate": 1.8419891577292247e-05, "loss": 0.0329, "step": 66820 }, { "epoch": 1.56, "learning_rate": 1.8419011522846216e-05, "loss": 0.0611, "step": 66825 }, { "epoch": 1.56, "learning_rate": 1.841813146840018e-05, "loss": 0.099, "step": 66830 }, { "epoch": 1.56, "learning_rate": 1.8417251413954142e-05, "loss": 0.061, "step": 66835 }, { "epoch": 1.56, "learning_rate": 1.8416371359508107e-05, "loss": 0.149, "step": 66840 }, { "epoch": 1.56, "learning_rate": 1.8415491305062076e-05, "loss": 0.1933, "step": 66845 }, { "epoch": 1.56, "learning_rate": 1.841461125061604e-05, "loss": 0.1724, "step": 66850 }, { "epoch": 1.56, "learning_rate": 1.8413731196170003e-05, "loss": 0.3023, "step": 66855 }, { "epoch": 1.56, "learning_rate": 1.8412851141723968e-05, "loss": 0.3011, "step": 66860 }, { "epoch": 1.56, "learning_rate": 1.8411971087277936e-05, "loss": 0.2182, "step": 66865 }, { "epoch": 1.56, "learning_rate": 1.8411091032831898e-05, "loss": 0.052, "step": 66870 }, { "epoch": 1.56, "learning_rate": 1.8410210978385863e-05, "loss": 0.0129, "step": 66875 }, { "epoch": 1.56, "learning_rate": 1.8409330923939828e-05, "loss": 0.0961, "step": 66880 }, { "epoch": 1.56, "learning_rate": 1.8408450869493796e-05, "loss": 0.0641, "step": 66885 }, { "epoch": 1.56, "learning_rate": 1.8407570815047758e-05, "loss": 0.0668, "step": 66890 }, { "epoch": 1.56, "learning_rate": 1.8406690760601723e-05, "loss": 0.1202, "step": 66895 }, { "epoch": 1.56, "learning_rate": 1.8405810706155688e-05, "loss": 0.2027, "step": 66900 }, { "epoch": 1.56, "learning_rate": 1.8404930651709653e-05, "loss": 0.1949, "step": 66905 }, { "epoch": 1.56, "learning_rate": 1.8404050597263618e-05, "loss": 0.2906, "step": 66910 }, { "epoch": 1.56, "learning_rate": 1.8403170542817583e-05, "loss": 0.3278, "step": 66915 }, { "epoch": 1.56, "learning_rate": 1.8402290488371548e-05, "loss": 0.0518, "step": 66920 }, { "epoch": 1.56, "learning_rate": 1.8401410433925513e-05, "loss": 0.0237, "step": 66925 }, { "epoch": 1.56, "learning_rate": 1.8400530379479478e-05, "loss": 0.0728, "step": 66930 }, { "epoch": 1.56, "learning_rate": 1.8399650325033443e-05, "loss": 0.0765, "step": 66935 }, { "epoch": 1.56, "learning_rate": 1.8398770270587405e-05, "loss": 0.209, "step": 66940 }, { "epoch": 1.56, "learning_rate": 1.8397890216141373e-05, "loss": 0.1617, "step": 66945 }, { "epoch": 1.56, "learning_rate": 1.8397010161695338e-05, "loss": 0.2206, "step": 66950 }, { "epoch": 1.56, "learning_rate": 1.8396130107249303e-05, "loss": 0.1824, "step": 66955 }, { "epoch": 1.56, "learning_rate": 1.8395250052803268e-05, "loss": 0.3269, "step": 66960 }, { "epoch": 1.56, "learning_rate": 1.8394369998357233e-05, "loss": 0.4103, "step": 66965 }, { "epoch": 1.56, "learning_rate": 1.8393489943911198e-05, "loss": 0.1198, "step": 66970 }, { "epoch": 1.56, "learning_rate": 1.839260988946516e-05, "loss": 0.0515, "step": 66975 }, { "epoch": 1.56, "learning_rate": 1.8391729835019128e-05, "loss": 0.1718, "step": 66980 }, { "epoch": 1.56, "learning_rate": 1.8390849780573093e-05, "loss": 0.0558, "step": 66985 }, { "epoch": 1.56, "learning_rate": 1.8389969726127058e-05, "loss": 0.1641, "step": 66990 }, { "epoch": 1.56, "learning_rate": 1.838908967168102e-05, "loss": 0.2622, "step": 66995 }, { "epoch": 1.56, "learning_rate": 1.8388209617234988e-05, "loss": 0.2942, "step": 67000 }, { "epoch": 1.56, "learning_rate": 1.8387329562788953e-05, "loss": 0.2718, "step": 67005 }, { "epoch": 1.56, "learning_rate": 1.8386449508342915e-05, "loss": 0.3075, "step": 67010 }, { "epoch": 1.56, "learning_rate": 1.838556945389688e-05, "loss": 0.2167, "step": 67015 }, { "epoch": 1.56, "learning_rate": 1.8384689399450848e-05, "loss": 0.0263, "step": 67020 }, { "epoch": 1.56, "learning_rate": 1.8383809345004813e-05, "loss": 0.0686, "step": 67025 }, { "epoch": 1.56, "learning_rate": 1.8382929290558775e-05, "loss": 0.0948, "step": 67030 }, { "epoch": 1.56, "learning_rate": 1.838204923611274e-05, "loss": 0.1068, "step": 67035 }, { "epoch": 1.56, "learning_rate": 1.8381169181666708e-05, "loss": 0.152, "step": 67040 }, { "epoch": 1.56, "learning_rate": 1.838028912722067e-05, "loss": 0.1331, "step": 67045 }, { "epoch": 1.56, "learning_rate": 1.8379409072774635e-05, "loss": 0.0747, "step": 67050 }, { "epoch": 1.56, "learning_rate": 1.83785290183286e-05, "loss": 0.3552, "step": 67055 }, { "epoch": 1.56, "learning_rate": 1.837764896388257e-05, "loss": 0.1838, "step": 67060 }, { "epoch": 1.56, "learning_rate": 1.837676890943653e-05, "loss": 0.2636, "step": 67065 }, { "epoch": 1.56, "learning_rate": 1.8375888854990495e-05, "loss": 0.0481, "step": 67070 }, { "epoch": 1.56, "learning_rate": 1.837500880054446e-05, "loss": 0.112, "step": 67075 }, { "epoch": 1.57, "learning_rate": 1.8374128746098425e-05, "loss": 0.0602, "step": 67080 }, { "epoch": 1.57, "learning_rate": 1.837324869165239e-05, "loss": 0.0784, "step": 67085 }, { "epoch": 1.57, "learning_rate": 1.8372368637206355e-05, "loss": 0.1359, "step": 67090 }, { "epoch": 1.57, "learning_rate": 1.8371488582760324e-05, "loss": 0.1534, "step": 67095 }, { "epoch": 1.57, "learning_rate": 1.8370608528314285e-05, "loss": 0.1296, "step": 67100 }, { "epoch": 1.57, "learning_rate": 1.836972847386825e-05, "loss": 0.106, "step": 67105 }, { "epoch": 1.57, "learning_rate": 1.8368848419422215e-05, "loss": 0.1854, "step": 67110 }, { "epoch": 1.57, "learning_rate": 1.836796836497618e-05, "loss": 0.1284, "step": 67115 }, { "epoch": 1.57, "learning_rate": 1.8367088310530145e-05, "loss": 0.0795, "step": 67120 }, { "epoch": 1.57, "learning_rate": 1.836620825608411e-05, "loss": 0.07, "step": 67125 }, { "epoch": 1.57, "learning_rate": 1.8365328201638075e-05, "loss": 0.045, "step": 67130 }, { "epoch": 1.57, "learning_rate": 1.836444814719204e-05, "loss": 0.0481, "step": 67135 }, { "epoch": 1.57, "learning_rate": 1.8363568092746005e-05, "loss": 0.1489, "step": 67140 }, { "epoch": 1.57, "learning_rate": 1.836268803829997e-05, "loss": 0.0618, "step": 67145 }, { "epoch": 1.57, "learning_rate": 1.8361807983853935e-05, "loss": 0.1712, "step": 67150 }, { "epoch": 1.57, "learning_rate": 1.83609279294079e-05, "loss": 0.2318, "step": 67155 }, { "epoch": 1.57, "learning_rate": 1.8360047874961865e-05, "loss": 0.3443, "step": 67160 }, { "epoch": 1.57, "learning_rate": 1.835916782051583e-05, "loss": 0.2973, "step": 67165 }, { "epoch": 1.57, "learning_rate": 1.8358287766069792e-05, "loss": 0.0908, "step": 67170 }, { "epoch": 1.57, "learning_rate": 1.835740771162376e-05, "loss": 0.0766, "step": 67175 }, { "epoch": 1.57, "learning_rate": 1.8356527657177725e-05, "loss": 0.1403, "step": 67180 }, { "epoch": 1.57, "learning_rate": 1.835564760273169e-05, "loss": 0.0751, "step": 67185 }, { "epoch": 1.57, "learning_rate": 1.8354767548285652e-05, "loss": 0.1638, "step": 67190 }, { "epoch": 1.57, "learning_rate": 1.835388749383962e-05, "loss": 0.1702, "step": 67195 }, { "epoch": 1.57, "learning_rate": 1.8353007439393586e-05, "loss": 0.2887, "step": 67200 }, { "epoch": 1.57, "learning_rate": 1.8352127384947547e-05, "loss": 0.1735, "step": 67205 }, { "epoch": 1.57, "learning_rate": 1.8351247330501512e-05, "loss": 0.3521, "step": 67210 }, { "epoch": 1.57, "learning_rate": 1.835036727605548e-05, "loss": 0.2527, "step": 67215 }, { "epoch": 1.57, "learning_rate": 1.8349487221609446e-05, "loss": 0.0519, "step": 67220 }, { "epoch": 1.57, "learning_rate": 1.8348607167163407e-05, "loss": 0.0419, "step": 67225 }, { "epoch": 1.57, "learning_rate": 1.8347727112717376e-05, "loss": 0.0195, "step": 67230 }, { "epoch": 1.57, "learning_rate": 1.834684705827134e-05, "loss": 0.1396, "step": 67235 }, { "epoch": 1.57, "learning_rate": 1.8345967003825302e-05, "loss": 0.0426, "step": 67240 }, { "epoch": 1.57, "learning_rate": 1.8345086949379267e-05, "loss": 0.1645, "step": 67245 }, { "epoch": 1.57, "learning_rate": 1.8344206894933236e-05, "loss": 0.2107, "step": 67250 }, { "epoch": 1.57, "learning_rate": 1.83433268404872e-05, "loss": 0.3535, "step": 67255 }, { "epoch": 1.57, "learning_rate": 1.8342446786041162e-05, "loss": 0.4493, "step": 67260 }, { "epoch": 1.57, "learning_rate": 1.8341566731595127e-05, "loss": 0.215, "step": 67265 }, { "epoch": 1.57, "learning_rate": 1.8340686677149096e-05, "loss": 0.1115, "step": 67270 }, { "epoch": 1.57, "learning_rate": 1.8339806622703057e-05, "loss": 0.0034, "step": 67275 }, { "epoch": 1.57, "learning_rate": 1.8338926568257023e-05, "loss": 0.0398, "step": 67280 }, { "epoch": 1.57, "learning_rate": 1.8338046513810988e-05, "loss": 0.0757, "step": 67285 }, { "epoch": 1.57, "learning_rate": 1.8337166459364956e-05, "loss": 0.1194, "step": 67290 }, { "epoch": 1.57, "learning_rate": 1.8336286404918918e-05, "loss": 0.1376, "step": 67295 }, { "epoch": 1.57, "learning_rate": 1.8335406350472883e-05, "loss": 0.1934, "step": 67300 }, { "epoch": 1.57, "learning_rate": 1.8334526296026848e-05, "loss": 0.2236, "step": 67305 }, { "epoch": 1.57, "learning_rate": 1.8333646241580813e-05, "loss": 0.3206, "step": 67310 }, { "epoch": 1.57, "learning_rate": 1.8332766187134778e-05, "loss": 0.2082, "step": 67315 }, { "epoch": 1.57, "learning_rate": 1.8331886132688743e-05, "loss": 0.0509, "step": 67320 }, { "epoch": 1.57, "learning_rate": 1.8331006078242708e-05, "loss": 0.0122, "step": 67325 }, { "epoch": 1.57, "learning_rate": 1.8330126023796673e-05, "loss": 0.0319, "step": 67330 }, { "epoch": 1.57, "learning_rate": 1.8329245969350638e-05, "loss": 0.0116, "step": 67335 }, { "epoch": 1.57, "learning_rate": 1.8328365914904603e-05, "loss": 0.2228, "step": 67340 }, { "epoch": 1.57, "learning_rate": 1.8327485860458564e-05, "loss": 0.3282, "step": 67345 }, { "epoch": 1.57, "learning_rate": 1.8326605806012533e-05, "loss": 0.0695, "step": 67350 }, { "epoch": 1.57, "learning_rate": 1.8325725751566498e-05, "loss": 0.1062, "step": 67355 }, { "epoch": 1.57, "learning_rate": 1.8324845697120463e-05, "loss": 0.2831, "step": 67360 }, { "epoch": 1.57, "learning_rate": 1.8323965642674428e-05, "loss": 0.2117, "step": 67365 }, { "epoch": 1.57, "learning_rate": 1.8323085588228393e-05, "loss": 0.0743, "step": 67370 }, { "epoch": 1.57, "learning_rate": 1.8322205533782358e-05, "loss": 0.074, "step": 67375 }, { "epoch": 1.57, "learning_rate": 1.832132547933632e-05, "loss": 0.1404, "step": 67380 }, { "epoch": 1.57, "learning_rate": 1.8320445424890288e-05, "loss": 0.0947, "step": 67385 }, { "epoch": 1.57, "learning_rate": 1.8319565370444253e-05, "loss": 0.1023, "step": 67390 }, { "epoch": 1.57, "learning_rate": 1.8318685315998218e-05, "loss": 0.127, "step": 67395 }, { "epoch": 1.57, "learning_rate": 1.831780526155218e-05, "loss": 0.261, "step": 67400 }, { "epoch": 1.57, "learning_rate": 1.8316925207106148e-05, "loss": 0.1477, "step": 67405 }, { "epoch": 1.57, "learning_rate": 1.8316045152660113e-05, "loss": 0.2258, "step": 67410 }, { "epoch": 1.57, "learning_rate": 1.8315165098214075e-05, "loss": 0.2913, "step": 67415 }, { "epoch": 1.57, "learning_rate": 1.831428504376804e-05, "loss": 0.0962, "step": 67420 }, { "epoch": 1.57, "learning_rate": 1.8313404989322008e-05, "loss": 0.0244, "step": 67425 }, { "epoch": 1.57, "learning_rate": 1.8312524934875973e-05, "loss": 0.0421, "step": 67430 }, { "epoch": 1.57, "learning_rate": 1.8311644880429935e-05, "loss": 0.1589, "step": 67435 }, { "epoch": 1.57, "learning_rate": 1.83107648259839e-05, "loss": 0.1027, "step": 67440 }, { "epoch": 1.57, "learning_rate": 1.8309884771537868e-05, "loss": 0.1286, "step": 67445 }, { "epoch": 1.57, "learning_rate": 1.830900471709183e-05, "loss": 0.281, "step": 67450 }, { "epoch": 1.57, "learning_rate": 1.8308124662645795e-05, "loss": 0.2689, "step": 67455 }, { "epoch": 1.57, "learning_rate": 1.830724460819976e-05, "loss": 0.3135, "step": 67460 }, { "epoch": 1.57, "learning_rate": 1.8306364553753728e-05, "loss": 0.2522, "step": 67465 }, { "epoch": 1.57, "learning_rate": 1.830548449930769e-05, "loss": 0.0266, "step": 67470 }, { "epoch": 1.57, "learning_rate": 1.8304604444861655e-05, "loss": 0.0057, "step": 67475 }, { "epoch": 1.57, "learning_rate": 1.830372439041562e-05, "loss": 0.0774, "step": 67480 }, { "epoch": 1.57, "learning_rate": 1.830284433596959e-05, "loss": 0.1285, "step": 67485 }, { "epoch": 1.57, "learning_rate": 1.830196428152355e-05, "loss": 0.1818, "step": 67490 }, { "epoch": 1.57, "learning_rate": 1.8301084227077515e-05, "loss": 0.141, "step": 67495 }, { "epoch": 1.57, "learning_rate": 1.830020417263148e-05, "loss": 0.2906, "step": 67500 }, { "epoch": 1.57, "learning_rate": 1.8299324118185445e-05, "loss": 0.1633, "step": 67505 }, { "epoch": 1.58, "learning_rate": 1.829844406373941e-05, "loss": 0.2584, "step": 67510 }, { "epoch": 1.58, "learning_rate": 1.8297564009293375e-05, "loss": 0.3714, "step": 67515 }, { "epoch": 1.58, "learning_rate": 1.8296683954847344e-05, "loss": 0.084, "step": 67520 }, { "epoch": 1.58, "learning_rate": 1.8295803900401305e-05, "loss": 0.0594, "step": 67525 }, { "epoch": 1.58, "learning_rate": 1.829492384595527e-05, "loss": 0.049, "step": 67530 }, { "epoch": 1.58, "learning_rate": 1.8294043791509235e-05, "loss": 0.159, "step": 67535 }, { "epoch": 1.58, "learning_rate": 1.82931637370632e-05, "loss": 0.0385, "step": 67540 }, { "epoch": 1.58, "learning_rate": 1.8292283682617165e-05, "loss": 0.1373, "step": 67545 }, { "epoch": 1.58, "learning_rate": 1.829140362817113e-05, "loss": 0.1207, "step": 67550 }, { "epoch": 1.58, "learning_rate": 1.8290523573725095e-05, "loss": 0.1993, "step": 67555 }, { "epoch": 1.58, "learning_rate": 1.828964351927906e-05, "loss": 0.1763, "step": 67560 }, { "epoch": 1.58, "learning_rate": 1.8288763464833025e-05, "loss": 0.195, "step": 67565 }, { "epoch": 1.58, "learning_rate": 1.828788341038699e-05, "loss": 0.0319, "step": 67570 }, { "epoch": 1.58, "learning_rate": 1.8287003355940952e-05, "loss": 0.0534, "step": 67575 }, { "epoch": 1.58, "learning_rate": 1.828612330149492e-05, "loss": 0.0147, "step": 67580 }, { "epoch": 1.58, "learning_rate": 1.8285243247048885e-05, "loss": 0.0823, "step": 67585 }, { "epoch": 1.58, "learning_rate": 1.828436319260285e-05, "loss": 0.1305, "step": 67590 }, { "epoch": 1.58, "learning_rate": 1.8283483138156812e-05, "loss": 0.1716, "step": 67595 }, { "epoch": 1.58, "learning_rate": 1.828260308371078e-05, "loss": 0.1998, "step": 67600 }, { "epoch": 1.58, "learning_rate": 1.8281723029264745e-05, "loss": 0.2392, "step": 67605 }, { "epoch": 1.58, "learning_rate": 1.8280842974818707e-05, "loss": 0.23, "step": 67610 }, { "epoch": 1.58, "learning_rate": 1.8279962920372672e-05, "loss": 0.1474, "step": 67615 }, { "epoch": 1.58, "learning_rate": 1.827908286592664e-05, "loss": 0.008, "step": 67620 }, { "epoch": 1.58, "learning_rate": 1.8278202811480606e-05, "loss": 0.0631, "step": 67625 }, { "epoch": 1.58, "learning_rate": 1.8277322757034567e-05, "loss": 0.0959, "step": 67630 }, { "epoch": 1.58, "learning_rate": 1.8276442702588536e-05, "loss": 0.1488, "step": 67635 }, { "epoch": 1.58, "learning_rate": 1.82755626481425e-05, "loss": 0.0641, "step": 67640 }, { "epoch": 1.58, "learning_rate": 1.8274682593696462e-05, "loss": 0.1894, "step": 67645 }, { "epoch": 1.58, "learning_rate": 1.8273802539250427e-05, "loss": 0.1496, "step": 67650 }, { "epoch": 1.58, "learning_rate": 1.8272922484804396e-05, "loss": 0.3329, "step": 67655 }, { "epoch": 1.58, "learning_rate": 1.827204243035836e-05, "loss": 0.3487, "step": 67660 }, { "epoch": 1.58, "learning_rate": 1.8271162375912322e-05, "loss": 0.2765, "step": 67665 }, { "epoch": 1.58, "learning_rate": 1.8270282321466287e-05, "loss": 0.0605, "step": 67670 }, { "epoch": 1.58, "learning_rate": 1.8269402267020256e-05, "loss": 0.0221, "step": 67675 }, { "epoch": 1.58, "learning_rate": 1.8268522212574217e-05, "loss": 0.0988, "step": 67680 }, { "epoch": 1.58, "learning_rate": 1.8267642158128182e-05, "loss": 0.054, "step": 67685 }, { "epoch": 1.58, "learning_rate": 1.8266762103682147e-05, "loss": 0.0316, "step": 67690 }, { "epoch": 1.58, "learning_rate": 1.8265882049236116e-05, "loss": 0.0577, "step": 67695 }, { "epoch": 1.58, "learning_rate": 1.8265001994790077e-05, "loss": 0.2199, "step": 67700 }, { "epoch": 1.58, "learning_rate": 1.8264121940344042e-05, "loss": 0.2494, "step": 67705 }, { "epoch": 1.58, "learning_rate": 1.8263241885898008e-05, "loss": 0.2316, "step": 67710 }, { "epoch": 1.58, "learning_rate": 1.8262361831451973e-05, "loss": 0.2779, "step": 67715 }, { "epoch": 1.58, "learning_rate": 1.8261481777005938e-05, "loss": 0.0829, "step": 67720 }, { "epoch": 1.58, "learning_rate": 1.8260601722559903e-05, "loss": 0.0345, "step": 67725 }, { "epoch": 1.58, "learning_rate": 1.8259721668113868e-05, "loss": 0.041, "step": 67730 }, { "epoch": 1.58, "learning_rate": 1.8258841613667833e-05, "loss": 0.0547, "step": 67735 }, { "epoch": 1.58, "learning_rate": 1.8257961559221798e-05, "loss": 0.0374, "step": 67740 }, { "epoch": 1.58, "learning_rate": 1.8257081504775763e-05, "loss": 0.197, "step": 67745 }, { "epoch": 1.58, "learning_rate": 1.8256201450329724e-05, "loss": 0.2218, "step": 67750 }, { "epoch": 1.58, "learning_rate": 1.8255321395883693e-05, "loss": 0.1625, "step": 67755 }, { "epoch": 1.58, "learning_rate": 1.8254441341437658e-05, "loss": 0.3527, "step": 67760 }, { "epoch": 1.58, "learning_rate": 1.8253561286991623e-05, "loss": 0.1736, "step": 67765 }, { "epoch": 1.58, "learning_rate": 1.8252681232545588e-05, "loss": 0.0459, "step": 67770 }, { "epoch": 1.58, "learning_rate": 1.8251801178099553e-05, "loss": 0.055, "step": 67775 }, { "epoch": 1.58, "learning_rate": 1.8250921123653518e-05, "loss": 0.0265, "step": 67780 }, { "epoch": 1.58, "learning_rate": 1.8250041069207483e-05, "loss": 0.1013, "step": 67785 }, { "epoch": 1.58, "learning_rate": 1.8249161014761448e-05, "loss": 0.1616, "step": 67790 }, { "epoch": 1.58, "learning_rate": 1.8248280960315413e-05, "loss": 0.0936, "step": 67795 }, { "epoch": 1.58, "learning_rate": 1.8247400905869378e-05, "loss": 0.3206, "step": 67800 }, { "epoch": 1.58, "learning_rate": 1.824652085142334e-05, "loss": 0.2902, "step": 67805 }, { "epoch": 1.58, "learning_rate": 1.8245640796977308e-05, "loss": 0.341, "step": 67810 }, { "epoch": 1.58, "learning_rate": 1.8244760742531273e-05, "loss": 0.3134, "step": 67815 }, { "epoch": 1.58, "learning_rate": 1.8243880688085238e-05, "loss": 0.0634, "step": 67820 }, { "epoch": 1.58, "learning_rate": 1.82430006336392e-05, "loss": 0.1691, "step": 67825 }, { "epoch": 1.58, "learning_rate": 1.8242120579193168e-05, "loss": 0.0348, "step": 67830 }, { "epoch": 1.58, "learning_rate": 1.8241240524747133e-05, "loss": 0.0697, "step": 67835 }, { "epoch": 1.58, "learning_rate": 1.8240360470301095e-05, "loss": 0.1334, "step": 67840 }, { "epoch": 1.58, "learning_rate": 1.823948041585506e-05, "loss": 0.1856, "step": 67845 }, { "epoch": 1.58, "learning_rate": 1.8238600361409028e-05, "loss": 0.0971, "step": 67850 }, { "epoch": 1.58, "learning_rate": 1.8237720306962993e-05, "loss": 0.1856, "step": 67855 }, { "epoch": 1.58, "learning_rate": 1.8236840252516955e-05, "loss": 0.3593, "step": 67860 }, { "epoch": 1.58, "learning_rate": 1.823596019807092e-05, "loss": 0.2422, "step": 67865 }, { "epoch": 1.58, "learning_rate": 1.8235080143624888e-05, "loss": 0.0512, "step": 67870 }, { "epoch": 1.58, "learning_rate": 1.823420008917885e-05, "loss": 0.0692, "step": 67875 }, { "epoch": 1.58, "learning_rate": 1.8233320034732815e-05, "loss": 0.0826, "step": 67880 }, { "epoch": 1.58, "learning_rate": 1.823243998028678e-05, "loss": 0.0374, "step": 67885 }, { "epoch": 1.58, "learning_rate": 1.8231559925840748e-05, "loss": 0.0684, "step": 67890 }, { "epoch": 1.58, "learning_rate": 1.823067987139471e-05, "loss": 0.1047, "step": 67895 }, { "epoch": 1.58, "learning_rate": 1.8229799816948675e-05, "loss": 0.1828, "step": 67900 }, { "epoch": 1.58, "learning_rate": 1.822891976250264e-05, "loss": 0.1842, "step": 67905 }, { "epoch": 1.58, "learning_rate": 1.8228039708056605e-05, "loss": 0.4952, "step": 67910 }, { "epoch": 1.58, "learning_rate": 1.822715965361057e-05, "loss": 0.224, "step": 67915 }, { "epoch": 1.58, "learning_rate": 1.8226279599164535e-05, "loss": 0.0492, "step": 67920 }, { "epoch": 1.58, "learning_rate": 1.8225399544718503e-05, "loss": 0.0847, "step": 67925 }, { "epoch": 1.58, "learning_rate": 1.8224519490272465e-05, "loss": 0.0717, "step": 67930 }, { "epoch": 1.59, "learning_rate": 1.822363943582643e-05, "loss": 0.0843, "step": 67935 }, { "epoch": 1.59, "learning_rate": 1.8222759381380395e-05, "loss": 0.0935, "step": 67940 }, { "epoch": 1.59, "learning_rate": 1.822187932693436e-05, "loss": 0.0889, "step": 67945 }, { "epoch": 1.59, "learning_rate": 1.8220999272488325e-05, "loss": 0.1684, "step": 67950 }, { "epoch": 1.59, "learning_rate": 1.822011921804229e-05, "loss": 0.2231, "step": 67955 }, { "epoch": 1.59, "learning_rate": 1.8219239163596255e-05, "loss": 0.2244, "step": 67960 }, { "epoch": 1.59, "learning_rate": 1.821835910915022e-05, "loss": 0.34, "step": 67965 }, { "epoch": 1.59, "learning_rate": 1.8217479054704185e-05, "loss": 0.1386, "step": 67970 }, { "epoch": 1.59, "learning_rate": 1.821659900025815e-05, "loss": 0.0279, "step": 67975 }, { "epoch": 1.59, "learning_rate": 1.8215718945812112e-05, "loss": 0.0058, "step": 67980 }, { "epoch": 1.59, "learning_rate": 1.821483889136608e-05, "loss": 0.0173, "step": 67985 }, { "epoch": 1.59, "learning_rate": 1.8213958836920045e-05, "loss": 0.0611, "step": 67990 }, { "epoch": 1.59, "learning_rate": 1.821307878247401e-05, "loss": 0.1524, "step": 67995 }, { "epoch": 1.59, "learning_rate": 1.8212198728027972e-05, "loss": 0.0514, "step": 68000 }, { "epoch": 1.59, "learning_rate": 1.821131867358194e-05, "loss": 0.2633, "step": 68005 }, { "epoch": 1.59, "learning_rate": 1.8210438619135905e-05, "loss": 0.3726, "step": 68010 }, { "epoch": 1.59, "learning_rate": 1.8209558564689867e-05, "loss": 0.3373, "step": 68015 }, { "epoch": 1.59, "learning_rate": 1.8208678510243832e-05, "loss": 0.0704, "step": 68020 }, { "epoch": 1.59, "learning_rate": 1.82077984557978e-05, "loss": 0.0767, "step": 68025 }, { "epoch": 1.59, "learning_rate": 1.8206918401351765e-05, "loss": 0.1409, "step": 68030 }, { "epoch": 1.59, "learning_rate": 1.8206038346905727e-05, "loss": 0.0582, "step": 68035 }, { "epoch": 1.59, "learning_rate": 1.8205158292459692e-05, "loss": 0.0919, "step": 68040 }, { "epoch": 1.59, "learning_rate": 1.820427823801366e-05, "loss": 0.1427, "step": 68045 }, { "epoch": 1.59, "learning_rate": 1.8203398183567622e-05, "loss": 0.2002, "step": 68050 }, { "epoch": 1.59, "learning_rate": 1.8202518129121587e-05, "loss": 0.2381, "step": 68055 }, { "epoch": 1.59, "learning_rate": 1.8201638074675556e-05, "loss": 0.2842, "step": 68060 }, { "epoch": 1.59, "learning_rate": 1.820075802022952e-05, "loss": 0.3423, "step": 68065 }, { "epoch": 1.59, "learning_rate": 1.8199877965783482e-05, "loss": 0.0668, "step": 68070 }, { "epoch": 1.59, "learning_rate": 1.8198997911337447e-05, "loss": 0.0428, "step": 68075 }, { "epoch": 1.59, "learning_rate": 1.8198117856891416e-05, "loss": 0.1036, "step": 68080 }, { "epoch": 1.59, "learning_rate": 1.819723780244538e-05, "loss": 0.1219, "step": 68085 }, { "epoch": 1.59, "learning_rate": 1.8196357747999342e-05, "loss": 0.0694, "step": 68090 }, { "epoch": 1.59, "learning_rate": 1.8195477693553307e-05, "loss": 0.3391, "step": 68095 }, { "epoch": 1.59, "learning_rate": 1.8194597639107276e-05, "loss": 0.1371, "step": 68100 }, { "epoch": 1.59, "learning_rate": 1.8193717584661237e-05, "loss": 0.1572, "step": 68105 }, { "epoch": 1.59, "learning_rate": 1.8192837530215202e-05, "loss": 0.3467, "step": 68110 }, { "epoch": 1.59, "learning_rate": 1.8191957475769167e-05, "loss": 0.2345, "step": 68115 }, { "epoch": 1.59, "learning_rate": 1.8191077421323136e-05, "loss": 0.0154, "step": 68120 }, { "epoch": 1.59, "learning_rate": 1.8190197366877097e-05, "loss": 0.064, "step": 68125 }, { "epoch": 1.59, "learning_rate": 1.8189317312431062e-05, "loss": 0.0368, "step": 68130 }, { "epoch": 1.59, "learning_rate": 1.8188437257985027e-05, "loss": 0.0769, "step": 68135 }, { "epoch": 1.59, "learning_rate": 1.8187557203538992e-05, "loss": 0.137, "step": 68140 }, { "epoch": 1.59, "learning_rate": 1.8186677149092958e-05, "loss": 0.0328, "step": 68145 }, { "epoch": 1.59, "learning_rate": 1.8185797094646923e-05, "loss": 0.1788, "step": 68150 }, { "epoch": 1.59, "learning_rate": 1.8184917040200888e-05, "loss": 0.1984, "step": 68155 }, { "epoch": 1.59, "learning_rate": 1.8184036985754853e-05, "loss": 0.2431, "step": 68160 }, { "epoch": 1.59, "learning_rate": 1.8183156931308818e-05, "loss": 0.2417, "step": 68165 }, { "epoch": 1.59, "learning_rate": 1.8182276876862783e-05, "loss": 0.0448, "step": 68170 }, { "epoch": 1.59, "learning_rate": 1.8181396822416748e-05, "loss": 0.1144, "step": 68175 }, { "epoch": 1.59, "learning_rate": 1.8180516767970713e-05, "loss": 0.0279, "step": 68180 }, { "epoch": 1.59, "learning_rate": 1.8179636713524678e-05, "loss": 0.0296, "step": 68185 }, { "epoch": 1.59, "learning_rate": 1.8178756659078643e-05, "loss": 0.1077, "step": 68190 }, { "epoch": 1.59, "learning_rate": 1.8177876604632608e-05, "loss": 0.128, "step": 68195 }, { "epoch": 1.59, "learning_rate": 1.8176996550186573e-05, "loss": 0.1812, "step": 68200 }, { "epoch": 1.59, "learning_rate": 1.8176116495740538e-05, "loss": 0.1766, "step": 68205 }, { "epoch": 1.59, "learning_rate": 1.81752364412945e-05, "loss": 0.2526, "step": 68210 }, { "epoch": 1.59, "learning_rate": 1.8174356386848468e-05, "loss": 0.2397, "step": 68215 }, { "epoch": 1.59, "learning_rate": 1.8173476332402433e-05, "loss": 0.1022, "step": 68220 }, { "epoch": 1.59, "learning_rate": 1.8172596277956398e-05, "loss": 0.0525, "step": 68225 }, { "epoch": 1.59, "learning_rate": 1.817171622351036e-05, "loss": 0.1494, "step": 68230 }, { "epoch": 1.59, "learning_rate": 1.8170836169064328e-05, "loss": 0.2018, "step": 68235 }, { "epoch": 1.59, "learning_rate": 1.8169956114618293e-05, "loss": 0.0818, "step": 68240 }, { "epoch": 1.59, "learning_rate": 1.8169076060172255e-05, "loss": 0.0993, "step": 68245 }, { "epoch": 1.59, "learning_rate": 1.816819600572622e-05, "loss": 0.2286, "step": 68250 }, { "epoch": 1.59, "learning_rate": 1.8167491962169394e-05, "loss": 0.3136, "step": 68255 }, { "epoch": 1.59, "learning_rate": 1.816661190772336e-05, "loss": 0.3009, "step": 68260 }, { "epoch": 1.59, "learning_rate": 1.816573185327732e-05, "loss": 0.2824, "step": 68265 }, { "epoch": 1.59, "learning_rate": 1.816485179883129e-05, "loss": 0.1331, "step": 68270 }, { "epoch": 1.59, "learning_rate": 1.8163971744385254e-05, "loss": 0.0544, "step": 68275 }, { "epoch": 1.59, "learning_rate": 1.816309168993922e-05, "loss": 0.0779, "step": 68280 }, { "epoch": 1.59, "learning_rate": 1.816221163549318e-05, "loss": 0.0486, "step": 68285 }, { "epoch": 1.59, "learning_rate": 1.816133158104715e-05, "loss": 0.1724, "step": 68290 }, { "epoch": 1.59, "learning_rate": 1.8160451526601114e-05, "loss": 0.0094, "step": 68295 }, { "epoch": 1.59, "learning_rate": 1.8159571472155076e-05, "loss": 0.1264, "step": 68300 }, { "epoch": 1.59, "learning_rate": 1.815869141770904e-05, "loss": 0.2551, "step": 68305 }, { "epoch": 1.59, "learning_rate": 1.815781136326301e-05, "loss": 0.2627, "step": 68310 }, { "epoch": 1.59, "learning_rate": 1.8156931308816975e-05, "loss": 0.2687, "step": 68315 }, { "epoch": 1.59, "learning_rate": 1.8156051254370936e-05, "loss": 0.005, "step": 68320 }, { "epoch": 1.59, "learning_rate": 1.81551711999249e-05, "loss": 0.0171, "step": 68325 }, { "epoch": 1.59, "learning_rate": 1.815429114547887e-05, "loss": 0.2001, "step": 68330 }, { "epoch": 1.59, "learning_rate": 1.815341109103283e-05, "loss": 0.1473, "step": 68335 }, { "epoch": 1.59, "learning_rate": 1.8152531036586796e-05, "loss": 0.0372, "step": 68340 }, { "epoch": 1.59, "learning_rate": 1.815165098214076e-05, "loss": 0.0897, "step": 68345 }, { "epoch": 1.59, "learning_rate": 1.815077092769473e-05, "loss": 0.0651, "step": 68350 }, { "epoch": 1.59, "learning_rate": 1.814989087324869e-05, "loss": 0.2751, "step": 68355 }, { "epoch": 1.59, "learning_rate": 1.8149010818802656e-05, "loss": 0.1099, "step": 68360 }, { "epoch": 1.6, "learning_rate": 1.814813076435662e-05, "loss": 0.1461, "step": 68365 }, { "epoch": 1.6, "learning_rate": 1.8147250709910586e-05, "loss": 0.0046, "step": 68370 }, { "epoch": 1.6, "learning_rate": 1.814637065546455e-05, "loss": 0.0325, "step": 68375 }, { "epoch": 1.6, "learning_rate": 1.8145490601018516e-05, "loss": 0.0121, "step": 68380 }, { "epoch": 1.6, "learning_rate": 1.8144610546572485e-05, "loss": 0.0128, "step": 68385 }, { "epoch": 1.6, "learning_rate": 1.8143730492126446e-05, "loss": 0.2371, "step": 68390 }, { "epoch": 1.6, "learning_rate": 1.814285043768041e-05, "loss": 0.2401, "step": 68395 }, { "epoch": 1.6, "learning_rate": 1.8141970383234376e-05, "loss": 0.2417, "step": 68400 }, { "epoch": 1.6, "learning_rate": 1.814109032878834e-05, "loss": 0.2245, "step": 68405 }, { "epoch": 1.6, "learning_rate": 1.8140210274342307e-05, "loss": 0.6545, "step": 68410 }, { "epoch": 1.6, "learning_rate": 1.813933021989627e-05, "loss": 0.1407, "step": 68415 }, { "epoch": 1.6, "learning_rate": 1.8138450165450237e-05, "loss": 0.0171, "step": 68420 }, { "epoch": 1.6, "learning_rate": 1.81375701110042e-05, "loss": 0.0059, "step": 68425 }, { "epoch": 1.6, "learning_rate": 1.8136690056558167e-05, "loss": 0.0499, "step": 68430 }, { "epoch": 1.6, "learning_rate": 1.813581000211213e-05, "loss": 0.096, "step": 68435 }, { "epoch": 1.6, "learning_rate": 1.8134929947666093e-05, "loss": 0.1225, "step": 68440 }, { "epoch": 1.6, "learning_rate": 1.813404989322006e-05, "loss": 0.0971, "step": 68445 }, { "epoch": 1.6, "learning_rate": 1.8133169838774027e-05, "loss": 0.1712, "step": 68450 }, { "epoch": 1.6, "learning_rate": 1.8132289784327992e-05, "loss": 0.3323, "step": 68455 }, { "epoch": 1.6, "learning_rate": 1.8131409729881953e-05, "loss": 0.2419, "step": 68460 }, { "epoch": 1.6, "learning_rate": 1.8130529675435922e-05, "loss": 0.2078, "step": 68465 }, { "epoch": 1.6, "learning_rate": 1.8129649620989887e-05, "loss": 0.0557, "step": 68470 }, { "epoch": 1.6, "learning_rate": 1.812876956654385e-05, "loss": 0.0514, "step": 68475 }, { "epoch": 1.6, "learning_rate": 1.8127889512097813e-05, "loss": 0.057, "step": 68480 }, { "epoch": 1.6, "learning_rate": 1.8127009457651782e-05, "loss": 0.0647, "step": 68485 }, { "epoch": 1.6, "learning_rate": 1.8126129403205747e-05, "loss": 0.0542, "step": 68490 }, { "epoch": 1.6, "learning_rate": 1.812524934875971e-05, "loss": 0.0858, "step": 68495 }, { "epoch": 1.6, "learning_rate": 1.8124369294313674e-05, "loss": 0.1483, "step": 68500 }, { "epoch": 1.6, "learning_rate": 1.8123489239867642e-05, "loss": 0.2056, "step": 68505 }, { "epoch": 1.6, "learning_rate": 1.8122609185421604e-05, "loss": 0.5636, "step": 68510 }, { "epoch": 1.6, "learning_rate": 1.812172913097557e-05, "loss": 0.2627, "step": 68515 }, { "epoch": 1.6, "learning_rate": 1.8120849076529537e-05, "loss": 0.0187, "step": 68520 }, { "epoch": 1.6, "learning_rate": 1.8119969022083502e-05, "loss": 0.0093, "step": 68525 }, { "epoch": 1.6, "learning_rate": 1.8119088967637464e-05, "loss": 0.0799, "step": 68530 }, { "epoch": 1.6, "learning_rate": 1.811820891319143e-05, "loss": 0.0646, "step": 68535 }, { "epoch": 1.6, "learning_rate": 1.8117328858745397e-05, "loss": 0.1694, "step": 68540 }, { "epoch": 1.6, "learning_rate": 1.811644880429936e-05, "loss": 0.1943, "step": 68545 }, { "epoch": 1.6, "learning_rate": 1.8115568749853324e-05, "loss": 0.2372, "step": 68550 }, { "epoch": 1.6, "learning_rate": 1.811468869540729e-05, "loss": 0.168, "step": 68555 }, { "epoch": 1.6, "learning_rate": 1.8113808640961257e-05, "loss": 0.1962, "step": 68560 }, { "epoch": 1.6, "learning_rate": 1.811292858651522e-05, "loss": 0.1238, "step": 68565 }, { "epoch": 1.6, "learning_rate": 1.8112048532069184e-05, "loss": 0.0677, "step": 68570 }, { "epoch": 1.6, "learning_rate": 1.811116847762315e-05, "loss": 0.0282, "step": 68575 }, { "epoch": 1.6, "learning_rate": 1.8110288423177117e-05, "loss": 0.0146, "step": 68580 }, { "epoch": 1.6, "learning_rate": 1.810940836873108e-05, "loss": 0.0585, "step": 68585 }, { "epoch": 1.6, "learning_rate": 1.8108528314285044e-05, "loss": 0.0936, "step": 68590 }, { "epoch": 1.6, "learning_rate": 1.810764825983901e-05, "loss": 0.019, "step": 68595 }, { "epoch": 1.6, "learning_rate": 1.8106768205392974e-05, "loss": 0.0852, "step": 68600 }, { "epoch": 1.6, "learning_rate": 1.810588815094694e-05, "loss": 0.1453, "step": 68605 }, { "epoch": 1.6, "learning_rate": 1.8105008096500904e-05, "loss": 0.4433, "step": 68610 }, { "epoch": 1.6, "learning_rate": 1.810412804205487e-05, "loss": 0.264, "step": 68615 }, { "epoch": 1.6, "learning_rate": 1.8103247987608834e-05, "loss": 0.0523, "step": 68620 }, { "epoch": 1.6, "learning_rate": 1.81023679331628e-05, "loss": 0.0316, "step": 68625 }, { "epoch": 1.6, "learning_rate": 1.8101487878716764e-05, "loss": 0.0577, "step": 68630 }, { "epoch": 1.6, "learning_rate": 1.8100607824270726e-05, "loss": 0.0263, "step": 68635 }, { "epoch": 1.6, "learning_rate": 1.8099727769824694e-05, "loss": 0.096, "step": 68640 }, { "epoch": 1.6, "learning_rate": 1.809884771537866e-05, "loss": 0.0526, "step": 68645 }, { "epoch": 1.6, "learning_rate": 1.8097967660932624e-05, "loss": 0.0478, "step": 68650 }, { "epoch": 1.6, "learning_rate": 1.809708760648659e-05, "loss": 0.1241, "step": 68655 }, { "epoch": 1.6, "learning_rate": 1.8096207552040554e-05, "loss": 0.0985, "step": 68660 }, { "epoch": 1.6, "learning_rate": 1.809532749759452e-05, "loss": 0.3628, "step": 68665 }, { "epoch": 1.6, "learning_rate": 1.809444744314848e-05, "loss": 0.0289, "step": 68670 }, { "epoch": 1.6, "learning_rate": 1.809356738870245e-05, "loss": 0.0192, "step": 68675 }, { "epoch": 1.6, "learning_rate": 1.8092687334256414e-05, "loss": 0.1066, "step": 68680 }, { "epoch": 1.6, "learning_rate": 1.809180727981038e-05, "loss": 0.0443, "step": 68685 }, { "epoch": 1.6, "learning_rate": 1.809092722536434e-05, "loss": 0.0348, "step": 68690 }, { "epoch": 1.6, "learning_rate": 1.809004717091831e-05, "loss": 0.1514, "step": 68695 }, { "epoch": 1.6, "learning_rate": 1.8089167116472274e-05, "loss": 0.204, "step": 68700 }, { "epoch": 1.6, "learning_rate": 1.8088287062026236e-05, "loss": 0.3144, "step": 68705 }, { "epoch": 1.6, "learning_rate": 1.80874070075802e-05, "loss": 0.2824, "step": 68710 }, { "epoch": 1.6, "learning_rate": 1.808652695313417e-05, "loss": 0.2191, "step": 68715 }, { "epoch": 1.6, "learning_rate": 1.8085646898688134e-05, "loss": 0.0409, "step": 68720 }, { "epoch": 1.6, "learning_rate": 1.8084766844242096e-05, "loss": 0.0995, "step": 68725 }, { "epoch": 1.6, "learning_rate": 1.808388678979606e-05, "loss": 0.0957, "step": 68730 }, { "epoch": 1.6, "learning_rate": 1.808300673535003e-05, "loss": 0.0952, "step": 68735 }, { "epoch": 1.6, "learning_rate": 1.808212668090399e-05, "loss": 0.0251, "step": 68740 }, { "epoch": 1.6, "learning_rate": 1.8081246626457956e-05, "loss": 0.1456, "step": 68745 }, { "epoch": 1.6, "learning_rate": 1.808036657201192e-05, "loss": 0.1611, "step": 68750 }, { "epoch": 1.6, "learning_rate": 1.807948651756589e-05, "loss": 0.238, "step": 68755 }, { "epoch": 1.6, "learning_rate": 1.807860646311985e-05, "loss": 0.3254, "step": 68760 }, { "epoch": 1.6, "learning_rate": 1.8077726408673816e-05, "loss": 0.0908, "step": 68765 }, { "epoch": 1.6, "learning_rate": 1.807684635422778e-05, "loss": 0.0426, "step": 68770 }, { "epoch": 1.6, "learning_rate": 1.8075966299781746e-05, "loss": 0.0355, "step": 68775 }, { "epoch": 1.6, "learning_rate": 1.807508624533571e-05, "loss": 0.2575, "step": 68780 }, { "epoch": 1.6, "learning_rate": 1.8074206190889676e-05, "loss": 0.1307, "step": 68785 }, { "epoch": 1.6, "learning_rate": 1.8073326136443645e-05, "loss": 0.1507, "step": 68790 }, { "epoch": 1.61, "learning_rate": 1.8072446081997606e-05, "loss": 0.0938, "step": 68795 }, { "epoch": 1.61, "learning_rate": 1.807156602755157e-05, "loss": 0.2438, "step": 68800 }, { "epoch": 1.61, "learning_rate": 1.8070685973105536e-05, "loss": 0.1765, "step": 68805 }, { "epoch": 1.61, "learning_rate": 1.80698059186595e-05, "loss": 0.3242, "step": 68810 }, { "epoch": 1.61, "learning_rate": 1.8068925864213466e-05, "loss": 0.2341, "step": 68815 }, { "epoch": 1.61, "learning_rate": 1.806804580976743e-05, "loss": 0.0355, "step": 68820 }, { "epoch": 1.61, "learning_rate": 1.8067165755321396e-05, "loss": 0.0448, "step": 68825 }, { "epoch": 1.61, "learning_rate": 1.806628570087536e-05, "loss": 0.0546, "step": 68830 }, { "epoch": 1.61, "learning_rate": 1.8065405646429327e-05, "loss": 0.0653, "step": 68835 }, { "epoch": 1.61, "learning_rate": 1.806452559198329e-05, "loss": 0.0801, "step": 68840 }, { "epoch": 1.61, "learning_rate": 1.8063645537537253e-05, "loss": 0.1612, "step": 68845 }, { "epoch": 1.61, "learning_rate": 1.806276548309122e-05, "loss": 0.1967, "step": 68850 }, { "epoch": 1.61, "learning_rate": 1.8061885428645187e-05, "loss": 0.149, "step": 68855 }, { "epoch": 1.61, "learning_rate": 1.806100537419915e-05, "loss": 0.2764, "step": 68860 }, { "epoch": 1.61, "learning_rate": 1.8060125319753113e-05, "loss": 0.4151, "step": 68865 }, { "epoch": 1.61, "learning_rate": 1.805924526530708e-05, "loss": 0.0484, "step": 68870 }, { "epoch": 1.61, "learning_rate": 1.8058365210861047e-05, "loss": 0.2995, "step": 68875 }, { "epoch": 1.61, "learning_rate": 1.805748515641501e-05, "loss": 0.0481, "step": 68880 }, { "epoch": 1.61, "learning_rate": 1.8056605101968973e-05, "loss": 0.3074, "step": 68885 }, { "epoch": 1.61, "learning_rate": 1.8055725047522942e-05, "loss": 0.0826, "step": 68890 }, { "epoch": 1.61, "learning_rate": 1.8054844993076907e-05, "loss": 0.168, "step": 68895 }, { "epoch": 1.61, "learning_rate": 1.805396493863087e-05, "loss": 0.1617, "step": 68900 }, { "epoch": 1.61, "learning_rate": 1.8053084884184833e-05, "loss": 0.1788, "step": 68905 }, { "epoch": 1.61, "learning_rate": 1.8052204829738802e-05, "loss": 0.4469, "step": 68910 }, { "epoch": 1.61, "learning_rate": 1.8051324775292767e-05, "loss": 0.2411, "step": 68915 }, { "epoch": 1.61, "learning_rate": 1.805044472084673e-05, "loss": 0.0527, "step": 68920 }, { "epoch": 1.61, "learning_rate": 1.8049564666400697e-05, "loss": 0.0157, "step": 68925 }, { "epoch": 1.61, "learning_rate": 1.8048684611954662e-05, "loss": 0.0207, "step": 68930 }, { "epoch": 1.61, "learning_rate": 1.8047804557508624e-05, "loss": 0.1538, "step": 68935 }, { "epoch": 1.61, "learning_rate": 1.804692450306259e-05, "loss": 0.0972, "step": 68940 }, { "epoch": 1.61, "learning_rate": 1.8046044448616557e-05, "loss": 0.1488, "step": 68945 }, { "epoch": 1.61, "learning_rate": 1.8045164394170522e-05, "loss": 0.1072, "step": 68950 }, { "epoch": 1.61, "learning_rate": 1.8044284339724484e-05, "loss": 0.2143, "step": 68955 }, { "epoch": 1.61, "learning_rate": 1.804340428527845e-05, "loss": 0.3003, "step": 68960 }, { "epoch": 1.61, "learning_rate": 1.8042524230832417e-05, "loss": 0.2204, "step": 68965 }, { "epoch": 1.61, "learning_rate": 1.804164417638638e-05, "loss": 0.0662, "step": 68970 }, { "epoch": 1.61, "learning_rate": 1.8040764121940344e-05, "loss": 0.0732, "step": 68975 }, { "epoch": 1.61, "learning_rate": 1.803988406749431e-05, "loss": 0.0792, "step": 68980 }, { "epoch": 1.61, "learning_rate": 1.8039004013048277e-05, "loss": 0.0103, "step": 68985 }, { "epoch": 1.61, "learning_rate": 1.803812395860224e-05, "loss": 0.1762, "step": 68990 }, { "epoch": 1.61, "learning_rate": 1.8037243904156204e-05, "loss": 0.1525, "step": 68995 }, { "epoch": 1.61, "learning_rate": 1.803636384971017e-05, "loss": 0.1526, "step": 69000 }, { "epoch": 1.61, "learning_rate": 1.8035483795264134e-05, "loss": 0.2542, "step": 69005 }, { "epoch": 1.61, "learning_rate": 1.80346037408181e-05, "loss": 0.1729, "step": 69010 }, { "epoch": 1.61, "learning_rate": 1.8033723686372064e-05, "loss": 0.3584, "step": 69015 }, { "epoch": 1.61, "learning_rate": 1.803284363192603e-05, "loss": 0.099, "step": 69020 }, { "epoch": 1.61, "learning_rate": 1.8031963577479994e-05, "loss": 0.1347, "step": 69025 }, { "epoch": 1.61, "learning_rate": 1.803108352303396e-05, "loss": 0.1825, "step": 69030 }, { "epoch": 1.61, "learning_rate": 1.8030203468587924e-05, "loss": 0.1209, "step": 69035 }, { "epoch": 1.61, "learning_rate": 1.8029323414141886e-05, "loss": 0.0379, "step": 69040 }, { "epoch": 1.61, "learning_rate": 1.8028443359695854e-05, "loss": 0.0409, "step": 69045 }, { "epoch": 1.61, "learning_rate": 1.802756330524982e-05, "loss": 0.4829, "step": 69050 }, { "epoch": 1.61, "learning_rate": 1.8026683250803784e-05, "loss": 0.2358, "step": 69055 }, { "epoch": 1.61, "learning_rate": 1.802580319635775e-05, "loss": 0.4106, "step": 69060 }, { "epoch": 1.61, "learning_rate": 1.8024923141911714e-05, "loss": 0.1373, "step": 69065 }, { "epoch": 1.61, "learning_rate": 1.802404308746568e-05, "loss": 0.0481, "step": 69070 }, { "epoch": 1.61, "learning_rate": 1.802316303301964e-05, "loss": 0.0069, "step": 69075 }, { "epoch": 1.61, "learning_rate": 1.802228297857361e-05, "loss": 0.0635, "step": 69080 }, { "epoch": 1.61, "learning_rate": 1.8021402924127574e-05, "loss": 0.1935, "step": 69085 }, { "epoch": 1.61, "learning_rate": 1.802052286968154e-05, "loss": 0.0674, "step": 69090 }, { "epoch": 1.61, "learning_rate": 1.80196428152355e-05, "loss": 0.0488, "step": 69095 }, { "epoch": 1.61, "learning_rate": 1.801876276078947e-05, "loss": 0.1376, "step": 69100 }, { "epoch": 1.61, "learning_rate": 1.8017882706343434e-05, "loss": 0.2721, "step": 69105 }, { "epoch": 1.61, "learning_rate": 1.8017002651897396e-05, "loss": 0.2919, "step": 69110 }, { "epoch": 1.61, "learning_rate": 1.801612259745136e-05, "loss": 0.2235, "step": 69115 }, { "epoch": 1.61, "learning_rate": 1.801524254300533e-05, "loss": 0.0209, "step": 69120 }, { "epoch": 1.61, "learning_rate": 1.8014362488559294e-05, "loss": 0.0234, "step": 69125 }, { "epoch": 1.61, "learning_rate": 1.8013482434113256e-05, "loss": 0.1793, "step": 69130 }, { "epoch": 1.61, "learning_rate": 1.801260237966722e-05, "loss": 0.1493, "step": 69135 }, { "epoch": 1.61, "learning_rate": 1.801172232522119e-05, "loss": 0.144, "step": 69140 }, { "epoch": 1.61, "learning_rate": 1.801084227077515e-05, "loss": 0.1008, "step": 69145 }, { "epoch": 1.61, "learning_rate": 1.8009962216329116e-05, "loss": 0.2248, "step": 69150 }, { "epoch": 1.61, "learning_rate": 1.800908216188308e-05, "loss": 0.1001, "step": 69155 }, { "epoch": 1.61, "learning_rate": 1.800820210743705e-05, "loss": 0.379, "step": 69160 }, { "epoch": 1.61, "learning_rate": 1.800732205299101e-05, "loss": 0.2176, "step": 69165 }, { "epoch": 1.61, "learning_rate": 1.8006441998544976e-05, "loss": 0.0784, "step": 69170 }, { "epoch": 1.61, "learning_rate": 1.800556194409894e-05, "loss": 0.0092, "step": 69175 }, { "epoch": 1.61, "learning_rate": 1.800468188965291e-05, "loss": 0.0337, "step": 69180 }, { "epoch": 1.61, "learning_rate": 1.800380183520687e-05, "loss": 0.1113, "step": 69185 }, { "epoch": 1.61, "learning_rate": 1.8002921780760836e-05, "loss": 0.0962, "step": 69190 }, { "epoch": 1.61, "learning_rate": 1.8002041726314805e-05, "loss": 0.0655, "step": 69195 }, { "epoch": 1.61, "learning_rate": 1.8001161671868766e-05, "loss": 0.204, "step": 69200 }, { "epoch": 1.61, "learning_rate": 1.800028161742273e-05, "loss": 0.2675, "step": 69205 }, { "epoch": 1.61, "learning_rate": 1.7999401562976696e-05, "loss": 0.2066, "step": 69210 }, { "epoch": 1.61, "learning_rate": 1.7998521508530665e-05, "loss": 0.1406, "step": 69215 }, { "epoch": 1.61, "learning_rate": 1.7997641454084626e-05, "loss": 0.0789, "step": 69220 }, { "epoch": 1.62, "learning_rate": 1.799676139963859e-05, "loss": 0.067, "step": 69225 }, { "epoch": 1.62, "learning_rate": 1.7995881345192556e-05, "loss": 0.1269, "step": 69230 }, { "epoch": 1.62, "learning_rate": 1.799500129074652e-05, "loss": 0.0517, "step": 69235 }, { "epoch": 1.62, "learning_rate": 1.7994121236300486e-05, "loss": 0.0662, "step": 69240 }, { "epoch": 1.62, "learning_rate": 1.799324118185445e-05, "loss": 0.1132, "step": 69245 }, { "epoch": 1.62, "learning_rate": 1.7992361127408416e-05, "loss": 0.1001, "step": 69250 }, { "epoch": 1.62, "learning_rate": 1.799148107296238e-05, "loss": 0.3561, "step": 69255 }, { "epoch": 1.62, "learning_rate": 1.7990601018516346e-05, "loss": 0.3375, "step": 69260 }, { "epoch": 1.62, "learning_rate": 1.798972096407031e-05, "loss": 0.3007, "step": 69265 }, { "epoch": 1.62, "learning_rate": 1.7988840909624273e-05, "loss": 0.0647, "step": 69270 }, { "epoch": 1.62, "learning_rate": 1.798796085517824e-05, "loss": 0.0079, "step": 69275 }, { "epoch": 1.62, "learning_rate": 1.7987080800732207e-05, "loss": 0.0622, "step": 69280 }, { "epoch": 1.62, "learning_rate": 1.798620074628617e-05, "loss": 0.1244, "step": 69285 }, { "epoch": 1.62, "learning_rate": 1.7985320691840133e-05, "loss": 0.0961, "step": 69290 }, { "epoch": 1.62, "learning_rate": 1.79844406373941e-05, "loss": 0.2191, "step": 69295 }, { "epoch": 1.62, "learning_rate": 1.7983560582948067e-05, "loss": 0.1789, "step": 69300 }, { "epoch": 1.62, "learning_rate": 1.7982680528502028e-05, "loss": 0.2608, "step": 69305 }, { "epoch": 1.62, "learning_rate": 1.7981800474055993e-05, "loss": 0.3759, "step": 69310 }, { "epoch": 1.62, "learning_rate": 1.7980920419609962e-05, "loss": 0.2493, "step": 69315 }, { "epoch": 1.62, "learning_rate": 1.7980040365163927e-05, "loss": 0.0496, "step": 69320 }, { "epoch": 1.62, "learning_rate": 1.797916031071789e-05, "loss": 0.0776, "step": 69325 }, { "epoch": 1.62, "learning_rate": 1.7978280256271857e-05, "loss": 0.1038, "step": 69330 }, { "epoch": 1.62, "learning_rate": 1.7977400201825822e-05, "loss": 0.1343, "step": 69335 }, { "epoch": 1.62, "learning_rate": 1.7976520147379783e-05, "loss": 0.1252, "step": 69340 }, { "epoch": 1.62, "learning_rate": 1.797564009293375e-05, "loss": 0.1521, "step": 69345 }, { "epoch": 1.62, "learning_rate": 1.7974760038487717e-05, "loss": 0.1395, "step": 69350 }, { "epoch": 1.62, "learning_rate": 1.7973879984041682e-05, "loss": 0.0888, "step": 69355 }, { "epoch": 1.62, "learning_rate": 1.7972999929595643e-05, "loss": 0.2979, "step": 69360 }, { "epoch": 1.62, "learning_rate": 1.797211987514961e-05, "loss": 0.279, "step": 69365 }, { "epoch": 1.62, "learning_rate": 1.7971239820703577e-05, "loss": 0.0334, "step": 69370 }, { "epoch": 1.62, "learning_rate": 1.797035976625754e-05, "loss": 0.0466, "step": 69375 }, { "epoch": 1.62, "learning_rate": 1.7969479711811504e-05, "loss": 0.1964, "step": 69380 }, { "epoch": 1.62, "learning_rate": 1.796859965736547e-05, "loss": 0.1049, "step": 69385 }, { "epoch": 1.62, "learning_rate": 1.7967719602919437e-05, "loss": 0.0687, "step": 69390 }, { "epoch": 1.62, "learning_rate": 1.79668395484734e-05, "loss": 0.0923, "step": 69395 }, { "epoch": 1.62, "learning_rate": 1.7965959494027364e-05, "loss": 0.1861, "step": 69400 }, { "epoch": 1.62, "learning_rate": 1.796507943958133e-05, "loss": 0.1211, "step": 69405 }, { "epoch": 1.62, "learning_rate": 1.7964199385135294e-05, "loss": 0.3742, "step": 69410 }, { "epoch": 1.62, "learning_rate": 1.796331933068926e-05, "loss": 0.1668, "step": 69415 }, { "epoch": 1.62, "learning_rate": 1.7962439276243224e-05, "loss": 0.0198, "step": 69420 }, { "epoch": 1.62, "learning_rate": 1.796155922179719e-05, "loss": 0.1242, "step": 69425 }, { "epoch": 1.62, "learning_rate": 1.7960679167351154e-05, "loss": 0.0906, "step": 69430 }, { "epoch": 1.62, "learning_rate": 1.795979911290512e-05, "loss": 0.0939, "step": 69435 }, { "epoch": 1.62, "learning_rate": 1.7958919058459084e-05, "loss": 0.1175, "step": 69440 }, { "epoch": 1.62, "learning_rate": 1.7958039004013045e-05, "loss": 0.0392, "step": 69445 }, { "epoch": 1.62, "learning_rate": 1.7957158949567014e-05, "loss": 0.2012, "step": 69450 }, { "epoch": 1.62, "learning_rate": 1.795627889512098e-05, "loss": 0.1339, "step": 69455 }, { "epoch": 1.62, "learning_rate": 1.7955398840674944e-05, "loss": 0.3891, "step": 69460 }, { "epoch": 1.62, "learning_rate": 1.795451878622891e-05, "loss": 0.1095, "step": 69465 }, { "epoch": 1.62, "learning_rate": 1.7953638731782874e-05, "loss": 0.0384, "step": 69470 }, { "epoch": 1.62, "learning_rate": 1.795275867733684e-05, "loss": 0.0616, "step": 69475 }, { "epoch": 1.62, "learning_rate": 1.7951878622890804e-05, "loss": 0.034, "step": 69480 }, { "epoch": 1.62, "learning_rate": 1.795099856844477e-05, "loss": 0.1555, "step": 69485 }, { "epoch": 1.62, "learning_rate": 1.7950118513998734e-05, "loss": 0.1441, "step": 69490 }, { "epoch": 1.62, "learning_rate": 1.79492384595527e-05, "loss": 0.0937, "step": 69495 }, { "epoch": 1.62, "learning_rate": 1.794835840510666e-05, "loss": 0.1003, "step": 69500 }, { "epoch": 1.62, "learning_rate": 1.794747835066063e-05, "loss": 0.0977, "step": 69505 }, { "epoch": 1.62, "learning_rate": 1.7946598296214594e-05, "loss": 0.2584, "step": 69510 }, { "epoch": 1.62, "learning_rate": 1.794571824176856e-05, "loss": 0.3689, "step": 69515 }, { "epoch": 1.62, "learning_rate": 1.794483818732252e-05, "loss": 0.0795, "step": 69520 }, { "epoch": 1.62, "learning_rate": 1.794395813287649e-05, "loss": 0.0423, "step": 69525 }, { "epoch": 1.62, "learning_rate": 1.7943078078430454e-05, "loss": 0.0565, "step": 69530 }, { "epoch": 1.62, "learning_rate": 1.7942198023984416e-05, "loss": 0.1119, "step": 69535 }, { "epoch": 1.62, "learning_rate": 1.794131796953838e-05, "loss": 0.0545, "step": 69540 }, { "epoch": 1.62, "learning_rate": 1.794043791509235e-05, "loss": 0.2771, "step": 69545 }, { "epoch": 1.62, "learning_rate": 1.7939557860646314e-05, "loss": 0.2045, "step": 69550 }, { "epoch": 1.62, "learning_rate": 1.7938677806200276e-05, "loss": 0.2578, "step": 69555 }, { "epoch": 1.62, "learning_rate": 1.793779775175424e-05, "loss": 0.41, "step": 69560 }, { "epoch": 1.62, "learning_rate": 1.793691769730821e-05, "loss": 0.2225, "step": 69565 }, { "epoch": 1.62, "learning_rate": 1.793603764286217e-05, "loss": 0.0462, "step": 69570 }, { "epoch": 1.62, "learning_rate": 1.7935157588416136e-05, "loss": 0.0937, "step": 69575 }, { "epoch": 1.62, "learning_rate": 1.79342775339701e-05, "loss": 0.0519, "step": 69580 }, { "epoch": 1.62, "learning_rate": 1.793339747952407e-05, "loss": 0.0243, "step": 69585 }, { "epoch": 1.62, "learning_rate": 1.793251742507803e-05, "loss": 0.1179, "step": 69590 }, { "epoch": 1.62, "learning_rate": 1.7931637370631996e-05, "loss": 0.0827, "step": 69595 }, { "epoch": 1.62, "learning_rate": 1.7930757316185964e-05, "loss": 0.1671, "step": 69600 }, { "epoch": 1.62, "learning_rate": 1.7929877261739926e-05, "loss": 0.3877, "step": 69605 }, { "epoch": 1.62, "learning_rate": 1.792899720729389e-05, "loss": 0.4612, "step": 69610 }, { "epoch": 1.62, "learning_rate": 1.7928117152847856e-05, "loss": 0.216, "step": 69615 }, { "epoch": 1.62, "learning_rate": 1.7927237098401825e-05, "loss": 0.0523, "step": 69620 }, { "epoch": 1.62, "learning_rate": 1.7926357043955786e-05, "loss": 0.0876, "step": 69625 }, { "epoch": 1.62, "learning_rate": 1.792547698950975e-05, "loss": 0.2479, "step": 69630 }, { "epoch": 1.62, "learning_rate": 1.7924596935063716e-05, "loss": 0.0783, "step": 69635 }, { "epoch": 1.62, "learning_rate": 1.792371688061768e-05, "loss": 0.1275, "step": 69640 }, { "epoch": 1.62, "learning_rate": 1.7922836826171646e-05, "loss": 0.2602, "step": 69645 }, { "epoch": 1.63, "learning_rate": 1.792195677172561e-05, "loss": 0.1418, "step": 69650 }, { "epoch": 1.63, "learning_rate": 1.7921076717279576e-05, "loss": 0.1835, "step": 69655 }, { "epoch": 1.63, "learning_rate": 1.792019666283354e-05, "loss": 0.228, "step": 69660 }, { "epoch": 1.63, "learning_rate": 1.7919316608387506e-05, "loss": 0.2593, "step": 69665 }, { "epoch": 1.63, "learning_rate": 1.791843655394147e-05, "loss": 0.0333, "step": 69670 }, { "epoch": 1.63, "learning_rate": 1.7917556499495433e-05, "loss": 0.0534, "step": 69675 }, { "epoch": 1.63, "learning_rate": 1.79166764450494e-05, "loss": 0.0355, "step": 69680 }, { "epoch": 1.63, "learning_rate": 1.7915796390603366e-05, "loss": 0.1337, "step": 69685 }, { "epoch": 1.63, "learning_rate": 1.791491633615733e-05, "loss": 0.2231, "step": 69690 }, { "epoch": 1.63, "learning_rate": 1.7914036281711293e-05, "loss": 0.1934, "step": 69695 }, { "epoch": 1.63, "learning_rate": 1.791315622726526e-05, "loss": 0.1615, "step": 69700 }, { "epoch": 1.63, "learning_rate": 1.7912276172819227e-05, "loss": 0.2828, "step": 69705 }, { "epoch": 1.63, "learning_rate": 1.7911396118373188e-05, "loss": 0.4036, "step": 69710 }, { "epoch": 1.63, "learning_rate": 1.7910516063927153e-05, "loss": 0.2364, "step": 69715 }, { "epoch": 1.63, "learning_rate": 1.790963600948112e-05, "loss": 0.0388, "step": 69720 }, { "epoch": 1.63, "learning_rate": 1.7908755955035087e-05, "loss": 0.0533, "step": 69725 }, { "epoch": 1.63, "learning_rate": 1.7907875900589048e-05, "loss": 0.0791, "step": 69730 }, { "epoch": 1.63, "learning_rate": 1.7906995846143017e-05, "loss": 0.032, "step": 69735 }, { "epoch": 1.63, "learning_rate": 1.790611579169698e-05, "loss": 0.0962, "step": 69740 }, { "epoch": 1.63, "learning_rate": 1.7905235737250943e-05, "loss": 0.1371, "step": 69745 }, { "epoch": 1.63, "learning_rate": 1.790435568280491e-05, "loss": 0.1005, "step": 69750 }, { "epoch": 1.63, "learning_rate": 1.7903475628358877e-05, "loss": 0.3033, "step": 69755 }, { "epoch": 1.63, "learning_rate": 1.7902595573912842e-05, "loss": 0.223, "step": 69760 }, { "epoch": 1.63, "learning_rate": 1.7901715519466803e-05, "loss": 0.282, "step": 69765 }, { "epoch": 1.63, "learning_rate": 1.790083546502077e-05, "loss": 0.0157, "step": 69770 }, { "epoch": 1.63, "learning_rate": 1.7899955410574737e-05, "loss": 0.1987, "step": 69775 }, { "epoch": 1.63, "learning_rate": 1.7899075356128702e-05, "loss": 0.0932, "step": 69780 }, { "epoch": 1.63, "learning_rate": 1.7898195301682663e-05, "loss": 0.0466, "step": 69785 }, { "epoch": 1.63, "learning_rate": 1.789731524723663e-05, "loss": 0.0689, "step": 69790 }, { "epoch": 1.63, "learning_rate": 1.7896435192790597e-05, "loss": 0.1091, "step": 69795 }, { "epoch": 1.63, "learning_rate": 1.789555513834456e-05, "loss": 0.1418, "step": 69800 }, { "epoch": 1.63, "learning_rate": 1.7894675083898524e-05, "loss": 0.1709, "step": 69805 }, { "epoch": 1.63, "learning_rate": 1.789379502945249e-05, "loss": 0.2332, "step": 69810 }, { "epoch": 1.63, "learning_rate": 1.7892914975006457e-05, "loss": 0.1993, "step": 69815 }, { "epoch": 1.63, "learning_rate": 1.789203492056042e-05, "loss": 0.0593, "step": 69820 }, { "epoch": 1.63, "learning_rate": 1.7891154866114384e-05, "loss": 0.122, "step": 69825 }, { "epoch": 1.63, "learning_rate": 1.789027481166835e-05, "loss": 0.1301, "step": 69830 }, { "epoch": 1.63, "learning_rate": 1.7889394757222314e-05, "loss": 0.0913, "step": 69835 }, { "epoch": 1.63, "learning_rate": 1.788851470277628e-05, "loss": 0.0591, "step": 69840 }, { "epoch": 1.63, "learning_rate": 1.7887634648330244e-05, "loss": 0.1129, "step": 69845 }, { "epoch": 1.63, "learning_rate": 1.788675459388421e-05, "loss": 0.0937, "step": 69850 }, { "epoch": 1.63, "learning_rate": 1.7885874539438174e-05, "loss": 0.3281, "step": 69855 }, { "epoch": 1.63, "learning_rate": 1.788499448499214e-05, "loss": 0.1868, "step": 69860 }, { "epoch": 1.63, "learning_rate": 1.7884114430546104e-05, "loss": 0.2173, "step": 69865 }, { "epoch": 1.63, "learning_rate": 1.788323437610007e-05, "loss": 0.0891, "step": 69870 }, { "epoch": 1.63, "learning_rate": 1.7882354321654034e-05, "loss": 0.0623, "step": 69875 }, { "epoch": 1.63, "learning_rate": 1.7881474267208e-05, "loss": 0.0399, "step": 69880 }, { "epoch": 1.63, "learning_rate": 1.7880594212761964e-05, "loss": 0.0962, "step": 69885 }, { "epoch": 1.63, "learning_rate": 1.787971415831593e-05, "loss": 0.0846, "step": 69890 }, { "epoch": 1.63, "learning_rate": 1.7878834103869894e-05, "loss": 0.1456, "step": 69895 }, { "epoch": 1.63, "learning_rate": 1.787795404942386e-05, "loss": 0.1538, "step": 69900 }, { "epoch": 1.63, "learning_rate": 1.787707399497782e-05, "loss": 0.2, "step": 69905 }, { "epoch": 1.63, "learning_rate": 1.787619394053179e-05, "loss": 0.2147, "step": 69910 }, { "epoch": 1.63, "learning_rate": 1.7875313886085754e-05, "loss": 0.3302, "step": 69915 }, { "epoch": 1.63, "learning_rate": 1.787443383163972e-05, "loss": 0.0255, "step": 69920 }, { "epoch": 1.63, "learning_rate": 1.787355377719368e-05, "loss": 0.0412, "step": 69925 }, { "epoch": 1.63, "learning_rate": 1.787267372274765e-05, "loss": 0.1238, "step": 69930 }, { "epoch": 1.63, "learning_rate": 1.7871793668301614e-05, "loss": 0.0535, "step": 69935 }, { "epoch": 1.63, "learning_rate": 1.7870913613855576e-05, "loss": 0.223, "step": 69940 }, { "epoch": 1.63, "learning_rate": 1.787003355940954e-05, "loss": 0.1959, "step": 69945 }, { "epoch": 1.63, "learning_rate": 1.786915350496351e-05, "loss": 0.0396, "step": 69950 }, { "epoch": 1.63, "learning_rate": 1.7868273450517474e-05, "loss": 0.1738, "step": 69955 }, { "epoch": 1.63, "learning_rate": 1.7867393396071436e-05, "loss": 0.4321, "step": 69960 }, { "epoch": 1.63, "learning_rate": 1.78665133416254e-05, "loss": 0.2095, "step": 69965 }, { "epoch": 1.63, "learning_rate": 1.786563328717937e-05, "loss": 0.0291, "step": 69970 }, { "epoch": 1.63, "learning_rate": 1.786475323273333e-05, "loss": 0.0509, "step": 69975 }, { "epoch": 1.63, "learning_rate": 1.7863873178287296e-05, "loss": 0.0412, "step": 69980 }, { "epoch": 1.63, "learning_rate": 1.786299312384126e-05, "loss": 0.1407, "step": 69985 }, { "epoch": 1.63, "learning_rate": 1.786211306939523e-05, "loss": 0.0703, "step": 69990 }, { "epoch": 1.63, "learning_rate": 1.786123301494919e-05, "loss": 0.1127, "step": 69995 }, { "epoch": 1.63, "learning_rate": 1.7860352960503156e-05, "loss": 0.3262, "step": 70000 }, { "epoch": 1.63, "learning_rate": 1.7859472906057124e-05, "loss": 0.2901, "step": 70005 }, { "epoch": 1.63, "learning_rate": 1.7858592851611086e-05, "loss": 0.3014, "step": 70010 }, { "epoch": 1.63, "learning_rate": 1.785771279716505e-05, "loss": 0.2284, "step": 70015 }, { "epoch": 1.63, "learning_rate": 1.7856832742719016e-05, "loss": 0.0953, "step": 70020 }, { "epoch": 1.63, "learning_rate": 1.7855952688272984e-05, "loss": 0.1113, "step": 70025 }, { "epoch": 1.63, "learning_rate": 1.7855072633826946e-05, "loss": 0.0596, "step": 70030 }, { "epoch": 1.63, "learning_rate": 1.785419257938091e-05, "loss": 0.1236, "step": 70035 }, { "epoch": 1.63, "learning_rate": 1.7853312524934876e-05, "loss": 0.1454, "step": 70040 }, { "epoch": 1.63, "learning_rate": 1.785243247048884e-05, "loss": 0.1427, "step": 70045 }, { "epoch": 1.63, "learning_rate": 1.7851552416042806e-05, "loss": 0.1978, "step": 70050 }, { "epoch": 1.63, "learning_rate": 1.785067236159677e-05, "loss": 0.1207, "step": 70055 }, { "epoch": 1.63, "learning_rate": 1.7849792307150736e-05, "loss": 0.4057, "step": 70060 }, { "epoch": 1.63, "learning_rate": 1.78489122527047e-05, "loss": 0.2033, "step": 70065 }, { "epoch": 1.63, "learning_rate": 1.7848032198258666e-05, "loss": 0.0731, "step": 70070 }, { "epoch": 1.63, "learning_rate": 1.784715214381263e-05, "loss": 0.0353, "step": 70075 }, { "epoch": 1.64, "learning_rate": 1.7846272089366593e-05, "loss": 0.0891, "step": 70080 }, { "epoch": 1.64, "learning_rate": 1.784539203492056e-05, "loss": 0.0571, "step": 70085 }, { "epoch": 1.64, "learning_rate": 1.7844511980474526e-05, "loss": 0.1914, "step": 70090 }, { "epoch": 1.64, "learning_rate": 1.784363192602849e-05, "loss": 0.1393, "step": 70095 }, { "epoch": 1.64, "learning_rate": 1.7842751871582453e-05, "loss": 0.1998, "step": 70100 }, { "epoch": 1.64, "learning_rate": 1.784187181713642e-05, "loss": 0.3238, "step": 70105 }, { "epoch": 1.64, "learning_rate": 1.7840991762690386e-05, "loss": 0.2528, "step": 70110 }, { "epoch": 1.64, "learning_rate": 1.784011170824435e-05, "loss": 0.208, "step": 70115 }, { "epoch": 1.64, "learning_rate": 1.7839231653798313e-05, "loss": 0.0701, "step": 70120 }, { "epoch": 1.64, "learning_rate": 1.783835159935228e-05, "loss": 0.0892, "step": 70125 }, { "epoch": 1.64, "learning_rate": 1.7837471544906246e-05, "loss": 0.1117, "step": 70130 }, { "epoch": 1.64, "learning_rate": 1.7836591490460208e-05, "loss": 0.0133, "step": 70135 }, { "epoch": 1.64, "learning_rate": 1.7835711436014177e-05, "loss": 0.0365, "step": 70140 }, { "epoch": 1.64, "learning_rate": 1.783483138156814e-05, "loss": 0.069, "step": 70145 }, { "epoch": 1.64, "learning_rate": 1.7833951327122107e-05, "loss": 0.0817, "step": 70150 }, { "epoch": 1.64, "learning_rate": 1.7833071272676068e-05, "loss": 0.2674, "step": 70155 }, { "epoch": 1.64, "learning_rate": 1.7832191218230037e-05, "loss": 0.1772, "step": 70160 }, { "epoch": 1.64, "learning_rate": 1.7831311163784e-05, "loss": 0.3384, "step": 70165 }, { "epoch": 1.64, "learning_rate": 1.7830431109337963e-05, "loss": 0.0441, "step": 70170 }, { "epoch": 1.64, "learning_rate": 1.7829551054891928e-05, "loss": 0.0689, "step": 70175 }, { "epoch": 1.64, "learning_rate": 1.7828671000445897e-05, "loss": 0.1645, "step": 70180 }, { "epoch": 1.64, "learning_rate": 1.7827790945999862e-05, "loss": 0.1407, "step": 70185 }, { "epoch": 1.64, "learning_rate": 1.7826910891553823e-05, "loss": 0.1091, "step": 70190 }, { "epoch": 1.64, "learning_rate": 1.782603083710779e-05, "loss": 0.1446, "step": 70195 }, { "epoch": 1.64, "learning_rate": 1.7825150782661757e-05, "loss": 0.152, "step": 70200 }, { "epoch": 1.64, "learning_rate": 1.782427072821572e-05, "loss": 0.2073, "step": 70205 }, { "epoch": 1.64, "learning_rate": 1.7823390673769683e-05, "loss": 0.2178, "step": 70210 }, { "epoch": 1.64, "learning_rate": 1.782251061932365e-05, "loss": 0.3621, "step": 70215 }, { "epoch": 1.64, "learning_rate": 1.7821630564877617e-05, "loss": 0.0171, "step": 70220 }, { "epoch": 1.64, "learning_rate": 1.782075051043158e-05, "loss": 0.0634, "step": 70225 }, { "epoch": 1.64, "learning_rate": 1.7819870455985544e-05, "loss": 0.0784, "step": 70230 }, { "epoch": 1.64, "learning_rate": 1.781899040153951e-05, "loss": 0.0425, "step": 70235 }, { "epoch": 1.64, "learning_rate": 1.7818110347093474e-05, "loss": 0.0723, "step": 70240 }, { "epoch": 1.64, "learning_rate": 1.781723029264744e-05, "loss": 0.1718, "step": 70245 }, { "epoch": 1.64, "learning_rate": 1.7816350238201404e-05, "loss": 0.1461, "step": 70250 }, { "epoch": 1.64, "learning_rate": 1.781547018375537e-05, "loss": 0.1083, "step": 70255 }, { "epoch": 1.64, "learning_rate": 1.7814590129309334e-05, "loss": 0.1814, "step": 70260 }, { "epoch": 1.64, "learning_rate": 1.78137100748633e-05, "loss": 0.141, "step": 70265 }, { "epoch": 1.64, "learning_rate": 1.7812830020417264e-05, "loss": 0.0235, "step": 70270 }, { "epoch": 1.64, "learning_rate": 1.781194996597123e-05, "loss": 0.084, "step": 70275 }, { "epoch": 1.64, "learning_rate": 1.7811069911525194e-05, "loss": 0.0815, "step": 70280 }, { "epoch": 1.64, "learning_rate": 1.781018985707916e-05, "loss": 0.0681, "step": 70285 }, { "epoch": 1.64, "learning_rate": 1.7809309802633124e-05, "loss": 0.1558, "step": 70290 }, { "epoch": 1.64, "learning_rate": 1.780842974818709e-05, "loss": 0.0591, "step": 70295 }, { "epoch": 1.64, "learning_rate": 1.7807549693741054e-05, "loss": 0.1396, "step": 70300 }, { "epoch": 1.64, "learning_rate": 1.780666963929502e-05, "loss": 0.376, "step": 70305 }, { "epoch": 1.64, "learning_rate": 1.780578958484898e-05, "loss": 0.2901, "step": 70310 }, { "epoch": 1.64, "learning_rate": 1.780490953040295e-05, "loss": 0.1855, "step": 70315 }, { "epoch": 1.64, "learning_rate": 1.7804029475956914e-05, "loss": 0.0713, "step": 70320 }, { "epoch": 1.64, "learning_rate": 1.780314942151088e-05, "loss": 0.1196, "step": 70325 }, { "epoch": 1.64, "learning_rate": 1.780226936706484e-05, "loss": 0.0249, "step": 70330 }, { "epoch": 1.64, "learning_rate": 1.780138931261881e-05, "loss": 0.1363, "step": 70335 }, { "epoch": 1.64, "learning_rate": 1.7800509258172774e-05, "loss": 0.0439, "step": 70340 }, { "epoch": 1.64, "learning_rate": 1.7799629203726736e-05, "loss": 0.2961, "step": 70345 }, { "epoch": 1.64, "learning_rate": 1.77987491492807e-05, "loss": 0.2181, "step": 70350 }, { "epoch": 1.64, "learning_rate": 1.779786909483467e-05, "loss": 0.2379, "step": 70355 }, { "epoch": 1.64, "learning_rate": 1.7796989040388634e-05, "loss": 0.2971, "step": 70360 }, { "epoch": 1.64, "learning_rate": 1.7796108985942596e-05, "loss": 0.3031, "step": 70365 }, { "epoch": 1.64, "learning_rate": 1.779522893149656e-05, "loss": 0.0097, "step": 70370 }, { "epoch": 1.64, "learning_rate": 1.779434887705053e-05, "loss": 0.03, "step": 70375 }, { "epoch": 1.64, "learning_rate": 1.779346882260449e-05, "loss": 0.1233, "step": 70380 }, { "epoch": 1.64, "learning_rate": 1.7792588768158456e-05, "loss": 0.0258, "step": 70385 }, { "epoch": 1.64, "learning_rate": 1.779170871371242e-05, "loss": 0.0367, "step": 70390 }, { "epoch": 1.64, "learning_rate": 1.779082865926639e-05, "loss": 0.2141, "step": 70395 }, { "epoch": 1.64, "learning_rate": 1.778994860482035e-05, "loss": 0.1395, "step": 70400 }, { "epoch": 1.64, "learning_rate": 1.7789068550374316e-05, "loss": 0.0847, "step": 70405 }, { "epoch": 1.64, "learning_rate": 1.7788188495928284e-05, "loss": 0.3257, "step": 70410 }, { "epoch": 1.64, "learning_rate": 1.778730844148225e-05, "loss": 0.4263, "step": 70415 }, { "epoch": 1.64, "learning_rate": 1.778642838703621e-05, "loss": 0.1214, "step": 70420 }, { "epoch": 1.64, "learning_rate": 1.7785724343479382e-05, "loss": 0.0265, "step": 70425 }, { "epoch": 1.64, "learning_rate": 1.778484428903335e-05, "loss": 0.0653, "step": 70430 }, { "epoch": 1.64, "learning_rate": 1.7783964234587312e-05, "loss": 0.1036, "step": 70435 }, { "epoch": 1.64, "learning_rate": 1.7783084180141277e-05, "loss": 0.1446, "step": 70440 }, { "epoch": 1.64, "learning_rate": 1.7782204125695242e-05, "loss": 0.1055, "step": 70445 }, { "epoch": 1.64, "learning_rate": 1.778132407124921e-05, "loss": 0.214, "step": 70450 }, { "epoch": 1.64, "learning_rate": 1.7780444016803172e-05, "loss": 0.1728, "step": 70455 }, { "epoch": 1.64, "learning_rate": 1.7779563962357137e-05, "loss": 0.4554, "step": 70460 }, { "epoch": 1.64, "learning_rate": 1.7778683907911102e-05, "loss": 0.3847, "step": 70465 }, { "epoch": 1.64, "learning_rate": 1.7777803853465067e-05, "loss": 0.0617, "step": 70470 }, { "epoch": 1.64, "learning_rate": 1.7776923799019032e-05, "loss": 0.0211, "step": 70475 }, { "epoch": 1.64, "learning_rate": 1.7776043744572997e-05, "loss": 0.1398, "step": 70480 }, { "epoch": 1.64, "learning_rate": 1.7775163690126966e-05, "loss": 0.0899, "step": 70485 }, { "epoch": 1.64, "learning_rate": 1.7774283635680928e-05, "loss": 0.104, "step": 70490 }, { "epoch": 1.64, "learning_rate": 1.7773403581234893e-05, "loss": 0.128, "step": 70495 }, { "epoch": 1.64, "learning_rate": 1.7772523526788858e-05, "loss": 0.1612, "step": 70500 }, { "epoch": 1.64, "learning_rate": 1.7771643472342823e-05, "loss": 0.3817, "step": 70505 }, { "epoch": 1.65, "learning_rate": 1.7770763417896788e-05, "loss": 0.2948, "step": 70510 }, { "epoch": 1.65, "learning_rate": 1.7769883363450753e-05, "loss": 0.3017, "step": 70515 }, { "epoch": 1.65, "learning_rate": 1.7769003309004718e-05, "loss": 0.0483, "step": 70520 }, { "epoch": 1.65, "learning_rate": 1.7768123254558683e-05, "loss": 0.0642, "step": 70525 }, { "epoch": 1.65, "learning_rate": 1.7767243200112648e-05, "loss": 0.0442, "step": 70530 }, { "epoch": 1.65, "learning_rate": 1.7766363145666613e-05, "loss": 0.0825, "step": 70535 }, { "epoch": 1.65, "learning_rate": 1.7765483091220574e-05, "loss": 0.1355, "step": 70540 }, { "epoch": 1.65, "learning_rate": 1.7764603036774543e-05, "loss": 0.0707, "step": 70545 }, { "epoch": 1.65, "learning_rate": 1.7763722982328508e-05, "loss": 0.3207, "step": 70550 }, { "epoch": 1.65, "learning_rate": 1.7762842927882473e-05, "loss": 0.1078, "step": 70555 }, { "epoch": 1.65, "learning_rate": 1.7761962873436434e-05, "loss": 0.3157, "step": 70560 }, { "epoch": 1.65, "learning_rate": 1.7761082818990403e-05, "loss": 0.383, "step": 70565 }, { "epoch": 1.65, "learning_rate": 1.7760202764544368e-05, "loss": 0.0874, "step": 70570 }, { "epoch": 1.65, "learning_rate": 1.7759322710098333e-05, "loss": 0.0466, "step": 70575 }, { "epoch": 1.65, "learning_rate": 1.7758442655652294e-05, "loss": 0.0575, "step": 70580 }, { "epoch": 1.65, "learning_rate": 1.7757562601206263e-05, "loss": 0.0684, "step": 70585 }, { "epoch": 1.65, "learning_rate": 1.7756682546760228e-05, "loss": 0.1086, "step": 70590 }, { "epoch": 1.65, "learning_rate": 1.775580249231419e-05, "loss": 0.2588, "step": 70595 }, { "epoch": 1.65, "learning_rate": 1.7754922437868158e-05, "loss": 0.1539, "step": 70600 }, { "epoch": 1.65, "learning_rate": 1.7754042383422123e-05, "loss": 0.2462, "step": 70605 }, { "epoch": 1.65, "learning_rate": 1.7753162328976088e-05, "loss": 0.4064, "step": 70610 }, { "epoch": 1.65, "learning_rate": 1.775228227453005e-05, "loss": 0.2498, "step": 70615 }, { "epoch": 1.65, "learning_rate": 1.7751402220084018e-05, "loss": 0.0379, "step": 70620 }, { "epoch": 1.65, "learning_rate": 1.7750522165637983e-05, "loss": 0.1208, "step": 70625 }, { "epoch": 1.65, "learning_rate": 1.7749642111191945e-05, "loss": 0.1297, "step": 70630 }, { "epoch": 1.65, "learning_rate": 1.774876205674591e-05, "loss": 0.096, "step": 70635 }, { "epoch": 1.65, "learning_rate": 1.7747882002299878e-05, "loss": 0.1593, "step": 70640 }, { "epoch": 1.65, "learning_rate": 1.7747001947853843e-05, "loss": 0.0883, "step": 70645 }, { "epoch": 1.65, "learning_rate": 1.7746121893407805e-05, "loss": 0.2926, "step": 70650 }, { "epoch": 1.65, "learning_rate": 1.774524183896177e-05, "loss": 0.2116, "step": 70655 }, { "epoch": 1.65, "learning_rate": 1.7744361784515738e-05, "loss": 0.2276, "step": 70660 }, { "epoch": 1.65, "learning_rate": 1.77434817300697e-05, "loss": 0.1407, "step": 70665 }, { "epoch": 1.65, "learning_rate": 1.7742601675623665e-05, "loss": 0.0134, "step": 70670 }, { "epoch": 1.65, "learning_rate": 1.774172162117763e-05, "loss": 0.0645, "step": 70675 }, { "epoch": 1.65, "learning_rate": 1.7740841566731598e-05, "loss": 0.0977, "step": 70680 }, { "epoch": 1.65, "learning_rate": 1.773996151228556e-05, "loss": 0.0145, "step": 70685 }, { "epoch": 1.65, "learning_rate": 1.7739081457839525e-05, "loss": 0.2134, "step": 70690 }, { "epoch": 1.65, "learning_rate": 1.773820140339349e-05, "loss": 0.0711, "step": 70695 }, { "epoch": 1.65, "learning_rate": 1.7737321348947455e-05, "loss": 0.1755, "step": 70700 }, { "epoch": 1.65, "learning_rate": 1.773644129450142e-05, "loss": 0.1576, "step": 70705 }, { "epoch": 1.65, "learning_rate": 1.7735561240055385e-05, "loss": 0.279, "step": 70710 }, { "epoch": 1.65, "learning_rate": 1.773468118560935e-05, "loss": 0.3318, "step": 70715 }, { "epoch": 1.65, "learning_rate": 1.7733801131163315e-05, "loss": 0.0146, "step": 70720 }, { "epoch": 1.65, "learning_rate": 1.773292107671728e-05, "loss": 0.1343, "step": 70725 }, { "epoch": 1.65, "learning_rate": 1.7732041022271245e-05, "loss": 0.0068, "step": 70730 }, { "epoch": 1.65, "learning_rate": 1.773116096782521e-05, "loss": 0.1634, "step": 70735 }, { "epoch": 1.65, "learning_rate": 1.7730280913379175e-05, "loss": 0.1674, "step": 70740 }, { "epoch": 1.65, "learning_rate": 1.772940085893314e-05, "loss": 0.2316, "step": 70745 }, { "epoch": 1.65, "learning_rate": 1.7728520804487105e-05, "loss": 0.085, "step": 70750 }, { "epoch": 1.65, "learning_rate": 1.772764075004107e-05, "loss": 0.1296, "step": 70755 }, { "epoch": 1.65, "learning_rate": 1.7726760695595035e-05, "loss": 0.4217, "step": 70760 }, { "epoch": 1.65, "learning_rate": 1.7725880641149e-05, "loss": 0.1154, "step": 70765 }, { "epoch": 1.65, "learning_rate": 1.7725000586702962e-05, "loss": 0.0423, "step": 70770 }, { "epoch": 1.65, "learning_rate": 1.772412053225693e-05, "loss": 0.017, "step": 70775 }, { "epoch": 1.65, "learning_rate": 1.7723240477810895e-05, "loss": 0.0647, "step": 70780 }, { "epoch": 1.65, "learning_rate": 1.772236042336486e-05, "loss": 0.1611, "step": 70785 }, { "epoch": 1.65, "learning_rate": 1.7721480368918822e-05, "loss": 0.1368, "step": 70790 }, { "epoch": 1.65, "learning_rate": 1.772060031447279e-05, "loss": 0.1274, "step": 70795 }, { "epoch": 1.65, "learning_rate": 1.7719720260026755e-05, "loss": 0.255, "step": 70800 }, { "epoch": 1.65, "learning_rate": 1.7718840205580717e-05, "loss": 0.149, "step": 70805 }, { "epoch": 1.65, "learning_rate": 1.7717960151134682e-05, "loss": 0.1287, "step": 70810 }, { "epoch": 1.65, "learning_rate": 1.771708009668865e-05, "loss": 0.2198, "step": 70815 }, { "epoch": 1.65, "learning_rate": 1.7716200042242615e-05, "loss": 0.1131, "step": 70820 }, { "epoch": 1.65, "learning_rate": 1.7715319987796577e-05, "loss": 0.0887, "step": 70825 }, { "epoch": 1.65, "learning_rate": 1.7714439933350542e-05, "loss": 0.0537, "step": 70830 }, { "epoch": 1.65, "learning_rate": 1.771355987890451e-05, "loss": 0.105, "step": 70835 }, { "epoch": 1.65, "learning_rate": 1.7712679824458472e-05, "loss": 0.0762, "step": 70840 }, { "epoch": 1.65, "learning_rate": 1.7711799770012437e-05, "loss": 0.14, "step": 70845 }, { "epoch": 1.65, "learning_rate": 1.7710919715566402e-05, "loss": 0.2643, "step": 70850 }, { "epoch": 1.65, "learning_rate": 1.771003966112037e-05, "loss": 0.3158, "step": 70855 }, { "epoch": 1.65, "learning_rate": 1.7709159606674332e-05, "loss": 0.3512, "step": 70860 }, { "epoch": 1.65, "learning_rate": 1.7708279552228297e-05, "loss": 0.2182, "step": 70865 }, { "epoch": 1.65, "learning_rate": 1.7707399497782262e-05, "loss": 0.0525, "step": 70870 }, { "epoch": 1.65, "learning_rate": 1.7706519443336227e-05, "loss": 0.0334, "step": 70875 }, { "epoch": 1.65, "learning_rate": 1.7705639388890192e-05, "loss": 0.0297, "step": 70880 }, { "epoch": 1.65, "learning_rate": 1.7704759334444157e-05, "loss": 0.1347, "step": 70885 }, { "epoch": 1.65, "learning_rate": 1.7703879279998126e-05, "loss": 0.1442, "step": 70890 }, { "epoch": 1.65, "learning_rate": 1.7702999225552087e-05, "loss": 0.2259, "step": 70895 }, { "epoch": 1.65, "learning_rate": 1.7702119171106052e-05, "loss": 0.1672, "step": 70900 }, { "epoch": 1.65, "learning_rate": 1.7701239116660017e-05, "loss": 0.2278, "step": 70905 }, { "epoch": 1.65, "learning_rate": 1.7700359062213986e-05, "loss": 0.5756, "step": 70910 }, { "epoch": 1.65, "learning_rate": 1.7699479007767947e-05, "loss": 0.1486, "step": 70915 }, { "epoch": 1.65, "learning_rate": 1.7698598953321912e-05, "loss": 0.0531, "step": 70920 }, { "epoch": 1.65, "learning_rate": 1.7697718898875878e-05, "loss": 0.1149, "step": 70925 }, { "epoch": 1.65, "learning_rate": 1.7696838844429843e-05, "loss": 0.1082, "step": 70930 }, { "epoch": 1.65, "learning_rate": 1.7695958789983808e-05, "loss": 0.2277, "step": 70935 }, { "epoch": 1.66, "learning_rate": 1.7695078735537773e-05, "loss": 0.0856, "step": 70940 }, { "epoch": 1.66, "learning_rate": 1.7694198681091738e-05, "loss": 0.1593, "step": 70945 }, { "epoch": 1.66, "learning_rate": 1.7693318626645703e-05, "loss": 0.1376, "step": 70950 }, { "epoch": 1.66, "learning_rate": 1.7692438572199668e-05, "loss": 0.1161, "step": 70955 }, { "epoch": 1.66, "learning_rate": 1.7691558517753633e-05, "loss": 0.2419, "step": 70960 }, { "epoch": 1.66, "learning_rate": 1.7690678463307594e-05, "loss": 0.1649, "step": 70965 }, { "epoch": 1.66, "learning_rate": 1.7689798408861563e-05, "loss": 0.037, "step": 70970 }, { "epoch": 1.66, "learning_rate": 1.7688918354415528e-05, "loss": 0.0396, "step": 70975 }, { "epoch": 1.66, "learning_rate": 1.7688038299969493e-05, "loss": 0.0326, "step": 70980 }, { "epoch": 1.66, "learning_rate": 1.7687158245523454e-05, "loss": 0.1491, "step": 70985 }, { "epoch": 1.66, "learning_rate": 1.7686278191077423e-05, "loss": 0.1493, "step": 70990 }, { "epoch": 1.66, "learning_rate": 1.7685398136631388e-05, "loss": 0.1681, "step": 70995 }, { "epoch": 1.66, "learning_rate": 1.768451808218535e-05, "loss": 0.1799, "step": 71000 }, { "epoch": 1.66, "learning_rate": 1.7683638027739314e-05, "loss": 0.1603, "step": 71005 }, { "epoch": 1.66, "learning_rate": 1.7682757973293283e-05, "loss": 0.2111, "step": 71010 }, { "epoch": 1.66, "learning_rate": 1.7681877918847248e-05, "loss": 0.2713, "step": 71015 }, { "epoch": 1.66, "learning_rate": 1.768099786440121e-05, "loss": 0.0288, "step": 71020 }, { "epoch": 1.66, "learning_rate": 1.7680117809955178e-05, "loss": 0.1085, "step": 71025 }, { "epoch": 1.66, "learning_rate": 1.7679237755509143e-05, "loss": 0.0118, "step": 71030 }, { "epoch": 1.66, "learning_rate": 1.7678357701063105e-05, "loss": 0.0263, "step": 71035 }, { "epoch": 1.66, "learning_rate": 1.767747764661707e-05, "loss": 0.051, "step": 71040 }, { "epoch": 1.66, "learning_rate": 1.7676597592171038e-05, "loss": 0.1268, "step": 71045 }, { "epoch": 1.66, "learning_rate": 1.7675717537725003e-05, "loss": 0.066, "step": 71050 }, { "epoch": 1.66, "learning_rate": 1.7674837483278965e-05, "loss": 0.1391, "step": 71055 }, { "epoch": 1.66, "learning_rate": 1.767395742883293e-05, "loss": 0.0845, "step": 71060 }, { "epoch": 1.66, "learning_rate": 1.7673077374386898e-05, "loss": 0.2518, "step": 71065 }, { "epoch": 1.66, "learning_rate": 1.767219731994086e-05, "loss": 0.023, "step": 71070 }, { "epoch": 1.66, "learning_rate": 1.7671317265494825e-05, "loss": 0.0906, "step": 71075 }, { "epoch": 1.66, "learning_rate": 1.767043721104879e-05, "loss": 0.0621, "step": 71080 }, { "epoch": 1.66, "learning_rate": 1.7669557156602758e-05, "loss": 0.0084, "step": 71085 }, { "epoch": 1.66, "learning_rate": 1.766867710215672e-05, "loss": 0.1103, "step": 71090 }, { "epoch": 1.66, "learning_rate": 1.7667797047710685e-05, "loss": 0.1746, "step": 71095 }, { "epoch": 1.66, "learning_rate": 1.766691699326465e-05, "loss": 0.1218, "step": 71100 }, { "epoch": 1.66, "learning_rate": 1.7666036938818615e-05, "loss": 0.3172, "step": 71105 }, { "epoch": 1.66, "learning_rate": 1.766515688437258e-05, "loss": 0.1901, "step": 71110 }, { "epoch": 1.66, "learning_rate": 1.7664276829926545e-05, "loss": 0.2349, "step": 71115 }, { "epoch": 1.66, "learning_rate": 1.766339677548051e-05, "loss": 0.0041, "step": 71120 }, { "epoch": 1.66, "learning_rate": 1.7662516721034475e-05, "loss": 0.1095, "step": 71125 }, { "epoch": 1.66, "learning_rate": 1.766163666658844e-05, "loss": 0.0765, "step": 71130 }, { "epoch": 1.66, "learning_rate": 1.7660756612142405e-05, "loss": 0.1268, "step": 71135 }, { "epoch": 1.66, "learning_rate": 1.765987655769637e-05, "loss": 0.0218, "step": 71140 }, { "epoch": 1.66, "learning_rate": 1.7658996503250335e-05, "loss": 0.1778, "step": 71145 }, { "epoch": 1.66, "learning_rate": 1.76581164488043e-05, "loss": 0.0665, "step": 71150 }, { "epoch": 1.66, "learning_rate": 1.7657236394358265e-05, "loss": 0.2589, "step": 71155 }, { "epoch": 1.66, "learning_rate": 1.765635633991223e-05, "loss": 0.2813, "step": 71160 }, { "epoch": 1.66, "learning_rate": 1.7655476285466195e-05, "loss": 0.2062, "step": 71165 }, { "epoch": 1.66, "learning_rate": 1.765459623102016e-05, "loss": 0.0368, "step": 71170 }, { "epoch": 1.66, "learning_rate": 1.7653716176574122e-05, "loss": 0.0398, "step": 71175 }, { "epoch": 1.66, "learning_rate": 1.765283612212809e-05, "loss": 0.0419, "step": 71180 }, { "epoch": 1.66, "learning_rate": 1.7651956067682055e-05, "loss": 0.0776, "step": 71185 }, { "epoch": 1.66, "learning_rate": 1.765107601323602e-05, "loss": 0.1326, "step": 71190 }, { "epoch": 1.66, "learning_rate": 1.7650195958789982e-05, "loss": 0.1437, "step": 71195 }, { "epoch": 1.66, "learning_rate": 1.764931590434395e-05, "loss": 0.1966, "step": 71200 }, { "epoch": 1.66, "learning_rate": 1.7648435849897915e-05, "loss": 0.2325, "step": 71205 }, { "epoch": 1.66, "learning_rate": 1.764755579545188e-05, "loss": 0.4026, "step": 71210 }, { "epoch": 1.66, "learning_rate": 1.7646675741005842e-05, "loss": 0.301, "step": 71215 }, { "epoch": 1.66, "learning_rate": 1.764579568655981e-05, "loss": 0.0396, "step": 71220 }, { "epoch": 1.66, "learning_rate": 1.7644915632113775e-05, "loss": 0.016, "step": 71225 }, { "epoch": 1.66, "learning_rate": 1.7644035577667737e-05, "loss": 0.0826, "step": 71230 }, { "epoch": 1.66, "learning_rate": 1.7643155523221702e-05, "loss": 0.1267, "step": 71235 }, { "epoch": 1.66, "learning_rate": 1.764227546877567e-05, "loss": 0.1831, "step": 71240 }, { "epoch": 1.66, "learning_rate": 1.7641395414329635e-05, "loss": 0.099, "step": 71245 }, { "epoch": 1.66, "learning_rate": 1.7640515359883597e-05, "loss": 0.0525, "step": 71250 }, { "epoch": 1.66, "learning_rate": 1.7639635305437562e-05, "loss": 0.4222, "step": 71255 }, { "epoch": 1.66, "learning_rate": 1.763875525099153e-05, "loss": 0.4118, "step": 71260 }, { "epoch": 1.66, "learning_rate": 1.7637875196545492e-05, "loss": 0.4685, "step": 71265 }, { "epoch": 1.66, "learning_rate": 1.7636995142099457e-05, "loss": 0.067, "step": 71270 }, { "epoch": 1.66, "learning_rate": 1.7636115087653422e-05, "loss": 0.0243, "step": 71275 }, { "epoch": 1.66, "learning_rate": 1.763523503320739e-05, "loss": 0.0978, "step": 71280 }, { "epoch": 1.66, "learning_rate": 1.7634354978761352e-05, "loss": 0.1363, "step": 71285 }, { "epoch": 1.66, "learning_rate": 1.7633474924315317e-05, "loss": 0.1131, "step": 71290 }, { "epoch": 1.66, "learning_rate": 1.7632594869869286e-05, "loss": 0.1727, "step": 71295 }, { "epoch": 1.66, "learning_rate": 1.7631714815423247e-05, "loss": 0.1182, "step": 71300 }, { "epoch": 1.66, "learning_rate": 1.7630834760977212e-05, "loss": 0.2113, "step": 71305 }, { "epoch": 1.66, "learning_rate": 1.7629954706531177e-05, "loss": 0.3791, "step": 71310 }, { "epoch": 1.66, "learning_rate": 1.7629074652085146e-05, "loss": 0.19, "step": 71315 }, { "epoch": 1.66, "learning_rate": 1.7628194597639107e-05, "loss": 0.0629, "step": 71320 }, { "epoch": 1.66, "learning_rate": 1.7627314543193072e-05, "loss": 0.0175, "step": 71325 }, { "epoch": 1.66, "learning_rate": 1.7626434488747037e-05, "loss": 0.0405, "step": 71330 }, { "epoch": 1.66, "learning_rate": 1.7625554434301002e-05, "loss": 0.1644, "step": 71335 }, { "epoch": 1.66, "learning_rate": 1.7624674379854967e-05, "loss": 0.0683, "step": 71340 }, { "epoch": 1.66, "learning_rate": 1.7623794325408932e-05, "loss": 0.2177, "step": 71345 }, { "epoch": 1.66, "learning_rate": 1.7622914270962897e-05, "loss": 0.1053, "step": 71350 }, { "epoch": 1.66, "learning_rate": 1.7622034216516863e-05, "loss": 0.1994, "step": 71355 }, { "epoch": 1.66, "learning_rate": 1.7621154162070828e-05, "loss": 0.2064, "step": 71360 }, { "epoch": 1.67, "learning_rate": 1.7620274107624793e-05, "loss": 0.2953, "step": 71365 }, { "epoch": 1.67, "learning_rate": 1.7619394053178754e-05, "loss": 0.1084, "step": 71370 }, { "epoch": 1.67, "learning_rate": 1.7618513998732723e-05, "loss": 0.0438, "step": 71375 }, { "epoch": 1.67, "learning_rate": 1.7617633944286688e-05, "loss": 0.0561, "step": 71380 }, { "epoch": 1.67, "learning_rate": 1.7616753889840653e-05, "loss": 0.1635, "step": 71385 }, { "epoch": 1.67, "learning_rate": 1.7615873835394614e-05, "loss": 0.0603, "step": 71390 }, { "epoch": 1.67, "learning_rate": 1.7614993780948583e-05, "loss": 0.1651, "step": 71395 }, { "epoch": 1.67, "learning_rate": 1.7614113726502548e-05, "loss": 0.1396, "step": 71400 }, { "epoch": 1.67, "learning_rate": 1.761323367205651e-05, "loss": 0.1324, "step": 71405 }, { "epoch": 1.67, "learning_rate": 1.7612353617610474e-05, "loss": 0.2587, "step": 71410 }, { "epoch": 1.67, "learning_rate": 1.7611473563164443e-05, "loss": 0.1733, "step": 71415 }, { "epoch": 1.67, "learning_rate": 1.7610593508718408e-05, "loss": 0.0129, "step": 71420 }, { "epoch": 1.67, "learning_rate": 1.760971345427237e-05, "loss": 0.1178, "step": 71425 }, { "epoch": 1.67, "learning_rate": 1.7608833399826338e-05, "loss": 0.1216, "step": 71430 }, { "epoch": 1.67, "learning_rate": 1.7607953345380303e-05, "loss": 0.0778, "step": 71435 }, { "epoch": 1.67, "learning_rate": 1.7607073290934264e-05, "loss": 0.1056, "step": 71440 }, { "epoch": 1.67, "learning_rate": 1.760619323648823e-05, "loss": 0.1378, "step": 71445 }, { "epoch": 1.67, "learning_rate": 1.7605313182042198e-05, "loss": 0.1544, "step": 71450 }, { "epoch": 1.67, "learning_rate": 1.7604433127596163e-05, "loss": 0.1753, "step": 71455 }, { "epoch": 1.67, "learning_rate": 1.7603553073150125e-05, "loss": 0.2507, "step": 71460 }, { "epoch": 1.67, "learning_rate": 1.760267301870409e-05, "loss": 0.1647, "step": 71465 }, { "epoch": 1.67, "learning_rate": 1.7601792964258058e-05, "loss": 0.0536, "step": 71470 }, { "epoch": 1.67, "learning_rate": 1.760091290981202e-05, "loss": 0.0353, "step": 71475 }, { "epoch": 1.67, "learning_rate": 1.7600032855365985e-05, "loss": 0.0635, "step": 71480 }, { "epoch": 1.67, "learning_rate": 1.759915280091995e-05, "loss": 0.1397, "step": 71485 }, { "epoch": 1.67, "learning_rate": 1.7598272746473918e-05, "loss": 0.0828, "step": 71490 }, { "epoch": 1.67, "learning_rate": 1.759739269202788e-05, "loss": 0.137, "step": 71495 }, { "epoch": 1.67, "learning_rate": 1.7596512637581845e-05, "loss": 0.1167, "step": 71500 }, { "epoch": 1.67, "learning_rate": 1.759563258313581e-05, "loss": 0.2824, "step": 71505 }, { "epoch": 1.67, "learning_rate": 1.7594752528689778e-05, "loss": 0.3682, "step": 71510 }, { "epoch": 1.67, "learning_rate": 1.759387247424374e-05, "loss": 0.3628, "step": 71515 }, { "epoch": 1.67, "learning_rate": 1.7592992419797705e-05, "loss": 0.0491, "step": 71520 }, { "epoch": 1.67, "learning_rate": 1.759211236535167e-05, "loss": 0.0547, "step": 71525 }, { "epoch": 1.67, "learning_rate": 1.7591232310905635e-05, "loss": 0.0438, "step": 71530 }, { "epoch": 1.67, "learning_rate": 1.75903522564596e-05, "loss": 0.0871, "step": 71535 }, { "epoch": 1.67, "learning_rate": 1.7589472202013565e-05, "loss": 0.0593, "step": 71540 }, { "epoch": 1.67, "learning_rate": 1.758859214756753e-05, "loss": 0.1205, "step": 71545 }, { "epoch": 1.67, "learning_rate": 1.7587712093121495e-05, "loss": 0.1203, "step": 71550 }, { "epoch": 1.67, "learning_rate": 1.758683203867546e-05, "loss": 0.1363, "step": 71555 }, { "epoch": 1.67, "learning_rate": 1.7585951984229425e-05, "loss": 0.4331, "step": 71560 }, { "epoch": 1.67, "learning_rate": 1.758507192978339e-05, "loss": 0.2283, "step": 71565 }, { "epoch": 1.67, "learning_rate": 1.7584191875337355e-05, "loss": 0.0225, "step": 71570 }, { "epoch": 1.67, "learning_rate": 1.758331182089132e-05, "loss": 0.0704, "step": 71575 }, { "epoch": 1.67, "learning_rate": 1.7582431766445285e-05, "loss": 0.0776, "step": 71580 }, { "epoch": 1.67, "learning_rate": 1.758155171199925e-05, "loss": 0.0405, "step": 71585 }, { "epoch": 1.67, "learning_rate": 1.7580671657553215e-05, "loss": 0.1485, "step": 71590 }, { "epoch": 1.67, "learning_rate": 1.757979160310718e-05, "loss": 0.1566, "step": 71595 }, { "epoch": 1.67, "learning_rate": 1.7578911548661142e-05, "loss": 0.1565, "step": 71600 }, { "epoch": 1.67, "learning_rate": 1.757803149421511e-05, "loss": 0.2519, "step": 71605 }, { "epoch": 1.67, "learning_rate": 1.7577151439769075e-05, "loss": 0.2061, "step": 71610 }, { "epoch": 1.67, "learning_rate": 1.757627138532304e-05, "loss": 0.1716, "step": 71615 }, { "epoch": 1.67, "learning_rate": 1.7575391330877002e-05, "loss": 0.0452, "step": 71620 }, { "epoch": 1.67, "learning_rate": 1.757451127643097e-05, "loss": 0.0096, "step": 71625 }, { "epoch": 1.67, "learning_rate": 1.7573631221984935e-05, "loss": 0.3128, "step": 71630 }, { "epoch": 1.67, "learning_rate": 1.7572751167538897e-05, "loss": 0.0796, "step": 71635 }, { "epoch": 1.67, "learning_rate": 1.7571871113092862e-05, "loss": 0.1759, "step": 71640 }, { "epoch": 1.67, "learning_rate": 1.757099105864683e-05, "loss": 0.1272, "step": 71645 }, { "epoch": 1.67, "learning_rate": 1.7570111004200795e-05, "loss": 0.1697, "step": 71650 }, { "epoch": 1.67, "learning_rate": 1.7569230949754757e-05, "loss": 0.1827, "step": 71655 }, { "epoch": 1.67, "learning_rate": 1.7568350895308722e-05, "loss": 0.2381, "step": 71660 }, { "epoch": 1.67, "learning_rate": 1.756747084086269e-05, "loss": 0.1224, "step": 71665 }, { "epoch": 1.67, "learning_rate": 1.7566590786416652e-05, "loss": 0.0779, "step": 71670 }, { "epoch": 1.67, "learning_rate": 1.7565710731970617e-05, "loss": 0.0857, "step": 71675 }, { "epoch": 1.67, "learning_rate": 1.7564830677524582e-05, "loss": 0.0491, "step": 71680 }, { "epoch": 1.67, "learning_rate": 1.756395062307855e-05, "loss": 0.0237, "step": 71685 }, { "epoch": 1.67, "learning_rate": 1.7563070568632512e-05, "loss": 0.1224, "step": 71690 }, { "epoch": 1.67, "learning_rate": 1.7562190514186477e-05, "loss": 0.0876, "step": 71695 }, { "epoch": 1.67, "learning_rate": 1.7561310459740446e-05, "loss": 0.1031, "step": 71700 }, { "epoch": 1.67, "learning_rate": 1.7560430405294407e-05, "loss": 0.1105, "step": 71705 }, { "epoch": 1.67, "learning_rate": 1.7559550350848372e-05, "loss": 0.3289, "step": 71710 }, { "epoch": 1.67, "learning_rate": 1.7558670296402337e-05, "loss": 0.2575, "step": 71715 }, { "epoch": 1.67, "learning_rate": 1.7557790241956306e-05, "loss": 0.0408, "step": 71720 }, { "epoch": 1.67, "learning_rate": 1.7556910187510267e-05, "loss": 0.1123, "step": 71725 }, { "epoch": 1.67, "learning_rate": 1.7556030133064232e-05, "loss": 0.0469, "step": 71730 }, { "epoch": 1.67, "learning_rate": 1.7555150078618197e-05, "loss": 0.0499, "step": 71735 }, { "epoch": 1.67, "learning_rate": 1.7554270024172162e-05, "loss": 0.0639, "step": 71740 }, { "epoch": 1.67, "learning_rate": 1.7553389969726127e-05, "loss": 0.1456, "step": 71745 }, { "epoch": 1.67, "learning_rate": 1.7552509915280092e-05, "loss": 0.1561, "step": 71750 }, { "epoch": 1.67, "learning_rate": 1.7551629860834057e-05, "loss": 0.201, "step": 71755 }, { "epoch": 1.67, "learning_rate": 1.7550749806388022e-05, "loss": 0.3642, "step": 71760 }, { "epoch": 1.67, "learning_rate": 1.7549869751941987e-05, "loss": 0.4458, "step": 71765 }, { "epoch": 1.67, "learning_rate": 1.7548989697495952e-05, "loss": 0.0469, "step": 71770 }, { "epoch": 1.67, "learning_rate": 1.7548109643049914e-05, "loss": 0.0108, "step": 71775 }, { "epoch": 1.67, "learning_rate": 1.7547229588603882e-05, "loss": 0.0158, "step": 71780 }, { "epoch": 1.67, "learning_rate": 1.7546349534157847e-05, "loss": 0.1203, "step": 71785 }, { "epoch": 1.67, "learning_rate": 1.7545469479711813e-05, "loss": 0.1594, "step": 71790 }, { "epoch": 1.68, "learning_rate": 1.7544589425265774e-05, "loss": 0.1475, "step": 71795 }, { "epoch": 1.68, "learning_rate": 1.7543709370819743e-05, "loss": 0.0421, "step": 71800 }, { "epoch": 1.68, "learning_rate": 1.7542829316373708e-05, "loss": 0.2468, "step": 71805 }, { "epoch": 1.68, "learning_rate": 1.7541949261927673e-05, "loss": 0.5781, "step": 71810 }, { "epoch": 1.68, "learning_rate": 1.7541069207481634e-05, "loss": 0.3332, "step": 71815 }, { "epoch": 1.68, "learning_rate": 1.7540189153035603e-05, "loss": 0.0449, "step": 71820 }, { "epoch": 1.68, "learning_rate": 1.7539309098589568e-05, "loss": 0.0606, "step": 71825 }, { "epoch": 1.68, "learning_rate": 1.753842904414353e-05, "loss": 0.0847, "step": 71830 }, { "epoch": 1.68, "learning_rate": 1.7537548989697498e-05, "loss": 0.0701, "step": 71835 }, { "epoch": 1.68, "learning_rate": 1.7536668935251463e-05, "loss": 0.0708, "step": 71840 }, { "epoch": 1.68, "learning_rate": 1.7535788880805428e-05, "loss": 0.1394, "step": 71845 }, { "epoch": 1.68, "learning_rate": 1.753490882635939e-05, "loss": 0.1933, "step": 71850 }, { "epoch": 1.68, "learning_rate": 1.7534028771913358e-05, "loss": 0.1062, "step": 71855 }, { "epoch": 1.68, "learning_rate": 1.7533148717467323e-05, "loss": 0.1126, "step": 71860 }, { "epoch": 1.68, "learning_rate": 1.7532268663021284e-05, "loss": 0.3413, "step": 71865 }, { "epoch": 1.68, "learning_rate": 1.753138860857525e-05, "loss": 0.0705, "step": 71870 }, { "epoch": 1.68, "learning_rate": 1.7530508554129218e-05, "loss": 0.0623, "step": 71875 }, { "epoch": 1.68, "learning_rate": 1.7529628499683183e-05, "loss": 0.0623, "step": 71880 }, { "epoch": 1.68, "learning_rate": 1.7528748445237145e-05, "loss": 0.0201, "step": 71885 }, { "epoch": 1.68, "learning_rate": 1.752786839079111e-05, "loss": 0.1657, "step": 71890 }, { "epoch": 1.68, "learning_rate": 1.7526988336345078e-05, "loss": 0.0315, "step": 71895 }, { "epoch": 1.68, "learning_rate": 1.752610828189904e-05, "loss": 0.1397, "step": 71900 }, { "epoch": 1.68, "learning_rate": 1.7525228227453005e-05, "loss": 0.2996, "step": 71905 }, { "epoch": 1.68, "learning_rate": 1.752434817300697e-05, "loss": 0.3789, "step": 71910 }, { "epoch": 1.68, "learning_rate": 1.7523468118560938e-05, "loss": 0.2121, "step": 71915 }, { "epoch": 1.68, "learning_rate": 1.75225880641149e-05, "loss": 0.08, "step": 71920 }, { "epoch": 1.68, "learning_rate": 1.7521708009668865e-05, "loss": 0.0716, "step": 71925 }, { "epoch": 1.68, "learning_rate": 1.752082795522283e-05, "loss": 0.1299, "step": 71930 }, { "epoch": 1.68, "learning_rate": 1.7519947900776795e-05, "loss": 0.0683, "step": 71935 }, { "epoch": 1.68, "learning_rate": 1.751906784633076e-05, "loss": 0.0994, "step": 71940 }, { "epoch": 1.68, "learning_rate": 1.7518187791884725e-05, "loss": 0.1125, "step": 71945 }, { "epoch": 1.68, "learning_rate": 1.751730773743869e-05, "loss": 0.0374, "step": 71950 }, { "epoch": 1.68, "learning_rate": 1.7516427682992655e-05, "loss": 0.3135, "step": 71955 }, { "epoch": 1.68, "learning_rate": 1.751554762854662e-05, "loss": 0.4645, "step": 71960 }, { "epoch": 1.68, "learning_rate": 1.7514667574100585e-05, "loss": 0.3063, "step": 71965 }, { "epoch": 1.68, "learning_rate": 1.751378751965455e-05, "loss": 0.0971, "step": 71970 }, { "epoch": 1.68, "learning_rate": 1.7512907465208515e-05, "loss": 0.056, "step": 71975 }, { "epoch": 1.68, "learning_rate": 1.751202741076248e-05, "loss": 0.0554, "step": 71980 }, { "epoch": 1.68, "learning_rate": 1.7511147356316445e-05, "loss": 0.0471, "step": 71985 }, { "epoch": 1.68, "learning_rate": 1.751026730187041e-05, "loss": 0.0826, "step": 71990 }, { "epoch": 1.68, "learning_rate": 1.7509387247424375e-05, "loss": 0.1233, "step": 71995 }, { "epoch": 1.68, "learning_rate": 1.750850719297834e-05, "loss": 0.0997, "step": 72000 }, { "epoch": 1.68, "learning_rate": 1.75076271385323e-05, "loss": 0.3032, "step": 72005 }, { "epoch": 1.68, "learning_rate": 1.750674708408627e-05, "loss": 0.3664, "step": 72010 }, { "epoch": 1.68, "learning_rate": 1.7505867029640235e-05, "loss": 0.1917, "step": 72015 }, { "epoch": 1.68, "learning_rate": 1.75049869751942e-05, "loss": 0.0695, "step": 72020 }, { "epoch": 1.68, "learning_rate": 1.7504106920748162e-05, "loss": 0.019, "step": 72025 }, { "epoch": 1.68, "learning_rate": 1.750322686630213e-05, "loss": 0.0625, "step": 72030 }, { "epoch": 1.68, "learning_rate": 1.7502346811856095e-05, "loss": 0.0592, "step": 72035 }, { "epoch": 1.68, "learning_rate": 1.7501466757410057e-05, "loss": 0.1607, "step": 72040 }, { "epoch": 1.68, "learning_rate": 1.7500586702964022e-05, "loss": 0.0878, "step": 72045 }, { "epoch": 1.68, "learning_rate": 1.749970664851799e-05, "loss": 0.1244, "step": 72050 }, { "epoch": 1.68, "learning_rate": 1.7498826594071955e-05, "loss": 0.1566, "step": 72055 }, { "epoch": 1.68, "learning_rate": 1.7497946539625917e-05, "loss": 0.5511, "step": 72060 }, { "epoch": 1.68, "learning_rate": 1.7497066485179882e-05, "loss": 0.3172, "step": 72065 }, { "epoch": 1.68, "learning_rate": 1.749618643073385e-05, "loss": 0.0093, "step": 72070 }, { "epoch": 1.68, "learning_rate": 1.7495306376287812e-05, "loss": 0.0287, "step": 72075 }, { "epoch": 1.68, "learning_rate": 1.7494426321841777e-05, "loss": 0.0796, "step": 72080 }, { "epoch": 1.68, "learning_rate": 1.7493546267395742e-05, "loss": 0.0848, "step": 72085 }, { "epoch": 1.68, "learning_rate": 1.749266621294971e-05, "loss": 0.0859, "step": 72090 }, { "epoch": 1.68, "learning_rate": 1.7491786158503672e-05, "loss": 0.1956, "step": 72095 }, { "epoch": 1.68, "learning_rate": 1.7490906104057637e-05, "loss": 0.1632, "step": 72100 }, { "epoch": 1.68, "learning_rate": 1.7490026049611605e-05, "loss": 0.1287, "step": 72105 }, { "epoch": 1.68, "learning_rate": 1.748914599516557e-05, "loss": 0.2881, "step": 72110 }, { "epoch": 1.68, "learning_rate": 1.7488265940719532e-05, "loss": 0.2134, "step": 72115 }, { "epoch": 1.68, "learning_rate": 1.7487385886273497e-05, "loss": 0.0336, "step": 72120 }, { "epoch": 1.68, "learning_rate": 1.7486505831827466e-05, "loss": 0.0296, "step": 72125 }, { "epoch": 1.68, "learning_rate": 1.7485625777381427e-05, "loss": 0.1242, "step": 72130 }, { "epoch": 1.68, "learning_rate": 1.7484745722935392e-05, "loss": 0.0522, "step": 72135 }, { "epoch": 1.68, "learning_rate": 1.7483865668489357e-05, "loss": 0.1008, "step": 72140 }, { "epoch": 1.68, "learning_rate": 1.7482985614043326e-05, "loss": 0.1357, "step": 72145 }, { "epoch": 1.68, "learning_rate": 1.7482105559597287e-05, "loss": 0.1584, "step": 72150 }, { "epoch": 1.68, "learning_rate": 1.7481225505151252e-05, "loss": 0.1281, "step": 72155 }, { "epoch": 1.68, "learning_rate": 1.7480345450705217e-05, "loss": 0.6564, "step": 72160 }, { "epoch": 1.68, "learning_rate": 1.7479465396259182e-05, "loss": 0.2614, "step": 72165 }, { "epoch": 1.68, "learning_rate": 1.7478585341813147e-05, "loss": 0.0594, "step": 72170 }, { "epoch": 1.68, "learning_rate": 1.7477705287367112e-05, "loss": 0.0256, "step": 72175 }, { "epoch": 1.68, "learning_rate": 1.7476825232921077e-05, "loss": 0.0874, "step": 72180 }, { "epoch": 1.68, "learning_rate": 1.7475945178475042e-05, "loss": 0.1824, "step": 72185 }, { "epoch": 1.68, "learning_rate": 1.7475065124029007e-05, "loss": 0.0865, "step": 72190 }, { "epoch": 1.68, "learning_rate": 1.7474185069582972e-05, "loss": 0.1937, "step": 72195 }, { "epoch": 1.68, "learning_rate": 1.7473305015136934e-05, "loss": 0.2075, "step": 72200 }, { "epoch": 1.68, "learning_rate": 1.7472424960690902e-05, "loss": 0.1904, "step": 72205 }, { "epoch": 1.68, "learning_rate": 1.7471544906244867e-05, "loss": 0.3172, "step": 72210 }, { "epoch": 1.68, "learning_rate": 1.7470664851798832e-05, "loss": 0.1678, "step": 72215 }, { "epoch": 1.68, "learning_rate": 1.7469784797352794e-05, "loss": 0.0291, "step": 72220 }, { "epoch": 1.69, "learning_rate": 1.7468904742906763e-05, "loss": 0.0596, "step": 72225 }, { "epoch": 1.69, "learning_rate": 1.7468024688460728e-05, "loss": 0.0275, "step": 72230 }, { "epoch": 1.69, "learning_rate": 1.746714463401469e-05, "loss": 0.0834, "step": 72235 }, { "epoch": 1.69, "learning_rate": 1.7466264579568658e-05, "loss": 0.1038, "step": 72240 }, { "epoch": 1.69, "learning_rate": 1.7465384525122623e-05, "loss": 0.0993, "step": 72245 }, { "epoch": 1.69, "learning_rate": 1.7464504470676588e-05, "loss": 0.1837, "step": 72250 }, { "epoch": 1.69, "learning_rate": 1.746362441623055e-05, "loss": 0.2535, "step": 72255 }, { "epoch": 1.69, "learning_rate": 1.7462744361784518e-05, "loss": 0.3694, "step": 72260 }, { "epoch": 1.69, "learning_rate": 1.7461864307338483e-05, "loss": 0.1605, "step": 72265 }, { "epoch": 1.69, "learning_rate": 1.7460984252892444e-05, "loss": 0.081, "step": 72270 }, { "epoch": 1.69, "learning_rate": 1.746010419844641e-05, "loss": 0.04, "step": 72275 }, { "epoch": 1.69, "learning_rate": 1.7459224144000378e-05, "loss": 0.0274, "step": 72280 }, { "epoch": 1.69, "learning_rate": 1.7458344089554343e-05, "loss": 0.0655, "step": 72285 }, { "epoch": 1.69, "learning_rate": 1.7457464035108304e-05, "loss": 0.1064, "step": 72290 }, { "epoch": 1.69, "learning_rate": 1.745658398066227e-05, "loss": 0.0459, "step": 72295 }, { "epoch": 1.69, "learning_rate": 1.7455703926216238e-05, "loss": 0.2314, "step": 72300 }, { "epoch": 1.69, "learning_rate": 1.74548238717702e-05, "loss": 0.1783, "step": 72305 }, { "epoch": 1.69, "learning_rate": 1.7453943817324164e-05, "loss": 0.3264, "step": 72310 }, { "epoch": 1.69, "learning_rate": 1.745306376287813e-05, "loss": 0.288, "step": 72315 }, { "epoch": 1.69, "learning_rate": 1.7452183708432098e-05, "loss": 0.0047, "step": 72320 }, { "epoch": 1.69, "learning_rate": 1.745130365398606e-05, "loss": 0.116, "step": 72325 }, { "epoch": 1.69, "learning_rate": 1.7450423599540025e-05, "loss": 0.1062, "step": 72330 }, { "epoch": 1.69, "learning_rate": 1.744954354509399e-05, "loss": 0.0803, "step": 72335 }, { "epoch": 1.69, "learning_rate": 1.7448663490647955e-05, "loss": 0.0938, "step": 72340 }, { "epoch": 1.69, "learning_rate": 1.744778343620192e-05, "loss": 0.1079, "step": 72345 }, { "epoch": 1.69, "learning_rate": 1.7446903381755885e-05, "loss": 0.309, "step": 72350 }, { "epoch": 1.69, "learning_rate": 1.744602332730985e-05, "loss": 0.3113, "step": 72355 }, { "epoch": 1.69, "learning_rate": 1.7445143272863815e-05, "loss": 0.2317, "step": 72360 }, { "epoch": 1.69, "learning_rate": 1.744426321841778e-05, "loss": 0.244, "step": 72365 }, { "epoch": 1.69, "learning_rate": 1.7443383163971745e-05, "loss": 0.0694, "step": 72370 }, { "epoch": 1.69, "learning_rate": 1.744250310952571e-05, "loss": 0.0832, "step": 72375 }, { "epoch": 1.69, "learning_rate": 1.7441623055079675e-05, "loss": 0.0408, "step": 72380 }, { "epoch": 1.69, "learning_rate": 1.744074300063364e-05, "loss": 0.1308, "step": 72385 }, { "epoch": 1.69, "learning_rate": 1.7439862946187605e-05, "loss": 0.0149, "step": 72390 }, { "epoch": 1.69, "learning_rate": 1.743898289174157e-05, "loss": 0.0668, "step": 72395 }, { "epoch": 1.69, "learning_rate": 1.7438102837295535e-05, "loss": 0.1225, "step": 72400 }, { "epoch": 1.69, "learning_rate": 1.74372227828495e-05, "loss": 0.2087, "step": 72405 }, { "epoch": 1.69, "learning_rate": 1.7436342728403465e-05, "loss": 0.392, "step": 72410 }, { "epoch": 1.69, "learning_rate": 1.743546267395743e-05, "loss": 0.3978, "step": 72415 }, { "epoch": 1.69, "learning_rate": 1.7434582619511395e-05, "loss": 0.0306, "step": 72420 }, { "epoch": 1.69, "learning_rate": 1.743370256506536e-05, "loss": 0.0551, "step": 72425 }, { "epoch": 1.69, "learning_rate": 1.743282251061932e-05, "loss": 0.0543, "step": 72430 }, { "epoch": 1.69, "learning_rate": 1.743194245617329e-05, "loss": 0.057, "step": 72435 }, { "epoch": 1.69, "learning_rate": 1.7431062401727255e-05, "loss": 0.1698, "step": 72440 }, { "epoch": 1.69, "learning_rate": 1.743018234728122e-05, "loss": 0.1334, "step": 72445 }, { "epoch": 1.69, "learning_rate": 1.742930229283518e-05, "loss": 0.0936, "step": 72450 }, { "epoch": 1.69, "learning_rate": 1.742842223838915e-05, "loss": 0.1935, "step": 72455 }, { "epoch": 1.69, "learning_rate": 1.7427542183943115e-05, "loss": 0.1546, "step": 72460 }, { "epoch": 1.69, "learning_rate": 1.7426662129497077e-05, "loss": 0.148, "step": 72465 }, { "epoch": 1.69, "learning_rate": 1.7425782075051042e-05, "loss": 0.0264, "step": 72470 }, { "epoch": 1.69, "learning_rate": 1.742490202060501e-05, "loss": 0.034, "step": 72475 }, { "epoch": 1.69, "learning_rate": 1.7424021966158975e-05, "loss": 0.0939, "step": 72480 }, { "epoch": 1.69, "learning_rate": 1.7423141911712937e-05, "loss": 0.2346, "step": 72485 }, { "epoch": 1.69, "learning_rate": 1.7422261857266902e-05, "loss": 0.056, "step": 72490 }, { "epoch": 1.69, "learning_rate": 1.742138180282087e-05, "loss": 0.0844, "step": 72495 }, { "epoch": 1.69, "learning_rate": 1.7420501748374832e-05, "loss": 0.2256, "step": 72500 }, { "epoch": 1.69, "learning_rate": 1.7419621693928797e-05, "loss": 0.142, "step": 72505 }, { "epoch": 1.69, "learning_rate": 1.7418741639482765e-05, "loss": 0.269, "step": 72510 }, { "epoch": 1.69, "learning_rate": 1.741786158503673e-05, "loss": 0.1818, "step": 72515 }, { "epoch": 1.69, "learning_rate": 1.7416981530590692e-05, "loss": 0.053, "step": 72520 }, { "epoch": 1.69, "learning_rate": 1.7416101476144657e-05, "loss": 0.0803, "step": 72525 }, { "epoch": 1.69, "learning_rate": 1.7415221421698625e-05, "loss": 0.0399, "step": 72530 }, { "epoch": 1.69, "learning_rate": 1.7414341367252587e-05, "loss": 0.0491, "step": 72535 }, { "epoch": 1.69, "learning_rate": 1.7413461312806552e-05, "loss": 0.0976, "step": 72540 }, { "epoch": 1.69, "learning_rate": 1.7412581258360517e-05, "loss": 0.1302, "step": 72545 }, { "epoch": 1.69, "learning_rate": 1.7411701203914485e-05, "loss": 0.1872, "step": 72550 }, { "epoch": 1.69, "learning_rate": 1.7410821149468447e-05, "loss": 0.3811, "step": 72555 }, { "epoch": 1.69, "learning_rate": 1.7409941095022412e-05, "loss": 0.3246, "step": 72560 }, { "epoch": 1.69, "learning_rate": 1.7409061040576377e-05, "loss": 0.2831, "step": 72565 }, { "epoch": 1.69, "learning_rate": 1.7408180986130342e-05, "loss": 0.0009, "step": 72570 }, { "epoch": 1.69, "learning_rate": 1.7407300931684307e-05, "loss": 0.0271, "step": 72575 }, { "epoch": 1.69, "learning_rate": 1.7406420877238272e-05, "loss": 0.0512, "step": 72580 }, { "epoch": 1.69, "learning_rate": 1.7405540822792237e-05, "loss": 0.0797, "step": 72585 }, { "epoch": 1.69, "learning_rate": 1.7404660768346202e-05, "loss": 0.1153, "step": 72590 }, { "epoch": 1.69, "learning_rate": 1.7403780713900167e-05, "loss": 0.2276, "step": 72595 }, { "epoch": 1.69, "learning_rate": 1.7402900659454132e-05, "loss": 0.1187, "step": 72600 }, { "epoch": 1.69, "learning_rate": 1.7402020605008094e-05, "loss": 0.2969, "step": 72605 }, { "epoch": 1.69, "learning_rate": 1.7401140550562062e-05, "loss": 0.4148, "step": 72610 }, { "epoch": 1.69, "learning_rate": 1.7400260496116027e-05, "loss": 0.1933, "step": 72615 }, { "epoch": 1.69, "learning_rate": 1.7399380441669992e-05, "loss": 0.0878, "step": 72620 }, { "epoch": 1.69, "learning_rate": 1.7398500387223954e-05, "loss": 0.1023, "step": 72625 }, { "epoch": 1.69, "learning_rate": 1.7397620332777922e-05, "loss": 0.0668, "step": 72630 }, { "epoch": 1.69, "learning_rate": 1.7396740278331887e-05, "loss": 0.0136, "step": 72635 }, { "epoch": 1.69, "learning_rate": 1.739586022388585e-05, "loss": 0.1118, "step": 72640 }, { "epoch": 1.69, "learning_rate": 1.7394980169439817e-05, "loss": 0.0908, "step": 72645 }, { "epoch": 1.7, "learning_rate": 1.7394100114993782e-05, "loss": 0.1546, "step": 72650 }, { "epoch": 1.7, "learning_rate": 1.7393220060547748e-05, "loss": 0.102, "step": 72655 }, { "epoch": 1.7, "learning_rate": 1.739234000610171e-05, "loss": 0.312, "step": 72660 }, { "epoch": 1.7, "learning_rate": 1.7391459951655678e-05, "loss": 0.423, "step": 72665 }, { "epoch": 1.7, "learning_rate": 1.7390579897209643e-05, "loss": 0.0993, "step": 72670 }, { "epoch": 1.7, "learning_rate": 1.7389699842763604e-05, "loss": 0.0866, "step": 72675 }, { "epoch": 1.7, "learning_rate": 1.738881978831757e-05, "loss": 0.0726, "step": 72680 }, { "epoch": 1.7, "learning_rate": 1.7387939733871538e-05, "loss": 0.106, "step": 72685 }, { "epoch": 1.7, "learning_rate": 1.7387059679425503e-05, "loss": 0.0855, "step": 72690 }, { "epoch": 1.7, "learning_rate": 1.7386179624979464e-05, "loss": 0.0642, "step": 72695 }, { "epoch": 1.7, "learning_rate": 1.738529957053343e-05, "loss": 0.1493, "step": 72700 }, { "epoch": 1.7, "learning_rate": 1.7384419516087398e-05, "loss": 0.245, "step": 72705 }, { "epoch": 1.7, "learning_rate": 1.738353946164136e-05, "loss": 0.2552, "step": 72710 }, { "epoch": 1.7, "learning_rate": 1.7382659407195324e-05, "loss": 0.1976, "step": 72715 }, { "epoch": 1.7, "learning_rate": 1.738177935274929e-05, "loss": 0.1226, "step": 72720 }, { "epoch": 1.7, "learning_rate": 1.7380899298303258e-05, "loss": 0.0302, "step": 72725 }, { "epoch": 1.7, "learning_rate": 1.738001924385722e-05, "loss": 0.0364, "step": 72730 }, { "epoch": 1.7, "learning_rate": 1.7379139189411184e-05, "loss": 0.0732, "step": 72735 }, { "epoch": 1.7, "learning_rate": 1.737825913496515e-05, "loss": 0.1038, "step": 72740 }, { "epoch": 1.7, "learning_rate": 1.7377379080519118e-05, "loss": 0.0613, "step": 72745 }, { "epoch": 1.7, "learning_rate": 1.737649902607308e-05, "loss": 0.0437, "step": 72750 }, { "epoch": 1.7, "learning_rate": 1.7375618971627045e-05, "loss": 0.1634, "step": 72755 }, { "epoch": 1.7, "learning_rate": 1.737473891718101e-05, "loss": 0.3913, "step": 72760 }, { "epoch": 1.7, "learning_rate": 1.7373858862734975e-05, "loss": 0.2238, "step": 72765 }, { "epoch": 1.7, "learning_rate": 1.737297880828894e-05, "loss": 0.0892, "step": 72770 }, { "epoch": 1.7, "learning_rate": 1.7372098753842905e-05, "loss": 0.0471, "step": 72775 }, { "epoch": 1.7, "learning_rate": 1.7371218699396873e-05, "loss": 0.1005, "step": 72780 }, { "epoch": 1.7, "learning_rate": 1.7370338644950835e-05, "loss": 0.1092, "step": 72785 }, { "epoch": 1.7, "learning_rate": 1.73694585905048e-05, "loss": 0.0284, "step": 72790 }, { "epoch": 1.7, "learning_rate": 1.7368578536058765e-05, "loss": 0.0318, "step": 72795 }, { "epoch": 1.7, "learning_rate": 1.736769848161273e-05, "loss": 0.3433, "step": 72800 }, { "epoch": 1.7, "learning_rate": 1.7366818427166695e-05, "loss": 0.1008, "step": 72805 }, { "epoch": 1.7, "learning_rate": 1.736593837272066e-05, "loss": 0.3252, "step": 72810 }, { "epoch": 1.7, "learning_rate": 1.7365058318274625e-05, "loss": 0.2092, "step": 72815 }, { "epoch": 1.7, "learning_rate": 1.736417826382859e-05, "loss": 0.0028, "step": 72820 }, { "epoch": 1.7, "learning_rate": 1.7363298209382555e-05, "loss": 0.1173, "step": 72825 }, { "epoch": 1.7, "learning_rate": 1.736241815493652e-05, "loss": 0.0884, "step": 72830 }, { "epoch": 1.7, "learning_rate": 1.736153810049048e-05, "loss": 0.1348, "step": 72835 }, { "epoch": 1.7, "learning_rate": 1.736065804604445e-05, "loss": 0.0412, "step": 72840 }, { "epoch": 1.7, "learning_rate": 1.7359777991598415e-05, "loss": 0.0567, "step": 72845 }, { "epoch": 1.7, "learning_rate": 1.735889793715238e-05, "loss": 0.1148, "step": 72850 }, { "epoch": 1.7, "learning_rate": 1.735801788270634e-05, "loss": 0.1931, "step": 72855 }, { "epoch": 1.7, "learning_rate": 1.735713782826031e-05, "loss": 0.3413, "step": 72860 }, { "epoch": 1.7, "learning_rate": 1.7356257773814275e-05, "loss": 0.1881, "step": 72865 }, { "epoch": 1.7, "learning_rate": 1.7355377719368237e-05, "loss": 0.0023, "step": 72870 }, { "epoch": 1.7, "learning_rate": 1.73544976649222e-05, "loss": 0.0389, "step": 72875 }, { "epoch": 1.7, "learning_rate": 1.735361761047617e-05, "loss": 0.0533, "step": 72880 }, { "epoch": 1.7, "learning_rate": 1.7352737556030135e-05, "loss": 0.134, "step": 72885 }, { "epoch": 1.7, "learning_rate": 1.7351857501584097e-05, "loss": 0.1498, "step": 72890 }, { "epoch": 1.7, "learning_rate": 1.7350977447138062e-05, "loss": 0.0713, "step": 72895 }, { "epoch": 1.7, "learning_rate": 1.735009739269203e-05, "loss": 0.2101, "step": 72900 }, { "epoch": 1.7, "learning_rate": 1.7349217338245992e-05, "loss": 0.2206, "step": 72905 }, { "epoch": 1.7, "learning_rate": 1.7348337283799957e-05, "loss": 0.3379, "step": 72910 }, { "epoch": 1.7, "learning_rate": 1.7347457229353925e-05, "loss": 0.2253, "step": 72915 }, { "epoch": 1.7, "learning_rate": 1.734657717490789e-05, "loss": 0.014, "step": 72920 }, { "epoch": 1.7, "learning_rate": 1.7345697120461852e-05, "loss": 0.0605, "step": 72925 }, { "epoch": 1.7, "learning_rate": 1.7344817066015817e-05, "loss": 0.1045, "step": 72930 }, { "epoch": 1.7, "learning_rate": 1.7343937011569785e-05, "loss": 0.0401, "step": 72935 }, { "epoch": 1.7, "learning_rate": 1.7343056957123747e-05, "loss": 0.1299, "step": 72940 }, { "epoch": 1.7, "learning_rate": 1.7342176902677712e-05, "loss": 0.0767, "step": 72945 }, { "epoch": 1.7, "learning_rate": 1.7341296848231677e-05, "loss": 0.2017, "step": 72950 }, { "epoch": 1.7, "learning_rate": 1.7340416793785645e-05, "loss": 0.2994, "step": 72955 }, { "epoch": 1.7, "learning_rate": 1.7339536739339607e-05, "loss": 0.2012, "step": 72960 }, { "epoch": 1.7, "learning_rate": 1.7338656684893572e-05, "loss": 0.213, "step": 72965 }, { "epoch": 1.7, "learning_rate": 1.7337776630447537e-05, "loss": 0.0216, "step": 72970 }, { "epoch": 1.7, "learning_rate": 1.7336896576001502e-05, "loss": 0.0696, "step": 72975 }, { "epoch": 1.7, "learning_rate": 1.7336016521555467e-05, "loss": 0.0159, "step": 72980 }, { "epoch": 1.7, "learning_rate": 1.7335136467109432e-05, "loss": 0.0801, "step": 72985 }, { "epoch": 1.7, "learning_rate": 1.7334256412663397e-05, "loss": 0.0472, "step": 72990 }, { "epoch": 1.7, "learning_rate": 1.7333376358217362e-05, "loss": 0.0967, "step": 72995 }, { "epoch": 1.7, "learning_rate": 1.7332496303771327e-05, "loss": 0.0557, "step": 73000 }, { "epoch": 1.7, "learning_rate": 1.7331616249325292e-05, "loss": 0.2252, "step": 73005 }, { "epoch": 1.7, "learning_rate": 1.7330736194879254e-05, "loss": 0.3392, "step": 73010 }, { "epoch": 1.7, "learning_rate": 1.7329856140433222e-05, "loss": 0.2639, "step": 73015 }, { "epoch": 1.7, "learning_rate": 1.7328976085987187e-05, "loss": 0.1121, "step": 73020 }, { "epoch": 1.7, "learning_rate": 1.7328096031541152e-05, "loss": 0.0669, "step": 73025 }, { "epoch": 1.7, "learning_rate": 1.7327215977095114e-05, "loss": 0.0744, "step": 73030 }, { "epoch": 1.7, "learning_rate": 1.7326335922649082e-05, "loss": 0.0354, "step": 73035 }, { "epoch": 1.7, "learning_rate": 1.7325455868203047e-05, "loss": 0.0924, "step": 73040 }, { "epoch": 1.7, "learning_rate": 1.7324575813757012e-05, "loss": 0.0776, "step": 73045 }, { "epoch": 1.7, "learning_rate": 1.7323695759310977e-05, "loss": 0.0972, "step": 73050 }, { "epoch": 1.7, "learning_rate": 1.7322815704864942e-05, "loss": 0.2812, "step": 73055 }, { "epoch": 1.7, "learning_rate": 1.7321935650418907e-05, "loss": 0.3606, "step": 73060 }, { "epoch": 1.7, "learning_rate": 1.732105559597287e-05, "loss": 0.3316, "step": 73065 }, { "epoch": 1.7, "learning_rate": 1.7320175541526837e-05, "loss": 0.1134, "step": 73070 }, { "epoch": 1.7, "learning_rate": 1.7319295487080802e-05, "loss": 0.0372, "step": 73075 }, { "epoch": 1.71, "learning_rate": 1.7318415432634767e-05, "loss": 0.0532, "step": 73080 }, { "epoch": 1.71, "learning_rate": 1.731753537818873e-05, "loss": 0.1238, "step": 73085 }, { "epoch": 1.71, "learning_rate": 1.7316655323742698e-05, "loss": 0.1038, "step": 73090 }, { "epoch": 1.71, "learning_rate": 1.7315775269296663e-05, "loss": 0.1175, "step": 73095 }, { "epoch": 1.71, "learning_rate": 1.7314895214850624e-05, "loss": 0.1567, "step": 73100 }, { "epoch": 1.71, "learning_rate": 1.731401516040459e-05, "loss": 0.0774, "step": 73105 }, { "epoch": 1.71, "learning_rate": 1.7313135105958558e-05, "loss": 0.3624, "step": 73110 }, { "epoch": 1.71, "learning_rate": 1.7312255051512523e-05, "loss": 0.3178, "step": 73115 }, { "epoch": 1.71, "learning_rate": 1.7311374997066484e-05, "loss": 0.0701, "step": 73120 }, { "epoch": 1.71, "learning_rate": 1.731049494262045e-05, "loss": 0.0946, "step": 73125 }, { "epoch": 1.71, "learning_rate": 1.7309614888174418e-05, "loss": 0.0564, "step": 73130 }, { "epoch": 1.71, "learning_rate": 1.730873483372838e-05, "loss": 0.0598, "step": 73135 }, { "epoch": 1.71, "learning_rate": 1.7307854779282344e-05, "loss": 0.1565, "step": 73140 }, { "epoch": 1.71, "learning_rate": 1.730697472483631e-05, "loss": 0.1712, "step": 73145 }, { "epoch": 1.71, "learning_rate": 1.7306094670390278e-05, "loss": 0.206, "step": 73150 }, { "epoch": 1.71, "learning_rate": 1.730521461594424e-05, "loss": 0.1744, "step": 73155 }, { "epoch": 1.71, "learning_rate": 1.7304334561498204e-05, "loss": 0.2255, "step": 73160 }, { "epoch": 1.71, "learning_rate": 1.730345450705217e-05, "loss": 0.2127, "step": 73165 }, { "epoch": 1.71, "learning_rate": 1.7302574452606134e-05, "loss": 0.067, "step": 73170 }, { "epoch": 1.71, "learning_rate": 1.73016943981601e-05, "loss": 0.0146, "step": 73175 }, { "epoch": 1.71, "learning_rate": 1.7300814343714065e-05, "loss": 0.1229, "step": 73180 }, { "epoch": 1.71, "learning_rate": 1.7299934289268033e-05, "loss": 0.0855, "step": 73185 }, { "epoch": 1.71, "learning_rate": 1.7299054234821995e-05, "loss": 0.2276, "step": 73190 }, { "epoch": 1.71, "learning_rate": 1.729817418037596e-05, "loss": 0.03, "step": 73195 }, { "epoch": 1.71, "learning_rate": 1.7297294125929925e-05, "loss": 0.3583, "step": 73200 }, { "epoch": 1.71, "learning_rate": 1.729641407148389e-05, "loss": 0.204, "step": 73205 }, { "epoch": 1.71, "learning_rate": 1.7295534017037855e-05, "loss": 0.2604, "step": 73210 }, { "epoch": 1.71, "learning_rate": 1.729465396259182e-05, "loss": 0.2477, "step": 73215 }, { "epoch": 1.71, "learning_rate": 1.7293773908145785e-05, "loss": 0.0356, "step": 73220 }, { "epoch": 1.71, "learning_rate": 1.729289385369975e-05, "loss": 0.0044, "step": 73225 }, { "epoch": 1.71, "learning_rate": 1.7292013799253715e-05, "loss": 0.062, "step": 73230 }, { "epoch": 1.71, "learning_rate": 1.729113374480768e-05, "loss": 0.1114, "step": 73235 }, { "epoch": 1.71, "learning_rate": 1.729025369036164e-05, "loss": 0.1165, "step": 73240 }, { "epoch": 1.71, "learning_rate": 1.728937363591561e-05, "loss": 0.3736, "step": 73245 }, { "epoch": 1.71, "learning_rate": 1.7288493581469575e-05, "loss": 0.3183, "step": 73250 }, { "epoch": 1.71, "learning_rate": 1.728761352702354e-05, "loss": 0.0854, "step": 73255 }, { "epoch": 1.71, "learning_rate": 1.72867334725775e-05, "loss": 0.3014, "step": 73260 }, { "epoch": 1.71, "learning_rate": 1.728585341813147e-05, "loss": 0.2394, "step": 73265 }, { "epoch": 1.71, "learning_rate": 1.7284973363685435e-05, "loss": 0.0262, "step": 73270 }, { "epoch": 1.71, "learning_rate": 1.7284093309239397e-05, "loss": 0.0344, "step": 73275 }, { "epoch": 1.71, "learning_rate": 1.728321325479336e-05, "loss": 0.1167, "step": 73280 }, { "epoch": 1.71, "learning_rate": 1.728233320034733e-05, "loss": 0.0277, "step": 73285 }, { "epoch": 1.71, "learning_rate": 1.7281453145901295e-05, "loss": 0.2069, "step": 73290 }, { "epoch": 1.71, "learning_rate": 1.7280573091455257e-05, "loss": 0.2107, "step": 73295 }, { "epoch": 1.71, "learning_rate": 1.727969303700922e-05, "loss": 0.0964, "step": 73300 }, { "epoch": 1.71, "learning_rate": 1.727881298256319e-05, "loss": 0.1775, "step": 73305 }, { "epoch": 1.71, "learning_rate": 1.727793292811715e-05, "loss": 0.2236, "step": 73310 }, { "epoch": 1.71, "learning_rate": 1.7277052873671117e-05, "loss": 0.3572, "step": 73315 }, { "epoch": 1.71, "learning_rate": 1.7276172819225085e-05, "loss": 0.0172, "step": 73320 }, { "epoch": 1.71, "learning_rate": 1.727529276477905e-05, "loss": 0.1587, "step": 73325 }, { "epoch": 1.71, "learning_rate": 1.7274412710333012e-05, "loss": 0.0404, "step": 73330 }, { "epoch": 1.71, "learning_rate": 1.7273532655886977e-05, "loss": 0.2052, "step": 73335 }, { "epoch": 1.71, "learning_rate": 1.7272652601440945e-05, "loss": 0.1459, "step": 73340 }, { "epoch": 1.71, "learning_rate": 1.727177254699491e-05, "loss": 0.0704, "step": 73345 }, { "epoch": 1.71, "learning_rate": 1.7270892492548872e-05, "loss": 0.1465, "step": 73350 }, { "epoch": 1.71, "learning_rate": 1.7270012438102837e-05, "loss": 0.2553, "step": 73355 }, { "epoch": 1.71, "learning_rate": 1.7269132383656805e-05, "loss": 0.2511, "step": 73360 }, { "epoch": 1.71, "learning_rate": 1.7268252329210767e-05, "loss": 0.3611, "step": 73365 }, { "epoch": 1.71, "learning_rate": 1.7267372274764732e-05, "loss": 0.0197, "step": 73370 }, { "epoch": 1.71, "learning_rate": 1.7266492220318697e-05, "loss": 0.0048, "step": 73375 }, { "epoch": 1.71, "learning_rate": 1.7265612165872665e-05, "loss": 0.0248, "step": 73380 }, { "epoch": 1.71, "learning_rate": 1.7264732111426627e-05, "loss": 0.1054, "step": 73385 }, { "epoch": 1.71, "learning_rate": 1.7263852056980592e-05, "loss": 0.0583, "step": 73390 }, { "epoch": 1.71, "learning_rate": 1.7262972002534557e-05, "loss": 0.309, "step": 73395 }, { "epoch": 1.71, "learning_rate": 1.7262091948088522e-05, "loss": 0.2149, "step": 73400 }, { "epoch": 1.71, "learning_rate": 1.7261211893642487e-05, "loss": 0.2479, "step": 73405 }, { "epoch": 1.71, "learning_rate": 1.7260331839196452e-05, "loss": 0.3594, "step": 73410 }, { "epoch": 1.71, "learning_rate": 1.7259451784750417e-05, "loss": 0.2578, "step": 73415 }, { "epoch": 1.71, "learning_rate": 1.7258571730304382e-05, "loss": 0.0439, "step": 73420 }, { "epoch": 1.71, "learning_rate": 1.7257691675858347e-05, "loss": 0.015, "step": 73425 }, { "epoch": 1.71, "learning_rate": 1.7256811621412312e-05, "loss": 0.0147, "step": 73430 }, { "epoch": 1.71, "learning_rate": 1.7255931566966274e-05, "loss": 0.0341, "step": 73435 }, { "epoch": 1.71, "learning_rate": 1.7255051512520242e-05, "loss": 0.1173, "step": 73440 }, { "epoch": 1.71, "learning_rate": 1.7254171458074207e-05, "loss": 0.0603, "step": 73445 }, { "epoch": 1.71, "learning_rate": 1.7253291403628172e-05, "loss": 0.074, "step": 73450 }, { "epoch": 1.71, "learning_rate": 1.7252411349182137e-05, "loss": 0.248, "step": 73455 }, { "epoch": 1.71, "learning_rate": 1.7251531294736102e-05, "loss": 0.4114, "step": 73460 }, { "epoch": 1.71, "learning_rate": 1.7250651240290067e-05, "loss": 0.5037, "step": 73465 }, { "epoch": 1.71, "learning_rate": 1.724977118584403e-05, "loss": 0.0038, "step": 73470 }, { "epoch": 1.71, "learning_rate": 1.7248891131397997e-05, "loss": 0.0554, "step": 73475 }, { "epoch": 1.71, "learning_rate": 1.7248011076951962e-05, "loss": 0.0813, "step": 73480 }, { "epoch": 1.71, "learning_rate": 1.7247131022505927e-05, "loss": 0.0675, "step": 73485 }, { "epoch": 1.71, "learning_rate": 1.724625096805989e-05, "loss": 0.1162, "step": 73490 }, { "epoch": 1.71, "learning_rate": 1.7245370913613857e-05, "loss": 0.132, "step": 73495 }, { "epoch": 1.71, "learning_rate": 1.7244490859167822e-05, "loss": 0.0969, "step": 73500 }, { "epoch": 1.71, "learning_rate": 1.7243610804721784e-05, "loss": 0.2109, "step": 73505 }, { "epoch": 1.72, "learning_rate": 1.724273075027575e-05, "loss": 0.5495, "step": 73510 }, { "epoch": 1.72, "learning_rate": 1.7241850695829717e-05, "loss": 0.3285, "step": 73515 }, { "epoch": 1.72, "learning_rate": 1.7240970641383683e-05, "loss": 0.1322, "step": 73520 }, { "epoch": 1.72, "learning_rate": 1.7240090586937644e-05, "loss": 0.0371, "step": 73525 }, { "epoch": 1.72, "learning_rate": 1.723921053249161e-05, "loss": 0.073, "step": 73530 }, { "epoch": 1.72, "learning_rate": 1.7238330478045578e-05, "loss": 0.0367, "step": 73535 }, { "epoch": 1.72, "learning_rate": 1.723745042359954e-05, "loss": 0.1855, "step": 73540 }, { "epoch": 1.72, "learning_rate": 1.7236570369153504e-05, "loss": 0.074, "step": 73545 }, { "epoch": 1.72, "learning_rate": 1.723569031470747e-05, "loss": 0.0552, "step": 73550 }, { "epoch": 1.72, "learning_rate": 1.7234810260261438e-05, "loss": 0.2634, "step": 73555 }, { "epoch": 1.72, "learning_rate": 1.72339302058154e-05, "loss": 0.3146, "step": 73560 }, { "epoch": 1.72, "learning_rate": 1.7233050151369364e-05, "loss": 0.216, "step": 73565 }, { "epoch": 1.72, "learning_rate": 1.723217009692333e-05, "loss": 0.0902, "step": 73570 }, { "epoch": 1.72, "learning_rate": 1.7231290042477294e-05, "loss": 0.0371, "step": 73575 }, { "epoch": 1.72, "learning_rate": 1.723040998803126e-05, "loss": 0.2187, "step": 73580 }, { "epoch": 1.72, "learning_rate": 1.7229529933585224e-05, "loss": 0.1366, "step": 73585 }, { "epoch": 1.72, "learning_rate": 1.722864987913919e-05, "loss": 0.0796, "step": 73590 }, { "epoch": 1.72, "learning_rate": 1.7227769824693154e-05, "loss": 0.0946, "step": 73595 }, { "epoch": 1.72, "learning_rate": 1.722688977024712e-05, "loss": 0.1114, "step": 73600 }, { "epoch": 1.72, "learning_rate": 1.7226009715801084e-05, "loss": 0.0991, "step": 73605 }, { "epoch": 1.72, "learning_rate": 1.722512966135505e-05, "loss": 0.2021, "step": 73610 }, { "epoch": 1.72, "learning_rate": 1.7224249606909015e-05, "loss": 0.2962, "step": 73615 }, { "epoch": 1.72, "learning_rate": 1.722336955246298e-05, "loss": 0.0063, "step": 73620 }, { "epoch": 1.72, "learning_rate": 1.7222489498016945e-05, "loss": 0.019, "step": 73625 }, { "epoch": 1.72, "learning_rate": 1.722160944357091e-05, "loss": 0.069, "step": 73630 }, { "epoch": 1.72, "learning_rate": 1.7220729389124875e-05, "loss": 0.1559, "step": 73635 }, { "epoch": 1.72, "learning_rate": 1.721984933467884e-05, "loss": 0.1221, "step": 73640 }, { "epoch": 1.72, "learning_rate": 1.7218969280232805e-05, "loss": 0.1637, "step": 73645 }, { "epoch": 1.72, "learning_rate": 1.721808922578677e-05, "loss": 0.1199, "step": 73650 }, { "epoch": 1.72, "learning_rate": 1.7217209171340735e-05, "loss": 0.2005, "step": 73655 }, { "epoch": 1.72, "learning_rate": 1.72163291168947e-05, "loss": 0.3866, "step": 73660 }, { "epoch": 1.72, "learning_rate": 1.721544906244866e-05, "loss": 0.3243, "step": 73665 }, { "epoch": 1.72, "learning_rate": 1.721456900800263e-05, "loss": 0.0428, "step": 73670 }, { "epoch": 1.72, "learning_rate": 1.7213688953556595e-05, "loss": 0.111, "step": 73675 }, { "epoch": 1.72, "learning_rate": 1.721280889911056e-05, "loss": 0.1591, "step": 73680 }, { "epoch": 1.72, "learning_rate": 1.721192884466452e-05, "loss": 0.0764, "step": 73685 }, { "epoch": 1.72, "learning_rate": 1.721104879021849e-05, "loss": 0.1694, "step": 73690 }, { "epoch": 1.72, "learning_rate": 1.7210168735772455e-05, "loss": 0.0917, "step": 73695 }, { "epoch": 1.72, "learning_rate": 1.7209288681326416e-05, "loss": 0.0505, "step": 73700 }, { "epoch": 1.72, "learning_rate": 1.720840862688038e-05, "loss": 0.2807, "step": 73705 }, { "epoch": 1.72, "learning_rate": 1.720752857243435e-05, "loss": 0.325, "step": 73710 }, { "epoch": 1.72, "learning_rate": 1.7206648517988315e-05, "loss": 0.1776, "step": 73715 }, { "epoch": 1.72, "learning_rate": 1.7205768463542277e-05, "loss": 0.082, "step": 73720 }, { "epoch": 1.72, "learning_rate": 1.7204888409096245e-05, "loss": 0.0533, "step": 73725 }, { "epoch": 1.72, "learning_rate": 1.720400835465021e-05, "loss": 0.0657, "step": 73730 }, { "epoch": 1.72, "learning_rate": 1.720312830020417e-05, "loss": 0.0317, "step": 73735 }, { "epoch": 1.72, "learning_rate": 1.7202248245758137e-05, "loss": 0.0503, "step": 73740 }, { "epoch": 1.72, "learning_rate": 1.7201368191312105e-05, "loss": 0.1783, "step": 73745 }, { "epoch": 1.72, "learning_rate": 1.720048813686607e-05, "loss": 0.0512, "step": 73750 }, { "epoch": 1.72, "learning_rate": 1.7199608082420032e-05, "loss": 0.1365, "step": 73755 }, { "epoch": 1.72, "learning_rate": 1.7198728027973997e-05, "loss": 0.4207, "step": 73760 }, { "epoch": 1.72, "learning_rate": 1.7197847973527965e-05, "loss": 0.1161, "step": 73765 }, { "epoch": 1.72, "learning_rate": 1.7196967919081927e-05, "loss": 0.0253, "step": 73770 }, { "epoch": 1.72, "learning_rate": 1.7196087864635892e-05, "loss": 0.0667, "step": 73775 }, { "epoch": 1.72, "learning_rate": 1.7195207810189857e-05, "loss": 0.0666, "step": 73780 }, { "epoch": 1.72, "learning_rate": 1.7194327755743825e-05, "loss": 0.0289, "step": 73785 }, { "epoch": 1.72, "learning_rate": 1.7193447701297787e-05, "loss": 0.1621, "step": 73790 }, { "epoch": 1.72, "learning_rate": 1.7192567646851752e-05, "loss": 0.2915, "step": 73795 }, { "epoch": 1.72, "learning_rate": 1.7191687592405717e-05, "loss": 0.1975, "step": 73800 }, { "epoch": 1.72, "learning_rate": 1.7190807537959682e-05, "loss": 0.2204, "step": 73805 }, { "epoch": 1.72, "learning_rate": 1.7189927483513647e-05, "loss": 0.3344, "step": 73810 }, { "epoch": 1.72, "learning_rate": 1.7189047429067612e-05, "loss": 0.116, "step": 73815 }, { "epoch": 1.72, "learning_rate": 1.7188167374621577e-05, "loss": 0.0051, "step": 73820 }, { "epoch": 1.72, "learning_rate": 1.7187287320175542e-05, "loss": 0.0498, "step": 73825 }, { "epoch": 1.72, "learning_rate": 1.7186407265729507e-05, "loss": 0.0113, "step": 73830 }, { "epoch": 1.72, "learning_rate": 1.7185527211283472e-05, "loss": 0.0314, "step": 73835 }, { "epoch": 1.72, "learning_rate": 1.7184647156837434e-05, "loss": 0.0831, "step": 73840 }, { "epoch": 1.72, "learning_rate": 1.7183767102391402e-05, "loss": 0.0273, "step": 73845 }, { "epoch": 1.72, "learning_rate": 1.7182887047945367e-05, "loss": 0.134, "step": 73850 }, { "epoch": 1.72, "learning_rate": 1.7182006993499332e-05, "loss": 0.05, "step": 73855 }, { "epoch": 1.72, "learning_rate": 1.7181126939053297e-05, "loss": 0.3533, "step": 73860 }, { "epoch": 1.72, "learning_rate": 1.7180246884607262e-05, "loss": 0.2009, "step": 73865 }, { "epoch": 1.72, "learning_rate": 1.7179366830161227e-05, "loss": 0.0206, "step": 73870 }, { "epoch": 1.72, "learning_rate": 1.717848677571519e-05, "loss": 0.0797, "step": 73875 }, { "epoch": 1.72, "learning_rate": 1.7177606721269157e-05, "loss": 0.009, "step": 73880 }, { "epoch": 1.72, "learning_rate": 1.7176726666823122e-05, "loss": 0.1297, "step": 73885 }, { "epoch": 1.72, "learning_rate": 1.7175846612377087e-05, "loss": 0.1726, "step": 73890 }, { "epoch": 1.72, "learning_rate": 1.717496655793105e-05, "loss": 0.1258, "step": 73895 }, { "epoch": 1.72, "learning_rate": 1.7174086503485017e-05, "loss": 0.1116, "step": 73900 }, { "epoch": 1.72, "learning_rate": 1.7173206449038982e-05, "loss": 0.1453, "step": 73905 }, { "epoch": 1.72, "learning_rate": 1.7172326394592944e-05, "loss": 0.2393, "step": 73910 }, { "epoch": 1.72, "learning_rate": 1.717144634014691e-05, "loss": 0.1698, "step": 73915 }, { "epoch": 1.72, "learning_rate": 1.7170566285700877e-05, "loss": 0.0393, "step": 73920 }, { "epoch": 1.72, "learning_rate": 1.7169686231254842e-05, "loss": 0.0053, "step": 73925 }, { "epoch": 1.72, "learning_rate": 1.7168806176808804e-05, "loss": 0.0202, "step": 73930 }, { "epoch": 1.72, "learning_rate": 1.716792612236277e-05, "loss": 0.0187, "step": 73935 }, { "epoch": 1.73, "learning_rate": 1.7167046067916737e-05, "loss": 0.1521, "step": 73940 }, { "epoch": 1.73, "learning_rate": 1.7166166013470702e-05, "loss": 0.0957, "step": 73945 }, { "epoch": 1.73, "learning_rate": 1.7165285959024664e-05, "loss": 0.0939, "step": 73950 }, { "epoch": 1.73, "learning_rate": 1.716440590457863e-05, "loss": 0.2528, "step": 73955 }, { "epoch": 1.73, "learning_rate": 1.7163525850132598e-05, "loss": 0.243, "step": 73960 }, { "epoch": 1.73, "learning_rate": 1.716264579568656e-05, "loss": 0.2822, "step": 73965 }, { "epoch": 1.73, "learning_rate": 1.7161765741240524e-05, "loss": 0.039, "step": 73970 }, { "epoch": 1.73, "learning_rate": 1.716088568679449e-05, "loss": 0.0125, "step": 73975 }, { "epoch": 1.73, "learning_rate": 1.7160005632348458e-05, "loss": 0.1283, "step": 73980 }, { "epoch": 1.73, "learning_rate": 1.715912557790242e-05, "loss": 0.1066, "step": 73985 }, { "epoch": 1.73, "learning_rate": 1.7158245523456384e-05, "loss": 0.0386, "step": 73990 }, { "epoch": 1.73, "learning_rate": 1.715736546901035e-05, "loss": 0.1268, "step": 73995 }, { "epoch": 1.73, "learning_rate": 1.7156485414564314e-05, "loss": 0.1739, "step": 74000 }, { "epoch": 1.73, "learning_rate": 1.715560536011828e-05, "loss": 0.248, "step": 74005 }, { "epoch": 1.73, "learning_rate": 1.7154725305672244e-05, "loss": 0.1775, "step": 74010 }, { "epoch": 1.73, "learning_rate": 1.7153845251226213e-05, "loss": 0.2628, "step": 74015 }, { "epoch": 1.73, "learning_rate": 1.7152965196780174e-05, "loss": 0.0206, "step": 74020 }, { "epoch": 1.73, "learning_rate": 1.715208514233414e-05, "loss": 0.0568, "step": 74025 }, { "epoch": 1.73, "learning_rate": 1.7151205087888104e-05, "loss": 0.1161, "step": 74030 }, { "epoch": 1.73, "learning_rate": 1.715032503344207e-05, "loss": 0.0824, "step": 74035 }, { "epoch": 1.73, "learning_rate": 1.7149444978996034e-05, "loss": 0.0345, "step": 74040 }, { "epoch": 1.73, "learning_rate": 1.714856492455e-05, "loss": 0.1209, "step": 74045 }, { "epoch": 1.73, "learning_rate": 1.7147684870103965e-05, "loss": 0.2666, "step": 74050 }, { "epoch": 1.73, "learning_rate": 1.714680481565793e-05, "loss": 0.1575, "step": 74055 }, { "epoch": 1.73, "learning_rate": 1.7145924761211895e-05, "loss": 0.2954, "step": 74060 }, { "epoch": 1.73, "learning_rate": 1.714504470676586e-05, "loss": 0.2207, "step": 74065 }, { "epoch": 1.73, "learning_rate": 1.714416465231982e-05, "loss": 0.0122, "step": 74070 }, { "epoch": 1.73, "learning_rate": 1.714328459787379e-05, "loss": 0.0246, "step": 74075 }, { "epoch": 1.73, "learning_rate": 1.7142404543427755e-05, "loss": 0.0834, "step": 74080 }, { "epoch": 1.73, "learning_rate": 1.714152448898172e-05, "loss": 0.1014, "step": 74085 }, { "epoch": 1.73, "learning_rate": 1.714064443453568e-05, "loss": 0.1977, "step": 74090 }, { "epoch": 1.73, "learning_rate": 1.713976438008965e-05, "loss": 0.0323, "step": 74095 }, { "epoch": 1.73, "learning_rate": 1.7138884325643615e-05, "loss": 0.3452, "step": 74100 }, { "epoch": 1.73, "learning_rate": 1.7138004271197576e-05, "loss": 0.0339, "step": 74105 }, { "epoch": 1.73, "learning_rate": 1.713712421675154e-05, "loss": 0.2574, "step": 74110 }, { "epoch": 1.73, "learning_rate": 1.713624416230551e-05, "loss": 0.195, "step": 74115 }, { "epoch": 1.73, "learning_rate": 1.7135364107859475e-05, "loss": 0.0234, "step": 74120 }, { "epoch": 1.73, "learning_rate": 1.7134484053413436e-05, "loss": 0.013, "step": 74125 }, { "epoch": 1.73, "learning_rate": 1.71336039989674e-05, "loss": 0.0959, "step": 74130 }, { "epoch": 1.73, "learning_rate": 1.713272394452137e-05, "loss": 0.0205, "step": 74135 }, { "epoch": 1.73, "learning_rate": 1.713184389007533e-05, "loss": 0.045, "step": 74140 }, { "epoch": 1.73, "learning_rate": 1.7130963835629297e-05, "loss": 0.3765, "step": 74145 }, { "epoch": 1.73, "learning_rate": 1.7130083781183265e-05, "loss": 0.1243, "step": 74150 }, { "epoch": 1.73, "learning_rate": 1.712920372673723e-05, "loss": 0.1443, "step": 74155 }, { "epoch": 1.73, "learning_rate": 1.712832367229119e-05, "loss": 0.3267, "step": 74160 }, { "epoch": 1.73, "learning_rate": 1.7127443617845157e-05, "loss": 0.3853, "step": 74165 }, { "epoch": 1.73, "learning_rate": 1.7126563563399125e-05, "loss": 0.0445, "step": 74170 }, { "epoch": 1.73, "learning_rate": 1.7125683508953087e-05, "loss": 0.0138, "step": 74175 }, { "epoch": 1.73, "learning_rate": 1.712480345450705e-05, "loss": 0.0579, "step": 74180 }, { "epoch": 1.73, "learning_rate": 1.7123923400061017e-05, "loss": 0.0564, "step": 74185 }, { "epoch": 1.73, "learning_rate": 1.7123043345614985e-05, "loss": 0.0534, "step": 74190 }, { "epoch": 1.73, "learning_rate": 1.7122163291168947e-05, "loss": 0.107, "step": 74195 }, { "epoch": 1.73, "learning_rate": 1.7121283236722912e-05, "loss": 0.1816, "step": 74200 }, { "epoch": 1.73, "learning_rate": 1.7120403182276877e-05, "loss": 0.0533, "step": 74205 }, { "epoch": 1.73, "learning_rate": 1.7119523127830842e-05, "loss": 0.253, "step": 74210 }, { "epoch": 1.73, "learning_rate": 1.7118643073384807e-05, "loss": 0.2357, "step": 74215 }, { "epoch": 1.73, "learning_rate": 1.7117763018938772e-05, "loss": 0.0394, "step": 74220 }, { "epoch": 1.73, "learning_rate": 1.7116882964492737e-05, "loss": 0.0235, "step": 74225 }, { "epoch": 1.73, "learning_rate": 1.7116002910046702e-05, "loss": 0.0423, "step": 74230 }, { "epoch": 1.73, "learning_rate": 1.7115122855600667e-05, "loss": 0.0867, "step": 74235 }, { "epoch": 1.73, "learning_rate": 1.7114242801154632e-05, "loss": 0.0752, "step": 74240 }, { "epoch": 1.73, "learning_rate": 1.7113362746708597e-05, "loss": 0.1842, "step": 74245 }, { "epoch": 1.73, "learning_rate": 1.7112482692262562e-05, "loss": 0.1881, "step": 74250 }, { "epoch": 1.73, "learning_rate": 1.7111778648705733e-05, "loss": 0.2248, "step": 74255 }, { "epoch": 1.73, "learning_rate": 1.71108985942597e-05, "loss": 0.2357, "step": 74260 }, { "epoch": 1.73, "learning_rate": 1.7110018539813663e-05, "loss": 0.211, "step": 74265 }, { "epoch": 1.73, "learning_rate": 1.710913848536763e-05, "loss": 0.0163, "step": 74270 }, { "epoch": 1.73, "learning_rate": 1.7108258430921593e-05, "loss": 0.0301, "step": 74275 }, { "epoch": 1.73, "learning_rate": 1.710737837647556e-05, "loss": 0.1304, "step": 74280 }, { "epoch": 1.73, "learning_rate": 1.7106498322029523e-05, "loss": 0.0806, "step": 74285 }, { "epoch": 1.73, "learning_rate": 1.710561826758349e-05, "loss": 0.1435, "step": 74290 }, { "epoch": 1.73, "learning_rate": 1.7104738213137453e-05, "loss": 0.1259, "step": 74295 }, { "epoch": 1.73, "learning_rate": 1.7103858158691415e-05, "loss": 0.1081, "step": 74300 }, { "epoch": 1.73, "learning_rate": 1.7102978104245384e-05, "loss": 0.3042, "step": 74305 }, { "epoch": 1.73, "learning_rate": 1.710209804979935e-05, "loss": 0.2495, "step": 74310 }, { "epoch": 1.73, "learning_rate": 1.7101217995353314e-05, "loss": 0.2483, "step": 74315 }, { "epoch": 1.73, "learning_rate": 1.7100337940907275e-05, "loss": 0.0023, "step": 74320 }, { "epoch": 1.73, "learning_rate": 1.7099457886461244e-05, "loss": 0.0171, "step": 74325 }, { "epoch": 1.73, "learning_rate": 1.709857783201521e-05, "loss": 0.0666, "step": 74330 }, { "epoch": 1.73, "learning_rate": 1.709769777756917e-05, "loss": 0.1104, "step": 74335 }, { "epoch": 1.73, "learning_rate": 1.709681772312314e-05, "loss": 0.1823, "step": 74340 }, { "epoch": 1.73, "learning_rate": 1.7095937668677104e-05, "loss": 0.12, "step": 74345 }, { "epoch": 1.73, "learning_rate": 1.709505761423107e-05, "loss": 0.2174, "step": 74350 }, { "epoch": 1.73, "learning_rate": 1.709417755978503e-05, "loss": 0.3021, "step": 74355 }, { "epoch": 1.73, "learning_rate": 1.7093297505339e-05, "loss": 0.3683, "step": 74360 }, { "epoch": 1.74, "learning_rate": 1.7092417450892964e-05, "loss": 0.2723, "step": 74365 }, { "epoch": 1.74, "learning_rate": 1.7091537396446925e-05, "loss": 0.0798, "step": 74370 }, { "epoch": 1.74, "learning_rate": 1.709065734200089e-05, "loss": 0.0548, "step": 74375 }, { "epoch": 1.74, "learning_rate": 1.708977728755486e-05, "loss": 0.1193, "step": 74380 }, { "epoch": 1.74, "learning_rate": 1.7088897233108824e-05, "loss": 0.1107, "step": 74385 }, { "epoch": 1.74, "learning_rate": 1.7088017178662785e-05, "loss": 0.0647, "step": 74390 }, { "epoch": 1.74, "learning_rate": 1.708713712421675e-05, "loss": 0.1707, "step": 74395 }, { "epoch": 1.74, "learning_rate": 1.708625706977072e-05, "loss": 0.1266, "step": 74400 }, { "epoch": 1.74, "learning_rate": 1.708537701532468e-05, "loss": 0.1777, "step": 74405 }, { "epoch": 1.74, "learning_rate": 1.7084496960878646e-05, "loss": 0.3395, "step": 74410 }, { "epoch": 1.74, "learning_rate": 1.708361690643261e-05, "loss": 0.3769, "step": 74415 }, { "epoch": 1.74, "learning_rate": 1.708273685198658e-05, "loss": 0.0057, "step": 74420 }, { "epoch": 1.74, "learning_rate": 1.708185679754054e-05, "loss": 0.0774, "step": 74425 }, { "epoch": 1.74, "learning_rate": 1.7080976743094506e-05, "loss": 0.1124, "step": 74430 }, { "epoch": 1.74, "learning_rate": 1.708009668864847e-05, "loss": 0.0244, "step": 74435 }, { "epoch": 1.74, "learning_rate": 1.707921663420244e-05, "loss": 0.0507, "step": 74440 }, { "epoch": 1.74, "learning_rate": 1.70783365797564e-05, "loss": 0.1267, "step": 74445 }, { "epoch": 1.74, "learning_rate": 1.7077456525310366e-05, "loss": 0.1813, "step": 74450 }, { "epoch": 1.74, "learning_rate": 1.707657647086433e-05, "loss": 0.2311, "step": 74455 }, { "epoch": 1.74, "learning_rate": 1.7075696416418296e-05, "loss": 0.2326, "step": 74460 }, { "epoch": 1.74, "learning_rate": 1.707481636197226e-05, "loss": 0.4089, "step": 74465 }, { "epoch": 1.74, "learning_rate": 1.7073936307526226e-05, "loss": 0.061, "step": 74470 }, { "epoch": 1.74, "learning_rate": 1.7073056253080194e-05, "loss": 0.0644, "step": 74475 }, { "epoch": 1.74, "learning_rate": 1.7072176198634156e-05, "loss": 0.0939, "step": 74480 }, { "epoch": 1.74, "learning_rate": 1.707129614418812e-05, "loss": 0.0668, "step": 74485 }, { "epoch": 1.74, "learning_rate": 1.7070416089742086e-05, "loss": 0.1278, "step": 74490 }, { "epoch": 1.74, "learning_rate": 1.706953603529605e-05, "loss": 0.141, "step": 74495 }, { "epoch": 1.74, "learning_rate": 1.7068655980850016e-05, "loss": 0.1796, "step": 74500 }, { "epoch": 1.74, "learning_rate": 1.706777592640398e-05, "loss": 0.166, "step": 74505 }, { "epoch": 1.74, "learning_rate": 1.7066895871957946e-05, "loss": 0.3014, "step": 74510 }, { "epoch": 1.74, "learning_rate": 1.706601581751191e-05, "loss": 0.2437, "step": 74515 }, { "epoch": 1.74, "learning_rate": 1.7065135763065876e-05, "loss": 0.0236, "step": 74520 }, { "epoch": 1.74, "learning_rate": 1.706425570861984e-05, "loss": 0.0421, "step": 74525 }, { "epoch": 1.74, "learning_rate": 1.7063375654173803e-05, "loss": 0.0937, "step": 74530 }, { "epoch": 1.74, "learning_rate": 1.706249559972777e-05, "loss": 0.1548, "step": 74535 }, { "epoch": 1.74, "learning_rate": 1.7061615545281736e-05, "loss": 0.0543, "step": 74540 }, { "epoch": 1.74, "learning_rate": 1.70607354908357e-05, "loss": 0.0861, "step": 74545 }, { "epoch": 1.74, "learning_rate": 1.7059855436389663e-05, "loss": 0.1406, "step": 74550 }, { "epoch": 1.74, "learning_rate": 1.705897538194363e-05, "loss": 0.1357, "step": 74555 }, { "epoch": 1.74, "learning_rate": 1.7058095327497596e-05, "loss": 0.154, "step": 74560 }, { "epoch": 1.74, "learning_rate": 1.7057215273051558e-05, "loss": 0.2927, "step": 74565 }, { "epoch": 1.74, "learning_rate": 1.7056335218605523e-05, "loss": 0.0138, "step": 74570 }, { "epoch": 1.74, "learning_rate": 1.705545516415949e-05, "loss": 0.011, "step": 74575 }, { "epoch": 1.74, "learning_rate": 1.7054575109713456e-05, "loss": 0.0595, "step": 74580 }, { "epoch": 1.74, "learning_rate": 1.7053695055267418e-05, "loss": 0.1403, "step": 74585 }, { "epoch": 1.74, "learning_rate": 1.7052815000821383e-05, "loss": 0.0574, "step": 74590 }, { "epoch": 1.74, "learning_rate": 1.705193494637535e-05, "loss": 0.0622, "step": 74595 }, { "epoch": 1.74, "learning_rate": 1.7051054891929313e-05, "loss": 0.1889, "step": 74600 }, { "epoch": 1.74, "learning_rate": 1.7050174837483278e-05, "loss": 0.262, "step": 74605 }, { "epoch": 1.74, "learning_rate": 1.7049294783037246e-05, "loss": 0.1855, "step": 74610 }, { "epoch": 1.74, "learning_rate": 1.704841472859121e-05, "loss": 0.2753, "step": 74615 }, { "epoch": 1.74, "learning_rate": 1.7047534674145173e-05, "loss": 0.0479, "step": 74620 }, { "epoch": 1.74, "learning_rate": 1.7046654619699138e-05, "loss": 0.0085, "step": 74625 }, { "epoch": 1.74, "learning_rate": 1.7045774565253106e-05, "loss": 0.0409, "step": 74630 }, { "epoch": 1.74, "learning_rate": 1.7044894510807068e-05, "loss": 0.0667, "step": 74635 }, { "epoch": 1.74, "learning_rate": 1.7044014456361033e-05, "loss": 0.1502, "step": 74640 }, { "epoch": 1.74, "learning_rate": 1.7043134401914998e-05, "loss": 0.104, "step": 74645 }, { "epoch": 1.74, "learning_rate": 1.7042254347468967e-05, "loss": 0.1418, "step": 74650 }, { "epoch": 1.74, "learning_rate": 1.7041374293022928e-05, "loss": 0.1123, "step": 74655 }, { "epoch": 1.74, "learning_rate": 1.7040494238576893e-05, "loss": 0.3105, "step": 74660 }, { "epoch": 1.74, "learning_rate": 1.7039614184130858e-05, "loss": 0.2199, "step": 74665 }, { "epoch": 1.74, "learning_rate": 1.7038734129684823e-05, "loss": 0.0171, "step": 74670 }, { "epoch": 1.74, "learning_rate": 1.7037854075238788e-05, "loss": 0.0646, "step": 74675 }, { "epoch": 1.74, "learning_rate": 1.7036974020792753e-05, "loss": 0.066, "step": 74680 }, { "epoch": 1.74, "learning_rate": 1.7036093966346718e-05, "loss": 0.0093, "step": 74685 }, { "epoch": 1.74, "learning_rate": 1.7035213911900683e-05, "loss": 0.0627, "step": 74690 }, { "epoch": 1.74, "learning_rate": 1.703433385745465e-05, "loss": 0.0662, "step": 74695 }, { "epoch": 1.74, "learning_rate": 1.7033453803008613e-05, "loss": 0.0463, "step": 74700 }, { "epoch": 1.74, "learning_rate": 1.7032573748562575e-05, "loss": 0.1293, "step": 74705 }, { "epoch": 1.74, "learning_rate": 1.7031693694116543e-05, "loss": 0.1845, "step": 74710 }, { "epoch": 1.74, "learning_rate": 1.703081363967051e-05, "loss": 0.3053, "step": 74715 }, { "epoch": 1.74, "learning_rate": 1.7029933585224473e-05, "loss": 0.0046, "step": 74720 }, { "epoch": 1.74, "learning_rate": 1.7029053530778435e-05, "loss": 0.013, "step": 74725 }, { "epoch": 1.74, "learning_rate": 1.7028173476332403e-05, "loss": 0.0336, "step": 74730 }, { "epoch": 1.74, "learning_rate": 1.702729342188637e-05, "loss": 0.0754, "step": 74735 }, { "epoch": 1.74, "learning_rate": 1.7026413367440334e-05, "loss": 0.1413, "step": 74740 }, { "epoch": 1.74, "learning_rate": 1.70255333129943e-05, "loss": 0.1694, "step": 74745 }, { "epoch": 1.74, "learning_rate": 1.7024653258548264e-05, "loss": 0.2382, "step": 74750 }, { "epoch": 1.74, "learning_rate": 1.702377320410223e-05, "loss": 0.1384, "step": 74755 }, { "epoch": 1.74, "learning_rate": 1.702289314965619e-05, "loss": 0.2374, "step": 74760 }, { "epoch": 1.74, "learning_rate": 1.702201309521016e-05, "loss": 0.2682, "step": 74765 }, { "epoch": 1.74, "learning_rate": 1.7021133040764124e-05, "loss": 0.0262, "step": 74770 }, { "epoch": 1.74, "learning_rate": 1.702025298631809e-05, "loss": 0.0384, "step": 74775 }, { "epoch": 1.74, "learning_rate": 1.701937293187205e-05, "loss": 0.1327, "step": 74780 }, { "epoch": 1.74, "learning_rate": 1.701849287742602e-05, "loss": 0.198, "step": 74785 }, { "epoch": 1.74, "learning_rate": 1.7017612822979984e-05, "loss": 0.2797, "step": 74790 }, { "epoch": 1.75, "learning_rate": 1.7016732768533945e-05, "loss": 0.225, "step": 74795 }, { "epoch": 1.75, "learning_rate": 1.701585271408791e-05, "loss": 0.1823, "step": 74800 }, { "epoch": 1.75, "learning_rate": 1.701497265964188e-05, "loss": 0.1692, "step": 74805 }, { "epoch": 1.75, "learning_rate": 1.7014092605195844e-05, "loss": 0.0669, "step": 74810 }, { "epoch": 1.75, "learning_rate": 1.7013212550749805e-05, "loss": 0.1628, "step": 74815 }, { "epoch": 1.75, "learning_rate": 1.701233249630377e-05, "loss": 0.1315, "step": 74820 }, { "epoch": 1.75, "learning_rate": 1.701145244185774e-05, "loss": 0.0887, "step": 74825 }, { "epoch": 1.75, "learning_rate": 1.70105723874117e-05, "loss": 0.0973, "step": 74830 }, { "epoch": 1.75, "learning_rate": 1.7009692332965666e-05, "loss": 0.1121, "step": 74835 }, { "epoch": 1.75, "learning_rate": 1.700881227851963e-05, "loss": 0.0813, "step": 74840 }, { "epoch": 1.75, "learning_rate": 1.70079322240736e-05, "loss": 0.113, "step": 74845 }, { "epoch": 1.75, "learning_rate": 1.700705216962756e-05, "loss": 0.0521, "step": 74850 }, { "epoch": 1.75, "learning_rate": 1.7006172115181526e-05, "loss": 0.2241, "step": 74855 }, { "epoch": 1.75, "learning_rate": 1.700529206073549e-05, "loss": 0.1954, "step": 74860 }, { "epoch": 1.75, "learning_rate": 1.7004412006289456e-05, "loss": 0.1938, "step": 74865 }, { "epoch": 1.75, "learning_rate": 1.700353195184342e-05, "loss": 0.0272, "step": 74870 }, { "epoch": 1.75, "learning_rate": 1.7002651897397386e-05, "loss": 0.0304, "step": 74875 }, { "epoch": 1.75, "learning_rate": 1.7001771842951354e-05, "loss": 0.0168, "step": 74880 }, { "epoch": 1.75, "learning_rate": 1.7000891788505316e-05, "loss": 0.1048, "step": 74885 }, { "epoch": 1.75, "learning_rate": 1.700001173405928e-05, "loss": 0.1093, "step": 74890 }, { "epoch": 1.75, "learning_rate": 1.6999131679613246e-05, "loss": 0.0986, "step": 74895 }, { "epoch": 1.75, "learning_rate": 1.699825162516721e-05, "loss": 0.2483, "step": 74900 }, { "epoch": 1.75, "learning_rate": 1.6997371570721176e-05, "loss": 0.2668, "step": 74905 }, { "epoch": 1.75, "learning_rate": 1.699649151627514e-05, "loss": 0.3933, "step": 74910 }, { "epoch": 1.75, "learning_rate": 1.6995611461829106e-05, "loss": 0.2766, "step": 74915 }, { "epoch": 1.75, "learning_rate": 1.699473140738307e-05, "loss": 0.0192, "step": 74920 }, { "epoch": 1.75, "learning_rate": 1.6993851352937036e-05, "loss": 0.1273, "step": 74925 }, { "epoch": 1.75, "learning_rate": 1.6992971298491e-05, "loss": 0.0605, "step": 74930 }, { "epoch": 1.75, "learning_rate": 1.6992091244044963e-05, "loss": 0.0459, "step": 74935 }, { "epoch": 1.75, "learning_rate": 1.699121118959893e-05, "loss": 0.1356, "step": 74940 }, { "epoch": 1.75, "learning_rate": 1.6990331135152896e-05, "loss": 0.162, "step": 74945 }, { "epoch": 1.75, "learning_rate": 1.698945108070686e-05, "loss": 0.083, "step": 74950 }, { "epoch": 1.75, "learning_rate": 1.6988571026260823e-05, "loss": 0.1864, "step": 74955 }, { "epoch": 1.75, "learning_rate": 1.698769097181479e-05, "loss": 0.4208, "step": 74960 }, { "epoch": 1.75, "learning_rate": 1.6986810917368756e-05, "loss": 0.445, "step": 74965 }, { "epoch": 1.75, "learning_rate": 1.6985930862922718e-05, "loss": 0.0347, "step": 74970 }, { "epoch": 1.75, "learning_rate": 1.6985050808476683e-05, "loss": 0.0745, "step": 74975 }, { "epoch": 1.75, "learning_rate": 1.698417075403065e-05, "loss": 0.0774, "step": 74980 }, { "epoch": 1.75, "learning_rate": 1.6983290699584616e-05, "loss": 0.0572, "step": 74985 }, { "epoch": 1.75, "learning_rate": 1.6982410645138578e-05, "loss": 0.03, "step": 74990 }, { "epoch": 1.75, "learning_rate": 1.6981530590692543e-05, "loss": 0.1867, "step": 74995 }, { "epoch": 1.75, "learning_rate": 1.698065053624651e-05, "loss": 0.1361, "step": 75000 }, { "epoch": 1.75, "learning_rate": 1.6979770481800473e-05, "loss": 0.322, "step": 75005 }, { "epoch": 1.75, "learning_rate": 1.6978890427354438e-05, "loss": 0.2756, "step": 75010 }, { "epoch": 1.75, "learning_rate": 1.6978010372908406e-05, "loss": 0.2584, "step": 75015 }, { "epoch": 1.75, "learning_rate": 1.697713031846237e-05, "loss": 0.0136, "step": 75020 }, { "epoch": 1.75, "learning_rate": 1.6976250264016333e-05, "loss": 0.1062, "step": 75025 }, { "epoch": 1.75, "learning_rate": 1.6975370209570298e-05, "loss": 0.0868, "step": 75030 }, { "epoch": 1.75, "learning_rate": 1.6974490155124266e-05, "loss": 0.1445, "step": 75035 }, { "epoch": 1.75, "learning_rate": 1.697361010067823e-05, "loss": 0.0827, "step": 75040 }, { "epoch": 1.75, "learning_rate": 1.6972730046232193e-05, "loss": 0.0856, "step": 75045 }, { "epoch": 1.75, "learning_rate": 1.6971849991786158e-05, "loss": 0.1778, "step": 75050 }, { "epoch": 1.75, "learning_rate": 1.6970969937340126e-05, "loss": 0.0766, "step": 75055 }, { "epoch": 1.75, "learning_rate": 1.6970089882894088e-05, "loss": 0.2323, "step": 75060 }, { "epoch": 1.75, "learning_rate": 1.6969209828448053e-05, "loss": 0.5269, "step": 75065 }, { "epoch": 1.75, "learning_rate": 1.6968329774002018e-05, "loss": 0.0112, "step": 75070 }, { "epoch": 1.75, "learning_rate": 1.6967449719555986e-05, "loss": 0.0481, "step": 75075 }, { "epoch": 1.75, "learning_rate": 1.6966569665109948e-05, "loss": 0.0656, "step": 75080 }, { "epoch": 1.75, "learning_rate": 1.6965689610663913e-05, "loss": 0.1267, "step": 75085 }, { "epoch": 1.75, "learning_rate": 1.6964809556217878e-05, "loss": 0.1824, "step": 75090 }, { "epoch": 1.75, "learning_rate": 1.6963929501771843e-05, "loss": 0.1091, "step": 75095 }, { "epoch": 1.75, "learning_rate": 1.6963049447325808e-05, "loss": 0.0788, "step": 75100 }, { "epoch": 1.75, "learning_rate": 1.6962169392879773e-05, "loss": 0.2346, "step": 75105 }, { "epoch": 1.75, "learning_rate": 1.6961289338433738e-05, "loss": 0.2039, "step": 75110 }, { "epoch": 1.75, "learning_rate": 1.6960409283987703e-05, "loss": 0.1332, "step": 75115 }, { "epoch": 1.75, "learning_rate": 1.6959529229541668e-05, "loss": 0.0052, "step": 75120 }, { "epoch": 1.75, "learning_rate": 1.6958649175095633e-05, "loss": 0.0148, "step": 75125 }, { "epoch": 1.75, "learning_rate": 1.6957769120649595e-05, "loss": 0.038, "step": 75130 }, { "epoch": 1.75, "learning_rate": 1.6956889066203563e-05, "loss": 0.0334, "step": 75135 }, { "epoch": 1.75, "learning_rate": 1.695600901175753e-05, "loss": 0.0674, "step": 75140 }, { "epoch": 1.75, "learning_rate": 1.6955128957311493e-05, "loss": 0.1096, "step": 75145 }, { "epoch": 1.75, "learning_rate": 1.695424890286546e-05, "loss": 0.0442, "step": 75150 }, { "epoch": 1.75, "learning_rate": 1.6953368848419423e-05, "loss": 0.2385, "step": 75155 }, { "epoch": 1.75, "learning_rate": 1.695248879397339e-05, "loss": 0.2952, "step": 75160 }, { "epoch": 1.75, "learning_rate": 1.695160873952735e-05, "loss": 0.1513, "step": 75165 }, { "epoch": 1.75, "learning_rate": 1.695072868508132e-05, "loss": 0.0154, "step": 75170 }, { "epoch": 1.75, "learning_rate": 1.6949848630635284e-05, "loss": 0.0057, "step": 75175 }, { "epoch": 1.75, "learning_rate": 1.694896857618925e-05, "loss": 0.0301, "step": 75180 }, { "epoch": 1.75, "learning_rate": 1.694808852174321e-05, "loss": 0.1306, "step": 75185 }, { "epoch": 1.75, "learning_rate": 1.694720846729718e-05, "loss": 0.035, "step": 75190 }, { "epoch": 1.75, "learning_rate": 1.6946328412851144e-05, "loss": 0.1087, "step": 75195 }, { "epoch": 1.75, "learning_rate": 1.6945448358405105e-05, "loss": 0.2006, "step": 75200 }, { "epoch": 1.75, "learning_rate": 1.694456830395907e-05, "loss": 0.0706, "step": 75205 }, { "epoch": 1.75, "learning_rate": 1.694368824951304e-05, "loss": 0.2364, "step": 75210 }, { "epoch": 1.75, "learning_rate": 1.6942808195067004e-05, "loss": 0.1684, "step": 75215 }, { "epoch": 1.75, "learning_rate": 1.6941928140620965e-05, "loss": 0.0499, "step": 75220 }, { "epoch": 1.76, "learning_rate": 1.694104808617493e-05, "loss": 0.0465, "step": 75225 }, { "epoch": 1.76, "learning_rate": 1.69401680317289e-05, "loss": 0.0814, "step": 75230 }, { "epoch": 1.76, "learning_rate": 1.693928797728286e-05, "loss": 0.1504, "step": 75235 }, { "epoch": 1.76, "learning_rate": 1.6938407922836825e-05, "loss": 0.046, "step": 75240 }, { "epoch": 1.76, "learning_rate": 1.693752786839079e-05, "loss": 0.1585, "step": 75245 }, { "epoch": 1.76, "learning_rate": 1.693664781394476e-05, "loss": 0.1474, "step": 75250 }, { "epoch": 1.76, "learning_rate": 1.693576775949872e-05, "loss": 0.3313, "step": 75255 }, { "epoch": 1.76, "learning_rate": 1.6934887705052685e-05, "loss": 0.2527, "step": 75260 }, { "epoch": 1.76, "learning_rate": 1.693400765060665e-05, "loss": 0.34, "step": 75265 }, { "epoch": 1.76, "learning_rate": 1.6933127596160616e-05, "loss": 0.0628, "step": 75270 }, { "epoch": 1.76, "learning_rate": 1.693224754171458e-05, "loss": 0.0363, "step": 75275 }, { "epoch": 1.76, "learning_rate": 1.6931367487268546e-05, "loss": 0.0676, "step": 75280 }, { "epoch": 1.76, "learning_rate": 1.6930487432822514e-05, "loss": 0.1915, "step": 75285 }, { "epoch": 1.76, "learning_rate": 1.6929607378376476e-05, "loss": 0.061, "step": 75290 }, { "epoch": 1.76, "learning_rate": 1.692872732393044e-05, "loss": 0.154, "step": 75295 }, { "epoch": 1.76, "learning_rate": 1.6927847269484406e-05, "loss": 0.0717, "step": 75300 }, { "epoch": 1.76, "learning_rate": 1.692696721503837e-05, "loss": 0.1218, "step": 75305 }, { "epoch": 1.76, "learning_rate": 1.6926087160592336e-05, "loss": 0.3594, "step": 75310 }, { "epoch": 1.76, "learning_rate": 1.69252071061463e-05, "loss": 0.2641, "step": 75315 }, { "epoch": 1.76, "learning_rate": 1.6924327051700266e-05, "loss": 0.0085, "step": 75320 }, { "epoch": 1.76, "learning_rate": 1.692344699725423e-05, "loss": 0.0034, "step": 75325 }, { "epoch": 1.76, "learning_rate": 1.6922566942808196e-05, "loss": 0.0324, "step": 75330 }, { "epoch": 1.76, "learning_rate": 1.692168688836216e-05, "loss": 0.1997, "step": 75335 }, { "epoch": 1.76, "learning_rate": 1.6920806833916122e-05, "loss": 0.0589, "step": 75340 }, { "epoch": 1.76, "learning_rate": 1.691992677947009e-05, "loss": 0.2232, "step": 75345 }, { "epoch": 1.76, "learning_rate": 1.6919046725024056e-05, "loss": 0.1092, "step": 75350 }, { "epoch": 1.76, "learning_rate": 1.691816667057802e-05, "loss": 0.0792, "step": 75355 }, { "epoch": 1.76, "learning_rate": 1.6917286616131983e-05, "loss": 0.3518, "step": 75360 }, { "epoch": 1.76, "learning_rate": 1.691640656168595e-05, "loss": 0.2592, "step": 75365 }, { "epoch": 1.76, "learning_rate": 1.6915526507239916e-05, "loss": 0.048, "step": 75370 }, { "epoch": 1.76, "learning_rate": 1.691464645279388e-05, "loss": 0.0541, "step": 75375 }, { "epoch": 1.76, "learning_rate": 1.6913766398347843e-05, "loss": 0.1376, "step": 75380 }, { "epoch": 1.76, "learning_rate": 1.691288634390181e-05, "loss": 0.1325, "step": 75385 }, { "epoch": 1.76, "learning_rate": 1.6912006289455776e-05, "loss": 0.0253, "step": 75390 }, { "epoch": 1.76, "learning_rate": 1.6911126235009738e-05, "loss": 0.0641, "step": 75395 }, { "epoch": 1.76, "learning_rate": 1.6910246180563703e-05, "loss": 0.0998, "step": 75400 }, { "epoch": 1.76, "learning_rate": 1.690936612611767e-05, "loss": 0.19, "step": 75405 }, { "epoch": 1.76, "learning_rate": 1.6908486071671636e-05, "loss": 0.3386, "step": 75410 }, { "epoch": 1.76, "learning_rate": 1.6907606017225598e-05, "loss": 0.106, "step": 75415 }, { "epoch": 1.76, "learning_rate": 1.6906725962779566e-05, "loss": 0.0045, "step": 75420 }, { "epoch": 1.76, "learning_rate": 1.690584590833353e-05, "loss": 0.1321, "step": 75425 }, { "epoch": 1.76, "learning_rate": 1.6904965853887493e-05, "loss": 0.0388, "step": 75430 }, { "epoch": 1.76, "learning_rate": 1.6904085799441458e-05, "loss": 0.0492, "step": 75435 }, { "epoch": 1.76, "learning_rate": 1.6903205744995426e-05, "loss": 0.1057, "step": 75440 }, { "epoch": 1.76, "learning_rate": 1.690232569054939e-05, "loss": 0.102, "step": 75445 }, { "epoch": 1.76, "learning_rate": 1.6901445636103353e-05, "loss": 0.0687, "step": 75450 }, { "epoch": 1.76, "learning_rate": 1.6900565581657318e-05, "loss": 0.2302, "step": 75455 }, { "epoch": 1.76, "learning_rate": 1.6899685527211286e-05, "loss": 0.534, "step": 75460 }, { "epoch": 1.76, "learning_rate": 1.6898805472765248e-05, "loss": 0.3236, "step": 75465 }, { "epoch": 1.76, "learning_rate": 1.6897925418319213e-05, "loss": 0.005, "step": 75470 }, { "epoch": 1.76, "learning_rate": 1.6897045363873178e-05, "loss": 0.036, "step": 75475 }, { "epoch": 1.76, "learning_rate": 1.6896165309427146e-05, "loss": 0.0661, "step": 75480 }, { "epoch": 1.76, "learning_rate": 1.6895285254981108e-05, "loss": 0.2045, "step": 75485 }, { "epoch": 1.76, "learning_rate": 1.6894405200535073e-05, "loss": 0.0547, "step": 75490 }, { "epoch": 1.76, "learning_rate": 1.6893525146089038e-05, "loss": 0.1308, "step": 75495 }, { "epoch": 1.76, "learning_rate": 1.6892645091643003e-05, "loss": 0.2255, "step": 75500 }, { "epoch": 1.76, "learning_rate": 1.6891765037196968e-05, "loss": 0.1093, "step": 75505 }, { "epoch": 1.76, "learning_rate": 1.6890884982750933e-05, "loss": 0.1662, "step": 75510 }, { "epoch": 1.76, "learning_rate": 1.6890004928304898e-05, "loss": 0.2223, "step": 75515 }, { "epoch": 1.76, "learning_rate": 1.6889124873858863e-05, "loss": 0.147, "step": 75520 }, { "epoch": 1.76, "learning_rate": 1.6888244819412828e-05, "loss": 0.0931, "step": 75525 }, { "epoch": 1.76, "learning_rate": 1.6887364764966793e-05, "loss": 0.021, "step": 75530 }, { "epoch": 1.76, "learning_rate": 1.6886484710520755e-05, "loss": 0.0429, "step": 75535 }, { "epoch": 1.76, "learning_rate": 1.6885604656074723e-05, "loss": 0.0683, "step": 75540 }, { "epoch": 1.76, "learning_rate": 1.6884724601628688e-05, "loss": 0.1426, "step": 75545 }, { "epoch": 1.76, "learning_rate": 1.6883844547182653e-05, "loss": 0.1409, "step": 75550 }, { "epoch": 1.76, "learning_rate": 1.6882964492736618e-05, "loss": 0.2426, "step": 75555 }, { "epoch": 1.76, "learning_rate": 1.6882084438290583e-05, "loss": 0.5352, "step": 75560 }, { "epoch": 1.76, "learning_rate": 1.688120438384455e-05, "loss": 0.1571, "step": 75565 }, { "epoch": 1.76, "learning_rate": 1.688032432939851e-05, "loss": 0.0549, "step": 75570 }, { "epoch": 1.76, "learning_rate": 1.687944427495248e-05, "loss": 0.0729, "step": 75575 }, { "epoch": 1.76, "learning_rate": 1.6878564220506443e-05, "loss": 0.0501, "step": 75580 }, { "epoch": 1.76, "learning_rate": 1.687768416606041e-05, "loss": 0.0283, "step": 75585 }, { "epoch": 1.76, "learning_rate": 1.687680411161437e-05, "loss": 0.0793, "step": 75590 }, { "epoch": 1.76, "learning_rate": 1.687592405716834e-05, "loss": 0.0372, "step": 75595 }, { "epoch": 1.76, "learning_rate": 1.6875044002722303e-05, "loss": 0.2332, "step": 75600 }, { "epoch": 1.76, "learning_rate": 1.6874163948276265e-05, "loss": 0.0914, "step": 75605 }, { "epoch": 1.76, "learning_rate": 1.687328389383023e-05, "loss": 0.2508, "step": 75610 }, { "epoch": 1.76, "learning_rate": 1.68724038393842e-05, "loss": 0.1211, "step": 75615 }, { "epoch": 1.76, "learning_rate": 1.6871523784938164e-05, "loss": 0.056, "step": 75620 }, { "epoch": 1.76, "learning_rate": 1.6870643730492125e-05, "loss": 0.0293, "step": 75625 }, { "epoch": 1.76, "learning_rate": 1.686976367604609e-05, "loss": 0.0939, "step": 75630 }, { "epoch": 1.76, "learning_rate": 1.686888362160006e-05, "loss": 0.0695, "step": 75635 }, { "epoch": 1.76, "learning_rate": 1.686800356715402e-05, "loss": 0.0465, "step": 75640 }, { "epoch": 1.76, "learning_rate": 1.6867123512707985e-05, "loss": 0.1277, "step": 75645 }, { "epoch": 1.76, "learning_rate": 1.686624345826195e-05, "loss": 0.1628, "step": 75650 }, { "epoch": 1.77, "learning_rate": 1.686536340381592e-05, "loss": 0.0832, "step": 75655 }, { "epoch": 1.77, "learning_rate": 1.686448334936988e-05, "loss": 0.5762, "step": 75660 }, { "epoch": 1.77, "learning_rate": 1.6863603294923845e-05, "loss": 0.4458, "step": 75665 }, { "epoch": 1.77, "learning_rate": 1.686272324047781e-05, "loss": 0.0241, "step": 75670 }, { "epoch": 1.77, "learning_rate": 1.686184318603178e-05, "loss": 0.1017, "step": 75675 }, { "epoch": 1.77, "learning_rate": 1.686096313158574e-05, "loss": 0.0518, "step": 75680 }, { "epoch": 1.77, "learning_rate": 1.6860083077139705e-05, "loss": 0.1181, "step": 75685 }, { "epoch": 1.77, "learning_rate": 1.6859203022693674e-05, "loss": 0.0468, "step": 75690 }, { "epoch": 1.77, "learning_rate": 1.6858322968247635e-05, "loss": 0.102, "step": 75695 }, { "epoch": 1.77, "learning_rate": 1.68574429138016e-05, "loss": 0.2518, "step": 75700 }, { "epoch": 1.77, "learning_rate": 1.6856562859355566e-05, "loss": 0.1733, "step": 75705 }, { "epoch": 1.77, "learning_rate": 1.6855682804909534e-05, "loss": 0.2763, "step": 75710 }, { "epoch": 1.77, "learning_rate": 1.6854802750463496e-05, "loss": 0.2979, "step": 75715 }, { "epoch": 1.77, "learning_rate": 1.685392269601746e-05, "loss": 0.0008, "step": 75720 }, { "epoch": 1.77, "learning_rate": 1.6853042641571426e-05, "loss": 0.1104, "step": 75725 }, { "epoch": 1.77, "learning_rate": 1.685216258712539e-05, "loss": 0.0491, "step": 75730 }, { "epoch": 1.77, "learning_rate": 1.6851282532679356e-05, "loss": 0.0925, "step": 75735 }, { "epoch": 1.77, "learning_rate": 1.685040247823332e-05, "loss": 0.0712, "step": 75740 }, { "epoch": 1.77, "learning_rate": 1.6849522423787286e-05, "loss": 0.0945, "step": 75745 }, { "epoch": 1.77, "learning_rate": 1.684864236934125e-05, "loss": 0.11, "step": 75750 }, { "epoch": 1.77, "learning_rate": 1.6847762314895216e-05, "loss": 0.0402, "step": 75755 }, { "epoch": 1.77, "learning_rate": 1.684688226044918e-05, "loss": 0.2614, "step": 75760 }, { "epoch": 1.77, "learning_rate": 1.6846002206003142e-05, "loss": 0.2425, "step": 75765 }, { "epoch": 1.77, "learning_rate": 1.684512215155711e-05, "loss": 0.0195, "step": 75770 }, { "epoch": 1.77, "learning_rate": 1.6844242097111076e-05, "loss": 0.0843, "step": 75775 }, { "epoch": 1.77, "learning_rate": 1.684336204266504e-05, "loss": 0.0408, "step": 75780 }, { "epoch": 1.77, "learning_rate": 1.6842481988219002e-05, "loss": 0.0813, "step": 75785 }, { "epoch": 1.77, "learning_rate": 1.684160193377297e-05, "loss": 0.134, "step": 75790 }, { "epoch": 1.77, "learning_rate": 1.6840721879326936e-05, "loss": 0.2527, "step": 75795 }, { "epoch": 1.77, "learning_rate": 1.6839841824880898e-05, "loss": 0.0546, "step": 75800 }, { "epoch": 1.77, "learning_rate": 1.6838961770434863e-05, "loss": 0.2913, "step": 75805 }, { "epoch": 1.77, "learning_rate": 1.683808171598883e-05, "loss": 0.4638, "step": 75810 }, { "epoch": 1.77, "learning_rate": 1.6837201661542796e-05, "loss": 0.2546, "step": 75815 }, { "epoch": 1.77, "learning_rate": 1.6836321607096758e-05, "loss": 0.0386, "step": 75820 }, { "epoch": 1.77, "learning_rate": 1.6835441552650726e-05, "loss": 0.1247, "step": 75825 }, { "epoch": 1.77, "learning_rate": 1.683456149820469e-05, "loss": 0.0426, "step": 75830 }, { "epoch": 1.77, "learning_rate": 1.6833681443758653e-05, "loss": 0.1479, "step": 75835 }, { "epoch": 1.77, "learning_rate": 1.6832801389312618e-05, "loss": 0.1024, "step": 75840 }, { "epoch": 1.77, "learning_rate": 1.6831921334866586e-05, "loss": 0.2215, "step": 75845 }, { "epoch": 1.77, "learning_rate": 1.683104128042055e-05, "loss": 0.2159, "step": 75850 }, { "epoch": 1.77, "learning_rate": 1.6830161225974513e-05, "loss": 0.3408, "step": 75855 }, { "epoch": 1.77, "learning_rate": 1.6829281171528478e-05, "loss": 0.3811, "step": 75860 }, { "epoch": 1.77, "learning_rate": 1.6828401117082446e-05, "loss": 0.1913, "step": 75865 }, { "epoch": 1.77, "learning_rate": 1.6827521062636408e-05, "loss": 0.0568, "step": 75870 }, { "epoch": 1.77, "learning_rate": 1.6826641008190373e-05, "loss": 0.0878, "step": 75875 }, { "epoch": 1.77, "learning_rate": 1.6825760953744338e-05, "loss": 0.0427, "step": 75880 }, { "epoch": 1.77, "learning_rate": 1.6824880899298306e-05, "loss": 0.0874, "step": 75885 }, { "epoch": 1.77, "learning_rate": 1.6824000844852268e-05, "loss": 0.0284, "step": 75890 }, { "epoch": 1.77, "learning_rate": 1.6823120790406233e-05, "loss": 0.0861, "step": 75895 }, { "epoch": 1.77, "learning_rate": 1.6822240735960198e-05, "loss": 0.1961, "step": 75900 }, { "epoch": 1.77, "learning_rate": 1.6821360681514163e-05, "loss": 0.1402, "step": 75905 }, { "epoch": 1.77, "learning_rate": 1.6820480627068128e-05, "loss": 0.323, "step": 75910 }, { "epoch": 1.77, "learning_rate": 1.6819600572622093e-05, "loss": 0.2163, "step": 75915 }, { "epoch": 1.77, "learning_rate": 1.6818720518176058e-05, "loss": 0.1121, "step": 75920 }, { "epoch": 1.77, "learning_rate": 1.6817840463730023e-05, "loss": 0.1145, "step": 75925 }, { "epoch": 1.77, "learning_rate": 1.6816960409283988e-05, "loss": 0.0215, "step": 75930 }, { "epoch": 1.77, "learning_rate": 1.6816080354837953e-05, "loss": 0.0891, "step": 75935 }, { "epoch": 1.77, "learning_rate": 1.6815200300391915e-05, "loss": 0.176, "step": 75940 }, { "epoch": 1.77, "learning_rate": 1.6814320245945883e-05, "loss": 0.1341, "step": 75945 }, { "epoch": 1.77, "learning_rate": 1.6813440191499848e-05, "loss": 0.0414, "step": 75950 }, { "epoch": 1.77, "learning_rate": 1.6812560137053813e-05, "loss": 0.1378, "step": 75955 }, { "epoch": 1.77, "learning_rate": 1.6811680082607778e-05, "loss": 0.158, "step": 75960 }, { "epoch": 1.77, "learning_rate": 1.6810800028161743e-05, "loss": 0.2096, "step": 75965 }, { "epoch": 1.77, "learning_rate": 1.6809919973715708e-05, "loss": 0.0502, "step": 75970 }, { "epoch": 1.77, "learning_rate": 1.6809039919269673e-05, "loss": 0.0007, "step": 75975 }, { "epoch": 1.77, "learning_rate": 1.6808159864823638e-05, "loss": 0.135, "step": 75980 }, { "epoch": 1.77, "learning_rate": 1.6807279810377603e-05, "loss": 0.0499, "step": 75985 }, { "epoch": 1.77, "learning_rate": 1.680639975593157e-05, "loss": 0.0758, "step": 75990 }, { "epoch": 1.77, "learning_rate": 1.680551970148553e-05, "loss": 0.1006, "step": 75995 }, { "epoch": 1.77, "learning_rate": 1.68046396470395e-05, "loss": 0.0602, "step": 76000 }, { "epoch": 1.77, "learning_rate": 1.6803759592593463e-05, "loss": 0.2699, "step": 76005 }, { "epoch": 1.77, "learning_rate": 1.680287953814743e-05, "loss": 0.2404, "step": 76010 }, { "epoch": 1.77, "learning_rate": 1.680199948370139e-05, "loss": 0.1787, "step": 76015 }, { "epoch": 1.77, "learning_rate": 1.680111942925536e-05, "loss": 0.1576, "step": 76020 }, { "epoch": 1.77, "learning_rate": 1.6800239374809323e-05, "loss": 0.1127, "step": 76025 }, { "epoch": 1.77, "learning_rate": 1.6799359320363285e-05, "loss": 0.0773, "step": 76030 }, { "epoch": 1.77, "learning_rate": 1.679847926591725e-05, "loss": 0.112, "step": 76035 }, { "epoch": 1.77, "learning_rate": 1.679759921147122e-05, "loss": 0.2615, "step": 76040 }, { "epoch": 1.77, "learning_rate": 1.6796719157025184e-05, "loss": 0.0634, "step": 76045 }, { "epoch": 1.77, "learning_rate": 1.6795839102579145e-05, "loss": 0.1196, "step": 76050 }, { "epoch": 1.77, "learning_rate": 1.679495904813311e-05, "loss": 0.1821, "step": 76055 }, { "epoch": 1.77, "learning_rate": 1.679407899368708e-05, "loss": 0.25, "step": 76060 }, { "epoch": 1.77, "learning_rate": 1.679319893924104e-05, "loss": 0.2805, "step": 76065 }, { "epoch": 1.77, "learning_rate": 1.6792318884795005e-05, "loss": 0.0981, "step": 76070 }, { "epoch": 1.77, "learning_rate": 1.679143883034897e-05, "loss": 0.0367, "step": 76075 }, { "epoch": 1.78, "learning_rate": 1.679055877590294e-05, "loss": 0.1113, "step": 76080 }, { "epoch": 1.78, "learning_rate": 1.67896787214569e-05, "loss": 0.0904, "step": 76085 }, { "epoch": 1.78, "learning_rate": 1.6788798667010865e-05, "loss": 0.1607, "step": 76090 }, { "epoch": 1.78, "learning_rate": 1.6787918612564834e-05, "loss": 0.056, "step": 76095 }, { "epoch": 1.78, "learning_rate": 1.6787038558118795e-05, "loss": 0.2981, "step": 76100 }, { "epoch": 1.78, "learning_rate": 1.678615850367276e-05, "loss": 0.2726, "step": 76105 }, { "epoch": 1.78, "learning_rate": 1.6785278449226725e-05, "loss": 0.2154, "step": 76110 }, { "epoch": 1.78, "learning_rate": 1.6784398394780694e-05, "loss": 0.2479, "step": 76115 }, { "epoch": 1.78, "learning_rate": 1.6783518340334655e-05, "loss": 0.0512, "step": 76120 }, { "epoch": 1.78, "learning_rate": 1.678263828588862e-05, "loss": 0.0605, "step": 76125 }, { "epoch": 1.78, "learning_rate": 1.6781758231442586e-05, "loss": 0.025, "step": 76130 }, { "epoch": 1.78, "learning_rate": 1.678087817699655e-05, "loss": 0.0735, "step": 76135 }, { "epoch": 1.78, "learning_rate": 1.6779998122550516e-05, "loss": 0.1484, "step": 76140 }, { "epoch": 1.78, "learning_rate": 1.677911806810448e-05, "loss": 0.1876, "step": 76145 }, { "epoch": 1.78, "learning_rate": 1.6778238013658446e-05, "loss": 0.0942, "step": 76150 }, { "epoch": 1.78, "learning_rate": 1.677735795921241e-05, "loss": 0.2316, "step": 76155 }, { "epoch": 1.78, "learning_rate": 1.6776477904766376e-05, "loss": 0.2623, "step": 76160 }, { "epoch": 1.78, "learning_rate": 1.677559785032034e-05, "loss": 0.2532, "step": 76165 }, { "epoch": 1.78, "learning_rate": 1.6774717795874302e-05, "loss": 0.0523, "step": 76170 }, { "epoch": 1.78, "learning_rate": 1.677383774142827e-05, "loss": 0.0225, "step": 76175 }, { "epoch": 1.78, "learning_rate": 1.6772957686982236e-05, "loss": 0.1082, "step": 76180 }, { "epoch": 1.78, "learning_rate": 1.67720776325362e-05, "loss": 0.1363, "step": 76185 }, { "epoch": 1.78, "learning_rate": 1.6771197578090162e-05, "loss": 0.0474, "step": 76190 }, { "epoch": 1.78, "learning_rate": 1.677031752364413e-05, "loss": 0.5694, "step": 76195 }, { "epoch": 1.78, "learning_rate": 1.6769437469198096e-05, "loss": 0.1416, "step": 76200 }, { "epoch": 1.78, "learning_rate": 1.6768557414752057e-05, "loss": 0.2346, "step": 76205 }, { "epoch": 1.78, "learning_rate": 1.6767677360306022e-05, "loss": 0.3463, "step": 76210 }, { "epoch": 1.78, "learning_rate": 1.676679730585999e-05, "loss": 0.2458, "step": 76215 }, { "epoch": 1.78, "learning_rate": 1.6765917251413956e-05, "loss": 0.0321, "step": 76220 }, { "epoch": 1.78, "learning_rate": 1.6765037196967918e-05, "loss": 0.0942, "step": 76225 }, { "epoch": 1.78, "learning_rate": 1.6764157142521886e-05, "loss": 0.0393, "step": 76230 }, { "epoch": 1.78, "learning_rate": 1.676327708807585e-05, "loss": 0.0573, "step": 76235 }, { "epoch": 1.78, "learning_rate": 1.6762397033629813e-05, "loss": 0.0681, "step": 76240 }, { "epoch": 1.78, "learning_rate": 1.6761516979183778e-05, "loss": 0.3095, "step": 76245 }, { "epoch": 1.78, "learning_rate": 1.6760636924737746e-05, "loss": 0.0903, "step": 76250 }, { "epoch": 1.78, "learning_rate": 1.675975687029171e-05, "loss": 0.222, "step": 76255 }, { "epoch": 1.78, "learning_rate": 1.6758876815845673e-05, "loss": 0.47, "step": 76260 }, { "epoch": 1.78, "learning_rate": 1.6757996761399638e-05, "loss": 0.2106, "step": 76265 }, { "epoch": 1.78, "learning_rate": 1.6757116706953606e-05, "loss": 0.0419, "step": 76270 }, { "epoch": 1.78, "learning_rate": 1.675623665250757e-05, "loss": 0.074, "step": 76275 }, { "epoch": 1.78, "learning_rate": 1.6755356598061533e-05, "loss": 0.0089, "step": 76280 }, { "epoch": 1.78, "learning_rate": 1.6754476543615498e-05, "loss": 0.0555, "step": 76285 }, { "epoch": 1.78, "learning_rate": 1.6753596489169466e-05, "loss": 0.2504, "step": 76290 }, { "epoch": 1.78, "learning_rate": 1.6752716434723428e-05, "loss": 0.123, "step": 76295 }, { "epoch": 1.78, "learning_rate": 1.6751836380277393e-05, "loss": 0.0778, "step": 76300 }, { "epoch": 1.78, "learning_rate": 1.6750956325831358e-05, "loss": 0.3065, "step": 76305 }, { "epoch": 1.78, "learning_rate": 1.6750076271385326e-05, "loss": 0.1214, "step": 76310 }, { "epoch": 1.78, "learning_rate": 1.6749196216939288e-05, "loss": 0.256, "step": 76315 }, { "epoch": 1.78, "learning_rate": 1.6748316162493253e-05, "loss": 0.0512, "step": 76320 }, { "epoch": 1.78, "learning_rate": 1.6747436108047218e-05, "loss": 0.0339, "step": 76325 }, { "epoch": 1.78, "learning_rate": 1.6746556053601183e-05, "loss": 0.0087, "step": 76330 }, { "epoch": 1.78, "learning_rate": 1.6745675999155148e-05, "loss": 0.0371, "step": 76335 }, { "epoch": 1.78, "learning_rate": 1.6744795944709113e-05, "loss": 0.0406, "step": 76340 }, { "epoch": 1.78, "learning_rate": 1.6743915890263078e-05, "loss": 0.1085, "step": 76345 }, { "epoch": 1.78, "learning_rate": 1.6743035835817043e-05, "loss": 0.0605, "step": 76350 }, { "epoch": 1.78, "learning_rate": 1.6742155781371008e-05, "loss": 0.4502, "step": 76355 }, { "epoch": 1.78, "learning_rate": 1.6741275726924973e-05, "loss": 0.3208, "step": 76360 }, { "epoch": 1.78, "learning_rate": 1.6740395672478938e-05, "loss": 0.1961, "step": 76365 }, { "epoch": 1.78, "learning_rate": 1.6739515618032903e-05, "loss": 0.0303, "step": 76370 }, { "epoch": 1.78, "learning_rate": 1.6738635563586868e-05, "loss": 0.0713, "step": 76375 }, { "epoch": 1.78, "learning_rate": 1.6737755509140833e-05, "loss": 0.0081, "step": 76380 }, { "epoch": 1.78, "learning_rate": 1.6736875454694798e-05, "loss": 0.1605, "step": 76385 }, { "epoch": 1.78, "learning_rate": 1.6735995400248763e-05, "loss": 0.0944, "step": 76390 }, { "epoch": 1.78, "learning_rate": 1.6735115345802728e-05, "loss": 0.0979, "step": 76395 }, { "epoch": 1.78, "learning_rate": 1.673423529135669e-05, "loss": 0.1735, "step": 76400 }, { "epoch": 1.78, "learning_rate": 1.6733355236910658e-05, "loss": 0.1853, "step": 76405 }, { "epoch": 1.78, "learning_rate": 1.6732475182464623e-05, "loss": 0.3832, "step": 76410 }, { "epoch": 1.78, "learning_rate": 1.6731595128018588e-05, "loss": 0.2677, "step": 76415 }, { "epoch": 1.78, "learning_rate": 1.673071507357255e-05, "loss": 0.0152, "step": 76420 }, { "epoch": 1.78, "learning_rate": 1.672983501912652e-05, "loss": 0.0105, "step": 76425 }, { "epoch": 1.78, "learning_rate": 1.6728954964680483e-05, "loss": 0.0588, "step": 76430 }, { "epoch": 1.78, "learning_rate": 1.6728074910234445e-05, "loss": 0.0912, "step": 76435 }, { "epoch": 1.78, "learning_rate": 1.672719485578841e-05, "loss": 0.1501, "step": 76440 }, { "epoch": 1.78, "learning_rate": 1.672631480134238e-05, "loss": 0.0791, "step": 76445 }, { "epoch": 1.78, "learning_rate": 1.6725434746896343e-05, "loss": 0.1436, "step": 76450 }, { "epoch": 1.78, "learning_rate": 1.6724554692450305e-05, "loss": 0.1093, "step": 76455 }, { "epoch": 1.78, "learning_rate": 1.672367463800427e-05, "loss": 0.1706, "step": 76460 }, { "epoch": 1.78, "learning_rate": 1.672279458355824e-05, "loss": 0.128, "step": 76465 }, { "epoch": 1.78, "learning_rate": 1.67219145291122e-05, "loss": 0.0935, "step": 76470 }, { "epoch": 1.78, "learning_rate": 1.6721034474666165e-05, "loss": 0.0731, "step": 76475 }, { "epoch": 1.78, "learning_rate": 1.672015442022013e-05, "loss": 0.1226, "step": 76480 }, { "epoch": 1.78, "learning_rate": 1.67192743657741e-05, "loss": 0.1076, "step": 76485 }, { "epoch": 1.78, "learning_rate": 1.671839431132806e-05, "loss": 0.0558, "step": 76490 }, { "epoch": 1.78, "learning_rate": 1.6717514256882025e-05, "loss": 0.0937, "step": 76495 }, { "epoch": 1.78, "learning_rate": 1.6716634202435994e-05, "loss": 0.1643, "step": 76500 }, { "epoch": 1.78, "learning_rate": 1.6715754147989955e-05, "loss": 0.2405, "step": 76505 }, { "epoch": 1.79, "learning_rate": 1.671487409354392e-05, "loss": 0.3575, "step": 76510 }, { "epoch": 1.79, "learning_rate": 1.6713994039097885e-05, "loss": 0.3247, "step": 76515 }, { "epoch": 1.79, "learning_rate": 1.6713113984651854e-05, "loss": 0.0318, "step": 76520 }, { "epoch": 1.79, "learning_rate": 1.6712233930205815e-05, "loss": 0.1159, "step": 76525 }, { "epoch": 1.79, "learning_rate": 1.671135387575978e-05, "loss": 0.0222, "step": 76530 }, { "epoch": 1.79, "learning_rate": 1.6710473821313745e-05, "loss": 0.1683, "step": 76535 }, { "epoch": 1.79, "learning_rate": 1.670959376686771e-05, "loss": 0.1824, "step": 76540 }, { "epoch": 1.79, "learning_rate": 1.6708713712421675e-05, "loss": 0.0813, "step": 76545 }, { "epoch": 1.79, "learning_rate": 1.670783365797564e-05, "loss": 0.159, "step": 76550 }, { "epoch": 1.79, "learning_rate": 1.6706953603529605e-05, "loss": 0.3134, "step": 76555 }, { "epoch": 1.79, "learning_rate": 1.670607354908357e-05, "loss": 0.3502, "step": 76560 }, { "epoch": 1.79, "learning_rate": 1.6705193494637536e-05, "loss": 0.3719, "step": 76565 }, { "epoch": 1.79, "learning_rate": 1.67043134401915e-05, "loss": 0.0034, "step": 76570 }, { "epoch": 1.79, "learning_rate": 1.6703433385745466e-05, "loss": 0.0246, "step": 76575 }, { "epoch": 1.79, "learning_rate": 1.6702729342188637e-05, "loss": 0.0453, "step": 76580 }, { "epoch": 1.79, "learning_rate": 1.6701849287742602e-05, "loss": 0.0396, "step": 76585 }, { "epoch": 1.79, "learning_rate": 1.6700969233296567e-05, "loss": 0.0819, "step": 76590 }, { "epoch": 1.79, "learning_rate": 1.6700089178850532e-05, "loss": 0.134, "step": 76595 }, { "epoch": 1.79, "learning_rate": 1.6699209124404497e-05, "loss": 0.0752, "step": 76600 }, { "epoch": 1.79, "learning_rate": 1.6698329069958462e-05, "loss": 0.2023, "step": 76605 }, { "epoch": 1.79, "learning_rate": 1.6697449015512427e-05, "loss": 0.417, "step": 76610 }, { "epoch": 1.79, "learning_rate": 1.6696568961066392e-05, "loss": 0.2928, "step": 76615 }, { "epoch": 1.79, "learning_rate": 1.6695688906620357e-05, "loss": 0.0363, "step": 76620 }, { "epoch": 1.79, "learning_rate": 1.6694808852174322e-05, "loss": 0.0062, "step": 76625 }, { "epoch": 1.79, "learning_rate": 1.6693928797728284e-05, "loss": 0.0142, "step": 76630 }, { "epoch": 1.79, "learning_rate": 1.6693048743282252e-05, "loss": 0.0624, "step": 76635 }, { "epoch": 1.79, "learning_rate": 1.6692168688836217e-05, "loss": 0.1059, "step": 76640 }, { "epoch": 1.79, "learning_rate": 1.6691288634390182e-05, "loss": 0.082, "step": 76645 }, { "epoch": 1.79, "learning_rate": 1.6690408579944144e-05, "loss": 0.0382, "step": 76650 }, { "epoch": 1.79, "learning_rate": 1.6689528525498112e-05, "loss": 0.2727, "step": 76655 }, { "epoch": 1.79, "learning_rate": 1.6688648471052077e-05, "loss": 0.1608, "step": 76660 }, { "epoch": 1.79, "learning_rate": 1.668776841660604e-05, "loss": 0.1146, "step": 76665 }, { "epoch": 1.79, "learning_rate": 1.6686888362160004e-05, "loss": 0.0122, "step": 76670 }, { "epoch": 1.79, "learning_rate": 1.6686008307713972e-05, "loss": 0.0609, "step": 76675 }, { "epoch": 1.79, "learning_rate": 1.6685128253267937e-05, "loss": 0.0291, "step": 76680 }, { "epoch": 1.79, "learning_rate": 1.66842481988219e-05, "loss": 0.0292, "step": 76685 }, { "epoch": 1.79, "learning_rate": 1.6683368144375867e-05, "loss": 0.0886, "step": 76690 }, { "epoch": 1.79, "learning_rate": 1.6682488089929832e-05, "loss": 0.0985, "step": 76695 }, { "epoch": 1.79, "learning_rate": 1.6681608035483794e-05, "loss": 0.0298, "step": 76700 }, { "epoch": 1.79, "learning_rate": 1.668072798103776e-05, "loss": 0.303, "step": 76705 }, { "epoch": 1.79, "learning_rate": 1.6679847926591727e-05, "loss": 0.2347, "step": 76710 }, { "epoch": 1.79, "learning_rate": 1.6678967872145692e-05, "loss": 0.1787, "step": 76715 }, { "epoch": 1.79, "learning_rate": 1.6678087817699654e-05, "loss": 0.0639, "step": 76720 }, { "epoch": 1.79, "learning_rate": 1.667720776325362e-05, "loss": 0.0253, "step": 76725 }, { "epoch": 1.79, "learning_rate": 1.6676327708807588e-05, "loss": 0.0845, "step": 76730 }, { "epoch": 1.79, "learning_rate": 1.667544765436155e-05, "loss": 0.1213, "step": 76735 }, { "epoch": 1.79, "learning_rate": 1.6674567599915514e-05, "loss": 0.1266, "step": 76740 }, { "epoch": 1.79, "learning_rate": 1.667368754546948e-05, "loss": 0.199, "step": 76745 }, { "epoch": 1.79, "learning_rate": 1.6672807491023448e-05, "loss": 0.143, "step": 76750 }, { "epoch": 1.79, "learning_rate": 1.667192743657741e-05, "loss": 0.1951, "step": 76755 }, { "epoch": 1.79, "learning_rate": 1.6671047382131374e-05, "loss": 0.385, "step": 76760 }, { "epoch": 1.79, "learning_rate": 1.667016732768534e-05, "loss": 0.1994, "step": 76765 }, { "epoch": 1.79, "learning_rate": 1.6669287273239308e-05, "loss": 0.057, "step": 76770 }, { "epoch": 1.79, "learning_rate": 1.666840721879327e-05, "loss": 0.0484, "step": 76775 }, { "epoch": 1.79, "learning_rate": 1.6667527164347234e-05, "loss": 0.0409, "step": 76780 }, { "epoch": 1.79, "learning_rate": 1.66666471099012e-05, "loss": 0.1007, "step": 76785 }, { "epoch": 1.79, "learning_rate": 1.6665767055455164e-05, "loss": 0.1179, "step": 76790 }, { "epoch": 1.79, "learning_rate": 1.666488700100913e-05, "loss": 0.1734, "step": 76795 }, { "epoch": 1.79, "learning_rate": 1.6664006946563094e-05, "loss": 0.1848, "step": 76800 }, { "epoch": 1.79, "learning_rate": 1.666312689211706e-05, "loss": 0.0958, "step": 76805 }, { "epoch": 1.79, "learning_rate": 1.6662246837671024e-05, "loss": 0.2379, "step": 76810 }, { "epoch": 1.79, "learning_rate": 1.666136678322499e-05, "loss": 0.2557, "step": 76815 }, { "epoch": 1.79, "learning_rate": 1.6660486728778954e-05, "loss": 0.0709, "step": 76820 }, { "epoch": 1.79, "learning_rate": 1.665960667433292e-05, "loss": 0.0106, "step": 76825 }, { "epoch": 1.79, "learning_rate": 1.6658726619886885e-05, "loss": 0.004, "step": 76830 }, { "epoch": 1.79, "learning_rate": 1.665784656544085e-05, "loss": 0.211, "step": 76835 }, { "epoch": 1.79, "learning_rate": 1.6656966510994815e-05, "loss": 0.1332, "step": 76840 }, { "epoch": 1.79, "learning_rate": 1.665608645654878e-05, "loss": 0.1072, "step": 76845 }, { "epoch": 1.79, "learning_rate": 1.6655206402102745e-05, "loss": 0.1714, "step": 76850 }, { "epoch": 1.79, "learning_rate": 1.665432634765671e-05, "loss": 0.2474, "step": 76855 }, { "epoch": 1.79, "learning_rate": 1.665344629321067e-05, "loss": 0.2823, "step": 76860 }, { "epoch": 1.79, "learning_rate": 1.665256623876464e-05, "loss": 0.2551, "step": 76865 }, { "epoch": 1.79, "learning_rate": 1.6651686184318605e-05, "loss": 0.0453, "step": 76870 }, { "epoch": 1.79, "learning_rate": 1.665080612987257e-05, "loss": 0.0407, "step": 76875 }, { "epoch": 1.79, "learning_rate": 1.664992607542653e-05, "loss": 0.1111, "step": 76880 }, { "epoch": 1.79, "learning_rate": 1.66490460209805e-05, "loss": 0.0704, "step": 76885 }, { "epoch": 1.79, "learning_rate": 1.6648165966534465e-05, "loss": 0.1927, "step": 76890 }, { "epoch": 1.79, "learning_rate": 1.6647285912088426e-05, "loss": 0.205, "step": 76895 }, { "epoch": 1.79, "learning_rate": 1.664640585764239e-05, "loss": 0.2543, "step": 76900 }, { "epoch": 1.79, "learning_rate": 1.664552580319636e-05, "loss": 0.2612, "step": 76905 }, { "epoch": 1.79, "learning_rate": 1.6644645748750325e-05, "loss": 0.2502, "step": 76910 }, { "epoch": 1.79, "learning_rate": 1.6643765694304286e-05, "loss": 0.2585, "step": 76915 }, { "epoch": 1.79, "learning_rate": 1.664288563985825e-05, "loss": 0.1173, "step": 76920 }, { "epoch": 1.79, "learning_rate": 1.664200558541222e-05, "loss": 0.0216, "step": 76925 }, { "epoch": 1.79, "learning_rate": 1.664112553096618e-05, "loss": 0.0471, "step": 76930 }, { "epoch": 1.79, "learning_rate": 1.6640245476520147e-05, "loss": 0.0459, "step": 76935 }, { "epoch": 1.8, "learning_rate": 1.663936542207411e-05, "loss": 0.1764, "step": 76940 }, { "epoch": 1.8, "learning_rate": 1.663848536762808e-05, "loss": 0.1843, "step": 76945 }, { "epoch": 1.8, "learning_rate": 1.663760531318204e-05, "loss": 0.2806, "step": 76950 }, { "epoch": 1.8, "learning_rate": 1.6636725258736007e-05, "loss": 0.4041, "step": 76955 }, { "epoch": 1.8, "learning_rate": 1.663584520428997e-05, "loss": 0.3523, "step": 76960 }, { "epoch": 1.8, "learning_rate": 1.6634965149843937e-05, "loss": 0.3318, "step": 76965 }, { "epoch": 1.8, "learning_rate": 1.6634085095397902e-05, "loss": 0.0549, "step": 76970 }, { "epoch": 1.8, "learning_rate": 1.6633205040951867e-05, "loss": 0.0969, "step": 76975 }, { "epoch": 1.8, "learning_rate": 1.6632324986505835e-05, "loss": 0.1415, "step": 76980 }, { "epoch": 1.8, "learning_rate": 1.6631444932059797e-05, "loss": 0.0757, "step": 76985 }, { "epoch": 1.8, "learning_rate": 1.6630564877613762e-05, "loss": 0.1156, "step": 76990 }, { "epoch": 1.8, "learning_rate": 1.6629684823167727e-05, "loss": 0.1004, "step": 76995 }, { "epoch": 1.8, "learning_rate": 1.6628804768721692e-05, "loss": 0.1237, "step": 77000 }, { "epoch": 1.8, "learning_rate": 1.6627924714275657e-05, "loss": 0.3455, "step": 77005 }, { "epoch": 1.8, "learning_rate": 1.6627044659829622e-05, "loss": 0.2628, "step": 77010 }, { "epoch": 1.8, "learning_rate": 1.6626164605383587e-05, "loss": 0.093, "step": 77015 }, { "epoch": 1.8, "learning_rate": 1.6625284550937552e-05, "loss": 0.0123, "step": 77020 }, { "epoch": 1.8, "learning_rate": 1.6624404496491517e-05, "loss": 0.012, "step": 77025 }, { "epoch": 1.8, "learning_rate": 1.6623524442045482e-05, "loss": 0.0137, "step": 77030 }, { "epoch": 1.8, "learning_rate": 1.6622644387599444e-05, "loss": 0.2208, "step": 77035 }, { "epoch": 1.8, "learning_rate": 1.6621764333153412e-05, "loss": 0.171, "step": 77040 }, { "epoch": 1.8, "learning_rate": 1.6620884278707377e-05, "loss": 0.2164, "step": 77045 }, { "epoch": 1.8, "learning_rate": 1.6620004224261342e-05, "loss": 0.138, "step": 77050 }, { "epoch": 1.8, "learning_rate": 1.6619124169815304e-05, "loss": 0.1374, "step": 77055 }, { "epoch": 1.8, "learning_rate": 1.6618244115369272e-05, "loss": 0.2348, "step": 77060 }, { "epoch": 1.8, "learning_rate": 1.6617364060923237e-05, "loss": 0.321, "step": 77065 }, { "epoch": 1.8, "learning_rate": 1.6616484006477202e-05, "loss": 0.0646, "step": 77070 }, { "epoch": 1.8, "learning_rate": 1.6615603952031164e-05, "loss": 0.0803, "step": 77075 }, { "epoch": 1.8, "learning_rate": 1.6614723897585132e-05, "loss": 0.1439, "step": 77080 }, { "epoch": 1.8, "learning_rate": 1.6613843843139097e-05, "loss": 0.1169, "step": 77085 }, { "epoch": 1.8, "learning_rate": 1.661296378869306e-05, "loss": 0.0341, "step": 77090 }, { "epoch": 1.8, "learning_rate": 1.6612083734247024e-05, "loss": 0.1736, "step": 77095 }, { "epoch": 1.8, "learning_rate": 1.6611203679800992e-05, "loss": 0.1909, "step": 77100 }, { "epoch": 1.8, "learning_rate": 1.6610323625354957e-05, "loss": 0.3442, "step": 77105 }, { "epoch": 1.8, "learning_rate": 1.660944357090892e-05, "loss": 0.4299, "step": 77110 }, { "epoch": 1.8, "learning_rate": 1.6608563516462887e-05, "loss": 0.1865, "step": 77115 }, { "epoch": 1.8, "learning_rate": 1.6607683462016852e-05, "loss": 0.0306, "step": 77120 }, { "epoch": 1.8, "learning_rate": 1.6606803407570814e-05, "loss": 0.0584, "step": 77125 }, { "epoch": 1.8, "learning_rate": 1.660592335312478e-05, "loss": 0.0218, "step": 77130 }, { "epoch": 1.8, "learning_rate": 1.6605043298678747e-05, "loss": 0.0562, "step": 77135 }, { "epoch": 1.8, "learning_rate": 1.6604163244232712e-05, "loss": 0.1739, "step": 77140 }, { "epoch": 1.8, "learning_rate": 1.6603283189786674e-05, "loss": 0.1697, "step": 77145 }, { "epoch": 1.8, "learning_rate": 1.660240313534064e-05, "loss": 0.0487, "step": 77150 }, { "epoch": 1.8, "learning_rate": 1.6601523080894607e-05, "loss": 0.1397, "step": 77155 }, { "epoch": 1.8, "learning_rate": 1.660064302644857e-05, "loss": 0.2416, "step": 77160 }, { "epoch": 1.8, "learning_rate": 1.6599762972002534e-05, "loss": 0.2344, "step": 77165 }, { "epoch": 1.8, "learning_rate": 1.65988829175565e-05, "loss": 0.0358, "step": 77170 }, { "epoch": 1.8, "learning_rate": 1.6598002863110468e-05, "loss": 0.054, "step": 77175 }, { "epoch": 1.8, "learning_rate": 1.659712280866443e-05, "loss": 0.0462, "step": 77180 }, { "epoch": 1.8, "learning_rate": 1.6596242754218394e-05, "loss": 0.1052, "step": 77185 }, { "epoch": 1.8, "learning_rate": 1.659536269977236e-05, "loss": 0.2072, "step": 77190 }, { "epoch": 1.8, "learning_rate": 1.6594482645326324e-05, "loss": 0.0966, "step": 77195 }, { "epoch": 1.8, "learning_rate": 1.659360259088029e-05, "loss": 0.15, "step": 77200 }, { "epoch": 1.8, "learning_rate": 1.6592722536434254e-05, "loss": 0.1465, "step": 77205 }, { "epoch": 1.8, "learning_rate": 1.659184248198822e-05, "loss": 0.3009, "step": 77210 }, { "epoch": 1.8, "learning_rate": 1.6590962427542184e-05, "loss": 0.236, "step": 77215 }, { "epoch": 1.8, "learning_rate": 1.659008237309615e-05, "loss": 0.0708, "step": 77220 }, { "epoch": 1.8, "learning_rate": 1.6589202318650114e-05, "loss": 0.0592, "step": 77225 }, { "epoch": 1.8, "learning_rate": 1.658832226420408e-05, "loss": 0.0552, "step": 77230 }, { "epoch": 1.8, "learning_rate": 1.6587442209758044e-05, "loss": 0.0304, "step": 77235 }, { "epoch": 1.8, "learning_rate": 1.658656215531201e-05, "loss": 0.0541, "step": 77240 }, { "epoch": 1.8, "learning_rate": 1.6585682100865974e-05, "loss": 0.1223, "step": 77245 }, { "epoch": 1.8, "learning_rate": 1.658480204641994e-05, "loss": 0.2094, "step": 77250 }, { "epoch": 1.8, "learning_rate": 1.6583921991973904e-05, "loss": 0.1793, "step": 77255 }, { "epoch": 1.8, "learning_rate": 1.658304193752787e-05, "loss": 0.3164, "step": 77260 }, { "epoch": 1.8, "learning_rate": 1.658216188308183e-05, "loss": 0.3735, "step": 77265 }, { "epoch": 1.8, "learning_rate": 1.65812818286358e-05, "loss": 0.0006, "step": 77270 }, { "epoch": 1.8, "learning_rate": 1.6580401774189765e-05, "loss": 0.0599, "step": 77275 }, { "epoch": 1.8, "learning_rate": 1.657952171974373e-05, "loss": 0.0545, "step": 77280 }, { "epoch": 1.8, "learning_rate": 1.657864166529769e-05, "loss": 0.0304, "step": 77285 }, { "epoch": 1.8, "learning_rate": 1.657776161085166e-05, "loss": 0.0441, "step": 77290 }, { "epoch": 1.8, "learning_rate": 1.6576881556405625e-05, "loss": 0.1592, "step": 77295 }, { "epoch": 1.8, "learning_rate": 1.6576001501959586e-05, "loss": 0.0615, "step": 77300 }, { "epoch": 1.8, "learning_rate": 1.657512144751355e-05, "loss": 0.2767, "step": 77305 }, { "epoch": 1.8, "learning_rate": 1.657424139306752e-05, "loss": 0.1038, "step": 77310 }, { "epoch": 1.8, "learning_rate": 1.6573361338621485e-05, "loss": 0.286, "step": 77315 }, { "epoch": 1.8, "learning_rate": 1.6572481284175446e-05, "loss": 0.0555, "step": 77320 }, { "epoch": 1.8, "learning_rate": 1.657160122972941e-05, "loss": 0.0346, "step": 77325 }, { "epoch": 1.8, "learning_rate": 1.657072117528338e-05, "loss": 0.049, "step": 77330 }, { "epoch": 1.8, "learning_rate": 1.656984112083734e-05, "loss": 0.0501, "step": 77335 }, { "epoch": 1.8, "learning_rate": 1.6568961066391306e-05, "loss": 0.0443, "step": 77340 }, { "epoch": 1.8, "learning_rate": 1.656808101194527e-05, "loss": 0.0992, "step": 77345 }, { "epoch": 1.8, "learning_rate": 1.656720095749924e-05, "loss": 0.1515, "step": 77350 }, { "epoch": 1.8, "learning_rate": 1.65663209030532e-05, "loss": 0.1175, "step": 77355 }, { "epoch": 1.8, "learning_rate": 1.6565440848607167e-05, "loss": 0.5541, "step": 77360 }, { "epoch": 1.81, "learning_rate": 1.656456079416113e-05, "loss": 0.2131, "step": 77365 }, { "epoch": 1.81, "learning_rate": 1.65636807397151e-05, "loss": 0.0369, "step": 77370 }, { "epoch": 1.81, "learning_rate": 1.656280068526906e-05, "loss": 0.0491, "step": 77375 }, { "epoch": 1.81, "learning_rate": 1.6561920630823027e-05, "loss": 0.1036, "step": 77380 }, { "epoch": 1.81, "learning_rate": 1.6561040576376995e-05, "loss": 0.0383, "step": 77385 }, { "epoch": 1.81, "learning_rate": 1.6560160521930957e-05, "loss": 0.2794, "step": 77390 }, { "epoch": 1.81, "learning_rate": 1.655928046748492e-05, "loss": 0.154, "step": 77395 }, { "epoch": 1.81, "learning_rate": 1.6558400413038887e-05, "loss": 0.1038, "step": 77400 }, { "epoch": 1.81, "learning_rate": 1.6557520358592855e-05, "loss": 0.1553, "step": 77405 }, { "epoch": 1.81, "learning_rate": 1.6556640304146817e-05, "loss": 0.2674, "step": 77410 }, { "epoch": 1.81, "learning_rate": 1.6555760249700782e-05, "loss": 0.1717, "step": 77415 }, { "epoch": 1.81, "learning_rate": 1.6554880195254747e-05, "loss": 0.0365, "step": 77420 }, { "epoch": 1.81, "learning_rate": 1.6554000140808712e-05, "loss": 0.038, "step": 77425 }, { "epoch": 1.81, "learning_rate": 1.6553120086362677e-05, "loss": 0.0784, "step": 77430 }, { "epoch": 1.81, "learning_rate": 1.6552240031916642e-05, "loss": 0.1308, "step": 77435 }, { "epoch": 1.81, "learning_rate": 1.6551359977470607e-05, "loss": 0.1376, "step": 77440 }, { "epoch": 1.81, "learning_rate": 1.6550479923024572e-05, "loss": 0.0408, "step": 77445 }, { "epoch": 1.81, "learning_rate": 1.6549599868578537e-05, "loss": 0.183, "step": 77450 }, { "epoch": 1.81, "learning_rate": 1.6548719814132502e-05, "loss": 0.2296, "step": 77455 }, { "epoch": 1.81, "learning_rate": 1.6547839759686464e-05, "loss": 0.2194, "step": 77460 }, { "epoch": 1.81, "learning_rate": 1.6546959705240432e-05, "loss": 0.3556, "step": 77465 }, { "epoch": 1.81, "learning_rate": 1.6546079650794397e-05, "loss": 0.1146, "step": 77470 }, { "epoch": 1.81, "learning_rate": 1.6545199596348362e-05, "loss": 0.0065, "step": 77475 }, { "epoch": 1.81, "learning_rate": 1.6544319541902324e-05, "loss": 0.0861, "step": 77480 }, { "epoch": 1.81, "learning_rate": 1.6543439487456292e-05, "loss": 0.1124, "step": 77485 }, { "epoch": 1.81, "learning_rate": 1.6542559433010257e-05, "loss": 0.0556, "step": 77490 }, { "epoch": 1.81, "learning_rate": 1.654167937856422e-05, "loss": 0.0464, "step": 77495 }, { "epoch": 1.81, "learning_rate": 1.6540799324118184e-05, "loss": 0.1832, "step": 77500 }, { "epoch": 1.81, "learning_rate": 1.6539919269672152e-05, "loss": 0.1781, "step": 77505 }, { "epoch": 1.81, "learning_rate": 1.6539039215226117e-05, "loss": 0.3588, "step": 77510 }, { "epoch": 1.81, "learning_rate": 1.653815916078008e-05, "loss": 0.1766, "step": 77515 }, { "epoch": 1.81, "learning_rate": 1.6537279106334047e-05, "loss": 0.0767, "step": 77520 }, { "epoch": 1.81, "learning_rate": 1.6536399051888012e-05, "loss": 0.0772, "step": 77525 }, { "epoch": 1.81, "learning_rate": 1.6535518997441974e-05, "loss": 0.1172, "step": 77530 }, { "epoch": 1.81, "learning_rate": 1.653463894299594e-05, "loss": 0.0517, "step": 77535 }, { "epoch": 1.81, "learning_rate": 1.6533758888549907e-05, "loss": 0.194, "step": 77540 }, { "epoch": 1.81, "learning_rate": 1.6532878834103872e-05, "loss": 0.0598, "step": 77545 }, { "epoch": 1.81, "learning_rate": 1.6531998779657834e-05, "loss": 0.0402, "step": 77550 }, { "epoch": 1.81, "learning_rate": 1.65311187252118e-05, "loss": 0.0695, "step": 77555 }, { "epoch": 1.81, "learning_rate": 1.6530238670765767e-05, "loss": 0.3305, "step": 77560 }, { "epoch": 1.81, "learning_rate": 1.652935861631973e-05, "loss": 0.2652, "step": 77565 }, { "epoch": 1.81, "learning_rate": 1.6528478561873694e-05, "loss": 0.0155, "step": 77570 }, { "epoch": 1.81, "learning_rate": 1.652759850742766e-05, "loss": 0.0421, "step": 77575 }, { "epoch": 1.81, "learning_rate": 1.6526718452981627e-05, "loss": 0.0435, "step": 77580 }, { "epoch": 1.81, "learning_rate": 1.652583839853559e-05, "loss": 0.0545, "step": 77585 }, { "epoch": 1.81, "learning_rate": 1.6524958344089554e-05, "loss": 0.0793, "step": 77590 }, { "epoch": 1.81, "learning_rate": 1.652407828964352e-05, "loss": 0.2129, "step": 77595 }, { "epoch": 1.81, "learning_rate": 1.6523198235197484e-05, "loss": 0.1426, "step": 77600 }, { "epoch": 1.81, "learning_rate": 1.652231818075145e-05, "loss": 0.2185, "step": 77605 }, { "epoch": 1.81, "learning_rate": 1.6521438126305414e-05, "loss": 0.2715, "step": 77610 }, { "epoch": 1.81, "learning_rate": 1.652055807185938e-05, "loss": 0.1931, "step": 77615 }, { "epoch": 1.81, "learning_rate": 1.6519678017413344e-05, "loss": 0.0874, "step": 77620 }, { "epoch": 1.81, "learning_rate": 1.651879796296731e-05, "loss": 0.0774, "step": 77625 }, { "epoch": 1.81, "learning_rate": 1.6517917908521274e-05, "loss": 0.1274, "step": 77630 }, { "epoch": 1.81, "learning_rate": 1.651703785407524e-05, "loss": 0.1425, "step": 77635 }, { "epoch": 1.81, "learning_rate": 1.6516157799629204e-05, "loss": 0.0777, "step": 77640 }, { "epoch": 1.81, "learning_rate": 1.651527774518317e-05, "loss": 0.0605, "step": 77645 }, { "epoch": 1.81, "learning_rate": 1.6514397690737134e-05, "loss": 0.0252, "step": 77650 }, { "epoch": 1.81, "learning_rate": 1.65135176362911e-05, "loss": 0.173, "step": 77655 }, { "epoch": 1.81, "learning_rate": 1.6512637581845064e-05, "loss": 0.2917, "step": 77660 }, { "epoch": 1.81, "learning_rate": 1.651175752739903e-05, "loss": 0.3362, "step": 77665 }, { "epoch": 1.81, "learning_rate": 1.6510877472952994e-05, "loss": 0.0561, "step": 77670 }, { "epoch": 1.81, "learning_rate": 1.650999741850696e-05, "loss": 0.0262, "step": 77675 }, { "epoch": 1.81, "learning_rate": 1.6509117364060924e-05, "loss": 0.202, "step": 77680 }, { "epoch": 1.81, "learning_rate": 1.650823730961489e-05, "loss": 0.0284, "step": 77685 }, { "epoch": 1.81, "learning_rate": 1.650735725516885e-05, "loss": 0.0598, "step": 77690 }, { "epoch": 1.81, "learning_rate": 1.650647720072282e-05, "loss": 0.1831, "step": 77695 }, { "epoch": 1.81, "learning_rate": 1.6505597146276785e-05, "loss": 0.1893, "step": 77700 }, { "epoch": 1.81, "learning_rate": 1.650471709183075e-05, "loss": 0.2067, "step": 77705 }, { "epoch": 1.81, "learning_rate": 1.650383703738471e-05, "loss": 0.2323, "step": 77710 }, { "epoch": 1.81, "learning_rate": 1.650295698293868e-05, "loss": 0.1928, "step": 77715 }, { "epoch": 1.81, "learning_rate": 1.6502076928492645e-05, "loss": 0.0519, "step": 77720 }, { "epoch": 1.81, "learning_rate": 1.6501196874046606e-05, "loss": 0.0626, "step": 77725 }, { "epoch": 1.81, "learning_rate": 1.650031681960057e-05, "loss": 0.1104, "step": 77730 }, { "epoch": 1.81, "learning_rate": 1.649943676515454e-05, "loss": 0.0578, "step": 77735 }, { "epoch": 1.81, "learning_rate": 1.6498556710708505e-05, "loss": 0.092, "step": 77740 }, { "epoch": 1.81, "learning_rate": 1.6497676656262466e-05, "loss": 0.1508, "step": 77745 }, { "epoch": 1.81, "learning_rate": 1.649679660181643e-05, "loss": 0.1024, "step": 77750 }, { "epoch": 1.81, "learning_rate": 1.64959165473704e-05, "loss": 0.1151, "step": 77755 }, { "epoch": 1.81, "learning_rate": 1.649503649292436e-05, "loss": 0.258, "step": 77760 }, { "epoch": 1.81, "learning_rate": 1.6494156438478326e-05, "loss": 0.3235, "step": 77765 }, { "epoch": 1.81, "learning_rate": 1.649327638403229e-05, "loss": 0.0074, "step": 77770 }, { "epoch": 1.81, "learning_rate": 1.649239632958626e-05, "loss": 0.0043, "step": 77775 }, { "epoch": 1.81, "learning_rate": 1.649151627514022e-05, "loss": 0.0357, "step": 77780 }, { "epoch": 1.81, "learning_rate": 1.6490636220694187e-05, "loss": 0.0215, "step": 77785 }, { "epoch": 1.81, "learning_rate": 1.6489756166248155e-05, "loss": 0.0391, "step": 77790 }, { "epoch": 1.82, "learning_rate": 1.6488876111802117e-05, "loss": 0.1735, "step": 77795 }, { "epoch": 1.82, "learning_rate": 1.648799605735608e-05, "loss": 0.1178, "step": 77800 }, { "epoch": 1.82, "learning_rate": 1.6487116002910047e-05, "loss": 0.2294, "step": 77805 }, { "epoch": 1.82, "learning_rate": 1.6486235948464015e-05, "loss": 0.4323, "step": 77810 }, { "epoch": 1.82, "learning_rate": 1.6485355894017977e-05, "loss": 0.1844, "step": 77815 }, { "epoch": 1.82, "learning_rate": 1.648447583957194e-05, "loss": 0.046, "step": 77820 }, { "epoch": 1.82, "learning_rate": 1.6483595785125907e-05, "loss": 0.0545, "step": 77825 }, { "epoch": 1.82, "learning_rate": 1.648271573067987e-05, "loss": 0.0596, "step": 77830 }, { "epoch": 1.82, "learning_rate": 1.6481835676233837e-05, "loss": 0.0667, "step": 77835 }, { "epoch": 1.82, "learning_rate": 1.6480955621787802e-05, "loss": 0.0472, "step": 77840 }, { "epoch": 1.82, "learning_rate": 1.6480075567341767e-05, "loss": 0.1476, "step": 77845 }, { "epoch": 1.82, "learning_rate": 1.6479195512895732e-05, "loss": 0.0867, "step": 77850 }, { "epoch": 1.82, "learning_rate": 1.6478315458449697e-05, "loss": 0.3813, "step": 77855 }, { "epoch": 1.82, "learning_rate": 1.6477435404003662e-05, "loss": 0.4149, "step": 77860 }, { "epoch": 1.82, "learning_rate": 1.6476555349557623e-05, "loss": 0.3361, "step": 77865 }, { "epoch": 1.82, "learning_rate": 1.6475675295111592e-05, "loss": 0.1317, "step": 77870 }, { "epoch": 1.82, "learning_rate": 1.6474795240665557e-05, "loss": 0.0279, "step": 77875 }, { "epoch": 1.82, "learning_rate": 1.6473915186219522e-05, "loss": 0.0114, "step": 77880 }, { "epoch": 1.82, "learning_rate": 1.6473035131773484e-05, "loss": 0.1153, "step": 77885 }, { "epoch": 1.82, "learning_rate": 1.6472155077327452e-05, "loss": 0.0394, "step": 77890 }, { "epoch": 1.82, "learning_rate": 1.6471275022881417e-05, "loss": 0.1997, "step": 77895 }, { "epoch": 1.82, "learning_rate": 1.647039496843538e-05, "loss": 0.072, "step": 77900 }, { "epoch": 1.82, "learning_rate": 1.6469514913989344e-05, "loss": 0.1122, "step": 77905 }, { "epoch": 1.82, "learning_rate": 1.6468634859543312e-05, "loss": 0.392, "step": 77910 }, { "epoch": 1.82, "learning_rate": 1.6467754805097277e-05, "loss": 0.2897, "step": 77915 }, { "epoch": 1.82, "learning_rate": 1.646687475065124e-05, "loss": 0.0601, "step": 77920 }, { "epoch": 1.82, "learning_rate": 1.6465994696205207e-05, "loss": 0.0988, "step": 77925 }, { "epoch": 1.82, "learning_rate": 1.6465114641759172e-05, "loss": 0.1028, "step": 77930 }, { "epoch": 1.82, "learning_rate": 1.6464234587313134e-05, "loss": 0.0918, "step": 77935 }, { "epoch": 1.82, "learning_rate": 1.64633545328671e-05, "loss": 0.0356, "step": 77940 }, { "epoch": 1.82, "learning_rate": 1.6462474478421067e-05, "loss": 0.1464, "step": 77945 }, { "epoch": 1.82, "learning_rate": 1.6461594423975032e-05, "loss": 0.3235, "step": 77950 }, { "epoch": 1.82, "learning_rate": 1.6460714369528994e-05, "loss": 0.3033, "step": 77955 }, { "epoch": 1.82, "learning_rate": 1.645983431508296e-05, "loss": 0.1708, "step": 77960 }, { "epoch": 1.82, "learning_rate": 1.6458954260636927e-05, "loss": 0.143, "step": 77965 }, { "epoch": 1.82, "learning_rate": 1.645807420619089e-05, "loss": 0.1013, "step": 77970 }, { "epoch": 1.82, "learning_rate": 1.6457194151744854e-05, "loss": 0.026, "step": 77975 }, { "epoch": 1.82, "learning_rate": 1.645631409729882e-05, "loss": 0.1369, "step": 77980 }, { "epoch": 1.82, "learning_rate": 1.6455434042852787e-05, "loss": 0.0657, "step": 77985 }, { "epoch": 1.82, "learning_rate": 1.645455398840675e-05, "loss": 0.0673, "step": 77990 }, { "epoch": 1.82, "learning_rate": 1.6453673933960714e-05, "loss": 0.1078, "step": 77995 }, { "epoch": 1.82, "learning_rate": 1.645279387951468e-05, "loss": 0.2593, "step": 78000 }, { "epoch": 1.82, "learning_rate": 1.6451913825068647e-05, "loss": 0.1138, "step": 78005 }, { "epoch": 1.82, "learning_rate": 1.645103377062261e-05, "loss": 0.496, "step": 78010 }, { "epoch": 1.82, "learning_rate": 1.6450153716176574e-05, "loss": 0.4282, "step": 78015 }, { "epoch": 1.82, "learning_rate": 1.644927366173054e-05, "loss": 0.0448, "step": 78020 }, { "epoch": 1.82, "learning_rate": 1.6448393607284504e-05, "loss": 0.0194, "step": 78025 }, { "epoch": 1.82, "learning_rate": 1.644751355283847e-05, "loss": 0.1179, "step": 78030 }, { "epoch": 1.82, "learning_rate": 1.6446633498392434e-05, "loss": 0.0147, "step": 78035 }, { "epoch": 1.82, "learning_rate": 1.64457534439464e-05, "loss": 0.1218, "step": 78040 }, { "epoch": 1.82, "learning_rate": 1.6444873389500364e-05, "loss": 0.1277, "step": 78045 }, { "epoch": 1.82, "learning_rate": 1.644399333505433e-05, "loss": 0.1494, "step": 78050 }, { "epoch": 1.82, "learning_rate": 1.6443113280608294e-05, "loss": 0.2336, "step": 78055 }, { "epoch": 1.82, "learning_rate": 1.644223322616226e-05, "loss": 0.3703, "step": 78060 }, { "epoch": 1.82, "learning_rate": 1.6441353171716224e-05, "loss": 0.3647, "step": 78065 }, { "epoch": 1.82, "learning_rate": 1.644047311727019e-05, "loss": 0.057, "step": 78070 }, { "epoch": 1.82, "learning_rate": 1.6439593062824154e-05, "loss": 0.0397, "step": 78075 }, { "epoch": 1.82, "learning_rate": 1.643871300837812e-05, "loss": 0.2145, "step": 78080 }, { "epoch": 1.82, "learning_rate": 1.6437832953932084e-05, "loss": 0.1249, "step": 78085 }, { "epoch": 1.82, "learning_rate": 1.643695289948605e-05, "loss": 0.1063, "step": 78090 }, { "epoch": 1.82, "learning_rate": 1.643607284504001e-05, "loss": 0.0992, "step": 78095 }, { "epoch": 1.82, "learning_rate": 1.643519279059398e-05, "loss": 0.0701, "step": 78100 }, { "epoch": 1.82, "learning_rate": 1.6434312736147944e-05, "loss": 0.3116, "step": 78105 }, { "epoch": 1.82, "learning_rate": 1.643343268170191e-05, "loss": 0.396, "step": 78110 }, { "epoch": 1.82, "learning_rate": 1.643255262725587e-05, "loss": 0.1926, "step": 78115 }, { "epoch": 1.82, "learning_rate": 1.643167257280984e-05, "loss": 0.0573, "step": 78120 }, { "epoch": 1.82, "learning_rate": 1.6430792518363805e-05, "loss": 0.0433, "step": 78125 }, { "epoch": 1.82, "learning_rate": 1.6429912463917766e-05, "loss": 0.0242, "step": 78130 }, { "epoch": 1.82, "learning_rate": 1.642903240947173e-05, "loss": 0.0856, "step": 78135 }, { "epoch": 1.82, "learning_rate": 1.64281523550257e-05, "loss": 0.0808, "step": 78140 }, { "epoch": 1.82, "learning_rate": 1.6427272300579665e-05, "loss": 0.0317, "step": 78145 }, { "epoch": 1.82, "learning_rate": 1.6426392246133626e-05, "loss": 0.2809, "step": 78150 }, { "epoch": 1.82, "learning_rate": 1.642551219168759e-05, "loss": 0.0907, "step": 78155 }, { "epoch": 1.82, "learning_rate": 1.642463213724156e-05, "loss": 0.1741, "step": 78160 }, { "epoch": 1.82, "learning_rate": 1.642375208279552e-05, "loss": 0.3186, "step": 78165 }, { "epoch": 1.82, "learning_rate": 1.6422872028349486e-05, "loss": 0.0334, "step": 78170 }, { "epoch": 1.82, "learning_rate": 1.642199197390345e-05, "loss": 0.0324, "step": 78175 }, { "epoch": 1.82, "learning_rate": 1.642111191945742e-05, "loss": 0.0322, "step": 78180 }, { "epoch": 1.82, "learning_rate": 1.642023186501138e-05, "loss": 0.1114, "step": 78185 }, { "epoch": 1.82, "learning_rate": 1.6419351810565346e-05, "loss": 0.1402, "step": 78190 }, { "epoch": 1.82, "learning_rate": 1.6418471756119315e-05, "loss": 0.1654, "step": 78195 }, { "epoch": 1.82, "learning_rate": 1.6417591701673276e-05, "loss": 0.1186, "step": 78200 }, { "epoch": 1.82, "learning_rate": 1.641671164722724e-05, "loss": 0.0603, "step": 78205 }, { "epoch": 1.82, "learning_rate": 1.6415831592781206e-05, "loss": 0.257, "step": 78210 }, { "epoch": 1.82, "learning_rate": 1.6414951538335175e-05, "loss": 0.1907, "step": 78215 }, { "epoch": 1.82, "learning_rate": 1.6414071483889137e-05, "loss": 0.0047, "step": 78220 }, { "epoch": 1.83, "learning_rate": 1.64131914294431e-05, "loss": 0.0195, "step": 78225 }, { "epoch": 1.83, "learning_rate": 1.6412311374997067e-05, "loss": 0.0947, "step": 78230 }, { "epoch": 1.83, "learning_rate": 1.641143132055103e-05, "loss": 0.0152, "step": 78235 }, { "epoch": 1.83, "learning_rate": 1.6410551266104997e-05, "loss": 0.1181, "step": 78240 }, { "epoch": 1.83, "learning_rate": 1.640967121165896e-05, "loss": 0.0886, "step": 78245 }, { "epoch": 1.83, "learning_rate": 1.6408791157212927e-05, "loss": 0.2893, "step": 78250 }, { "epoch": 1.83, "learning_rate": 1.640791110276689e-05, "loss": 0.1462, "step": 78255 }, { "epoch": 1.83, "learning_rate": 1.6407031048320857e-05, "loss": 0.2103, "step": 78260 }, { "epoch": 1.83, "learning_rate": 1.6406150993874822e-05, "loss": 0.2139, "step": 78265 }, { "epoch": 1.83, "learning_rate": 1.6405270939428783e-05, "loss": 0.0414, "step": 78270 }, { "epoch": 1.83, "learning_rate": 1.6404390884982752e-05, "loss": 0.0325, "step": 78275 }, { "epoch": 1.83, "learning_rate": 1.6403510830536717e-05, "loss": 0.0352, "step": 78280 }, { "epoch": 1.83, "learning_rate": 1.6402630776090682e-05, "loss": 0.0313, "step": 78285 }, { "epoch": 1.83, "learning_rate": 1.6401750721644643e-05, "loss": 0.0544, "step": 78290 }, { "epoch": 1.83, "learning_rate": 1.6400870667198612e-05, "loss": 0.1102, "step": 78295 }, { "epoch": 1.83, "learning_rate": 1.6399990612752577e-05, "loss": 0.1869, "step": 78300 }, { "epoch": 1.83, "learning_rate": 1.6399110558306542e-05, "loss": 0.2194, "step": 78305 }, { "epoch": 1.83, "learning_rate": 1.6398230503860503e-05, "loss": 0.1661, "step": 78310 }, { "epoch": 1.83, "learning_rate": 1.6397350449414472e-05, "loss": 0.3934, "step": 78315 }, { "epoch": 1.83, "learning_rate": 1.6396470394968437e-05, "loss": 0.0797, "step": 78320 }, { "epoch": 1.83, "learning_rate": 1.63955903405224e-05, "loss": 0.0068, "step": 78325 }, { "epoch": 1.83, "learning_rate": 1.6394710286076367e-05, "loss": 0.1076, "step": 78330 }, { "epoch": 1.83, "learning_rate": 1.6393830231630332e-05, "loss": 0.1089, "step": 78335 }, { "epoch": 1.83, "learning_rate": 1.6392950177184297e-05, "loss": 0.1594, "step": 78340 }, { "epoch": 1.83, "learning_rate": 1.639207012273826e-05, "loss": 0.1336, "step": 78345 }, { "epoch": 1.83, "learning_rate": 1.6391190068292227e-05, "loss": 0.129, "step": 78350 }, { "epoch": 1.83, "learning_rate": 1.6390310013846192e-05, "loss": 0.3125, "step": 78355 }, { "epoch": 1.83, "learning_rate": 1.6389429959400154e-05, "loss": 0.2371, "step": 78360 }, { "epoch": 1.83, "learning_rate": 1.638854990495412e-05, "loss": 0.313, "step": 78365 }, { "epoch": 1.83, "learning_rate": 1.6387669850508087e-05, "loss": 0.0128, "step": 78370 }, { "epoch": 1.83, "learning_rate": 1.6386789796062052e-05, "loss": 0.1065, "step": 78375 }, { "epoch": 1.83, "learning_rate": 1.6385909741616014e-05, "loss": 0.017, "step": 78380 }, { "epoch": 1.83, "learning_rate": 1.638502968716998e-05, "loss": 0.0959, "step": 78385 }, { "epoch": 1.83, "learning_rate": 1.6384149632723947e-05, "loss": 0.0968, "step": 78390 }, { "epoch": 1.83, "learning_rate": 1.638326957827791e-05, "loss": 0.1653, "step": 78395 }, { "epoch": 1.83, "learning_rate": 1.6382389523831874e-05, "loss": 0.0816, "step": 78400 }, { "epoch": 1.83, "learning_rate": 1.638150946938584e-05, "loss": 0.0508, "step": 78405 }, { "epoch": 1.83, "learning_rate": 1.6380629414939807e-05, "loss": 0.2163, "step": 78410 }, { "epoch": 1.83, "learning_rate": 1.637974936049377e-05, "loss": 0.2619, "step": 78415 }, { "epoch": 1.83, "learning_rate": 1.6378869306047734e-05, "loss": 0.0028, "step": 78420 }, { "epoch": 1.83, "learning_rate": 1.63779892516017e-05, "loss": 0.0476, "step": 78425 }, { "epoch": 1.83, "learning_rate": 1.6377109197155664e-05, "loss": 0.0049, "step": 78430 }, { "epoch": 1.83, "learning_rate": 1.637622914270963e-05, "loss": 0.1433, "step": 78435 }, { "epoch": 1.83, "learning_rate": 1.6375349088263594e-05, "loss": 0.2124, "step": 78440 }, { "epoch": 1.83, "learning_rate": 1.637446903381756e-05, "loss": 0.1882, "step": 78445 }, { "epoch": 1.83, "learning_rate": 1.6373588979371524e-05, "loss": 0.1138, "step": 78450 }, { "epoch": 1.83, "learning_rate": 1.637270892492549e-05, "loss": 0.0894, "step": 78455 }, { "epoch": 1.83, "learning_rate": 1.6371828870479454e-05, "loss": 0.2825, "step": 78460 }, { "epoch": 1.83, "learning_rate": 1.637094881603342e-05, "loss": 0.2599, "step": 78465 }, { "epoch": 1.83, "learning_rate": 1.6370068761587384e-05, "loss": 0.0942, "step": 78470 }, { "epoch": 1.83, "learning_rate": 1.636918870714135e-05, "loss": 0.0916, "step": 78475 }, { "epoch": 1.83, "learning_rate": 1.6368308652695314e-05, "loss": 0.1548, "step": 78480 }, { "epoch": 1.83, "learning_rate": 1.636742859824928e-05, "loss": 0.2115, "step": 78485 }, { "epoch": 1.83, "learning_rate": 1.6366548543803244e-05, "loss": 0.1475, "step": 78490 }, { "epoch": 1.83, "learning_rate": 1.636566848935721e-05, "loss": 0.1986, "step": 78495 }, { "epoch": 1.83, "learning_rate": 1.636478843491117e-05, "loss": 0.1345, "step": 78500 }, { "epoch": 1.83, "learning_rate": 1.636390838046514e-05, "loss": 0.1855, "step": 78505 }, { "epoch": 1.83, "learning_rate": 1.6363028326019104e-05, "loss": 0.1772, "step": 78510 }, { "epoch": 1.83, "learning_rate": 1.636214827157307e-05, "loss": 0.247, "step": 78515 }, { "epoch": 1.83, "learning_rate": 1.636126821712703e-05, "loss": 0.1237, "step": 78520 }, { "epoch": 1.83, "learning_rate": 1.6360388162681e-05, "loss": 0.0359, "step": 78525 }, { "epoch": 1.83, "learning_rate": 1.6359508108234964e-05, "loss": 0.0897, "step": 78530 }, { "epoch": 1.83, "learning_rate": 1.6358628053788926e-05, "loss": 0.2498, "step": 78535 }, { "epoch": 1.83, "learning_rate": 1.635774799934289e-05, "loss": 0.0197, "step": 78540 }, { "epoch": 1.83, "learning_rate": 1.635686794489686e-05, "loss": 0.182, "step": 78545 }, { "epoch": 1.83, "learning_rate": 1.6355987890450824e-05, "loss": 0.2047, "step": 78550 }, { "epoch": 1.83, "learning_rate": 1.6355107836004786e-05, "loss": 0.2086, "step": 78555 }, { "epoch": 1.83, "learning_rate": 1.635422778155875e-05, "loss": 0.208, "step": 78560 }, { "epoch": 1.83, "learning_rate": 1.635334772711272e-05, "loss": 0.2554, "step": 78565 }, { "epoch": 1.83, "learning_rate": 1.635246767266668e-05, "loss": 0.0522, "step": 78570 }, { "epoch": 1.83, "learning_rate": 1.6351587618220646e-05, "loss": 0.1175, "step": 78575 }, { "epoch": 1.83, "learning_rate": 1.635070756377461e-05, "loss": 0.0327, "step": 78580 }, { "epoch": 1.83, "learning_rate": 1.634982750932858e-05, "loss": 0.0751, "step": 78585 }, { "epoch": 1.83, "learning_rate": 1.634894745488254e-05, "loss": 0.0597, "step": 78590 }, { "epoch": 1.83, "learning_rate": 1.6348067400436506e-05, "loss": 0.1986, "step": 78595 }, { "epoch": 1.83, "learning_rate": 1.6347187345990475e-05, "loss": 0.0992, "step": 78600 }, { "epoch": 1.83, "learning_rate": 1.634630729154444e-05, "loss": 0.1133, "step": 78605 }, { "epoch": 1.83, "learning_rate": 1.63454272370984e-05, "loss": 0.4278, "step": 78610 }, { "epoch": 1.83, "learning_rate": 1.6344547182652366e-05, "loss": 0.133, "step": 78615 }, { "epoch": 1.83, "learning_rate": 1.6343667128206335e-05, "loss": 0.061, "step": 78620 }, { "epoch": 1.83, "learning_rate": 1.6342787073760296e-05, "loss": 0.0555, "step": 78625 }, { "epoch": 1.83, "learning_rate": 1.634190701931426e-05, "loss": 0.1009, "step": 78630 }, { "epoch": 1.83, "learning_rate": 1.6341026964868226e-05, "loss": 0.1086, "step": 78635 }, { "epoch": 1.83, "learning_rate": 1.6340146910422195e-05, "loss": 0.101, "step": 78640 }, { "epoch": 1.83, "learning_rate": 1.6339266855976156e-05, "loss": 0.0717, "step": 78645 }, { "epoch": 1.83, "learning_rate": 1.633838680153012e-05, "loss": 0.1718, "step": 78650 }, { "epoch": 1.84, "learning_rate": 1.6337506747084087e-05, "loss": 0.1228, "step": 78655 }, { "epoch": 1.84, "learning_rate": 1.633662669263805e-05, "loss": 0.2195, "step": 78660 }, { "epoch": 1.84, "learning_rate": 1.6335746638192017e-05, "loss": 0.134, "step": 78665 }, { "epoch": 1.84, "learning_rate": 1.633486658374598e-05, "loss": 0.0465, "step": 78670 }, { "epoch": 1.84, "learning_rate": 1.6333986529299947e-05, "loss": 0.1433, "step": 78675 }, { "epoch": 1.84, "learning_rate": 1.633310647485391e-05, "loss": 0.1321, "step": 78680 }, { "epoch": 1.84, "learning_rate": 1.6332226420407877e-05, "loss": 0.0337, "step": 78685 }, { "epoch": 1.84, "learning_rate": 1.633134636596184e-05, "loss": 0.1999, "step": 78690 }, { "epoch": 1.84, "learning_rate": 1.6330466311515803e-05, "loss": 0.0941, "step": 78695 }, { "epoch": 1.84, "learning_rate": 1.6329586257069772e-05, "loss": 0.0845, "step": 78700 }, { "epoch": 1.84, "learning_rate": 1.6328706202623737e-05, "loss": 0.169, "step": 78705 }, { "epoch": 1.84, "learning_rate": 1.6327826148177702e-05, "loss": 0.3198, "step": 78710 }, { "epoch": 1.84, "learning_rate": 1.6326946093731663e-05, "loss": 0.2985, "step": 78715 }, { "epoch": 1.84, "learning_rate": 1.6326066039285632e-05, "loss": 0.0064, "step": 78720 }, { "epoch": 1.84, "learning_rate": 1.6325185984839597e-05, "loss": 0.1262, "step": 78725 }, { "epoch": 1.84, "learning_rate": 1.632430593039356e-05, "loss": 0.0352, "step": 78730 }, { "epoch": 1.84, "learning_rate": 1.6323425875947527e-05, "loss": 0.151, "step": 78735 }, { "epoch": 1.84, "learning_rate": 1.6322545821501492e-05, "loss": 0.1146, "step": 78740 }, { "epoch": 1.84, "learning_rate": 1.6321665767055457e-05, "loss": 0.1027, "step": 78745 }, { "epoch": 1.84, "learning_rate": 1.632078571260942e-05, "loss": 0.1469, "step": 78750 }, { "epoch": 1.84, "learning_rate": 1.6319905658163387e-05, "loss": 0.1902, "step": 78755 }, { "epoch": 1.84, "learning_rate": 1.6319025603717352e-05, "loss": 0.2487, "step": 78760 }, { "epoch": 1.84, "learning_rate": 1.6318145549271314e-05, "loss": 0.3164, "step": 78765 }, { "epoch": 1.84, "learning_rate": 1.631726549482528e-05, "loss": 0.0897, "step": 78770 }, { "epoch": 1.84, "learning_rate": 1.6316385440379247e-05, "loss": 0.0339, "step": 78775 }, { "epoch": 1.84, "learning_rate": 1.6315505385933212e-05, "loss": 0.0601, "step": 78780 }, { "epoch": 1.84, "learning_rate": 1.6314625331487174e-05, "loss": 0.0635, "step": 78785 }, { "epoch": 1.84, "learning_rate": 1.631374527704114e-05, "loss": 0.0788, "step": 78790 }, { "epoch": 1.84, "learning_rate": 1.6312865222595107e-05, "loss": 0.0846, "step": 78795 }, { "epoch": 1.84, "learning_rate": 1.631198516814907e-05, "loss": 0.1929, "step": 78800 }, { "epoch": 1.84, "learning_rate": 1.6311105113703034e-05, "loss": 0.2321, "step": 78805 }, { "epoch": 1.84, "learning_rate": 1.6310225059257e-05, "loss": 0.2993, "step": 78810 }, { "epoch": 1.84, "learning_rate": 1.6309345004810967e-05, "loss": 0.1664, "step": 78815 }, { "epoch": 1.84, "learning_rate": 1.630846495036493e-05, "loss": 0.0595, "step": 78820 }, { "epoch": 1.84, "learning_rate": 1.6307584895918894e-05, "loss": 0.0931, "step": 78825 }, { "epoch": 1.84, "learning_rate": 1.630670484147286e-05, "loss": 0.0612, "step": 78830 }, { "epoch": 1.84, "learning_rate": 1.6305824787026824e-05, "loss": 0.0492, "step": 78835 }, { "epoch": 1.84, "learning_rate": 1.630494473258079e-05, "loss": 0.0248, "step": 78840 }, { "epoch": 1.84, "learning_rate": 1.6304064678134754e-05, "loss": 0.0932, "step": 78845 }, { "epoch": 1.84, "learning_rate": 1.630318462368872e-05, "loss": 0.1121, "step": 78850 }, { "epoch": 1.84, "learning_rate": 1.6302304569242684e-05, "loss": 0.2487, "step": 78855 }, { "epoch": 1.84, "learning_rate": 1.630142451479665e-05, "loss": 0.3318, "step": 78860 }, { "epoch": 1.84, "learning_rate": 1.6300544460350614e-05, "loss": 0.1854, "step": 78865 }, { "epoch": 1.84, "learning_rate": 1.629966440590458e-05, "loss": 0.1244, "step": 78870 }, { "epoch": 1.84, "learning_rate": 1.6298784351458544e-05, "loss": 0.0388, "step": 78875 }, { "epoch": 1.84, "learning_rate": 1.629790429701251e-05, "loss": 0.0316, "step": 78880 }, { "epoch": 1.84, "learning_rate": 1.6297024242566474e-05, "loss": 0.0474, "step": 78885 }, { "epoch": 1.84, "learning_rate": 1.629614418812044e-05, "loss": 0.0268, "step": 78890 }, { "epoch": 1.84, "learning_rate": 1.6295264133674404e-05, "loss": 0.0821, "step": 78895 }, { "epoch": 1.84, "learning_rate": 1.629438407922837e-05, "loss": 0.1907, "step": 78900 }, { "epoch": 1.84, "learning_rate": 1.6293504024782334e-05, "loss": 0.1387, "step": 78905 }, { "epoch": 1.84, "learning_rate": 1.62926239703363e-05, "loss": 0.3901, "step": 78910 }, { "epoch": 1.84, "learning_rate": 1.6291743915890264e-05, "loss": 0.2539, "step": 78915 }, { "epoch": 1.84, "learning_rate": 1.629086386144423e-05, "loss": 0.115, "step": 78920 }, { "epoch": 1.84, "learning_rate": 1.628998380699819e-05, "loss": 0.0278, "step": 78925 }, { "epoch": 1.84, "learning_rate": 1.628910375255216e-05, "loss": 0.0332, "step": 78930 }, { "epoch": 1.84, "learning_rate": 1.6288223698106124e-05, "loss": 0.1772, "step": 78935 }, { "epoch": 1.84, "learning_rate": 1.628734364366009e-05, "loss": 0.0619, "step": 78940 }, { "epoch": 1.84, "learning_rate": 1.628646358921405e-05, "loss": 0.1163, "step": 78945 }, { "epoch": 1.84, "learning_rate": 1.628558353476802e-05, "loss": 0.2472, "step": 78950 }, { "epoch": 1.84, "learning_rate": 1.6284703480321984e-05, "loss": 0.0717, "step": 78955 }, { "epoch": 1.84, "learning_rate": 1.6283823425875946e-05, "loss": 0.3349, "step": 78960 }, { "epoch": 1.84, "learning_rate": 1.628294337142991e-05, "loss": 0.1542, "step": 78965 }, { "epoch": 1.84, "learning_rate": 1.628206331698388e-05, "loss": 0.0498, "step": 78970 }, { "epoch": 1.84, "learning_rate": 1.6281183262537844e-05, "loss": 0.0741, "step": 78975 }, { "epoch": 1.84, "learning_rate": 1.6280303208091806e-05, "loss": 0.0829, "step": 78980 }, { "epoch": 1.84, "learning_rate": 1.627942315364577e-05, "loss": 0.2759, "step": 78985 }, { "epoch": 1.84, "learning_rate": 1.627854309919974e-05, "loss": 0.1949, "step": 78990 }, { "epoch": 1.84, "learning_rate": 1.62776630447537e-05, "loss": 0.0418, "step": 78995 }, { "epoch": 1.84, "learning_rate": 1.6276782990307666e-05, "loss": 0.2271, "step": 79000 }, { "epoch": 1.84, "learning_rate": 1.6275902935861635e-05, "loss": 0.1107, "step": 79005 }, { "epoch": 1.84, "learning_rate": 1.62750228814156e-05, "loss": 0.0822, "step": 79010 }, { "epoch": 1.84, "learning_rate": 1.627414282696956e-05, "loss": 0.2126, "step": 79015 }, { "epoch": 1.84, "learning_rate": 1.6273262772523526e-05, "loss": 0.2805, "step": 79020 }, { "epoch": 1.84, "learning_rate": 1.6272382718077495e-05, "loss": 0.0063, "step": 79025 }, { "epoch": 1.84, "learning_rate": 1.6271502663631456e-05, "loss": 0.0873, "step": 79030 }, { "epoch": 1.84, "learning_rate": 1.627062260918542e-05, "loss": 0.0168, "step": 79035 }, { "epoch": 1.84, "learning_rate": 1.6269742554739386e-05, "loss": 0.135, "step": 79040 }, { "epoch": 1.84, "learning_rate": 1.6268862500293355e-05, "loss": 0.0843, "step": 79045 }, { "epoch": 1.84, "learning_rate": 1.6267982445847316e-05, "loss": 0.1419, "step": 79050 }, { "epoch": 1.84, "learning_rate": 1.626710239140128e-05, "loss": 0.238, "step": 79055 }, { "epoch": 1.84, "learning_rate": 1.6266222336955246e-05, "loss": 0.4387, "step": 79060 }, { "epoch": 1.84, "learning_rate": 1.626534228250921e-05, "loss": 0.4766, "step": 79065 }, { "epoch": 1.84, "learning_rate": 1.6264462228063176e-05, "loss": 0.0824, "step": 79070 }, { "epoch": 1.84, "learning_rate": 1.626358217361714e-05, "loss": 0.1273, "step": 79075 }, { "epoch": 1.85, "learning_rate": 1.6262702119171106e-05, "loss": 0.071, "step": 79080 }, { "epoch": 1.85, "learning_rate": 1.626182206472507e-05, "loss": 0.0599, "step": 79085 }, { "epoch": 1.85, "learning_rate": 1.6260942010279037e-05, "loss": 0.0891, "step": 79090 }, { "epoch": 1.85, "learning_rate": 1.6260061955833e-05, "loss": 0.1343, "step": 79095 }, { "epoch": 1.85, "learning_rate": 1.6259181901386963e-05, "loss": 0.1037, "step": 79100 }, { "epoch": 1.85, "learning_rate": 1.625830184694093e-05, "loss": 0.1602, "step": 79105 }, { "epoch": 1.85, "learning_rate": 1.6257421792494897e-05, "loss": 0.2448, "step": 79110 }, { "epoch": 1.85, "learning_rate": 1.625654173804886e-05, "loss": 0.2489, "step": 79115 }, { "epoch": 1.85, "learning_rate": 1.6255661683602823e-05, "loss": 0.0189, "step": 79120 }, { "epoch": 1.85, "learning_rate": 1.625478162915679e-05, "loss": 0.0382, "step": 79125 }, { "epoch": 1.85, "learning_rate": 1.6253901574710757e-05, "loss": 0.0256, "step": 79130 }, { "epoch": 1.85, "learning_rate": 1.625302152026472e-05, "loss": 0.2398, "step": 79135 }, { "epoch": 1.85, "learning_rate": 1.6252141465818687e-05, "loss": 0.0253, "step": 79140 }, { "epoch": 1.85, "learning_rate": 1.6251261411372652e-05, "loss": 0.093, "step": 79145 }, { "epoch": 1.85, "learning_rate": 1.6250381356926617e-05, "loss": 0.1303, "step": 79150 }, { "epoch": 1.85, "learning_rate": 1.624950130248058e-05, "loss": 0.2933, "step": 79155 }, { "epoch": 1.85, "learning_rate": 1.6248621248034547e-05, "loss": 0.2659, "step": 79160 }, { "epoch": 1.85, "learning_rate": 1.6247741193588512e-05, "loss": 0.128, "step": 79165 }, { "epoch": 1.85, "learning_rate": 1.6246861139142473e-05, "loss": 0.0245, "step": 79170 }, { "epoch": 1.85, "learning_rate": 1.624598108469644e-05, "loss": 0.0894, "step": 79175 }, { "epoch": 1.85, "learning_rate": 1.6245101030250407e-05, "loss": 0.0469, "step": 79180 }, { "epoch": 1.85, "learning_rate": 1.6244220975804372e-05, "loss": 0.0331, "step": 79185 }, { "epoch": 1.85, "learning_rate": 1.6243340921358334e-05, "loss": 0.0349, "step": 79190 }, { "epoch": 1.85, "learning_rate": 1.62424608669123e-05, "loss": 0.0808, "step": 79195 }, { "epoch": 1.85, "learning_rate": 1.6241580812466267e-05, "loss": 0.4287, "step": 79200 }, { "epoch": 1.85, "learning_rate": 1.6240700758020232e-05, "loss": 0.1108, "step": 79205 }, { "epoch": 1.85, "learning_rate": 1.6239820703574194e-05, "loss": 0.5069, "step": 79210 }, { "epoch": 1.85, "learning_rate": 1.623894064912816e-05, "loss": 0.2012, "step": 79215 }, { "epoch": 1.85, "learning_rate": 1.6238060594682127e-05, "loss": 0.1833, "step": 79220 }, { "epoch": 1.85, "learning_rate": 1.623718054023609e-05, "loss": 0.0051, "step": 79225 }, { "epoch": 1.85, "learning_rate": 1.6236300485790054e-05, "loss": 0.0731, "step": 79230 }, { "epoch": 1.85, "learning_rate": 1.623542043134402e-05, "loss": 0.1806, "step": 79235 }, { "epoch": 1.85, "learning_rate": 1.6234540376897987e-05, "loss": 0.1083, "step": 79240 }, { "epoch": 1.85, "learning_rate": 1.623366032245195e-05, "loss": 0.1405, "step": 79245 }, { "epoch": 1.85, "learning_rate": 1.6232780268005914e-05, "loss": 0.2659, "step": 79250 }, { "epoch": 1.85, "learning_rate": 1.623190021355988e-05, "loss": 0.2371, "step": 79255 }, { "epoch": 1.85, "learning_rate": 1.6231020159113844e-05, "loss": 0.485, "step": 79260 }, { "epoch": 1.85, "learning_rate": 1.623014010466781e-05, "loss": 0.1761, "step": 79265 }, { "epoch": 1.85, "learning_rate": 1.6229260050221774e-05, "loss": 0.0289, "step": 79270 }, { "epoch": 1.85, "learning_rate": 1.6228379995775742e-05, "loss": 0.0249, "step": 79275 }, { "epoch": 1.85, "learning_rate": 1.6227499941329704e-05, "loss": 0.0603, "step": 79280 }, { "epoch": 1.85, "learning_rate": 1.622661988688367e-05, "loss": 0.0809, "step": 79285 }, { "epoch": 1.85, "learning_rate": 1.6225739832437634e-05, "loss": 0.0887, "step": 79290 }, { "epoch": 1.85, "learning_rate": 1.62248597779916e-05, "loss": 0.0469, "step": 79295 }, { "epoch": 1.85, "learning_rate": 1.6223979723545564e-05, "loss": 0.1811, "step": 79300 }, { "epoch": 1.85, "learning_rate": 1.622309966909953e-05, "loss": 0.1622, "step": 79305 }, { "epoch": 1.85, "learning_rate": 1.6222219614653494e-05, "loss": 0.1026, "step": 79310 }, { "epoch": 1.85, "learning_rate": 1.622133956020746e-05, "loss": 0.2643, "step": 79315 }, { "epoch": 1.85, "learning_rate": 1.6220459505761424e-05, "loss": 0.0277, "step": 79320 }, { "epoch": 1.85, "learning_rate": 1.621957945131539e-05, "loss": 0.0314, "step": 79325 }, { "epoch": 1.85, "learning_rate": 1.621869939686935e-05, "loss": 0.041, "step": 79330 }, { "epoch": 1.85, "learning_rate": 1.621781934242332e-05, "loss": 0.2409, "step": 79335 }, { "epoch": 1.85, "learning_rate": 1.6216939287977284e-05, "loss": 0.2235, "step": 79340 }, { "epoch": 1.85, "learning_rate": 1.621605923353125e-05, "loss": 0.1216, "step": 79345 }, { "epoch": 1.85, "learning_rate": 1.621517917908521e-05, "loss": 0.073, "step": 79350 }, { "epoch": 1.85, "learning_rate": 1.621429912463918e-05, "loss": 0.25, "step": 79355 }, { "epoch": 1.85, "learning_rate": 1.6213419070193144e-05, "loss": 0.4415, "step": 79360 }, { "epoch": 1.85, "learning_rate": 1.6212539015747106e-05, "loss": 0.1902, "step": 79365 }, { "epoch": 1.85, "learning_rate": 1.621165896130107e-05, "loss": 0.0171, "step": 79370 }, { "epoch": 1.85, "learning_rate": 1.621077890685504e-05, "loss": 0.0816, "step": 79375 }, { "epoch": 1.85, "learning_rate": 1.6209898852409004e-05, "loss": 0.0756, "step": 79380 }, { "epoch": 1.85, "learning_rate": 1.6209018797962966e-05, "loss": 0.0599, "step": 79385 }, { "epoch": 1.85, "learning_rate": 1.620813874351693e-05, "loss": 0.0293, "step": 79390 }, { "epoch": 1.85, "learning_rate": 1.62072586890709e-05, "loss": 0.1153, "step": 79395 }, { "epoch": 1.85, "learning_rate": 1.620637863462486e-05, "loss": 0.1097, "step": 79400 }, { "epoch": 1.85, "learning_rate": 1.6205498580178826e-05, "loss": 0.2558, "step": 79405 }, { "epoch": 1.85, "learning_rate": 1.6204618525732794e-05, "loss": 0.3608, "step": 79410 }, { "epoch": 1.85, "learning_rate": 1.620373847128676e-05, "loss": 0.2351, "step": 79415 }, { "epoch": 1.85, "learning_rate": 1.620285841684072e-05, "loss": 0.049, "step": 79420 }, { "epoch": 1.85, "learning_rate": 1.6201978362394686e-05, "loss": 0.0103, "step": 79425 }, { "epoch": 1.85, "learning_rate": 1.6201098307948655e-05, "loss": 0.0135, "step": 79430 }, { "epoch": 1.85, "learning_rate": 1.6200218253502616e-05, "loss": 0.0335, "step": 79435 }, { "epoch": 1.85, "learning_rate": 1.619933819905658e-05, "loss": 0.0415, "step": 79440 }, { "epoch": 1.85, "learning_rate": 1.6198458144610546e-05, "loss": 0.0803, "step": 79445 }, { "epoch": 1.85, "learning_rate": 1.6197578090164515e-05, "loss": 0.1797, "step": 79450 }, { "epoch": 1.85, "learning_rate": 1.6196698035718476e-05, "loss": 0.1766, "step": 79455 }, { "epoch": 1.85, "learning_rate": 1.619581798127244e-05, "loss": 0.2958, "step": 79460 }, { "epoch": 1.85, "learning_rate": 1.6194937926826406e-05, "loss": 0.2661, "step": 79465 }, { "epoch": 1.85, "learning_rate": 1.619405787238037e-05, "loss": 0.0444, "step": 79470 }, { "epoch": 1.85, "learning_rate": 1.6193177817934336e-05, "loss": 0.0969, "step": 79475 }, { "epoch": 1.85, "learning_rate": 1.61922977634883e-05, "loss": 0.0992, "step": 79480 }, { "epoch": 1.85, "learning_rate": 1.6191417709042266e-05, "loss": 0.1106, "step": 79485 }, { "epoch": 1.85, "learning_rate": 1.619053765459623e-05, "loss": 0.1153, "step": 79490 }, { "epoch": 1.85, "learning_rate": 1.6189657600150196e-05, "loss": 0.0772, "step": 79495 }, { "epoch": 1.85, "learning_rate": 1.618877754570416e-05, "loss": 0.2272, "step": 79500 }, { "epoch": 1.85, "learning_rate": 1.6187897491258126e-05, "loss": 0.1253, "step": 79505 }, { "epoch": 1.86, "learning_rate": 1.618701743681209e-05, "loss": 0.3848, "step": 79510 }, { "epoch": 1.86, "learning_rate": 1.6186137382366057e-05, "loss": 0.1232, "step": 79515 }, { "epoch": 1.86, "learning_rate": 1.618525732792002e-05, "loss": 0.0513, "step": 79520 }, { "epoch": 1.86, "learning_rate": 1.6184377273473983e-05, "loss": 0.041, "step": 79525 }, { "epoch": 1.86, "learning_rate": 1.618349721902795e-05, "loss": 0.0669, "step": 79530 }, { "epoch": 1.86, "learning_rate": 1.6182617164581917e-05, "loss": 0.0192, "step": 79535 }, { "epoch": 1.86, "learning_rate": 1.618173711013588e-05, "loss": 0.1089, "step": 79540 }, { "epoch": 1.86, "learning_rate": 1.6180857055689847e-05, "loss": 0.0501, "step": 79545 }, { "epoch": 1.86, "learning_rate": 1.617997700124381e-05, "loss": 0.0202, "step": 79550 }, { "epoch": 1.86, "learning_rate": 1.6179096946797777e-05, "loss": 0.2827, "step": 79555 }, { "epoch": 1.86, "learning_rate": 1.6178216892351738e-05, "loss": 0.1646, "step": 79560 }, { "epoch": 1.86, "learning_rate": 1.6177336837905707e-05, "loss": 0.42, "step": 79565 }, { "epoch": 1.86, "learning_rate": 1.6176456783459672e-05, "loss": 0.1382, "step": 79570 }, { "epoch": 1.86, "learning_rate": 1.6175576729013637e-05, "loss": 0.0116, "step": 79575 }, { "epoch": 1.86, "learning_rate": 1.61746966745676e-05, "loss": 0.1187, "step": 79580 }, { "epoch": 1.86, "learning_rate": 1.6173816620121567e-05, "loss": 0.1247, "step": 79585 }, { "epoch": 1.86, "learning_rate": 1.6172936565675532e-05, "loss": 0.0678, "step": 79590 }, { "epoch": 1.86, "learning_rate": 1.6172056511229493e-05, "loss": 0.1734, "step": 79595 }, { "epoch": 1.86, "learning_rate": 1.617117645678346e-05, "loss": 0.2045, "step": 79600 }, { "epoch": 1.86, "learning_rate": 1.6170296402337427e-05, "loss": 0.264, "step": 79605 }, { "epoch": 1.86, "learning_rate": 1.6169416347891392e-05, "loss": 0.1862, "step": 79610 }, { "epoch": 1.86, "learning_rate": 1.6168536293445354e-05, "loss": 0.2806, "step": 79615 }, { "epoch": 1.86, "learning_rate": 1.616765623899932e-05, "loss": 0.0432, "step": 79620 }, { "epoch": 1.86, "learning_rate": 1.6166776184553287e-05, "loss": 0.0133, "step": 79625 }, { "epoch": 1.86, "learning_rate": 1.616589613010725e-05, "loss": 0.1146, "step": 79630 }, { "epoch": 1.86, "learning_rate": 1.6165016075661214e-05, "loss": 0.0484, "step": 79635 }, { "epoch": 1.86, "learning_rate": 1.616413602121518e-05, "loss": 0.1153, "step": 79640 }, { "epoch": 1.86, "learning_rate": 1.6163255966769147e-05, "loss": 0.1561, "step": 79645 }, { "epoch": 1.86, "learning_rate": 1.616237591232311e-05, "loss": 0.1059, "step": 79650 }, { "epoch": 1.86, "learning_rate": 1.6161495857877074e-05, "loss": 0.2602, "step": 79655 }, { "epoch": 1.86, "learning_rate": 1.616061580343104e-05, "loss": 0.4033, "step": 79660 }, { "epoch": 1.86, "learning_rate": 1.6159735748985004e-05, "loss": 0.2624, "step": 79665 }, { "epoch": 1.86, "learning_rate": 1.615885569453897e-05, "loss": 0.023, "step": 79670 }, { "epoch": 1.86, "learning_rate": 1.6157975640092934e-05, "loss": 0.0844, "step": 79675 }, { "epoch": 1.86, "learning_rate": 1.61570955856469e-05, "loss": 0.1025, "step": 79680 }, { "epoch": 1.86, "learning_rate": 1.6156215531200864e-05, "loss": 0.0609, "step": 79685 }, { "epoch": 1.86, "learning_rate": 1.615533547675483e-05, "loss": 0.0824, "step": 79690 }, { "epoch": 1.86, "learning_rate": 1.6154455422308794e-05, "loss": 0.14, "step": 79695 }, { "epoch": 1.86, "learning_rate": 1.615357536786276e-05, "loss": 0.1251, "step": 79700 }, { "epoch": 1.86, "learning_rate": 1.6152695313416724e-05, "loss": 0.1011, "step": 79705 }, { "epoch": 1.86, "learning_rate": 1.615181525897069e-05, "loss": 0.5816, "step": 79710 }, { "epoch": 1.86, "learning_rate": 1.6150935204524654e-05, "loss": 0.2413, "step": 79715 }, { "epoch": 1.86, "learning_rate": 1.615005515007862e-05, "loss": 0.0408, "step": 79720 }, { "epoch": 1.86, "learning_rate": 1.6149175095632584e-05, "loss": 0.0258, "step": 79725 }, { "epoch": 1.86, "learning_rate": 1.614829504118655e-05, "loss": 0.0455, "step": 79730 }, { "epoch": 1.86, "learning_rate": 1.614741498674051e-05, "loss": 0.0391, "step": 79735 }, { "epoch": 1.86, "learning_rate": 1.614653493229448e-05, "loss": 0.0524, "step": 79740 }, { "epoch": 1.86, "learning_rate": 1.6145654877848444e-05, "loss": 0.2066, "step": 79745 }, { "epoch": 1.86, "learning_rate": 1.614477482340241e-05, "loss": 0.1518, "step": 79750 }, { "epoch": 1.86, "learning_rate": 1.614389476895637e-05, "loss": 0.1557, "step": 79755 }, { "epoch": 1.86, "learning_rate": 1.614301471451034e-05, "loss": 0.2961, "step": 79760 }, { "epoch": 1.86, "learning_rate": 1.6142134660064304e-05, "loss": 0.1755, "step": 79765 }, { "epoch": 1.86, "learning_rate": 1.6141254605618266e-05, "loss": 0.0167, "step": 79770 }, { "epoch": 1.86, "learning_rate": 1.614037455117223e-05, "loss": 0.0348, "step": 79775 }, { "epoch": 1.86, "learning_rate": 1.61394944967262e-05, "loss": 0.0559, "step": 79780 }, { "epoch": 1.86, "learning_rate": 1.6138614442280164e-05, "loss": 0.0541, "step": 79785 }, { "epoch": 1.86, "learning_rate": 1.6137734387834126e-05, "loss": 0.0136, "step": 79790 }, { "epoch": 1.86, "learning_rate": 1.613685433338809e-05, "loss": 0.1315, "step": 79795 }, { "epoch": 1.86, "learning_rate": 1.613597427894206e-05, "loss": 0.1176, "step": 79800 }, { "epoch": 1.86, "learning_rate": 1.6135094224496024e-05, "loss": 0.3128, "step": 79805 }, { "epoch": 1.86, "learning_rate": 1.6134214170049986e-05, "loss": 0.3447, "step": 79810 }, { "epoch": 1.86, "learning_rate": 1.6133334115603954e-05, "loss": 0.2427, "step": 79815 }, { "epoch": 1.86, "learning_rate": 1.613245406115792e-05, "loss": 0.0823, "step": 79820 }, { "epoch": 1.86, "learning_rate": 1.613157400671188e-05, "loss": 0.0824, "step": 79825 }, { "epoch": 1.86, "learning_rate": 1.6130693952265846e-05, "loss": 0.0296, "step": 79830 }, { "epoch": 1.86, "learning_rate": 1.6129813897819814e-05, "loss": 0.0446, "step": 79835 }, { "epoch": 1.86, "learning_rate": 1.612893384337378e-05, "loss": 0.1586, "step": 79840 }, { "epoch": 1.86, "learning_rate": 1.612805378892774e-05, "loss": 0.0057, "step": 79845 }, { "epoch": 1.86, "learning_rate": 1.6127173734481706e-05, "loss": 0.1183, "step": 79850 }, { "epoch": 1.86, "learning_rate": 1.6126293680035675e-05, "loss": 0.1643, "step": 79855 }, { "epoch": 1.86, "learning_rate": 1.6125413625589636e-05, "loss": 0.2025, "step": 79860 }, { "epoch": 1.86, "learning_rate": 1.61245335711436e-05, "loss": 0.1554, "step": 79865 }, { "epoch": 1.86, "learning_rate": 1.6123653516697566e-05, "loss": 0.0923, "step": 79870 }, { "epoch": 1.86, "learning_rate": 1.6122773462251535e-05, "loss": 0.0308, "step": 79875 }, { "epoch": 1.86, "learning_rate": 1.6121893407805496e-05, "loss": 0.0379, "step": 79880 }, { "epoch": 1.86, "learning_rate": 1.612101335335946e-05, "loss": 0.109, "step": 79885 }, { "epoch": 1.86, "learning_rate": 1.6120133298913426e-05, "loss": 0.085, "step": 79890 }, { "epoch": 1.86, "learning_rate": 1.611925324446739e-05, "loss": 0.0621, "step": 79895 }, { "epoch": 1.86, "learning_rate": 1.6118373190021356e-05, "loss": 0.0526, "step": 79900 }, { "epoch": 1.86, "learning_rate": 1.611749313557532e-05, "loss": 0.0861, "step": 79905 }, { "epoch": 1.86, "learning_rate": 1.6116613081129286e-05, "loss": 0.4734, "step": 79910 }, { "epoch": 1.86, "learning_rate": 1.611573302668325e-05, "loss": 0.2154, "step": 79915 }, { "epoch": 1.86, "learning_rate": 1.6114852972237216e-05, "loss": 0.0993, "step": 79920 }, { "epoch": 1.86, "learning_rate": 1.611397291779118e-05, "loss": 0.003, "step": 79925 }, { "epoch": 1.86, "learning_rate": 1.6113092863345143e-05, "loss": 0.0772, "step": 79930 }, { "epoch": 1.86, "learning_rate": 1.611221280889911e-05, "loss": 0.0938, "step": 79935 }, { "epoch": 1.87, "learning_rate": 1.6111332754453076e-05, "loss": 0.2013, "step": 79940 }, { "epoch": 1.87, "learning_rate": 1.611045270000704e-05, "loss": 0.1677, "step": 79945 }, { "epoch": 1.87, "learning_rate": 1.6109572645561007e-05, "loss": 0.14, "step": 79950 }, { "epoch": 1.87, "learning_rate": 1.610869259111497e-05, "loss": 0.1934, "step": 79955 }, { "epoch": 1.87, "learning_rate": 1.6107812536668937e-05, "loss": 0.1929, "step": 79960 }, { "epoch": 1.87, "learning_rate": 1.6106932482222898e-05, "loss": 0.2006, "step": 79965 }, { "epoch": 1.87, "learning_rate": 1.6106052427776867e-05, "loss": 0.0427, "step": 79970 }, { "epoch": 1.87, "learning_rate": 1.610517237333083e-05, "loss": 0.0291, "step": 79975 }, { "epoch": 1.87, "learning_rate": 1.6104292318884797e-05, "loss": 0.0585, "step": 79980 }, { "epoch": 1.87, "learning_rate": 1.6103412264438758e-05, "loss": 0.0594, "step": 79985 }, { "epoch": 1.87, "learning_rate": 1.6102532209992727e-05, "loss": 0.0511, "step": 79990 }, { "epoch": 1.87, "learning_rate": 1.6101652155546692e-05, "loss": 0.0673, "step": 79995 }, { "epoch": 1.87, "learning_rate": 1.6100772101100653e-05, "loss": 0.3042, "step": 80000 }, { "epoch": 1.87, "learning_rate": 1.609989204665462e-05, "loss": 0.2588, "step": 80005 }, { "epoch": 1.87, "learning_rate": 1.6099011992208587e-05, "loss": 0.3003, "step": 80010 }, { "epoch": 1.87, "learning_rate": 1.6098131937762552e-05, "loss": 0.4484, "step": 80015 }, { "epoch": 1.87, "learning_rate": 1.6097251883316513e-05, "loss": 0.0877, "step": 80020 }, { "epoch": 1.87, "learning_rate": 1.609637182887048e-05, "loss": 0.0217, "step": 80025 }, { "epoch": 1.87, "learning_rate": 1.6095491774424447e-05, "loss": 0.0564, "step": 80030 }, { "epoch": 1.87, "learning_rate": 1.609461171997841e-05, "loss": 0.0616, "step": 80035 }, { "epoch": 1.87, "learning_rate": 1.6093731665532373e-05, "loss": 0.1037, "step": 80040 }, { "epoch": 1.87, "learning_rate": 1.609285161108634e-05, "loss": 0.1702, "step": 80045 }, { "epoch": 1.87, "learning_rate": 1.6091971556640307e-05, "loss": 0.1175, "step": 80050 }, { "epoch": 1.87, "learning_rate": 1.609109150219427e-05, "loss": 0.0802, "step": 80055 }, { "epoch": 1.87, "learning_rate": 1.6090211447748234e-05, "loss": 0.3026, "step": 80060 }, { "epoch": 1.87, "learning_rate": 1.60893313933022e-05, "loss": 0.1974, "step": 80065 }, { "epoch": 1.87, "learning_rate": 1.6088451338856164e-05, "loss": 0.0249, "step": 80070 }, { "epoch": 1.87, "learning_rate": 1.608757128441013e-05, "loss": 0.043, "step": 80075 }, { "epoch": 1.87, "learning_rate": 1.6086691229964094e-05, "loss": 0.0214, "step": 80080 }, { "epoch": 1.87, "learning_rate": 1.608581117551806e-05, "loss": 0.1546, "step": 80085 }, { "epoch": 1.87, "learning_rate": 1.6084931121072024e-05, "loss": 0.0768, "step": 80090 }, { "epoch": 1.87, "learning_rate": 1.608405106662599e-05, "loss": 0.0963, "step": 80095 }, { "epoch": 1.87, "learning_rate": 1.6083171012179954e-05, "loss": 0.0775, "step": 80100 }, { "epoch": 1.87, "learning_rate": 1.608229095773392e-05, "loss": 0.0708, "step": 80105 }, { "epoch": 1.87, "learning_rate": 1.6081410903287884e-05, "loss": 0.1653, "step": 80110 }, { "epoch": 1.87, "learning_rate": 1.608053084884185e-05, "loss": 0.3877, "step": 80115 }, { "epoch": 1.87, "learning_rate": 1.6079650794395814e-05, "loss": 0.1632, "step": 80120 }, { "epoch": 1.87, "learning_rate": 1.607877073994978e-05, "loss": 0.0208, "step": 80125 }, { "epoch": 1.87, "learning_rate": 1.6077890685503744e-05, "loss": 0.0829, "step": 80130 }, { "epoch": 1.87, "learning_rate": 1.607701063105771e-05, "loss": 0.1847, "step": 80135 }, { "epoch": 1.87, "learning_rate": 1.6076130576611674e-05, "loss": 0.0567, "step": 80140 }, { "epoch": 1.87, "learning_rate": 1.607525052216564e-05, "loss": 0.0695, "step": 80145 }, { "epoch": 1.87, "learning_rate": 1.6074370467719604e-05, "loss": 0.1844, "step": 80150 }, { "epoch": 1.87, "learning_rate": 1.607349041327357e-05, "loss": 0.2439, "step": 80155 }, { "epoch": 1.87, "learning_rate": 1.607261035882753e-05, "loss": 0.2559, "step": 80160 }, { "epoch": 1.87, "learning_rate": 1.60717303043815e-05, "loss": 0.2111, "step": 80165 }, { "epoch": 1.87, "learning_rate": 1.6070850249935464e-05, "loss": 0.0661, "step": 80170 }, { "epoch": 1.87, "learning_rate": 1.606997019548943e-05, "loss": 0.0711, "step": 80175 }, { "epoch": 1.87, "learning_rate": 1.606909014104339e-05, "loss": 0.0783, "step": 80180 }, { "epoch": 1.87, "learning_rate": 1.606821008659736e-05, "loss": 0.0029, "step": 80185 }, { "epoch": 1.87, "learning_rate": 1.6067330032151324e-05, "loss": 0.1114, "step": 80190 }, { "epoch": 1.87, "learning_rate": 1.6066449977705286e-05, "loss": 0.0865, "step": 80195 }, { "epoch": 1.87, "learning_rate": 1.606556992325925e-05, "loss": 0.1569, "step": 80200 }, { "epoch": 1.87, "learning_rate": 1.606468986881322e-05, "loss": 0.2332, "step": 80205 }, { "epoch": 1.87, "learning_rate": 1.6063809814367184e-05, "loss": 0.2265, "step": 80210 }, { "epoch": 1.87, "learning_rate": 1.6062929759921146e-05, "loss": 0.1063, "step": 80215 }, { "epoch": 1.87, "learning_rate": 1.606204970547511e-05, "loss": 0.0131, "step": 80220 }, { "epoch": 1.87, "learning_rate": 1.606116965102908e-05, "loss": 0.013, "step": 80225 }, { "epoch": 1.87, "learning_rate": 1.606028959658304e-05, "loss": 0.0816, "step": 80230 }, { "epoch": 1.87, "learning_rate": 1.6059409542137006e-05, "loss": 0.0728, "step": 80235 }, { "epoch": 1.87, "learning_rate": 1.6058529487690974e-05, "loss": 0.0945, "step": 80240 }, { "epoch": 1.87, "learning_rate": 1.605764943324494e-05, "loss": 0.2101, "step": 80245 }, { "epoch": 1.87, "learning_rate": 1.60567693787989e-05, "loss": 0.1578, "step": 80250 }, { "epoch": 1.87, "learning_rate": 1.6055889324352866e-05, "loss": 0.0814, "step": 80255 }, { "epoch": 1.87, "learning_rate": 1.6055009269906834e-05, "loss": 0.1141, "step": 80260 }, { "epoch": 1.87, "learning_rate": 1.6054129215460796e-05, "loss": 0.2271, "step": 80265 }, { "epoch": 1.87, "learning_rate": 1.605324916101476e-05, "loss": 0.0055, "step": 80270 }, { "epoch": 1.87, "learning_rate": 1.6052369106568726e-05, "loss": 0.0073, "step": 80275 }, { "epoch": 1.87, "learning_rate": 1.6051489052122694e-05, "loss": 0.1335, "step": 80280 }, { "epoch": 1.87, "learning_rate": 1.6050608997676656e-05, "loss": 0.0682, "step": 80285 }, { "epoch": 1.87, "learning_rate": 1.604972894323062e-05, "loss": 0.0965, "step": 80290 }, { "epoch": 1.87, "learning_rate": 1.6048848888784586e-05, "loss": 0.0479, "step": 80295 }, { "epoch": 1.87, "learning_rate": 1.604796883433855e-05, "loss": 0.0437, "step": 80300 }, { "epoch": 1.87, "learning_rate": 1.6047088779892516e-05, "loss": 0.1607, "step": 80305 }, { "epoch": 1.87, "learning_rate": 1.604620872544648e-05, "loss": 0.4025, "step": 80310 }, { "epoch": 1.87, "learning_rate": 1.6045328671000446e-05, "loss": 0.1191, "step": 80315 }, { "epoch": 1.87, "learning_rate": 1.604444861655441e-05, "loss": 0.0492, "step": 80320 }, { "epoch": 1.87, "learning_rate": 1.6043568562108376e-05, "loss": 0.0804, "step": 80325 }, { "epoch": 1.87, "learning_rate": 1.604268850766234e-05, "loss": 0.0609, "step": 80330 }, { "epoch": 1.87, "learning_rate": 1.6041808453216303e-05, "loss": 0.0343, "step": 80335 }, { "epoch": 1.87, "learning_rate": 1.604092839877027e-05, "loss": 0.1218, "step": 80340 }, { "epoch": 1.87, "learning_rate": 1.6040048344324236e-05, "loss": 0.0755, "step": 80345 }, { "epoch": 1.87, "learning_rate": 1.60391682898782e-05, "loss": 0.0698, "step": 80350 }, { "epoch": 1.87, "learning_rate": 1.6038288235432166e-05, "loss": 0.2427, "step": 80355 }, { "epoch": 1.87, "learning_rate": 1.603740818098613e-05, "loss": 0.2301, "step": 80360 }, { "epoch": 1.88, "learning_rate": 1.6036528126540096e-05, "loss": 0.3279, "step": 80365 }, { "epoch": 1.88, "learning_rate": 1.6035648072094058e-05, "loss": 0.0763, "step": 80370 }, { "epoch": 1.88, "learning_rate": 1.6034768017648026e-05, "loss": 0.0391, "step": 80375 }, { "epoch": 1.88, "learning_rate": 1.603388796320199e-05, "loss": 0.0814, "step": 80380 }, { "epoch": 1.88, "learning_rate": 1.6033007908755957e-05, "loss": 0.0131, "step": 80385 }, { "epoch": 1.88, "learning_rate": 1.6032127854309918e-05, "loss": 0.0672, "step": 80390 }, { "epoch": 1.88, "learning_rate": 1.6031247799863887e-05, "loss": 0.2004, "step": 80395 }, { "epoch": 1.88, "learning_rate": 1.603036774541785e-05, "loss": 0.1189, "step": 80400 }, { "epoch": 1.88, "learning_rate": 1.6029487690971813e-05, "loss": 0.2778, "step": 80405 }, { "epoch": 1.88, "learning_rate": 1.6028607636525778e-05, "loss": 0.3217, "step": 80410 }, { "epoch": 1.88, "learning_rate": 1.6027727582079747e-05, "loss": 0.1585, "step": 80415 }, { "epoch": 1.88, "learning_rate": 1.602684752763371e-05, "loss": 0.0526, "step": 80420 }, { "epoch": 1.88, "learning_rate": 1.6025967473187673e-05, "loss": 0.0158, "step": 80425 }, { "epoch": 1.88, "learning_rate": 1.602508741874164e-05, "loss": 0.0816, "step": 80430 }, { "epoch": 1.88, "learning_rate": 1.6024207364295607e-05, "loss": 0.0532, "step": 80435 }, { "epoch": 1.88, "learning_rate": 1.6023327309849572e-05, "loss": 0.0847, "step": 80440 }, { "epoch": 1.88, "learning_rate": 1.6022447255403533e-05, "loss": 0.2646, "step": 80445 }, { "epoch": 1.88, "learning_rate": 1.60215672009575e-05, "loss": 0.1558, "step": 80450 }, { "epoch": 1.88, "learning_rate": 1.6020687146511467e-05, "loss": 0.164, "step": 80455 }, { "epoch": 1.88, "learning_rate": 1.601980709206543e-05, "loss": 0.4011, "step": 80460 }, { "epoch": 1.88, "learning_rate": 1.6018927037619393e-05, "loss": 0.3035, "step": 80465 }, { "epoch": 1.88, "learning_rate": 1.601804698317336e-05, "loss": 0.1056, "step": 80470 }, { "epoch": 1.88, "learning_rate": 1.6017166928727327e-05, "loss": 0.0703, "step": 80475 }, { "epoch": 1.88, "learning_rate": 1.601628687428129e-05, "loss": 0.0306, "step": 80480 }, { "epoch": 1.88, "learning_rate": 1.6015406819835254e-05, "loss": 0.0322, "step": 80485 }, { "epoch": 1.88, "learning_rate": 1.601452676538922e-05, "loss": 0.086, "step": 80490 }, { "epoch": 1.88, "learning_rate": 1.6013646710943184e-05, "loss": 0.1386, "step": 80495 }, { "epoch": 1.88, "learning_rate": 1.601276665649715e-05, "loss": 0.1247, "step": 80500 }, { "epoch": 1.88, "learning_rate": 1.6011886602051114e-05, "loss": 0.2694, "step": 80505 }, { "epoch": 1.88, "learning_rate": 1.6011006547605082e-05, "loss": 0.2123, "step": 80510 }, { "epoch": 1.88, "learning_rate": 1.6010126493159044e-05, "loss": 0.1646, "step": 80515 }, { "epoch": 1.88, "learning_rate": 1.600924643871301e-05, "loss": 0.0585, "step": 80520 }, { "epoch": 1.88, "learning_rate": 1.6008366384266974e-05, "loss": 0.0736, "step": 80525 }, { "epoch": 1.88, "learning_rate": 1.600748632982094e-05, "loss": 0.0568, "step": 80530 }, { "epoch": 1.88, "learning_rate": 1.6006606275374904e-05, "loss": 0.0082, "step": 80535 }, { "epoch": 1.88, "learning_rate": 1.600572622092887e-05, "loss": 0.0333, "step": 80540 }, { "epoch": 1.88, "learning_rate": 1.6004846166482834e-05, "loss": 0.0017, "step": 80545 }, { "epoch": 1.88, "learning_rate": 1.60039661120368e-05, "loss": 0.0855, "step": 80550 }, { "epoch": 1.88, "learning_rate": 1.6003086057590764e-05, "loss": 0.0571, "step": 80555 }, { "epoch": 1.88, "learning_rate": 1.600220600314473e-05, "loss": 0.1683, "step": 80560 }, { "epoch": 1.88, "learning_rate": 1.600132594869869e-05, "loss": 0.1789, "step": 80565 }, { "epoch": 1.88, "learning_rate": 1.600044589425266e-05, "loss": 0.0216, "step": 80570 }, { "epoch": 1.88, "learning_rate": 1.5999565839806624e-05, "loss": 0.0646, "step": 80575 }, { "epoch": 1.88, "learning_rate": 1.599868578536059e-05, "loss": 0.0586, "step": 80580 }, { "epoch": 1.88, "learning_rate": 1.599780573091455e-05, "loss": 0.0702, "step": 80585 }, { "epoch": 1.88, "learning_rate": 1.599692567646852e-05, "loss": 0.0631, "step": 80590 }, { "epoch": 1.88, "learning_rate": 1.5996045622022484e-05, "loss": 0.0773, "step": 80595 }, { "epoch": 1.88, "learning_rate": 1.5995165567576446e-05, "loss": 0.1915, "step": 80600 }, { "epoch": 1.88, "learning_rate": 1.599428551313041e-05, "loss": 0.2036, "step": 80605 }, { "epoch": 1.88, "learning_rate": 1.599340545868438e-05, "loss": 0.3546, "step": 80610 }, { "epoch": 1.88, "learning_rate": 1.5992525404238344e-05, "loss": 0.3159, "step": 80615 }, { "epoch": 1.88, "learning_rate": 1.5991645349792306e-05, "loss": 0.0162, "step": 80620 }, { "epoch": 1.88, "learning_rate": 1.599076529534627e-05, "loss": 0.0658, "step": 80625 }, { "epoch": 1.88, "learning_rate": 1.598988524090024e-05, "loss": 0.0603, "step": 80630 }, { "epoch": 1.88, "learning_rate": 1.59890051864542e-05, "loss": 0.0717, "step": 80635 }, { "epoch": 1.88, "learning_rate": 1.5988125132008166e-05, "loss": 0.0901, "step": 80640 }, { "epoch": 1.88, "learning_rate": 1.5987245077562134e-05, "loss": 0.1571, "step": 80645 }, { "epoch": 1.88, "learning_rate": 1.59863650231161e-05, "loss": 0.083, "step": 80650 }, { "epoch": 1.88, "learning_rate": 1.598548496867006e-05, "loss": 0.3501, "step": 80655 }, { "epoch": 1.88, "learning_rate": 1.5984604914224026e-05, "loss": 0.4651, "step": 80660 }, { "epoch": 1.88, "learning_rate": 1.5983724859777994e-05, "loss": 0.4055, "step": 80665 }, { "epoch": 1.88, "learning_rate": 1.5982844805331956e-05, "loss": 0.1176, "step": 80670 }, { "epoch": 1.88, "learning_rate": 1.598196475088592e-05, "loss": 0.0443, "step": 80675 }, { "epoch": 1.88, "learning_rate": 1.5981084696439886e-05, "loss": 0.077, "step": 80680 }, { "epoch": 1.88, "learning_rate": 1.5980204641993854e-05, "loss": 0.1716, "step": 80685 }, { "epoch": 1.88, "learning_rate": 1.5979324587547816e-05, "loss": 0.0329, "step": 80690 }, { "epoch": 1.88, "learning_rate": 1.597844453310178e-05, "loss": 0.0641, "step": 80695 }, { "epoch": 1.88, "learning_rate": 1.5977564478655746e-05, "loss": 0.2754, "step": 80700 }, { "epoch": 1.88, "learning_rate": 1.597668442420971e-05, "loss": 0.2126, "step": 80705 }, { "epoch": 1.88, "learning_rate": 1.5975804369763676e-05, "loss": 0.236, "step": 80710 }, { "epoch": 1.88, "learning_rate": 1.597492431531764e-05, "loss": 0.1722, "step": 80715 }, { "epoch": 1.88, "learning_rate": 1.5974044260871606e-05, "loss": 0.0307, "step": 80720 }, { "epoch": 1.88, "learning_rate": 1.597316420642557e-05, "loss": 0.0274, "step": 80725 }, { "epoch": 1.88, "learning_rate": 1.5972284151979536e-05, "loss": 0.0808, "step": 80730 }, { "epoch": 1.88, "learning_rate": 1.59714040975335e-05, "loss": 0.0865, "step": 80735 }, { "epoch": 1.88, "learning_rate": 1.5970524043087466e-05, "loss": 0.0678, "step": 80740 }, { "epoch": 1.88, "learning_rate": 1.596964398864143e-05, "loss": 0.1636, "step": 80745 }, { "epoch": 1.88, "learning_rate": 1.5968763934195396e-05, "loss": 0.1664, "step": 80750 }, { "epoch": 1.88, "learning_rate": 1.596788387974936e-05, "loss": 0.2556, "step": 80755 }, { "epoch": 1.88, "learning_rate": 1.5967003825303323e-05, "loss": 0.2148, "step": 80760 }, { "epoch": 1.88, "learning_rate": 1.596612377085729e-05, "loss": 0.5528, "step": 80765 }, { "epoch": 1.88, "learning_rate": 1.5965243716411256e-05, "loss": 0.0352, "step": 80770 }, { "epoch": 1.88, "learning_rate": 1.596436366196522e-05, "loss": 0.0715, "step": 80775 }, { "epoch": 1.88, "learning_rate": 1.5963483607519186e-05, "loss": 0.0781, "step": 80780 }, { "epoch": 1.88, "learning_rate": 1.596260355307315e-05, "loss": 0.0339, "step": 80785 }, { "epoch": 1.88, "learning_rate": 1.5961723498627116e-05, "loss": 0.1898, "step": 80790 }, { "epoch": 1.89, "learning_rate": 1.5960843444181078e-05, "loss": 0.1548, "step": 80795 }, { "epoch": 1.89, "learning_rate": 1.5959963389735046e-05, "loss": 0.2466, "step": 80800 }, { "epoch": 1.89, "learning_rate": 1.595908333528901e-05, "loss": 0.1627, "step": 80805 }, { "epoch": 1.89, "learning_rate": 1.5958203280842976e-05, "loss": 0.2302, "step": 80810 }, { "epoch": 1.89, "learning_rate": 1.5957323226396938e-05, "loss": 0.213, "step": 80815 }, { "epoch": 1.89, "learning_rate": 1.5956443171950907e-05, "loss": 0.0607, "step": 80820 }, { "epoch": 1.89, "learning_rate": 1.595556311750487e-05, "loss": 0.0129, "step": 80825 }, { "epoch": 1.89, "learning_rate": 1.5954683063058833e-05, "loss": 0.0997, "step": 80830 }, { "epoch": 1.89, "learning_rate": 1.5953803008612798e-05, "loss": 0.0844, "step": 80835 }, { "epoch": 1.89, "learning_rate": 1.5952922954166767e-05, "loss": 0.0934, "step": 80840 }, { "epoch": 1.89, "learning_rate": 1.595204289972073e-05, "loss": 0.1583, "step": 80845 }, { "epoch": 1.89, "learning_rate": 1.5951162845274693e-05, "loss": 0.1051, "step": 80850 }, { "epoch": 1.89, "learning_rate": 1.5950282790828658e-05, "loss": 0.1647, "step": 80855 }, { "epoch": 1.89, "learning_rate": 1.5949402736382627e-05, "loss": 0.289, "step": 80860 }, { "epoch": 1.89, "learning_rate": 1.594852268193659e-05, "loss": 0.1269, "step": 80865 }, { "epoch": 1.89, "learning_rate": 1.5947642627490553e-05, "loss": 0.0279, "step": 80870 }, { "epoch": 1.89, "learning_rate": 1.594676257304452e-05, "loss": 0.0792, "step": 80875 }, { "epoch": 1.89, "learning_rate": 1.5945882518598487e-05, "loss": 0.0579, "step": 80880 }, { "epoch": 1.89, "learning_rate": 1.594500246415245e-05, "loss": 0.0363, "step": 80885 }, { "epoch": 1.89, "learning_rate": 1.5944122409706413e-05, "loss": 0.0464, "step": 80890 }, { "epoch": 1.89, "learning_rate": 1.594324235526038e-05, "loss": 0.1928, "step": 80895 }, { "epoch": 1.89, "learning_rate": 1.5942362300814343e-05, "loss": 0.0193, "step": 80900 }, { "epoch": 1.89, "learning_rate": 1.594148224636831e-05, "loss": 0.2704, "step": 80905 }, { "epoch": 1.89, "learning_rate": 1.5940602191922274e-05, "loss": 0.3451, "step": 80910 }, { "epoch": 1.89, "learning_rate": 1.5939722137476242e-05, "loss": 0.3403, "step": 80915 }, { "epoch": 1.89, "learning_rate": 1.5938842083030204e-05, "loss": 0.0348, "step": 80920 }, { "epoch": 1.89, "learning_rate": 1.593796202858417e-05, "loss": 0.0029, "step": 80925 }, { "epoch": 1.89, "learning_rate": 1.5937081974138134e-05, "loss": 0.0029, "step": 80930 }, { "epoch": 1.89, "learning_rate": 1.59362019196921e-05, "loss": 0.1062, "step": 80935 }, { "epoch": 1.89, "learning_rate": 1.5935321865246064e-05, "loss": 0.0296, "step": 80940 }, { "epoch": 1.89, "learning_rate": 1.593444181080003e-05, "loss": 0.1883, "step": 80945 }, { "epoch": 1.89, "learning_rate": 1.5933561756353994e-05, "loss": 0.1837, "step": 80950 }, { "epoch": 1.89, "learning_rate": 1.593268170190796e-05, "loss": 0.0903, "step": 80955 }, { "epoch": 1.89, "learning_rate": 1.5931801647461924e-05, "loss": 0.2454, "step": 80960 }, { "epoch": 1.89, "learning_rate": 1.593092159301589e-05, "loss": 0.1849, "step": 80965 }, { "epoch": 1.89, "learning_rate": 1.593004153856985e-05, "loss": 0.084, "step": 80970 }, { "epoch": 1.89, "learning_rate": 1.592916148412382e-05, "loss": 0.0746, "step": 80975 }, { "epoch": 1.89, "learning_rate": 1.5928281429677784e-05, "loss": 0.0561, "step": 80980 }, { "epoch": 1.89, "learning_rate": 1.592740137523175e-05, "loss": 0.0834, "step": 80985 }, { "epoch": 1.89, "learning_rate": 1.592652132078571e-05, "loss": 0.0777, "step": 80990 }, { "epoch": 1.89, "learning_rate": 1.592564126633968e-05, "loss": 0.1288, "step": 80995 }, { "epoch": 1.89, "learning_rate": 1.5924761211893644e-05, "loss": 0.196, "step": 81000 }, { "epoch": 1.89, "learning_rate": 1.5923881157447606e-05, "loss": 0.1359, "step": 81005 }, { "epoch": 1.89, "learning_rate": 1.592300110300157e-05, "loss": 0.5108, "step": 81010 }, { "epoch": 1.89, "learning_rate": 1.592212104855554e-05, "loss": 0.2217, "step": 81015 }, { "epoch": 1.89, "learning_rate": 1.5921240994109504e-05, "loss": 0.1355, "step": 81020 }, { "epoch": 1.89, "learning_rate": 1.5920360939663466e-05, "loss": 0.0061, "step": 81025 }, { "epoch": 1.89, "learning_rate": 1.591948088521743e-05, "loss": 0.0486, "step": 81030 }, { "epoch": 1.89, "learning_rate": 1.59186008307714e-05, "loss": 0.0476, "step": 81035 }, { "epoch": 1.89, "learning_rate": 1.5917720776325364e-05, "loss": 0.1127, "step": 81040 }, { "epoch": 1.89, "learning_rate": 1.5916840721879326e-05, "loss": 0.1921, "step": 81045 }, { "epoch": 1.89, "learning_rate": 1.5915960667433294e-05, "loss": 0.1205, "step": 81050 }, { "epoch": 1.89, "learning_rate": 1.591508061298726e-05, "loss": 0.2675, "step": 81055 }, { "epoch": 1.89, "learning_rate": 1.591420055854122e-05, "loss": 0.2871, "step": 81060 }, { "epoch": 1.89, "learning_rate": 1.5913320504095186e-05, "loss": 0.1674, "step": 81065 }, { "epoch": 1.89, "learning_rate": 1.5912440449649154e-05, "loss": 0.0132, "step": 81070 }, { "epoch": 1.89, "learning_rate": 1.591156039520312e-05, "loss": 0.0151, "step": 81075 }, { "epoch": 1.89, "learning_rate": 1.591068034075708e-05, "loss": 0.0909, "step": 81080 }, { "epoch": 1.89, "learning_rate": 1.5909800286311046e-05, "loss": 0.0868, "step": 81085 }, { "epoch": 1.89, "learning_rate": 1.5908920231865014e-05, "loss": 0.1327, "step": 81090 }, { "epoch": 1.89, "learning_rate": 1.5908040177418976e-05, "loss": 0.1236, "step": 81095 }, { "epoch": 1.89, "learning_rate": 1.5907336133862147e-05, "loss": 0.2413, "step": 81100 }, { "epoch": 1.89, "learning_rate": 1.5906456079416116e-05, "loss": 0.3135, "step": 81105 }, { "epoch": 1.89, "learning_rate": 1.590557602497008e-05, "loss": 0.267, "step": 81110 }, { "epoch": 1.89, "learning_rate": 1.5904695970524042e-05, "loss": 0.2715, "step": 81115 }, { "epoch": 1.89, "learning_rate": 1.5903815916078007e-05, "loss": 0.0459, "step": 81120 }, { "epoch": 1.89, "learning_rate": 1.5902935861631976e-05, "loss": 0.0395, "step": 81125 }, { "epoch": 1.89, "learning_rate": 1.5902055807185937e-05, "loss": 0.0664, "step": 81130 }, { "epoch": 1.89, "learning_rate": 1.5901175752739902e-05, "loss": 0.0102, "step": 81135 }, { "epoch": 1.89, "learning_rate": 1.5900295698293867e-05, "loss": 0.0482, "step": 81140 }, { "epoch": 1.89, "learning_rate": 1.5899415643847836e-05, "loss": 0.2547, "step": 81145 }, { "epoch": 1.89, "learning_rate": 1.5898535589401797e-05, "loss": 0.088, "step": 81150 }, { "epoch": 1.89, "learning_rate": 1.5897655534955762e-05, "loss": 0.1828, "step": 81155 }, { "epoch": 1.89, "learning_rate": 1.5896775480509727e-05, "loss": 0.3202, "step": 81160 }, { "epoch": 1.89, "learning_rate": 1.5895895426063692e-05, "loss": 0.2503, "step": 81165 }, { "epoch": 1.89, "learning_rate": 1.5895015371617658e-05, "loss": 0.0102, "step": 81170 }, { "epoch": 1.89, "learning_rate": 1.5894135317171623e-05, "loss": 0.0294, "step": 81175 }, { "epoch": 1.89, "learning_rate": 1.5893255262725588e-05, "loss": 0.0633, "step": 81180 }, { "epoch": 1.89, "learning_rate": 1.5892375208279553e-05, "loss": 0.0565, "step": 81185 }, { "epoch": 1.89, "learning_rate": 1.5891495153833518e-05, "loss": 0.1654, "step": 81190 }, { "epoch": 1.89, "learning_rate": 1.5890615099387483e-05, "loss": 0.0399, "step": 81195 }, { "epoch": 1.89, "learning_rate": 1.5889735044941444e-05, "loss": 0.0895, "step": 81200 }, { "epoch": 1.89, "learning_rate": 1.5888854990495413e-05, "loss": 0.1803, "step": 81205 }, { "epoch": 1.89, "learning_rate": 1.5887974936049378e-05, "loss": 0.2756, "step": 81210 }, { "epoch": 1.89, "learning_rate": 1.5887094881603343e-05, "loss": 0.2724, "step": 81215 }, { "epoch": 1.89, "learning_rate": 1.5886214827157304e-05, "loss": 0.017, "step": 81220 }, { "epoch": 1.9, "learning_rate": 1.5885334772711273e-05, "loss": 0.0358, "step": 81225 }, { "epoch": 1.9, "learning_rate": 1.5884454718265238e-05, "loss": 0.0195, "step": 81230 }, { "epoch": 1.9, "learning_rate": 1.5883574663819203e-05, "loss": 0.0511, "step": 81235 }, { "epoch": 1.9, "learning_rate": 1.5882694609373168e-05, "loss": 0.076, "step": 81240 }, { "epoch": 1.9, "learning_rate": 1.5881814554927133e-05, "loss": 0.1937, "step": 81245 }, { "epoch": 1.9, "learning_rate": 1.5880934500481098e-05, "loss": 0.234, "step": 81250 }, { "epoch": 1.9, "learning_rate": 1.588005444603506e-05, "loss": 0.2045, "step": 81255 }, { "epoch": 1.9, "learning_rate": 1.5879174391589028e-05, "loss": 0.1145, "step": 81260 }, { "epoch": 1.9, "learning_rate": 1.5878294337142993e-05, "loss": 0.2791, "step": 81265 }, { "epoch": 1.9, "learning_rate": 1.5877414282696958e-05, "loss": 0.0383, "step": 81270 }, { "epoch": 1.9, "learning_rate": 1.587653422825092e-05, "loss": 0.0264, "step": 81275 }, { "epoch": 1.9, "learning_rate": 1.5875654173804888e-05, "loss": 0.0976, "step": 81280 }, { "epoch": 1.9, "learning_rate": 1.5874774119358853e-05, "loss": 0.0405, "step": 81285 }, { "epoch": 1.9, "learning_rate": 1.5873894064912815e-05, "loss": 0.0523, "step": 81290 }, { "epoch": 1.9, "learning_rate": 1.587301401046678e-05, "loss": 0.1283, "step": 81295 }, { "epoch": 1.9, "learning_rate": 1.5872133956020748e-05, "loss": 0.127, "step": 81300 }, { "epoch": 1.9, "learning_rate": 1.5871253901574713e-05, "loss": 0.1872, "step": 81305 }, { "epoch": 1.9, "learning_rate": 1.5870373847128675e-05, "loss": 0.3574, "step": 81310 }, { "epoch": 1.9, "learning_rate": 1.586949379268264e-05, "loss": 0.2223, "step": 81315 }, { "epoch": 1.9, "learning_rate": 1.5868613738236608e-05, "loss": 0.1192, "step": 81320 }, { "epoch": 1.9, "learning_rate": 1.586773368379057e-05, "loss": 0.07, "step": 81325 }, { "epoch": 1.9, "learning_rate": 1.5866853629344535e-05, "loss": 0.047, "step": 81330 }, { "epoch": 1.9, "learning_rate": 1.58659735748985e-05, "loss": 0.0051, "step": 81335 }, { "epoch": 1.9, "learning_rate": 1.5865093520452468e-05, "loss": 0.0915, "step": 81340 }, { "epoch": 1.9, "learning_rate": 1.586421346600643e-05, "loss": 0.0253, "step": 81345 }, { "epoch": 1.9, "learning_rate": 1.5863333411560395e-05, "loss": 0.0556, "step": 81350 }, { "epoch": 1.9, "learning_rate": 1.586245335711436e-05, "loss": 0.2999, "step": 81355 }, { "epoch": 1.9, "learning_rate": 1.5861573302668325e-05, "loss": 0.387, "step": 81360 }, { "epoch": 1.9, "learning_rate": 1.586069324822229e-05, "loss": 0.3123, "step": 81365 }, { "epoch": 1.9, "learning_rate": 1.5859813193776255e-05, "loss": 0.03, "step": 81370 }, { "epoch": 1.9, "learning_rate": 1.5858933139330223e-05, "loss": 0.0578, "step": 81375 }, { "epoch": 1.9, "learning_rate": 1.585822909577339e-05, "loss": 0.1745, "step": 81380 }, { "epoch": 1.9, "learning_rate": 1.5857349041327356e-05, "loss": 0.0943, "step": 81385 }, { "epoch": 1.9, "learning_rate": 1.585646898688132e-05, "loss": 0.2174, "step": 81390 }, { "epoch": 1.9, "learning_rate": 1.5855588932435286e-05, "loss": 0.137, "step": 81395 }, { "epoch": 1.9, "learning_rate": 1.585470887798925e-05, "loss": 0.1076, "step": 81400 }, { "epoch": 1.9, "learning_rate": 1.5853828823543216e-05, "loss": 0.1685, "step": 81405 }, { "epoch": 1.9, "learning_rate": 1.585294876909718e-05, "loss": 0.1073, "step": 81410 }, { "epoch": 1.9, "learning_rate": 1.5852068714651146e-05, "loss": 0.1916, "step": 81415 }, { "epoch": 1.9, "learning_rate": 1.585118866020511e-05, "loss": 0.0434, "step": 81420 }, { "epoch": 1.9, "learning_rate": 1.5850308605759076e-05, "loss": 0.0319, "step": 81425 }, { "epoch": 1.9, "learning_rate": 1.5849428551313045e-05, "loss": 0.0442, "step": 81430 }, { "epoch": 1.9, "learning_rate": 1.5848548496867007e-05, "loss": 0.0224, "step": 81435 }, { "epoch": 1.9, "learning_rate": 1.584766844242097e-05, "loss": 0.0523, "step": 81440 }, { "epoch": 1.9, "learning_rate": 1.5846788387974937e-05, "loss": 0.2548, "step": 81445 }, { "epoch": 1.9, "learning_rate": 1.58459083335289e-05, "loss": 0.0725, "step": 81450 }, { "epoch": 1.9, "learning_rate": 1.5845028279082867e-05, "loss": 0.0744, "step": 81455 }, { "epoch": 1.9, "learning_rate": 1.584414822463683e-05, "loss": 0.303, "step": 81460 }, { "epoch": 1.9, "learning_rate": 1.5843268170190797e-05, "loss": 0.2577, "step": 81465 }, { "epoch": 1.9, "learning_rate": 1.584238811574476e-05, "loss": 0.0799, "step": 81470 }, { "epoch": 1.9, "learning_rate": 1.5841508061298727e-05, "loss": 0.0177, "step": 81475 }, { "epoch": 1.9, "learning_rate": 1.5840628006852692e-05, "loss": 0.0539, "step": 81480 }, { "epoch": 1.9, "learning_rate": 1.5839747952406653e-05, "loss": 0.0722, "step": 81485 }, { "epoch": 1.9, "learning_rate": 1.5838867897960622e-05, "loss": 0.0164, "step": 81490 }, { "epoch": 1.9, "learning_rate": 1.5837987843514587e-05, "loss": 0.1857, "step": 81495 }, { "epoch": 1.9, "learning_rate": 1.5837107789068552e-05, "loss": 0.2002, "step": 81500 }, { "epoch": 1.9, "learning_rate": 1.5836227734622513e-05, "loss": 0.1618, "step": 81505 }, { "epoch": 1.9, "learning_rate": 1.5835347680176482e-05, "loss": 0.3, "step": 81510 }, { "epoch": 1.9, "learning_rate": 1.5834467625730447e-05, "loss": 0.1901, "step": 81515 }, { "epoch": 1.9, "learning_rate": 1.583358757128441e-05, "loss": 0.016, "step": 81520 }, { "epoch": 1.9, "learning_rate": 1.5832707516838374e-05, "loss": 0.0683, "step": 81525 }, { "epoch": 1.9, "learning_rate": 1.5831827462392342e-05, "loss": 0.0115, "step": 81530 }, { "epoch": 1.9, "learning_rate": 1.5830947407946307e-05, "loss": 0.0152, "step": 81535 }, { "epoch": 1.9, "learning_rate": 1.583006735350027e-05, "loss": 0.103, "step": 81540 }, { "epoch": 1.9, "learning_rate": 1.5829187299054234e-05, "loss": 0.1557, "step": 81545 }, { "epoch": 1.9, "learning_rate": 1.5828307244608202e-05, "loss": 0.0641, "step": 81550 }, { "epoch": 1.9, "learning_rate": 1.5827427190162164e-05, "loss": 0.2051, "step": 81555 }, { "epoch": 1.9, "learning_rate": 1.582654713571613e-05, "loss": 0.272, "step": 81560 }, { "epoch": 1.9, "learning_rate": 1.5825667081270097e-05, "loss": 0.329, "step": 81565 }, { "epoch": 1.9, "learning_rate": 1.5824787026824062e-05, "loss": 0.0367, "step": 81570 }, { "epoch": 1.9, "learning_rate": 1.5823906972378024e-05, "loss": 0.1126, "step": 81575 }, { "epoch": 1.9, "learning_rate": 1.582302691793199e-05, "loss": 0.0913, "step": 81580 }, { "epoch": 1.9, "learning_rate": 1.5822146863485957e-05, "loss": 0.1525, "step": 81585 }, { "epoch": 1.9, "learning_rate": 1.582126680903992e-05, "loss": 0.0703, "step": 81590 }, { "epoch": 1.9, "learning_rate": 1.5820386754593884e-05, "loss": 0.1219, "step": 81595 }, { "epoch": 1.9, "learning_rate": 1.581950670014785e-05, "loss": 0.0488, "step": 81600 }, { "epoch": 1.9, "learning_rate": 1.5818626645701817e-05, "loss": 0.1668, "step": 81605 }, { "epoch": 1.9, "learning_rate": 1.581774659125578e-05, "loss": 0.2141, "step": 81610 }, { "epoch": 1.9, "learning_rate": 1.5816866536809744e-05, "loss": 0.4408, "step": 81615 }, { "epoch": 1.9, "learning_rate": 1.581598648236371e-05, "loss": 0.0239, "step": 81620 }, { "epoch": 1.9, "learning_rate": 1.5815106427917674e-05, "loss": 0.0186, "step": 81625 }, { "epoch": 1.9, "learning_rate": 1.581422637347164e-05, "loss": 0.0238, "step": 81630 }, { "epoch": 1.9, "learning_rate": 1.5813346319025604e-05, "loss": 0.0513, "step": 81635 }, { "epoch": 1.9, "learning_rate": 1.581246626457957e-05, "loss": 0.163, "step": 81640 }, { "epoch": 1.9, "learning_rate": 1.5811586210133534e-05, "loss": 0.0098, "step": 81645 }, { "epoch": 1.9, "learning_rate": 1.58107061556875e-05, "loss": 0.1659, "step": 81650 }, { "epoch": 1.91, "learning_rate": 1.5809826101241464e-05, "loss": 0.1403, "step": 81655 }, { "epoch": 1.91, "learning_rate": 1.5808946046795426e-05, "loss": 0.1641, "step": 81660 }, { "epoch": 1.91, "learning_rate": 1.5808065992349394e-05, "loss": 0.202, "step": 81665 }, { "epoch": 1.91, "learning_rate": 1.580718593790336e-05, "loss": 0.0692, "step": 81670 }, { "epoch": 1.91, "learning_rate": 1.5806305883457324e-05, "loss": 0.0118, "step": 81675 }, { "epoch": 1.91, "learning_rate": 1.5805425829011286e-05, "loss": 0.0577, "step": 81680 }, { "epoch": 1.91, "learning_rate": 1.5804545774565254e-05, "loss": 0.0667, "step": 81685 }, { "epoch": 1.91, "learning_rate": 1.580366572011922e-05, "loss": 0.1466, "step": 81690 }, { "epoch": 1.91, "learning_rate": 1.580278566567318e-05, "loss": 0.0453, "step": 81695 }, { "epoch": 1.91, "learning_rate": 1.580190561122715e-05, "loss": 0.1364, "step": 81700 }, { "epoch": 1.91, "learning_rate": 1.5801025556781114e-05, "loss": 0.2246, "step": 81705 }, { "epoch": 1.91, "learning_rate": 1.580014550233508e-05, "loss": 0.2743, "step": 81710 }, { "epoch": 1.91, "learning_rate": 1.579926544788904e-05, "loss": 0.357, "step": 81715 }, { "epoch": 1.91, "learning_rate": 1.579838539344301e-05, "loss": 0.0544, "step": 81720 }, { "epoch": 1.91, "learning_rate": 1.5797505338996974e-05, "loss": 0.1409, "step": 81725 }, { "epoch": 1.91, "learning_rate": 1.579662528455094e-05, "loss": 0.061, "step": 81730 }, { "epoch": 1.91, "learning_rate": 1.57957452301049e-05, "loss": 0.1406, "step": 81735 }, { "epoch": 1.91, "learning_rate": 1.579486517565887e-05, "loss": 0.1516, "step": 81740 }, { "epoch": 1.91, "learning_rate": 1.5793985121212834e-05, "loss": 0.1109, "step": 81745 }, { "epoch": 1.91, "learning_rate": 1.5793105066766796e-05, "loss": 0.0954, "step": 81750 }, { "epoch": 1.91, "learning_rate": 1.579222501232076e-05, "loss": 0.1592, "step": 81755 }, { "epoch": 1.91, "learning_rate": 1.579134495787473e-05, "loss": 0.1928, "step": 81760 }, { "epoch": 1.91, "learning_rate": 1.5790464903428694e-05, "loss": 0.3713, "step": 81765 }, { "epoch": 1.91, "learning_rate": 1.5789584848982656e-05, "loss": 0.012, "step": 81770 }, { "epoch": 1.91, "learning_rate": 1.578870479453662e-05, "loss": 0.0281, "step": 81775 }, { "epoch": 1.91, "learning_rate": 1.578782474009059e-05, "loss": 0.0392, "step": 81780 }, { "epoch": 1.91, "learning_rate": 1.578694468564455e-05, "loss": 0.0577, "step": 81785 }, { "epoch": 1.91, "learning_rate": 1.5786064631198516e-05, "loss": 0.2178, "step": 81790 }, { "epoch": 1.91, "learning_rate": 1.578518457675248e-05, "loss": 0.1446, "step": 81795 }, { "epoch": 1.91, "learning_rate": 1.578430452230645e-05, "loss": 0.1491, "step": 81800 }, { "epoch": 1.91, "learning_rate": 1.578342446786041e-05, "loss": 0.1919, "step": 81805 }, { "epoch": 1.91, "learning_rate": 1.5782544413414376e-05, "loss": 0.2936, "step": 81810 }, { "epoch": 1.91, "learning_rate": 1.578166435896834e-05, "loss": 0.2299, "step": 81815 }, { "epoch": 1.91, "learning_rate": 1.5780784304522306e-05, "loss": 0.02, "step": 81820 }, { "epoch": 1.91, "learning_rate": 1.577990425007627e-05, "loss": 0.0243, "step": 81825 }, { "epoch": 1.91, "learning_rate": 1.5779024195630236e-05, "loss": 0.0301, "step": 81830 }, { "epoch": 1.91, "learning_rate": 1.5778144141184205e-05, "loss": 0.0708, "step": 81835 }, { "epoch": 1.91, "learning_rate": 1.5777264086738166e-05, "loss": 0.0463, "step": 81840 }, { "epoch": 1.91, "learning_rate": 1.577638403229213e-05, "loss": 0.1446, "step": 81845 }, { "epoch": 1.91, "learning_rate": 1.5775503977846096e-05, "loss": 0.1471, "step": 81850 }, { "epoch": 1.91, "learning_rate": 1.577462392340006e-05, "loss": 0.1749, "step": 81855 }, { "epoch": 1.91, "learning_rate": 1.5773743868954026e-05, "loss": 0.0886, "step": 81860 }, { "epoch": 1.91, "learning_rate": 1.577286381450799e-05, "loss": 0.1983, "step": 81865 }, { "epoch": 1.91, "learning_rate": 1.5771983760061957e-05, "loss": 0.0432, "step": 81870 }, { "epoch": 1.91, "learning_rate": 1.577110370561592e-05, "loss": 0.0021, "step": 81875 }, { "epoch": 1.91, "learning_rate": 1.5770223651169887e-05, "loss": 0.1551, "step": 81880 }, { "epoch": 1.91, "learning_rate": 1.576934359672385e-05, "loss": 0.0458, "step": 81885 }, { "epoch": 1.91, "learning_rate": 1.5768463542277813e-05, "loss": 0.0976, "step": 81890 }, { "epoch": 1.91, "learning_rate": 1.576758348783178e-05, "loss": 0.0297, "step": 81895 }, { "epoch": 1.91, "learning_rate": 1.5766703433385747e-05, "loss": 0.1553, "step": 81900 }, { "epoch": 1.91, "learning_rate": 1.576582337893971e-05, "loss": 0.1297, "step": 81905 }, { "epoch": 1.91, "learning_rate": 1.5764943324493673e-05, "loss": 0.2142, "step": 81910 }, { "epoch": 1.91, "learning_rate": 1.5764063270047642e-05, "loss": 0.1508, "step": 81915 }, { "epoch": 1.91, "learning_rate": 1.5763183215601607e-05, "loss": 0.0408, "step": 81920 }, { "epoch": 1.91, "learning_rate": 1.576230316115557e-05, "loss": 0.0412, "step": 81925 }, { "epoch": 1.91, "learning_rate": 1.5761423106709533e-05, "loss": 0.0203, "step": 81930 }, { "epoch": 1.91, "learning_rate": 1.5760543052263502e-05, "loss": 0.0577, "step": 81935 }, { "epoch": 1.91, "learning_rate": 1.5759662997817467e-05, "loss": 0.0743, "step": 81940 }, { "epoch": 1.91, "learning_rate": 1.575878294337143e-05, "loss": 0.1519, "step": 81945 }, { "epoch": 1.91, "learning_rate": 1.5757902888925393e-05, "loss": 0.0783, "step": 81950 }, { "epoch": 1.91, "learning_rate": 1.5757022834479362e-05, "loss": 0.1019, "step": 81955 }, { "epoch": 1.91, "learning_rate": 1.5756142780033324e-05, "loss": 0.1405, "step": 81960 }, { "epoch": 1.91, "learning_rate": 1.575526272558729e-05, "loss": 0.0848, "step": 81965 }, { "epoch": 1.91, "learning_rate": 1.5754382671141257e-05, "loss": 0.0405, "step": 81970 }, { "epoch": 1.91, "learning_rate": 1.5753502616695222e-05, "loss": 0.0272, "step": 81975 }, { "epoch": 1.91, "learning_rate": 1.5752622562249184e-05, "loss": 0.0305, "step": 81980 }, { "epoch": 1.91, "learning_rate": 1.575174250780315e-05, "loss": 0.0567, "step": 81985 }, { "epoch": 1.91, "learning_rate": 1.5750862453357117e-05, "loss": 0.0928, "step": 81990 }, { "epoch": 1.91, "learning_rate": 1.574998239891108e-05, "loss": 0.2398, "step": 81995 }, { "epoch": 1.91, "learning_rate": 1.5749102344465044e-05, "loss": 0.1104, "step": 82000 }, { "epoch": 1.91, "learning_rate": 1.574822229001901e-05, "loss": 0.1147, "step": 82005 }, { "epoch": 1.91, "learning_rate": 1.5747342235572977e-05, "loss": 0.2789, "step": 82010 }, { "epoch": 1.91, "learning_rate": 1.574646218112694e-05, "loss": 0.1877, "step": 82015 }, { "epoch": 1.91, "learning_rate": 1.5745582126680904e-05, "loss": 0.0053, "step": 82020 }, { "epoch": 1.91, "learning_rate": 1.574470207223487e-05, "loss": 0.0382, "step": 82025 }, { "epoch": 1.91, "learning_rate": 1.5743822017788837e-05, "loss": 0.0721, "step": 82030 }, { "epoch": 1.91, "learning_rate": 1.57429419633428e-05, "loss": 0.055, "step": 82035 }, { "epoch": 1.91, "learning_rate": 1.5742061908896764e-05, "loss": 0.0257, "step": 82040 }, { "epoch": 1.91, "learning_rate": 1.574118185445073e-05, "loss": 0.0579, "step": 82045 }, { "epoch": 1.91, "learning_rate": 1.5740301800004694e-05, "loss": 0.2128, "step": 82050 }, { "epoch": 1.91, "learning_rate": 1.573942174555866e-05, "loss": 0.1346, "step": 82055 }, { "epoch": 1.91, "learning_rate": 1.5738541691112624e-05, "loss": 0.3942, "step": 82060 }, { "epoch": 1.91, "learning_rate": 1.573766163666659e-05, "loss": 0.3479, "step": 82065 }, { "epoch": 1.91, "learning_rate": 1.5736781582220554e-05, "loss": 0.0447, "step": 82070 }, { "epoch": 1.91, "learning_rate": 1.573590152777452e-05, "loss": 0.0076, "step": 82075 }, { "epoch": 1.92, "learning_rate": 1.5735021473328484e-05, "loss": 0.0706, "step": 82080 }, { "epoch": 1.92, "learning_rate": 1.5734141418882446e-05, "loss": 0.0314, "step": 82085 }, { "epoch": 1.92, "learning_rate": 1.5733261364436414e-05, "loss": 0.0357, "step": 82090 }, { "epoch": 1.92, "learning_rate": 1.573238130999038e-05, "loss": 0.0986, "step": 82095 }, { "epoch": 1.92, "learning_rate": 1.5731501255544344e-05, "loss": 0.1768, "step": 82100 }, { "epoch": 1.92, "learning_rate": 1.573062120109831e-05, "loss": 0.1093, "step": 82105 }, { "epoch": 1.92, "learning_rate": 1.5729741146652274e-05, "loss": 0.223, "step": 82110 }, { "epoch": 1.92, "learning_rate": 1.572886109220624e-05, "loss": 0.2313, "step": 82115 }, { "epoch": 1.92, "learning_rate": 1.57279810377602e-05, "loss": 0.0431, "step": 82120 }, { "epoch": 1.92, "learning_rate": 1.572710098331417e-05, "loss": 0.034, "step": 82125 }, { "epoch": 1.92, "learning_rate": 1.5726220928868134e-05, "loss": 0.0462, "step": 82130 }, { "epoch": 1.92, "learning_rate": 1.57253408744221e-05, "loss": 0.1237, "step": 82135 }, { "epoch": 1.92, "learning_rate": 1.572446081997606e-05, "loss": 0.0566, "step": 82140 }, { "epoch": 1.92, "learning_rate": 1.572358076553003e-05, "loss": 0.013, "step": 82145 }, { "epoch": 1.92, "learning_rate": 1.5722700711083994e-05, "loss": 0.1431, "step": 82150 }, { "epoch": 1.92, "learning_rate": 1.5721820656637956e-05, "loss": 0.1703, "step": 82155 }, { "epoch": 1.92, "learning_rate": 1.572094060219192e-05, "loss": 0.2691, "step": 82160 }, { "epoch": 1.92, "learning_rate": 1.572006054774589e-05, "loss": 0.2651, "step": 82165 }, { "epoch": 1.92, "learning_rate": 1.5719180493299854e-05, "loss": 0.0674, "step": 82170 }, { "epoch": 1.92, "learning_rate": 1.5718300438853816e-05, "loss": 0.0509, "step": 82175 }, { "epoch": 1.92, "learning_rate": 1.571742038440778e-05, "loss": 0.1591, "step": 82180 }, { "epoch": 1.92, "learning_rate": 1.571654032996175e-05, "loss": 0.059, "step": 82185 }, { "epoch": 1.92, "learning_rate": 1.571566027551571e-05, "loss": 0.1355, "step": 82190 }, { "epoch": 1.92, "learning_rate": 1.5714780221069676e-05, "loss": 0.08, "step": 82195 }, { "epoch": 1.92, "learning_rate": 1.571390016662364e-05, "loss": 0.1157, "step": 82200 }, { "epoch": 1.92, "learning_rate": 1.571302011217761e-05, "loss": 0.0953, "step": 82205 }, { "epoch": 1.92, "learning_rate": 1.571214005773157e-05, "loss": 0.3337, "step": 82210 }, { "epoch": 1.92, "learning_rate": 1.5711260003285536e-05, "loss": 0.3357, "step": 82215 }, { "epoch": 1.92, "learning_rate": 1.57103799488395e-05, "loss": 0.0179, "step": 82220 }, { "epoch": 1.92, "learning_rate": 1.5709499894393466e-05, "loss": 0.0297, "step": 82225 }, { "epoch": 1.92, "learning_rate": 1.570861983994743e-05, "loss": 0.0641, "step": 82230 }, { "epoch": 1.92, "learning_rate": 1.5707739785501396e-05, "loss": 0.1221, "step": 82235 }, { "epoch": 1.92, "learning_rate": 1.5706859731055365e-05, "loss": 0.0794, "step": 82240 }, { "epoch": 1.92, "learning_rate": 1.5705979676609326e-05, "loss": 0.1191, "step": 82245 }, { "epoch": 1.92, "learning_rate": 1.570509962216329e-05, "loss": 0.0753, "step": 82250 }, { "epoch": 1.92, "learning_rate": 1.5704219567717256e-05, "loss": 0.1493, "step": 82255 }, { "epoch": 1.92, "learning_rate": 1.570333951327122e-05, "loss": 0.273, "step": 82260 }, { "epoch": 1.92, "learning_rate": 1.5702459458825186e-05, "loss": 0.3897, "step": 82265 }, { "epoch": 1.92, "learning_rate": 1.570157940437915e-05, "loss": 0.0246, "step": 82270 }, { "epoch": 1.92, "learning_rate": 1.5700699349933116e-05, "loss": 0.0127, "step": 82275 }, { "epoch": 1.92, "learning_rate": 1.569981929548708e-05, "loss": 0.0069, "step": 82280 }, { "epoch": 1.92, "learning_rate": 1.5698939241041046e-05, "loss": 0.1769, "step": 82285 }, { "epoch": 1.92, "learning_rate": 1.569805918659501e-05, "loss": 0.0813, "step": 82290 }, { "epoch": 1.92, "learning_rate": 1.5697179132148973e-05, "loss": 0.0294, "step": 82295 }, { "epoch": 1.92, "learning_rate": 1.569629907770294e-05, "loss": 0.164, "step": 82300 }, { "epoch": 1.92, "learning_rate": 1.5695419023256907e-05, "loss": 0.2387, "step": 82305 }, { "epoch": 1.92, "learning_rate": 1.569453896881087e-05, "loss": 0.1948, "step": 82310 }, { "epoch": 1.92, "learning_rate": 1.5693658914364833e-05, "loss": 0.24, "step": 82315 }, { "epoch": 1.92, "learning_rate": 1.56927788599188e-05, "loss": 0.0765, "step": 82320 }, { "epoch": 1.92, "learning_rate": 1.5691898805472767e-05, "loss": 0.0317, "step": 82325 }, { "epoch": 1.92, "learning_rate": 1.569101875102673e-05, "loss": 0.0994, "step": 82330 }, { "epoch": 1.92, "learning_rate": 1.5690138696580693e-05, "loss": 0.1444, "step": 82335 }, { "epoch": 1.92, "learning_rate": 1.5689258642134662e-05, "loss": 0.0424, "step": 82340 }, { "epoch": 1.92, "learning_rate": 1.5688378587688627e-05, "loss": 0.1589, "step": 82345 }, { "epoch": 1.92, "learning_rate": 1.568749853324259e-05, "loss": 0.1757, "step": 82350 }, { "epoch": 1.92, "learning_rate": 1.5686618478796553e-05, "loss": 0.119, "step": 82355 }, { "epoch": 1.92, "learning_rate": 1.5685738424350522e-05, "loss": 0.2494, "step": 82360 }, { "epoch": 1.92, "learning_rate": 1.5684858369904487e-05, "loss": 0.1862, "step": 82365 }, { "epoch": 1.92, "learning_rate": 1.568397831545845e-05, "loss": 0.0106, "step": 82370 }, { "epoch": 1.92, "learning_rate": 1.5683098261012417e-05, "loss": 0.0407, "step": 82375 }, { "epoch": 1.92, "learning_rate": 1.5682218206566382e-05, "loss": 0.0306, "step": 82380 }, { "epoch": 1.92, "learning_rate": 1.5681338152120343e-05, "loss": 0.0747, "step": 82385 }, { "epoch": 1.92, "learning_rate": 1.568045809767431e-05, "loss": 0.1513, "step": 82390 }, { "epoch": 1.92, "learning_rate": 1.5679578043228277e-05, "loss": 0.0716, "step": 82395 }, { "epoch": 1.92, "learning_rate": 1.5678697988782242e-05, "loss": 0.1704, "step": 82400 }, { "epoch": 1.92, "learning_rate": 1.5677817934336204e-05, "loss": 0.1633, "step": 82405 }, { "epoch": 1.92, "learning_rate": 1.567693787989017e-05, "loss": 0.2406, "step": 82410 }, { "epoch": 1.92, "learning_rate": 1.5676057825444137e-05, "loss": 0.3364, "step": 82415 }, { "epoch": 1.92, "learning_rate": 1.56751777709981e-05, "loss": 0.0315, "step": 82420 }, { "epoch": 1.92, "learning_rate": 1.5674297716552064e-05, "loss": 0.0962, "step": 82425 }, { "epoch": 1.92, "learning_rate": 1.567341766210603e-05, "loss": 0.0691, "step": 82430 }, { "epoch": 1.92, "learning_rate": 1.5672537607659997e-05, "loss": 0.084, "step": 82435 }, { "epoch": 1.92, "learning_rate": 1.567165755321396e-05, "loss": 0.1251, "step": 82440 }, { "epoch": 1.92, "learning_rate": 1.5670777498767924e-05, "loss": 0.06, "step": 82445 }, { "epoch": 1.92, "learning_rate": 1.566989744432189e-05, "loss": 0.0847, "step": 82450 }, { "epoch": 1.92, "learning_rate": 1.5669017389875854e-05, "loss": 0.1123, "step": 82455 }, { "epoch": 1.92, "learning_rate": 1.566813733542982e-05, "loss": 0.2792, "step": 82460 }, { "epoch": 1.92, "learning_rate": 1.5667257280983784e-05, "loss": 0.2442, "step": 82465 }, { "epoch": 1.92, "learning_rate": 1.566637722653775e-05, "loss": 0.0346, "step": 82470 }, { "epoch": 1.92, "learning_rate": 1.5665497172091714e-05, "loss": 0.0223, "step": 82475 }, { "epoch": 1.92, "learning_rate": 1.566461711764568e-05, "loss": 0.0552, "step": 82480 }, { "epoch": 1.92, "learning_rate": 1.5663737063199644e-05, "loss": 0.0558, "step": 82485 }, { "epoch": 1.92, "learning_rate": 1.5662857008753606e-05, "loss": 0.0531, "step": 82490 }, { "epoch": 1.92, "learning_rate": 1.5661976954307574e-05, "loss": 0.2505, "step": 82495 }, { "epoch": 1.92, "learning_rate": 1.566109689986154e-05, "loss": 0.024, "step": 82500 }, { "epoch": 1.92, "learning_rate": 1.5660216845415504e-05, "loss": 0.0734, "step": 82505 }, { "epoch": 1.93, "learning_rate": 1.5659512801858675e-05, "loss": 0.23, "step": 82510 }, { "epoch": 1.93, "learning_rate": 1.565863274741264e-05, "loss": 0.4113, "step": 82515 }, { "epoch": 1.93, "learning_rate": 1.5657752692966605e-05, "loss": 0.05, "step": 82520 }, { "epoch": 1.93, "learning_rate": 1.565687263852057e-05, "loss": 0.0017, "step": 82525 }, { "epoch": 1.93, "learning_rate": 1.5655992584074535e-05, "loss": 0.0008, "step": 82530 }, { "epoch": 1.93, "learning_rate": 1.56551125296285e-05, "loss": 0.077, "step": 82535 }, { "epoch": 1.93, "learning_rate": 1.5654232475182465e-05, "loss": 0.0246, "step": 82540 }, { "epoch": 1.93, "learning_rate": 1.5653352420736427e-05, "loss": 0.1694, "step": 82545 }, { "epoch": 1.93, "learning_rate": 1.5652472366290395e-05, "loss": 0.0773, "step": 82550 }, { "epoch": 1.93, "learning_rate": 1.565159231184436e-05, "loss": 0.2547, "step": 82555 }, { "epoch": 1.93, "learning_rate": 1.5650712257398326e-05, "loss": 0.4914, "step": 82560 }, { "epoch": 1.93, "learning_rate": 1.564983220295229e-05, "loss": 0.1674, "step": 82565 }, { "epoch": 1.93, "learning_rate": 1.5648952148506256e-05, "loss": 0.1579, "step": 82570 }, { "epoch": 1.93, "learning_rate": 1.564807209406022e-05, "loss": 0.0397, "step": 82575 }, { "epoch": 1.93, "learning_rate": 1.5647192039614182e-05, "loss": 0.1197, "step": 82580 }, { "epoch": 1.93, "learning_rate": 1.564631198516815e-05, "loss": 0.0661, "step": 82585 }, { "epoch": 1.93, "learning_rate": 1.5645431930722116e-05, "loss": 0.0685, "step": 82590 }, { "epoch": 1.93, "learning_rate": 1.564455187627608e-05, "loss": 0.0539, "step": 82595 }, { "epoch": 1.93, "learning_rate": 1.5643671821830042e-05, "loss": 0.0952, "step": 82600 }, { "epoch": 1.93, "learning_rate": 1.564279176738401e-05, "loss": 0.2363, "step": 82605 }, { "epoch": 1.93, "learning_rate": 1.5641911712937976e-05, "loss": 0.2122, "step": 82610 }, { "epoch": 1.93, "learning_rate": 1.5641031658491937e-05, "loss": 0.3198, "step": 82615 }, { "epoch": 1.93, "learning_rate": 1.5640151604045902e-05, "loss": 0.0012, "step": 82620 }, { "epoch": 1.93, "learning_rate": 1.563927154959987e-05, "loss": 0.0183, "step": 82625 }, { "epoch": 1.93, "learning_rate": 1.5638391495153836e-05, "loss": 0.0315, "step": 82630 }, { "epoch": 1.93, "learning_rate": 1.5637511440707797e-05, "loss": 0.1255, "step": 82635 }, { "epoch": 1.93, "learning_rate": 1.5636631386261762e-05, "loss": 0.1393, "step": 82640 }, { "epoch": 1.93, "learning_rate": 1.563575133181573e-05, "loss": 0.0418, "step": 82645 }, { "epoch": 1.93, "learning_rate": 1.5634871277369693e-05, "loss": 0.1032, "step": 82650 }, { "epoch": 1.93, "learning_rate": 1.5633991222923658e-05, "loss": 0.1799, "step": 82655 }, { "epoch": 1.93, "learning_rate": 1.5633111168477623e-05, "loss": 0.2655, "step": 82660 }, { "epoch": 1.93, "learning_rate": 1.563223111403159e-05, "loss": 0.2426, "step": 82665 }, { "epoch": 1.93, "learning_rate": 1.5631351059585553e-05, "loss": 0.0318, "step": 82670 }, { "epoch": 1.93, "learning_rate": 1.5630471005139518e-05, "loss": 0.0293, "step": 82675 }, { "epoch": 1.93, "learning_rate": 1.5629590950693483e-05, "loss": 0.0199, "step": 82680 }, { "epoch": 1.93, "learning_rate": 1.5628710896247448e-05, "loss": 0.1427, "step": 82685 }, { "epoch": 1.93, "learning_rate": 1.5627830841801413e-05, "loss": 0.0939, "step": 82690 }, { "epoch": 1.93, "learning_rate": 1.5626950787355378e-05, "loss": 0.1593, "step": 82695 }, { "epoch": 1.93, "learning_rate": 1.5626070732909343e-05, "loss": 0.1579, "step": 82700 }, { "epoch": 1.93, "learning_rate": 1.5625190678463308e-05, "loss": 0.1445, "step": 82705 }, { "epoch": 1.93, "learning_rate": 1.5624310624017273e-05, "loss": 0.2457, "step": 82710 }, { "epoch": 1.93, "learning_rate": 1.5623430569571238e-05, "loss": 0.1935, "step": 82715 }, { "epoch": 1.93, "learning_rate": 1.5622550515125203e-05, "loss": 0.0618, "step": 82720 }, { "epoch": 1.93, "learning_rate": 1.5621670460679168e-05, "loss": 0.0582, "step": 82725 }, { "epoch": 1.93, "learning_rate": 1.5620790406233133e-05, "loss": 0.0342, "step": 82730 }, { "epoch": 1.93, "learning_rate": 1.5619910351787098e-05, "loss": 0.0975, "step": 82735 }, { "epoch": 1.93, "learning_rate": 1.5619030297341063e-05, "loss": 0.0502, "step": 82740 }, { "epoch": 1.93, "learning_rate": 1.5618150242895028e-05, "loss": 0.146, "step": 82745 }, { "epoch": 1.93, "learning_rate": 1.5617270188448993e-05, "loss": 0.1485, "step": 82750 }, { "epoch": 1.93, "learning_rate": 1.5616390134002955e-05, "loss": 0.104, "step": 82755 }, { "epoch": 1.93, "learning_rate": 1.5615510079556923e-05, "loss": 0.4562, "step": 82760 }, { "epoch": 1.93, "learning_rate": 1.5614630025110888e-05, "loss": 0.5022, "step": 82765 }, { "epoch": 1.93, "learning_rate": 1.5613749970664853e-05, "loss": 0.0532, "step": 82770 }, { "epoch": 1.93, "learning_rate": 1.5612869916218815e-05, "loss": 0.1032, "step": 82775 }, { "epoch": 1.93, "learning_rate": 1.5611989861772783e-05, "loss": 0.0738, "step": 82780 }, { "epoch": 1.93, "learning_rate": 1.5611109807326748e-05, "loss": 0.0421, "step": 82785 }, { "epoch": 1.93, "learning_rate": 1.561022975288071e-05, "loss": 0.0678, "step": 82790 }, { "epoch": 1.93, "learning_rate": 1.5609349698434675e-05, "loss": 0.0117, "step": 82795 }, { "epoch": 1.93, "learning_rate": 1.5608469643988643e-05, "loss": 0.2063, "step": 82800 }, { "epoch": 1.93, "learning_rate": 1.5607589589542608e-05, "loss": 0.3399, "step": 82805 }, { "epoch": 1.93, "learning_rate": 1.560670953509657e-05, "loss": 0.2389, "step": 82810 }, { "epoch": 1.93, "learning_rate": 1.5605829480650535e-05, "loss": 0.2606, "step": 82815 }, { "epoch": 1.93, "learning_rate": 1.5604949426204503e-05, "loss": 0.0873, "step": 82820 }, { "epoch": 1.93, "learning_rate": 1.5604069371758468e-05, "loss": 0.0394, "step": 82825 }, { "epoch": 1.93, "learning_rate": 1.560318931731243e-05, "loss": 0.0848, "step": 82830 }, { "epoch": 1.93, "learning_rate": 1.5602309262866398e-05, "loss": 0.1273, "step": 82835 }, { "epoch": 1.93, "learning_rate": 1.5601429208420363e-05, "loss": 0.1186, "step": 82840 }, { "epoch": 1.93, "learning_rate": 1.5600549153974325e-05, "loss": 0.1448, "step": 82845 }, { "epoch": 1.93, "learning_rate": 1.559966909952829e-05, "loss": 0.1487, "step": 82850 }, { "epoch": 1.93, "learning_rate": 1.559878904508226e-05, "loss": 0.3485, "step": 82855 }, { "epoch": 1.93, "learning_rate": 1.5597908990636223e-05, "loss": 0.2451, "step": 82860 }, { "epoch": 1.93, "learning_rate": 1.5597028936190185e-05, "loss": 0.2834, "step": 82865 }, { "epoch": 1.93, "learning_rate": 1.559614888174415e-05, "loss": 0.076, "step": 82870 }, { "epoch": 1.93, "learning_rate": 1.559526882729812e-05, "loss": 0.0559, "step": 82875 }, { "epoch": 1.93, "learning_rate": 1.559438877285208e-05, "loss": 0.0772, "step": 82880 }, { "epoch": 1.93, "learning_rate": 1.5593508718406045e-05, "loss": 0.065, "step": 82885 }, { "epoch": 1.93, "learning_rate": 1.559262866396001e-05, "loss": 0.037, "step": 82890 }, { "epoch": 1.93, "learning_rate": 1.559174860951398e-05, "loss": 0.0586, "step": 82895 }, { "epoch": 1.93, "learning_rate": 1.559086855506794e-05, "loss": 0.0778, "step": 82900 }, { "epoch": 1.93, "learning_rate": 1.5589988500621905e-05, "loss": 0.1534, "step": 82905 }, { "epoch": 1.93, "learning_rate": 1.558910844617587e-05, "loss": 0.2417, "step": 82910 }, { "epoch": 1.93, "learning_rate": 1.5588228391729835e-05, "loss": 0.2312, "step": 82915 }, { "epoch": 1.93, "learning_rate": 1.55873483372838e-05, "loss": 0.0025, "step": 82920 }, { "epoch": 1.93, "learning_rate": 1.5586468282837765e-05, "loss": 0.0586, "step": 82925 }, { "epoch": 1.93, "learning_rate": 1.558558822839173e-05, "loss": 0.0408, "step": 82930 }, { "epoch": 1.93, "learning_rate": 1.5584708173945695e-05, "loss": 0.1914, "step": 82935 }, { "epoch": 1.94, "learning_rate": 1.558382811949966e-05, "loss": 0.1823, "step": 82940 }, { "epoch": 1.94, "learning_rate": 1.5582948065053625e-05, "loss": 0.0288, "step": 82945 }, { "epoch": 1.94, "learning_rate": 1.5582068010607587e-05, "loss": 0.1121, "step": 82950 }, { "epoch": 1.94, "learning_rate": 1.5581187956161555e-05, "loss": 0.097, "step": 82955 }, { "epoch": 1.94, "learning_rate": 1.558030790171552e-05, "loss": 0.4391, "step": 82960 }, { "epoch": 1.94, "learning_rate": 1.5579427847269485e-05, "loss": 0.0975, "step": 82965 }, { "epoch": 1.94, "learning_rate": 1.557854779282345e-05, "loss": 0.1073, "step": 82970 }, { "epoch": 1.94, "learning_rate": 1.5577667738377415e-05, "loss": 0.0034, "step": 82975 }, { "epoch": 1.94, "learning_rate": 1.557678768393138e-05, "loss": 0.021, "step": 82980 }, { "epoch": 1.94, "learning_rate": 1.5575907629485342e-05, "loss": 0.0311, "step": 82985 }, { "epoch": 1.94, "learning_rate": 1.557502757503931e-05, "loss": 0.0387, "step": 82990 }, { "epoch": 1.94, "learning_rate": 1.5574147520593276e-05, "loss": 0.1892, "step": 82995 }, { "epoch": 1.94, "learning_rate": 1.557326746614724e-05, "loss": 0.1247, "step": 83000 }, { "epoch": 1.94, "learning_rate": 1.5572387411701202e-05, "loss": 0.1819, "step": 83005 }, { "epoch": 1.94, "learning_rate": 1.557150735725517e-05, "loss": 0.3684, "step": 83010 }, { "epoch": 1.94, "learning_rate": 1.5570627302809136e-05, "loss": 0.1665, "step": 83015 }, { "epoch": 1.94, "learning_rate": 1.5569747248363097e-05, "loss": 0.0385, "step": 83020 }, { "epoch": 1.94, "learning_rate": 1.5568867193917062e-05, "loss": 0.1306, "step": 83025 }, { "epoch": 1.94, "learning_rate": 1.556798713947103e-05, "loss": 0.1164, "step": 83030 }, { "epoch": 1.94, "learning_rate": 1.5567107085024996e-05, "loss": 0.0493, "step": 83035 }, { "epoch": 1.94, "learning_rate": 1.5566227030578957e-05, "loss": 0.0418, "step": 83040 }, { "epoch": 1.94, "learning_rate": 1.5565346976132922e-05, "loss": 0.0882, "step": 83045 }, { "epoch": 1.94, "learning_rate": 1.556446692168689e-05, "loss": 0.2033, "step": 83050 }, { "epoch": 1.94, "learning_rate": 1.5563586867240852e-05, "loss": 0.2319, "step": 83055 }, { "epoch": 1.94, "learning_rate": 1.5562706812794817e-05, "loss": 0.3298, "step": 83060 }, { "epoch": 1.94, "learning_rate": 1.5561826758348782e-05, "loss": 0.1548, "step": 83065 }, { "epoch": 1.94, "learning_rate": 1.556094670390275e-05, "loss": 0.0432, "step": 83070 }, { "epoch": 1.94, "learning_rate": 1.5560066649456712e-05, "loss": 0.0988, "step": 83075 }, { "epoch": 1.94, "learning_rate": 1.5559186595010677e-05, "loss": 0.0242, "step": 83080 }, { "epoch": 1.94, "learning_rate": 1.5558306540564643e-05, "loss": 0.0525, "step": 83085 }, { "epoch": 1.94, "learning_rate": 1.5557426486118608e-05, "loss": 0.0759, "step": 83090 }, { "epoch": 1.94, "learning_rate": 1.5556546431672573e-05, "loss": 0.1097, "step": 83095 }, { "epoch": 1.94, "learning_rate": 1.5555666377226538e-05, "loss": 0.0391, "step": 83100 }, { "epoch": 1.94, "learning_rate": 1.5554786322780503e-05, "loss": 0.083, "step": 83105 }, { "epoch": 1.94, "learning_rate": 1.5553906268334468e-05, "loss": 0.1299, "step": 83110 }, { "epoch": 1.94, "learning_rate": 1.5553026213888433e-05, "loss": 0.1179, "step": 83115 }, { "epoch": 1.94, "learning_rate": 1.5552146159442398e-05, "loss": 0.1014, "step": 83120 }, { "epoch": 1.94, "learning_rate": 1.5551266104996366e-05, "loss": 0.0173, "step": 83125 }, { "epoch": 1.94, "learning_rate": 1.5550386050550328e-05, "loss": 0.1225, "step": 83130 }, { "epoch": 1.94, "learning_rate": 1.5549505996104293e-05, "loss": 0.0359, "step": 83135 }, { "epoch": 1.94, "learning_rate": 1.5548625941658258e-05, "loss": 0.0608, "step": 83140 }, { "epoch": 1.94, "learning_rate": 1.5547745887212223e-05, "loss": 0.0715, "step": 83145 }, { "epoch": 1.94, "learning_rate": 1.5546865832766188e-05, "loss": 0.1834, "step": 83150 }, { "epoch": 1.94, "learning_rate": 1.5545985778320153e-05, "loss": 0.2694, "step": 83155 }, { "epoch": 1.94, "learning_rate": 1.5545105723874118e-05, "loss": 0.3063, "step": 83160 }, { "epoch": 1.94, "learning_rate": 1.5544225669428083e-05, "loss": 0.1712, "step": 83165 }, { "epoch": 1.94, "learning_rate": 1.5543345614982048e-05, "loss": 0.0067, "step": 83170 }, { "epoch": 1.94, "learning_rate": 1.5542465560536013e-05, "loss": 0.0478, "step": 83175 }, { "epoch": 1.94, "learning_rate": 1.5541585506089975e-05, "loss": 0.089, "step": 83180 }, { "epoch": 1.94, "learning_rate": 1.5540705451643943e-05, "loss": 0.1226, "step": 83185 }, { "epoch": 1.94, "learning_rate": 1.5539825397197908e-05, "loss": 0.155, "step": 83190 }, { "epoch": 1.94, "learning_rate": 1.5538945342751873e-05, "loss": 0.0876, "step": 83195 }, { "epoch": 1.94, "learning_rate": 1.5538065288305835e-05, "loss": 0.0938, "step": 83200 }, { "epoch": 1.94, "learning_rate": 1.5537185233859803e-05, "loss": 0.2453, "step": 83205 }, { "epoch": 1.94, "learning_rate": 1.5536305179413768e-05, "loss": 0.4302, "step": 83210 }, { "epoch": 1.94, "learning_rate": 1.553542512496773e-05, "loss": 0.4542, "step": 83215 }, { "epoch": 1.94, "learning_rate": 1.5534545070521695e-05, "loss": 0.0399, "step": 83220 }, { "epoch": 1.94, "learning_rate": 1.5533665016075663e-05, "loss": 0.0491, "step": 83225 }, { "epoch": 1.94, "learning_rate": 1.5532784961629628e-05, "loss": 0.0605, "step": 83230 }, { "epoch": 1.94, "learning_rate": 1.553190490718359e-05, "loss": 0.0533, "step": 83235 }, { "epoch": 1.94, "learning_rate": 1.5531024852737555e-05, "loss": 0.1004, "step": 83240 }, { "epoch": 1.94, "learning_rate": 1.5530144798291523e-05, "loss": 0.0854, "step": 83245 }, { "epoch": 1.94, "learning_rate": 1.5529264743845485e-05, "loss": 0.2053, "step": 83250 }, { "epoch": 1.94, "learning_rate": 1.552838468939945e-05, "loss": 0.1626, "step": 83255 }, { "epoch": 1.94, "learning_rate": 1.5527504634953418e-05, "loss": 0.4052, "step": 83260 }, { "epoch": 1.94, "learning_rate": 1.5526624580507383e-05, "loss": 0.1917, "step": 83265 }, { "epoch": 1.94, "learning_rate": 1.5525744526061345e-05, "loss": 0.0085, "step": 83270 }, { "epoch": 1.94, "learning_rate": 1.552486447161531e-05, "loss": 0.026, "step": 83275 }, { "epoch": 1.94, "learning_rate": 1.552398441716928e-05, "loss": 0.0916, "step": 83280 }, { "epoch": 1.94, "learning_rate": 1.552310436272324e-05, "loss": 0.0868, "step": 83285 }, { "epoch": 1.94, "learning_rate": 1.5522224308277205e-05, "loss": 0.0934, "step": 83290 }, { "epoch": 1.94, "learning_rate": 1.552134425383117e-05, "loss": 0.0811, "step": 83295 }, { "epoch": 1.94, "learning_rate": 1.552046419938514e-05, "loss": 0.1988, "step": 83300 }, { "epoch": 1.94, "learning_rate": 1.55195841449391e-05, "loss": 0.1043, "step": 83305 }, { "epoch": 1.94, "learning_rate": 1.5518704090493065e-05, "loss": 0.1999, "step": 83310 }, { "epoch": 1.94, "learning_rate": 1.551782403604703e-05, "loss": 0.0897, "step": 83315 }, { "epoch": 1.94, "learning_rate": 1.5516943981600995e-05, "loss": 0.0062, "step": 83320 }, { "epoch": 1.94, "learning_rate": 1.551606392715496e-05, "loss": 0.0438, "step": 83325 }, { "epoch": 1.94, "learning_rate": 1.5515183872708925e-05, "loss": 0.084, "step": 83330 }, { "epoch": 1.94, "learning_rate": 1.551430381826289e-05, "loss": 0.0226, "step": 83335 }, { "epoch": 1.94, "learning_rate": 1.5513423763816855e-05, "loss": 0.1589, "step": 83340 }, { "epoch": 1.94, "learning_rate": 1.551254370937082e-05, "loss": 0.0892, "step": 83345 }, { "epoch": 1.94, "learning_rate": 1.5511663654924785e-05, "loss": 0.1054, "step": 83350 }, { "epoch": 1.94, "learning_rate": 1.5510783600478747e-05, "loss": 0.2004, "step": 83355 }, { "epoch": 1.94, "learning_rate": 1.5509903546032715e-05, "loss": 0.2917, "step": 83360 }, { "epoch": 1.94, "learning_rate": 1.550902349158668e-05, "loss": 0.1779, "step": 83365 }, { "epoch": 1.95, "learning_rate": 1.5508143437140645e-05, "loss": 0.086, "step": 83370 }, { "epoch": 1.95, "learning_rate": 1.550726338269461e-05, "loss": 0.0152, "step": 83375 }, { "epoch": 1.95, "learning_rate": 1.5506383328248575e-05, "loss": 0.0053, "step": 83380 }, { "epoch": 1.95, "learning_rate": 1.550550327380254e-05, "loss": 0.0754, "step": 83385 }, { "epoch": 1.95, "learning_rate": 1.5504623219356502e-05, "loss": 0.0781, "step": 83390 }, { "epoch": 1.95, "learning_rate": 1.550374316491047e-05, "loss": 0.0155, "step": 83395 }, { "epoch": 1.95, "learning_rate": 1.5502863110464435e-05, "loss": 0.0692, "step": 83400 }, { "epoch": 1.95, "learning_rate": 1.55019830560184e-05, "loss": 0.0858, "step": 83405 }, { "epoch": 1.95, "learning_rate": 1.5501103001572362e-05, "loss": 0.252, "step": 83410 }, { "epoch": 1.95, "learning_rate": 1.550022294712633e-05, "loss": 0.2042, "step": 83415 }, { "epoch": 1.95, "learning_rate": 1.5499342892680296e-05, "loss": 0.0554, "step": 83420 }, { "epoch": 1.95, "learning_rate": 1.549846283823426e-05, "loss": 0.0953, "step": 83425 }, { "epoch": 1.95, "learning_rate": 1.5497582783788222e-05, "loss": 0.1584, "step": 83430 }, { "epoch": 1.95, "learning_rate": 1.549670272934219e-05, "loss": 0.0419, "step": 83435 }, { "epoch": 1.95, "learning_rate": 1.5495822674896156e-05, "loss": 0.0947, "step": 83440 }, { "epoch": 1.95, "learning_rate": 1.5494942620450117e-05, "loss": 0.0637, "step": 83445 }, { "epoch": 1.95, "learning_rate": 1.5494062566004082e-05, "loss": 0.1315, "step": 83450 }, { "epoch": 1.95, "learning_rate": 1.549318251155805e-05, "loss": 0.3543, "step": 83455 }, { "epoch": 1.95, "learning_rate": 1.5492302457112016e-05, "loss": 0.1855, "step": 83460 }, { "epoch": 1.95, "learning_rate": 1.5491422402665977e-05, "loss": 0.4025, "step": 83465 }, { "epoch": 1.95, "learning_rate": 1.5490542348219942e-05, "loss": 0.0311, "step": 83470 }, { "epoch": 1.95, "learning_rate": 1.548966229377391e-05, "loss": 0.0195, "step": 83475 }, { "epoch": 1.95, "learning_rate": 1.5488782239327872e-05, "loss": 0.0624, "step": 83480 }, { "epoch": 1.95, "learning_rate": 1.5487902184881837e-05, "loss": 0.0985, "step": 83485 }, { "epoch": 1.95, "learning_rate": 1.5487022130435802e-05, "loss": 0.0968, "step": 83490 }, { "epoch": 1.95, "learning_rate": 1.548614207598977e-05, "loss": 0.1378, "step": 83495 }, { "epoch": 1.95, "learning_rate": 1.5485262021543732e-05, "loss": 0.1037, "step": 83500 }, { "epoch": 1.95, "learning_rate": 1.5484381967097697e-05, "loss": 0.0815, "step": 83505 }, { "epoch": 1.95, "learning_rate": 1.5483501912651662e-05, "loss": 0.2534, "step": 83510 }, { "epoch": 1.95, "learning_rate": 1.5482621858205628e-05, "loss": 0.0893, "step": 83515 }, { "epoch": 1.95, "learning_rate": 1.5481741803759593e-05, "loss": 0.0298, "step": 83520 }, { "epoch": 1.95, "learning_rate": 1.5480861749313558e-05, "loss": 0.0695, "step": 83525 }, { "epoch": 1.95, "learning_rate": 1.5479981694867526e-05, "loss": 0.0074, "step": 83530 }, { "epoch": 1.95, "learning_rate": 1.5479101640421488e-05, "loss": 0.08, "step": 83535 }, { "epoch": 1.95, "learning_rate": 1.5478221585975453e-05, "loss": 0.0503, "step": 83540 }, { "epoch": 1.95, "learning_rate": 1.5477341531529418e-05, "loss": 0.1032, "step": 83545 }, { "epoch": 1.95, "learning_rate": 1.5476461477083383e-05, "loss": 0.1406, "step": 83550 }, { "epoch": 1.95, "learning_rate": 1.5475581422637348e-05, "loss": 0.189, "step": 83555 }, { "epoch": 1.95, "learning_rate": 1.5474701368191313e-05, "loss": 0.2061, "step": 83560 }, { "epoch": 1.95, "learning_rate": 1.5473821313745278e-05, "loss": 0.1632, "step": 83565 }, { "epoch": 1.95, "learning_rate": 1.5472941259299243e-05, "loss": 0.0662, "step": 83570 }, { "epoch": 1.95, "learning_rate": 1.5472061204853208e-05, "loss": 0.0078, "step": 83575 }, { "epoch": 1.95, "learning_rate": 1.5471181150407173e-05, "loss": 0.0357, "step": 83580 }, { "epoch": 1.95, "learning_rate": 1.5470301095961134e-05, "loss": 0.0639, "step": 83585 }, { "epoch": 1.95, "learning_rate": 1.5469421041515103e-05, "loss": 0.0928, "step": 83590 }, { "epoch": 1.95, "learning_rate": 1.5468540987069068e-05, "loss": 0.0973, "step": 83595 }, { "epoch": 1.95, "learning_rate": 1.5467660932623033e-05, "loss": 0.0091, "step": 83600 }, { "epoch": 1.95, "learning_rate": 1.5466780878176994e-05, "loss": 0.1064, "step": 83605 }, { "epoch": 1.95, "learning_rate": 1.5465900823730963e-05, "loss": 0.0591, "step": 83610 }, { "epoch": 1.95, "learning_rate": 1.5465020769284928e-05, "loss": 0.1802, "step": 83615 }, { "epoch": 1.95, "learning_rate": 1.546414071483889e-05, "loss": 0.1327, "step": 83620 }, { "epoch": 1.95, "learning_rate": 1.5463260660392855e-05, "loss": 0.1652, "step": 83625 }, { "epoch": 1.95, "learning_rate": 1.5462380605946823e-05, "loss": 0.0469, "step": 83630 }, { "epoch": 1.95, "learning_rate": 1.5461500551500788e-05, "loss": 0.0872, "step": 83635 }, { "epoch": 1.95, "learning_rate": 1.546062049705475e-05, "loss": 0.0409, "step": 83640 }, { "epoch": 1.95, "learning_rate": 1.5459740442608715e-05, "loss": 0.2714, "step": 83645 }, { "epoch": 1.95, "learning_rate": 1.5458860388162683e-05, "loss": 0.0452, "step": 83650 }, { "epoch": 1.95, "learning_rate": 1.5457980333716645e-05, "loss": 0.1087, "step": 83655 }, { "epoch": 1.95, "learning_rate": 1.545710027927061e-05, "loss": 0.1754, "step": 83660 }, { "epoch": 1.95, "learning_rate": 1.5456220224824578e-05, "loss": 0.2825, "step": 83665 }, { "epoch": 1.95, "learning_rate": 1.5455340170378543e-05, "loss": 0.0192, "step": 83670 }, { "epoch": 1.95, "learning_rate": 1.5454460115932505e-05, "loss": 0.0387, "step": 83675 }, { "epoch": 1.95, "learning_rate": 1.545358006148647e-05, "loss": 0.1381, "step": 83680 }, { "epoch": 1.95, "learning_rate": 1.5452700007040438e-05, "loss": 0.0813, "step": 83685 }, { "epoch": 1.95, "learning_rate": 1.54518199525944e-05, "loss": 0.1124, "step": 83690 }, { "epoch": 1.95, "learning_rate": 1.5450939898148365e-05, "loss": 0.132, "step": 83695 }, { "epoch": 1.95, "learning_rate": 1.545005984370233e-05, "loss": 0.1615, "step": 83700 }, { "epoch": 1.95, "learning_rate": 1.5449179789256298e-05, "loss": 0.1146, "step": 83705 }, { "epoch": 1.95, "learning_rate": 1.544829973481026e-05, "loss": 0.4876, "step": 83710 }, { "epoch": 1.95, "learning_rate": 1.5447419680364225e-05, "loss": 0.1723, "step": 83715 }, { "epoch": 1.95, "learning_rate": 1.544653962591819e-05, "loss": 0.1153, "step": 83720 }, { "epoch": 1.95, "learning_rate": 1.544565957147216e-05, "loss": 0.0262, "step": 83725 }, { "epoch": 1.95, "learning_rate": 1.544477951702612e-05, "loss": 0.0356, "step": 83730 }, { "epoch": 1.95, "learning_rate": 1.5443899462580085e-05, "loss": 0.0684, "step": 83735 }, { "epoch": 1.95, "learning_rate": 1.544301940813405e-05, "loss": 0.0237, "step": 83740 }, { "epoch": 1.95, "learning_rate": 1.5442139353688015e-05, "loss": 0.1154, "step": 83745 }, { "epoch": 1.95, "learning_rate": 1.544125929924198e-05, "loss": 0.0938, "step": 83750 }, { "epoch": 1.95, "learning_rate": 1.5440379244795945e-05, "loss": 0.1505, "step": 83755 }, { "epoch": 1.95, "learning_rate": 1.543949919034991e-05, "loss": 0.4108, "step": 83760 }, { "epoch": 1.95, "learning_rate": 1.5438619135903875e-05, "loss": 0.3079, "step": 83765 }, { "epoch": 1.95, "learning_rate": 1.543773908145784e-05, "loss": 0.0449, "step": 83770 }, { "epoch": 1.95, "learning_rate": 1.5436859027011805e-05, "loss": 0.1926, "step": 83775 }, { "epoch": 1.95, "learning_rate": 1.5435978972565767e-05, "loss": 0.0882, "step": 83780 }, { "epoch": 1.95, "learning_rate": 1.5435098918119735e-05, "loss": 0.0077, "step": 83785 }, { "epoch": 1.95, "learning_rate": 1.54342188636737e-05, "loss": 0.0342, "step": 83790 }, { "epoch": 1.96, "learning_rate": 1.5433338809227665e-05, "loss": 0.0665, "step": 83795 }, { "epoch": 1.96, "learning_rate": 1.543245875478163e-05, "loss": 0.2147, "step": 83800 }, { "epoch": 1.96, "learning_rate": 1.5431578700335595e-05, "loss": 0.1477, "step": 83805 }, { "epoch": 1.96, "learning_rate": 1.543069864588956e-05, "loss": 0.1421, "step": 83810 }, { "epoch": 1.96, "learning_rate": 1.5429818591443522e-05, "loss": 0.2111, "step": 83815 }, { "epoch": 1.96, "learning_rate": 1.542893853699749e-05, "loss": 0.0334, "step": 83820 }, { "epoch": 1.96, "learning_rate": 1.5428058482551455e-05, "loss": 0.0353, "step": 83825 }, { "epoch": 1.96, "learning_rate": 1.542717842810542e-05, "loss": 0.0962, "step": 83830 }, { "epoch": 1.96, "learning_rate": 1.5426298373659382e-05, "loss": 0.0993, "step": 83835 }, { "epoch": 1.96, "learning_rate": 1.542541831921335e-05, "loss": 0.1562, "step": 83840 }, { "epoch": 1.96, "learning_rate": 1.5424538264767315e-05, "loss": 0.0439, "step": 83845 }, { "epoch": 1.96, "learning_rate": 1.5423658210321277e-05, "loss": 0.1826, "step": 83850 }, { "epoch": 1.96, "learning_rate": 1.5422778155875242e-05, "loss": 0.1569, "step": 83855 }, { "epoch": 1.96, "learning_rate": 1.542189810142921e-05, "loss": 0.3114, "step": 83860 }, { "epoch": 1.96, "learning_rate": 1.5421018046983176e-05, "loss": 0.0903, "step": 83865 }, { "epoch": 1.96, "learning_rate": 1.5420137992537137e-05, "loss": 0.0381, "step": 83870 }, { "epoch": 1.96, "learning_rate": 1.5419257938091102e-05, "loss": 0.128, "step": 83875 }, { "epoch": 1.96, "learning_rate": 1.541837788364507e-05, "loss": 0.0425, "step": 83880 }, { "epoch": 1.96, "learning_rate": 1.5417497829199032e-05, "loss": 0.1346, "step": 83885 }, { "epoch": 1.96, "learning_rate": 1.5416617774752997e-05, "loss": 0.0901, "step": 83890 }, { "epoch": 1.96, "learning_rate": 1.5415737720306962e-05, "loss": 0.101, "step": 83895 }, { "epoch": 1.96, "learning_rate": 1.541485766586093e-05, "loss": 0.0933, "step": 83900 }, { "epoch": 1.96, "learning_rate": 1.5413977611414892e-05, "loss": 0.1209, "step": 83905 }, { "epoch": 1.96, "learning_rate": 1.5413097556968857e-05, "loss": 0.298, "step": 83910 }, { "epoch": 1.96, "learning_rate": 1.5412217502522822e-05, "loss": 0.3647, "step": 83915 }, { "epoch": 1.96, "learning_rate": 1.5411337448076787e-05, "loss": 0.0111, "step": 83920 }, { "epoch": 1.96, "learning_rate": 1.5410457393630752e-05, "loss": 0.0911, "step": 83925 }, { "epoch": 1.96, "learning_rate": 1.5409577339184717e-05, "loss": 0.0879, "step": 83930 }, { "epoch": 1.96, "learning_rate": 1.5408697284738686e-05, "loss": 0.1124, "step": 83935 }, { "epoch": 1.96, "learning_rate": 1.5407817230292647e-05, "loss": 0.0174, "step": 83940 }, { "epoch": 1.96, "learning_rate": 1.5406937175846612e-05, "loss": 0.1783, "step": 83945 }, { "epoch": 1.96, "learning_rate": 1.5406057121400578e-05, "loss": 0.2273, "step": 83950 }, { "epoch": 1.96, "learning_rate": 1.5405177066954543e-05, "loss": 0.0806, "step": 83955 }, { "epoch": 1.96, "learning_rate": 1.5404297012508508e-05, "loss": 0.0988, "step": 83960 }, { "epoch": 1.96, "learning_rate": 1.5403416958062473e-05, "loss": 0.1926, "step": 83965 }, { "epoch": 1.96, "learning_rate": 1.5402536903616438e-05, "loss": 0.0111, "step": 83970 }, { "epoch": 1.96, "learning_rate": 1.5401656849170403e-05, "loss": 0.0963, "step": 83975 }, { "epoch": 1.96, "learning_rate": 1.5400776794724368e-05, "loss": 0.1169, "step": 83980 }, { "epoch": 1.96, "learning_rate": 1.5399896740278333e-05, "loss": 0.1086, "step": 83985 }, { "epoch": 1.96, "learning_rate": 1.5399016685832294e-05, "loss": 0.1454, "step": 83990 }, { "epoch": 1.96, "learning_rate": 1.5398136631386263e-05, "loss": 0.041, "step": 83995 }, { "epoch": 1.96, "learning_rate": 1.5397256576940228e-05, "loss": 0.0781, "step": 84000 }, { "epoch": 1.96, "learning_rate": 1.5396376522494193e-05, "loss": 0.1307, "step": 84005 }, { "epoch": 1.96, "learning_rate": 1.5395496468048154e-05, "loss": 0.2125, "step": 84010 }, { "epoch": 1.96, "learning_rate": 1.5394616413602123e-05, "loss": 0.2481, "step": 84015 }, { "epoch": 1.96, "learning_rate": 1.5393736359156088e-05, "loss": 0.0736, "step": 84020 }, { "epoch": 1.96, "learning_rate": 1.539285630471005e-05, "loss": 0.07, "step": 84025 }, { "epoch": 1.96, "learning_rate": 1.5391976250264014e-05, "loss": 0.0127, "step": 84030 }, { "epoch": 1.96, "learning_rate": 1.5391096195817983e-05, "loss": 0.0452, "step": 84035 }, { "epoch": 1.96, "learning_rate": 1.5390216141371948e-05, "loss": 0.0707, "step": 84040 }, { "epoch": 1.96, "learning_rate": 1.538933608692591e-05, "loss": 0.0613, "step": 84045 }, { "epoch": 1.96, "learning_rate": 1.5388456032479875e-05, "loss": 0.1188, "step": 84050 }, { "epoch": 1.96, "learning_rate": 1.5387575978033843e-05, "loss": 0.0796, "step": 84055 }, { "epoch": 1.96, "learning_rate": 1.5386695923587808e-05, "loss": 0.1456, "step": 84060 }, { "epoch": 1.96, "learning_rate": 1.538581586914177e-05, "loss": 0.1841, "step": 84065 }, { "epoch": 1.96, "learning_rate": 1.5384935814695738e-05, "loss": 0.0856, "step": 84070 }, { "epoch": 1.96, "learning_rate": 1.5384055760249703e-05, "loss": 0.0249, "step": 84075 }, { "epoch": 1.96, "learning_rate": 1.5383175705803665e-05, "loss": 0.0696, "step": 84080 }, { "epoch": 1.96, "learning_rate": 1.538229565135763e-05, "loss": 0.1243, "step": 84085 }, { "epoch": 1.96, "learning_rate": 1.5381415596911598e-05, "loss": 0.0695, "step": 84090 }, { "epoch": 1.96, "learning_rate": 1.5380535542465563e-05, "loss": 0.0622, "step": 84095 }, { "epoch": 1.96, "learning_rate": 1.5379655488019525e-05, "loss": 0.1856, "step": 84100 }, { "epoch": 1.96, "learning_rate": 1.537877543357349e-05, "loss": 0.13, "step": 84105 }, { "epoch": 1.96, "learning_rate": 1.5377895379127458e-05, "loss": 0.1521, "step": 84110 }, { "epoch": 1.96, "learning_rate": 1.537701532468142e-05, "loss": 0.2981, "step": 84115 }, { "epoch": 1.96, "learning_rate": 1.5376135270235385e-05, "loss": 0.0328, "step": 84120 }, { "epoch": 1.96, "learning_rate": 1.537525521578935e-05, "loss": 0.0934, "step": 84125 }, { "epoch": 1.96, "learning_rate": 1.5374375161343318e-05, "loss": 0.1258, "step": 84130 }, { "epoch": 1.96, "learning_rate": 1.537349510689728e-05, "loss": 0.1092, "step": 84135 }, { "epoch": 1.96, "learning_rate": 1.5372615052451245e-05, "loss": 0.0393, "step": 84140 }, { "epoch": 1.96, "learning_rate": 1.537173499800521e-05, "loss": 0.1083, "step": 84145 }, { "epoch": 1.96, "learning_rate": 1.5370854943559175e-05, "loss": 0.092, "step": 84150 }, { "epoch": 1.96, "learning_rate": 1.536997488911314e-05, "loss": 0.2066, "step": 84155 }, { "epoch": 1.96, "learning_rate": 1.5369094834667105e-05, "loss": 0.1837, "step": 84160 }, { "epoch": 1.96, "learning_rate": 1.536821478022107e-05, "loss": 0.126, "step": 84165 }, { "epoch": 1.96, "learning_rate": 1.5367334725775035e-05, "loss": 0.044, "step": 84170 }, { "epoch": 1.96, "learning_rate": 1.5366454671329e-05, "loss": 0.0462, "step": 84175 }, { "epoch": 1.96, "learning_rate": 1.5365574616882965e-05, "loss": 0.0158, "step": 84180 }, { "epoch": 1.96, "learning_rate": 1.5364694562436927e-05, "loss": 0.0564, "step": 84185 }, { "epoch": 1.96, "learning_rate": 1.5363814507990895e-05, "loss": 0.1502, "step": 84190 }, { "epoch": 1.96, "learning_rate": 1.536293445354486e-05, "loss": 0.094, "step": 84195 }, { "epoch": 1.96, "learning_rate": 1.5362054399098825e-05, "loss": 0.1962, "step": 84200 }, { "epoch": 1.96, "learning_rate": 1.536117434465279e-05, "loss": 0.03, "step": 84205 }, { "epoch": 1.96, "learning_rate": 1.5360294290206755e-05, "loss": 0.3635, "step": 84210 }, { "epoch": 1.96, "learning_rate": 1.535941423576072e-05, "loss": 0.3176, "step": 84215 }, { "epoch": 1.96, "learning_rate": 1.5358534181314682e-05, "loss": 0.017, "step": 84220 }, { "epoch": 1.97, "learning_rate": 1.535765412686865e-05, "loss": 0.0301, "step": 84225 }, { "epoch": 1.97, "learning_rate": 1.5356774072422615e-05, "loss": 0.0347, "step": 84230 }, { "epoch": 1.97, "learning_rate": 1.535589401797658e-05, "loss": 0.1092, "step": 84235 }, { "epoch": 1.97, "learning_rate": 1.5355013963530542e-05, "loss": 0.0915, "step": 84240 }, { "epoch": 1.97, "learning_rate": 1.535413390908451e-05, "loss": 0.0479, "step": 84245 }, { "epoch": 1.97, "learning_rate": 1.5353253854638475e-05, "loss": 0.1494, "step": 84250 }, { "epoch": 1.97, "learning_rate": 1.5352373800192437e-05, "loss": 0.0594, "step": 84255 }, { "epoch": 1.97, "learning_rate": 1.5351493745746402e-05, "loss": 0.1338, "step": 84260 }, { "epoch": 1.97, "learning_rate": 1.535061369130037e-05, "loss": 0.0941, "step": 84265 }, { "epoch": 1.97, "learning_rate": 1.5349733636854335e-05, "loss": 0.0443, "step": 84270 }, { "epoch": 1.97, "learning_rate": 1.5348853582408297e-05, "loss": 0.0042, "step": 84275 }, { "epoch": 1.97, "learning_rate": 1.5347973527962262e-05, "loss": 0.2186, "step": 84280 }, { "epoch": 1.97, "learning_rate": 1.534709347351623e-05, "loss": 0.1102, "step": 84285 }, { "epoch": 1.97, "learning_rate": 1.5346213419070192e-05, "loss": 0.1576, "step": 84290 }, { "epoch": 1.97, "learning_rate": 1.5345333364624157e-05, "loss": 0.0568, "step": 84295 }, { "epoch": 1.97, "learning_rate": 1.5344453310178122e-05, "loss": 0.1559, "step": 84300 }, { "epoch": 1.97, "learning_rate": 1.534357325573209e-05, "loss": 0.1974, "step": 84305 }, { "epoch": 1.97, "learning_rate": 1.5342693201286052e-05, "loss": 0.1032, "step": 84310 }, { "epoch": 1.97, "learning_rate": 1.5341813146840017e-05, "loss": 0.1388, "step": 84315 }, { "epoch": 1.97, "learning_rate": 1.5340933092393982e-05, "loss": 0.0423, "step": 84320 }, { "epoch": 1.97, "learning_rate": 1.5340053037947947e-05, "loss": 0.0213, "step": 84325 }, { "epoch": 1.97, "learning_rate": 1.5339172983501912e-05, "loss": 0.0558, "step": 84330 }, { "epoch": 1.97, "learning_rate": 1.5338292929055877e-05, "loss": 0.0112, "step": 84335 }, { "epoch": 1.97, "learning_rate": 1.5337412874609846e-05, "loss": 0.0326, "step": 84340 }, { "epoch": 1.97, "learning_rate": 1.5336532820163807e-05, "loss": 0.0428, "step": 84345 }, { "epoch": 1.97, "learning_rate": 1.5335652765717772e-05, "loss": 0.0957, "step": 84350 }, { "epoch": 1.97, "learning_rate": 1.5334772711271737e-05, "loss": 0.1075, "step": 84355 }, { "epoch": 1.97, "learning_rate": 1.5333892656825706e-05, "loss": 0.1083, "step": 84360 }, { "epoch": 1.97, "learning_rate": 1.5333012602379667e-05, "loss": 0.3033, "step": 84365 }, { "epoch": 1.97, "learning_rate": 1.5332132547933632e-05, "loss": 0.0007, "step": 84370 }, { "epoch": 1.97, "learning_rate": 1.5331252493487597e-05, "loss": 0.1397, "step": 84375 }, { "epoch": 1.97, "learning_rate": 1.5330372439041563e-05, "loss": 0.0486, "step": 84380 }, { "epoch": 1.97, "learning_rate": 1.5329492384595528e-05, "loss": 0.1327, "step": 84385 }, { "epoch": 1.97, "learning_rate": 1.5328612330149493e-05, "loss": 0.1093, "step": 84390 }, { "epoch": 1.97, "learning_rate": 1.5327732275703458e-05, "loss": 0.0408, "step": 84395 }, { "epoch": 1.97, "learning_rate": 1.5326852221257423e-05, "loss": 0.094, "step": 84400 }, { "epoch": 1.97, "learning_rate": 1.5325972166811388e-05, "loss": 0.2336, "step": 84405 }, { "epoch": 1.97, "learning_rate": 1.5325092112365353e-05, "loss": 0.1053, "step": 84410 }, { "epoch": 1.97, "learning_rate": 1.5324212057919314e-05, "loss": 0.1773, "step": 84415 }, { "epoch": 1.97, "learning_rate": 1.5323332003473283e-05, "loss": 0.0505, "step": 84420 }, { "epoch": 1.97, "learning_rate": 1.5322451949027248e-05, "loss": 0.0689, "step": 84425 }, { "epoch": 1.97, "learning_rate": 1.5321571894581213e-05, "loss": 0.0522, "step": 84430 }, { "epoch": 1.97, "learning_rate": 1.5320691840135174e-05, "loss": 0.1311, "step": 84435 }, { "epoch": 1.97, "learning_rate": 1.5319811785689143e-05, "loss": 0.0957, "step": 84440 }, { "epoch": 1.97, "learning_rate": 1.5318931731243108e-05, "loss": 0.0892, "step": 84445 }, { "epoch": 1.97, "learning_rate": 1.531805167679707e-05, "loss": 0.1444, "step": 84450 }, { "epoch": 1.97, "learning_rate": 1.5317171622351034e-05, "loss": 0.1708, "step": 84455 }, { "epoch": 1.97, "learning_rate": 1.5316291567905003e-05, "loss": 0.3356, "step": 84460 }, { "epoch": 1.97, "learning_rate": 1.5315411513458968e-05, "loss": 0.1249, "step": 84465 }, { "epoch": 1.97, "learning_rate": 1.531453145901293e-05, "loss": 0.0127, "step": 84470 }, { "epoch": 1.97, "learning_rate": 1.5313651404566898e-05, "loss": 0.0483, "step": 84475 }, { "epoch": 1.97, "learning_rate": 1.5312771350120863e-05, "loss": 0.0042, "step": 84480 }, { "epoch": 1.97, "learning_rate": 1.5311891295674825e-05, "loss": 0.0731, "step": 84485 }, { "epoch": 1.97, "learning_rate": 1.531101124122879e-05, "loss": 0.1213, "step": 84490 }, { "epoch": 1.97, "learning_rate": 1.5310131186782758e-05, "loss": 0.0769, "step": 84495 }, { "epoch": 1.97, "learning_rate": 1.5309251132336723e-05, "loss": 0.2166, "step": 84500 }, { "epoch": 1.97, "learning_rate": 1.5308371077890685e-05, "loss": 0.2013, "step": 84505 }, { "epoch": 1.97, "learning_rate": 1.530749102344465e-05, "loss": 0.3932, "step": 84510 }, { "epoch": 1.97, "learning_rate": 1.5306610968998618e-05, "loss": 0.4241, "step": 84515 }, { "epoch": 1.97, "learning_rate": 1.530573091455258e-05, "loss": 0.0147, "step": 84520 }, { "epoch": 1.97, "learning_rate": 1.5304850860106545e-05, "loss": 0.0504, "step": 84525 }, { "epoch": 1.97, "learning_rate": 1.530397080566051e-05, "loss": 0.0928, "step": 84530 }, { "epoch": 1.97, "learning_rate": 1.5303090751214478e-05, "loss": 0.0235, "step": 84535 }, { "epoch": 1.97, "learning_rate": 1.530221069676844e-05, "loss": 0.0217, "step": 84540 }, { "epoch": 1.97, "learning_rate": 1.5301330642322405e-05, "loss": 0.0528, "step": 84545 }, { "epoch": 1.97, "learning_rate": 1.530045058787637e-05, "loss": 0.1771, "step": 84550 }, { "epoch": 1.97, "learning_rate": 1.5299570533430335e-05, "loss": 0.2543, "step": 84555 }, { "epoch": 1.97, "learning_rate": 1.52986904789843e-05, "loss": 0.3009, "step": 84560 }, { "epoch": 1.97, "learning_rate": 1.5297810424538265e-05, "loss": 0.2166, "step": 84565 }, { "epoch": 1.97, "learning_rate": 1.529693037009223e-05, "loss": 0.0698, "step": 84570 }, { "epoch": 1.97, "learning_rate": 1.5296050315646195e-05, "loss": 0.0306, "step": 84575 }, { "epoch": 1.97, "learning_rate": 1.529517026120016e-05, "loss": 0.0601, "step": 84580 }, { "epoch": 1.97, "learning_rate": 1.5294290206754125e-05, "loss": 0.0789, "step": 84585 }, { "epoch": 1.97, "learning_rate": 1.5293410152308087e-05, "loss": 0.0082, "step": 84590 }, { "epoch": 1.97, "learning_rate": 1.5292530097862055e-05, "loss": 0.0969, "step": 84595 }, { "epoch": 1.97, "learning_rate": 1.529165004341602e-05, "loss": 0.129, "step": 84600 }, { "epoch": 1.97, "learning_rate": 1.5290769988969985e-05, "loss": 0.3489, "step": 84605 }, { "epoch": 1.97, "learning_rate": 1.528988993452395e-05, "loss": 0.1846, "step": 84610 }, { "epoch": 1.97, "learning_rate": 1.5289009880077915e-05, "loss": 0.4098, "step": 84615 }, { "epoch": 1.97, "learning_rate": 1.528812982563188e-05, "loss": 0.0505, "step": 84620 }, { "epoch": 1.97, "learning_rate": 1.5287249771185842e-05, "loss": 0.0921, "step": 84625 }, { "epoch": 1.97, "learning_rate": 1.528636971673981e-05, "loss": 0.1086, "step": 84630 }, { "epoch": 1.97, "learning_rate": 1.5285489662293775e-05, "loss": 0.147, "step": 84635 }, { "epoch": 1.97, "learning_rate": 1.528460960784774e-05, "loss": 0.1419, "step": 84640 }, { "epoch": 1.97, "learning_rate": 1.5283729553401702e-05, "loss": 0.1685, "step": 84645 }, { "epoch": 1.97, "learning_rate": 1.528284949895567e-05, "loss": 0.198, "step": 84650 }, { "epoch": 1.98, "learning_rate": 1.5281969444509635e-05, "loss": 0.0805, "step": 84655 }, { "epoch": 1.98, "learning_rate": 1.52810893900636e-05, "loss": 0.1573, "step": 84660 }, { "epoch": 1.98, "learning_rate": 1.5280209335617562e-05, "loss": 0.1873, "step": 84665 }, { "epoch": 1.98, "learning_rate": 1.527932928117153e-05, "loss": 0.0387, "step": 84670 }, { "epoch": 1.98, "learning_rate": 1.5278449226725495e-05, "loss": 0.0893, "step": 84675 }, { "epoch": 1.98, "learning_rate": 1.5277569172279457e-05, "loss": 0.1183, "step": 84680 }, { "epoch": 1.98, "learning_rate": 1.5276689117833422e-05, "loss": 0.0978, "step": 84685 }, { "epoch": 1.98, "learning_rate": 1.527580906338739e-05, "loss": 0.0895, "step": 84690 }, { "epoch": 1.98, "learning_rate": 1.5274929008941355e-05, "loss": 0.0368, "step": 84695 }, { "epoch": 1.98, "learning_rate": 1.5274048954495317e-05, "loss": 0.1388, "step": 84700 }, { "epoch": 1.98, "learning_rate": 1.5273168900049282e-05, "loss": 0.1122, "step": 84705 }, { "epoch": 1.98, "learning_rate": 1.527228884560325e-05, "loss": 0.4212, "step": 84710 }, { "epoch": 1.98, "learning_rate": 1.5271408791157212e-05, "loss": 0.1698, "step": 84715 }, { "epoch": 1.98, "learning_rate": 1.5270528736711177e-05, "loss": 0.021, "step": 84720 }, { "epoch": 1.98, "learning_rate": 1.5269648682265142e-05, "loss": 0.0073, "step": 84725 }, { "epoch": 1.98, "learning_rate": 1.526876862781911e-05, "loss": 0.0203, "step": 84730 }, { "epoch": 1.98, "learning_rate": 1.5267888573373072e-05, "loss": 0.0708, "step": 84735 }, { "epoch": 1.98, "learning_rate": 1.5267008518927037e-05, "loss": 0.0791, "step": 84740 }, { "epoch": 1.98, "learning_rate": 1.5266128464481006e-05, "loss": 0.2275, "step": 84745 }, { "epoch": 1.98, "learning_rate": 1.5265248410034967e-05, "loss": 0.1637, "step": 84750 }, { "epoch": 1.98, "learning_rate": 1.5264368355588932e-05, "loss": 0.2699, "step": 84755 }, { "epoch": 1.98, "learning_rate": 1.5263488301142897e-05, "loss": 0.2988, "step": 84760 }, { "epoch": 1.98, "learning_rate": 1.5262608246696866e-05, "loss": 0.4932, "step": 84765 }, { "epoch": 1.98, "learning_rate": 1.5261728192250827e-05, "loss": 0.0019, "step": 84770 }, { "epoch": 1.98, "learning_rate": 1.5260848137804792e-05, "loss": 0.0242, "step": 84775 }, { "epoch": 1.98, "learning_rate": 1.5259968083358757e-05, "loss": 0.0638, "step": 84780 }, { "epoch": 1.98, "learning_rate": 1.5259088028912722e-05, "loss": 0.0414, "step": 84785 }, { "epoch": 1.98, "learning_rate": 1.5258207974466687e-05, "loss": 0.0321, "step": 84790 }, { "epoch": 1.98, "learning_rate": 1.5257327920020652e-05, "loss": 0.0743, "step": 84795 }, { "epoch": 1.98, "learning_rate": 1.5256447865574619e-05, "loss": 0.1795, "step": 84800 }, { "epoch": 1.98, "learning_rate": 1.525556781112858e-05, "loss": 0.3857, "step": 84805 }, { "epoch": 1.98, "learning_rate": 1.5254687756682547e-05, "loss": 0.1652, "step": 84810 }, { "epoch": 1.98, "learning_rate": 1.5253807702236513e-05, "loss": 0.4482, "step": 84815 }, { "epoch": 1.98, "learning_rate": 1.5252927647790476e-05, "loss": 0.1584, "step": 84820 }, { "epoch": 1.98, "learning_rate": 1.5252047593344441e-05, "loss": 0.0335, "step": 84825 }, { "epoch": 1.98, "learning_rate": 1.5251167538898408e-05, "loss": 0.0903, "step": 84830 }, { "epoch": 1.98, "learning_rate": 1.5250287484452373e-05, "loss": 0.1991, "step": 84835 }, { "epoch": 1.98, "learning_rate": 1.5249407430006336e-05, "loss": 0.2152, "step": 84840 }, { "epoch": 1.98, "learning_rate": 1.5248527375560301e-05, "loss": 0.037, "step": 84845 }, { "epoch": 1.98, "learning_rate": 1.5247647321114268e-05, "loss": 0.2178, "step": 84850 }, { "epoch": 1.98, "learning_rate": 1.524676726666823e-05, "loss": 0.1938, "step": 84855 }, { "epoch": 1.98, "learning_rate": 1.5245887212222196e-05, "loss": 0.2164, "step": 84860 }, { "epoch": 1.98, "learning_rate": 1.5245007157776161e-05, "loss": 0.4828, "step": 84865 }, { "epoch": 1.98, "learning_rate": 1.5244127103330128e-05, "loss": 0.0206, "step": 84870 }, { "epoch": 1.98, "learning_rate": 1.5243247048884091e-05, "loss": 0.0222, "step": 84875 }, { "epoch": 1.98, "learning_rate": 1.5242366994438056e-05, "loss": 0.0629, "step": 84880 }, { "epoch": 1.98, "learning_rate": 1.5241486939992023e-05, "loss": 0.0204, "step": 84885 }, { "epoch": 1.98, "learning_rate": 1.5240606885545984e-05, "loss": 0.0875, "step": 84890 }, { "epoch": 1.98, "learning_rate": 1.5239726831099951e-05, "loss": 0.318, "step": 84895 }, { "epoch": 1.98, "learning_rate": 1.5238846776653916e-05, "loss": 0.066, "step": 84900 }, { "epoch": 1.98, "learning_rate": 1.5237966722207883e-05, "loss": 0.2532, "step": 84905 }, { "epoch": 1.98, "learning_rate": 1.5237086667761845e-05, "loss": 0.3294, "step": 84910 }, { "epoch": 1.98, "learning_rate": 1.5236206613315811e-05, "loss": 0.2452, "step": 84915 }, { "epoch": 1.98, "learning_rate": 1.5235326558869776e-05, "loss": 0.0916, "step": 84920 }, { "epoch": 1.98, "learning_rate": 1.523444650442374e-05, "loss": 0.0404, "step": 84925 }, { "epoch": 1.98, "learning_rate": 1.5233566449977705e-05, "loss": 0.0437, "step": 84930 }, { "epoch": 1.98, "learning_rate": 1.5232686395531671e-05, "loss": 0.0971, "step": 84935 }, { "epoch": 1.98, "learning_rate": 1.5231806341085636e-05, "loss": 0.0428, "step": 84940 }, { "epoch": 1.98, "learning_rate": 1.52309262866396e-05, "loss": 0.0909, "step": 84945 }, { "epoch": 1.98, "learning_rate": 1.5230046232193565e-05, "loss": 0.1445, "step": 84950 }, { "epoch": 1.98, "learning_rate": 1.5229166177747531e-05, "loss": 0.1732, "step": 84955 }, { "epoch": 1.98, "learning_rate": 1.5228286123301496e-05, "loss": 0.2252, "step": 84960 }, { "epoch": 1.98, "learning_rate": 1.522740606885546e-05, "loss": 0.1873, "step": 84965 }, { "epoch": 1.98, "learning_rate": 1.5226526014409425e-05, "loss": 0.0249, "step": 84970 }, { "epoch": 1.98, "learning_rate": 1.5225645959963391e-05, "loss": 0.0293, "step": 84975 }, { "epoch": 1.98, "learning_rate": 1.5224765905517353e-05, "loss": 0.0514, "step": 84980 }, { "epoch": 1.98, "learning_rate": 1.522388585107132e-05, "loss": 0.1088, "step": 84985 }, { "epoch": 1.98, "learning_rate": 1.5223005796625285e-05, "loss": 0.0431, "step": 84990 }, { "epoch": 1.98, "learning_rate": 1.5222125742179252e-05, "loss": 0.0057, "step": 84995 }, { "epoch": 1.98, "learning_rate": 1.5221245687733213e-05, "loss": 0.1136, "step": 85000 }, { "epoch": 1.98, "learning_rate": 1.522036563328718e-05, "loss": 0.136, "step": 85005 }, { "epoch": 1.98, "learning_rate": 1.5219485578841147e-05, "loss": 0.1989, "step": 85010 }, { "epoch": 1.98, "learning_rate": 1.5218605524395108e-05, "loss": 0.1411, "step": 85015 }, { "epoch": 1.98, "learning_rate": 1.5217725469949075e-05, "loss": 0.0272, "step": 85020 }, { "epoch": 1.98, "learning_rate": 1.521684541550304e-05, "loss": 0.0322, "step": 85025 }, { "epoch": 1.98, "learning_rate": 1.5215965361057007e-05, "loss": 0.0742, "step": 85030 }, { "epoch": 1.98, "learning_rate": 1.5215085306610968e-05, "loss": 0.1097, "step": 85035 }, { "epoch": 1.98, "learning_rate": 1.5214205252164935e-05, "loss": 0.0511, "step": 85040 }, { "epoch": 1.98, "learning_rate": 1.52133251977189e-05, "loss": 0.0925, "step": 85045 }, { "epoch": 1.98, "learning_rate": 1.5212445143272863e-05, "loss": 0.2354, "step": 85050 }, { "epoch": 1.98, "learning_rate": 1.5211565088826828e-05, "loss": 0.1803, "step": 85055 }, { "epoch": 1.98, "learning_rate": 1.5210685034380795e-05, "loss": 0.1763, "step": 85060 }, { "epoch": 1.98, "learning_rate": 1.520980497993476e-05, "loss": 0.2124, "step": 85065 }, { "epoch": 1.98, "learning_rate": 1.5208924925488723e-05, "loss": 0.0308, "step": 85070 }, { "epoch": 1.98, "learning_rate": 1.5208044871042689e-05, "loss": 0.0354, "step": 85075 }, { "epoch": 1.99, "learning_rate": 1.5207164816596655e-05, "loss": 0.06, "step": 85080 }, { "epoch": 1.99, "learning_rate": 1.5206284762150617e-05, "loss": 0.0939, "step": 85085 }, { "epoch": 1.99, "learning_rate": 1.5205404707704584e-05, "loss": 0.0569, "step": 85090 }, { "epoch": 1.99, "learning_rate": 1.5204524653258549e-05, "loss": 0.0975, "step": 85095 }, { "epoch": 1.99, "learning_rate": 1.5203644598812515e-05, "loss": 0.1086, "step": 85100 }, { "epoch": 1.99, "learning_rate": 1.5202764544366477e-05, "loss": 0.2788, "step": 85105 }, { "epoch": 1.99, "learning_rate": 1.5201884489920444e-05, "loss": 0.3067, "step": 85110 }, { "epoch": 1.99, "learning_rate": 1.5201004435474409e-05, "loss": 0.2681, "step": 85115 }, { "epoch": 1.99, "learning_rate": 1.5200124381028372e-05, "loss": 0.0255, "step": 85120 }, { "epoch": 1.99, "learning_rate": 1.5199244326582337e-05, "loss": 0.0246, "step": 85125 }, { "epoch": 1.99, "learning_rate": 1.5198364272136304e-05, "loss": 0.0527, "step": 85130 }, { "epoch": 1.99, "learning_rate": 1.5197484217690269e-05, "loss": 0.0355, "step": 85135 }, { "epoch": 1.99, "learning_rate": 1.5196604163244232e-05, "loss": 0.0227, "step": 85140 }, { "epoch": 1.99, "learning_rate": 1.5195724108798199e-05, "loss": 0.2449, "step": 85145 }, { "epoch": 1.99, "learning_rate": 1.5194844054352164e-05, "loss": 0.1381, "step": 85150 }, { "epoch": 1.99, "learning_rate": 1.5193963999906127e-05, "loss": 0.0516, "step": 85155 }, { "epoch": 1.99, "learning_rate": 1.5193083945460092e-05, "loss": 0.3156, "step": 85160 }, { "epoch": 1.99, "learning_rate": 1.5192203891014059e-05, "loss": 0.3532, "step": 85165 }, { "epoch": 1.99, "learning_rate": 1.5191323836568024e-05, "loss": 0.0065, "step": 85170 }, { "epoch": 1.99, "learning_rate": 1.5190443782121987e-05, "loss": 0.0153, "step": 85175 }, { "epoch": 1.99, "learning_rate": 1.5189563727675952e-05, "loss": 0.031, "step": 85180 }, { "epoch": 1.99, "learning_rate": 1.5188683673229919e-05, "loss": 0.0344, "step": 85185 }, { "epoch": 1.99, "learning_rate": 1.518780361878388e-05, "loss": 0.0612, "step": 85190 }, { "epoch": 1.99, "learning_rate": 1.5186923564337847e-05, "loss": 0.0994, "step": 85195 }, { "epoch": 1.99, "learning_rate": 1.5186043509891812e-05, "loss": 0.1545, "step": 85200 }, { "epoch": 1.99, "learning_rate": 1.5185163455445779e-05, "loss": 0.1158, "step": 85205 }, { "epoch": 1.99, "learning_rate": 1.518428340099974e-05, "loss": 0.2251, "step": 85210 }, { "epoch": 1.99, "learning_rate": 1.5183403346553707e-05, "loss": 0.2459, "step": 85215 }, { "epoch": 1.99, "learning_rate": 1.5182523292107672e-05, "loss": 0.0145, "step": 85220 }, { "epoch": 1.99, "learning_rate": 1.5181643237661636e-05, "loss": 0.0131, "step": 85225 }, { "epoch": 1.99, "learning_rate": 1.51807631832156e-05, "loss": 0.0468, "step": 85230 }, { "epoch": 1.99, "learning_rate": 1.5179883128769567e-05, "loss": 0.1764, "step": 85235 }, { "epoch": 1.99, "learning_rate": 1.5179003074323532e-05, "loss": 0.0589, "step": 85240 }, { "epoch": 1.99, "learning_rate": 1.5178123019877496e-05, "loss": 0.1659, "step": 85245 }, { "epoch": 1.99, "learning_rate": 1.517724296543146e-05, "loss": 0.2706, "step": 85250 }, { "epoch": 1.99, "learning_rate": 1.5176362910985428e-05, "loss": 0.2591, "step": 85255 }, { "epoch": 1.99, "learning_rate": 1.5175482856539393e-05, "loss": 0.3099, "step": 85260 }, { "epoch": 1.99, "learning_rate": 1.5174602802093356e-05, "loss": 0.276, "step": 85265 }, { "epoch": 1.99, "learning_rate": 1.5173722747647321e-05, "loss": 0.0004, "step": 85270 }, { "epoch": 1.99, "learning_rate": 1.5172842693201288e-05, "loss": 0.0673, "step": 85275 }, { "epoch": 1.99, "learning_rate": 1.5171962638755251e-05, "loss": 0.0018, "step": 85280 }, { "epoch": 1.99, "learning_rate": 1.5171082584309216e-05, "loss": 0.105, "step": 85285 }, { "epoch": 1.99, "learning_rate": 1.5170202529863183e-05, "loss": 0.0855, "step": 85290 }, { "epoch": 1.99, "learning_rate": 1.5169322475417148e-05, "loss": 0.1426, "step": 85295 }, { "epoch": 1.99, "learning_rate": 1.5168442420971111e-05, "loss": 0.1256, "step": 85300 }, { "epoch": 1.99, "learning_rate": 1.5167562366525076e-05, "loss": 0.2461, "step": 85305 }, { "epoch": 1.99, "learning_rate": 1.5166682312079043e-05, "loss": 0.1093, "step": 85310 }, { "epoch": 1.99, "learning_rate": 1.5165802257633004e-05, "loss": 0.1795, "step": 85315 }, { "epoch": 1.99, "learning_rate": 1.5164922203186971e-05, "loss": 0.0442, "step": 85320 }, { "epoch": 1.99, "learning_rate": 1.5164042148740936e-05, "loss": 0.0018, "step": 85325 }, { "epoch": 1.99, "learning_rate": 1.5163162094294903e-05, "loss": 0.0681, "step": 85330 }, { "epoch": 1.99, "learning_rate": 1.5162282039848864e-05, "loss": 0.0067, "step": 85335 }, { "epoch": 1.99, "learning_rate": 1.5161401985402831e-05, "loss": 0.1219, "step": 85340 }, { "epoch": 1.99, "learning_rate": 1.5160521930956796e-05, "loss": 0.1273, "step": 85345 }, { "epoch": 1.99, "learning_rate": 1.515964187651076e-05, "loss": 0.1573, "step": 85350 }, { "epoch": 1.99, "learning_rate": 1.5158761822064725e-05, "loss": 0.2245, "step": 85355 }, { "epoch": 1.99, "learning_rate": 1.5157881767618691e-05, "loss": 0.2223, "step": 85360 }, { "epoch": 1.99, "learning_rate": 1.5157001713172656e-05, "loss": 0.1896, "step": 85365 }, { "epoch": 1.99, "learning_rate": 1.515612165872662e-05, "loss": 0.0343, "step": 85370 }, { "epoch": 1.99, "learning_rate": 1.5155241604280585e-05, "loss": 0.0586, "step": 85375 }, { "epoch": 1.99, "learning_rate": 1.5154361549834551e-05, "loss": 0.0133, "step": 85380 }, { "epoch": 1.99, "learning_rate": 1.5153481495388513e-05, "loss": 0.0321, "step": 85385 }, { "epoch": 1.99, "learning_rate": 1.515260144094248e-05, "loss": 0.1243, "step": 85390 }, { "epoch": 1.99, "learning_rate": 1.5151721386496445e-05, "loss": 0.1391, "step": 85395 }, { "epoch": 1.99, "learning_rate": 1.5150841332050411e-05, "loss": 0.1445, "step": 85400 }, { "epoch": 1.99, "learning_rate": 1.5149961277604373e-05, "loss": 0.1315, "step": 85405 }, { "epoch": 1.99, "learning_rate": 1.514908122315834e-05, "loss": 0.2999, "step": 85410 }, { "epoch": 1.99, "learning_rate": 1.5148201168712305e-05, "loss": 0.3249, "step": 85415 }, { "epoch": 1.99, "learning_rate": 1.5147321114266268e-05, "loss": 0.0185, "step": 85420 }, { "epoch": 1.99, "learning_rate": 1.5146441059820235e-05, "loss": 0.0754, "step": 85425 }, { "epoch": 1.99, "learning_rate": 1.51455610053742e-05, "loss": 0.0188, "step": 85430 }, { "epoch": 1.99, "learning_rate": 1.5144680950928167e-05, "loss": 0.0195, "step": 85435 }, { "epoch": 1.99, "learning_rate": 1.5143800896482128e-05, "loss": 0.1629, "step": 85440 }, { "epoch": 1.99, "learning_rate": 1.5142920842036095e-05, "loss": 0.1109, "step": 85445 }, { "epoch": 1.99, "learning_rate": 1.514204078759006e-05, "loss": 0.0431, "step": 85450 }, { "epoch": 1.99, "learning_rate": 1.5141160733144023e-05, "loss": 0.2354, "step": 85455 }, { "epoch": 1.99, "learning_rate": 1.5140280678697988e-05, "loss": 0.2372, "step": 85460 }, { "epoch": 1.99, "learning_rate": 1.5139400624251955e-05, "loss": 0.2203, "step": 85465 }, { "epoch": 1.99, "learning_rate": 1.513852056980592e-05, "loss": 0.0118, "step": 85470 }, { "epoch": 1.99, "learning_rate": 1.5137640515359883e-05, "loss": 0.0379, "step": 85475 }, { "epoch": 1.99, "learning_rate": 1.5136760460913848e-05, "loss": 0.0791, "step": 85480 }, { "epoch": 1.99, "learning_rate": 1.5135880406467815e-05, "loss": 0.0931, "step": 85485 }, { "epoch": 1.99, "learning_rate": 1.5135000352021777e-05, "loss": 0.0589, "step": 85490 }, { "epoch": 1.99, "learning_rate": 1.5134120297575743e-05, "loss": 0.0535, "step": 85495 }, { "epoch": 1.99, "learning_rate": 1.5133240243129708e-05, "loss": 0.1993, "step": 85500 }, { "epoch": 1.99, "learning_rate": 1.5132360188683675e-05, "loss": 0.1959, "step": 85505 }, { "epoch": 2.0, "learning_rate": 1.5131480134237637e-05, "loss": 0.201, "step": 85510 }, { "epoch": 2.0, "learning_rate": 1.5130600079791604e-05, "loss": 0.1237, "step": 85515 }, { "epoch": 2.0, "learning_rate": 1.5129720025345569e-05, "loss": 0.03, "step": 85520 }, { "epoch": 2.0, "learning_rate": 1.5128839970899532e-05, "loss": 0.0607, "step": 85525 }, { "epoch": 2.0, "learning_rate": 1.5127959916453497e-05, "loss": 0.0277, "step": 85530 }, { "epoch": 2.0, "learning_rate": 1.5127079862007464e-05, "loss": 0.0418, "step": 85535 }, { "epoch": 2.0, "learning_rate": 1.5126199807561429e-05, "loss": 0.0439, "step": 85540 }, { "epoch": 2.0, "learning_rate": 1.5125319753115392e-05, "loss": 0.1398, "step": 85545 }, { "epoch": 2.0, "learning_rate": 1.5124439698669359e-05, "loss": 0.1389, "step": 85550 }, { "epoch": 2.0, "learning_rate": 1.5123559644223324e-05, "loss": 0.1641, "step": 85555 }, { "epoch": 2.0, "learning_rate": 1.512267958977729e-05, "loss": 0.3692, "step": 85560 }, { "epoch": 2.0, "learning_rate": 1.5121799535331252e-05, "loss": 0.2628, "step": 85565 }, { "epoch": 2.0, "learning_rate": 1.5120919480885219e-05, "loss": 0.0459, "step": 85570 }, { "epoch": 2.0, "learning_rate": 1.5120039426439184e-05, "loss": 0.0469, "step": 85575 }, { "epoch": 2.0, "learning_rate": 1.5119159371993147e-05, "loss": 0.0223, "step": 85580 }, { "epoch": 2.0, "learning_rate": 1.5118279317547112e-05, "loss": 0.0326, "step": 85585 }, { "epoch": 2.0, "learning_rate": 1.5117399263101079e-05, "loss": 0.0393, "step": 85590 }, { "epoch": 2.0, "learning_rate": 1.5116519208655044e-05, "loss": 0.022, "step": 85595 }, { "epoch": 2.0, "learning_rate": 1.5115639154209007e-05, "loss": 0.0415, "step": 85600 }, { "epoch": 2.0, "learning_rate": 1.5114759099762972e-05, "loss": 0.1661, "step": 85605 }, { "epoch": 2.0, "learning_rate": 1.5113879045316939e-05, "loss": 0.3693, "step": 85610 }, { "epoch": 2.0, "learning_rate": 1.51129989908709e-05, "loss": 0.2407, "step": 85615 }, { "epoch": 2.0, "learning_rate": 1.5112118936424867e-05, "loss": 0.0494, "step": 85620 }, { "epoch": 2.0, "learning_rate": 1.5111238881978832e-05, "loss": 0.0442, "step": 85625 }, { "epoch": 2.0, "learning_rate": 1.5110358827532799e-05, "loss": 0.014, "step": 85630 }, { "epoch": 2.0, "learning_rate": 1.510947877308676e-05, "loss": 0.1126, "step": 85635 }, { "epoch": 2.0, "learning_rate": 1.5108598718640727e-05, "loss": 0.2787, "step": 85640 }, { "epoch": 2.0, "learning_rate": 1.5107718664194692e-05, "loss": 0.2342, "step": 85645 }, { "epoch": 2.0, "learning_rate": 1.5106838609748656e-05, "loss": 0.3036, "step": 85650 }, { "epoch": 2.0, "learning_rate": 1.510595855530262e-05, "loss": 0.273, "step": 85655 }, { "epoch": 2.0, "learning_rate": 1.5105078500856587e-05, "loss": 0.2986, "step": 85660 }, { "epoch": 2.0, "learning_rate": 1.5104198446410552e-05, "loss": 0.3031, "step": 85665 }, { "epoch": 2.0, "learning_rate": 1.5103318391964516e-05, "loss": 0.0183, "step": 85670 }, { "epoch": 2.0, "learning_rate": 1.510243833751848e-05, "loss": 0.033, "step": 85675 }, { "epoch": 2.0, "learning_rate": 1.5101558283072448e-05, "loss": 0.0447, "step": 85680 }, { "epoch": 2.0, "learning_rate": 1.510067822862641e-05, "loss": 0.1199, "step": 85685 }, { "epoch": 2.0, "learning_rate": 1.5099798174180376e-05, "loss": 0.0691, "step": 85690 }, { "epoch": 2.0, "learning_rate": 1.5098918119734343e-05, "loss": 0.0686, "step": 85695 }, { "epoch": 2.0, "learning_rate": 1.5098038065288308e-05, "loss": 0.0989, "step": 85700 }, { "epoch": 2.0, "learning_rate": 1.5097158010842271e-05, "loss": 0.1587, "step": 85705 }, { "epoch": 2.0, "learning_rate": 1.5096277956396236e-05, "loss": 0.3262, "step": 85710 }, { "epoch": 2.0, "learning_rate": 1.5095397901950203e-05, "loss": 0.1186, "step": 85715 }, { "epoch": 2.0, "learning_rate": 1.5094517847504164e-05, "loss": 0.0269, "step": 85720 }, { "epoch": 2.0, "eval_cer": 0.013293742058594479, "eval_loss": 0.03494327887892723, "eval_runtime": 703.1566, "eval_samples_per_second": 27.092, "eval_steps_per_second": 6.774, "eval_wer": 0.11067297581493166, "step": 85722 }, { "epoch": 2.0, "learning_rate": 1.5093637793058131e-05, "loss": 0.2609, "step": 85725 }, { "epoch": 2.0, "learning_rate": 1.5092757738612096e-05, "loss": 0.0618, "step": 85730 }, { "epoch": 2.0, "learning_rate": 1.5091877684166063e-05, "loss": 0.0225, "step": 85735 }, { "epoch": 2.0, "learning_rate": 1.5090997629720024e-05, "loss": 0.0414, "step": 85740 }, { "epoch": 2.0, "learning_rate": 1.5090117575273991e-05, "loss": 0.1621, "step": 85745 }, { "epoch": 2.0, "learning_rate": 1.5089237520827956e-05, "loss": 0.0841, "step": 85750 }, { "epoch": 2.0, "learning_rate": 1.508835746638192e-05, "loss": 0.1311, "step": 85755 }, { "epoch": 2.0, "learning_rate": 1.5087477411935884e-05, "loss": 0.1054, "step": 85760 }, { "epoch": 2.0, "learning_rate": 1.5086597357489851e-05, "loss": 0.2742, "step": 85765 }, { "epoch": 2.0, "learning_rate": 1.5085717303043816e-05, "loss": 0.2411, "step": 85770 }, { "epoch": 2.0, "learning_rate": 1.508483724859778e-05, "loss": 0.277, "step": 85775 }, { "epoch": 2.0, "learning_rate": 1.5083957194151745e-05, "loss": 0.0882, "step": 85780 }, { "epoch": 2.0, "learning_rate": 1.5083077139705711e-05, "loss": 0.0208, "step": 85785 }, { "epoch": 2.0, "learning_rate": 1.5082197085259673e-05, "loss": 0.0034, "step": 85790 }, { "epoch": 2.0, "learning_rate": 1.508131703081364e-05, "loss": 0.0229, "step": 85795 }, { "epoch": 2.0, "learning_rate": 1.5080436976367605e-05, "loss": 0.0988, "step": 85800 }, { "epoch": 2.0, "learning_rate": 1.5079556921921571e-05, "loss": 0.0268, "step": 85805 }, { "epoch": 2.0, "learning_rate": 1.5078676867475533e-05, "loss": 0.1123, "step": 85810 }, { "epoch": 2.0, "learning_rate": 1.50777968130295e-05, "loss": 0.144, "step": 85815 }, { "epoch": 2.0, "learning_rate": 1.5076916758583465e-05, "loss": 0.2621, "step": 85820 }, { "epoch": 2.0, "learning_rate": 1.5076036704137428e-05, "loss": 0.2609, "step": 85825 }, { "epoch": 2.0, "learning_rate": 1.5075156649691395e-05, "loss": 0.0871, "step": 85830 }, { "epoch": 2.0, "learning_rate": 1.507427659524536e-05, "loss": 0.007, "step": 85835 }, { "epoch": 2.0, "learning_rate": 1.5073396540799326e-05, "loss": 0.1239, "step": 85840 }, { "epoch": 2.0, "learning_rate": 1.5072516486353288e-05, "loss": 0.0421, "step": 85845 }, { "epoch": 2.0, "learning_rate": 1.5071636431907255e-05, "loss": 0.0776, "step": 85850 }, { "epoch": 2.0, "learning_rate": 1.507075637746122e-05, "loss": 0.1149, "step": 85855 }, { "epoch": 2.0, "learning_rate": 1.5069876323015187e-05, "loss": 0.0339, "step": 85860 }, { "epoch": 2.0, "learning_rate": 1.5068996268569148e-05, "loss": 0.1115, "step": 85865 }, { "epoch": 2.0, "learning_rate": 1.5068116214123115e-05, "loss": 0.1816, "step": 85870 }, { "epoch": 2.0, "learning_rate": 1.506723615967708e-05, "loss": 0.5314, "step": 85875 }, { "epoch": 2.0, "learning_rate": 1.5066356105231043e-05, "loss": 0.0148, "step": 85880 }, { "epoch": 2.0, "learning_rate": 1.5065476050785008e-05, "loss": 0.037, "step": 85885 }, { "epoch": 2.0, "learning_rate": 1.5064595996338975e-05, "loss": 0.0567, "step": 85890 }, { "epoch": 2.0, "learning_rate": 1.506371594189294e-05, "loss": 0.0492, "step": 85895 }, { "epoch": 2.0, "learning_rate": 1.5062835887446903e-05, "loss": 0.2458, "step": 85900 }, { "epoch": 2.0, "learning_rate": 1.5061955833000868e-05, "loss": 0.0775, "step": 85905 }, { "epoch": 2.0, "learning_rate": 1.5061075778554835e-05, "loss": 0.0344, "step": 85910 }, { "epoch": 2.0, "learning_rate": 1.5060195724108797e-05, "loss": 0.1243, "step": 85915 }, { "epoch": 2.0, "learning_rate": 1.5059315669662763e-05, "loss": 0.1605, "step": 85920 }, { "epoch": 2.0, "learning_rate": 1.5058435615216728e-05, "loss": 0.3229, "step": 85925 }, { "epoch": 2.0, "learning_rate": 1.5057555560770695e-05, "loss": 0.0173, "step": 85930 }, { "epoch": 2.0, "learning_rate": 1.5056675506324657e-05, "loss": 0.0854, "step": 85935 }, { "epoch": 2.01, "learning_rate": 1.5055795451878623e-05, "loss": 0.0249, "step": 85940 }, { "epoch": 2.01, "learning_rate": 1.5054915397432589e-05, "loss": 0.0536, "step": 85945 }, { "epoch": 2.01, "learning_rate": 1.5054035342986552e-05, "loss": 0.0168, "step": 85950 }, { "epoch": 2.01, "learning_rate": 1.5053155288540517e-05, "loss": 0.0249, "step": 85955 }, { "epoch": 2.01, "learning_rate": 1.5052275234094484e-05, "loss": 0.1403, "step": 85960 }, { "epoch": 2.01, "learning_rate": 1.505139517964845e-05, "loss": 0.0944, "step": 85965 }, { "epoch": 2.01, "learning_rate": 1.5050515125202412e-05, "loss": 0.1732, "step": 85970 }, { "epoch": 2.01, "learning_rate": 1.5049635070756379e-05, "loss": 0.1499, "step": 85975 }, { "epoch": 2.01, "learning_rate": 1.5048755016310344e-05, "loss": 0.0495, "step": 85980 }, { "epoch": 2.01, "learning_rate": 1.5047874961864307e-05, "loss": 0.0003, "step": 85985 }, { "epoch": 2.01, "learning_rate": 1.5046994907418272e-05, "loss": 0.065, "step": 85990 }, { "epoch": 2.01, "learning_rate": 1.5046114852972239e-05, "loss": 0.1208, "step": 85995 }, { "epoch": 2.01, "learning_rate": 1.5045234798526204e-05, "loss": 0.1199, "step": 86000 }, { "epoch": 2.01, "learning_rate": 1.5044354744080167e-05, "loss": 0.1099, "step": 86005 }, { "epoch": 2.01, "learning_rate": 1.5043474689634132e-05, "loss": 0.2233, "step": 86010 }, { "epoch": 2.01, "learning_rate": 1.5042594635188099e-05, "loss": 0.1671, "step": 86015 }, { "epoch": 2.01, "learning_rate": 1.504171458074206e-05, "loss": 0.3953, "step": 86020 }, { "epoch": 2.01, "learning_rate": 1.5040834526296027e-05, "loss": 0.4684, "step": 86025 }, { "epoch": 2.01, "learning_rate": 1.5039954471849992e-05, "loss": 0.0128, "step": 86030 }, { "epoch": 2.01, "learning_rate": 1.5039074417403959e-05, "loss": 0.0226, "step": 86035 }, { "epoch": 2.01, "learning_rate": 1.503819436295792e-05, "loss": 0.0234, "step": 86040 }, { "epoch": 2.01, "learning_rate": 1.5037314308511887e-05, "loss": 0.0654, "step": 86045 }, { "epoch": 2.01, "learning_rate": 1.5036434254065852e-05, "loss": 0.1265, "step": 86050 }, { "epoch": 2.01, "learning_rate": 1.5035554199619816e-05, "loss": 0.0862, "step": 86055 }, { "epoch": 2.01, "learning_rate": 1.503467414517378e-05, "loss": 0.1047, "step": 86060 }, { "epoch": 2.01, "learning_rate": 1.5033794090727747e-05, "loss": 0.1772, "step": 86065 }, { "epoch": 2.01, "learning_rate": 1.5032914036281712e-05, "loss": 0.1897, "step": 86070 }, { "epoch": 2.01, "learning_rate": 1.5032033981835676e-05, "loss": 0.4734, "step": 86075 }, { "epoch": 2.01, "learning_rate": 1.503115392738964e-05, "loss": 0.1263, "step": 86080 }, { "epoch": 2.01, "learning_rate": 1.5030273872943607e-05, "loss": 0.0271, "step": 86085 }, { "epoch": 2.01, "learning_rate": 1.502939381849757e-05, "loss": 0.1919, "step": 86090 }, { "epoch": 2.01, "learning_rate": 1.5028513764051536e-05, "loss": 0.0203, "step": 86095 }, { "epoch": 2.01, "learning_rate": 1.5027633709605502e-05, "loss": 0.1254, "step": 86100 }, { "epoch": 2.01, "learning_rate": 1.5026753655159467e-05, "loss": 0.0657, "step": 86105 }, { "epoch": 2.01, "learning_rate": 1.502587360071343e-05, "loss": 0.0754, "step": 86110 }, { "epoch": 2.01, "learning_rate": 1.5024993546267396e-05, "loss": 0.3057, "step": 86115 }, { "epoch": 2.01, "learning_rate": 1.5024113491821363e-05, "loss": 0.1988, "step": 86120 }, { "epoch": 2.01, "learning_rate": 1.5023233437375324e-05, "loss": 0.4691, "step": 86125 }, { "epoch": 2.01, "learning_rate": 1.5022353382929291e-05, "loss": 0.0374, "step": 86130 }, { "epoch": 2.01, "learning_rate": 1.5021473328483256e-05, "loss": 0.0783, "step": 86135 }, { "epoch": 2.01, "learning_rate": 1.5020593274037223e-05, "loss": 0.0226, "step": 86140 }, { "epoch": 2.01, "learning_rate": 1.5019713219591184e-05, "loss": 0.1966, "step": 86145 }, { "epoch": 2.01, "learning_rate": 1.5018833165145151e-05, "loss": 0.0291, "step": 86150 }, { "epoch": 2.01, "learning_rate": 1.5017953110699116e-05, "loss": 0.089, "step": 86155 }, { "epoch": 2.01, "learning_rate": 1.501707305625308e-05, "loss": 0.05, "step": 86160 }, { "epoch": 2.01, "learning_rate": 1.5016193001807044e-05, "loss": 0.1232, "step": 86165 }, { "epoch": 2.01, "learning_rate": 1.5015312947361011e-05, "loss": 0.2625, "step": 86170 }, { "epoch": 2.01, "learning_rate": 1.5014432892914976e-05, "loss": 0.2104, "step": 86175 }, { "epoch": 2.01, "learning_rate": 1.501355283846894e-05, "loss": 0.056, "step": 86180 }, { "epoch": 2.01, "learning_rate": 1.5012672784022904e-05, "loss": 0.1399, "step": 86185 }, { "epoch": 2.01, "learning_rate": 1.5011792729576871e-05, "loss": 0.111, "step": 86190 }, { "epoch": 2.01, "learning_rate": 1.5010912675130836e-05, "loss": 0.0968, "step": 86195 }, { "epoch": 2.01, "learning_rate": 1.50100326206848e-05, "loss": 0.1505, "step": 86200 }, { "epoch": 2.01, "learning_rate": 1.5009152566238765e-05, "loss": 0.1728, "step": 86205 }, { "epoch": 2.01, "learning_rate": 1.5008272511792731e-05, "loss": 0.202, "step": 86210 }, { "epoch": 2.01, "learning_rate": 1.5007392457346693e-05, "loss": 0.0914, "step": 86215 }, { "epoch": 2.01, "learning_rate": 1.500651240290066e-05, "loss": 0.1004, "step": 86220 }, { "epoch": 2.01, "learning_rate": 1.5005632348454625e-05, "loss": 0.3078, "step": 86225 }, { "epoch": 2.01, "learning_rate": 1.5004752294008591e-05, "loss": 0.0871, "step": 86230 }, { "epoch": 2.01, "learning_rate": 1.5003872239562555e-05, "loss": 0.0477, "step": 86235 }, { "epoch": 2.01, "learning_rate": 1.500299218511652e-05, "loss": 0.072, "step": 86240 }, { "epoch": 2.01, "learning_rate": 1.5002112130670486e-05, "loss": 0.0042, "step": 86245 }, { "epoch": 2.01, "learning_rate": 1.5001232076224448e-05, "loss": 0.0249, "step": 86250 }, { "epoch": 2.01, "learning_rate": 1.5000352021778415e-05, "loss": 0.1723, "step": 86255 }, { "epoch": 2.01, "learning_rate": 1.499947196733238e-05, "loss": 0.1513, "step": 86260 }, { "epoch": 2.01, "learning_rate": 1.4998591912886345e-05, "loss": 0.1208, "step": 86265 }, { "epoch": 2.01, "learning_rate": 1.499771185844031e-05, "loss": 0.2603, "step": 86270 }, { "epoch": 2.01, "learning_rate": 1.4996831803994275e-05, "loss": 0.232, "step": 86275 }, { "epoch": 2.01, "learning_rate": 1.4995951749548238e-05, "loss": 0.054, "step": 86280 }, { "epoch": 2.01, "learning_rate": 1.4995071695102205e-05, "loss": 0.0664, "step": 86285 }, { "epoch": 2.01, "learning_rate": 1.4994191640656168e-05, "loss": 0.0378, "step": 86290 }, { "epoch": 2.01, "learning_rate": 1.4993311586210135e-05, "loss": 0.0654, "step": 86295 }, { "epoch": 2.01, "learning_rate": 1.4992431531764098e-05, "loss": 0.0766, "step": 86300 }, { "epoch": 2.01, "learning_rate": 1.4991551477318065e-05, "loss": 0.086, "step": 86305 }, { "epoch": 2.01, "learning_rate": 1.4990671422872028e-05, "loss": 0.1824, "step": 86310 }, { "epoch": 2.01, "learning_rate": 1.4989791368425993e-05, "loss": 0.0713, "step": 86315 }, { "epoch": 2.01, "learning_rate": 1.4988911313979958e-05, "loss": 0.3435, "step": 86320 }, { "epoch": 2.01, "learning_rate": 1.4988031259533923e-05, "loss": 0.2613, "step": 86325 }, { "epoch": 2.01, "learning_rate": 1.4987151205087888e-05, "loss": 0.0176, "step": 86330 }, { "epoch": 2.01, "learning_rate": 1.4986271150641853e-05, "loss": 0.0034, "step": 86335 }, { "epoch": 2.01, "learning_rate": 1.4985391096195818e-05, "loss": 0.0967, "step": 86340 }, { "epoch": 2.01, "learning_rate": 1.4984511041749783e-05, "loss": 0.0971, "step": 86345 }, { "epoch": 2.01, "learning_rate": 1.4983630987303747e-05, "loss": 0.0108, "step": 86350 }, { "epoch": 2.01, "learning_rate": 1.4982750932857713e-05, "loss": 0.0826, "step": 86355 }, { "epoch": 2.01, "learning_rate": 1.4981870878411677e-05, "loss": 0.0801, "step": 86360 }, { "epoch": 2.02, "learning_rate": 1.4980990823965643e-05, "loss": 0.3535, "step": 86365 }, { "epoch": 2.02, "learning_rate": 1.4980110769519608e-05, "loss": 0.3692, "step": 86370 }, { "epoch": 2.02, "learning_rate": 1.4979230715073574e-05, "loss": 0.5901, "step": 86375 }, { "epoch": 2.02, "learning_rate": 1.4978350660627539e-05, "loss": 0.0184, "step": 86380 }, { "epoch": 2.02, "learning_rate": 1.4977470606181502e-05, "loss": 0.0647, "step": 86385 }, { "epoch": 2.02, "learning_rate": 1.4976590551735469e-05, "loss": 0.0225, "step": 86390 }, { "epoch": 2.02, "learning_rate": 1.4975710497289432e-05, "loss": 0.203, "step": 86395 }, { "epoch": 2.02, "learning_rate": 1.4974830442843399e-05, "loss": 0.0378, "step": 86400 }, { "epoch": 2.02, "learning_rate": 1.4973950388397362e-05, "loss": 0.0673, "step": 86405 }, { "epoch": 2.02, "learning_rate": 1.4973070333951329e-05, "loss": 0.0753, "step": 86410 }, { "epoch": 2.02, "learning_rate": 1.4972190279505292e-05, "loss": 0.2182, "step": 86415 }, { "epoch": 2.02, "learning_rate": 1.4971310225059259e-05, "loss": 0.399, "step": 86420 }, { "epoch": 2.02, "learning_rate": 1.4970430170613222e-05, "loss": 0.2298, "step": 86425 }, { "epoch": 2.02, "learning_rate": 1.4969550116167187e-05, "loss": 0.0272, "step": 86430 }, { "epoch": 2.02, "learning_rate": 1.4968670061721152e-05, "loss": 0.0779, "step": 86435 }, { "epoch": 2.02, "learning_rate": 1.4967790007275117e-05, "loss": 0.0193, "step": 86440 }, { "epoch": 2.02, "learning_rate": 1.4966909952829082e-05, "loss": 0.0233, "step": 86445 }, { "epoch": 2.02, "learning_rate": 1.4966029898383047e-05, "loss": 0.1518, "step": 86450 }, { "epoch": 2.02, "learning_rate": 1.4965149843937012e-05, "loss": 0.0255, "step": 86455 }, { "epoch": 2.02, "learning_rate": 1.4964269789490977e-05, "loss": 0.0947, "step": 86460 }, { "epoch": 2.02, "learning_rate": 1.496338973504494e-05, "loss": 0.0223, "step": 86465 }, { "epoch": 2.02, "learning_rate": 1.4962509680598907e-05, "loss": 0.1965, "step": 86470 }, { "epoch": 2.02, "learning_rate": 1.496162962615287e-05, "loss": 0.2652, "step": 86475 }, { "epoch": 2.02, "learning_rate": 1.4960749571706837e-05, "loss": 0.1121, "step": 86480 }, { "epoch": 2.02, "learning_rate": 1.49598695172608e-05, "loss": 0.0324, "step": 86485 }, { "epoch": 2.02, "learning_rate": 1.4958989462814767e-05, "loss": 0.0454, "step": 86490 }, { "epoch": 2.02, "learning_rate": 1.495810940836873e-05, "loss": 0.0945, "step": 86495 }, { "epoch": 2.02, "learning_rate": 1.4957229353922696e-05, "loss": 0.0542, "step": 86500 }, { "epoch": 2.02, "learning_rate": 1.4956349299476662e-05, "loss": 0.1227, "step": 86505 }, { "epoch": 2.02, "learning_rate": 1.4955469245030626e-05, "loss": 0.2207, "step": 86510 }, { "epoch": 2.02, "learning_rate": 1.4954589190584592e-05, "loss": 0.3301, "step": 86515 }, { "epoch": 2.02, "learning_rate": 1.4953709136138556e-05, "loss": 0.1677, "step": 86520 }, { "epoch": 2.02, "learning_rate": 1.4952829081692522e-05, "loss": 0.2315, "step": 86525 }, { "epoch": 2.02, "learning_rate": 1.4951949027246486e-05, "loss": 0.0781, "step": 86530 }, { "epoch": 2.02, "learning_rate": 1.495106897280045e-05, "loss": 0.0269, "step": 86535 }, { "epoch": 2.02, "learning_rate": 1.4950188918354416e-05, "loss": 0.0562, "step": 86540 }, { "epoch": 2.02, "learning_rate": 1.494930886390838e-05, "loss": 0.0919, "step": 86545 }, { "epoch": 2.02, "learning_rate": 1.4948428809462346e-05, "loss": 0.0333, "step": 86550 }, { "epoch": 2.02, "learning_rate": 1.4947548755016311e-05, "loss": 0.1148, "step": 86555 }, { "epoch": 2.02, "learning_rate": 1.4946668700570276e-05, "loss": 0.1704, "step": 86560 }, { "epoch": 2.02, "learning_rate": 1.4945788646124241e-05, "loss": 0.0887, "step": 86565 }, { "epoch": 2.02, "learning_rate": 1.4944908591678206e-05, "loss": 0.2093, "step": 86570 }, { "epoch": 2.02, "learning_rate": 1.4944028537232171e-05, "loss": 0.2959, "step": 86575 }, { "epoch": 2.02, "learning_rate": 1.4943148482786134e-05, "loss": 0.0436, "step": 86580 }, { "epoch": 2.02, "learning_rate": 1.4942268428340101e-05, "loss": 0.0073, "step": 86585 }, { "epoch": 2.02, "learning_rate": 1.4941388373894064e-05, "loss": 0.1329, "step": 86590 }, { "epoch": 2.02, "learning_rate": 1.4940508319448031e-05, "loss": 0.1183, "step": 86595 }, { "epoch": 2.02, "learning_rate": 1.4939628265001994e-05, "loss": 0.0789, "step": 86600 }, { "epoch": 2.02, "learning_rate": 1.4938748210555961e-05, "loss": 0.0361, "step": 86605 }, { "epoch": 2.02, "learning_rate": 1.4937868156109924e-05, "loss": 0.3917, "step": 86610 }, { "epoch": 2.02, "learning_rate": 1.493698810166389e-05, "loss": 0.2049, "step": 86615 }, { "epoch": 2.02, "learning_rate": 1.4936108047217854e-05, "loss": 0.0959, "step": 86620 }, { "epoch": 2.02, "learning_rate": 1.493522799277182e-05, "loss": 0.2059, "step": 86625 }, { "epoch": 2.02, "learning_rate": 1.4934347938325784e-05, "loss": 0.0264, "step": 86630 }, { "epoch": 2.02, "learning_rate": 1.493346788387975e-05, "loss": 0.0072, "step": 86635 }, { "epoch": 2.02, "learning_rate": 1.4932587829433716e-05, "loss": 0.034, "step": 86640 }, { "epoch": 2.02, "learning_rate": 1.493170777498768e-05, "loss": 0.09, "step": 86645 }, { "epoch": 2.02, "learning_rate": 1.4930827720541645e-05, "loss": 0.0968, "step": 86650 }, { "epoch": 2.02, "learning_rate": 1.4930123676984816e-05, "loss": 0.0956, "step": 86655 }, { "epoch": 2.02, "learning_rate": 1.4929243622538781e-05, "loss": 0.0907, "step": 86660 }, { "epoch": 2.02, "learning_rate": 1.4928363568092746e-05, "loss": 0.2147, "step": 86665 }, { "epoch": 2.02, "learning_rate": 1.4927483513646711e-05, "loss": 0.3102, "step": 86670 }, { "epoch": 2.02, "learning_rate": 1.4926603459200676e-05, "loss": 0.2485, "step": 86675 }, { "epoch": 2.02, "learning_rate": 1.4925723404754641e-05, "loss": 0.0733, "step": 86680 }, { "epoch": 2.02, "learning_rate": 1.4924843350308606e-05, "loss": 0.0886, "step": 86685 }, { "epoch": 2.02, "learning_rate": 1.4923963295862571e-05, "loss": 0.0235, "step": 86690 }, { "epoch": 2.02, "learning_rate": 1.4923083241416536e-05, "loss": 0.2041, "step": 86695 }, { "epoch": 2.02, "learning_rate": 1.4922203186970501e-05, "loss": 0.058, "step": 86700 }, { "epoch": 2.02, "learning_rate": 1.4921323132524466e-05, "loss": 0.1166, "step": 86705 }, { "epoch": 2.02, "learning_rate": 1.4920443078078431e-05, "loss": 0.0931, "step": 86710 }, { "epoch": 2.02, "learning_rate": 1.4919563023632396e-05, "loss": 0.1571, "step": 86715 }, { "epoch": 2.02, "learning_rate": 1.4918682969186361e-05, "loss": 0.2798, "step": 86720 }, { "epoch": 2.02, "learning_rate": 1.4917802914740326e-05, "loss": 0.236, "step": 86725 }, { "epoch": 2.02, "learning_rate": 1.491692286029429e-05, "loss": 0.025, "step": 86730 }, { "epoch": 2.02, "learning_rate": 1.4916042805848256e-05, "loss": 0.0092, "step": 86735 }, { "epoch": 2.02, "learning_rate": 1.491516275140222e-05, "loss": 0.0289, "step": 86740 }, { "epoch": 2.02, "learning_rate": 1.4914282696956186e-05, "loss": 0.0797, "step": 86745 }, { "epoch": 2.02, "learning_rate": 1.491340264251015e-05, "loss": 0.0424, "step": 86750 }, { "epoch": 2.02, "learning_rate": 1.4912522588064116e-05, "loss": 0.0696, "step": 86755 }, { "epoch": 2.02, "learning_rate": 1.491164253361808e-05, "loss": 0.0495, "step": 86760 }, { "epoch": 2.02, "learning_rate": 1.4910762479172046e-05, "loss": 0.0889, "step": 86765 }, { "epoch": 2.02, "learning_rate": 1.490988242472601e-05, "loss": 0.139, "step": 86770 }, { "epoch": 2.02, "learning_rate": 1.4909002370279975e-05, "loss": 0.2006, "step": 86775 }, { "epoch": 2.02, "learning_rate": 1.490812231583394e-05, "loss": 0.0413, "step": 86780 }, { "epoch": 2.02, "learning_rate": 1.4907242261387905e-05, "loss": 0.0344, "step": 86785 }, { "epoch": 2.02, "learning_rate": 1.490636220694187e-05, "loss": 0.068, "step": 86790 }, { "epoch": 2.03, "learning_rate": 1.4905482152495835e-05, "loss": 0.0849, "step": 86795 }, { "epoch": 2.03, "learning_rate": 1.49046020980498e-05, "loss": 0.0805, "step": 86800 }, { "epoch": 2.03, "learning_rate": 1.4903722043603765e-05, "loss": 0.0865, "step": 86805 }, { "epoch": 2.03, "learning_rate": 1.4902841989157728e-05, "loss": 0.0418, "step": 86810 }, { "epoch": 2.03, "learning_rate": 1.4901961934711695e-05, "loss": 0.1308, "step": 86815 }, { "epoch": 2.03, "learning_rate": 1.4901081880265658e-05, "loss": 0.3256, "step": 86820 }, { "epoch": 2.03, "learning_rate": 1.4900201825819625e-05, "loss": 0.4039, "step": 86825 }, { "epoch": 2.03, "learning_rate": 1.489932177137359e-05, "loss": 0.0409, "step": 86830 }, { "epoch": 2.03, "learning_rate": 1.4898441716927555e-05, "loss": 0.1476, "step": 86835 }, { "epoch": 2.03, "learning_rate": 1.489756166248152e-05, "loss": 0.0262, "step": 86840 }, { "epoch": 2.03, "learning_rate": 1.4896681608035483e-05, "loss": 0.0137, "step": 86845 }, { "epoch": 2.03, "learning_rate": 1.489580155358945e-05, "loss": 0.0913, "step": 86850 }, { "epoch": 2.03, "learning_rate": 1.4894921499143413e-05, "loss": 0.09, "step": 86855 }, { "epoch": 2.03, "learning_rate": 1.489404144469738e-05, "loss": 0.3434, "step": 86860 }, { "epoch": 2.03, "learning_rate": 1.4893161390251343e-05, "loss": 0.0862, "step": 86865 }, { "epoch": 2.03, "learning_rate": 1.489228133580531e-05, "loss": 0.4435, "step": 86870 }, { "epoch": 2.03, "learning_rate": 1.4891401281359273e-05, "loss": 0.1614, "step": 86875 }, { "epoch": 2.03, "learning_rate": 1.4890521226913238e-05, "loss": 0.0229, "step": 86880 }, { "epoch": 2.03, "learning_rate": 1.4889641172467203e-05, "loss": 0.0626, "step": 86885 }, { "epoch": 2.03, "learning_rate": 1.4888761118021168e-05, "loss": 0.0491, "step": 86890 }, { "epoch": 2.03, "learning_rate": 1.4887881063575133e-05, "loss": 0.0255, "step": 86895 }, { "epoch": 2.03, "learning_rate": 1.4887001009129099e-05, "loss": 0.1125, "step": 86900 }, { "epoch": 2.03, "learning_rate": 1.4886120954683064e-05, "loss": 0.1242, "step": 86905 }, { "epoch": 2.03, "learning_rate": 1.4885240900237029e-05, "loss": 0.0827, "step": 86910 }, { "epoch": 2.03, "learning_rate": 1.4884360845790994e-05, "loss": 0.0714, "step": 86915 }, { "epoch": 2.03, "learning_rate": 1.4883480791344959e-05, "loss": 0.3017, "step": 86920 }, { "epoch": 2.03, "learning_rate": 1.4882600736898922e-05, "loss": 0.3286, "step": 86925 }, { "epoch": 2.03, "learning_rate": 1.4881720682452889e-05, "loss": 0.1154, "step": 86930 }, { "epoch": 2.03, "learning_rate": 1.4880840628006852e-05, "loss": 0.0339, "step": 86935 }, { "epoch": 2.03, "learning_rate": 1.4879960573560819e-05, "loss": 0.0418, "step": 86940 }, { "epoch": 2.03, "learning_rate": 1.4879080519114782e-05, "loss": 0.1151, "step": 86945 }, { "epoch": 2.03, "learning_rate": 1.4878200464668749e-05, "loss": 0.1009, "step": 86950 }, { "epoch": 2.03, "learning_rate": 1.4877320410222712e-05, "loss": 0.1149, "step": 86955 }, { "epoch": 2.03, "learning_rate": 1.4876440355776677e-05, "loss": 0.2499, "step": 86960 }, { "epoch": 2.03, "learning_rate": 1.4875560301330642e-05, "loss": 0.1372, "step": 86965 }, { "epoch": 2.03, "learning_rate": 1.4874680246884607e-05, "loss": 0.2307, "step": 86970 }, { "epoch": 2.03, "learning_rate": 1.4873800192438574e-05, "loss": 0.2929, "step": 86975 }, { "epoch": 2.03, "learning_rate": 1.4872920137992537e-05, "loss": 0.0491, "step": 86980 }, { "epoch": 2.03, "learning_rate": 1.4872040083546504e-05, "loss": 0.0389, "step": 86985 }, { "epoch": 2.03, "learning_rate": 1.4871160029100467e-05, "loss": 0.066, "step": 86990 }, { "epoch": 2.03, "learning_rate": 1.4870279974654432e-05, "loss": 0.2365, "step": 86995 }, { "epoch": 2.03, "learning_rate": 1.4869399920208397e-05, "loss": 0.0612, "step": 87000 }, { "epoch": 2.03, "learning_rate": 1.4868519865762362e-05, "loss": 0.0799, "step": 87005 }, { "epoch": 2.03, "learning_rate": 1.4867639811316327e-05, "loss": 0.0219, "step": 87010 }, { "epoch": 2.03, "learning_rate": 1.4866759756870292e-05, "loss": 0.1342, "step": 87015 }, { "epoch": 2.03, "learning_rate": 1.4865879702424257e-05, "loss": 0.1714, "step": 87020 }, { "epoch": 2.03, "learning_rate": 1.4864999647978222e-05, "loss": 0.2704, "step": 87025 }, { "epoch": 2.03, "learning_rate": 1.4864119593532186e-05, "loss": 0.0306, "step": 87030 }, { "epoch": 2.03, "learning_rate": 1.4863239539086152e-05, "loss": 0.0724, "step": 87035 }, { "epoch": 2.03, "learning_rate": 1.4862359484640116e-05, "loss": 0.0324, "step": 87040 }, { "epoch": 2.03, "learning_rate": 1.4861479430194082e-05, "loss": 0.1166, "step": 87045 }, { "epoch": 2.03, "learning_rate": 1.4860599375748046e-05, "loss": 0.0178, "step": 87050 }, { "epoch": 2.03, "learning_rate": 1.4859719321302012e-05, "loss": 0.0528, "step": 87055 }, { "epoch": 2.03, "learning_rate": 1.4858839266855976e-05, "loss": 0.1274, "step": 87060 }, { "epoch": 2.03, "learning_rate": 1.4857959212409942e-05, "loss": 0.1487, "step": 87065 }, { "epoch": 2.03, "learning_rate": 1.4857079157963906e-05, "loss": 0.303, "step": 87070 }, { "epoch": 2.03, "learning_rate": 1.485619910351787e-05, "loss": 0.1305, "step": 87075 }, { "epoch": 2.03, "learning_rate": 1.4855319049071836e-05, "loss": 0.0792, "step": 87080 }, { "epoch": 2.03, "learning_rate": 1.4854438994625801e-05, "loss": 0.0345, "step": 87085 }, { "epoch": 2.03, "learning_rate": 1.4853558940179766e-05, "loss": 0.0634, "step": 87090 }, { "epoch": 2.03, "learning_rate": 1.4852678885733731e-05, "loss": 0.1096, "step": 87095 }, { "epoch": 2.03, "learning_rate": 1.4851798831287696e-05, "loss": 0.2392, "step": 87100 }, { "epoch": 2.03, "learning_rate": 1.4850918776841661e-05, "loss": 0.1173, "step": 87105 }, { "epoch": 2.03, "learning_rate": 1.4850038722395626e-05, "loss": 0.0377, "step": 87110 }, { "epoch": 2.03, "learning_rate": 1.4849158667949591e-05, "loss": 0.0344, "step": 87115 }, { "epoch": 2.03, "learning_rate": 1.4848278613503556e-05, "loss": 0.1784, "step": 87120 }, { "epoch": 2.03, "learning_rate": 1.4847398559057521e-05, "loss": 0.2768, "step": 87125 }, { "epoch": 2.03, "learning_rate": 1.4846518504611486e-05, "loss": 0.0109, "step": 87130 }, { "epoch": 2.03, "learning_rate": 1.4845638450165451e-05, "loss": 0.0592, "step": 87135 }, { "epoch": 2.03, "learning_rate": 1.4844758395719416e-05, "loss": 0.0974, "step": 87140 }, { "epoch": 2.03, "learning_rate": 1.484387834127338e-05, "loss": 0.0377, "step": 87145 }, { "epoch": 2.03, "learning_rate": 1.4842998286827346e-05, "loss": 0.14, "step": 87150 }, { "epoch": 2.03, "learning_rate": 1.484211823238131e-05, "loss": 0.0501, "step": 87155 }, { "epoch": 2.03, "learning_rate": 1.4841238177935276e-05, "loss": 0.0767, "step": 87160 }, { "epoch": 2.03, "learning_rate": 1.484035812348924e-05, "loss": 0.0952, "step": 87165 }, { "epoch": 2.03, "learning_rate": 1.4839478069043206e-05, "loss": 0.1933, "step": 87170 }, { "epoch": 2.03, "learning_rate": 1.483859801459717e-05, "loss": 0.2693, "step": 87175 }, { "epoch": 2.03, "learning_rate": 1.4837717960151135e-05, "loss": 0.0103, "step": 87180 }, { "epoch": 2.03, "learning_rate": 1.48368379057051e-05, "loss": 0.1011, "step": 87185 }, { "epoch": 2.03, "learning_rate": 1.4835957851259065e-05, "loss": 0.1143, "step": 87190 }, { "epoch": 2.03, "learning_rate": 1.483507779681303e-05, "loss": 0.0836, "step": 87195 }, { "epoch": 2.03, "learning_rate": 1.4834197742366995e-05, "loss": 0.1034, "step": 87200 }, { "epoch": 2.03, "learning_rate": 1.483331768792096e-05, "loss": 0.0574, "step": 87205 }, { "epoch": 2.03, "learning_rate": 1.4832437633474925e-05, "loss": 0.1278, "step": 87210 }, { "epoch": 2.03, "learning_rate": 1.483155757902889e-05, "loss": 0.1552, "step": 87215 }, { "epoch": 2.03, "learning_rate": 1.4830677524582855e-05, "loss": 0.3295, "step": 87220 }, { "epoch": 2.04, "learning_rate": 1.4829797470136818e-05, "loss": 0.3557, "step": 87225 }, { "epoch": 2.04, "learning_rate": 1.4828917415690785e-05, "loss": 0.0066, "step": 87230 }, { "epoch": 2.04, "learning_rate": 1.4828037361244748e-05, "loss": 0.0976, "step": 87235 }, { "epoch": 2.04, "learning_rate": 1.4827157306798715e-05, "loss": 0.0053, "step": 87240 }, { "epoch": 2.04, "learning_rate": 1.482627725235268e-05, "loss": 0.1811, "step": 87245 }, { "epoch": 2.04, "learning_rate": 1.4825397197906645e-05, "loss": 0.045, "step": 87250 }, { "epoch": 2.04, "learning_rate": 1.482451714346061e-05, "loss": 0.1395, "step": 87255 }, { "epoch": 2.04, "learning_rate": 1.4823637089014573e-05, "loss": 0.1455, "step": 87260 }, { "epoch": 2.04, "learning_rate": 1.482275703456854e-05, "loss": 0.1087, "step": 87265 }, { "epoch": 2.04, "learning_rate": 1.4821876980122503e-05, "loss": 0.1063, "step": 87270 }, { "epoch": 2.04, "learning_rate": 1.482099692567647e-05, "loss": 0.121, "step": 87275 }, { "epoch": 2.04, "learning_rate": 1.4820116871230433e-05, "loss": 0.0804, "step": 87280 }, { "epoch": 2.04, "learning_rate": 1.48192368167844e-05, "loss": 0.0919, "step": 87285 }, { "epoch": 2.04, "learning_rate": 1.4818356762338363e-05, "loss": 0.1, "step": 87290 }, { "epoch": 2.04, "learning_rate": 1.4817476707892328e-05, "loss": 0.0618, "step": 87295 }, { "epoch": 2.04, "learning_rate": 1.4816596653446293e-05, "loss": 0.0974, "step": 87300 }, { "epoch": 2.04, "learning_rate": 1.4815716599000258e-05, "loss": 0.3235, "step": 87305 }, { "epoch": 2.04, "learning_rate": 1.4814836544554223e-05, "loss": 0.1801, "step": 87310 }, { "epoch": 2.04, "learning_rate": 1.4813956490108188e-05, "loss": 0.1722, "step": 87315 }, { "epoch": 2.04, "learning_rate": 1.4813076435662153e-05, "loss": 0.1723, "step": 87320 }, { "epoch": 2.04, "learning_rate": 1.4812196381216118e-05, "loss": 0.3629, "step": 87325 }, { "epoch": 2.04, "learning_rate": 1.4811316326770082e-05, "loss": 0.0211, "step": 87330 }, { "epoch": 2.04, "learning_rate": 1.4810436272324049e-05, "loss": 0.0034, "step": 87335 }, { "epoch": 2.04, "learning_rate": 1.4809556217878012e-05, "loss": 0.1093, "step": 87340 }, { "epoch": 2.04, "learning_rate": 1.4808676163431979e-05, "loss": 0.1423, "step": 87345 }, { "epoch": 2.04, "learning_rate": 1.4807796108985942e-05, "loss": 0.0255, "step": 87350 }, { "epoch": 2.04, "learning_rate": 1.4806916054539909e-05, "loss": 0.0085, "step": 87355 }, { "epoch": 2.04, "learning_rate": 1.4806036000093872e-05, "loss": 0.0675, "step": 87360 }, { "epoch": 2.04, "learning_rate": 1.4805155945647839e-05, "loss": 0.1006, "step": 87365 }, { "epoch": 2.04, "learning_rate": 1.4804275891201802e-05, "loss": 0.3022, "step": 87370 }, { "epoch": 2.04, "learning_rate": 1.4803395836755767e-05, "loss": 0.2625, "step": 87375 }, { "epoch": 2.04, "learning_rate": 1.4802515782309734e-05, "loss": 0.0302, "step": 87380 }, { "epoch": 2.04, "learning_rate": 1.4801635727863697e-05, "loss": 0.0707, "step": 87385 }, { "epoch": 2.04, "learning_rate": 1.4800755673417664e-05, "loss": 0.0305, "step": 87390 }, { "epoch": 2.04, "learning_rate": 1.4799875618971627e-05, "loss": 0.1112, "step": 87395 }, { "epoch": 2.04, "learning_rate": 1.4798995564525594e-05, "loss": 0.0268, "step": 87400 }, { "epoch": 2.04, "learning_rate": 1.4798115510079557e-05, "loss": 0.0682, "step": 87405 }, { "epoch": 2.04, "learning_rate": 1.4797235455633522e-05, "loss": 0.1229, "step": 87410 }, { "epoch": 2.04, "learning_rate": 1.4796355401187487e-05, "loss": 0.2432, "step": 87415 }, { "epoch": 2.04, "learning_rate": 1.4795475346741452e-05, "loss": 0.2087, "step": 87420 }, { "epoch": 2.04, "learning_rate": 1.4794595292295417e-05, "loss": 0.2759, "step": 87425 }, { "epoch": 2.04, "learning_rate": 1.4793715237849382e-05, "loss": 0.0054, "step": 87430 }, { "epoch": 2.04, "learning_rate": 1.4792835183403347e-05, "loss": 0.0312, "step": 87435 }, { "epoch": 2.04, "learning_rate": 1.4791955128957312e-05, "loss": 0.0173, "step": 87440 }, { "epoch": 2.04, "learning_rate": 1.4791075074511276e-05, "loss": 0.0074, "step": 87445 }, { "epoch": 2.04, "learning_rate": 1.4790195020065242e-05, "loss": 0.0918, "step": 87450 }, { "epoch": 2.04, "learning_rate": 1.4789314965619206e-05, "loss": 0.1291, "step": 87455 }, { "epoch": 2.04, "learning_rate": 1.4788434911173172e-05, "loss": 0.1873, "step": 87460 }, { "epoch": 2.04, "learning_rate": 1.4787554856727136e-05, "loss": 0.1179, "step": 87465 }, { "epoch": 2.04, "learning_rate": 1.4786674802281102e-05, "loss": 0.4887, "step": 87470 }, { "epoch": 2.04, "learning_rate": 1.4785794747835066e-05, "loss": 0.1915, "step": 87475 }, { "epoch": 2.04, "learning_rate": 1.478491469338903e-05, "loss": 0.0318, "step": 87480 }, { "epoch": 2.04, "learning_rate": 1.4784034638942996e-05, "loss": 0.0099, "step": 87485 }, { "epoch": 2.04, "learning_rate": 1.478315458449696e-05, "loss": 0.0831, "step": 87490 }, { "epoch": 2.04, "learning_rate": 1.4782274530050926e-05, "loss": 0.0881, "step": 87495 }, { "epoch": 2.04, "learning_rate": 1.478139447560489e-05, "loss": 0.0523, "step": 87500 }, { "epoch": 2.04, "learning_rate": 1.4780514421158856e-05, "loss": 0.1306, "step": 87505 }, { "epoch": 2.04, "learning_rate": 1.4779634366712821e-05, "loss": 0.2338, "step": 87510 }, { "epoch": 2.04, "learning_rate": 1.4778754312266788e-05, "loss": 0.1881, "step": 87515 }, { "epoch": 2.04, "learning_rate": 1.4777874257820751e-05, "loss": 0.2307, "step": 87520 }, { "epoch": 2.04, "learning_rate": 1.4776994203374716e-05, "loss": 0.2692, "step": 87525 }, { "epoch": 2.04, "learning_rate": 1.4776114148928681e-05, "loss": 0.0626, "step": 87530 }, { "epoch": 2.04, "learning_rate": 1.4775234094482646e-05, "loss": 0.0426, "step": 87535 }, { "epoch": 2.04, "learning_rate": 1.4774354040036611e-05, "loss": 0.0513, "step": 87540 }, { "epoch": 2.04, "learning_rate": 1.4773473985590576e-05, "loss": 0.0616, "step": 87545 }, { "epoch": 2.04, "learning_rate": 1.4772593931144541e-05, "loss": 0.029, "step": 87550 }, { "epoch": 2.04, "learning_rate": 1.4771713876698506e-05, "loss": 0.1571, "step": 87555 }, { "epoch": 2.04, "learning_rate": 1.477083382225247e-05, "loss": 0.1218, "step": 87560 }, { "epoch": 2.04, "learning_rate": 1.4769953767806436e-05, "loss": 0.2177, "step": 87565 }, { "epoch": 2.04, "learning_rate": 1.47690737133604e-05, "loss": 0.2672, "step": 87570 }, { "epoch": 2.04, "learning_rate": 1.4768193658914366e-05, "loss": 0.245, "step": 87575 }, { "epoch": 2.04, "learning_rate": 1.476731360446833e-05, "loss": 0.049, "step": 87580 }, { "epoch": 2.04, "learning_rate": 1.4766433550022296e-05, "loss": 0.1174, "step": 87585 }, { "epoch": 2.04, "learning_rate": 1.476555349557626e-05, "loss": 0.1089, "step": 87590 }, { "epoch": 2.04, "learning_rate": 1.4764673441130225e-05, "loss": 0.0229, "step": 87595 }, { "epoch": 2.04, "learning_rate": 1.476379338668419e-05, "loss": 0.1206, "step": 87600 }, { "epoch": 2.04, "learning_rate": 1.4762913332238155e-05, "loss": 0.1487, "step": 87605 }, { "epoch": 2.04, "learning_rate": 1.476203327779212e-05, "loss": 0.0805, "step": 87610 }, { "epoch": 2.04, "learning_rate": 1.4761153223346085e-05, "loss": 0.0589, "step": 87615 }, { "epoch": 2.04, "learning_rate": 1.476027316890005e-05, "loss": 0.3187, "step": 87620 }, { "epoch": 2.04, "learning_rate": 1.4759393114454015e-05, "loss": 0.3039, "step": 87625 }, { "epoch": 2.04, "learning_rate": 1.4758513060007978e-05, "loss": 0.0231, "step": 87630 }, { "epoch": 2.04, "learning_rate": 1.4757633005561945e-05, "loss": 0.0519, "step": 87635 }, { "epoch": 2.04, "learning_rate": 1.4756752951115908e-05, "loss": 0.083, "step": 87640 }, { "epoch": 2.04, "learning_rate": 1.4755872896669875e-05, "loss": 0.1532, "step": 87645 }, { "epoch": 2.04, "learning_rate": 1.475499284222384e-05, "loss": 0.0792, "step": 87650 }, { "epoch": 2.05, "learning_rate": 1.4754112787777805e-05, "loss": 0.1938, "step": 87655 }, { "epoch": 2.05, "learning_rate": 1.475323273333177e-05, "loss": 0.1437, "step": 87660 }, { "epoch": 2.05, "learning_rate": 1.4752352678885735e-05, "loss": 0.1895, "step": 87665 }, { "epoch": 2.05, "learning_rate": 1.47514726244397e-05, "loss": 0.3918, "step": 87670 }, { "epoch": 2.05, "learning_rate": 1.4750592569993663e-05, "loss": 0.3154, "step": 87675 }, { "epoch": 2.05, "learning_rate": 1.474971251554763e-05, "loss": 0.01, "step": 87680 }, { "epoch": 2.05, "learning_rate": 1.4748832461101593e-05, "loss": 0.0267, "step": 87685 }, { "epoch": 2.05, "learning_rate": 1.474795240665556e-05, "loss": 0.0342, "step": 87690 }, { "epoch": 2.05, "learning_rate": 1.4747072352209523e-05, "loss": 0.0385, "step": 87695 }, { "epoch": 2.05, "learning_rate": 1.474619229776349e-05, "loss": 0.0129, "step": 87700 }, { "epoch": 2.05, "learning_rate": 1.4745312243317453e-05, "loss": 0.0242, "step": 87705 }, { "epoch": 2.05, "learning_rate": 1.4744432188871418e-05, "loss": 0.1523, "step": 87710 }, { "epoch": 2.05, "learning_rate": 1.4743552134425383e-05, "loss": 0.312, "step": 87715 }, { "epoch": 2.05, "learning_rate": 1.4742672079979348e-05, "loss": 0.1736, "step": 87720 }, { "epoch": 2.05, "learning_rate": 1.4741792025533313e-05, "loss": 0.3298, "step": 87725 }, { "epoch": 2.05, "learning_rate": 1.4740911971087278e-05, "loss": 0.1248, "step": 87730 }, { "epoch": 2.05, "learning_rate": 1.4740031916641243e-05, "loss": 0.0211, "step": 87735 }, { "epoch": 2.05, "learning_rate": 1.4739151862195208e-05, "loss": 0.0889, "step": 87740 }, { "epoch": 2.05, "learning_rate": 1.4738271807749172e-05, "loss": 0.1555, "step": 87745 }, { "epoch": 2.05, "learning_rate": 1.4737391753303138e-05, "loss": 0.104, "step": 87750 }, { "epoch": 2.05, "learning_rate": 1.4736511698857102e-05, "loss": 0.0283, "step": 87755 }, { "epoch": 2.05, "learning_rate": 1.4735631644411068e-05, "loss": 0.2214, "step": 87760 }, { "epoch": 2.05, "learning_rate": 1.4734751589965032e-05, "loss": 0.3968, "step": 87765 }, { "epoch": 2.05, "learning_rate": 1.4733871535518999e-05, "loss": 0.0942, "step": 87770 }, { "epoch": 2.05, "learning_rate": 1.4732991481072962e-05, "loss": 0.3292, "step": 87775 }, { "epoch": 2.05, "learning_rate": 1.4732111426626927e-05, "loss": 0.0175, "step": 87780 }, { "epoch": 2.05, "learning_rate": 1.4731231372180894e-05, "loss": 0.0184, "step": 87785 }, { "epoch": 2.05, "learning_rate": 1.4730351317734857e-05, "loss": 0.0395, "step": 87790 }, { "epoch": 2.05, "learning_rate": 1.4729471263288824e-05, "loss": 0.0736, "step": 87795 }, { "epoch": 2.05, "learning_rate": 1.4728591208842787e-05, "loss": 0.0687, "step": 87800 }, { "epoch": 2.05, "learning_rate": 1.4727711154396754e-05, "loss": 0.0963, "step": 87805 }, { "epoch": 2.05, "learning_rate": 1.4726831099950717e-05, "loss": 0.0563, "step": 87810 }, { "epoch": 2.05, "learning_rate": 1.4725951045504684e-05, "loss": 0.0723, "step": 87815 }, { "epoch": 2.05, "learning_rate": 1.4725070991058647e-05, "loss": 0.3973, "step": 87820 }, { "epoch": 2.05, "learning_rate": 1.4724190936612612e-05, "loss": 0.3193, "step": 87825 }, { "epoch": 2.05, "learning_rate": 1.4723310882166577e-05, "loss": 0.0876, "step": 87830 }, { "epoch": 2.05, "learning_rate": 1.4722430827720542e-05, "loss": 0.0599, "step": 87835 }, { "epoch": 2.05, "learning_rate": 1.4721550773274507e-05, "loss": 0.0217, "step": 87840 }, { "epoch": 2.05, "learning_rate": 1.4720670718828472e-05, "loss": 0.1285, "step": 87845 }, { "epoch": 2.05, "learning_rate": 1.4719790664382437e-05, "loss": 0.2184, "step": 87850 }, { "epoch": 2.05, "learning_rate": 1.4718910609936402e-05, "loss": 0.1124, "step": 87855 }, { "epoch": 2.05, "learning_rate": 1.4718030555490366e-05, "loss": 0.0663, "step": 87860 }, { "epoch": 2.05, "learning_rate": 1.4717150501044332e-05, "loss": 0.2932, "step": 87865 }, { "epoch": 2.05, "learning_rate": 1.4716270446598296e-05, "loss": 0.3166, "step": 87870 }, { "epoch": 2.05, "learning_rate": 1.4715390392152262e-05, "loss": 0.3035, "step": 87875 }, { "epoch": 2.05, "learning_rate": 1.4714510337706226e-05, "loss": 0.1126, "step": 87880 }, { "epoch": 2.05, "learning_rate": 1.4713630283260192e-05, "loss": 0.0868, "step": 87885 }, { "epoch": 2.05, "learning_rate": 1.4712750228814156e-05, "loss": 0.0543, "step": 87890 }, { "epoch": 2.05, "learning_rate": 1.471187017436812e-05, "loss": 0.028, "step": 87895 }, { "epoch": 2.05, "learning_rate": 1.4710990119922086e-05, "loss": 0.0671, "step": 87900 }, { "epoch": 2.05, "learning_rate": 1.471011006547605e-05, "loss": 0.0991, "step": 87905 }, { "epoch": 2.05, "learning_rate": 1.4709230011030016e-05, "loss": 0.138, "step": 87910 }, { "epoch": 2.05, "learning_rate": 1.470834995658398e-05, "loss": 0.0767, "step": 87915 }, { "epoch": 2.05, "learning_rate": 1.4707469902137947e-05, "loss": 0.1077, "step": 87920 }, { "epoch": 2.05, "learning_rate": 1.470658984769191e-05, "loss": 0.2771, "step": 87925 }, { "epoch": 2.05, "learning_rate": 1.4705709793245876e-05, "loss": 0.0368, "step": 87930 }, { "epoch": 2.05, "learning_rate": 1.470482973879984e-05, "loss": 0.0292, "step": 87935 }, { "epoch": 2.05, "learning_rate": 1.4703949684353806e-05, "loss": 0.04, "step": 87940 }, { "epoch": 2.05, "learning_rate": 1.4703069629907771e-05, "loss": 0.0656, "step": 87945 }, { "epoch": 2.05, "learning_rate": 1.4702189575461736e-05, "loss": 0.1085, "step": 87950 }, { "epoch": 2.05, "learning_rate": 1.4701309521015701e-05, "loss": 0.2151, "step": 87955 }, { "epoch": 2.05, "learning_rate": 1.4700429466569666e-05, "loss": 0.1966, "step": 87960 }, { "epoch": 2.05, "learning_rate": 1.4699549412123631e-05, "loss": 0.0799, "step": 87965 }, { "epoch": 2.05, "learning_rate": 1.4698669357677596e-05, "loss": 0.1652, "step": 87970 }, { "epoch": 2.05, "learning_rate": 1.469778930323156e-05, "loss": 0.2074, "step": 87975 }, { "epoch": 2.05, "learning_rate": 1.4696909248785526e-05, "loss": 0.0063, "step": 87980 }, { "epoch": 2.05, "learning_rate": 1.469602919433949e-05, "loss": 0.0354, "step": 87985 }, { "epoch": 2.05, "learning_rate": 1.4695149139893456e-05, "loss": 0.0793, "step": 87990 }, { "epoch": 2.05, "learning_rate": 1.469426908544742e-05, "loss": 0.0536, "step": 87995 }, { "epoch": 2.05, "learning_rate": 1.4693389031001386e-05, "loss": 0.0426, "step": 88000 }, { "epoch": 2.05, "learning_rate": 1.469250897655535e-05, "loss": 0.102, "step": 88005 }, { "epoch": 2.05, "learning_rate": 1.4691628922109314e-05, "loss": 0.182, "step": 88010 }, { "epoch": 2.05, "learning_rate": 1.469074886766328e-05, "loss": 0.1889, "step": 88015 }, { "epoch": 2.05, "learning_rate": 1.4689868813217244e-05, "loss": 0.2034, "step": 88020 }, { "epoch": 2.05, "learning_rate": 1.468898875877121e-05, "loss": 0.365, "step": 88025 }, { "epoch": 2.05, "learning_rate": 1.4688108704325175e-05, "loss": 0.0332, "step": 88030 }, { "epoch": 2.05, "learning_rate": 1.468722864987914e-05, "loss": 0.0143, "step": 88035 }, { "epoch": 2.05, "learning_rate": 1.4686348595433105e-05, "loss": 0.0426, "step": 88040 }, { "epoch": 2.05, "learning_rate": 1.4685468540987068e-05, "loss": 0.1085, "step": 88045 }, { "epoch": 2.05, "learning_rate": 1.4684588486541035e-05, "loss": 0.1936, "step": 88050 }, { "epoch": 2.05, "learning_rate": 1.4683708432095e-05, "loss": 0.0413, "step": 88055 }, { "epoch": 2.05, "learning_rate": 1.4682828377648965e-05, "loss": 0.0974, "step": 88060 }, { "epoch": 2.05, "learning_rate": 1.468194832320293e-05, "loss": 0.108, "step": 88065 }, { "epoch": 2.05, "learning_rate": 1.4681068268756895e-05, "loss": 0.4516, "step": 88070 }, { "epoch": 2.05, "learning_rate": 1.468018821431086e-05, "loss": 0.3423, "step": 88075 }, { "epoch": 2.06, "learning_rate": 1.4679308159864823e-05, "loss": 0.0336, "step": 88080 }, { "epoch": 2.06, "learning_rate": 1.467842810541879e-05, "loss": 0.0819, "step": 88085 }, { "epoch": 2.06, "learning_rate": 1.4677548050972753e-05, "loss": 0.0078, "step": 88090 }, { "epoch": 2.06, "learning_rate": 1.467666799652672e-05, "loss": 0.0566, "step": 88095 }, { "epoch": 2.06, "learning_rate": 1.4675787942080683e-05, "loss": 0.0853, "step": 88100 }, { "epoch": 2.06, "learning_rate": 1.467490788763465e-05, "loss": 0.3602, "step": 88105 }, { "epoch": 2.06, "learning_rate": 1.4674027833188613e-05, "loss": 0.1258, "step": 88110 }, { "epoch": 2.06, "learning_rate": 1.467314777874258e-05, "loss": 0.1429, "step": 88115 }, { "epoch": 2.06, "learning_rate": 1.4672267724296543e-05, "loss": 0.3816, "step": 88120 }, { "epoch": 2.06, "learning_rate": 1.4671387669850508e-05, "loss": 0.2398, "step": 88125 }, { "epoch": 2.06, "learning_rate": 1.4670507615404473e-05, "loss": 0.0705, "step": 88130 }, { "epoch": 2.06, "learning_rate": 1.4669627560958438e-05, "loss": 0.0065, "step": 88135 }, { "epoch": 2.06, "learning_rate": 1.4668747506512403e-05, "loss": 0.0562, "step": 88140 }, { "epoch": 2.06, "learning_rate": 1.4667867452066368e-05, "loss": 0.0213, "step": 88145 }, { "epoch": 2.06, "learning_rate": 1.4666987397620333e-05, "loss": 0.0636, "step": 88150 }, { "epoch": 2.06, "learning_rate": 1.4666107343174298e-05, "loss": 0.0843, "step": 88155 }, { "epoch": 2.06, "learning_rate": 1.4665227288728262e-05, "loss": 0.0814, "step": 88160 }, { "epoch": 2.06, "learning_rate": 1.4664347234282228e-05, "loss": 0.2243, "step": 88165 }, { "epoch": 2.06, "learning_rate": 1.4663467179836192e-05, "loss": 0.3175, "step": 88170 }, { "epoch": 2.06, "learning_rate": 1.4662587125390158e-05, "loss": 0.1793, "step": 88175 }, { "epoch": 2.06, "learning_rate": 1.4661707070944122e-05, "loss": 0.0339, "step": 88180 }, { "epoch": 2.06, "learning_rate": 1.4660827016498088e-05, "loss": 0.0142, "step": 88185 }, { "epoch": 2.06, "learning_rate": 1.4659946962052053e-05, "loss": 0.0625, "step": 88190 }, { "epoch": 2.06, "learning_rate": 1.4659066907606017e-05, "loss": 0.0165, "step": 88195 }, { "epoch": 2.06, "learning_rate": 1.4658186853159984e-05, "loss": 0.0392, "step": 88200 }, { "epoch": 2.06, "learning_rate": 1.4657306798713947e-05, "loss": 0.1136, "step": 88205 }, { "epoch": 2.06, "learning_rate": 1.4656426744267914e-05, "loss": 0.129, "step": 88210 }, { "epoch": 2.06, "learning_rate": 1.4655546689821877e-05, "loss": 0.2566, "step": 88215 }, { "epoch": 2.06, "learning_rate": 1.4654666635375844e-05, "loss": 0.1933, "step": 88220 }, { "epoch": 2.06, "learning_rate": 1.4653786580929807e-05, "loss": 0.3663, "step": 88225 }, { "epoch": 2.06, "learning_rate": 1.4652906526483772e-05, "loss": 0.0278, "step": 88230 }, { "epoch": 2.06, "learning_rate": 1.4652026472037737e-05, "loss": 0.0215, "step": 88235 }, { "epoch": 2.06, "learning_rate": 1.4651146417591702e-05, "loss": 0.0053, "step": 88240 }, { "epoch": 2.06, "learning_rate": 1.4650266363145667e-05, "loss": 0.1, "step": 88245 }, { "epoch": 2.06, "learning_rate": 1.4649386308699632e-05, "loss": 0.0348, "step": 88250 }, { "epoch": 2.06, "learning_rate": 1.4648506254253597e-05, "loss": 0.1026, "step": 88255 }, { "epoch": 2.06, "learning_rate": 1.4647626199807562e-05, "loss": 0.1015, "step": 88260 }, { "epoch": 2.06, "learning_rate": 1.4646746145361527e-05, "loss": 0.2531, "step": 88265 }, { "epoch": 2.06, "learning_rate": 1.4645866090915492e-05, "loss": 0.5245, "step": 88270 }, { "epoch": 2.06, "learning_rate": 1.4644986036469455e-05, "loss": 0.2976, "step": 88275 }, { "epoch": 2.06, "learning_rate": 1.4644105982023422e-05, "loss": 0.0131, "step": 88280 }, { "epoch": 2.06, "learning_rate": 1.4643225927577385e-05, "loss": 0.0533, "step": 88285 }, { "epoch": 2.06, "learning_rate": 1.4642345873131352e-05, "loss": 0.0838, "step": 88290 }, { "epoch": 2.06, "learning_rate": 1.4641465818685316e-05, "loss": 0.1315, "step": 88295 }, { "epoch": 2.06, "learning_rate": 1.4640585764239282e-05, "loss": 0.08, "step": 88300 }, { "epoch": 2.06, "learning_rate": 1.4639705709793246e-05, "loss": 0.0999, "step": 88305 }, { "epoch": 2.06, "learning_rate": 1.463882565534721e-05, "loss": 0.2794, "step": 88310 }, { "epoch": 2.06, "learning_rate": 1.4637945600901176e-05, "loss": 0.215, "step": 88315 }, { "epoch": 2.06, "learning_rate": 1.463706554645514e-05, "loss": 0.1282, "step": 88320 }, { "epoch": 2.06, "learning_rate": 1.4636185492009107e-05, "loss": 0.2212, "step": 88325 }, { "epoch": 2.06, "learning_rate": 1.463530543756307e-05, "loss": 0.0707, "step": 88330 }, { "epoch": 2.06, "learning_rate": 1.4634425383117037e-05, "loss": 0.0012, "step": 88335 }, { "epoch": 2.06, "learning_rate": 1.4633545328671e-05, "loss": 0.0527, "step": 88340 }, { "epoch": 2.06, "learning_rate": 1.4632665274224966e-05, "loss": 0.0495, "step": 88345 }, { "epoch": 2.06, "learning_rate": 1.463178521977893e-05, "loss": 0.1249, "step": 88350 }, { "epoch": 2.06, "learning_rate": 1.4630905165332896e-05, "loss": 0.108, "step": 88355 }, { "epoch": 2.06, "learning_rate": 1.463002511088686e-05, "loss": 0.1036, "step": 88360 }, { "epoch": 2.06, "learning_rate": 1.4629145056440826e-05, "loss": 0.1513, "step": 88365 }, { "epoch": 2.06, "learning_rate": 1.462826500199479e-05, "loss": 0.3517, "step": 88370 }, { "epoch": 2.06, "learning_rate": 1.4627384947548756e-05, "loss": 0.3465, "step": 88375 }, { "epoch": 2.06, "learning_rate": 1.462650489310272e-05, "loss": 0.0299, "step": 88380 }, { "epoch": 2.06, "learning_rate": 1.4625624838656686e-05, "loss": 0.0531, "step": 88385 }, { "epoch": 2.06, "learning_rate": 1.462474478421065e-05, "loss": 0.1885, "step": 88390 }, { "epoch": 2.06, "learning_rate": 1.4623864729764616e-05, "loss": 0.0644, "step": 88395 }, { "epoch": 2.06, "learning_rate": 1.462298467531858e-05, "loss": 0.0351, "step": 88400 }, { "epoch": 2.06, "learning_rate": 1.4622104620872546e-05, "loss": 0.2008, "step": 88405 }, { "epoch": 2.06, "learning_rate": 1.462122456642651e-05, "loss": 0.1243, "step": 88410 }, { "epoch": 2.06, "learning_rate": 1.4620344511980476e-05, "loss": 0.3384, "step": 88415 }, { "epoch": 2.06, "learning_rate": 1.461946445753444e-05, "loss": 0.2003, "step": 88420 }, { "epoch": 2.06, "learning_rate": 1.4618584403088404e-05, "loss": 0.2308, "step": 88425 }, { "epoch": 2.06, "learning_rate": 1.461770434864237e-05, "loss": 0.1033, "step": 88430 }, { "epoch": 2.06, "learning_rate": 1.4616824294196334e-05, "loss": 0.1187, "step": 88435 }, { "epoch": 2.06, "learning_rate": 1.46159442397503e-05, "loss": 0.0349, "step": 88440 }, { "epoch": 2.06, "learning_rate": 1.4615064185304264e-05, "loss": 0.0846, "step": 88445 }, { "epoch": 2.06, "learning_rate": 1.461418413085823e-05, "loss": 0.1011, "step": 88450 }, { "epoch": 2.06, "learning_rate": 1.4613304076412194e-05, "loss": 0.1339, "step": 88455 }, { "epoch": 2.06, "learning_rate": 1.461242402196616e-05, "loss": 0.0933, "step": 88460 }, { "epoch": 2.06, "learning_rate": 1.4611543967520125e-05, "loss": 0.141, "step": 88465 }, { "epoch": 2.06, "learning_rate": 1.461066391307409e-05, "loss": 0.1904, "step": 88470 }, { "epoch": 2.06, "learning_rate": 1.4609783858628055e-05, "loss": 0.3512, "step": 88475 }, { "epoch": 2.06, "learning_rate": 1.460890380418202e-05, "loss": 0.0385, "step": 88480 }, { "epoch": 2.06, "learning_rate": 1.4608023749735985e-05, "loss": 0.0912, "step": 88485 }, { "epoch": 2.06, "learning_rate": 1.460714369528995e-05, "loss": 0.0492, "step": 88490 }, { "epoch": 2.06, "learning_rate": 1.4606263640843913e-05, "loss": 0.0374, "step": 88495 }, { "epoch": 2.06, "learning_rate": 1.460538358639788e-05, "loss": 0.0653, "step": 88500 }, { "epoch": 2.06, "learning_rate": 1.4604503531951843e-05, "loss": 0.0795, "step": 88505 }, { "epoch": 2.07, "learning_rate": 1.460362347750581e-05, "loss": 0.073, "step": 88510 }, { "epoch": 2.07, "learning_rate": 1.4602743423059773e-05, "loss": 0.131, "step": 88515 }, { "epoch": 2.07, "learning_rate": 1.460186336861374e-05, "loss": 0.2182, "step": 88520 }, { "epoch": 2.07, "learning_rate": 1.4600983314167703e-05, "loss": 0.5675, "step": 88525 }, { "epoch": 2.07, "learning_rate": 1.4600103259721668e-05, "loss": 0.0648, "step": 88530 }, { "epoch": 2.07, "learning_rate": 1.4599223205275633e-05, "loss": 0.0022, "step": 88535 }, { "epoch": 2.07, "learning_rate": 1.4598343150829598e-05, "loss": 0.0695, "step": 88540 }, { "epoch": 2.07, "learning_rate": 1.4597463096383563e-05, "loss": 0.0752, "step": 88545 }, { "epoch": 2.07, "learning_rate": 1.4596583041937528e-05, "loss": 0.0113, "step": 88550 }, { "epoch": 2.07, "learning_rate": 1.4595702987491493e-05, "loss": 0.1044, "step": 88555 }, { "epoch": 2.07, "learning_rate": 1.4594822933045458e-05, "loss": 0.1385, "step": 88560 }, { "epoch": 2.07, "learning_rate": 1.4593942878599422e-05, "loss": 0.1281, "step": 88565 }, { "epoch": 2.07, "learning_rate": 1.4593062824153388e-05, "loss": 0.3396, "step": 88570 }, { "epoch": 2.07, "learning_rate": 1.4592182769707352e-05, "loss": 0.2611, "step": 88575 }, { "epoch": 2.07, "learning_rate": 1.4591302715261318e-05, "loss": 0.0353, "step": 88580 }, { "epoch": 2.07, "learning_rate": 1.4590422660815282e-05, "loss": 0.0305, "step": 88585 }, { "epoch": 2.07, "learning_rate": 1.4589542606369248e-05, "loss": 0.0243, "step": 88590 }, { "epoch": 2.07, "learning_rate": 1.4588662551923213e-05, "loss": 0.0286, "step": 88595 }, { "epoch": 2.07, "learning_rate": 1.4587782497477178e-05, "loss": 0.107, "step": 88600 }, { "epoch": 2.07, "learning_rate": 1.4586902443031143e-05, "loss": 0.1478, "step": 88605 }, { "epoch": 2.07, "learning_rate": 1.4586022388585107e-05, "loss": 0.0855, "step": 88610 }, { "epoch": 2.07, "learning_rate": 1.4585142334139073e-05, "loss": 0.2512, "step": 88615 }, { "epoch": 2.07, "learning_rate": 1.4584262279693037e-05, "loss": 0.1398, "step": 88620 }, { "epoch": 2.07, "learning_rate": 1.4583382225247003e-05, "loss": 0.256, "step": 88625 }, { "epoch": 2.07, "learning_rate": 1.4582502170800967e-05, "loss": 0.0928, "step": 88630 }, { "epoch": 2.07, "learning_rate": 1.4581622116354934e-05, "loss": 0.0645, "step": 88635 }, { "epoch": 2.07, "learning_rate": 1.4580742061908897e-05, "loss": 0.0192, "step": 88640 }, { "epoch": 2.07, "learning_rate": 1.4579862007462862e-05, "loss": 0.0315, "step": 88645 }, { "epoch": 2.07, "learning_rate": 1.4578981953016827e-05, "loss": 0.1398, "step": 88650 }, { "epoch": 2.07, "learning_rate": 1.4578101898570792e-05, "loss": 0.0996, "step": 88655 }, { "epoch": 2.07, "learning_rate": 1.4577221844124757e-05, "loss": 0.1557, "step": 88660 }, { "epoch": 2.07, "learning_rate": 1.4576341789678722e-05, "loss": 0.1241, "step": 88665 }, { "epoch": 2.07, "learning_rate": 1.4575461735232687e-05, "loss": 0.3765, "step": 88670 }, { "epoch": 2.07, "learning_rate": 1.4574581680786652e-05, "loss": 0.3718, "step": 88675 }, { "epoch": 2.07, "learning_rate": 1.4573701626340615e-05, "loss": 0.0808, "step": 88680 }, { "epoch": 2.07, "learning_rate": 1.4572821571894582e-05, "loss": 0.048, "step": 88685 }, { "epoch": 2.07, "learning_rate": 1.4571941517448545e-05, "loss": 0.0566, "step": 88690 }, { "epoch": 2.07, "learning_rate": 1.4571061463002512e-05, "loss": 0.0352, "step": 88695 }, { "epoch": 2.07, "learning_rate": 1.4570181408556475e-05, "loss": 0.0769, "step": 88700 }, { "epoch": 2.07, "learning_rate": 1.4569301354110442e-05, "loss": 0.1044, "step": 88705 }, { "epoch": 2.07, "learning_rate": 1.4568421299664405e-05, "loss": 0.0452, "step": 88710 }, { "epoch": 2.07, "learning_rate": 1.456754124521837e-05, "loss": 0.1448, "step": 88715 }, { "epoch": 2.07, "learning_rate": 1.4566661190772335e-05, "loss": 0.2789, "step": 88720 }, { "epoch": 2.07, "learning_rate": 1.45657811363263e-05, "loss": 0.2029, "step": 88725 }, { "epoch": 2.07, "learning_rate": 1.4564901081880266e-05, "loss": 0.086, "step": 88730 }, { "epoch": 2.07, "learning_rate": 1.456402102743423e-05, "loss": 0.0816, "step": 88735 }, { "epoch": 2.07, "learning_rate": 1.4563140972988197e-05, "loss": 0.022, "step": 88740 }, { "epoch": 2.07, "learning_rate": 1.456226091854216e-05, "loss": 0.0285, "step": 88745 }, { "epoch": 2.07, "learning_rate": 1.4561380864096127e-05, "loss": 0.0379, "step": 88750 }, { "epoch": 2.07, "learning_rate": 1.456050080965009e-05, "loss": 0.0128, "step": 88755 }, { "epoch": 2.07, "learning_rate": 1.4559620755204056e-05, "loss": 0.0501, "step": 88760 }, { "epoch": 2.07, "learning_rate": 1.455874070075802e-05, "loss": 0.0679, "step": 88765 }, { "epoch": 2.07, "learning_rate": 1.4557860646311986e-05, "loss": 0.0338, "step": 88770 }, { "epoch": 2.07, "learning_rate": 1.455698059186595e-05, "loss": 0.2607, "step": 88775 }, { "epoch": 2.07, "learning_rate": 1.4556100537419916e-05, "loss": 0.1154, "step": 88780 }, { "epoch": 2.07, "learning_rate": 1.455522048297388e-05, "loss": 0.0047, "step": 88785 }, { "epoch": 2.07, "learning_rate": 1.4554340428527846e-05, "loss": 0.2118, "step": 88790 }, { "epoch": 2.07, "learning_rate": 1.4553460374081809e-05, "loss": 0.1133, "step": 88795 }, { "epoch": 2.07, "learning_rate": 1.4552580319635776e-05, "loss": 0.0471, "step": 88800 }, { "epoch": 2.07, "learning_rate": 1.4551700265189739e-05, "loss": 0.1729, "step": 88805 }, { "epoch": 2.07, "learning_rate": 1.4550820210743706e-05, "loss": 0.0686, "step": 88810 }, { "epoch": 2.07, "learning_rate": 1.454994015629767e-05, "loss": 0.0791, "step": 88815 }, { "epoch": 2.07, "learning_rate": 1.4549060101851636e-05, "loss": 0.3542, "step": 88820 }, { "epoch": 2.07, "learning_rate": 1.45481800474056e-05, "loss": 0.1274, "step": 88825 }, { "epoch": 2.07, "learning_rate": 1.4547299992959564e-05, "loss": 0.0297, "step": 88830 }, { "epoch": 2.07, "learning_rate": 1.454641993851353e-05, "loss": 0.1118, "step": 88835 }, { "epoch": 2.07, "learning_rate": 1.4545539884067494e-05, "loss": 0.0495, "step": 88840 }, { "epoch": 2.07, "learning_rate": 1.454465982962146e-05, "loss": 0.0362, "step": 88845 }, { "epoch": 2.07, "learning_rate": 1.4543779775175424e-05, "loss": 0.0437, "step": 88850 }, { "epoch": 2.07, "learning_rate": 1.454289972072939e-05, "loss": 0.088, "step": 88855 }, { "epoch": 2.07, "learning_rate": 1.4542019666283354e-05, "loss": 0.1058, "step": 88860 }, { "epoch": 2.07, "learning_rate": 1.454113961183732e-05, "loss": 0.1013, "step": 88865 }, { "epoch": 2.07, "learning_rate": 1.4540259557391284e-05, "loss": 0.191, "step": 88870 }, { "epoch": 2.07, "learning_rate": 1.453937950294525e-05, "loss": 0.2519, "step": 88875 }, { "epoch": 2.07, "learning_rate": 1.4538499448499214e-05, "loss": 0.0954, "step": 88880 }, { "epoch": 2.07, "learning_rate": 1.453761939405318e-05, "loss": 0.0575, "step": 88885 }, { "epoch": 2.07, "learning_rate": 1.4536739339607144e-05, "loss": 0.1091, "step": 88890 }, { "epoch": 2.07, "learning_rate": 1.453585928516111e-05, "loss": 0.0871, "step": 88895 }, { "epoch": 2.07, "learning_rate": 1.4534979230715075e-05, "loss": 0.0478, "step": 88900 }, { "epoch": 2.07, "learning_rate": 1.453409917626904e-05, "loss": 0.0983, "step": 88905 }, { "epoch": 2.07, "learning_rate": 1.4533219121823003e-05, "loss": 0.1307, "step": 88910 }, { "epoch": 2.07, "learning_rate": 1.453233906737697e-05, "loss": 0.1166, "step": 88915 }, { "epoch": 2.07, "learning_rate": 1.4531459012930933e-05, "loss": 0.1945, "step": 88920 }, { "epoch": 2.07, "learning_rate": 1.45305789584849e-05, "loss": 0.188, "step": 88925 }, { "epoch": 2.07, "learning_rate": 1.4529698904038863e-05, "loss": 0.0598, "step": 88930 }, { "epoch": 2.07, "learning_rate": 1.452881884959283e-05, "loss": 0.0606, "step": 88935 }, { "epoch": 2.08, "learning_rate": 1.4527938795146793e-05, "loss": 0.0788, "step": 88940 }, { "epoch": 2.08, "learning_rate": 1.4527058740700758e-05, "loss": 0.0806, "step": 88945 }, { "epoch": 2.08, "learning_rate": 1.4526178686254723e-05, "loss": 0.0459, "step": 88950 }, { "epoch": 2.08, "learning_rate": 1.4525298631808688e-05, "loss": 0.1017, "step": 88955 }, { "epoch": 2.08, "learning_rate": 1.4524418577362653e-05, "loss": 0.0878, "step": 88960 }, { "epoch": 2.08, "learning_rate": 1.4523538522916618e-05, "loss": 0.1638, "step": 88965 }, { "epoch": 2.08, "learning_rate": 1.4522658468470583e-05, "loss": 0.1197, "step": 88970 }, { "epoch": 2.08, "learning_rate": 1.4521778414024548e-05, "loss": 0.3, "step": 88975 }, { "epoch": 2.08, "learning_rate": 1.4520898359578511e-05, "loss": 0.0309, "step": 88980 }, { "epoch": 2.08, "learning_rate": 1.4520018305132478e-05, "loss": 0.0542, "step": 88985 }, { "epoch": 2.08, "learning_rate": 1.4519138250686442e-05, "loss": 0.0733, "step": 88990 }, { "epoch": 2.08, "learning_rate": 1.4518258196240408e-05, "loss": 0.0793, "step": 88995 }, { "epoch": 2.08, "learning_rate": 1.4517378141794372e-05, "loss": 0.038, "step": 89000 }, { "epoch": 2.08, "learning_rate": 1.4516498087348338e-05, "loss": 0.3077, "step": 89005 }, { "epoch": 2.08, "learning_rate": 1.4515618032902303e-05, "loss": 0.1763, "step": 89010 }, { "epoch": 2.08, "learning_rate": 1.4514737978456267e-05, "loss": 0.1046, "step": 89015 }, { "epoch": 2.08, "learning_rate": 1.451403393489944e-05, "loss": 0.416, "step": 89020 }, { "epoch": 2.08, "learning_rate": 1.4513153880453403e-05, "loss": 0.1861, "step": 89025 }, { "epoch": 2.08, "learning_rate": 1.451227382600737e-05, "loss": 0.036, "step": 89030 }, { "epoch": 2.08, "learning_rate": 1.4511393771561333e-05, "loss": 0.0008, "step": 89035 }, { "epoch": 2.08, "learning_rate": 1.45105137171153e-05, "loss": 0.0568, "step": 89040 }, { "epoch": 2.08, "learning_rate": 1.4509633662669263e-05, "loss": 0.0219, "step": 89045 }, { "epoch": 2.08, "learning_rate": 1.450875360822323e-05, "loss": 0.031, "step": 89050 }, { "epoch": 2.08, "learning_rate": 1.4507873553777193e-05, "loss": 0.0384, "step": 89055 }, { "epoch": 2.08, "learning_rate": 1.450699349933116e-05, "loss": 0.165, "step": 89060 }, { "epoch": 2.08, "learning_rate": 1.4506113444885125e-05, "loss": 0.1802, "step": 89065 }, { "epoch": 2.08, "learning_rate": 1.4505233390439088e-05, "loss": 0.1747, "step": 89070 }, { "epoch": 2.08, "learning_rate": 1.4504353335993055e-05, "loss": 0.1497, "step": 89075 }, { "epoch": 2.08, "learning_rate": 1.4503473281547018e-05, "loss": 0.0454, "step": 89080 }, { "epoch": 2.08, "learning_rate": 1.4502593227100985e-05, "loss": 0.0376, "step": 89085 }, { "epoch": 2.08, "learning_rate": 1.4501713172654948e-05, "loss": 0.0246, "step": 89090 }, { "epoch": 2.08, "learning_rate": 1.4500833118208915e-05, "loss": 0.0045, "step": 89095 }, { "epoch": 2.08, "learning_rate": 1.4499953063762878e-05, "loss": 0.0373, "step": 89100 }, { "epoch": 2.08, "learning_rate": 1.4499073009316843e-05, "loss": 0.0336, "step": 89105 }, { "epoch": 2.08, "learning_rate": 1.4498192954870808e-05, "loss": 0.0522, "step": 89110 }, { "epoch": 2.08, "learning_rate": 1.4497312900424773e-05, "loss": 0.1211, "step": 89115 }, { "epoch": 2.08, "learning_rate": 1.4496432845978738e-05, "loss": 0.2042, "step": 89120 }, { "epoch": 2.08, "learning_rate": 1.4495552791532703e-05, "loss": 0.2913, "step": 89125 }, { "epoch": 2.08, "learning_rate": 1.4494672737086668e-05, "loss": 0.0695, "step": 89130 }, { "epoch": 2.08, "learning_rate": 1.4493792682640633e-05, "loss": 0.0312, "step": 89135 }, { "epoch": 2.08, "learning_rate": 1.4492912628194597e-05, "loss": 0.0764, "step": 89140 }, { "epoch": 2.08, "learning_rate": 1.4492032573748563e-05, "loss": 0.0812, "step": 89145 }, { "epoch": 2.08, "learning_rate": 1.4491152519302527e-05, "loss": 0.161, "step": 89150 }, { "epoch": 2.08, "learning_rate": 1.4490272464856494e-05, "loss": 0.1067, "step": 89155 }, { "epoch": 2.08, "learning_rate": 1.4489392410410457e-05, "loss": 0.188, "step": 89160 }, { "epoch": 2.08, "learning_rate": 1.4488512355964424e-05, "loss": 0.026, "step": 89165 }, { "epoch": 2.08, "learning_rate": 1.4487632301518387e-05, "loss": 0.1324, "step": 89170 }, { "epoch": 2.08, "learning_rate": 1.4486752247072352e-05, "loss": 0.2056, "step": 89175 }, { "epoch": 2.08, "learning_rate": 1.4485872192626317e-05, "loss": 0.0565, "step": 89180 }, { "epoch": 2.08, "learning_rate": 1.4484992138180282e-05, "loss": 0.0731, "step": 89185 }, { "epoch": 2.08, "learning_rate": 1.4484112083734247e-05, "loss": 0.0498, "step": 89190 }, { "epoch": 2.08, "learning_rate": 1.4483232029288212e-05, "loss": 0.0911, "step": 89195 }, { "epoch": 2.08, "learning_rate": 1.4482351974842179e-05, "loss": 0.0784, "step": 89200 }, { "epoch": 2.08, "learning_rate": 1.4481471920396142e-05, "loss": 0.0464, "step": 89205 }, { "epoch": 2.08, "learning_rate": 1.4480591865950109e-05, "loss": 0.1359, "step": 89210 }, { "epoch": 2.08, "learning_rate": 1.4479711811504072e-05, "loss": 0.0831, "step": 89215 }, { "epoch": 2.08, "learning_rate": 1.4478831757058037e-05, "loss": 0.1637, "step": 89220 }, { "epoch": 2.08, "learning_rate": 1.4477951702612002e-05, "loss": 0.262, "step": 89225 }, { "epoch": 2.08, "learning_rate": 1.4477071648165967e-05, "loss": 0.1163, "step": 89230 }, { "epoch": 2.08, "learning_rate": 1.4476191593719932e-05, "loss": 0.032, "step": 89235 }, { "epoch": 2.08, "learning_rate": 1.4475311539273897e-05, "loss": 0.0324, "step": 89240 }, { "epoch": 2.08, "learning_rate": 1.4474431484827862e-05, "loss": 0.0204, "step": 89245 }, { "epoch": 2.08, "learning_rate": 1.4473551430381827e-05, "loss": 0.1105, "step": 89250 }, { "epoch": 2.08, "learning_rate": 1.447267137593579e-05, "loss": 0.0667, "step": 89255 }, { "epoch": 2.08, "learning_rate": 1.4471791321489757e-05, "loss": 0.1575, "step": 89260 }, { "epoch": 2.08, "learning_rate": 1.447091126704372e-05, "loss": 0.1551, "step": 89265 }, { "epoch": 2.08, "learning_rate": 1.4470031212597687e-05, "loss": 0.0409, "step": 89270 }, { "epoch": 2.08, "learning_rate": 1.446915115815165e-05, "loss": 0.1506, "step": 89275 }, { "epoch": 2.08, "learning_rate": 1.4468271103705617e-05, "loss": 0.1095, "step": 89280 }, { "epoch": 2.08, "learning_rate": 1.446739104925958e-05, "loss": 0.0007, "step": 89285 }, { "epoch": 2.08, "learning_rate": 1.4466510994813546e-05, "loss": 0.0173, "step": 89290 }, { "epoch": 2.08, "learning_rate": 1.446563094036751e-05, "loss": 0.0574, "step": 89295 }, { "epoch": 2.08, "learning_rate": 1.4464750885921476e-05, "loss": 0.0589, "step": 89300 }, { "epoch": 2.08, "learning_rate": 1.446387083147544e-05, "loss": 0.1841, "step": 89305 }, { "epoch": 2.08, "learning_rate": 1.4462990777029406e-05, "loss": 0.1676, "step": 89310 }, { "epoch": 2.08, "learning_rate": 1.446211072258337e-05, "loss": 0.1609, "step": 89315 }, { "epoch": 2.08, "learning_rate": 1.4461230668137336e-05, "loss": 0.2948, "step": 89320 }, { "epoch": 2.08, "learning_rate": 1.4460350613691299e-05, "loss": 0.2525, "step": 89325 }, { "epoch": 2.08, "learning_rate": 1.4459470559245266e-05, "loss": 0.0926, "step": 89330 }, { "epoch": 2.08, "learning_rate": 1.4458590504799231e-05, "loss": 0.0219, "step": 89335 }, { "epoch": 2.08, "learning_rate": 1.4457710450353196e-05, "loss": 0.0899, "step": 89340 }, { "epoch": 2.08, "learning_rate": 1.4456830395907161e-05, "loss": 0.1588, "step": 89345 }, { "epoch": 2.08, "learning_rate": 1.4455950341461126e-05, "loss": 0.1317, "step": 89350 }, { "epoch": 2.08, "learning_rate": 1.4455070287015091e-05, "loss": 0.1604, "step": 89355 }, { "epoch": 2.08, "learning_rate": 1.4454190232569054e-05, "loss": 0.0893, "step": 89360 }, { "epoch": 2.08, "learning_rate": 1.4453310178123021e-05, "loss": 0.0689, "step": 89365 }, { "epoch": 2.09, "learning_rate": 1.4452430123676984e-05, "loss": 0.2872, "step": 89370 }, { "epoch": 2.09, "learning_rate": 1.4451550069230951e-05, "loss": 0.2615, "step": 89375 }, { "epoch": 2.09, "learning_rate": 1.4450670014784914e-05, "loss": 0.0887, "step": 89380 }, { "epoch": 2.09, "learning_rate": 1.4449789960338881e-05, "loss": 0.0177, "step": 89385 }, { "epoch": 2.09, "learning_rate": 1.4448909905892844e-05, "loss": 0.0358, "step": 89390 }, { "epoch": 2.09, "learning_rate": 1.4448029851446811e-05, "loss": 0.0156, "step": 89395 }, { "epoch": 2.09, "learning_rate": 1.4447149797000774e-05, "loss": 0.0178, "step": 89400 }, { "epoch": 2.09, "learning_rate": 1.444626974255474e-05, "loss": 0.0725, "step": 89405 }, { "epoch": 2.09, "learning_rate": 1.4445389688108704e-05, "loss": 0.269, "step": 89410 }, { "epoch": 2.09, "learning_rate": 1.444450963366267e-05, "loss": 0.0587, "step": 89415 }, { "epoch": 2.09, "learning_rate": 1.4443629579216635e-05, "loss": 0.208, "step": 89420 }, { "epoch": 2.09, "learning_rate": 1.44427495247706e-05, "loss": 0.3226, "step": 89425 }, { "epoch": 2.09, "learning_rate": 1.4441869470324565e-05, "loss": 0.0269, "step": 89430 }, { "epoch": 2.09, "learning_rate": 1.444098941587853e-05, "loss": 0.0007, "step": 89435 }, { "epoch": 2.09, "learning_rate": 1.4440109361432493e-05, "loss": 0.0671, "step": 89440 }, { "epoch": 2.09, "learning_rate": 1.443922930698646e-05, "loss": 0.1235, "step": 89445 }, { "epoch": 2.09, "learning_rate": 1.4438349252540423e-05, "loss": 0.0838, "step": 89450 }, { "epoch": 2.09, "learning_rate": 1.443746919809439e-05, "loss": 0.0677, "step": 89455 }, { "epoch": 2.09, "learning_rate": 1.4436589143648353e-05, "loss": 0.2287, "step": 89460 }, { "epoch": 2.09, "learning_rate": 1.443570908920232e-05, "loss": 0.1086, "step": 89465 }, { "epoch": 2.09, "learning_rate": 1.4434829034756285e-05, "loss": 0.2412, "step": 89470 }, { "epoch": 2.09, "learning_rate": 1.4433948980310248e-05, "loss": 0.2953, "step": 89475 }, { "epoch": 2.09, "learning_rate": 1.4433068925864215e-05, "loss": 0.0791, "step": 89480 }, { "epoch": 2.09, "learning_rate": 1.4432188871418178e-05, "loss": 0.0751, "step": 89485 }, { "epoch": 2.09, "learning_rate": 1.4431308816972145e-05, "loss": 0.0229, "step": 89490 }, { "epoch": 2.09, "learning_rate": 1.4430428762526108e-05, "loss": 0.0895, "step": 89495 }, { "epoch": 2.09, "learning_rate": 1.4429548708080075e-05, "loss": 0.0845, "step": 89500 }, { "epoch": 2.09, "learning_rate": 1.4428668653634038e-05, "loss": 0.0162, "step": 89505 }, { "epoch": 2.09, "learning_rate": 1.4427788599188003e-05, "loss": 0.0865, "step": 89510 }, { "epoch": 2.09, "learning_rate": 1.4426908544741968e-05, "loss": 0.1057, "step": 89515 }, { "epoch": 2.09, "learning_rate": 1.4426028490295933e-05, "loss": 0.1716, "step": 89520 }, { "epoch": 2.09, "learning_rate": 1.4425148435849898e-05, "loss": 0.174, "step": 89525 }, { "epoch": 2.09, "learning_rate": 1.4424268381403863e-05, "loss": 0.0185, "step": 89530 }, { "epoch": 2.09, "learning_rate": 1.4423388326957828e-05, "loss": 0.062, "step": 89535 }, { "epoch": 2.09, "learning_rate": 1.4422508272511793e-05, "loss": 0.1572, "step": 89540 }, { "epoch": 2.09, "learning_rate": 1.4421628218065758e-05, "loss": 0.1388, "step": 89545 }, { "epoch": 2.09, "learning_rate": 1.4420748163619723e-05, "loss": 0.1224, "step": 89550 }, { "epoch": 2.09, "learning_rate": 1.4419868109173687e-05, "loss": 0.2071, "step": 89555 }, { "epoch": 2.09, "learning_rate": 1.4418988054727653e-05, "loss": 0.1888, "step": 89560 }, { "epoch": 2.09, "learning_rate": 1.4418108000281617e-05, "loss": 0.1504, "step": 89565 }, { "epoch": 2.09, "learning_rate": 1.4417227945835583e-05, "loss": 0.4289, "step": 89570 }, { "epoch": 2.09, "learning_rate": 1.4416347891389547e-05, "loss": 0.4284, "step": 89575 }, { "epoch": 2.09, "learning_rate": 1.4415467836943513e-05, "loss": 0.0156, "step": 89580 }, { "epoch": 2.09, "learning_rate": 1.4414587782497477e-05, "loss": 0.0408, "step": 89585 }, { "epoch": 2.09, "learning_rate": 1.4413707728051442e-05, "loss": 0.0908, "step": 89590 }, { "epoch": 2.09, "learning_rate": 1.4412827673605407e-05, "loss": 0.049, "step": 89595 }, { "epoch": 2.09, "learning_rate": 1.4411947619159372e-05, "loss": 0.0974, "step": 89600 }, { "epoch": 2.09, "learning_rate": 1.4411067564713339e-05, "loss": 0.0971, "step": 89605 }, { "epoch": 2.09, "learning_rate": 1.4410187510267302e-05, "loss": 0.173, "step": 89610 }, { "epoch": 2.09, "learning_rate": 1.4409307455821269e-05, "loss": 0.2746, "step": 89615 }, { "epoch": 2.09, "learning_rate": 1.4408427401375232e-05, "loss": 0.1315, "step": 89620 }, { "epoch": 2.09, "learning_rate": 1.4407547346929197e-05, "loss": 0.2886, "step": 89625 }, { "epoch": 2.09, "learning_rate": 1.4406667292483162e-05, "loss": 0.0383, "step": 89630 }, { "epoch": 2.09, "learning_rate": 1.4405787238037127e-05, "loss": 0.0212, "step": 89635 }, { "epoch": 2.09, "learning_rate": 1.4404907183591092e-05, "loss": 0.1146, "step": 89640 }, { "epoch": 2.09, "learning_rate": 1.4404027129145057e-05, "loss": 0.195, "step": 89645 }, { "epoch": 2.09, "learning_rate": 1.4403147074699022e-05, "loss": 0.0327, "step": 89650 }, { "epoch": 2.09, "learning_rate": 1.4402267020252987e-05, "loss": 0.094, "step": 89655 }, { "epoch": 2.09, "learning_rate": 1.440138696580695e-05, "loss": 0.1202, "step": 89660 }, { "epoch": 2.09, "learning_rate": 1.4400682922250123e-05, "loss": 0.2409, "step": 89665 }, { "epoch": 2.09, "learning_rate": 1.4399802867804088e-05, "loss": 0.2145, "step": 89670 }, { "epoch": 2.09, "learning_rate": 1.4398922813358053e-05, "loss": 0.2325, "step": 89675 }, { "epoch": 2.09, "learning_rate": 1.4398042758912019e-05, "loss": 0.0389, "step": 89680 }, { "epoch": 2.09, "learning_rate": 1.4397162704465984e-05, "loss": 0.0106, "step": 89685 }, { "epoch": 2.09, "learning_rate": 1.4396282650019949e-05, "loss": 0.1088, "step": 89690 }, { "epoch": 2.09, "learning_rate": 1.4395402595573914e-05, "loss": 0.1338, "step": 89695 }, { "epoch": 2.09, "learning_rate": 1.4394522541127879e-05, "loss": 0.0993, "step": 89700 }, { "epoch": 2.09, "learning_rate": 1.4393642486681844e-05, "loss": 0.0692, "step": 89705 }, { "epoch": 2.09, "learning_rate": 1.4392762432235809e-05, "loss": 0.0878, "step": 89710 }, { "epoch": 2.09, "learning_rate": 1.4391882377789772e-05, "loss": 0.2799, "step": 89715 }, { "epoch": 2.09, "learning_rate": 1.4391002323343739e-05, "loss": 0.3418, "step": 89720 }, { "epoch": 2.09, "learning_rate": 1.4390122268897702e-05, "loss": 0.4016, "step": 89725 }, { "epoch": 2.09, "learning_rate": 1.4389242214451669e-05, "loss": 0.0595, "step": 89730 }, { "epoch": 2.09, "learning_rate": 1.4388362160005632e-05, "loss": 0.0386, "step": 89735 }, { "epoch": 2.09, "learning_rate": 1.4387482105559599e-05, "loss": 0.0918, "step": 89740 }, { "epoch": 2.09, "learning_rate": 1.4386602051113562e-05, "loss": 0.0147, "step": 89745 }, { "epoch": 2.09, "learning_rate": 1.4385721996667527e-05, "loss": 0.0666, "step": 89750 }, { "epoch": 2.09, "learning_rate": 1.4384841942221492e-05, "loss": 0.0685, "step": 89755 }, { "epoch": 2.09, "learning_rate": 1.4383961887775457e-05, "loss": 0.0903, "step": 89760 }, { "epoch": 2.09, "learning_rate": 1.4383081833329422e-05, "loss": 0.1984, "step": 89765 }, { "epoch": 2.09, "learning_rate": 1.4382201778883387e-05, "loss": 0.203, "step": 89770 }, { "epoch": 2.09, "learning_rate": 1.4381321724437352e-05, "loss": 0.3033, "step": 89775 }, { "epoch": 2.09, "learning_rate": 1.4380441669991317e-05, "loss": 0.0374, "step": 89780 }, { "epoch": 2.09, "learning_rate": 1.437956161554528e-05, "loss": 0.0186, "step": 89785 }, { "epoch": 2.09, "learning_rate": 1.4378681561099247e-05, "loss": 0.0314, "step": 89790 }, { "epoch": 2.1, "learning_rate": 1.4377801506653212e-05, "loss": 0.069, "step": 89795 }, { "epoch": 2.1, "learning_rate": 1.4376921452207177e-05, "loss": 0.0217, "step": 89800 }, { "epoch": 2.1, "learning_rate": 1.4376041397761142e-05, "loss": 0.046, "step": 89805 }, { "epoch": 2.1, "learning_rate": 1.4375161343315107e-05, "loss": 0.2705, "step": 89810 }, { "epoch": 2.1, "learning_rate": 1.4374281288869072e-05, "loss": 0.0443, "step": 89815 }, { "epoch": 2.1, "learning_rate": 1.4373401234423036e-05, "loss": 0.1877, "step": 89820 }, { "epoch": 2.1, "learning_rate": 1.4372521179977002e-05, "loss": 0.2316, "step": 89825 }, { "epoch": 2.1, "learning_rate": 1.4371641125530966e-05, "loss": 0.1762, "step": 89830 }, { "epoch": 2.1, "learning_rate": 1.4370761071084932e-05, "loss": 0.0055, "step": 89835 }, { "epoch": 2.1, "learning_rate": 1.4369881016638896e-05, "loss": 0.0719, "step": 89840 }, { "epoch": 2.1, "learning_rate": 1.4369000962192862e-05, "loss": 0.0182, "step": 89845 }, { "epoch": 2.1, "learning_rate": 1.4368120907746826e-05, "loss": 0.094, "step": 89850 }, { "epoch": 2.1, "learning_rate": 1.4367240853300793e-05, "loss": 0.0078, "step": 89855 }, { "epoch": 2.1, "learning_rate": 1.4366360798854756e-05, "loss": 0.1718, "step": 89860 }, { "epoch": 2.1, "learning_rate": 1.4365480744408721e-05, "loss": 0.0911, "step": 89865 }, { "epoch": 2.1, "learning_rate": 1.4364600689962686e-05, "loss": 0.4466, "step": 89870 }, { "epoch": 2.1, "learning_rate": 1.4363720635516651e-05, "loss": 0.3079, "step": 89875 }, { "epoch": 2.1, "learning_rate": 1.4362840581070616e-05, "loss": 0.0273, "step": 89880 }, { "epoch": 2.1, "learning_rate": 1.4361960526624581e-05, "loss": 0.0468, "step": 89885 }, { "epoch": 2.1, "learning_rate": 1.4361080472178546e-05, "loss": 0.0618, "step": 89890 }, { "epoch": 2.1, "learning_rate": 1.4360200417732511e-05, "loss": 0.0625, "step": 89895 }, { "epoch": 2.1, "learning_rate": 1.4359320363286474e-05, "loss": 0.1405, "step": 89900 }, { "epoch": 2.1, "learning_rate": 1.4358440308840441e-05, "loss": 0.1271, "step": 89905 }, { "epoch": 2.1, "learning_rate": 1.4357560254394404e-05, "loss": 0.125, "step": 89910 }, { "epoch": 2.1, "learning_rate": 1.4356680199948371e-05, "loss": 0.2036, "step": 89915 }, { "epoch": 2.1, "learning_rate": 1.4355800145502334e-05, "loss": 0.2162, "step": 89920 }, { "epoch": 2.1, "learning_rate": 1.4354920091056301e-05, "loss": 0.2339, "step": 89925 }, { "epoch": 2.1, "learning_rate": 1.4354040036610266e-05, "loss": 0.034, "step": 89930 }, { "epoch": 2.1, "learning_rate": 1.435315998216423e-05, "loss": 0.0282, "step": 89935 }, { "epoch": 2.1, "learning_rate": 1.4352279927718196e-05, "loss": 0.0408, "step": 89940 }, { "epoch": 2.1, "learning_rate": 1.435139987327216e-05, "loss": 0.0708, "step": 89945 }, { "epoch": 2.1, "learning_rate": 1.4350519818826126e-05, "loss": 0.0584, "step": 89950 }, { "epoch": 2.1, "learning_rate": 1.434963976438009e-05, "loss": 0.0642, "step": 89955 }, { "epoch": 2.1, "learning_rate": 1.4348759709934056e-05, "loss": 0.091, "step": 89960 }, { "epoch": 2.1, "learning_rate": 1.434787965548802e-05, "loss": 0.1639, "step": 89965 }, { "epoch": 2.1, "learning_rate": 1.4346999601041985e-05, "loss": 0.1519, "step": 89970 }, { "epoch": 2.1, "learning_rate": 1.434611954659595e-05, "loss": 0.3158, "step": 89975 }, { "epoch": 2.1, "learning_rate": 1.4345239492149915e-05, "loss": 0.1349, "step": 89980 }, { "epoch": 2.1, "learning_rate": 1.434435943770388e-05, "loss": 0.0324, "step": 89985 }, { "epoch": 2.1, "learning_rate": 1.4343479383257845e-05, "loss": 0.0758, "step": 89990 }, { "epoch": 2.1, "learning_rate": 1.434259932881181e-05, "loss": 0.0124, "step": 89995 }, { "epoch": 2.1, "learning_rate": 1.4341719274365775e-05, "loss": 0.0481, "step": 90000 }, { "epoch": 2.1, "learning_rate": 1.434083921991974e-05, "loss": 0.2135, "step": 90005 }, { "epoch": 2.1, "learning_rate": 1.4339959165473705e-05, "loss": 0.1201, "step": 90010 }, { "epoch": 2.1, "learning_rate": 1.4339079111027668e-05, "loss": 0.1341, "step": 90015 }, { "epoch": 2.1, "learning_rate": 1.4338199056581635e-05, "loss": 0.2193, "step": 90020 }, { "epoch": 2.1, "learning_rate": 1.4337319002135598e-05, "loss": 0.2763, "step": 90025 }, { "epoch": 2.1, "learning_rate": 1.4336438947689565e-05, "loss": 0.0622, "step": 90030 }, { "epoch": 2.1, "learning_rate": 1.4335558893243528e-05, "loss": 0.0466, "step": 90035 }, { "epoch": 2.1, "learning_rate": 1.4334678838797495e-05, "loss": 0.033, "step": 90040 }, { "epoch": 2.1, "learning_rate": 1.4333798784351458e-05, "loss": 0.2442, "step": 90045 }, { "epoch": 2.1, "learning_rate": 1.4332918729905423e-05, "loss": 0.0645, "step": 90050 }, { "epoch": 2.1, "learning_rate": 1.4332038675459388e-05, "loss": 0.031, "step": 90055 }, { "epoch": 2.1, "learning_rate": 1.4331158621013353e-05, "loss": 0.1481, "step": 90060 }, { "epoch": 2.1, "learning_rate": 1.4330278566567318e-05, "loss": 0.1188, "step": 90065 }, { "epoch": 2.1, "learning_rate": 1.4329398512121283e-05, "loss": 0.2029, "step": 90070 }, { "epoch": 2.1, "learning_rate": 1.432851845767525e-05, "loss": 0.3187, "step": 90075 }, { "epoch": 2.1, "learning_rate": 1.4327638403229213e-05, "loss": 0.0214, "step": 90080 }, { "epoch": 2.1, "learning_rate": 1.4326758348783178e-05, "loss": 0.1322, "step": 90085 }, { "epoch": 2.1, "learning_rate": 1.4325878294337143e-05, "loss": 0.0575, "step": 90090 }, { "epoch": 2.1, "learning_rate": 1.4324998239891108e-05, "loss": 0.1153, "step": 90095 }, { "epoch": 2.1, "learning_rate": 1.4324118185445073e-05, "loss": 0.0853, "step": 90100 }, { "epoch": 2.1, "learning_rate": 1.4323238130999038e-05, "loss": 0.0543, "step": 90105 }, { "epoch": 2.1, "learning_rate": 1.4322358076553004e-05, "loss": 0.0176, "step": 90110 }, { "epoch": 2.1, "learning_rate": 1.4321478022106969e-05, "loss": 0.132, "step": 90115 }, { "epoch": 2.1, "learning_rate": 1.4320597967660932e-05, "loss": 0.1168, "step": 90120 }, { "epoch": 2.1, "learning_rate": 1.4319717913214899e-05, "loss": 0.1933, "step": 90125 }, { "epoch": 2.1, "learning_rate": 1.4318837858768862e-05, "loss": 0.0186, "step": 90130 }, { "epoch": 2.1, "learning_rate": 1.4317957804322829e-05, "loss": 0.0076, "step": 90135 }, { "epoch": 2.1, "learning_rate": 1.4317077749876792e-05, "loss": 0.0859, "step": 90140 }, { "epoch": 2.1, "learning_rate": 1.4316197695430759e-05, "loss": 0.1136, "step": 90145 }, { "epoch": 2.1, "learning_rate": 1.4315317640984722e-05, "loss": 0.078, "step": 90150 }, { "epoch": 2.1, "learning_rate": 1.4314437586538687e-05, "loss": 0.083, "step": 90155 }, { "epoch": 2.1, "learning_rate": 1.4313557532092652e-05, "loss": 0.137, "step": 90160 }, { "epoch": 2.1, "learning_rate": 1.4312677477646617e-05, "loss": 0.1241, "step": 90165 }, { "epoch": 2.1, "learning_rate": 1.4311797423200582e-05, "loss": 0.2083, "step": 90170 }, { "epoch": 2.1, "learning_rate": 1.4310917368754547e-05, "loss": 0.2461, "step": 90175 }, { "epoch": 2.1, "learning_rate": 1.4310037314308512e-05, "loss": 0.005, "step": 90180 }, { "epoch": 2.1, "learning_rate": 1.4309157259862477e-05, "loss": 0.0021, "step": 90185 }, { "epoch": 2.1, "learning_rate": 1.4308277205416442e-05, "loss": 0.1219, "step": 90190 }, { "epoch": 2.1, "learning_rate": 1.4307397150970407e-05, "loss": 0.0516, "step": 90195 }, { "epoch": 2.1, "learning_rate": 1.4306517096524372e-05, "loss": 0.0755, "step": 90200 }, { "epoch": 2.1, "learning_rate": 1.4305637042078337e-05, "loss": 0.1857, "step": 90205 }, { "epoch": 2.1, "learning_rate": 1.4304756987632302e-05, "loss": 0.1, "step": 90210 }, { "epoch": 2.1, "learning_rate": 1.4303876933186267e-05, "loss": 0.0971, "step": 90215 }, { "epoch": 2.1, "learning_rate": 1.4302996878740232e-05, "loss": 0.2222, "step": 90220 }, { "epoch": 2.11, "learning_rate": 1.4302116824294197e-05, "loss": 0.2387, "step": 90225 }, { "epoch": 2.11, "learning_rate": 1.4301236769848162e-05, "loss": 0.0325, "step": 90230 }, { "epoch": 2.11, "learning_rate": 1.4300356715402126e-05, "loss": 0.1304, "step": 90235 }, { "epoch": 2.11, "learning_rate": 1.4299476660956092e-05, "loss": 0.1251, "step": 90240 }, { "epoch": 2.11, "learning_rate": 1.4298596606510056e-05, "loss": 0.0816, "step": 90245 }, { "epoch": 2.11, "learning_rate": 1.4297716552064022e-05, "loss": 0.0508, "step": 90250 }, { "epoch": 2.11, "learning_rate": 1.4296836497617986e-05, "loss": 0.2061, "step": 90255 }, { "epoch": 2.11, "learning_rate": 1.4295956443171952e-05, "loss": 0.1363, "step": 90260 }, { "epoch": 2.11, "learning_rate": 1.4295076388725916e-05, "loss": 0.1528, "step": 90265 }, { "epoch": 2.11, "learning_rate": 1.429419633427988e-05, "loss": 0.2048, "step": 90270 }, { "epoch": 2.11, "learning_rate": 1.4293316279833846e-05, "loss": 0.2769, "step": 90275 }, { "epoch": 2.11, "learning_rate": 1.429243622538781e-05, "loss": 0.0128, "step": 90280 }, { "epoch": 2.11, "learning_rate": 1.4291556170941776e-05, "loss": 0.0723, "step": 90285 }, { "epoch": 2.11, "learning_rate": 1.4290676116495741e-05, "loss": 0.0292, "step": 90290 }, { "epoch": 2.11, "learning_rate": 1.4289796062049706e-05, "loss": 0.0038, "step": 90295 }, { "epoch": 2.11, "learning_rate": 1.4288916007603671e-05, "loss": 0.1234, "step": 90300 }, { "epoch": 2.11, "learning_rate": 1.4288035953157634e-05, "loss": 0.057, "step": 90305 }, { "epoch": 2.11, "learning_rate": 1.4287155898711601e-05, "loss": 0.1629, "step": 90310 }, { "epoch": 2.11, "learning_rate": 1.4286275844265564e-05, "loss": 0.1714, "step": 90315 }, { "epoch": 2.11, "learning_rate": 1.4285395789819531e-05, "loss": 0.1485, "step": 90320 }, { "epoch": 2.11, "learning_rate": 1.4284515735373494e-05, "loss": 0.2639, "step": 90325 }, { "epoch": 2.11, "learning_rate": 1.4283635680927461e-05, "loss": 0.052, "step": 90330 }, { "epoch": 2.11, "learning_rate": 1.4282755626481424e-05, "loss": 0.0203, "step": 90335 }, { "epoch": 2.11, "learning_rate": 1.4281875572035391e-05, "loss": 0.0603, "step": 90340 }, { "epoch": 2.11, "learning_rate": 1.4280995517589356e-05, "loss": 0.1021, "step": 90345 }, { "epoch": 2.11, "learning_rate": 1.428011546314332e-05, "loss": 0.0483, "step": 90350 }, { "epoch": 2.11, "learning_rate": 1.4279235408697286e-05, "loss": 0.1292, "step": 90355 }, { "epoch": 2.11, "learning_rate": 1.427835535425125e-05, "loss": 0.1348, "step": 90360 }, { "epoch": 2.11, "learning_rate": 1.4277475299805216e-05, "loss": 0.2471, "step": 90365 }, { "epoch": 2.11, "learning_rate": 1.427659524535918e-05, "loss": 0.2456, "step": 90370 }, { "epoch": 2.11, "learning_rate": 1.4275715190913146e-05, "loss": 0.4196, "step": 90375 }, { "epoch": 2.11, "learning_rate": 1.427483513646711e-05, "loss": 0.0845, "step": 90380 }, { "epoch": 2.11, "learning_rate": 1.4273955082021075e-05, "loss": 0.0109, "step": 90385 }, { "epoch": 2.11, "learning_rate": 1.427307502757504e-05, "loss": 0.073, "step": 90390 }, { "epoch": 2.11, "learning_rate": 1.4272194973129005e-05, "loss": 0.0098, "step": 90395 }, { "epoch": 2.11, "learning_rate": 1.427131491868297e-05, "loss": 0.1104, "step": 90400 }, { "epoch": 2.11, "learning_rate": 1.4270434864236935e-05, "loss": 0.1363, "step": 90405 }, { "epoch": 2.11, "learning_rate": 1.42695548097909e-05, "loss": 0.2085, "step": 90410 }, { "epoch": 2.11, "learning_rate": 1.4268674755344865e-05, "loss": 0.1348, "step": 90415 }, { "epoch": 2.11, "learning_rate": 1.4267794700898828e-05, "loss": 0.1327, "step": 90420 }, { "epoch": 2.11, "learning_rate": 1.4266914646452795e-05, "loss": 0.2449, "step": 90425 }, { "epoch": 2.11, "learning_rate": 1.4266034592006758e-05, "loss": 0.0009, "step": 90430 }, { "epoch": 2.11, "learning_rate": 1.4265154537560725e-05, "loss": 0.025, "step": 90435 }, { "epoch": 2.11, "learning_rate": 1.4264274483114688e-05, "loss": 0.0345, "step": 90440 }, { "epoch": 2.11, "learning_rate": 1.4263394428668655e-05, "loss": 0.0767, "step": 90445 }, { "epoch": 2.11, "learning_rate": 1.4262514374222618e-05, "loss": 0.0874, "step": 90450 }, { "epoch": 2.11, "learning_rate": 1.4261634319776583e-05, "loss": 0.1448, "step": 90455 }, { "epoch": 2.11, "learning_rate": 1.4260754265330548e-05, "loss": 0.1657, "step": 90460 }, { "epoch": 2.11, "learning_rate": 1.4259874210884513e-05, "loss": 0.1266, "step": 90465 }, { "epoch": 2.11, "learning_rate": 1.4258994156438478e-05, "loss": 0.0945, "step": 90470 }, { "epoch": 2.11, "learning_rate": 1.4258114101992443e-05, "loss": 0.1377, "step": 90475 }, { "epoch": 2.11, "learning_rate": 1.425723404754641e-05, "loss": 0.0889, "step": 90480 }, { "epoch": 2.11, "learning_rate": 1.4256353993100373e-05, "loss": 0.0099, "step": 90485 }, { "epoch": 2.11, "learning_rate": 1.425547393865434e-05, "loss": 0.1129, "step": 90490 }, { "epoch": 2.11, "learning_rate": 1.4254593884208303e-05, "loss": 0.0715, "step": 90495 }, { "epoch": 2.11, "learning_rate": 1.4253713829762268e-05, "loss": 0.0647, "step": 90500 }, { "epoch": 2.11, "learning_rate": 1.4252833775316233e-05, "loss": 0.1033, "step": 90505 }, { "epoch": 2.11, "learning_rate": 1.4251953720870198e-05, "loss": 0.1108, "step": 90510 }, { "epoch": 2.11, "learning_rate": 1.4251073666424163e-05, "loss": 0.0845, "step": 90515 }, { "epoch": 2.11, "learning_rate": 1.4250193611978128e-05, "loss": 0.2508, "step": 90520 }, { "epoch": 2.11, "learning_rate": 1.4249313557532093e-05, "loss": 0.1793, "step": 90525 }, { "epoch": 2.11, "learning_rate": 1.4248433503086058e-05, "loss": 0.0396, "step": 90530 }, { "epoch": 2.11, "learning_rate": 1.4247553448640022e-05, "loss": 0.0859, "step": 90535 }, { "epoch": 2.11, "learning_rate": 1.4246673394193988e-05, "loss": 0.0586, "step": 90540 }, { "epoch": 2.11, "learning_rate": 1.4245793339747952e-05, "loss": 0.024, "step": 90545 }, { "epoch": 2.11, "learning_rate": 1.4244913285301919e-05, "loss": 0.1089, "step": 90550 }, { "epoch": 2.11, "learning_rate": 1.4244033230855882e-05, "loss": 0.0878, "step": 90555 }, { "epoch": 2.11, "learning_rate": 1.4243153176409849e-05, "loss": 0.0699, "step": 90560 }, { "epoch": 2.11, "learning_rate": 1.4242273121963812e-05, "loss": 0.1907, "step": 90565 }, { "epoch": 2.11, "learning_rate": 1.4241393067517777e-05, "loss": 0.0778, "step": 90570 }, { "epoch": 2.11, "learning_rate": 1.4240513013071742e-05, "loss": 0.325, "step": 90575 }, { "epoch": 2.11, "learning_rate": 1.4239632958625707e-05, "loss": 0.1015, "step": 90580 }, { "epoch": 2.11, "learning_rate": 1.4238752904179672e-05, "loss": 0.0543, "step": 90585 }, { "epoch": 2.11, "learning_rate": 1.4237872849733637e-05, "loss": 0.0585, "step": 90590 }, { "epoch": 2.11, "learning_rate": 1.4236992795287602e-05, "loss": 0.0785, "step": 90595 }, { "epoch": 2.11, "learning_rate": 1.4236112740841567e-05, "loss": 0.0594, "step": 90600 }, { "epoch": 2.11, "learning_rate": 1.423523268639553e-05, "loss": 0.1685, "step": 90605 }, { "epoch": 2.11, "learning_rate": 1.4234352631949497e-05, "loss": 0.0959, "step": 90610 }, { "epoch": 2.11, "learning_rate": 1.4233472577503462e-05, "loss": 0.0979, "step": 90615 }, { "epoch": 2.11, "learning_rate": 1.4232592523057427e-05, "loss": 0.0912, "step": 90620 }, { "epoch": 2.11, "learning_rate": 1.4231712468611392e-05, "loss": 0.3138, "step": 90625 }, { "epoch": 2.11, "learning_rate": 1.4230832414165357e-05, "loss": 0.1254, "step": 90630 }, { "epoch": 2.11, "learning_rate": 1.4229952359719322e-05, "loss": 0.0475, "step": 90635 }, { "epoch": 2.11, "learning_rate": 1.4229072305273287e-05, "loss": 0.0444, "step": 90640 }, { "epoch": 2.11, "learning_rate": 1.4228192250827252e-05, "loss": 0.0044, "step": 90645 }, { "epoch": 2.11, "learning_rate": 1.4227312196381216e-05, "loss": 0.0158, "step": 90650 }, { "epoch": 2.12, "learning_rate": 1.4226432141935182e-05, "loss": 0.0206, "step": 90655 }, { "epoch": 2.12, "learning_rate": 1.4225552087489146e-05, "loss": 0.0913, "step": 90660 }, { "epoch": 2.12, "learning_rate": 1.4224672033043112e-05, "loss": 0.0833, "step": 90665 }, { "epoch": 2.12, "learning_rate": 1.4223791978597076e-05, "loss": 0.2521, "step": 90670 }, { "epoch": 2.12, "learning_rate": 1.4222911924151042e-05, "loss": 0.5824, "step": 90675 }, { "epoch": 2.12, "learning_rate": 1.4222031869705006e-05, "loss": 0.0613, "step": 90680 }, { "epoch": 2.12, "learning_rate": 1.422115181525897e-05, "loss": 0.0521, "step": 90685 }, { "epoch": 2.12, "learning_rate": 1.4220271760812936e-05, "loss": 0.0782, "step": 90690 }, { "epoch": 2.12, "learning_rate": 1.42193917063669e-05, "loss": 0.0162, "step": 90695 }, { "epoch": 2.12, "learning_rate": 1.4218511651920866e-05, "loss": 0.1348, "step": 90700 }, { "epoch": 2.12, "learning_rate": 1.421763159747483e-05, "loss": 0.0597, "step": 90705 }, { "epoch": 2.12, "learning_rate": 1.4216751543028796e-05, "loss": 0.1048, "step": 90710 }, { "epoch": 2.12, "learning_rate": 1.421587148858276e-05, "loss": 0.262, "step": 90715 }, { "epoch": 2.12, "learning_rate": 1.4214991434136724e-05, "loss": 0.0856, "step": 90720 }, { "epoch": 2.12, "learning_rate": 1.4214111379690691e-05, "loss": 0.15, "step": 90725 }, { "epoch": 2.12, "learning_rate": 1.4213231325244654e-05, "loss": 0.0486, "step": 90730 }, { "epoch": 2.12, "learning_rate": 1.4212351270798621e-05, "loss": 0.069, "step": 90735 }, { "epoch": 2.12, "learning_rate": 1.4211471216352584e-05, "loss": 0.0958, "step": 90740 }, { "epoch": 2.12, "learning_rate": 1.4210591161906551e-05, "loss": 0.1029, "step": 90745 }, { "epoch": 2.12, "learning_rate": 1.4209711107460516e-05, "loss": 0.0603, "step": 90750 }, { "epoch": 2.12, "learning_rate": 1.420883105301448e-05, "loss": 0.0896, "step": 90755 }, { "epoch": 2.12, "learning_rate": 1.4207950998568446e-05, "loss": 0.0913, "step": 90760 }, { "epoch": 2.12, "learning_rate": 1.420707094412241e-05, "loss": 0.1501, "step": 90765 }, { "epoch": 2.12, "learning_rate": 1.4206190889676376e-05, "loss": 0.1645, "step": 90770 }, { "epoch": 2.12, "learning_rate": 1.420531083523034e-05, "loss": 0.1534, "step": 90775 }, { "epoch": 2.12, "learning_rate": 1.4204430780784306e-05, "loss": 0.1726, "step": 90780 }, { "epoch": 2.12, "learning_rate": 1.420355072633827e-05, "loss": 0.0727, "step": 90785 }, { "epoch": 2.12, "learning_rate": 1.4202670671892236e-05, "loss": 0.0257, "step": 90790 }, { "epoch": 2.12, "learning_rate": 1.42017906174462e-05, "loss": 0.0497, "step": 90795 }, { "epoch": 2.12, "learning_rate": 1.4200910563000164e-05, "loss": 0.1284, "step": 90800 }, { "epoch": 2.12, "learning_rate": 1.420003050855413e-05, "loss": 0.1232, "step": 90805 }, { "epoch": 2.12, "learning_rate": 1.4199150454108095e-05, "loss": 0.2577, "step": 90810 }, { "epoch": 2.12, "learning_rate": 1.419827039966206e-05, "loss": 0.1724, "step": 90815 }, { "epoch": 2.12, "learning_rate": 1.4197390345216025e-05, "loss": 0.2381, "step": 90820 }, { "epoch": 2.12, "learning_rate": 1.419651029076999e-05, "loss": 0.1737, "step": 90825 }, { "epoch": 2.12, "learning_rate": 1.4195630236323955e-05, "loss": 0.0797, "step": 90830 }, { "epoch": 2.12, "learning_rate": 1.4194750181877918e-05, "loss": 0.0182, "step": 90835 }, { "epoch": 2.12, "learning_rate": 1.4193870127431885e-05, "loss": 0.0546, "step": 90840 }, { "epoch": 2.12, "learning_rate": 1.4192990072985848e-05, "loss": 0.029, "step": 90845 }, { "epoch": 2.12, "learning_rate": 1.4192110018539815e-05, "loss": 0.1826, "step": 90850 }, { "epoch": 2.12, "learning_rate": 1.4191229964093778e-05, "loss": 0.1213, "step": 90855 }, { "epoch": 2.12, "learning_rate": 1.4190349909647745e-05, "loss": 0.1663, "step": 90860 }, { "epoch": 2.12, "learning_rate": 1.4189469855201708e-05, "loss": 0.0159, "step": 90865 }, { "epoch": 2.12, "learning_rate": 1.4188589800755673e-05, "loss": 0.2645, "step": 90870 }, { "epoch": 2.12, "learning_rate": 1.4187709746309638e-05, "loss": 0.2864, "step": 90875 }, { "epoch": 2.12, "learning_rate": 1.4186829691863603e-05, "loss": 0.052, "step": 90880 }, { "epoch": 2.12, "learning_rate": 1.418594963741757e-05, "loss": 0.0288, "step": 90885 }, { "epoch": 2.12, "learning_rate": 1.4185069582971533e-05, "loss": 0.0445, "step": 90890 }, { "epoch": 2.12, "learning_rate": 1.41841895285255e-05, "loss": 0.0317, "step": 90895 }, { "epoch": 2.12, "learning_rate": 1.4183309474079463e-05, "loss": 0.0834, "step": 90900 }, { "epoch": 2.12, "learning_rate": 1.4182429419633428e-05, "loss": 0.0624, "step": 90905 }, { "epoch": 2.12, "learning_rate": 1.4181549365187393e-05, "loss": 0.0773, "step": 90910 }, { "epoch": 2.12, "learning_rate": 1.4180669310741358e-05, "loss": 0.1384, "step": 90915 }, { "epoch": 2.12, "learning_rate": 1.4179789256295323e-05, "loss": 0.2467, "step": 90920 }, { "epoch": 2.12, "learning_rate": 1.4178909201849288e-05, "loss": 0.2896, "step": 90925 }, { "epoch": 2.12, "learning_rate": 1.4178029147403253e-05, "loss": 0.0335, "step": 90930 }, { "epoch": 2.12, "learning_rate": 1.4177149092957218e-05, "loss": 0.0278, "step": 90935 }, { "epoch": 2.12, "learning_rate": 1.4176269038511183e-05, "loss": 0.0015, "step": 90940 }, { "epoch": 2.12, "learning_rate": 1.4175388984065148e-05, "loss": 0.0195, "step": 90945 }, { "epoch": 2.12, "learning_rate": 1.4174508929619112e-05, "loss": 0.0976, "step": 90950 }, { "epoch": 2.12, "learning_rate": 1.4173628875173078e-05, "loss": 0.0258, "step": 90955 }, { "epoch": 2.12, "learning_rate": 1.4172748820727042e-05, "loss": 0.1076, "step": 90960 }, { "epoch": 2.12, "learning_rate": 1.4171868766281008e-05, "loss": 0.2155, "step": 90965 }, { "epoch": 2.12, "learning_rate": 1.4170988711834972e-05, "loss": 0.3162, "step": 90970 }, { "epoch": 2.12, "learning_rate": 1.4170108657388939e-05, "loss": 0.3422, "step": 90975 }, { "epoch": 2.12, "learning_rate": 1.4169228602942902e-05, "loss": 0.0352, "step": 90980 }, { "epoch": 2.12, "learning_rate": 1.4168348548496867e-05, "loss": 0.013, "step": 90985 }, { "epoch": 2.12, "learning_rate": 1.4167468494050832e-05, "loss": 0.0115, "step": 90990 }, { "epoch": 2.12, "learning_rate": 1.4166588439604797e-05, "loss": 0.0685, "step": 90995 }, { "epoch": 2.12, "learning_rate": 1.4165708385158762e-05, "loss": 0.0858, "step": 91000 }, { "epoch": 2.12, "learning_rate": 1.4164828330712727e-05, "loss": 0.0172, "step": 91005 }, { "epoch": 2.12, "learning_rate": 1.4163948276266692e-05, "loss": 0.0053, "step": 91010 }, { "epoch": 2.12, "learning_rate": 1.4163068221820657e-05, "loss": 0.1776, "step": 91015 }, { "epoch": 2.12, "learning_rate": 1.4162188167374622e-05, "loss": 0.2142, "step": 91020 }, { "epoch": 2.12, "learning_rate": 1.4161308112928587e-05, "loss": 0.1642, "step": 91025 }, { "epoch": 2.12, "learning_rate": 1.4160428058482552e-05, "loss": 0.0714, "step": 91030 }, { "epoch": 2.12, "learning_rate": 1.4159548004036517e-05, "loss": 0.0431, "step": 91035 }, { "epoch": 2.12, "learning_rate": 1.4158667949590482e-05, "loss": 0.0209, "step": 91040 }, { "epoch": 2.12, "learning_rate": 1.4157787895144447e-05, "loss": 0.0527, "step": 91045 }, { "epoch": 2.12, "learning_rate": 1.4156907840698412e-05, "loss": 0.111, "step": 91050 }, { "epoch": 2.12, "learning_rate": 1.4156027786252375e-05, "loss": 0.0627, "step": 91055 }, { "epoch": 2.12, "learning_rate": 1.4155147731806342e-05, "loss": 0.0916, "step": 91060 }, { "epoch": 2.12, "learning_rate": 1.4154267677360305e-05, "loss": 0.1865, "step": 91065 }, { "epoch": 2.12, "learning_rate": 1.4153387622914272e-05, "loss": 0.3549, "step": 91070 }, { "epoch": 2.12, "learning_rate": 1.4152507568468236e-05, "loss": 0.1341, "step": 91075 }, { "epoch": 2.13, "learning_rate": 1.4151627514022202e-05, "loss": 0.0193, "step": 91080 }, { "epoch": 2.13, "learning_rate": 1.4150747459576166e-05, "loss": 0.1622, "step": 91085 }, { "epoch": 2.13, "learning_rate": 1.4149867405130132e-05, "loss": 0.0252, "step": 91090 }, { "epoch": 2.13, "learning_rate": 1.4148987350684096e-05, "loss": 0.0615, "step": 91095 }, { "epoch": 2.13, "learning_rate": 1.414810729623806e-05, "loss": 0.0588, "step": 91100 }, { "epoch": 2.13, "learning_rate": 1.4147227241792026e-05, "loss": 0.0855, "step": 91105 }, { "epoch": 2.13, "learning_rate": 1.414634718734599e-05, "loss": 0.0676, "step": 91110 }, { "epoch": 2.13, "learning_rate": 1.4145467132899956e-05, "loss": 0.1202, "step": 91115 }, { "epoch": 2.13, "learning_rate": 1.414458707845392e-05, "loss": 0.3165, "step": 91120 }, { "epoch": 2.13, "learning_rate": 1.4143707024007886e-05, "loss": 0.3093, "step": 91125 }, { "epoch": 2.13, "learning_rate": 1.414282696956185e-05, "loss": 0.1159, "step": 91130 }, { "epoch": 2.13, "learning_rate": 1.4141946915115814e-05, "loss": 0.022, "step": 91135 }, { "epoch": 2.13, "learning_rate": 1.414106686066978e-05, "loss": 0.0372, "step": 91140 }, { "epoch": 2.13, "learning_rate": 1.4140186806223744e-05, "loss": 0.0981, "step": 91145 }, { "epoch": 2.13, "learning_rate": 1.413930675177771e-05, "loss": 0.1793, "step": 91150 }, { "epoch": 2.13, "learning_rate": 1.4138426697331676e-05, "loss": 0.0948, "step": 91155 }, { "epoch": 2.13, "learning_rate": 1.4137546642885641e-05, "loss": 0.1353, "step": 91160 }, { "epoch": 2.13, "learning_rate": 1.4136666588439606e-05, "loss": 0.192, "step": 91165 }, { "epoch": 2.13, "learning_rate": 1.413578653399357e-05, "loss": 0.2145, "step": 91170 }, { "epoch": 2.13, "learning_rate": 1.4134906479547536e-05, "loss": 0.4316, "step": 91175 }, { "epoch": 2.13, "learning_rate": 1.41340264251015e-05, "loss": 0.098, "step": 91180 }, { "epoch": 2.13, "learning_rate": 1.4133146370655466e-05, "loss": 0.0275, "step": 91185 }, { "epoch": 2.13, "learning_rate": 1.413226631620943e-05, "loss": 0.0241, "step": 91190 }, { "epoch": 2.13, "learning_rate": 1.4131386261763396e-05, "loss": 0.0359, "step": 91195 }, { "epoch": 2.13, "learning_rate": 1.413050620731736e-05, "loss": 0.1154, "step": 91200 }, { "epoch": 2.13, "learning_rate": 1.4129626152871324e-05, "loss": 0.0657, "step": 91205 }, { "epoch": 2.13, "learning_rate": 1.412874609842529e-05, "loss": 0.1982, "step": 91210 }, { "epoch": 2.13, "learning_rate": 1.4127866043979254e-05, "loss": 0.245, "step": 91215 }, { "epoch": 2.13, "learning_rate": 1.412698598953322e-05, "loss": 0.1109, "step": 91220 }, { "epoch": 2.13, "learning_rate": 1.4126105935087184e-05, "loss": 0.1705, "step": 91225 }, { "epoch": 2.13, "learning_rate": 1.412522588064115e-05, "loss": 0.0425, "step": 91230 }, { "epoch": 2.13, "learning_rate": 1.4124345826195114e-05, "loss": 0.0047, "step": 91235 }, { "epoch": 2.13, "learning_rate": 1.412346577174908e-05, "loss": 0.064, "step": 91240 }, { "epoch": 2.13, "learning_rate": 1.4122585717303045e-05, "loss": 0.1257, "step": 91245 }, { "epoch": 2.13, "learning_rate": 1.4121705662857008e-05, "loss": 0.0281, "step": 91250 }, { "epoch": 2.13, "learning_rate": 1.4120825608410975e-05, "loss": 0.2415, "step": 91255 }, { "epoch": 2.13, "learning_rate": 1.4119945553964938e-05, "loss": 0.1173, "step": 91260 }, { "epoch": 2.13, "learning_rate": 1.4119065499518905e-05, "loss": 0.1179, "step": 91265 }, { "epoch": 2.13, "learning_rate": 1.4118185445072868e-05, "loss": 0.2828, "step": 91270 }, { "epoch": 2.13, "learning_rate": 1.4117305390626835e-05, "loss": 0.3063, "step": 91275 }, { "epoch": 2.13, "learning_rate": 1.4116425336180798e-05, "loss": 0.0806, "step": 91280 }, { "epoch": 2.13, "learning_rate": 1.4115545281734763e-05, "loss": 0.0154, "step": 91285 }, { "epoch": 2.13, "learning_rate": 1.411466522728873e-05, "loss": 0.2139, "step": 91290 }, { "epoch": 2.13, "learning_rate": 1.4113785172842693e-05, "loss": 0.0731, "step": 91295 }, { "epoch": 2.13, "learning_rate": 1.411290511839666e-05, "loss": 0.1324, "step": 91300 }, { "epoch": 2.13, "learning_rate": 1.4112025063950623e-05, "loss": 0.0739, "step": 91305 }, { "epoch": 2.13, "learning_rate": 1.411114500950459e-05, "loss": 0.0893, "step": 91310 }, { "epoch": 2.13, "learning_rate": 1.4110264955058553e-05, "loss": 0.1739, "step": 91315 }, { "epoch": 2.13, "learning_rate": 1.4109384900612518e-05, "loss": 0.3598, "step": 91320 }, { "epoch": 2.13, "learning_rate": 1.4108504846166483e-05, "loss": 0.1712, "step": 91325 }, { "epoch": 2.13, "learning_rate": 1.4107624791720448e-05, "loss": 0.0706, "step": 91330 }, { "epoch": 2.13, "learning_rate": 1.4106744737274413e-05, "loss": 0.0047, "step": 91335 }, { "epoch": 2.13, "learning_rate": 1.4105864682828378e-05, "loss": 0.0902, "step": 91340 }, { "epoch": 2.13, "learning_rate": 1.4104984628382343e-05, "loss": 0.062, "step": 91345 }, { "epoch": 2.13, "learning_rate": 1.4104104573936308e-05, "loss": 0.0687, "step": 91350 }, { "epoch": 2.13, "learning_rate": 1.4103224519490272e-05, "loss": 0.0538, "step": 91355 }, { "epoch": 2.13, "learning_rate": 1.4102344465044238e-05, "loss": 0.2444, "step": 91360 }, { "epoch": 2.13, "learning_rate": 1.4101464410598202e-05, "loss": 0.2371, "step": 91365 }, { "epoch": 2.13, "learning_rate": 1.4100584356152168e-05, "loss": 0.395, "step": 91370 }, { "epoch": 2.13, "learning_rate": 1.4099704301706132e-05, "loss": 0.2751, "step": 91375 }, { "epoch": 2.13, "learning_rate": 1.4098824247260098e-05, "loss": 0.0606, "step": 91380 }, { "epoch": 2.13, "learning_rate": 1.4097944192814062e-05, "loss": 0.0282, "step": 91385 }, { "epoch": 2.13, "learning_rate": 1.4097064138368028e-05, "loss": 0.0295, "step": 91390 }, { "epoch": 2.13, "learning_rate": 1.4096184083921992e-05, "loss": 0.0563, "step": 91395 }, { "epoch": 2.13, "learning_rate": 1.4095304029475957e-05, "loss": 0.0363, "step": 91400 }, { "epoch": 2.13, "learning_rate": 1.4094423975029922e-05, "loss": 0.0762, "step": 91405 }, { "epoch": 2.13, "learning_rate": 1.4093543920583887e-05, "loss": 0.0756, "step": 91410 }, { "epoch": 2.13, "learning_rate": 1.4092663866137852e-05, "loss": 0.0888, "step": 91415 }, { "epoch": 2.13, "learning_rate": 1.4091783811691817e-05, "loss": 0.2872, "step": 91420 }, { "epoch": 2.13, "learning_rate": 1.4090903757245782e-05, "loss": 0.2292, "step": 91425 }, { "epoch": 2.13, "learning_rate": 1.4090023702799747e-05, "loss": 0.0556, "step": 91430 }, { "epoch": 2.13, "learning_rate": 1.4089143648353712e-05, "loss": 0.0974, "step": 91435 }, { "epoch": 2.13, "learning_rate": 1.4088263593907677e-05, "loss": 0.083, "step": 91440 }, { "epoch": 2.13, "learning_rate": 1.4087383539461642e-05, "loss": 0.119, "step": 91445 }, { "epoch": 2.13, "learning_rate": 1.4086503485015607e-05, "loss": 0.1173, "step": 91450 }, { "epoch": 2.13, "learning_rate": 1.4085623430569572e-05, "loss": 0.0937, "step": 91455 }, { "epoch": 2.13, "learning_rate": 1.4084743376123537e-05, "loss": 0.0645, "step": 91460 }, { "epoch": 2.13, "learning_rate": 1.4083863321677502e-05, "loss": 0.0613, "step": 91465 }, { "epoch": 2.13, "learning_rate": 1.4082983267231465e-05, "loss": 0.2097, "step": 91470 }, { "epoch": 2.13, "learning_rate": 1.4082103212785432e-05, "loss": 0.261, "step": 91475 }, { "epoch": 2.13, "learning_rate": 1.4081223158339395e-05, "loss": 0.0219, "step": 91480 }, { "epoch": 2.13, "learning_rate": 1.4080343103893362e-05, "loss": 0.01, "step": 91485 }, { "epoch": 2.13, "learning_rate": 1.4079463049447325e-05, "loss": 0.0249, "step": 91490 }, { "epoch": 2.13, "learning_rate": 1.4078582995001292e-05, "loss": 0.0337, "step": 91495 }, { "epoch": 2.13, "learning_rate": 1.4077702940555255e-05, "loss": 0.0737, "step": 91500 }, { "epoch": 2.13, "learning_rate": 1.407682288610922e-05, "loss": 0.1337, "step": 91505 }, { "epoch": 2.14, "learning_rate": 1.4075942831663186e-05, "loss": 0.1416, "step": 91510 }, { "epoch": 2.14, "learning_rate": 1.407506277721715e-05, "loss": 0.1256, "step": 91515 }, { "epoch": 2.14, "learning_rate": 1.4074182722771116e-05, "loss": 0.1566, "step": 91520 }, { "epoch": 2.14, "learning_rate": 1.407330266832508e-05, "loss": 0.3277, "step": 91525 }, { "epoch": 2.14, "learning_rate": 1.4072422613879046e-05, "loss": 0.0385, "step": 91530 }, { "epoch": 2.14, "learning_rate": 1.407154255943301e-05, "loss": 0.0852, "step": 91535 }, { "epoch": 2.14, "learning_rate": 1.4070662504986976e-05, "loss": 0.0884, "step": 91540 }, { "epoch": 2.14, "learning_rate": 1.406978245054094e-05, "loss": 0.0973, "step": 91545 }, { "epoch": 2.14, "learning_rate": 1.4068902396094904e-05, "loss": 0.0521, "step": 91550 }, { "epoch": 2.14, "learning_rate": 1.406802234164887e-05, "loss": 0.0777, "step": 91555 }, { "epoch": 2.14, "learning_rate": 1.4067142287202836e-05, "loss": 0.1525, "step": 91560 }, { "epoch": 2.14, "learning_rate": 1.40662622327568e-05, "loss": 0.0509, "step": 91565 }, { "epoch": 2.14, "learning_rate": 1.4065382178310766e-05, "loss": 0.155, "step": 91570 }, { "epoch": 2.14, "learning_rate": 1.406450212386473e-05, "loss": 0.1613, "step": 91575 }, { "epoch": 2.14, "learning_rate": 1.4063622069418696e-05, "loss": 0.0386, "step": 91580 }, { "epoch": 2.14, "learning_rate": 1.4062742014972659e-05, "loss": 0.0564, "step": 91585 }, { "epoch": 2.14, "learning_rate": 1.4061861960526626e-05, "loss": 0.0889, "step": 91590 }, { "epoch": 2.14, "learning_rate": 1.406098190608059e-05, "loss": 0.0689, "step": 91595 }, { "epoch": 2.14, "learning_rate": 1.4060101851634556e-05, "loss": 0.0695, "step": 91600 }, { "epoch": 2.14, "learning_rate": 1.405922179718852e-05, "loss": 0.0888, "step": 91605 }, { "epoch": 2.14, "learning_rate": 1.4058341742742486e-05, "loss": 0.1362, "step": 91610 }, { "epoch": 2.14, "learning_rate": 1.405746168829645e-05, "loss": 0.0037, "step": 91615 }, { "epoch": 2.14, "learning_rate": 1.4056581633850414e-05, "loss": 0.0921, "step": 91620 }, { "epoch": 2.14, "learning_rate": 1.405570157940438e-05, "loss": 0.3966, "step": 91625 }, { "epoch": 2.14, "learning_rate": 1.4054821524958344e-05, "loss": 0.0421, "step": 91630 }, { "epoch": 2.14, "learning_rate": 1.405394147051231e-05, "loss": 0.0771, "step": 91635 }, { "epoch": 2.14, "learning_rate": 1.4053061416066274e-05, "loss": 0.0386, "step": 91640 }, { "epoch": 2.14, "learning_rate": 1.405218136162024e-05, "loss": 0.0346, "step": 91645 }, { "epoch": 2.14, "learning_rate": 1.4051301307174204e-05, "loss": 0.0715, "step": 91650 }, { "epoch": 2.14, "learning_rate": 1.4050421252728168e-05, "loss": 0.0433, "step": 91655 }, { "epoch": 2.14, "learning_rate": 1.4049541198282134e-05, "loss": 0.1567, "step": 91660 }, { "epoch": 2.14, "learning_rate": 1.4048661143836098e-05, "loss": 0.1595, "step": 91665 }, { "epoch": 2.14, "learning_rate": 1.4047781089390064e-05, "loss": 0.325, "step": 91670 }, { "epoch": 2.14, "learning_rate": 1.4046901034944028e-05, "loss": 0.4146, "step": 91675 }, { "epoch": 2.14, "learning_rate": 1.4046020980497995e-05, "loss": 0.1701, "step": 91680 }, { "epoch": 2.14, "learning_rate": 1.4045140926051958e-05, "loss": 0.0346, "step": 91685 }, { "epoch": 2.14, "learning_rate": 1.4044260871605925e-05, "loss": 0.0835, "step": 91690 }, { "epoch": 2.14, "learning_rate": 1.4043380817159888e-05, "loss": 0.0384, "step": 91695 }, { "epoch": 2.14, "learning_rate": 1.4042500762713853e-05, "loss": 0.0296, "step": 91700 }, { "epoch": 2.14, "learning_rate": 1.404162070826782e-05, "loss": 0.0837, "step": 91705 }, { "epoch": 2.14, "learning_rate": 1.4040740653821783e-05, "loss": 0.1512, "step": 91710 }, { "epoch": 2.14, "learning_rate": 1.403986059937575e-05, "loss": 0.2094, "step": 91715 }, { "epoch": 2.14, "learning_rate": 1.4038980544929713e-05, "loss": 0.2357, "step": 91720 }, { "epoch": 2.14, "learning_rate": 1.403810049048368e-05, "loss": 0.2797, "step": 91725 }, { "epoch": 2.14, "learning_rate": 1.4037220436037643e-05, "loss": 0.0624, "step": 91730 }, { "epoch": 2.14, "learning_rate": 1.4036340381591608e-05, "loss": 0.0417, "step": 91735 }, { "epoch": 2.14, "learning_rate": 1.4035460327145573e-05, "loss": 0.0295, "step": 91740 }, { "epoch": 2.14, "learning_rate": 1.4034580272699538e-05, "loss": 0.0788, "step": 91745 }, { "epoch": 2.14, "learning_rate": 1.4033700218253503e-05, "loss": 0.1097, "step": 91750 }, { "epoch": 2.14, "learning_rate": 1.4032820163807468e-05, "loss": 0.0252, "step": 91755 }, { "epoch": 2.14, "learning_rate": 1.4031940109361433e-05, "loss": 0.1432, "step": 91760 }, { "epoch": 2.14, "learning_rate": 1.4031060054915398e-05, "loss": 0.048, "step": 91765 }, { "epoch": 2.14, "learning_rate": 1.4030180000469362e-05, "loss": 0.3445, "step": 91770 }, { "epoch": 2.14, "learning_rate": 1.4029299946023328e-05, "loss": 0.3799, "step": 91775 }, { "epoch": 2.14, "learning_rate": 1.4028419891577292e-05, "loss": 0.0122, "step": 91780 }, { "epoch": 2.14, "learning_rate": 1.4027539837131258e-05, "loss": 0.0133, "step": 91785 }, { "epoch": 2.14, "learning_rate": 1.4026659782685222e-05, "loss": 0.0546, "step": 91790 }, { "epoch": 2.14, "learning_rate": 1.4025779728239188e-05, "loss": 0.0933, "step": 91795 }, { "epoch": 2.14, "learning_rate": 1.4024899673793152e-05, "loss": 0.0834, "step": 91800 }, { "epoch": 2.14, "learning_rate": 1.4024019619347117e-05, "loss": 0.0601, "step": 91805 }, { "epoch": 2.14, "learning_rate": 1.4023139564901082e-05, "loss": 0.1314, "step": 91810 }, { "epoch": 2.14, "learning_rate": 1.4022259510455047e-05, "loss": 0.1964, "step": 91815 }, { "epoch": 2.14, "learning_rate": 1.4021379456009012e-05, "loss": 0.2752, "step": 91820 }, { "epoch": 2.14, "learning_rate": 1.4020499401562977e-05, "loss": 0.2131, "step": 91825 }, { "epoch": 2.14, "learning_rate": 1.4019619347116942e-05, "loss": 0.0665, "step": 91830 }, { "epoch": 2.14, "learning_rate": 1.4018739292670907e-05, "loss": 0.1061, "step": 91835 }, { "epoch": 2.14, "learning_rate": 1.4017859238224874e-05, "loss": 0.0024, "step": 91840 }, { "epoch": 2.14, "learning_rate": 1.4016979183778837e-05, "loss": 0.0747, "step": 91845 }, { "epoch": 2.14, "learning_rate": 1.4016099129332802e-05, "loss": 0.1144, "step": 91850 }, { "epoch": 2.14, "learning_rate": 1.4015219074886767e-05, "loss": 0.0668, "step": 91855 }, { "epoch": 2.14, "learning_rate": 1.4014339020440732e-05, "loss": 0.0724, "step": 91860 }, { "epoch": 2.14, "learning_rate": 1.4013458965994697e-05, "loss": 0.1559, "step": 91865 }, { "epoch": 2.14, "learning_rate": 1.4012578911548662e-05, "loss": 0.3785, "step": 91870 }, { "epoch": 2.14, "learning_rate": 1.4011698857102627e-05, "loss": 0.475, "step": 91875 }, { "epoch": 2.14, "learning_rate": 1.4010818802656592e-05, "loss": 0.0512, "step": 91880 }, { "epoch": 2.14, "learning_rate": 1.4009938748210555e-05, "loss": 0.0033, "step": 91885 }, { "epoch": 2.14, "learning_rate": 1.4009058693764522e-05, "loss": 0.018, "step": 91890 }, { "epoch": 2.14, "learning_rate": 1.4008178639318485e-05, "loss": 0.003, "step": 91895 }, { "epoch": 2.14, "learning_rate": 1.4007298584872452e-05, "loss": 0.0888, "step": 91900 }, { "epoch": 2.14, "learning_rate": 1.4006418530426415e-05, "loss": 0.1566, "step": 91905 }, { "epoch": 2.14, "learning_rate": 1.4005538475980382e-05, "loss": 0.0861, "step": 91910 }, { "epoch": 2.14, "learning_rate": 1.4004658421534345e-05, "loss": 0.2098, "step": 91915 }, { "epoch": 2.14, "learning_rate": 1.400377836708831e-05, "loss": 0.1315, "step": 91920 }, { "epoch": 2.14, "learning_rate": 1.4002898312642275e-05, "loss": 0.1931, "step": 91925 }, { "epoch": 2.14, "learning_rate": 1.400201825819624e-05, "loss": 0.0448, "step": 91930 }, { "epoch": 2.14, "learning_rate": 1.4001138203750206e-05, "loss": 0.0141, "step": 91935 }, { "epoch": 2.15, "learning_rate": 1.400025814930417e-05, "loss": 0.0472, "step": 91940 }, { "epoch": 2.15, "learning_rate": 1.3999378094858136e-05, "loss": 0.0735, "step": 91945 }, { "epoch": 2.15, "learning_rate": 1.39984980404121e-05, "loss": 0.0804, "step": 91950 }, { "epoch": 2.15, "learning_rate": 1.3997617985966064e-05, "loss": 0.0409, "step": 91955 }, { "epoch": 2.15, "learning_rate": 1.399673793152003e-05, "loss": 0.11, "step": 91960 }, { "epoch": 2.15, "learning_rate": 1.3995857877073994e-05, "loss": 0.2966, "step": 91965 }, { "epoch": 2.15, "learning_rate": 1.399497782262796e-05, "loss": 0.1389, "step": 91970 }, { "epoch": 2.15, "learning_rate": 1.3994097768181926e-05, "loss": 0.3141, "step": 91975 }, { "epoch": 2.15, "learning_rate": 1.399321771373589e-05, "loss": 0.1186, "step": 91980 }, { "epoch": 2.15, "learning_rate": 1.3992337659289856e-05, "loss": 0.077, "step": 91985 }, { "epoch": 2.15, "learning_rate": 1.3991457604843819e-05, "loss": 0.0444, "step": 91990 }, { "epoch": 2.15, "learning_rate": 1.3990577550397786e-05, "loss": 0.0445, "step": 91995 }, { "epoch": 2.15, "learning_rate": 1.3989697495951749e-05, "loss": 0.1211, "step": 92000 }, { "epoch": 2.15, "learning_rate": 1.3988817441505716e-05, "loss": 0.1415, "step": 92005 }, { "epoch": 2.15, "learning_rate": 1.3987937387059679e-05, "loss": 0.0743, "step": 92010 }, { "epoch": 2.15, "learning_rate": 1.3987057332613646e-05, "loss": 0.3732, "step": 92015 }, { "epoch": 2.15, "learning_rate": 1.398617727816761e-05, "loss": 0.3529, "step": 92020 }, { "epoch": 2.15, "learning_rate": 1.3985297223721576e-05, "loss": 0.2729, "step": 92025 }, { "epoch": 2.15, "learning_rate": 1.398441716927554e-05, "loss": 0.0244, "step": 92030 }, { "epoch": 2.15, "learning_rate": 1.3983537114829504e-05, "loss": 0.0473, "step": 92035 }, { "epoch": 2.15, "learning_rate": 1.398265706038347e-05, "loss": 0.0102, "step": 92040 }, { "epoch": 2.15, "learning_rate": 1.3981777005937434e-05, "loss": 0.1107, "step": 92045 }, { "epoch": 2.15, "learning_rate": 1.39808969514914e-05, "loss": 0.018, "step": 92050 }, { "epoch": 2.15, "learning_rate": 1.3980016897045364e-05, "loss": 0.1014, "step": 92055 }, { "epoch": 2.15, "learning_rate": 1.397913684259933e-05, "loss": 0.0939, "step": 92060 }, { "epoch": 2.15, "learning_rate": 1.3978256788153294e-05, "loss": 0.2336, "step": 92065 }, { "epoch": 2.15, "learning_rate": 1.3977376733707258e-05, "loss": 0.2888, "step": 92070 }, { "epoch": 2.15, "learning_rate": 1.3976496679261224e-05, "loss": 0.1485, "step": 92075 }, { "epoch": 2.15, "learning_rate": 1.3975616624815188e-05, "loss": 0.1362, "step": 92080 }, { "epoch": 2.15, "learning_rate": 1.3974736570369154e-05, "loss": 0.0116, "step": 92085 }, { "epoch": 2.15, "learning_rate": 1.3973856515923118e-05, "loss": 0.008, "step": 92090 }, { "epoch": 2.15, "learning_rate": 1.3972976461477084e-05, "loss": 0.0576, "step": 92095 }, { "epoch": 2.15, "learning_rate": 1.3972096407031048e-05, "loss": 0.0024, "step": 92100 }, { "epoch": 2.15, "learning_rate": 1.3971216352585013e-05, "loss": 0.1418, "step": 92105 }, { "epoch": 2.15, "learning_rate": 1.397033629813898e-05, "loss": 0.139, "step": 92110 }, { "epoch": 2.15, "learning_rate": 1.3969456243692943e-05, "loss": 0.0862, "step": 92115 }, { "epoch": 2.15, "learning_rate": 1.396857618924691e-05, "loss": 0.2172, "step": 92120 }, { "epoch": 2.15, "learning_rate": 1.3967696134800873e-05, "loss": 0.4777, "step": 92125 }, { "epoch": 2.15, "learning_rate": 1.396681608035484e-05, "loss": 0.0617, "step": 92130 }, { "epoch": 2.15, "learning_rate": 1.3965936025908803e-05, "loss": 0.1026, "step": 92135 }, { "epoch": 2.15, "learning_rate": 1.3965055971462768e-05, "loss": 0.0787, "step": 92140 }, { "epoch": 2.15, "learning_rate": 1.3964175917016733e-05, "loss": 0.0588, "step": 92145 }, { "epoch": 2.15, "learning_rate": 1.3963295862570698e-05, "loss": 0.0331, "step": 92150 }, { "epoch": 2.15, "learning_rate": 1.3962415808124663e-05, "loss": 0.1207, "step": 92155 }, { "epoch": 2.15, "learning_rate": 1.3961535753678628e-05, "loss": 0.1693, "step": 92160 }, { "epoch": 2.15, "learning_rate": 1.3960655699232593e-05, "loss": 0.0296, "step": 92165 }, { "epoch": 2.15, "learning_rate": 1.3959775644786558e-05, "loss": 0.2204, "step": 92170 }, { "epoch": 2.15, "learning_rate": 1.3958895590340523e-05, "loss": 0.4087, "step": 92175 }, { "epoch": 2.15, "learning_rate": 1.3958015535894488e-05, "loss": 0.0824, "step": 92180 }, { "epoch": 2.15, "learning_rate": 1.3957135481448451e-05, "loss": 0.0918, "step": 92185 }, { "epoch": 2.15, "learning_rate": 1.3956255427002418e-05, "loss": 0.032, "step": 92190 }, { "epoch": 2.15, "learning_rate": 1.3955375372556381e-05, "loss": 0.0737, "step": 92195 }, { "epoch": 2.15, "learning_rate": 1.3954495318110348e-05, "loss": 0.0152, "step": 92200 }, { "epoch": 2.15, "learning_rate": 1.3953615263664312e-05, "loss": 0.0451, "step": 92205 }, { "epoch": 2.15, "learning_rate": 1.3952735209218278e-05, "loss": 0.04, "step": 92210 }, { "epoch": 2.15, "learning_rate": 1.3951855154772242e-05, "loss": 0.0946, "step": 92215 }, { "epoch": 2.15, "learning_rate": 1.3950975100326207e-05, "loss": 0.185, "step": 92220 }, { "epoch": 2.15, "learning_rate": 1.3950095045880172e-05, "loss": 0.3471, "step": 92225 }, { "epoch": 2.15, "learning_rate": 1.3949214991434137e-05, "loss": 0.0982, "step": 92230 }, { "epoch": 2.15, "learning_rate": 1.3948334936988102e-05, "loss": 0.0004, "step": 92235 }, { "epoch": 2.15, "learning_rate": 1.3947454882542067e-05, "loss": 0.0202, "step": 92240 }, { "epoch": 2.15, "learning_rate": 1.3946574828096033e-05, "loss": 0.0868, "step": 92245 }, { "epoch": 2.15, "learning_rate": 1.3945694773649997e-05, "loss": 0.086, "step": 92250 }, { "epoch": 2.15, "learning_rate": 1.3944814719203962e-05, "loss": 0.0181, "step": 92255 }, { "epoch": 2.15, "learning_rate": 1.3943934664757927e-05, "loss": 0.1192, "step": 92260 }, { "epoch": 2.15, "learning_rate": 1.3943054610311892e-05, "loss": 0.2409, "step": 92265 }, { "epoch": 2.15, "learning_rate": 1.3942174555865857e-05, "loss": 0.1867, "step": 92270 }, { "epoch": 2.15, "learning_rate": 1.3941294501419822e-05, "loss": 0.245, "step": 92275 }, { "epoch": 2.15, "learning_rate": 1.3940414446973787e-05, "loss": 0.0428, "step": 92280 }, { "epoch": 2.15, "learning_rate": 1.3939534392527752e-05, "loss": 0.0486, "step": 92285 }, { "epoch": 2.15, "learning_rate": 1.3938654338081715e-05, "loss": 0.0249, "step": 92290 }, { "epoch": 2.15, "learning_rate": 1.3937774283635682e-05, "loss": 0.0029, "step": 92295 }, { "epoch": 2.15, "learning_rate": 1.3936894229189645e-05, "loss": 0.1042, "step": 92300 }, { "epoch": 2.15, "learning_rate": 1.3936014174743612e-05, "loss": 0.0514, "step": 92305 }, { "epoch": 2.15, "learning_rate": 1.3935134120297575e-05, "loss": 0.1778, "step": 92310 }, { "epoch": 2.15, "learning_rate": 1.3934254065851542e-05, "loss": 0.3108, "step": 92315 }, { "epoch": 2.15, "learning_rate": 1.3933374011405505e-05, "loss": 0.312, "step": 92320 }, { "epoch": 2.15, "learning_rate": 1.3932493956959472e-05, "loss": 0.1447, "step": 92325 }, { "epoch": 2.15, "learning_rate": 1.3931613902513435e-05, "loss": 0.0834, "step": 92330 }, { "epoch": 2.15, "learning_rate": 1.39307338480674e-05, "loss": 0.0479, "step": 92335 }, { "epoch": 2.15, "learning_rate": 1.3929853793621365e-05, "loss": 0.0434, "step": 92340 }, { "epoch": 2.15, "learning_rate": 1.392897373917533e-05, "loss": 0.0351, "step": 92345 }, { "epoch": 2.15, "learning_rate": 1.3928093684729295e-05, "loss": 0.0876, "step": 92350 }, { "epoch": 2.15, "learning_rate": 1.392721363028326e-05, "loss": 0.0489, "step": 92355 }, { "epoch": 2.15, "learning_rate": 1.3926333575837225e-05, "loss": 0.1422, "step": 92360 }, { "epoch": 2.15, "learning_rate": 1.392545352139119e-05, "loss": 0.0828, "step": 92365 }, { "epoch": 2.16, "learning_rate": 1.3924573466945154e-05, "loss": 0.271, "step": 92370 }, { "epoch": 2.16, "learning_rate": 1.392369341249912e-05, "loss": 0.439, "step": 92375 }, { "epoch": 2.16, "learning_rate": 1.3922813358053086e-05, "loss": 0.0358, "step": 92380 }, { "epoch": 2.16, "learning_rate": 1.392193330360705e-05, "loss": 0.0073, "step": 92385 }, { "epoch": 2.16, "learning_rate": 1.3921053249161016e-05, "loss": 0.0216, "step": 92390 }, { "epoch": 2.16, "learning_rate": 1.392017319471498e-05, "loss": 0.0348, "step": 92395 }, { "epoch": 2.16, "learning_rate": 1.3919293140268946e-05, "loss": 0.0427, "step": 92400 }, { "epoch": 2.16, "learning_rate": 1.3918413085822909e-05, "loss": 0.2245, "step": 92405 }, { "epoch": 2.16, "learning_rate": 1.3917533031376876e-05, "loss": 0.142, "step": 92410 }, { "epoch": 2.16, "learning_rate": 1.3916652976930839e-05, "loss": 0.0685, "step": 92415 }, { "epoch": 2.16, "learning_rate": 1.3915772922484806e-05, "loss": 0.2907, "step": 92420 }, { "epoch": 2.16, "learning_rate": 1.3914892868038769e-05, "loss": 0.2025, "step": 92425 }, { "epoch": 2.16, "learning_rate": 1.3914012813592736e-05, "loss": 0.0594, "step": 92430 }, { "epoch": 2.16, "learning_rate": 1.3913132759146699e-05, "loss": 0.0103, "step": 92435 }, { "epoch": 2.16, "learning_rate": 1.3912252704700664e-05, "loss": 0.0547, "step": 92440 }, { "epoch": 2.16, "learning_rate": 1.3911372650254629e-05, "loss": 0.0585, "step": 92445 }, { "epoch": 2.16, "learning_rate": 1.3910492595808594e-05, "loss": 0.113, "step": 92450 }, { "epoch": 2.16, "learning_rate": 1.390961254136256e-05, "loss": 0.1272, "step": 92455 }, { "epoch": 2.16, "learning_rate": 1.3908732486916524e-05, "loss": 0.1807, "step": 92460 }, { "epoch": 2.16, "learning_rate": 1.390785243247049e-05, "loss": 0.1373, "step": 92465 }, { "epoch": 2.16, "learning_rate": 1.3906972378024454e-05, "loss": 0.1711, "step": 92470 }, { "epoch": 2.16, "learning_rate": 1.390609232357842e-05, "loss": 0.2144, "step": 92475 }, { "epoch": 2.16, "learning_rate": 1.3905212269132384e-05, "loss": 0.0531, "step": 92480 }, { "epoch": 2.16, "learning_rate": 1.3904332214686348e-05, "loss": 0.0472, "step": 92485 }, { "epoch": 2.16, "learning_rate": 1.3903452160240314e-05, "loss": 0.0475, "step": 92490 }, { "epoch": 2.16, "learning_rate": 1.3902572105794278e-05, "loss": 0.0466, "step": 92495 }, { "epoch": 2.16, "learning_rate": 1.3901692051348244e-05, "loss": 0.0748, "step": 92500 }, { "epoch": 2.16, "learning_rate": 1.3900811996902208e-05, "loss": 0.0152, "step": 92505 }, { "epoch": 2.16, "learning_rate": 1.3899931942456174e-05, "loss": 0.0494, "step": 92510 }, { "epoch": 2.16, "learning_rate": 1.389905188801014e-05, "loss": 0.1649, "step": 92515 }, { "epoch": 2.16, "learning_rate": 1.3898171833564103e-05, "loss": 0.378, "step": 92520 }, { "epoch": 2.16, "learning_rate": 1.389729177911807e-05, "loss": 0.354, "step": 92525 }, { "epoch": 2.16, "learning_rate": 1.3896411724672033e-05, "loss": 0.0913, "step": 92530 }, { "epoch": 2.16, "learning_rate": 1.3895531670226e-05, "loss": 0.1049, "step": 92535 }, { "epoch": 2.16, "learning_rate": 1.3894651615779963e-05, "loss": 0.0395, "step": 92540 }, { "epoch": 2.16, "learning_rate": 1.389377156133393e-05, "loss": 0.1372, "step": 92545 }, { "epoch": 2.16, "learning_rate": 1.3892891506887893e-05, "loss": 0.0165, "step": 92550 }, { "epoch": 2.16, "learning_rate": 1.3892011452441858e-05, "loss": 0.0561, "step": 92555 }, { "epoch": 2.16, "learning_rate": 1.3891131397995823e-05, "loss": 0.0664, "step": 92560 }, { "epoch": 2.16, "learning_rate": 1.3890251343549788e-05, "loss": 0.2018, "step": 92565 }, { "epoch": 2.16, "learning_rate": 1.3889371289103753e-05, "loss": 0.1705, "step": 92570 }, { "epoch": 2.16, "learning_rate": 1.3888491234657718e-05, "loss": 0.2147, "step": 92575 }, { "epoch": 2.16, "learning_rate": 1.3887611180211683e-05, "loss": 0.0485, "step": 92580 }, { "epoch": 2.16, "learning_rate": 1.3886731125765648e-05, "loss": 0.0331, "step": 92585 }, { "epoch": 2.16, "learning_rate": 1.3885851071319611e-05, "loss": 0.1147, "step": 92590 }, { "epoch": 2.16, "learning_rate": 1.3884971016873578e-05, "loss": 0.0731, "step": 92595 }, { "epoch": 2.16, "learning_rate": 1.3884090962427541e-05, "loss": 0.0601, "step": 92600 }, { "epoch": 2.16, "learning_rate": 1.3883210907981508e-05, "loss": 0.2593, "step": 92605 }, { "epoch": 2.16, "learning_rate": 1.3882330853535471e-05, "loss": 0.0801, "step": 92610 }, { "epoch": 2.16, "learning_rate": 1.3881450799089438e-05, "loss": 0.0907, "step": 92615 }, { "epoch": 2.16, "learning_rate": 1.3880570744643401e-05, "loss": 0.215, "step": 92620 }, { "epoch": 2.16, "learning_rate": 1.3879690690197368e-05, "loss": 0.2281, "step": 92625 }, { "epoch": 2.16, "learning_rate": 1.3878810635751331e-05, "loss": 0.1084, "step": 92630 }, { "epoch": 2.16, "learning_rate": 1.3877930581305297e-05, "loss": 0.0057, "step": 92635 }, { "epoch": 2.16, "learning_rate": 1.3877050526859262e-05, "loss": 0.0305, "step": 92640 }, { "epoch": 2.16, "learning_rate": 1.3876170472413227e-05, "loss": 0.0117, "step": 92645 }, { "epoch": 2.16, "learning_rate": 1.3875290417967193e-05, "loss": 0.0902, "step": 92650 }, { "epoch": 2.16, "learning_rate": 1.3874410363521157e-05, "loss": 0.1535, "step": 92655 }, { "epoch": 2.16, "learning_rate": 1.3873530309075123e-05, "loss": 0.1143, "step": 92660 }, { "epoch": 2.16, "learning_rate": 1.3872650254629087e-05, "loss": 0.0473, "step": 92665 }, { "epoch": 2.16, "learning_rate": 1.3871770200183052e-05, "loss": 0.1988, "step": 92670 }, { "epoch": 2.16, "learning_rate": 1.3870890145737017e-05, "loss": 0.1525, "step": 92675 }, { "epoch": 2.16, "learning_rate": 1.3870010091290982e-05, "loss": 0.0721, "step": 92680 }, { "epoch": 2.16, "learning_rate": 1.3869130036844947e-05, "loss": 0.0633, "step": 92685 }, { "epoch": 2.16, "learning_rate": 1.3868249982398912e-05, "loss": 0.0108, "step": 92690 }, { "epoch": 2.16, "learning_rate": 1.3867369927952877e-05, "loss": 0.0547, "step": 92695 }, { "epoch": 2.16, "learning_rate": 1.3866489873506842e-05, "loss": 0.0797, "step": 92700 }, { "epoch": 2.16, "learning_rate": 1.3865609819060805e-05, "loss": 0.0891, "step": 92705 }, { "epoch": 2.16, "learning_rate": 1.3864729764614772e-05, "loss": 0.0676, "step": 92710 }, { "epoch": 2.16, "learning_rate": 1.3863849710168735e-05, "loss": 0.1012, "step": 92715 }, { "epoch": 2.16, "learning_rate": 1.3862969655722702e-05, "loss": 0.2114, "step": 92720 }, { "epoch": 2.16, "learning_rate": 1.3862089601276665e-05, "loss": 0.2437, "step": 92725 }, { "epoch": 2.16, "learning_rate": 1.3861209546830632e-05, "loss": 0.1076, "step": 92730 }, { "epoch": 2.16, "learning_rate": 1.3860329492384595e-05, "loss": 0.0794, "step": 92735 }, { "epoch": 2.16, "learning_rate": 1.385944943793856e-05, "loss": 0.005, "step": 92740 }, { "epoch": 2.16, "learning_rate": 1.3858569383492525e-05, "loss": 0.019, "step": 92745 }, { "epoch": 2.16, "learning_rate": 1.385768932904649e-05, "loss": 0.0589, "step": 92750 }, { "epoch": 2.16, "learning_rate": 1.3856809274600455e-05, "loss": 0.1643, "step": 92755 }, { "epoch": 2.16, "learning_rate": 1.385592922015442e-05, "loss": 0.0254, "step": 92760 }, { "epoch": 2.16, "learning_rate": 1.3855049165708385e-05, "loss": 0.3485, "step": 92765 }, { "epoch": 2.16, "learning_rate": 1.385416911126235e-05, "loss": 0.3426, "step": 92770 }, { "epoch": 2.16, "learning_rate": 1.3853289056816315e-05, "loss": 0.291, "step": 92775 }, { "epoch": 2.16, "learning_rate": 1.385240900237028e-05, "loss": 0.0421, "step": 92780 }, { "epoch": 2.16, "learning_rate": 1.3851528947924245e-05, "loss": 0.0384, "step": 92785 }, { "epoch": 2.16, "learning_rate": 1.385064889347821e-05, "loss": 0.03, "step": 92790 }, { "epoch": 2.17, "learning_rate": 1.3849768839032175e-05, "loss": 0.0469, "step": 92795 }, { "epoch": 2.17, "learning_rate": 1.384888878458614e-05, "loss": 0.0579, "step": 92800 }, { "epoch": 2.17, "learning_rate": 1.3848008730140106e-05, "loss": 0.2252, "step": 92805 }, { "epoch": 2.17, "learning_rate": 1.384712867569407e-05, "loss": 0.1269, "step": 92810 }, { "epoch": 2.17, "learning_rate": 1.3846248621248036e-05, "loss": 0.1139, "step": 92815 }, { "epoch": 2.17, "learning_rate": 1.3845368566801999e-05, "loss": 0.0868, "step": 92820 }, { "epoch": 2.17, "learning_rate": 1.3844488512355966e-05, "loss": 0.2113, "step": 92825 }, { "epoch": 2.17, "learning_rate": 1.3843608457909929e-05, "loss": 0.0494, "step": 92830 }, { "epoch": 2.17, "learning_rate": 1.3842728403463896e-05, "loss": 0.0077, "step": 92835 }, { "epoch": 2.17, "learning_rate": 1.3841848349017859e-05, "loss": 0.0683, "step": 92840 }, { "epoch": 2.17, "learning_rate": 1.3840968294571826e-05, "loss": 0.0718, "step": 92845 }, { "epoch": 2.17, "learning_rate": 1.3840088240125789e-05, "loss": 0.0605, "step": 92850 }, { "epoch": 2.17, "learning_rate": 1.3839208185679754e-05, "loss": 0.1058, "step": 92855 }, { "epoch": 2.17, "learning_rate": 1.3838328131233719e-05, "loss": 0.1869, "step": 92860 }, { "epoch": 2.17, "learning_rate": 1.3837448076787684e-05, "loss": 0.0766, "step": 92865 }, { "epoch": 2.17, "learning_rate": 1.3836568022341649e-05, "loss": 0.1576, "step": 92870 }, { "epoch": 2.17, "learning_rate": 1.3835687967895614e-05, "loss": 0.403, "step": 92875 }, { "epoch": 2.17, "learning_rate": 1.3834807913449579e-05, "loss": 0.0466, "step": 92880 }, { "epoch": 2.17, "learning_rate": 1.3833927859003544e-05, "loss": 0.102, "step": 92885 }, { "epoch": 2.17, "learning_rate": 1.3833047804557507e-05, "loss": 0.0746, "step": 92890 }, { "epoch": 2.17, "learning_rate": 1.3832167750111474e-05, "loss": 0.0655, "step": 92895 }, { "epoch": 2.17, "learning_rate": 1.3831287695665438e-05, "loss": 0.0589, "step": 92900 }, { "epoch": 2.17, "learning_rate": 1.3830407641219404e-05, "loss": 0.0774, "step": 92905 }, { "epoch": 2.17, "learning_rate": 1.3829527586773368e-05, "loss": 0.0746, "step": 92910 }, { "epoch": 2.17, "learning_rate": 1.3828647532327334e-05, "loss": 0.0892, "step": 92915 }, { "epoch": 2.17, "learning_rate": 1.38277674778813e-05, "loss": 0.1878, "step": 92920 }, { "epoch": 2.17, "learning_rate": 1.3826887423435264e-05, "loss": 0.4114, "step": 92925 }, { "epoch": 2.17, "learning_rate": 1.382600736898923e-05, "loss": 0.028, "step": 92930 }, { "epoch": 2.17, "learning_rate": 1.3825127314543193e-05, "loss": 0.1039, "step": 92935 }, { "epoch": 2.17, "learning_rate": 1.382424726009716e-05, "loss": 0.0059, "step": 92940 }, { "epoch": 2.17, "learning_rate": 1.3823367205651123e-05, "loss": 0.1221, "step": 92945 }, { "epoch": 2.17, "learning_rate": 1.382248715120509e-05, "loss": 0.0381, "step": 92950 }, { "epoch": 2.17, "learning_rate": 1.3821607096759053e-05, "loss": 0.0768, "step": 92955 }, { "epoch": 2.17, "learning_rate": 1.382072704231302e-05, "loss": 0.1088, "step": 92960 }, { "epoch": 2.17, "learning_rate": 1.3819846987866983e-05, "loss": 0.0633, "step": 92965 }, { "epoch": 2.17, "learning_rate": 1.3818966933420948e-05, "loss": 0.1952, "step": 92970 }, { "epoch": 2.17, "learning_rate": 1.3818086878974913e-05, "loss": 0.2711, "step": 92975 }, { "epoch": 2.17, "learning_rate": 1.3817206824528878e-05, "loss": 0.044, "step": 92980 }, { "epoch": 2.17, "learning_rate": 1.3816326770082843e-05, "loss": 0.0349, "step": 92985 }, { "epoch": 2.17, "learning_rate": 1.3815446715636808e-05, "loss": 0.0609, "step": 92990 }, { "epoch": 2.17, "learning_rate": 1.3814566661190773e-05, "loss": 0.1068, "step": 92995 }, { "epoch": 2.17, "learning_rate": 1.3813686606744738e-05, "loss": 0.0352, "step": 93000 }, { "epoch": 2.17, "learning_rate": 1.3812806552298701e-05, "loss": 0.0314, "step": 93005 }, { "epoch": 2.17, "learning_rate": 1.3811926497852668e-05, "loss": 0.2665, "step": 93010 }, { "epoch": 2.17, "learning_rate": 1.3811046443406631e-05, "loss": 0.2787, "step": 93015 }, { "epoch": 2.17, "learning_rate": 1.3810166388960598e-05, "loss": 0.207, "step": 93020 }, { "epoch": 2.17, "learning_rate": 1.3809286334514561e-05, "loss": 0.4842, "step": 93025 }, { "epoch": 2.17, "learning_rate": 1.3808406280068528e-05, "loss": 0.0345, "step": 93030 }, { "epoch": 2.17, "learning_rate": 1.3807526225622491e-05, "loss": 0.0158, "step": 93035 }, { "epoch": 2.17, "learning_rate": 1.3806646171176456e-05, "loss": 0.0265, "step": 93040 }, { "epoch": 2.17, "learning_rate": 1.3805766116730421e-05, "loss": 0.014, "step": 93045 }, { "epoch": 2.17, "learning_rate": 1.3804886062284386e-05, "loss": 0.0602, "step": 93050 }, { "epoch": 2.17, "learning_rate": 1.3804006007838353e-05, "loss": 0.0201, "step": 93055 }, { "epoch": 2.17, "learning_rate": 1.3803125953392316e-05, "loss": 0.0794, "step": 93060 }, { "epoch": 2.17, "learning_rate": 1.3802245898946283e-05, "loss": 0.0972, "step": 93065 }, { "epoch": 2.17, "learning_rate": 1.3801365844500247e-05, "loss": 0.2699, "step": 93070 }, { "epoch": 2.17, "learning_rate": 1.3800485790054213e-05, "loss": 0.4515, "step": 93075 }, { "epoch": 2.17, "learning_rate": 1.3799605735608177e-05, "loss": 0.0514, "step": 93080 }, { "epoch": 2.17, "learning_rate": 1.3798725681162142e-05, "loss": 0.0092, "step": 93085 }, { "epoch": 2.17, "learning_rate": 1.3797845626716107e-05, "loss": 0.1472, "step": 93090 }, { "epoch": 2.17, "learning_rate": 1.3796965572270072e-05, "loss": 0.1027, "step": 93095 }, { "epoch": 2.17, "learning_rate": 1.3796085517824037e-05, "loss": 0.0238, "step": 93100 }, { "epoch": 2.17, "learning_rate": 1.3795205463378002e-05, "loss": 0.105, "step": 93105 }, { "epoch": 2.17, "learning_rate": 1.3794325408931967e-05, "loss": 0.0241, "step": 93110 }, { "epoch": 2.17, "learning_rate": 1.3793445354485932e-05, "loss": 0.1601, "step": 93115 }, { "epoch": 2.17, "learning_rate": 1.3792565300039895e-05, "loss": 0.3241, "step": 93120 }, { "epoch": 2.17, "learning_rate": 1.3791685245593862e-05, "loss": 0.2851, "step": 93125 }, { "epoch": 2.17, "learning_rate": 1.3790805191147825e-05, "loss": 0.0435, "step": 93130 }, { "epoch": 2.17, "learning_rate": 1.3789925136701792e-05, "loss": 0.0322, "step": 93135 }, { "epoch": 2.17, "learning_rate": 1.3789045082255755e-05, "loss": 0.0436, "step": 93140 }, { "epoch": 2.17, "learning_rate": 1.3788165027809722e-05, "loss": 0.0743, "step": 93145 }, { "epoch": 2.17, "learning_rate": 1.3787284973363685e-05, "loss": 0.1564, "step": 93150 }, { "epoch": 2.17, "learning_rate": 1.378640491891765e-05, "loss": 0.0892, "step": 93155 }, { "epoch": 2.17, "learning_rate": 1.3785524864471615e-05, "loss": 0.0801, "step": 93160 }, { "epoch": 2.17, "learning_rate": 1.378464481002558e-05, "loss": 0.1276, "step": 93165 }, { "epoch": 2.17, "learning_rate": 1.3783764755579545e-05, "loss": 0.1181, "step": 93170 }, { "epoch": 2.17, "learning_rate": 1.378288470113351e-05, "loss": 0.0982, "step": 93175 }, { "epoch": 2.17, "learning_rate": 1.3782004646687475e-05, "loss": 0.0206, "step": 93180 }, { "epoch": 2.17, "learning_rate": 1.378112459224144e-05, "loss": 0.0502, "step": 93185 }, { "epoch": 2.17, "learning_rate": 1.3780244537795405e-05, "loss": 0.0568, "step": 93190 }, { "epoch": 2.17, "learning_rate": 1.377936448334937e-05, "loss": 0.0452, "step": 93195 }, { "epoch": 2.17, "learning_rate": 1.3778484428903335e-05, "loss": 0.058, "step": 93200 }, { "epoch": 2.17, "learning_rate": 1.37776043744573e-05, "loss": 0.0483, "step": 93205 }, { "epoch": 2.17, "learning_rate": 1.3776724320011265e-05, "loss": 0.0512, "step": 93210 }, { "epoch": 2.17, "learning_rate": 1.377584426556523e-05, "loss": 0.1715, "step": 93215 }, { "epoch": 2.17, "learning_rate": 1.3774964211119195e-05, "loss": 0.3808, "step": 93220 }, { "epoch": 2.18, "learning_rate": 1.377408415667316e-05, "loss": 0.283, "step": 93225 }, { "epoch": 2.18, "learning_rate": 1.3773204102227125e-05, "loss": 0.0694, "step": 93230 }, { "epoch": 2.18, "learning_rate": 1.3772324047781089e-05, "loss": 0.0699, "step": 93235 }, { "epoch": 2.18, "learning_rate": 1.3771443993335056e-05, "loss": 0.0186, "step": 93240 }, { "epoch": 2.18, "learning_rate": 1.3770563938889019e-05, "loss": 0.0714, "step": 93245 }, { "epoch": 2.18, "learning_rate": 1.3769683884442986e-05, "loss": 0.0771, "step": 93250 }, { "epoch": 2.18, "learning_rate": 1.3768803829996949e-05, "loss": 0.1089, "step": 93255 }, { "epoch": 2.18, "learning_rate": 1.3767923775550916e-05, "loss": 0.0589, "step": 93260 }, { "epoch": 2.18, "learning_rate": 1.3767043721104879e-05, "loss": 0.1502, "step": 93265 }, { "epoch": 2.18, "learning_rate": 1.3766163666658844e-05, "loss": 0.1704, "step": 93270 }, { "epoch": 2.18, "learning_rate": 1.3765283612212809e-05, "loss": 0.3171, "step": 93275 }, { "epoch": 2.18, "learning_rate": 1.3764403557766774e-05, "loss": 0.1337, "step": 93280 }, { "epoch": 2.18, "learning_rate": 1.3763523503320739e-05, "loss": 0.052, "step": 93285 }, { "epoch": 2.18, "learning_rate": 1.3762643448874704e-05, "loss": 0.0431, "step": 93290 }, { "epoch": 2.18, "learning_rate": 1.3761763394428669e-05, "loss": 0.1335, "step": 93295 }, { "epoch": 2.18, "learning_rate": 1.3760883339982634e-05, "loss": 0.0928, "step": 93300 }, { "epoch": 2.18, "learning_rate": 1.3760003285536597e-05, "loss": 0.0385, "step": 93305 }, { "epoch": 2.18, "learning_rate": 1.3759123231090564e-05, "loss": 0.112, "step": 93310 }, { "epoch": 2.18, "learning_rate": 1.3758243176644527e-05, "loss": 0.158, "step": 93315 }, { "epoch": 2.18, "learning_rate": 1.3757363122198494e-05, "loss": 0.1396, "step": 93320 }, { "epoch": 2.18, "learning_rate": 1.375648306775246e-05, "loss": 0.1444, "step": 93325 }, { "epoch": 2.18, "learning_rate": 1.3755603013306424e-05, "loss": 0.0893, "step": 93330 }, { "epoch": 2.18, "learning_rate": 1.375472295886039e-05, "loss": 0.0023, "step": 93335 }, { "epoch": 2.18, "learning_rate": 1.3753842904414353e-05, "loss": 0.1066, "step": 93340 }, { "epoch": 2.18, "learning_rate": 1.375296284996832e-05, "loss": 0.1033, "step": 93345 }, { "epoch": 2.18, "learning_rate": 1.3752082795522283e-05, "loss": 0.0976, "step": 93350 }, { "epoch": 2.18, "learning_rate": 1.375120274107625e-05, "loss": 0.0727, "step": 93355 }, { "epoch": 2.18, "learning_rate": 1.3750322686630213e-05, "loss": 0.0656, "step": 93360 }, { "epoch": 2.18, "learning_rate": 1.374944263218418e-05, "loss": 0.0995, "step": 93365 }, { "epoch": 2.18, "learning_rate": 1.3748562577738143e-05, "loss": 0.0556, "step": 93370 }, { "epoch": 2.18, "learning_rate": 1.374768252329211e-05, "loss": 0.1964, "step": 93375 }, { "epoch": 2.18, "learning_rate": 1.3746802468846073e-05, "loss": 0.0008, "step": 93380 }, { "epoch": 2.18, "learning_rate": 1.3745922414400038e-05, "loss": 0.0167, "step": 93385 }, { "epoch": 2.18, "learning_rate": 1.3745042359954003e-05, "loss": 0.0851, "step": 93390 }, { "epoch": 2.18, "learning_rate": 1.3744162305507968e-05, "loss": 0.0377, "step": 93395 }, { "epoch": 2.18, "learning_rate": 1.3743282251061933e-05, "loss": 0.0694, "step": 93400 }, { "epoch": 2.18, "learning_rate": 1.3742402196615898e-05, "loss": 0.0896, "step": 93405 }, { "epoch": 2.18, "learning_rate": 1.3741522142169863e-05, "loss": 0.1659, "step": 93410 }, { "epoch": 2.18, "learning_rate": 1.3740642087723828e-05, "loss": 0.0996, "step": 93415 }, { "epoch": 2.18, "learning_rate": 1.3739762033277791e-05, "loss": 0.2189, "step": 93420 }, { "epoch": 2.18, "learning_rate": 1.3738881978831758e-05, "loss": 0.237, "step": 93425 }, { "epoch": 2.18, "learning_rate": 1.3738001924385721e-05, "loss": 0.0612, "step": 93430 }, { "epoch": 2.18, "learning_rate": 1.3737121869939688e-05, "loss": 0.0264, "step": 93435 }, { "epoch": 2.18, "learning_rate": 1.3736241815493651e-05, "loss": 0.1444, "step": 93440 }, { "epoch": 2.18, "learning_rate": 1.3735361761047618e-05, "loss": 0.0948, "step": 93445 }, { "epoch": 2.18, "learning_rate": 1.3734481706601581e-05, "loss": 0.0401, "step": 93450 }, { "epoch": 2.18, "learning_rate": 1.3733601652155546e-05, "loss": 0.1499, "step": 93455 }, { "epoch": 2.18, "learning_rate": 1.3732721597709511e-05, "loss": 0.1222, "step": 93460 }, { "epoch": 2.18, "learning_rate": 1.3731841543263476e-05, "loss": 0.3017, "step": 93465 }, { "epoch": 2.18, "learning_rate": 1.3730961488817443e-05, "loss": 0.2908, "step": 93470 }, { "epoch": 2.18, "learning_rate": 1.3730081434371406e-05, "loss": 0.2598, "step": 93475 }, { "epoch": 2.18, "learning_rate": 1.3729201379925373e-05, "loss": 0.0656, "step": 93480 }, { "epoch": 2.18, "learning_rate": 1.3728321325479336e-05, "loss": 0.0087, "step": 93485 }, { "epoch": 2.18, "learning_rate": 1.3727441271033301e-05, "loss": 0.0873, "step": 93490 }, { "epoch": 2.18, "learning_rate": 1.3726561216587266e-05, "loss": 0.0734, "step": 93495 }, { "epoch": 2.18, "learning_rate": 1.3725681162141232e-05, "loss": 0.0075, "step": 93500 }, { "epoch": 2.18, "learning_rate": 1.3724801107695197e-05, "loss": 0.0937, "step": 93505 }, { "epoch": 2.18, "learning_rate": 1.3723921053249162e-05, "loss": 0.1385, "step": 93510 }, { "epoch": 2.18, "learning_rate": 1.3723040998803127e-05, "loss": 0.0416, "step": 93515 }, { "epoch": 2.18, "learning_rate": 1.3722160944357092e-05, "loss": 0.1829, "step": 93520 }, { "epoch": 2.18, "learning_rate": 1.3721280889911057e-05, "loss": 0.3693, "step": 93525 }, { "epoch": 2.18, "learning_rate": 1.3720400835465022e-05, "loss": 0.0337, "step": 93530 }, { "epoch": 2.18, "learning_rate": 1.3719520781018985e-05, "loss": 0.0898, "step": 93535 }, { "epoch": 2.18, "learning_rate": 1.3718640726572952e-05, "loss": 0.0627, "step": 93540 }, { "epoch": 2.18, "learning_rate": 1.3717760672126915e-05, "loss": 0.0389, "step": 93545 }, { "epoch": 2.18, "learning_rate": 1.3716880617680882e-05, "loss": 0.0246, "step": 93550 }, { "epoch": 2.18, "learning_rate": 1.3716000563234845e-05, "loss": 0.1132, "step": 93555 }, { "epoch": 2.18, "learning_rate": 1.3715120508788812e-05, "loss": 0.1114, "step": 93560 }, { "epoch": 2.18, "learning_rate": 1.3714240454342775e-05, "loss": 0.1095, "step": 93565 }, { "epoch": 2.18, "learning_rate": 1.371336039989674e-05, "loss": 0.19, "step": 93570 }, { "epoch": 2.18, "learning_rate": 1.3712480345450705e-05, "loss": 0.2013, "step": 93575 }, { "epoch": 2.18, "learning_rate": 1.371160029100467e-05, "loss": 0.0518, "step": 93580 }, { "epoch": 2.18, "learning_rate": 1.3710720236558635e-05, "loss": 0.0492, "step": 93585 }, { "epoch": 2.18, "learning_rate": 1.37098401821126e-05, "loss": 0.0121, "step": 93590 }, { "epoch": 2.18, "learning_rate": 1.3708960127666565e-05, "loss": 0.0651, "step": 93595 }, { "epoch": 2.18, "learning_rate": 1.370808007322053e-05, "loss": 0.0583, "step": 93600 }, { "epoch": 2.18, "learning_rate": 1.3707200018774495e-05, "loss": 0.1469, "step": 93605 }, { "epoch": 2.18, "learning_rate": 1.370631996432846e-05, "loss": 0.1318, "step": 93610 }, { "epoch": 2.18, "learning_rate": 1.3705439909882425e-05, "loss": 0.1493, "step": 93615 }, { "epoch": 2.18, "learning_rate": 1.370455985543639e-05, "loss": 0.248, "step": 93620 }, { "epoch": 2.18, "learning_rate": 1.3703679800990355e-05, "loss": 0.1388, "step": 93625 }, { "epoch": 2.18, "learning_rate": 1.370279974654432e-05, "loss": 0.116, "step": 93630 }, { "epoch": 2.18, "learning_rate": 1.3701919692098285e-05, "loss": 0.011, "step": 93635 }, { "epoch": 2.18, "learning_rate": 1.3701039637652249e-05, "loss": 0.0328, "step": 93640 }, { "epoch": 2.18, "learning_rate": 1.3700159583206215e-05, "loss": 0.0227, "step": 93645 }, { "epoch": 2.18, "learning_rate": 1.3699279528760179e-05, "loss": 0.0765, "step": 93650 }, { "epoch": 2.19, "learning_rate": 1.3698399474314145e-05, "loss": 0.02, "step": 93655 }, { "epoch": 2.19, "learning_rate": 1.3697519419868109e-05, "loss": 0.138, "step": 93660 }, { "epoch": 2.19, "learning_rate": 1.3696639365422076e-05, "loss": 0.2498, "step": 93665 }, { "epoch": 2.19, "learning_rate": 1.3695759310976039e-05, "loss": 0.296, "step": 93670 }, { "epoch": 2.19, "learning_rate": 1.3694879256530006e-05, "loss": 0.163, "step": 93675 }, { "epoch": 2.19, "learning_rate": 1.3693999202083969e-05, "loss": 0.1234, "step": 93680 }, { "epoch": 2.19, "learning_rate": 1.3693119147637934e-05, "loss": 0.0696, "step": 93685 }, { "epoch": 2.19, "learning_rate": 1.3692239093191899e-05, "loss": 0.0429, "step": 93690 }, { "epoch": 2.19, "learning_rate": 1.3691359038745864e-05, "loss": 0.1214, "step": 93695 }, { "epoch": 2.19, "learning_rate": 1.3690478984299829e-05, "loss": 0.0427, "step": 93700 }, { "epoch": 2.19, "learning_rate": 1.3689598929853794e-05, "loss": 0.0722, "step": 93705 }, { "epoch": 2.19, "learning_rate": 1.3688718875407759e-05, "loss": 0.2101, "step": 93710 }, { "epoch": 2.19, "learning_rate": 1.3687838820961724e-05, "loss": 0.1177, "step": 93715 }, { "epoch": 2.19, "learning_rate": 1.3686958766515687e-05, "loss": 0.2495, "step": 93720 }, { "epoch": 2.19, "learning_rate": 1.3686078712069654e-05, "loss": 0.2236, "step": 93725 }, { "epoch": 2.19, "learning_rate": 1.3685198657623617e-05, "loss": 0.047, "step": 93730 }, { "epoch": 2.19, "learning_rate": 1.3684318603177584e-05, "loss": 0.0459, "step": 93735 }, { "epoch": 2.19, "learning_rate": 1.3683438548731549e-05, "loss": 0.02, "step": 93740 }, { "epoch": 2.19, "learning_rate": 1.3682558494285514e-05, "loss": 0.0522, "step": 93745 }, { "epoch": 2.19, "learning_rate": 1.368167843983948e-05, "loss": 0.0937, "step": 93750 }, { "epoch": 2.19, "learning_rate": 1.3680798385393442e-05, "loss": 0.03, "step": 93755 }, { "epoch": 2.19, "learning_rate": 1.367991833094741e-05, "loss": 0.1025, "step": 93760 }, { "epoch": 2.19, "learning_rate": 1.3679214287390579e-05, "loss": 0.2295, "step": 93765 }, { "epoch": 2.19, "learning_rate": 1.3678334232944546e-05, "loss": 0.2176, "step": 93770 }, { "epoch": 2.19, "learning_rate": 1.3677454178498509e-05, "loss": 0.3808, "step": 93775 }, { "epoch": 2.19, "learning_rate": 1.3676574124052476e-05, "loss": 0.0745, "step": 93780 }, { "epoch": 2.19, "learning_rate": 1.3675694069606439e-05, "loss": 0.0434, "step": 93785 }, { "epoch": 2.19, "learning_rate": 1.3674814015160406e-05, "loss": 0.0219, "step": 93790 }, { "epoch": 2.19, "learning_rate": 1.367393396071437e-05, "loss": 0.0122, "step": 93795 }, { "epoch": 2.19, "learning_rate": 1.3673053906268334e-05, "loss": 0.0471, "step": 93800 }, { "epoch": 2.19, "learning_rate": 1.36721738518223e-05, "loss": 0.1208, "step": 93805 }, { "epoch": 2.19, "learning_rate": 1.3671293797376264e-05, "loss": 0.0835, "step": 93810 }, { "epoch": 2.19, "learning_rate": 1.367041374293023e-05, "loss": 0.1383, "step": 93815 }, { "epoch": 2.19, "learning_rate": 1.3669533688484194e-05, "loss": 0.2926, "step": 93820 }, { "epoch": 2.19, "learning_rate": 1.366865363403816e-05, "loss": 0.2731, "step": 93825 }, { "epoch": 2.19, "learning_rate": 1.3667773579592124e-05, "loss": 0.0399, "step": 93830 }, { "epoch": 2.19, "learning_rate": 1.3666893525146089e-05, "loss": 0.0411, "step": 93835 }, { "epoch": 2.19, "learning_rate": 1.3666013470700054e-05, "loss": 0.0634, "step": 93840 }, { "epoch": 2.19, "learning_rate": 1.366513341625402e-05, "loss": 0.0252, "step": 93845 }, { "epoch": 2.19, "learning_rate": 1.3664253361807984e-05, "loss": 0.0707, "step": 93850 }, { "epoch": 2.19, "learning_rate": 1.366337330736195e-05, "loss": 0.0297, "step": 93855 }, { "epoch": 2.19, "learning_rate": 1.3662493252915914e-05, "loss": 0.1135, "step": 93860 }, { "epoch": 2.19, "learning_rate": 1.366161319846988e-05, "loss": 0.1664, "step": 93865 }, { "epoch": 2.19, "learning_rate": 1.3660733144023844e-05, "loss": 0.1307, "step": 93870 }, { "epoch": 2.19, "learning_rate": 1.365985308957781e-05, "loss": 0.2091, "step": 93875 }, { "epoch": 2.19, "learning_rate": 1.3658973035131773e-05, "loss": 0.0927, "step": 93880 }, { "epoch": 2.19, "learning_rate": 1.365809298068574e-05, "loss": 0.0011, "step": 93885 }, { "epoch": 2.19, "learning_rate": 1.3657212926239703e-05, "loss": 0.0086, "step": 93890 }, { "epoch": 2.19, "learning_rate": 1.365633287179367e-05, "loss": 0.0476, "step": 93895 }, { "epoch": 2.19, "learning_rate": 1.3655452817347633e-05, "loss": 0.0745, "step": 93900 }, { "epoch": 2.19, "learning_rate": 1.36545727629016e-05, "loss": 0.0698, "step": 93905 }, { "epoch": 2.19, "learning_rate": 1.3653692708455563e-05, "loss": 0.1289, "step": 93910 }, { "epoch": 2.19, "learning_rate": 1.3652812654009528e-05, "loss": 0.2388, "step": 93915 }, { "epoch": 2.19, "learning_rate": 1.3651932599563493e-05, "loss": 0.1267, "step": 93920 }, { "epoch": 2.19, "learning_rate": 1.3651052545117458e-05, "loss": 0.2441, "step": 93925 }, { "epoch": 2.19, "learning_rate": 1.3650172490671425e-05, "loss": 0.058, "step": 93930 }, { "epoch": 2.19, "learning_rate": 1.3649292436225388e-05, "loss": 0.0234, "step": 93935 }, { "epoch": 2.19, "learning_rate": 1.3648412381779355e-05, "loss": 0.0526, "step": 93940 }, { "epoch": 2.19, "learning_rate": 1.3647532327333318e-05, "loss": 0.0727, "step": 93945 }, { "epoch": 2.19, "learning_rate": 1.3646652272887283e-05, "loss": 0.1066, "step": 93950 }, { "epoch": 2.19, "learning_rate": 1.3645772218441248e-05, "loss": 0.0491, "step": 93955 }, { "epoch": 2.19, "learning_rate": 1.3644892163995213e-05, "loss": 0.1058, "step": 93960 }, { "epoch": 2.19, "learning_rate": 1.3644012109549178e-05, "loss": 0.1816, "step": 93965 }, { "epoch": 2.19, "learning_rate": 1.3643132055103143e-05, "loss": 0.3165, "step": 93970 }, { "epoch": 2.19, "learning_rate": 1.3642252000657108e-05, "loss": 0.2477, "step": 93975 }, { "epoch": 2.19, "learning_rate": 1.3641371946211073e-05, "loss": 0.0862, "step": 93980 }, { "epoch": 2.19, "learning_rate": 1.3640491891765036e-05, "loss": 0.0234, "step": 93985 }, { "epoch": 2.19, "learning_rate": 1.3639611837319003e-05, "loss": 0.0354, "step": 93990 }, { "epoch": 2.19, "learning_rate": 1.3638731782872966e-05, "loss": 0.1902, "step": 93995 }, { "epoch": 2.19, "learning_rate": 1.3637851728426933e-05, "loss": 0.1976, "step": 94000 }, { "epoch": 2.19, "learning_rate": 1.3636971673980896e-05, "loss": 0.1215, "step": 94005 }, { "epoch": 2.19, "learning_rate": 1.3636091619534863e-05, "loss": 0.207, "step": 94010 }, { "epoch": 2.19, "learning_rate": 1.3635211565088826e-05, "loss": 0.0833, "step": 94015 }, { "epoch": 2.19, "learning_rate": 1.3634331510642793e-05, "loss": 0.3215, "step": 94020 }, { "epoch": 2.19, "learning_rate": 1.3633451456196757e-05, "loss": 0.2237, "step": 94025 }, { "epoch": 2.19, "learning_rate": 1.3632571401750722e-05, "loss": 0.0246, "step": 94030 }, { "epoch": 2.19, "learning_rate": 1.3631691347304687e-05, "loss": 0.0437, "step": 94035 }, { "epoch": 2.19, "learning_rate": 1.3630811292858652e-05, "loss": 0.0229, "step": 94040 }, { "epoch": 2.19, "learning_rate": 1.3629931238412617e-05, "loss": 0.073, "step": 94045 }, { "epoch": 2.19, "learning_rate": 1.3629051183966582e-05, "loss": 0.0669, "step": 94050 }, { "epoch": 2.19, "learning_rate": 1.3628171129520547e-05, "loss": 0.1067, "step": 94055 }, { "epoch": 2.19, "learning_rate": 1.3627291075074512e-05, "loss": 0.0486, "step": 94060 }, { "epoch": 2.19, "learning_rate": 1.3626411020628477e-05, "loss": 0.0544, "step": 94065 }, { "epoch": 2.19, "learning_rate": 1.3625530966182442e-05, "loss": 0.3252, "step": 94070 }, { "epoch": 2.19, "learning_rate": 1.3624650911736407e-05, "loss": 0.2659, "step": 94075 }, { "epoch": 2.2, "learning_rate": 1.3623770857290372e-05, "loss": 0.0501, "step": 94080 }, { "epoch": 2.2, "learning_rate": 1.3622890802844337e-05, "loss": 0.0408, "step": 94085 }, { "epoch": 2.2, "learning_rate": 1.3622010748398302e-05, "loss": 0.0437, "step": 94090 }, { "epoch": 2.2, "learning_rate": 1.3621130693952267e-05, "loss": 0.0381, "step": 94095 }, { "epoch": 2.2, "learning_rate": 1.362025063950623e-05, "loss": 0.227, "step": 94100 }, { "epoch": 2.2, "learning_rate": 1.3619370585060197e-05, "loss": 0.0901, "step": 94105 }, { "epoch": 2.2, "learning_rate": 1.361849053061416e-05, "loss": 0.1226, "step": 94110 }, { "epoch": 2.2, "learning_rate": 1.3617610476168127e-05, "loss": 0.1024, "step": 94115 }, { "epoch": 2.2, "learning_rate": 1.361673042172209e-05, "loss": 0.1525, "step": 94120 }, { "epoch": 2.2, "learning_rate": 1.3615850367276057e-05, "loss": 0.1236, "step": 94125 }, { "epoch": 2.2, "learning_rate": 1.361497031283002e-05, "loss": 0.0025, "step": 94130 }, { "epoch": 2.2, "learning_rate": 1.3614090258383985e-05, "loss": 0.0447, "step": 94135 }, { "epoch": 2.2, "learning_rate": 1.361321020393795e-05, "loss": 0.0658, "step": 94140 }, { "epoch": 2.2, "learning_rate": 1.3612330149491915e-05, "loss": 0.0028, "step": 94145 }, { "epoch": 2.2, "learning_rate": 1.361145009504588e-05, "loss": 0.0746, "step": 94150 }, { "epoch": 2.2, "learning_rate": 1.3610570040599845e-05, "loss": 0.028, "step": 94155 }, { "epoch": 2.2, "learning_rate": 1.360968998615381e-05, "loss": 0.1215, "step": 94160 }, { "epoch": 2.2, "learning_rate": 1.3608809931707775e-05, "loss": 0.1626, "step": 94165 }, { "epoch": 2.2, "learning_rate": 1.360792987726174e-05, "loss": 0.2191, "step": 94170 }, { "epoch": 2.2, "learning_rate": 1.3607049822815705e-05, "loss": 0.1994, "step": 94175 }, { "epoch": 2.2, "learning_rate": 1.3606169768369669e-05, "loss": 0.1009, "step": 94180 }, { "epoch": 2.2, "learning_rate": 1.3605289713923635e-05, "loss": 0.0381, "step": 94185 }, { "epoch": 2.2, "learning_rate": 1.3604409659477599e-05, "loss": 0.0273, "step": 94190 }, { "epoch": 2.2, "learning_rate": 1.3603529605031566e-05, "loss": 0.0739, "step": 94195 }, { "epoch": 2.2, "learning_rate": 1.360264955058553e-05, "loss": 0.0123, "step": 94200 }, { "epoch": 2.2, "learning_rate": 1.3601769496139496e-05, "loss": 0.0829, "step": 94205 }, { "epoch": 2.2, "learning_rate": 1.360088944169346e-05, "loss": 0.198, "step": 94210 }, { "epoch": 2.2, "learning_rate": 1.3600009387247424e-05, "loss": 0.1379, "step": 94215 }, { "epoch": 2.2, "learning_rate": 1.359912933280139e-05, "loss": 0.2031, "step": 94220 }, { "epoch": 2.2, "learning_rate": 1.3598249278355354e-05, "loss": 0.2014, "step": 94225 }, { "epoch": 2.2, "learning_rate": 1.359736922390932e-05, "loss": 0.0714, "step": 94230 }, { "epoch": 2.2, "learning_rate": 1.3596489169463284e-05, "loss": 0.0311, "step": 94235 }, { "epoch": 2.2, "learning_rate": 1.359560911501725e-05, "loss": 0.0859, "step": 94240 }, { "epoch": 2.2, "learning_rate": 1.3594729060571214e-05, "loss": 0.054, "step": 94245 }, { "epoch": 2.2, "learning_rate": 1.3593849006125179e-05, "loss": 0.0072, "step": 94250 }, { "epoch": 2.2, "learning_rate": 1.3592968951679144e-05, "loss": 0.142, "step": 94255 }, { "epoch": 2.2, "learning_rate": 1.3592088897233109e-05, "loss": 0.2017, "step": 94260 }, { "epoch": 2.2, "learning_rate": 1.3591208842787074e-05, "loss": 0.2847, "step": 94265 }, { "epoch": 2.2, "learning_rate": 1.3590328788341039e-05, "loss": 0.1307, "step": 94270 }, { "epoch": 2.2, "learning_rate": 1.3589448733895004e-05, "loss": 0.1992, "step": 94275 }, { "epoch": 2.2, "learning_rate": 1.358856867944897e-05, "loss": 0.0178, "step": 94280 }, { "epoch": 2.2, "learning_rate": 1.3587688625002933e-05, "loss": 0.0115, "step": 94285 }, { "epoch": 2.2, "learning_rate": 1.35868085705569e-05, "loss": 0.0443, "step": 94290 }, { "epoch": 2.2, "learning_rate": 1.3585928516110863e-05, "loss": 0.0789, "step": 94295 }, { "epoch": 2.2, "learning_rate": 1.358504846166483e-05, "loss": 0.0738, "step": 94300 }, { "epoch": 2.2, "learning_rate": 1.3584168407218793e-05, "loss": 0.0529, "step": 94305 }, { "epoch": 2.2, "learning_rate": 1.358328835277276e-05, "loss": 0.0805, "step": 94310 }, { "epoch": 2.2, "learning_rate": 1.3582408298326723e-05, "loss": 0.1757, "step": 94315 }, { "epoch": 2.2, "learning_rate": 1.358152824388069e-05, "loss": 0.2537, "step": 94320 }, { "epoch": 2.2, "learning_rate": 1.3580648189434653e-05, "loss": 0.3296, "step": 94325 }, { "epoch": 2.2, "learning_rate": 1.3579768134988618e-05, "loss": 0.0516, "step": 94330 }, { "epoch": 2.2, "learning_rate": 1.3578888080542584e-05, "loss": 0.0358, "step": 94335 }, { "epoch": 2.2, "learning_rate": 1.3578008026096548e-05, "loss": 0.0825, "step": 94340 }, { "epoch": 2.2, "learning_rate": 1.3577127971650514e-05, "loss": 0.0269, "step": 94345 }, { "epoch": 2.2, "learning_rate": 1.3576247917204478e-05, "loss": 0.0348, "step": 94350 }, { "epoch": 2.2, "learning_rate": 1.3575367862758444e-05, "loss": 0.0992, "step": 94355 }, { "epoch": 2.2, "learning_rate": 1.3574487808312408e-05, "loss": 0.1396, "step": 94360 }, { "epoch": 2.2, "learning_rate": 1.3573783764755581e-05, "loss": 0.3207, "step": 94365 }, { "epoch": 2.2, "learning_rate": 1.3572903710309544e-05, "loss": 0.1368, "step": 94370 }, { "epoch": 2.2, "learning_rate": 1.357202365586351e-05, "loss": 0.1318, "step": 94375 }, { "epoch": 2.2, "learning_rate": 1.3571143601417474e-05, "loss": 0.0681, "step": 94380 }, { "epoch": 2.2, "learning_rate": 1.357026354697144e-05, "loss": 0.0205, "step": 94385 }, { "epoch": 2.2, "learning_rate": 1.3569383492525406e-05, "loss": 0.1177, "step": 94390 }, { "epoch": 2.2, "learning_rate": 1.356850343807937e-05, "loss": 0.0868, "step": 94395 }, { "epoch": 2.2, "learning_rate": 1.3567623383633336e-05, "loss": 0.02, "step": 94400 }, { "epoch": 2.2, "learning_rate": 1.35667433291873e-05, "loss": 0.0216, "step": 94405 }, { "epoch": 2.2, "learning_rate": 1.3565863274741264e-05, "loss": 0.1359, "step": 94410 }, { "epoch": 2.2, "learning_rate": 1.356498322029523e-05, "loss": 0.2149, "step": 94415 }, { "epoch": 2.2, "learning_rate": 1.3564103165849194e-05, "loss": 0.2045, "step": 94420 }, { "epoch": 2.2, "learning_rate": 1.356322311140316e-05, "loss": 0.3431, "step": 94425 }, { "epoch": 2.2, "learning_rate": 1.3562343056957124e-05, "loss": 0.0609, "step": 94430 }, { "epoch": 2.2, "learning_rate": 1.356146300251109e-05, "loss": 0.0237, "step": 94435 }, { "epoch": 2.2, "learning_rate": 1.3560582948065054e-05, "loss": 0.073, "step": 94440 }, { "epoch": 2.2, "learning_rate": 1.3559702893619018e-05, "loss": 0.0757, "step": 94445 }, { "epoch": 2.2, "learning_rate": 1.3558822839172985e-05, "loss": 0.0346, "step": 94450 }, { "epoch": 2.2, "learning_rate": 1.3557942784726948e-05, "loss": 0.1226, "step": 94455 }, { "epoch": 2.2, "learning_rate": 1.3557062730280915e-05, "loss": 0.1096, "step": 94460 }, { "epoch": 2.2, "learning_rate": 1.3556182675834878e-05, "loss": 0.1895, "step": 94465 }, { "epoch": 2.2, "learning_rate": 1.3555302621388845e-05, "loss": 0.3238, "step": 94470 }, { "epoch": 2.2, "learning_rate": 1.3554422566942808e-05, "loss": 0.4477, "step": 94475 }, { "epoch": 2.2, "learning_rate": 1.3553542512496773e-05, "loss": 0.133, "step": 94480 }, { "epoch": 2.2, "learning_rate": 1.3552662458050738e-05, "loss": 0.0713, "step": 94485 }, { "epoch": 2.2, "learning_rate": 1.3551782403604703e-05, "loss": 0.0309, "step": 94490 }, { "epoch": 2.2, "learning_rate": 1.3550902349158668e-05, "loss": 0.0192, "step": 94495 }, { "epoch": 2.2, "learning_rate": 1.3550022294712633e-05, "loss": 0.1362, "step": 94500 }, { "epoch": 2.2, "learning_rate": 1.3549142240266598e-05, "loss": 0.0454, "step": 94505 }, { "epoch": 2.21, "learning_rate": 1.3548262185820563e-05, "loss": 0.1277, "step": 94510 }, { "epoch": 2.21, "learning_rate": 1.3547382131374528e-05, "loss": 0.176, "step": 94515 }, { "epoch": 2.21, "learning_rate": 1.3546502076928493e-05, "loss": 0.3175, "step": 94520 }, { "epoch": 2.21, "learning_rate": 1.3545622022482458e-05, "loss": 0.4265, "step": 94525 }, { "epoch": 2.21, "learning_rate": 1.3544741968036423e-05, "loss": 0.0706, "step": 94530 }, { "epoch": 2.21, "learning_rate": 1.3543861913590388e-05, "loss": 0.0442, "step": 94535 }, { "epoch": 2.21, "learning_rate": 1.3542981859144353e-05, "loss": 0.0219, "step": 94540 }, { "epoch": 2.21, "learning_rate": 1.3542101804698318e-05, "loss": 0.1164, "step": 94545 }, { "epoch": 2.21, "learning_rate": 1.3541221750252283e-05, "loss": 0.1062, "step": 94550 }, { "epoch": 2.21, "learning_rate": 1.3540341695806248e-05, "loss": 0.0635, "step": 94555 }, { "epoch": 2.21, "learning_rate": 1.3539461641360212e-05, "loss": 0.1201, "step": 94560 }, { "epoch": 2.21, "learning_rate": 1.3538581586914178e-05, "loss": 0.1811, "step": 94565 }, { "epoch": 2.21, "learning_rate": 1.3537701532468142e-05, "loss": 0.1902, "step": 94570 }, { "epoch": 2.21, "learning_rate": 1.3536821478022108e-05, "loss": 0.205, "step": 94575 }, { "epoch": 2.21, "learning_rate": 1.3535941423576072e-05, "loss": 0.0326, "step": 94580 }, { "epoch": 2.21, "learning_rate": 1.3535061369130038e-05, "loss": 0.0297, "step": 94585 }, { "epoch": 2.21, "learning_rate": 1.3534181314684002e-05, "loss": 0.0722, "step": 94590 }, { "epoch": 2.21, "learning_rate": 1.3533301260237967e-05, "loss": 0.0124, "step": 94595 }, { "epoch": 2.21, "learning_rate": 1.3532421205791932e-05, "loss": 0.0995, "step": 94600 }, { "epoch": 2.21, "learning_rate": 1.3531541151345897e-05, "loss": 0.0261, "step": 94605 }, { "epoch": 2.21, "learning_rate": 1.3530661096899862e-05, "loss": 0.0871, "step": 94610 }, { "epoch": 2.21, "learning_rate": 1.3529781042453827e-05, "loss": 0.1534, "step": 94615 }, { "epoch": 2.21, "learning_rate": 1.3528900988007792e-05, "loss": 0.1763, "step": 94620 }, { "epoch": 2.21, "learning_rate": 1.3528020933561757e-05, "loss": 0.186, "step": 94625 }, { "epoch": 2.21, "learning_rate": 1.352714087911572e-05, "loss": 0.0769, "step": 94630 }, { "epoch": 2.21, "learning_rate": 1.3526260824669687e-05, "loss": 0.0625, "step": 94635 }, { "epoch": 2.21, "learning_rate": 1.352538077022365e-05, "loss": 0.0539, "step": 94640 }, { "epoch": 2.21, "learning_rate": 1.3524500715777617e-05, "loss": 0.0444, "step": 94645 }, { "epoch": 2.21, "learning_rate": 1.352362066133158e-05, "loss": 0.0574, "step": 94650 }, { "epoch": 2.21, "learning_rate": 1.3522740606885547e-05, "loss": 0.1014, "step": 94655 }, { "epoch": 2.21, "learning_rate": 1.3521860552439512e-05, "loss": 0.1286, "step": 94660 }, { "epoch": 2.21, "learning_rate": 1.3520980497993477e-05, "loss": 0.1201, "step": 94665 }, { "epoch": 2.21, "learning_rate": 1.3520100443547442e-05, "loss": 0.3092, "step": 94670 }, { "epoch": 2.21, "learning_rate": 1.3519220389101405e-05, "loss": 0.3885, "step": 94675 }, { "epoch": 2.21, "learning_rate": 1.3518340334655372e-05, "loss": 0.0362, "step": 94680 }, { "epoch": 2.21, "learning_rate": 1.3517460280209335e-05, "loss": 0.0701, "step": 94685 }, { "epoch": 2.21, "learning_rate": 1.3516580225763302e-05, "loss": 0.0166, "step": 94690 }, { "epoch": 2.21, "learning_rate": 1.3515700171317265e-05, "loss": 0.0192, "step": 94695 }, { "epoch": 2.21, "learning_rate": 1.3514820116871232e-05, "loss": 0.0687, "step": 94700 }, { "epoch": 2.21, "learning_rate": 1.3513940062425195e-05, "loss": 0.0169, "step": 94705 }, { "epoch": 2.21, "learning_rate": 1.351306000797916e-05, "loss": 0.0297, "step": 94710 }, { "epoch": 2.21, "learning_rate": 1.3512179953533126e-05, "loss": 0.088, "step": 94715 }, { "epoch": 2.21, "learning_rate": 1.351129989908709e-05, "loss": 0.363, "step": 94720 }, { "epoch": 2.21, "learning_rate": 1.3510419844641056e-05, "loss": 0.2636, "step": 94725 }, { "epoch": 2.21, "learning_rate": 1.350953979019502e-05, "loss": 0.1426, "step": 94730 }, { "epoch": 2.21, "learning_rate": 1.3508659735748986e-05, "loss": 0.0549, "step": 94735 }, { "epoch": 2.21, "learning_rate": 1.350777968130295e-05, "loss": 0.0059, "step": 94740 }, { "epoch": 2.21, "learning_rate": 1.3506899626856914e-05, "loss": 0.0334, "step": 94745 }, { "epoch": 2.21, "learning_rate": 1.350601957241088e-05, "loss": 0.03, "step": 94750 }, { "epoch": 2.21, "learning_rate": 1.3505139517964844e-05, "loss": 0.1534, "step": 94755 }, { "epoch": 2.21, "learning_rate": 1.350425946351881e-05, "loss": 0.3589, "step": 94760 }, { "epoch": 2.21, "learning_rate": 1.3503379409072774e-05, "loss": 0.1763, "step": 94765 }, { "epoch": 2.21, "learning_rate": 1.350249935462674e-05, "loss": 0.3242, "step": 94770 }, { "epoch": 2.21, "learning_rate": 1.3501619300180704e-05, "loss": 0.3285, "step": 94775 }, { "epoch": 2.21, "learning_rate": 1.3500739245734669e-05, "loss": 0.0645, "step": 94780 }, { "epoch": 2.21, "learning_rate": 1.3499859191288634e-05, "loss": 0.0791, "step": 94785 }, { "epoch": 2.21, "learning_rate": 1.3498979136842599e-05, "loss": 0.0702, "step": 94790 }, { "epoch": 2.21, "learning_rate": 1.3498099082396564e-05, "loss": 0.0196, "step": 94795 }, { "epoch": 2.21, "learning_rate": 1.349721902795053e-05, "loss": 0.0253, "step": 94800 }, { "epoch": 2.21, "learning_rate": 1.3496338973504496e-05, "loss": 0.1297, "step": 94805 }, { "epoch": 2.21, "learning_rate": 1.349545891905846e-05, "loss": 0.1073, "step": 94810 }, { "epoch": 2.21, "learning_rate": 1.3494578864612426e-05, "loss": 0.2304, "step": 94815 }, { "epoch": 2.21, "learning_rate": 1.349369881016639e-05, "loss": 0.4124, "step": 94820 }, { "epoch": 2.21, "learning_rate": 1.3492818755720354e-05, "loss": 0.2205, "step": 94825 }, { "epoch": 2.21, "learning_rate": 1.349193870127432e-05, "loss": 0.0754, "step": 94830 }, { "epoch": 2.21, "learning_rate": 1.3491058646828284e-05, "loss": 0.0581, "step": 94835 }, { "epoch": 2.21, "learning_rate": 1.349017859238225e-05, "loss": 0.0322, "step": 94840 }, { "epoch": 2.21, "learning_rate": 1.3489298537936214e-05, "loss": 0.0636, "step": 94845 }, { "epoch": 2.21, "learning_rate": 1.348841848349018e-05, "loss": 0.1377, "step": 94850 }, { "epoch": 2.21, "learning_rate": 1.3487538429044144e-05, "loss": 0.0555, "step": 94855 }, { "epoch": 2.21, "learning_rate": 1.3486658374598108e-05, "loss": 0.2181, "step": 94860 }, { "epoch": 2.21, "learning_rate": 1.3485778320152074e-05, "loss": 0.1843, "step": 94865 }, { "epoch": 2.21, "learning_rate": 1.3484898265706038e-05, "loss": 0.2121, "step": 94870 }, { "epoch": 2.21, "learning_rate": 1.3484018211260004e-05, "loss": 0.2892, "step": 94875 }, { "epoch": 2.21, "learning_rate": 1.3483138156813968e-05, "loss": 0.0602, "step": 94880 }, { "epoch": 2.21, "learning_rate": 1.3482258102367935e-05, "loss": 0.0607, "step": 94885 }, { "epoch": 2.21, "learning_rate": 1.3481378047921898e-05, "loss": 0.0491, "step": 94890 }, { "epoch": 2.21, "learning_rate": 1.3480497993475863e-05, "loss": 0.0344, "step": 94895 }, { "epoch": 2.21, "learning_rate": 1.3479617939029828e-05, "loss": 0.0705, "step": 94900 }, { "epoch": 2.21, "learning_rate": 1.3478737884583793e-05, "loss": 0.2344, "step": 94905 }, { "epoch": 2.21, "learning_rate": 1.3477857830137758e-05, "loss": 0.0721, "step": 94910 }, { "epoch": 2.21, "learning_rate": 1.3476977775691723e-05, "loss": 0.0875, "step": 94915 }, { "epoch": 2.21, "learning_rate": 1.3476097721245688e-05, "loss": 0.335, "step": 94920 }, { "epoch": 2.21, "learning_rate": 1.3475217666799653e-05, "loss": 0.2031, "step": 94925 }, { "epoch": 2.21, "learning_rate": 1.3474337612353618e-05, "loss": 0.0267, "step": 94930 }, { "epoch": 2.21, "learning_rate": 1.3473457557907583e-05, "loss": 0.0198, "step": 94935 }, { "epoch": 2.22, "learning_rate": 1.3472577503461548e-05, "loss": 0.3651, "step": 94940 }, { "epoch": 2.22, "learning_rate": 1.3471697449015513e-05, "loss": 0.145, "step": 94945 }, { "epoch": 2.22, "learning_rate": 1.3470817394569478e-05, "loss": 0.0402, "step": 94950 }, { "epoch": 2.22, "learning_rate": 1.3469937340123443e-05, "loss": 0.0746, "step": 94955 }, { "epoch": 2.22, "learning_rate": 1.3469057285677408e-05, "loss": 0.2106, "step": 94960 }, { "epoch": 2.22, "learning_rate": 1.3468177231231373e-05, "loss": 0.2678, "step": 94965 }, { "epoch": 2.22, "learning_rate": 1.3467297176785338e-05, "loss": 0.2669, "step": 94970 }, { "epoch": 2.22, "learning_rate": 1.3466417122339301e-05, "loss": 0.4525, "step": 94975 }, { "epoch": 2.22, "learning_rate": 1.3465537067893268e-05, "loss": 0.1065, "step": 94980 }, { "epoch": 2.22, "learning_rate": 1.3464657013447232e-05, "loss": 0.0255, "step": 94985 }, { "epoch": 2.22, "learning_rate": 1.3463776959001198e-05, "loss": 0.0933, "step": 94990 }, { "epoch": 2.22, "learning_rate": 1.3462896904555162e-05, "loss": 0.0477, "step": 94995 }, { "epoch": 2.22, "learning_rate": 1.3462016850109128e-05, "loss": 0.1028, "step": 95000 }, { "epoch": 2.22, "learning_rate": 1.3461136795663092e-05, "loss": 0.1048, "step": 95005 }, { "epoch": 2.22, "learning_rate": 1.3460256741217057e-05, "loss": 0.2118, "step": 95010 }, { "epoch": 2.22, "learning_rate": 1.3459376686771022e-05, "loss": 0.2409, "step": 95015 }, { "epoch": 2.22, "learning_rate": 1.3458496632324987e-05, "loss": 0.2311, "step": 95020 }, { "epoch": 2.22, "learning_rate": 1.3457616577878952e-05, "loss": 0.2398, "step": 95025 }, { "epoch": 2.22, "learning_rate": 1.3456736523432917e-05, "loss": 0.0576, "step": 95030 }, { "epoch": 2.22, "learning_rate": 1.3455856468986882e-05, "loss": 0.028, "step": 95035 }, { "epoch": 2.22, "learning_rate": 1.3454976414540847e-05, "loss": 0.0298, "step": 95040 }, { "epoch": 2.22, "learning_rate": 1.345409636009481e-05, "loss": 0.0067, "step": 95045 }, { "epoch": 2.22, "learning_rate": 1.3453216305648777e-05, "loss": 0.1667, "step": 95050 }, { "epoch": 2.22, "learning_rate": 1.345233625120274e-05, "loss": 0.1038, "step": 95055 }, { "epoch": 2.22, "learning_rate": 1.3451456196756707e-05, "loss": 0.0813, "step": 95060 }, { "epoch": 2.22, "learning_rate": 1.345057614231067e-05, "loss": 0.107, "step": 95065 }, { "epoch": 2.22, "learning_rate": 1.3449696087864637e-05, "loss": 0.1877, "step": 95070 }, { "epoch": 2.22, "learning_rate": 1.3448816033418602e-05, "loss": 0.1695, "step": 95075 }, { "epoch": 2.22, "learning_rate": 1.3447935978972565e-05, "loss": 0.1436, "step": 95080 }, { "epoch": 2.22, "learning_rate": 1.3447055924526532e-05, "loss": 0.1631, "step": 95085 }, { "epoch": 2.22, "learning_rate": 1.3446175870080495e-05, "loss": 0.0852, "step": 95090 }, { "epoch": 2.22, "learning_rate": 1.3445295815634462e-05, "loss": 0.0907, "step": 95095 }, { "epoch": 2.22, "learning_rate": 1.3444415761188425e-05, "loss": 0.1417, "step": 95100 }, { "epoch": 2.22, "learning_rate": 1.3443535706742392e-05, "loss": 0.1277, "step": 95105 }, { "epoch": 2.22, "learning_rate": 1.3442655652296355e-05, "loss": 0.3155, "step": 95110 }, { "epoch": 2.22, "learning_rate": 1.3441775597850322e-05, "loss": 0.151, "step": 95115 }, { "epoch": 2.22, "learning_rate": 1.3440895543404285e-05, "loss": 0.2243, "step": 95120 }, { "epoch": 2.22, "learning_rate": 1.344001548895825e-05, "loss": 0.2602, "step": 95125 }, { "epoch": 2.22, "learning_rate": 1.3439135434512215e-05, "loss": 0.0519, "step": 95130 }, { "epoch": 2.22, "learning_rate": 1.343825538006618e-05, "loss": 0.0009, "step": 95135 }, { "epoch": 2.22, "learning_rate": 1.3437375325620145e-05, "loss": 0.0035, "step": 95140 }, { "epoch": 2.22, "learning_rate": 1.343649527117411e-05, "loss": 0.0047, "step": 95145 }, { "epoch": 2.22, "learning_rate": 1.3435615216728076e-05, "loss": 0.0849, "step": 95150 }, { "epoch": 2.22, "learning_rate": 1.343473516228204e-05, "loss": 0.0247, "step": 95155 }, { "epoch": 2.22, "learning_rate": 1.3433855107836004e-05, "loss": 0.1352, "step": 95160 }, { "epoch": 2.22, "learning_rate": 1.343297505338997e-05, "loss": 0.1428, "step": 95165 }, { "epoch": 2.22, "learning_rate": 1.3432094998943934e-05, "loss": 0.2197, "step": 95170 }, { "epoch": 2.22, "learning_rate": 1.34312149444979e-05, "loss": 0.2254, "step": 95175 }, { "epoch": 2.22, "learning_rate": 1.3430334890051864e-05, "loss": 0.0423, "step": 95180 }, { "epoch": 2.22, "learning_rate": 1.342945483560583e-05, "loss": 0.107, "step": 95185 }, { "epoch": 2.22, "learning_rate": 1.3428574781159794e-05, "loss": 0.0584, "step": 95190 }, { "epoch": 2.22, "learning_rate": 1.3427694726713759e-05, "loss": 0.0971, "step": 95195 }, { "epoch": 2.22, "learning_rate": 1.3426814672267724e-05, "loss": 0.1172, "step": 95200 }, { "epoch": 2.22, "learning_rate": 1.3425934617821689e-05, "loss": 0.1069, "step": 95205 }, { "epoch": 2.22, "learning_rate": 1.3425054563375656e-05, "loss": 0.1425, "step": 95210 }, { "epoch": 2.22, "learning_rate": 1.3424174508929619e-05, "loss": 0.1553, "step": 95215 }, { "epoch": 2.22, "learning_rate": 1.3423294454483586e-05, "loss": 0.2055, "step": 95220 }, { "epoch": 2.22, "learning_rate": 1.3422414400037549e-05, "loss": 0.1865, "step": 95225 }, { "epoch": 2.22, "learning_rate": 1.3421534345591514e-05, "loss": 0.1265, "step": 95230 }, { "epoch": 2.22, "learning_rate": 1.342065429114548e-05, "loss": 0.0026, "step": 95235 }, { "epoch": 2.22, "learning_rate": 1.3419774236699444e-05, "loss": 0.0983, "step": 95240 }, { "epoch": 2.22, "learning_rate": 1.341889418225341e-05, "loss": 0.0443, "step": 95245 }, { "epoch": 2.22, "learning_rate": 1.3418014127807374e-05, "loss": 0.1044, "step": 95250 }, { "epoch": 2.22, "learning_rate": 1.341713407336134e-05, "loss": 0.0299, "step": 95255 }, { "epoch": 2.22, "learning_rate": 1.3416254018915304e-05, "loss": 0.1349, "step": 95260 }, { "epoch": 2.22, "learning_rate": 1.341537396446927e-05, "loss": 0.0908, "step": 95265 }, { "epoch": 2.22, "learning_rate": 1.3414493910023234e-05, "loss": 0.1452, "step": 95270 }, { "epoch": 2.22, "learning_rate": 1.3413613855577198e-05, "loss": 0.2757, "step": 95275 }, { "epoch": 2.22, "learning_rate": 1.3412733801131164e-05, "loss": 0.008, "step": 95280 }, { "epoch": 2.22, "learning_rate": 1.3411853746685128e-05, "loss": 0.0176, "step": 95285 }, { "epoch": 2.22, "learning_rate": 1.3410973692239094e-05, "loss": 0.0535, "step": 95290 }, { "epoch": 2.22, "learning_rate": 1.3410093637793058e-05, "loss": 0.097, "step": 95295 }, { "epoch": 2.22, "learning_rate": 1.3409213583347024e-05, "loss": 0.0679, "step": 95300 }, { "epoch": 2.22, "learning_rate": 1.3408333528900988e-05, "loss": 0.0884, "step": 95305 }, { "epoch": 2.22, "learning_rate": 1.3407453474454953e-05, "loss": 0.1639, "step": 95310 }, { "epoch": 2.22, "learning_rate": 1.3406573420008918e-05, "loss": 0.1262, "step": 95315 }, { "epoch": 2.22, "learning_rate": 1.3405693365562883e-05, "loss": 0.1134, "step": 95320 }, { "epoch": 2.22, "learning_rate": 1.3404813311116848e-05, "loss": 0.2527, "step": 95325 }, { "epoch": 2.22, "learning_rate": 1.3403933256670813e-05, "loss": 0.1011, "step": 95330 }, { "epoch": 2.22, "learning_rate": 1.3403053202224778e-05, "loss": 0.0006, "step": 95335 }, { "epoch": 2.22, "learning_rate": 1.3402173147778743e-05, "loss": 0.1054, "step": 95340 }, { "epoch": 2.22, "learning_rate": 1.3401293093332708e-05, "loss": 0.1129, "step": 95345 }, { "epoch": 2.22, "learning_rate": 1.3400413038886673e-05, "loss": 0.0328, "step": 95350 }, { "epoch": 2.22, "learning_rate": 1.3399532984440638e-05, "loss": 0.1589, "step": 95355 }, { "epoch": 2.22, "learning_rate": 1.3398652929994603e-05, "loss": 0.2, "step": 95360 }, { "epoch": 2.22, "learning_rate": 1.3397772875548568e-05, "loss": 0.226, "step": 95365 }, { "epoch": 2.23, "learning_rate": 1.3396892821102533e-05, "loss": 0.2045, "step": 95370 }, { "epoch": 2.23, "learning_rate": 1.3396012766656498e-05, "loss": 0.3941, "step": 95375 }, { "epoch": 2.23, "learning_rate": 1.3395132712210461e-05, "loss": 0.0845, "step": 95380 }, { "epoch": 2.23, "learning_rate": 1.3394252657764428e-05, "loss": 0.0026, "step": 95385 }, { "epoch": 2.23, "learning_rate": 1.3393372603318391e-05, "loss": 0.0165, "step": 95390 }, { "epoch": 2.23, "learning_rate": 1.3392492548872358e-05, "loss": 0.1025, "step": 95395 }, { "epoch": 2.23, "learning_rate": 1.3391612494426321e-05, "loss": 0.0834, "step": 95400 }, { "epoch": 2.23, "learning_rate": 1.3390732439980288e-05, "loss": 0.0615, "step": 95405 }, { "epoch": 2.23, "learning_rate": 1.3389852385534252e-05, "loss": 0.0612, "step": 95410 }, { "epoch": 2.23, "learning_rate": 1.3388972331088217e-05, "loss": 0.0774, "step": 95415 }, { "epoch": 2.23, "learning_rate": 1.3388092276642182e-05, "loss": 0.2668, "step": 95420 }, { "epoch": 2.23, "learning_rate": 1.3387212222196147e-05, "loss": 0.3042, "step": 95425 }, { "epoch": 2.23, "learning_rate": 1.3386332167750112e-05, "loss": 0.0507, "step": 95430 }, { "epoch": 2.23, "learning_rate": 1.3385452113304077e-05, "loss": 0.027, "step": 95435 }, { "epoch": 2.23, "learning_rate": 1.3384572058858042e-05, "loss": 0.0202, "step": 95440 }, { "epoch": 2.23, "learning_rate": 1.3383692004412007e-05, "loss": 0.0566, "step": 95445 }, { "epoch": 2.23, "learning_rate": 1.3382811949965972e-05, "loss": 0.0288, "step": 95450 }, { "epoch": 2.23, "learning_rate": 1.3381931895519937e-05, "loss": 0.1294, "step": 95455 }, { "epoch": 2.23, "learning_rate": 1.33810518410739e-05, "loss": 0.1022, "step": 95460 }, { "epoch": 2.23, "learning_rate": 1.3380171786627867e-05, "loss": 0.1067, "step": 95465 }, { "epoch": 2.23, "learning_rate": 1.337929173218183e-05, "loss": 0.321, "step": 95470 }, { "epoch": 2.23, "learning_rate": 1.3378411677735797e-05, "loss": 0.2618, "step": 95475 }, { "epoch": 2.23, "learning_rate": 1.3377531623289762e-05, "loss": 0.092, "step": 95480 }, { "epoch": 2.23, "learning_rate": 1.3376651568843727e-05, "loss": 0.0743, "step": 95485 }, { "epoch": 2.23, "learning_rate": 1.3375771514397692e-05, "loss": 0.0265, "step": 95490 }, { "epoch": 2.23, "learning_rate": 1.3374891459951655e-05, "loss": 0.1383, "step": 95495 }, { "epoch": 2.23, "learning_rate": 1.3374011405505622e-05, "loss": 0.097, "step": 95500 }, { "epoch": 2.23, "learning_rate": 1.3373131351059585e-05, "loss": 0.0731, "step": 95505 }, { "epoch": 2.23, "learning_rate": 1.3372251296613552e-05, "loss": 0.0415, "step": 95510 }, { "epoch": 2.23, "learning_rate": 1.3371371242167515e-05, "loss": 0.0822, "step": 95515 }, { "epoch": 2.23, "learning_rate": 1.3370491187721482e-05, "loss": 0.3577, "step": 95520 }, { "epoch": 2.23, "learning_rate": 1.3369611133275445e-05, "loss": 0.4319, "step": 95525 }, { "epoch": 2.23, "learning_rate": 1.336873107882941e-05, "loss": 0.0371, "step": 95530 }, { "epoch": 2.23, "learning_rate": 1.3367851024383375e-05, "loss": 0.0504, "step": 95535 }, { "epoch": 2.23, "learning_rate": 1.336697096993734e-05, "loss": 0.1052, "step": 95540 }, { "epoch": 2.23, "learning_rate": 1.3366090915491305e-05, "loss": 0.0244, "step": 95545 }, { "epoch": 2.23, "learning_rate": 1.336521086104527e-05, "loss": 0.0381, "step": 95550 }, { "epoch": 2.23, "learning_rate": 1.3364330806599235e-05, "loss": 0.2384, "step": 95555 }, { "epoch": 2.23, "learning_rate": 1.33634507521532e-05, "loss": 0.1241, "step": 95560 }, { "epoch": 2.23, "learning_rate": 1.3362570697707164e-05, "loss": 0.1838, "step": 95565 }, { "epoch": 2.23, "learning_rate": 1.336169064326113e-05, "loss": 0.3113, "step": 95570 }, { "epoch": 2.23, "learning_rate": 1.3360810588815094e-05, "loss": 0.2577, "step": 95575 }, { "epoch": 2.23, "learning_rate": 1.335993053436906e-05, "loss": 0.0879, "step": 95580 }, { "epoch": 2.23, "learning_rate": 1.3359050479923024e-05, "loss": 0.0296, "step": 95585 }, { "epoch": 2.23, "learning_rate": 1.335817042547699e-05, "loss": 0.0407, "step": 95590 }, { "epoch": 2.23, "learning_rate": 1.3357290371030954e-05, "loss": 0.0042, "step": 95595 }, { "epoch": 2.23, "learning_rate": 1.335641031658492e-05, "loss": 0.1054, "step": 95600 }, { "epoch": 2.23, "learning_rate": 1.3355530262138884e-05, "loss": 0.1248, "step": 95605 }, { "epoch": 2.23, "learning_rate": 1.3354650207692849e-05, "loss": 0.0286, "step": 95610 }, { "epoch": 2.23, "learning_rate": 1.3353770153246816e-05, "loss": 0.1323, "step": 95615 }, { "epoch": 2.23, "learning_rate": 1.3352890098800779e-05, "loss": 0.2198, "step": 95620 }, { "epoch": 2.23, "learning_rate": 1.3352010044354746e-05, "loss": 0.1798, "step": 95625 }, { "epoch": 2.23, "learning_rate": 1.3351129989908709e-05, "loss": 0.1032, "step": 95630 }, { "epoch": 2.23, "learning_rate": 1.3350249935462676e-05, "loss": 0.0744, "step": 95635 }, { "epoch": 2.23, "learning_rate": 1.3349369881016639e-05, "loss": 0.0386, "step": 95640 }, { "epoch": 2.23, "learning_rate": 1.3348489826570604e-05, "loss": 0.1039, "step": 95645 }, { "epoch": 2.23, "learning_rate": 1.3347609772124569e-05, "loss": 0.1283, "step": 95650 }, { "epoch": 2.23, "learning_rate": 1.3346729717678534e-05, "loss": 0.1548, "step": 95655 }, { "epoch": 2.23, "learning_rate": 1.3345849663232499e-05, "loss": 0.0445, "step": 95660 }, { "epoch": 2.23, "learning_rate": 1.3344969608786464e-05, "loss": 0.2019, "step": 95665 }, { "epoch": 2.23, "learning_rate": 1.334408955434043e-05, "loss": 0.2773, "step": 95670 }, { "epoch": 2.23, "learning_rate": 1.3343209499894394e-05, "loss": 0.2829, "step": 95675 }, { "epoch": 2.23, "learning_rate": 1.3342329445448358e-05, "loss": 0.0475, "step": 95680 }, { "epoch": 2.23, "learning_rate": 1.3341449391002324e-05, "loss": 0.0034, "step": 95685 }, { "epoch": 2.23, "learning_rate": 1.3340569336556288e-05, "loss": 0.0956, "step": 95690 }, { "epoch": 2.23, "learning_rate": 1.3339689282110254e-05, "loss": 0.0605, "step": 95695 }, { "epoch": 2.23, "learning_rate": 1.3338809227664218e-05, "loss": 0.0668, "step": 95700 }, { "epoch": 2.23, "learning_rate": 1.3337929173218184e-05, "loss": 0.0448, "step": 95705 }, { "epoch": 2.23, "learning_rate": 1.3337049118772148e-05, "loss": 0.0408, "step": 95710 }, { "epoch": 2.23, "learning_rate": 1.3336169064326113e-05, "loss": 0.2651, "step": 95715 }, { "epoch": 2.23, "learning_rate": 1.3335289009880078e-05, "loss": 0.0551, "step": 95720 }, { "epoch": 2.23, "learning_rate": 1.3334408955434043e-05, "loss": 0.3747, "step": 95725 }, { "epoch": 2.23, "learning_rate": 1.3333528900988008e-05, "loss": 0.0354, "step": 95730 }, { "epoch": 2.23, "learning_rate": 1.3332648846541973e-05, "loss": 0.0019, "step": 95735 }, { "epoch": 2.23, "learning_rate": 1.3331768792095938e-05, "loss": 0.2077, "step": 95740 }, { "epoch": 2.23, "learning_rate": 1.3330888737649903e-05, "loss": 0.0467, "step": 95745 }, { "epoch": 2.23, "learning_rate": 1.333000868320387e-05, "loss": 0.0246, "step": 95750 }, { "epoch": 2.23, "learning_rate": 1.3329128628757833e-05, "loss": 0.1533, "step": 95755 }, { "epoch": 2.23, "learning_rate": 1.3328248574311798e-05, "loss": 0.1664, "step": 95760 }, { "epoch": 2.23, "learning_rate": 1.3327368519865763e-05, "loss": 0.0685, "step": 95765 }, { "epoch": 2.23, "learning_rate": 1.3326488465419728e-05, "loss": 0.1554, "step": 95770 }, { "epoch": 2.23, "learning_rate": 1.3325608410973693e-05, "loss": 0.2572, "step": 95775 }, { "epoch": 2.23, "learning_rate": 1.3324728356527658e-05, "loss": 0.016, "step": 95780 }, { "epoch": 2.23, "learning_rate": 1.3323848302081623e-05, "loss": 0.0709, "step": 95785 }, { "epoch": 2.23, "learning_rate": 1.3322968247635588e-05, "loss": 0.0207, "step": 95790 }, { "epoch": 2.24, "learning_rate": 1.3322088193189551e-05, "loss": 0.0286, "step": 95795 }, { "epoch": 2.24, "learning_rate": 1.3321208138743518e-05, "loss": 0.061, "step": 95800 }, { "epoch": 2.24, "learning_rate": 1.3320328084297481e-05, "loss": 0.1133, "step": 95805 }, { "epoch": 2.24, "learning_rate": 1.3319448029851448e-05, "loss": 0.0814, "step": 95810 }, { "epoch": 2.24, "learning_rate": 1.3318567975405411e-05, "loss": 0.1619, "step": 95815 }, { "epoch": 2.24, "learning_rate": 1.3317687920959378e-05, "loss": 0.4079, "step": 95820 }, { "epoch": 2.24, "learning_rate": 1.3316807866513341e-05, "loss": 0.3941, "step": 95825 }, { "epoch": 2.24, "learning_rate": 1.3315927812067306e-05, "loss": 0.0653, "step": 95830 }, { "epoch": 2.24, "learning_rate": 1.3315047757621271e-05, "loss": 0.0404, "step": 95835 }, { "epoch": 2.24, "learning_rate": 1.3314167703175236e-05, "loss": 0.0404, "step": 95840 }, { "epoch": 2.24, "learning_rate": 1.3313287648729202e-05, "loss": 0.1059, "step": 95845 }, { "epoch": 2.24, "learning_rate": 1.3312407594283167e-05, "loss": 0.1217, "step": 95850 }, { "epoch": 2.24, "learning_rate": 1.3311527539837132e-05, "loss": 0.0524, "step": 95855 }, { "epoch": 2.24, "learning_rate": 1.3310647485391097e-05, "loss": 0.0618, "step": 95860 }, { "epoch": 2.24, "learning_rate": 1.330976743094506e-05, "loss": 0.1827, "step": 95865 }, { "epoch": 2.24, "learning_rate": 1.3308887376499027e-05, "loss": 0.2698, "step": 95870 }, { "epoch": 2.24, "learning_rate": 1.330800732205299e-05, "loss": 0.1501, "step": 95875 }, { "epoch": 2.24, "learning_rate": 1.3307127267606957e-05, "loss": 0.0409, "step": 95880 }, { "epoch": 2.24, "learning_rate": 1.3306247213160922e-05, "loss": 0.0335, "step": 95885 }, { "epoch": 2.24, "learning_rate": 1.3305367158714887e-05, "loss": 0.0645, "step": 95890 }, { "epoch": 2.24, "learning_rate": 1.3304487104268852e-05, "loss": 0.1019, "step": 95895 }, { "epoch": 2.24, "learning_rate": 1.3303607049822817e-05, "loss": 0.026, "step": 95900 }, { "epoch": 2.24, "learning_rate": 1.3302726995376782e-05, "loss": 0.049, "step": 95905 }, { "epoch": 2.24, "learning_rate": 1.3301846940930745e-05, "loss": 0.0609, "step": 95910 }, { "epoch": 2.24, "learning_rate": 1.3300966886484712e-05, "loss": 0.1311, "step": 95915 }, { "epoch": 2.24, "learning_rate": 1.3300086832038675e-05, "loss": 0.174, "step": 95920 }, { "epoch": 2.24, "learning_rate": 1.3299206777592642e-05, "loss": 0.2402, "step": 95925 }, { "epoch": 2.24, "learning_rate": 1.3298326723146605e-05, "loss": 0.0046, "step": 95930 }, { "epoch": 2.24, "learning_rate": 1.3297446668700572e-05, "loss": 0.0129, "step": 95935 }, { "epoch": 2.24, "learning_rate": 1.3296566614254535e-05, "loss": 0.0406, "step": 95940 }, { "epoch": 2.24, "learning_rate": 1.32956865598085e-05, "loss": 0.0208, "step": 95945 }, { "epoch": 2.24, "learning_rate": 1.3294806505362465e-05, "loss": 0.0997, "step": 95950 }, { "epoch": 2.24, "learning_rate": 1.329392645091643e-05, "loss": 0.0555, "step": 95955 }, { "epoch": 2.24, "learning_rate": 1.3293046396470395e-05, "loss": 0.1214, "step": 95960 }, { "epoch": 2.24, "learning_rate": 1.329216634202436e-05, "loss": 0.229, "step": 95965 }, { "epoch": 2.24, "learning_rate": 1.3291286287578325e-05, "loss": 0.3524, "step": 95970 }, { "epoch": 2.24, "learning_rate": 1.329040623313229e-05, "loss": 0.3234, "step": 95975 }, { "epoch": 2.24, "learning_rate": 1.3289526178686254e-05, "loss": 0.0362, "step": 95980 }, { "epoch": 2.24, "learning_rate": 1.328864612424022e-05, "loss": 0.0585, "step": 95985 }, { "epoch": 2.24, "learning_rate": 1.3287766069794184e-05, "loss": 0.1254, "step": 95990 }, { "epoch": 2.24, "learning_rate": 1.328688601534815e-05, "loss": 0.0688, "step": 95995 }, { "epoch": 2.24, "learning_rate": 1.3286005960902114e-05, "loss": 0.0455, "step": 96000 }, { "epoch": 2.24, "learning_rate": 1.328512590645608e-05, "loss": 0.079, "step": 96005 }, { "epoch": 2.24, "learning_rate": 1.3284245852010044e-05, "loss": 0.1778, "step": 96010 }, { "epoch": 2.24, "learning_rate": 1.3283365797564009e-05, "loss": 0.1149, "step": 96015 }, { "epoch": 2.24, "learning_rate": 1.3282485743117976e-05, "loss": 0.1931, "step": 96020 }, { "epoch": 2.24, "learning_rate": 1.3281605688671939e-05, "loss": 0.3909, "step": 96025 }, { "epoch": 2.24, "learning_rate": 1.3280725634225906e-05, "loss": 0.0847, "step": 96030 }, { "epoch": 2.24, "learning_rate": 1.3279845579779869e-05, "loss": 0.045, "step": 96035 }, { "epoch": 2.24, "learning_rate": 1.3278965525333836e-05, "loss": 0.0706, "step": 96040 }, { "epoch": 2.24, "learning_rate": 1.3278085470887799e-05, "loss": 0.0574, "step": 96045 }, { "epoch": 2.24, "learning_rate": 1.3277205416441766e-05, "loss": 0.1217, "step": 96050 }, { "epoch": 2.24, "learning_rate": 1.3276325361995729e-05, "loss": 0.1009, "step": 96055 }, { "epoch": 2.24, "learning_rate": 1.3275445307549694e-05, "loss": 0.1514, "step": 96060 }, { "epoch": 2.24, "learning_rate": 1.3274565253103659e-05, "loss": 0.117, "step": 96065 }, { "epoch": 2.24, "learning_rate": 1.3273685198657624e-05, "loss": 0.0907, "step": 96070 }, { "epoch": 2.24, "learning_rate": 1.3272805144211589e-05, "loss": 0.2497, "step": 96075 }, { "epoch": 2.24, "learning_rate": 1.3271925089765554e-05, "loss": 0.0772, "step": 96080 }, { "epoch": 2.24, "learning_rate": 1.3271045035319519e-05, "loss": 0.0273, "step": 96085 }, { "epoch": 2.24, "learning_rate": 1.3270164980873484e-05, "loss": 0.0648, "step": 96090 }, { "epoch": 2.24, "learning_rate": 1.3269284926427447e-05, "loss": 0.0587, "step": 96095 }, { "epoch": 2.24, "learning_rate": 1.3268404871981414e-05, "loss": 0.0664, "step": 96100 }, { "epoch": 2.24, "learning_rate": 1.3267524817535377e-05, "loss": 0.2472, "step": 96105 }, { "epoch": 2.24, "learning_rate": 1.3266644763089344e-05, "loss": 0.127, "step": 96110 }, { "epoch": 2.24, "learning_rate": 1.3265764708643308e-05, "loss": 0.2063, "step": 96115 }, { "epoch": 2.24, "learning_rate": 1.3264884654197274e-05, "loss": 0.2211, "step": 96120 }, { "epoch": 2.24, "learning_rate": 1.3264004599751238e-05, "loss": 0.1934, "step": 96125 }, { "epoch": 2.24, "learning_rate": 1.3263124545305203e-05, "loss": 0.1429, "step": 96130 }, { "epoch": 2.24, "learning_rate": 1.3262244490859168e-05, "loss": 0.0393, "step": 96135 }, { "epoch": 2.24, "learning_rate": 1.3261364436413133e-05, "loss": 0.0511, "step": 96140 }, { "epoch": 2.24, "learning_rate": 1.3260484381967098e-05, "loss": 0.0602, "step": 96145 }, { "epoch": 2.24, "learning_rate": 1.3259604327521063e-05, "loss": 0.0271, "step": 96150 }, { "epoch": 2.24, "learning_rate": 1.3258724273075028e-05, "loss": 0.1192, "step": 96155 }, { "epoch": 2.24, "learning_rate": 1.3257844218628993e-05, "loss": 0.1399, "step": 96160 }, { "epoch": 2.24, "learning_rate": 1.3256964164182958e-05, "loss": 0.0689, "step": 96165 }, { "epoch": 2.24, "learning_rate": 1.3256084109736923e-05, "loss": 0.21, "step": 96170 }, { "epoch": 2.24, "learning_rate": 1.3255204055290888e-05, "loss": 0.291, "step": 96175 }, { "epoch": 2.24, "learning_rate": 1.3254324000844853e-05, "loss": 0.0193, "step": 96180 }, { "epoch": 2.24, "learning_rate": 1.3253443946398818e-05, "loss": 0.0149, "step": 96185 }, { "epoch": 2.24, "learning_rate": 1.3252563891952783e-05, "loss": 0.0495, "step": 96190 }, { "epoch": 2.24, "learning_rate": 1.3251683837506748e-05, "loss": 0.1352, "step": 96195 }, { "epoch": 2.24, "learning_rate": 1.3250803783060713e-05, "loss": 0.1064, "step": 96200 }, { "epoch": 2.24, "learning_rate": 1.3249923728614678e-05, "loss": 0.0889, "step": 96205 }, { "epoch": 2.24, "learning_rate": 1.3249043674168641e-05, "loss": 0.0772, "step": 96210 }, { "epoch": 2.24, "learning_rate": 1.3248163619722608e-05, "loss": 0.2094, "step": 96215 }, { "epoch": 2.24, "learning_rate": 1.3247283565276571e-05, "loss": 0.1026, "step": 96220 }, { "epoch": 2.25, "learning_rate": 1.3246403510830538e-05, "loss": 0.1801, "step": 96225 }, { "epoch": 2.25, "learning_rate": 1.3245523456384501e-05, "loss": 0.0918, "step": 96230 }, { "epoch": 2.25, "learning_rate": 1.3244643401938468e-05, "loss": 0.0246, "step": 96235 }, { "epoch": 2.25, "learning_rate": 1.3243763347492431e-05, "loss": 0.0797, "step": 96240 }, { "epoch": 2.25, "learning_rate": 1.3242883293046396e-05, "loss": 0.1084, "step": 96245 }, { "epoch": 2.25, "learning_rate": 1.3242003238600361e-05, "loss": 0.0107, "step": 96250 }, { "epoch": 2.25, "learning_rate": 1.3241123184154326e-05, "loss": 0.0617, "step": 96255 }, { "epoch": 2.25, "learning_rate": 1.3240243129708291e-05, "loss": 0.0968, "step": 96260 }, { "epoch": 2.25, "learning_rate": 1.3239363075262256e-05, "loss": 0.054, "step": 96265 }, { "epoch": 2.25, "learning_rate": 1.3238483020816221e-05, "loss": 0.0669, "step": 96270 }, { "epoch": 2.25, "learning_rate": 1.3237602966370187e-05, "loss": 0.2048, "step": 96275 }, { "epoch": 2.25, "learning_rate": 1.323672291192415e-05, "loss": 0.0544, "step": 96280 }, { "epoch": 2.25, "learning_rate": 1.3235842857478117e-05, "loss": 0.04, "step": 96285 }, { "epoch": 2.25, "learning_rate": 1.3234962803032082e-05, "loss": 0.0512, "step": 96290 }, { "epoch": 2.25, "learning_rate": 1.3234082748586047e-05, "loss": 0.013, "step": 96295 }, { "epoch": 2.25, "learning_rate": 1.3233202694140012e-05, "loss": 0.0886, "step": 96300 }, { "epoch": 2.25, "learning_rate": 1.3232322639693977e-05, "loss": 0.0483, "step": 96305 }, { "epoch": 2.25, "learning_rate": 1.3231442585247942e-05, "loss": 0.1552, "step": 96310 }, { "epoch": 2.25, "learning_rate": 1.3230562530801905e-05, "loss": 0.1182, "step": 96315 }, { "epoch": 2.25, "learning_rate": 1.3229682476355872e-05, "loss": 0.095, "step": 96320 }, { "epoch": 2.25, "learning_rate": 1.3228802421909835e-05, "loss": 0.1006, "step": 96325 }, { "epoch": 2.25, "learning_rate": 1.3227922367463802e-05, "loss": 0.0738, "step": 96330 }, { "epoch": 2.25, "learning_rate": 1.3227042313017765e-05, "loss": 0.0732, "step": 96335 }, { "epoch": 2.25, "learning_rate": 1.3226162258571732e-05, "loss": 0.0433, "step": 96340 }, { "epoch": 2.25, "learning_rate": 1.3225282204125695e-05, "loss": 0.1422, "step": 96345 }, { "epoch": 2.25, "learning_rate": 1.3224402149679662e-05, "loss": 0.0375, "step": 96350 }, { "epoch": 2.25, "learning_rate": 1.3223522095233625e-05, "loss": 0.1088, "step": 96355 }, { "epoch": 2.25, "learning_rate": 1.322264204078759e-05, "loss": 0.0951, "step": 96360 }, { "epoch": 2.25, "learning_rate": 1.3221761986341555e-05, "loss": 0.1429, "step": 96365 }, { "epoch": 2.25, "learning_rate": 1.322088193189552e-05, "loss": 0.1431, "step": 96370 }, { "epoch": 2.25, "learning_rate": 1.3220001877449485e-05, "loss": 0.2513, "step": 96375 }, { "epoch": 2.25, "learning_rate": 1.321912182300345e-05, "loss": 0.0622, "step": 96380 }, { "epoch": 2.25, "learning_rate": 1.3218241768557415e-05, "loss": 0.0444, "step": 96385 }, { "epoch": 2.25, "learning_rate": 1.321736171411138e-05, "loss": 0.0317, "step": 96390 }, { "epoch": 2.25, "learning_rate": 1.3216481659665344e-05, "loss": 0.0313, "step": 96395 }, { "epoch": 2.25, "learning_rate": 1.321560160521931e-05, "loss": 0.1023, "step": 96400 }, { "epoch": 2.25, "learning_rate": 1.3214721550773274e-05, "loss": 0.0603, "step": 96405 }, { "epoch": 2.25, "learning_rate": 1.321384149632724e-05, "loss": 0.1188, "step": 96410 }, { "epoch": 2.25, "learning_rate": 1.3212961441881204e-05, "loss": 0.0167, "step": 96415 }, { "epoch": 2.25, "learning_rate": 1.321208138743517e-05, "loss": 0.0539, "step": 96420 }, { "epoch": 2.25, "learning_rate": 1.3211201332989134e-05, "loss": 0.3057, "step": 96425 }, { "epoch": 2.25, "learning_rate": 1.3210321278543099e-05, "loss": 0.0383, "step": 96430 }, { "epoch": 2.25, "learning_rate": 1.3209441224097065e-05, "loss": 0.1011, "step": 96435 }, { "epoch": 2.25, "learning_rate": 1.3208561169651029e-05, "loss": 0.0123, "step": 96440 }, { "epoch": 2.25, "learning_rate": 1.3207681115204996e-05, "loss": 0.0029, "step": 96445 }, { "epoch": 2.25, "learning_rate": 1.3206801060758959e-05, "loss": 0.0804, "step": 96450 }, { "epoch": 2.25, "learning_rate": 1.3205921006312926e-05, "loss": 0.1244, "step": 96455 }, { "epoch": 2.25, "learning_rate": 1.3205040951866889e-05, "loss": 0.0734, "step": 96460 }, { "epoch": 2.25, "learning_rate": 1.3204160897420854e-05, "loss": 0.1701, "step": 96465 }, { "epoch": 2.25, "learning_rate": 1.3203280842974819e-05, "loss": 0.0058, "step": 96470 }, { "epoch": 2.25, "learning_rate": 1.3202400788528784e-05, "loss": 0.2317, "step": 96475 }, { "epoch": 2.25, "learning_rate": 1.3201520734082749e-05, "loss": 0.0123, "step": 96480 }, { "epoch": 2.25, "learning_rate": 1.3200640679636714e-05, "loss": 0.1025, "step": 96485 }, { "epoch": 2.25, "learning_rate": 1.3199760625190679e-05, "loss": 0.0437, "step": 96490 }, { "epoch": 2.25, "learning_rate": 1.3198880570744644e-05, "loss": 0.055, "step": 96495 }, { "epoch": 2.25, "learning_rate": 1.3198000516298609e-05, "loss": 0.0232, "step": 96500 }, { "epoch": 2.25, "learning_rate": 1.3197120461852574e-05, "loss": 0.1436, "step": 96505 }, { "epoch": 2.25, "learning_rate": 1.3196240407406537e-05, "loss": 0.1044, "step": 96510 }, { "epoch": 2.25, "learning_rate": 1.3195360352960504e-05, "loss": 0.2478, "step": 96515 }, { "epoch": 2.25, "learning_rate": 1.3194480298514467e-05, "loss": 0.2102, "step": 96520 }, { "epoch": 2.25, "learning_rate": 1.3193600244068434e-05, "loss": 0.3935, "step": 96525 }, { "epoch": 2.25, "learning_rate": 1.3192720189622397e-05, "loss": 0.1286, "step": 96530 }, { "epoch": 2.25, "learning_rate": 1.3191840135176364e-05, "loss": 0.0782, "step": 96535 }, { "epoch": 2.25, "learning_rate": 1.3190960080730328e-05, "loss": 0.0013, "step": 96540 }, { "epoch": 2.25, "learning_rate": 1.3190080026284293e-05, "loss": 0.0937, "step": 96545 }, { "epoch": 2.25, "learning_rate": 1.3189199971838258e-05, "loss": 0.0979, "step": 96550 }, { "epoch": 2.25, "learning_rate": 1.3188319917392223e-05, "loss": 0.0816, "step": 96555 }, { "epoch": 2.25, "learning_rate": 1.3187439862946188e-05, "loss": 0.1349, "step": 96560 }, { "epoch": 2.25, "learning_rate": 1.3186559808500153e-05, "loss": 0.0953, "step": 96565 }, { "epoch": 2.25, "learning_rate": 1.318567975405412e-05, "loss": 0.2427, "step": 96570 }, { "epoch": 2.25, "learning_rate": 1.3184799699608083e-05, "loss": 0.2303, "step": 96575 }, { "epoch": 2.25, "learning_rate": 1.3183919645162048e-05, "loss": 0.0206, "step": 96580 }, { "epoch": 2.25, "learning_rate": 1.3183039590716013e-05, "loss": 0.0382, "step": 96585 }, { "epoch": 2.25, "learning_rate": 1.3182159536269978e-05, "loss": 0.091, "step": 96590 }, { "epoch": 2.25, "learning_rate": 1.3181279481823943e-05, "loss": 0.076, "step": 96595 }, { "epoch": 2.25, "learning_rate": 1.3180399427377908e-05, "loss": 0.0144, "step": 96600 }, { "epoch": 2.25, "learning_rate": 1.3179519372931873e-05, "loss": 0.1069, "step": 96605 }, { "epoch": 2.25, "learning_rate": 1.3178639318485838e-05, "loss": 0.1, "step": 96610 }, { "epoch": 2.25, "learning_rate": 1.3177759264039801e-05, "loss": 0.0745, "step": 96615 }, { "epoch": 2.25, "learning_rate": 1.3176879209593768e-05, "loss": 0.3267, "step": 96620 }, { "epoch": 2.25, "learning_rate": 1.3175999155147731e-05, "loss": 0.304, "step": 96625 }, { "epoch": 2.25, "learning_rate": 1.3175119100701698e-05, "loss": 0.0114, "step": 96630 }, { "epoch": 2.25, "learning_rate": 1.3174239046255661e-05, "loss": 0.0005, "step": 96635 }, { "epoch": 2.25, "learning_rate": 1.3173358991809628e-05, "loss": 0.0643, "step": 96640 }, { "epoch": 2.25, "learning_rate": 1.3172478937363591e-05, "loss": 0.0242, "step": 96645 }, { "epoch": 2.25, "learning_rate": 1.3171598882917558e-05, "loss": 0.0902, "step": 96650 }, { "epoch": 2.26, "learning_rate": 1.3170718828471521e-05, "loss": 0.1114, "step": 96655 }, { "epoch": 2.26, "learning_rate": 1.3169838774025486e-05, "loss": 0.167, "step": 96660 }, { "epoch": 2.26, "learning_rate": 1.3168958719579451e-05, "loss": 0.1483, "step": 96665 }, { "epoch": 2.26, "learning_rate": 1.3168078665133416e-05, "loss": 0.277, "step": 96670 }, { "epoch": 2.26, "learning_rate": 1.3167198610687381e-05, "loss": 0.5273, "step": 96675 }, { "epoch": 2.26, "learning_rate": 1.3166318556241346e-05, "loss": 0.1072, "step": 96680 }, { "epoch": 2.26, "learning_rate": 1.3165438501795311e-05, "loss": 0.0265, "step": 96685 }, { "epoch": 2.26, "learning_rate": 1.3164558447349276e-05, "loss": 0.1724, "step": 96690 }, { "epoch": 2.26, "learning_rate": 1.316367839290324e-05, "loss": 0.1045, "step": 96695 }, { "epoch": 2.26, "learning_rate": 1.3162798338457206e-05, "loss": 0.0577, "step": 96700 }, { "epoch": 2.26, "learning_rate": 1.3161918284011171e-05, "loss": 0.165, "step": 96705 }, { "epoch": 2.26, "learning_rate": 1.3161038229565137e-05, "loss": 0.154, "step": 96710 }, { "epoch": 2.26, "learning_rate": 1.3160158175119102e-05, "loss": 0.1018, "step": 96715 }, { "epoch": 2.26, "learning_rate": 1.3159278120673067e-05, "loss": 0.1508, "step": 96720 }, { "epoch": 2.26, "learning_rate": 1.3158398066227032e-05, "loss": 0.1937, "step": 96725 }, { "epoch": 2.26, "learning_rate": 1.3157518011780995e-05, "loss": 0.0122, "step": 96730 }, { "epoch": 2.26, "learning_rate": 1.3156637957334962e-05, "loss": 0.0493, "step": 96735 }, { "epoch": 2.26, "learning_rate": 1.3155757902888925e-05, "loss": 0.0794, "step": 96740 }, { "epoch": 2.26, "learning_rate": 1.3154877848442892e-05, "loss": 0.0955, "step": 96745 }, { "epoch": 2.26, "learning_rate": 1.3153997793996855e-05, "loss": 0.048, "step": 96750 }, { "epoch": 2.26, "learning_rate": 1.3153117739550822e-05, "loss": 0.0541, "step": 96755 }, { "epoch": 2.26, "learning_rate": 1.3152237685104785e-05, "loss": 0.1009, "step": 96760 }, { "epoch": 2.26, "learning_rate": 1.315135763065875e-05, "loss": 0.2069, "step": 96765 }, { "epoch": 2.26, "learning_rate": 1.3150477576212715e-05, "loss": 0.141, "step": 96770 }, { "epoch": 2.26, "learning_rate": 1.314959752176668e-05, "loss": 0.2072, "step": 96775 }, { "epoch": 2.26, "learning_rate": 1.3148717467320645e-05, "loss": 0.0852, "step": 96780 }, { "epoch": 2.26, "learning_rate": 1.314783741287461e-05, "loss": 0.0897, "step": 96785 }, { "epoch": 2.26, "learning_rate": 1.3146957358428575e-05, "loss": 0.1014, "step": 96790 }, { "epoch": 2.26, "learning_rate": 1.314607730398254e-05, "loss": 0.0023, "step": 96795 }, { "epoch": 2.26, "learning_rate": 1.3145197249536505e-05, "loss": 0.1035, "step": 96800 }, { "epoch": 2.26, "learning_rate": 1.314431719509047e-05, "loss": 0.0987, "step": 96805 }, { "epoch": 2.26, "learning_rate": 1.3143437140644434e-05, "loss": 0.0542, "step": 96810 }, { "epoch": 2.26, "learning_rate": 1.31425570861984e-05, "loss": 0.0544, "step": 96815 }, { "epoch": 2.26, "learning_rate": 1.3141677031752364e-05, "loss": 0.2782, "step": 96820 }, { "epoch": 2.26, "learning_rate": 1.314079697730633e-05, "loss": 0.3358, "step": 96825 }, { "epoch": 2.26, "learning_rate": 1.3139916922860294e-05, "loss": 0.0895, "step": 96830 }, { "epoch": 2.26, "learning_rate": 1.313903686841426e-05, "loss": 0.0331, "step": 96835 }, { "epoch": 2.26, "learning_rate": 1.3138156813968225e-05, "loss": 0.0369, "step": 96840 }, { "epoch": 2.26, "learning_rate": 1.3137276759522189e-05, "loss": 0.1508, "step": 96845 }, { "epoch": 2.26, "learning_rate": 1.3136396705076155e-05, "loss": 0.1073, "step": 96850 }, { "epoch": 2.26, "learning_rate": 1.3135516650630119e-05, "loss": 0.0955, "step": 96855 }, { "epoch": 2.26, "learning_rate": 1.3134636596184085e-05, "loss": 0.1056, "step": 96860 }, { "epoch": 2.26, "learning_rate": 1.3133756541738049e-05, "loss": 0.119, "step": 96865 }, { "epoch": 2.26, "learning_rate": 1.3132876487292015e-05, "loss": 0.1447, "step": 96870 }, { "epoch": 2.26, "learning_rate": 1.3131996432845979e-05, "loss": 0.2385, "step": 96875 }, { "epoch": 2.26, "learning_rate": 1.3131116378399944e-05, "loss": 0.0779, "step": 96880 }, { "epoch": 2.26, "learning_rate": 1.3130236323953909e-05, "loss": 0.0835, "step": 96885 }, { "epoch": 2.26, "learning_rate": 1.3129356269507874e-05, "loss": 0.0017, "step": 96890 }, { "epoch": 2.26, "learning_rate": 1.3128476215061839e-05, "loss": 0.0363, "step": 96895 }, { "epoch": 2.26, "learning_rate": 1.3127596160615804e-05, "loss": 0.0304, "step": 96900 }, { "epoch": 2.26, "learning_rate": 1.3126716106169769e-05, "loss": 0.0505, "step": 96905 }, { "epoch": 2.26, "learning_rate": 1.3125836051723734e-05, "loss": 0.1148, "step": 96910 }, { "epoch": 2.26, "learning_rate": 1.3124955997277697e-05, "loss": 0.115, "step": 96915 }, { "epoch": 2.26, "learning_rate": 1.3124075942831664e-05, "loss": 0.1084, "step": 96920 }, { "epoch": 2.26, "learning_rate": 1.3123195888385627e-05, "loss": 0.2462, "step": 96925 }, { "epoch": 2.26, "learning_rate": 1.3122315833939594e-05, "loss": 0.0778, "step": 96930 }, { "epoch": 2.26, "learning_rate": 1.3121435779493557e-05, "loss": 0.0358, "step": 96935 }, { "epoch": 2.26, "learning_rate": 1.3120555725047524e-05, "loss": 0.0608, "step": 96940 }, { "epoch": 2.26, "learning_rate": 1.3119675670601487e-05, "loss": 0.1111, "step": 96945 }, { "epoch": 2.26, "learning_rate": 1.3118795616155454e-05, "loss": 0.0683, "step": 96950 }, { "epoch": 2.26, "learning_rate": 1.3117915561709417e-05, "loss": 0.0805, "step": 96955 }, { "epoch": 2.26, "learning_rate": 1.3117035507263382e-05, "loss": 0.1381, "step": 96960 }, { "epoch": 2.26, "learning_rate": 1.3116155452817347e-05, "loss": 0.0553, "step": 96965 }, { "epoch": 2.26, "learning_rate": 1.3115275398371312e-05, "loss": 0.0717, "step": 96970 }, { "epoch": 2.26, "learning_rate": 1.311439534392528e-05, "loss": 0.369, "step": 96975 }, { "epoch": 2.26, "learning_rate": 1.3113515289479243e-05, "loss": 0.0458, "step": 96980 }, { "epoch": 2.26, "learning_rate": 1.311263523503321e-05, "loss": 0.023, "step": 96985 }, { "epoch": 2.26, "learning_rate": 1.3111755180587173e-05, "loss": 0.1055, "step": 96990 }, { "epoch": 2.26, "learning_rate": 1.3110875126141138e-05, "loss": 0.089, "step": 96995 }, { "epoch": 2.26, "learning_rate": 1.3109995071695103e-05, "loss": 0.045, "step": 97000 }, { "epoch": 2.26, "learning_rate": 1.3109115017249068e-05, "loss": 0.0939, "step": 97005 }, { "epoch": 2.26, "learning_rate": 1.3108234962803033e-05, "loss": 0.1439, "step": 97010 }, { "epoch": 2.26, "learning_rate": 1.3107354908356998e-05, "loss": 0.1671, "step": 97015 }, { "epoch": 2.26, "learning_rate": 1.3106474853910963e-05, "loss": 0.2415, "step": 97020 }, { "epoch": 2.26, "learning_rate": 1.3105594799464928e-05, "loss": 0.2841, "step": 97025 }, { "epoch": 2.26, "learning_rate": 1.3104714745018891e-05, "loss": 0.0545, "step": 97030 }, { "epoch": 2.26, "learning_rate": 1.3103834690572858e-05, "loss": 0.0273, "step": 97035 }, { "epoch": 2.26, "learning_rate": 1.3102954636126821e-05, "loss": 0.0342, "step": 97040 }, { "epoch": 2.26, "learning_rate": 1.3102074581680788e-05, "loss": 0.1507, "step": 97045 }, { "epoch": 2.26, "learning_rate": 1.3101194527234751e-05, "loss": 0.0252, "step": 97050 }, { "epoch": 2.26, "learning_rate": 1.3100314472788718e-05, "loss": 0.1525, "step": 97055 }, { "epoch": 2.26, "learning_rate": 1.3099434418342681e-05, "loss": 0.1059, "step": 97060 }, { "epoch": 2.26, "learning_rate": 1.3098554363896646e-05, "loss": 0.1586, "step": 97065 }, { "epoch": 2.26, "learning_rate": 1.3097674309450611e-05, "loss": 0.2367, "step": 97070 }, { "epoch": 2.26, "learning_rate": 1.3096794255004576e-05, "loss": 0.2169, "step": 97075 }, { "epoch": 2.26, "learning_rate": 1.3095914200558541e-05, "loss": 0.0178, "step": 97080 }, { "epoch": 2.27, "learning_rate": 1.3095034146112506e-05, "loss": 0.0582, "step": 97085 }, { "epoch": 2.27, "learning_rate": 1.3094154091666471e-05, "loss": 0.0125, "step": 97090 }, { "epoch": 2.27, "learning_rate": 1.3093274037220436e-05, "loss": 0.0368, "step": 97095 }, { "epoch": 2.27, "learning_rate": 1.3092393982774401e-05, "loss": 0.0739, "step": 97100 }, { "epoch": 2.27, "learning_rate": 1.3091513928328366e-05, "loss": 0.1461, "step": 97105 }, { "epoch": 2.27, "learning_rate": 1.3090633873882331e-05, "loss": 0.0669, "step": 97110 }, { "epoch": 2.27, "learning_rate": 1.3089753819436296e-05, "loss": 0.1013, "step": 97115 }, { "epoch": 2.27, "learning_rate": 1.3088873764990261e-05, "loss": 0.2562, "step": 97120 }, { "epoch": 2.27, "learning_rate": 1.3087993710544226e-05, "loss": 0.287, "step": 97125 }, { "epoch": 2.27, "learning_rate": 1.3087113656098191e-05, "loss": 0.0766, "step": 97130 }, { "epoch": 2.27, "learning_rate": 1.3086233601652156e-05, "loss": 0.008, "step": 97135 }, { "epoch": 2.27, "learning_rate": 1.3085353547206122e-05, "loss": 0.0995, "step": 97140 }, { "epoch": 2.27, "learning_rate": 1.3084473492760085e-05, "loss": 0.0724, "step": 97145 }, { "epoch": 2.27, "learning_rate": 1.3083593438314052e-05, "loss": 0.0428, "step": 97150 }, { "epoch": 2.27, "learning_rate": 1.3082713383868015e-05, "loss": 0.1458, "step": 97155 }, { "epoch": 2.27, "learning_rate": 1.3081833329421982e-05, "loss": 0.1567, "step": 97160 }, { "epoch": 2.27, "learning_rate": 1.3080953274975945e-05, "loss": 0.1467, "step": 97165 }, { "epoch": 2.27, "learning_rate": 1.3080073220529912e-05, "loss": 0.25, "step": 97170 }, { "epoch": 2.27, "learning_rate": 1.3079193166083875e-05, "loss": 0.2336, "step": 97175 }, { "epoch": 2.27, "learning_rate": 1.307831311163784e-05, "loss": 0.1068, "step": 97180 }, { "epoch": 2.27, "learning_rate": 1.3077433057191805e-05, "loss": 0.0038, "step": 97185 }, { "epoch": 2.27, "learning_rate": 1.307655300274577e-05, "loss": 0.0781, "step": 97190 }, { "epoch": 2.27, "learning_rate": 1.3075672948299735e-05, "loss": 0.0492, "step": 97195 }, { "epoch": 2.27, "learning_rate": 1.30747928938537e-05, "loss": 0.1408, "step": 97200 }, { "epoch": 2.27, "learning_rate": 1.3073912839407665e-05, "loss": 0.076, "step": 97205 }, { "epoch": 2.27, "learning_rate": 1.307303278496163e-05, "loss": 0.1071, "step": 97210 }, { "epoch": 2.27, "learning_rate": 1.3072152730515593e-05, "loss": 0.1558, "step": 97215 }, { "epoch": 2.27, "learning_rate": 1.307127267606956e-05, "loss": 0.2488, "step": 97220 }, { "epoch": 2.27, "learning_rate": 1.3070392621623523e-05, "loss": 0.2309, "step": 97225 }, { "epoch": 2.27, "learning_rate": 1.306951256717749e-05, "loss": 0.06, "step": 97230 }, { "epoch": 2.27, "learning_rate": 1.3068632512731454e-05, "loss": 0.0159, "step": 97235 }, { "epoch": 2.27, "learning_rate": 1.306775245828542e-05, "loss": 0.0793, "step": 97240 }, { "epoch": 2.27, "learning_rate": 1.3066872403839385e-05, "loss": 0.0719, "step": 97245 }, { "epoch": 2.27, "learning_rate": 1.3065992349393349e-05, "loss": 0.0646, "step": 97250 }, { "epoch": 2.27, "learning_rate": 1.3065112294947315e-05, "loss": 0.215, "step": 97255 }, { "epoch": 2.27, "learning_rate": 1.3064232240501279e-05, "loss": 0.0763, "step": 97260 }, { "epoch": 2.27, "learning_rate": 1.3063352186055245e-05, "loss": 0.0651, "step": 97265 }, { "epoch": 2.27, "learning_rate": 1.3062472131609209e-05, "loss": 0.1695, "step": 97270 }, { "epoch": 2.27, "learning_rate": 1.3061592077163175e-05, "loss": 0.3951, "step": 97275 }, { "epoch": 2.27, "learning_rate": 1.3060712022717139e-05, "loss": 0.0386, "step": 97280 }, { "epoch": 2.27, "learning_rate": 1.3059831968271105e-05, "loss": 0.0226, "step": 97285 }, { "epoch": 2.27, "learning_rate": 1.3058951913825069e-05, "loss": 0.1601, "step": 97290 }, { "epoch": 2.27, "learning_rate": 1.3058071859379034e-05, "loss": 0.0545, "step": 97295 }, { "epoch": 2.27, "learning_rate": 1.3057191804932999e-05, "loss": 0.0637, "step": 97300 }, { "epoch": 2.27, "learning_rate": 1.3056311750486964e-05, "loss": 0.0404, "step": 97305 }, { "epoch": 2.27, "learning_rate": 1.3055431696040929e-05, "loss": 0.0928, "step": 97310 }, { "epoch": 2.27, "learning_rate": 1.3054551641594894e-05, "loss": 0.2239, "step": 97315 }, { "epoch": 2.27, "learning_rate": 1.3053671587148859e-05, "loss": 0.3322, "step": 97320 }, { "epoch": 2.27, "learning_rate": 1.3052791532702824e-05, "loss": 0.3495, "step": 97325 }, { "epoch": 2.27, "learning_rate": 1.3051911478256787e-05, "loss": 0.0761, "step": 97330 }, { "epoch": 2.27, "learning_rate": 1.3051031423810754e-05, "loss": 0.0214, "step": 97335 }, { "epoch": 2.27, "learning_rate": 1.3050151369364717e-05, "loss": 0.0465, "step": 97340 }, { "epoch": 2.27, "learning_rate": 1.3049271314918684e-05, "loss": 0.025, "step": 97345 }, { "epoch": 2.27, "learning_rate": 1.3048391260472647e-05, "loss": 0.0331, "step": 97350 }, { "epoch": 2.27, "learning_rate": 1.3047511206026614e-05, "loss": 0.2081, "step": 97355 }, { "epoch": 2.27, "learning_rate": 1.3046631151580577e-05, "loss": 0.2161, "step": 97360 }, { "epoch": 2.27, "learning_rate": 1.3045751097134542e-05, "loss": 0.15, "step": 97365 }, { "epoch": 2.27, "learning_rate": 1.3044871042688507e-05, "loss": 0.1338, "step": 97370 }, { "epoch": 2.27, "learning_rate": 1.3043990988242472e-05, "loss": 0.3805, "step": 97375 }, { "epoch": 2.27, "learning_rate": 1.3043110933796439e-05, "loss": 0.0322, "step": 97380 }, { "epoch": 2.27, "learning_rate": 1.3042230879350402e-05, "loss": 0.0481, "step": 97385 }, { "epoch": 2.27, "learning_rate": 1.3041350824904369e-05, "loss": 0.0668, "step": 97390 }, { "epoch": 2.27, "learning_rate": 1.3040470770458332e-05, "loss": 0.0767, "step": 97395 }, { "epoch": 2.27, "learning_rate": 1.3039590716012297e-05, "loss": 0.0716, "step": 97400 }, { "epoch": 2.27, "learning_rate": 1.3038710661566263e-05, "loss": 0.0938, "step": 97405 }, { "epoch": 2.27, "learning_rate": 1.3037830607120228e-05, "loss": 0.1989, "step": 97410 }, { "epoch": 2.27, "learning_rate": 1.3036950552674193e-05, "loss": 0.2002, "step": 97415 }, { "epoch": 2.27, "learning_rate": 1.3036070498228158e-05, "loss": 0.0997, "step": 97420 }, { "epoch": 2.27, "learning_rate": 1.3035190443782123e-05, "loss": 0.2253, "step": 97425 }, { "epoch": 2.27, "learning_rate": 1.3034310389336088e-05, "loss": 0.0638, "step": 97430 }, { "epoch": 2.27, "learning_rate": 1.3033430334890053e-05, "loss": 0.004, "step": 97435 }, { "epoch": 2.27, "learning_rate": 1.3032550280444018e-05, "loss": 0.0873, "step": 97440 }, { "epoch": 2.27, "learning_rate": 1.3031670225997981e-05, "loss": 0.0283, "step": 97445 }, { "epoch": 2.27, "learning_rate": 1.3030790171551948e-05, "loss": 0.0174, "step": 97450 }, { "epoch": 2.27, "learning_rate": 1.3029910117105911e-05, "loss": 0.1778, "step": 97455 }, { "epoch": 2.27, "learning_rate": 1.3029030062659878e-05, "loss": 0.0582, "step": 97460 }, { "epoch": 2.27, "learning_rate": 1.3028150008213841e-05, "loss": 0.1641, "step": 97465 }, { "epoch": 2.27, "learning_rate": 1.3027269953767808e-05, "loss": 0.2345, "step": 97470 }, { "epoch": 2.27, "learning_rate": 1.3026389899321771e-05, "loss": 0.2193, "step": 97475 }, { "epoch": 2.27, "learning_rate": 1.3025509844875736e-05, "loss": 0.0711, "step": 97480 }, { "epoch": 2.27, "learning_rate": 1.3024629790429701e-05, "loss": 0.0278, "step": 97485 }, { "epoch": 2.27, "learning_rate": 1.3023749735983666e-05, "loss": 0.0219, "step": 97490 }, { "epoch": 2.27, "learning_rate": 1.3022869681537631e-05, "loss": 0.0782, "step": 97495 }, { "epoch": 2.27, "learning_rate": 1.3021989627091596e-05, "loss": 0.1269, "step": 97500 }, { "epoch": 2.27, "learning_rate": 1.3021109572645561e-05, "loss": 0.0318, "step": 97505 }, { "epoch": 2.28, "learning_rate": 1.3020229518199526e-05, "loss": 0.0672, "step": 97510 }, { "epoch": 2.28, "learning_rate": 1.3019349463753491e-05, "loss": 0.0887, "step": 97515 }, { "epoch": 2.28, "learning_rate": 1.3018469409307456e-05, "loss": 0.2593, "step": 97520 }, { "epoch": 2.28, "learning_rate": 1.3017589354861421e-05, "loss": 0.1816, "step": 97525 }, { "epoch": 2.28, "learning_rate": 1.3016709300415386e-05, "loss": 0.1056, "step": 97530 }, { "epoch": 2.28, "learning_rate": 1.3015829245969351e-05, "loss": 0.0486, "step": 97535 }, { "epoch": 2.28, "learning_rate": 1.3014949191523316e-05, "loss": 0.0068, "step": 97540 }, { "epoch": 2.28, "learning_rate": 1.3014069137077281e-05, "loss": 0.0518, "step": 97545 }, { "epoch": 2.28, "learning_rate": 1.3013189082631245e-05, "loss": 0.1739, "step": 97550 }, { "epoch": 2.28, "learning_rate": 1.3012309028185211e-05, "loss": 0.0169, "step": 97555 }, { "epoch": 2.28, "learning_rate": 1.3011428973739175e-05, "loss": 0.0412, "step": 97560 }, { "epoch": 2.28, "learning_rate": 1.3010548919293141e-05, "loss": 0.141, "step": 97565 }, { "epoch": 2.28, "learning_rate": 1.3009668864847105e-05, "loss": 0.0488, "step": 97570 }, { "epoch": 2.28, "learning_rate": 1.3008788810401072e-05, "loss": 0.1485, "step": 97575 }, { "epoch": 2.28, "learning_rate": 1.3007908755955035e-05, "loss": 0.0188, "step": 97580 }, { "epoch": 2.28, "learning_rate": 1.3007028701509002e-05, "loss": 0.0097, "step": 97585 }, { "epoch": 2.28, "learning_rate": 1.3006148647062965e-05, "loss": 0.0277, "step": 97590 }, { "epoch": 2.28, "learning_rate": 1.300526859261693e-05, "loss": 0.0122, "step": 97595 }, { "epoch": 2.28, "learning_rate": 1.3004388538170895e-05, "loss": 0.0128, "step": 97600 }, { "epoch": 2.28, "learning_rate": 1.300350848372486e-05, "loss": 0.1614, "step": 97605 }, { "epoch": 2.28, "learning_rate": 1.3002628429278825e-05, "loss": 0.1942, "step": 97610 }, { "epoch": 2.28, "learning_rate": 1.300174837483279e-05, "loss": 0.1615, "step": 97615 }, { "epoch": 2.28, "learning_rate": 1.3000868320386755e-05, "loss": 0.2529, "step": 97620 }, { "epoch": 2.28, "learning_rate": 1.299998826594072e-05, "loss": 0.333, "step": 97625 }, { "epoch": 2.28, "learning_rate": 1.2999108211494683e-05, "loss": 0.0291, "step": 97630 }, { "epoch": 2.28, "learning_rate": 1.299822815704865e-05, "loss": 0.0007, "step": 97635 }, { "epoch": 2.28, "learning_rate": 1.2997348102602613e-05, "loss": 0.053, "step": 97640 }, { "epoch": 2.28, "learning_rate": 1.299646804815658e-05, "loss": 0.091, "step": 97645 }, { "epoch": 2.28, "learning_rate": 1.2995587993710545e-05, "loss": 0.0056, "step": 97650 }, { "epoch": 2.28, "learning_rate": 1.299470793926451e-05, "loss": 0.0453, "step": 97655 }, { "epoch": 2.28, "learning_rate": 1.2993827884818475e-05, "loss": 0.1694, "step": 97660 }, { "epoch": 2.28, "learning_rate": 1.2992947830372438e-05, "loss": 0.2022, "step": 97665 }, { "epoch": 2.28, "learning_rate": 1.2992067775926405e-05, "loss": 0.4426, "step": 97670 }, { "epoch": 2.28, "learning_rate": 1.2991187721480369e-05, "loss": 0.0701, "step": 97675 }, { "epoch": 2.28, "learning_rate": 1.2990307667034335e-05, "loss": 0.064, "step": 97680 }, { "epoch": 2.28, "learning_rate": 1.2989427612588299e-05, "loss": 0.0665, "step": 97685 }, { "epoch": 2.28, "learning_rate": 1.2988547558142265e-05, "loss": 0.0951, "step": 97690 }, { "epoch": 2.28, "learning_rate": 1.2987667503696229e-05, "loss": 0.0776, "step": 97695 }, { "epoch": 2.28, "learning_rate": 1.2986787449250194e-05, "loss": 0.0542, "step": 97700 }, { "epoch": 2.28, "learning_rate": 1.2985907394804159e-05, "loss": 0.1707, "step": 97705 }, { "epoch": 2.28, "learning_rate": 1.2985027340358124e-05, "loss": 0.1441, "step": 97710 }, { "epoch": 2.28, "learning_rate": 1.2984147285912089e-05, "loss": 0.0441, "step": 97715 }, { "epoch": 2.28, "learning_rate": 1.2983267231466054e-05, "loss": 0.2114, "step": 97720 }, { "epoch": 2.28, "learning_rate": 1.2982387177020019e-05, "loss": 0.2222, "step": 97725 }, { "epoch": 2.28, "learning_rate": 1.2981507122573984e-05, "loss": 0.0124, "step": 97730 }, { "epoch": 2.28, "learning_rate": 1.2980627068127949e-05, "loss": 0.0663, "step": 97735 }, { "epoch": 2.28, "learning_rate": 1.2979747013681914e-05, "loss": 0.0346, "step": 97740 }, { "epoch": 2.28, "learning_rate": 1.2978866959235877e-05, "loss": 0.0377, "step": 97745 }, { "epoch": 2.28, "learning_rate": 1.2977986904789844e-05, "loss": 0.1811, "step": 97750 }, { "epoch": 2.28, "learning_rate": 1.2977106850343807e-05, "loss": 0.0525, "step": 97755 }, { "epoch": 2.28, "learning_rate": 1.2976226795897774e-05, "loss": 0.0706, "step": 97760 }, { "epoch": 2.28, "learning_rate": 1.2975346741451737e-05, "loss": 0.1291, "step": 97765 }, { "epoch": 2.28, "learning_rate": 1.2974466687005704e-05, "loss": 0.2327, "step": 97770 }, { "epoch": 2.28, "learning_rate": 1.2973586632559667e-05, "loss": 0.27, "step": 97775 }, { "epoch": 2.28, "learning_rate": 1.2972706578113632e-05, "loss": 0.0333, "step": 97780 }, { "epoch": 2.28, "learning_rate": 1.2971826523667597e-05, "loss": 0.0711, "step": 97785 }, { "epoch": 2.28, "learning_rate": 1.2970946469221562e-05, "loss": 0.027, "step": 97790 }, { "epoch": 2.28, "learning_rate": 1.2970066414775529e-05, "loss": 0.0578, "step": 97795 }, { "epoch": 2.28, "learning_rate": 1.2969186360329492e-05, "loss": 0.1332, "step": 97800 }, { "epoch": 2.28, "learning_rate": 1.2968306305883459e-05, "loss": 0.1807, "step": 97805 }, { "epoch": 2.28, "learning_rate": 1.2967426251437422e-05, "loss": 0.1461, "step": 97810 }, { "epoch": 2.28, "learning_rate": 1.2966546196991387e-05, "loss": 0.2363, "step": 97815 }, { "epoch": 2.28, "learning_rate": 1.2965666142545352e-05, "loss": 0.1637, "step": 97820 }, { "epoch": 2.28, "learning_rate": 1.2964786088099317e-05, "loss": 0.2751, "step": 97825 }, { "epoch": 2.28, "learning_rate": 1.2963906033653282e-05, "loss": 0.0243, "step": 97830 }, { "epoch": 2.28, "learning_rate": 1.2963025979207247e-05, "loss": 0.0013, "step": 97835 }, { "epoch": 2.28, "learning_rate": 1.2962145924761213e-05, "loss": 0.033, "step": 97840 }, { "epoch": 2.28, "learning_rate": 1.2961265870315178e-05, "loss": 0.0637, "step": 97845 }, { "epoch": 2.28, "learning_rate": 1.2960385815869141e-05, "loss": 0.3043, "step": 97850 }, { "epoch": 2.28, "learning_rate": 1.2959505761423108e-05, "loss": 0.2116, "step": 97855 }, { "epoch": 2.28, "learning_rate": 1.2958625706977071e-05, "loss": 0.0884, "step": 97860 }, { "epoch": 2.28, "learning_rate": 1.2957745652531038e-05, "loss": 0.1555, "step": 97865 }, { "epoch": 2.28, "learning_rate": 1.2956865598085001e-05, "loss": 0.1951, "step": 97870 }, { "epoch": 2.28, "learning_rate": 1.2955985543638968e-05, "loss": 0.1962, "step": 97875 }, { "epoch": 2.28, "learning_rate": 1.2955105489192931e-05, "loss": 0.0703, "step": 97880 }, { "epoch": 2.28, "learning_rate": 1.2954225434746898e-05, "loss": 0.0301, "step": 97885 }, { "epoch": 2.28, "learning_rate": 1.2953345380300861e-05, "loss": 0.029, "step": 97890 }, { "epoch": 2.28, "learning_rate": 1.2952465325854826e-05, "loss": 0.0613, "step": 97895 }, { "epoch": 2.28, "learning_rate": 1.2951585271408791e-05, "loss": 0.0527, "step": 97900 }, { "epoch": 2.28, "learning_rate": 1.2950705216962756e-05, "loss": 0.0561, "step": 97905 }, { "epoch": 2.28, "learning_rate": 1.2949825162516721e-05, "loss": 0.0543, "step": 97910 }, { "epoch": 2.28, "learning_rate": 1.2948945108070686e-05, "loss": 0.157, "step": 97915 }, { "epoch": 2.28, "learning_rate": 1.2948065053624651e-05, "loss": 0.2259, "step": 97920 }, { "epoch": 2.28, "learning_rate": 1.2947184999178616e-05, "loss": 0.21, "step": 97925 }, { "epoch": 2.28, "learning_rate": 1.2946304944732581e-05, "loss": 0.0736, "step": 97930 }, { "epoch": 2.28, "learning_rate": 1.2945424890286546e-05, "loss": 0.0893, "step": 97935 }, { "epoch": 2.29, "learning_rate": 1.2944544835840511e-05, "loss": 0.0709, "step": 97940 }, { "epoch": 2.29, "learning_rate": 1.2943664781394476e-05, "loss": 0.0107, "step": 97945 }, { "epoch": 2.29, "learning_rate": 1.2942784726948441e-05, "loss": 0.0548, "step": 97950 }, { "epoch": 2.29, "learning_rate": 1.2941904672502406e-05, "loss": 0.0332, "step": 97955 }, { "epoch": 2.29, "learning_rate": 1.2941024618056371e-05, "loss": 0.0246, "step": 97960 }, { "epoch": 2.29, "learning_rate": 1.2940144563610335e-05, "loss": 0.0709, "step": 97965 }, { "epoch": 2.29, "learning_rate": 1.2939264509164301e-05, "loss": 0.1557, "step": 97970 }, { "epoch": 2.29, "learning_rate": 1.2938384454718265e-05, "loss": 0.2417, "step": 97975 }, { "epoch": 2.29, "learning_rate": 1.2937504400272231e-05, "loss": 0.084, "step": 97980 }, { "epoch": 2.29, "learning_rate": 1.2936624345826195e-05, "loss": 0.0375, "step": 97985 }, { "epoch": 2.29, "learning_rate": 1.2935744291380161e-05, "loss": 0.0378, "step": 97990 }, { "epoch": 2.29, "learning_rate": 1.2934864236934125e-05, "loss": 0.001, "step": 97995 }, { "epoch": 2.29, "learning_rate": 1.293398418248809e-05, "loss": 0.0662, "step": 98000 }, { "epoch": 2.29, "learning_rate": 1.2933104128042055e-05, "loss": 0.044, "step": 98005 }, { "epoch": 2.29, "learning_rate": 1.293222407359602e-05, "loss": 0.0784, "step": 98010 }, { "epoch": 2.29, "learning_rate": 1.2931344019149985e-05, "loss": 0.1095, "step": 98015 }, { "epoch": 2.29, "learning_rate": 1.293046396470395e-05, "loss": 0.1488, "step": 98020 }, { "epoch": 2.29, "learning_rate": 1.2929583910257915e-05, "loss": 0.3022, "step": 98025 }, { "epoch": 2.29, "learning_rate": 1.292870385581188e-05, "loss": 0.0431, "step": 98030 }, { "epoch": 2.29, "learning_rate": 1.2927823801365845e-05, "loss": 0.0632, "step": 98035 }, { "epoch": 2.29, "learning_rate": 1.292694374691981e-05, "loss": 0.0201, "step": 98040 }, { "epoch": 2.29, "learning_rate": 1.2926063692473773e-05, "loss": 0.124, "step": 98045 }, { "epoch": 2.29, "learning_rate": 1.292518363802774e-05, "loss": 0.157, "step": 98050 }, { "epoch": 2.29, "learning_rate": 1.2924303583581705e-05, "loss": 0.1129, "step": 98055 }, { "epoch": 2.29, "learning_rate": 1.292342352913567e-05, "loss": 0.0553, "step": 98060 }, { "epoch": 2.29, "learning_rate": 1.2922543474689635e-05, "loss": 0.1922, "step": 98065 }, { "epoch": 2.29, "learning_rate": 1.29216634202436e-05, "loss": 0.2723, "step": 98070 }, { "epoch": 2.29, "learning_rate": 1.2920783365797565e-05, "loss": 0.2438, "step": 98075 }, { "epoch": 2.29, "learning_rate": 1.2919903311351528e-05, "loss": 0.0154, "step": 98080 }, { "epoch": 2.29, "learning_rate": 1.2919023256905495e-05, "loss": 0.0366, "step": 98085 }, { "epoch": 2.29, "learning_rate": 1.2918143202459458e-05, "loss": 0.0669, "step": 98090 }, { "epoch": 2.29, "learning_rate": 1.2917263148013425e-05, "loss": 0.0783, "step": 98095 }, { "epoch": 2.29, "learning_rate": 1.2916383093567389e-05, "loss": 0.0893, "step": 98100 }, { "epoch": 2.29, "learning_rate": 1.2915503039121355e-05, "loss": 0.0136, "step": 98105 }, { "epoch": 2.29, "learning_rate": 1.2914622984675319e-05, "loss": 0.205, "step": 98110 }, { "epoch": 2.29, "learning_rate": 1.2913742930229284e-05, "loss": 0.1721, "step": 98115 }, { "epoch": 2.29, "learning_rate": 1.2912862875783249e-05, "loss": 0.2449, "step": 98120 }, { "epoch": 2.29, "learning_rate": 1.2911982821337214e-05, "loss": 0.2334, "step": 98125 }, { "epoch": 2.29, "learning_rate": 1.2911102766891179e-05, "loss": 0.0561, "step": 98130 }, { "epoch": 2.29, "learning_rate": 1.2910222712445144e-05, "loss": 0.0698, "step": 98135 }, { "epoch": 2.29, "learning_rate": 1.2909342657999109e-05, "loss": 0.0434, "step": 98140 }, { "epoch": 2.29, "learning_rate": 1.2908462603553074e-05, "loss": 0.0895, "step": 98145 }, { "epoch": 2.29, "learning_rate": 1.2907582549107037e-05, "loss": 0.1374, "step": 98150 }, { "epoch": 2.29, "learning_rate": 1.2906702494661004e-05, "loss": 0.0951, "step": 98155 }, { "epoch": 2.29, "learning_rate": 1.2905822440214967e-05, "loss": 0.0988, "step": 98160 }, { "epoch": 2.29, "learning_rate": 1.2904942385768934e-05, "loss": 0.1212, "step": 98165 }, { "epoch": 2.29, "learning_rate": 1.2904062331322897e-05, "loss": 0.191, "step": 98170 }, { "epoch": 2.29, "learning_rate": 1.2903182276876864e-05, "loss": 0.2004, "step": 98175 }, { "epoch": 2.29, "learning_rate": 1.2902302222430827e-05, "loss": 0.0333, "step": 98180 }, { "epoch": 2.29, "learning_rate": 1.2901422167984794e-05, "loss": 0.0454, "step": 98185 }, { "epoch": 2.29, "learning_rate": 1.2900542113538757e-05, "loss": 0.0278, "step": 98190 }, { "epoch": 2.29, "learning_rate": 1.2899662059092722e-05, "loss": 0.1697, "step": 98195 }, { "epoch": 2.29, "learning_rate": 1.2898782004646689e-05, "loss": 0.1222, "step": 98200 }, { "epoch": 2.29, "learning_rate": 1.2897901950200652e-05, "loss": 0.1129, "step": 98205 }, { "epoch": 2.29, "learning_rate": 1.2897021895754619e-05, "loss": 0.081, "step": 98210 }, { "epoch": 2.29, "learning_rate": 1.2896141841308582e-05, "loss": 0.1419, "step": 98215 }, { "epoch": 2.29, "learning_rate": 1.2895261786862549e-05, "loss": 0.327, "step": 98220 }, { "epoch": 2.29, "learning_rate": 1.2894381732416512e-05, "loss": 0.3367, "step": 98225 }, { "epoch": 2.29, "learning_rate": 1.2893501677970477e-05, "loss": 0.067, "step": 98230 }, { "epoch": 2.29, "learning_rate": 1.2892621623524442e-05, "loss": 0.1236, "step": 98235 }, { "epoch": 2.29, "learning_rate": 1.2891741569078407e-05, "loss": 0.0769, "step": 98240 }, { "epoch": 2.29, "learning_rate": 1.2890861514632372e-05, "loss": 0.0755, "step": 98245 }, { "epoch": 2.29, "learning_rate": 1.2889981460186337e-05, "loss": 0.0961, "step": 98250 }, { "epoch": 2.29, "learning_rate": 1.2889101405740302e-05, "loss": 0.0454, "step": 98255 }, { "epoch": 2.29, "learning_rate": 1.2888221351294267e-05, "loss": 0.1569, "step": 98260 }, { "epoch": 2.29, "learning_rate": 1.288734129684823e-05, "loss": 0.2446, "step": 98265 }, { "epoch": 2.29, "learning_rate": 1.2886461242402198e-05, "loss": 0.3483, "step": 98270 }, { "epoch": 2.29, "learning_rate": 1.288558118795616e-05, "loss": 0.1936, "step": 98275 }, { "epoch": 2.29, "learning_rate": 1.2884701133510128e-05, "loss": 0.002, "step": 98280 }, { "epoch": 2.29, "learning_rate": 1.2883821079064091e-05, "loss": 0.0217, "step": 98285 }, { "epoch": 2.29, "learning_rate": 1.2882941024618058e-05, "loss": 0.0425, "step": 98290 }, { "epoch": 2.29, "learning_rate": 1.2882060970172021e-05, "loss": 0.1266, "step": 98295 }, { "epoch": 2.29, "learning_rate": 1.2881180915725986e-05, "loss": 0.026, "step": 98300 }, { "epoch": 2.29, "learning_rate": 1.2880300861279951e-05, "loss": 0.0442, "step": 98305 }, { "epoch": 2.29, "learning_rate": 1.2879420806833916e-05, "loss": 0.2387, "step": 98310 }, { "epoch": 2.29, "learning_rate": 1.2878540752387881e-05, "loss": 0.1712, "step": 98315 }, { "epoch": 2.29, "learning_rate": 1.2877660697941846e-05, "loss": 0.3543, "step": 98320 }, { "epoch": 2.29, "learning_rate": 1.2876780643495811e-05, "loss": 0.2238, "step": 98325 }, { "epoch": 2.29, "learning_rate": 1.2875900589049776e-05, "loss": 0.0577, "step": 98330 }, { "epoch": 2.29, "learning_rate": 1.2875020534603743e-05, "loss": 0.0648, "step": 98335 }, { "epoch": 2.29, "learning_rate": 1.2874140480157706e-05, "loss": 0.0159, "step": 98340 }, { "epoch": 2.29, "learning_rate": 1.2873260425711671e-05, "loss": 0.0783, "step": 98345 }, { "epoch": 2.29, "learning_rate": 1.2872380371265636e-05, "loss": 0.0848, "step": 98350 }, { "epoch": 2.29, "learning_rate": 1.2871500316819601e-05, "loss": 0.0701, "step": 98355 }, { "epoch": 2.29, "learning_rate": 1.2870620262373566e-05, "loss": 0.1951, "step": 98360 }, { "epoch": 2.29, "learning_rate": 1.2869740207927531e-05, "loss": 0.2215, "step": 98365 }, { "epoch": 2.3, "learning_rate": 1.2868860153481496e-05, "loss": 0.3758, "step": 98370 }, { "epoch": 2.3, "learning_rate": 1.2867980099035461e-05, "loss": 0.3343, "step": 98375 }, { "epoch": 2.3, "learning_rate": 1.2867100044589425e-05, "loss": 0.0396, "step": 98380 }, { "epoch": 2.3, "learning_rate": 1.2866219990143391e-05, "loss": 0.0358, "step": 98385 }, { "epoch": 2.3, "learning_rate": 1.2865339935697355e-05, "loss": 0.039, "step": 98390 }, { "epoch": 2.3, "learning_rate": 1.2864459881251321e-05, "loss": 0.0346, "step": 98395 }, { "epoch": 2.3, "learning_rate": 1.2863579826805285e-05, "loss": 0.0132, "step": 98400 }, { "epoch": 2.3, "learning_rate": 1.2862699772359251e-05, "loss": 0.0463, "step": 98405 }, { "epoch": 2.3, "learning_rate": 1.2861819717913215e-05, "loss": 0.1099, "step": 98410 }, { "epoch": 2.3, "learning_rate": 1.286093966346718e-05, "loss": 0.102, "step": 98415 }, { "epoch": 2.3, "learning_rate": 1.2860059609021145e-05, "loss": 0.3052, "step": 98420 }, { "epoch": 2.3, "learning_rate": 1.285917955457511e-05, "loss": 0.3684, "step": 98425 }, { "epoch": 2.3, "learning_rate": 1.2858299500129075e-05, "loss": 0.0089, "step": 98430 }, { "epoch": 2.3, "learning_rate": 1.285741944568304e-05, "loss": 0.0317, "step": 98435 }, { "epoch": 2.3, "learning_rate": 1.2856539391237005e-05, "loss": 0.1303, "step": 98440 }, { "epoch": 2.3, "learning_rate": 1.285565933679097e-05, "loss": 0.0073, "step": 98445 }, { "epoch": 2.3, "learning_rate": 1.2854779282344933e-05, "loss": 0.2009, "step": 98450 }, { "epoch": 2.3, "learning_rate": 1.28538992278989e-05, "loss": 0.0358, "step": 98455 }, { "epoch": 2.3, "learning_rate": 1.2853019173452863e-05, "loss": 0.1106, "step": 98460 }, { "epoch": 2.3, "learning_rate": 1.285213911900683e-05, "loss": 0.1801, "step": 98465 }, { "epoch": 2.3, "learning_rate": 1.2851259064560795e-05, "loss": 0.1492, "step": 98470 }, { "epoch": 2.3, "learning_rate": 1.285037901011476e-05, "loss": 0.1788, "step": 98475 }, { "epoch": 2.3, "learning_rate": 1.2849498955668725e-05, "loss": 0.114, "step": 98480 }, { "epoch": 2.3, "learning_rate": 1.284861890122269e-05, "loss": 0.0261, "step": 98485 }, { "epoch": 2.3, "learning_rate": 1.2847738846776655e-05, "loss": 0.0268, "step": 98490 }, { "epoch": 2.3, "learning_rate": 1.2846858792330618e-05, "loss": 0.0816, "step": 98495 }, { "epoch": 2.3, "learning_rate": 1.2845978737884585e-05, "loss": 0.0446, "step": 98500 }, { "epoch": 2.3, "learning_rate": 1.2845098683438548e-05, "loss": 0.0663, "step": 98505 }, { "epoch": 2.3, "learning_rate": 1.2844218628992515e-05, "loss": 0.2599, "step": 98510 }, { "epoch": 2.3, "learning_rate": 1.2843338574546478e-05, "loss": 0.1933, "step": 98515 }, { "epoch": 2.3, "learning_rate": 1.2842458520100445e-05, "loss": 0.1802, "step": 98520 }, { "epoch": 2.3, "learning_rate": 1.2841578465654408e-05, "loss": 0.1684, "step": 98525 }, { "epoch": 2.3, "learning_rate": 1.2840698411208373e-05, "loss": 0.0069, "step": 98530 }, { "epoch": 2.3, "learning_rate": 1.2839818356762339e-05, "loss": 0.0449, "step": 98535 }, { "epoch": 2.3, "learning_rate": 1.2838938302316304e-05, "loss": 0.0202, "step": 98540 }, { "epoch": 2.3, "learning_rate": 1.2838058247870269e-05, "loss": 0.1299, "step": 98545 }, { "epoch": 2.3, "learning_rate": 1.2837178193424234e-05, "loss": 0.0594, "step": 98550 }, { "epoch": 2.3, "learning_rate": 1.2836298138978199e-05, "loss": 0.0534, "step": 98555 }, { "epoch": 2.3, "learning_rate": 1.2835418084532164e-05, "loss": 0.0536, "step": 98560 }, { "epoch": 2.3, "learning_rate": 1.2834538030086127e-05, "loss": 0.1407, "step": 98565 }, { "epoch": 2.3, "learning_rate": 1.2833657975640094e-05, "loss": 0.3884, "step": 98570 }, { "epoch": 2.3, "learning_rate": 1.2832777921194057e-05, "loss": 0.2201, "step": 98575 }, { "epoch": 2.3, "learning_rate": 1.2831897866748024e-05, "loss": 0.027, "step": 98580 }, { "epoch": 2.3, "learning_rate": 1.2831017812301987e-05, "loss": 0.0046, "step": 98585 }, { "epoch": 2.3, "learning_rate": 1.2830137757855954e-05, "loss": 0.0733, "step": 98590 }, { "epoch": 2.3, "learning_rate": 1.2829257703409917e-05, "loss": 0.0868, "step": 98595 }, { "epoch": 2.3, "learning_rate": 1.2828377648963882e-05, "loss": 0.1371, "step": 98600 }, { "epoch": 2.3, "learning_rate": 1.2827497594517849e-05, "loss": 0.0623, "step": 98605 }, { "epoch": 2.3, "learning_rate": 1.2826617540071812e-05, "loss": 0.1152, "step": 98610 }, { "epoch": 2.3, "learning_rate": 1.2825737485625779e-05, "loss": 0.2051, "step": 98615 }, { "epoch": 2.3, "learning_rate": 1.2824857431179742e-05, "loss": 0.291, "step": 98620 }, { "epoch": 2.3, "learning_rate": 1.2823977376733709e-05, "loss": 0.255, "step": 98625 }, { "epoch": 2.3, "learning_rate": 1.2823097322287672e-05, "loss": 0.004, "step": 98630 }, { "epoch": 2.3, "learning_rate": 1.2822217267841639e-05, "loss": 0.0084, "step": 98635 }, { "epoch": 2.3, "learning_rate": 1.2821337213395602e-05, "loss": 0.1286, "step": 98640 }, { "epoch": 2.3, "learning_rate": 1.2820457158949567e-05, "loss": 0.0831, "step": 98645 }, { "epoch": 2.3, "learning_rate": 1.2819577104503532e-05, "loss": 0.0626, "step": 98650 }, { "epoch": 2.3, "learning_rate": 1.2818697050057497e-05, "loss": 0.0584, "step": 98655 }, { "epoch": 2.3, "learning_rate": 1.2817816995611462e-05, "loss": 0.0713, "step": 98660 }, { "epoch": 2.3, "learning_rate": 1.2816936941165427e-05, "loss": 0.2136, "step": 98665 }, { "epoch": 2.3, "learning_rate": 1.2816056886719392e-05, "loss": 0.3155, "step": 98670 }, { "epoch": 2.3, "learning_rate": 1.2815176832273357e-05, "loss": 0.228, "step": 98675 }, { "epoch": 2.3, "learning_rate": 1.281429677782732e-05, "loss": 0.0197, "step": 98680 }, { "epoch": 2.3, "learning_rate": 1.2813416723381287e-05, "loss": 0.0026, "step": 98685 }, { "epoch": 2.3, "learning_rate": 1.281253666893525e-05, "loss": 0.0056, "step": 98690 }, { "epoch": 2.3, "learning_rate": 1.2811656614489217e-05, "loss": 0.1925, "step": 98695 }, { "epoch": 2.3, "learning_rate": 1.281077656004318e-05, "loss": 0.1897, "step": 98700 }, { "epoch": 2.3, "learning_rate": 1.2809896505597148e-05, "loss": 0.0479, "step": 98705 }, { "epoch": 2.3, "learning_rate": 1.280901645115111e-05, "loss": 0.1297, "step": 98710 }, { "epoch": 2.3, "learning_rate": 1.2808136396705076e-05, "loss": 0.0777, "step": 98715 }, { "epoch": 2.3, "learning_rate": 1.2807256342259041e-05, "loss": 0.2618, "step": 98720 }, { "epoch": 2.3, "learning_rate": 1.2806376287813006e-05, "loss": 0.2255, "step": 98725 }, { "epoch": 2.3, "learning_rate": 1.2805496233366971e-05, "loss": 0.0424, "step": 98730 }, { "epoch": 2.3, "learning_rate": 1.2804616178920936e-05, "loss": 0.0279, "step": 98735 }, { "epoch": 2.3, "learning_rate": 1.2803736124474903e-05, "loss": 0.0142, "step": 98740 }, { "epoch": 2.3, "learning_rate": 1.2802856070028866e-05, "loss": 0.0548, "step": 98745 }, { "epoch": 2.3, "learning_rate": 1.2801976015582831e-05, "loss": 0.0512, "step": 98750 }, { "epoch": 2.3, "learning_rate": 1.2801095961136796e-05, "loss": 0.0818, "step": 98755 }, { "epoch": 2.3, "learning_rate": 1.2800215906690761e-05, "loss": 0.0778, "step": 98760 }, { "epoch": 2.3, "learning_rate": 1.2799335852244726e-05, "loss": 0.1998, "step": 98765 }, { "epoch": 2.3, "learning_rate": 1.2798455797798691e-05, "loss": 0.1916, "step": 98770 }, { "epoch": 2.3, "learning_rate": 1.2797575743352656e-05, "loss": 0.3161, "step": 98775 }, { "epoch": 2.3, "learning_rate": 1.2796695688906621e-05, "loss": 0.0915, "step": 98780 }, { "epoch": 2.3, "learning_rate": 1.2795815634460586e-05, "loss": 0.1005, "step": 98785 }, { "epoch": 2.3, "learning_rate": 1.2794935580014551e-05, "loss": 0.1326, "step": 98790 }, { "epoch": 2.31, "learning_rate": 1.2794055525568514e-05, "loss": 0.0784, "step": 98795 }, { "epoch": 2.31, "learning_rate": 1.2793175471122481e-05, "loss": 0.0955, "step": 98800 }, { "epoch": 2.31, "learning_rate": 1.2792295416676445e-05, "loss": 0.0702, "step": 98805 }, { "epoch": 2.31, "learning_rate": 1.2791415362230411e-05, "loss": 0.0776, "step": 98810 }, { "epoch": 2.31, "learning_rate": 1.2790535307784375e-05, "loss": 0.2379, "step": 98815 }, { "epoch": 2.31, "learning_rate": 1.2789655253338341e-05, "loss": 0.2942, "step": 98820 }, { "epoch": 2.31, "learning_rate": 1.2788775198892305e-05, "loss": 0.1065, "step": 98825 }, { "epoch": 2.31, "learning_rate": 1.278789514444627e-05, "loss": 0.0434, "step": 98830 }, { "epoch": 2.31, "learning_rate": 1.2787015090000235e-05, "loss": 0.0061, "step": 98835 }, { "epoch": 2.31, "learning_rate": 1.27861350355542e-05, "loss": 0.0324, "step": 98840 }, { "epoch": 2.31, "learning_rate": 1.2785254981108165e-05, "loss": 0.0092, "step": 98845 }, { "epoch": 2.31, "learning_rate": 1.278437492666213e-05, "loss": 0.0124, "step": 98850 }, { "epoch": 2.31, "learning_rate": 1.2783494872216095e-05, "loss": 0.0285, "step": 98855 }, { "epoch": 2.31, "learning_rate": 1.278261481777006e-05, "loss": 0.1949, "step": 98860 }, { "epoch": 2.31, "learning_rate": 1.2781734763324023e-05, "loss": 0.1892, "step": 98865 }, { "epoch": 2.31, "learning_rate": 1.278085470887799e-05, "loss": 0.1843, "step": 98870 }, { "epoch": 2.31, "learning_rate": 1.2779974654431955e-05, "loss": 0.1254, "step": 98875 }, { "epoch": 2.31, "learning_rate": 1.277909459998592e-05, "loss": 0.0185, "step": 98880 }, { "epoch": 2.31, "learning_rate": 1.2778214545539885e-05, "loss": 0.0926, "step": 98885 }, { "epoch": 2.31, "learning_rate": 1.277733449109385e-05, "loss": 0.0043, "step": 98890 }, { "epoch": 2.31, "learning_rate": 1.2776454436647815e-05, "loss": 0.0312, "step": 98895 }, { "epoch": 2.31, "learning_rate": 1.2775574382201778e-05, "loss": 0.0293, "step": 98900 }, { "epoch": 2.31, "learning_rate": 1.2774694327755745e-05, "loss": 0.1351, "step": 98905 }, { "epoch": 2.31, "learning_rate": 1.2773814273309708e-05, "loss": 0.174, "step": 98910 }, { "epoch": 2.31, "learning_rate": 1.2772934218863675e-05, "loss": 0.1126, "step": 98915 }, { "epoch": 2.31, "learning_rate": 1.2772054164417638e-05, "loss": 0.2571, "step": 98920 }, { "epoch": 2.31, "learning_rate": 1.2771350120860811e-05, "loss": 0.263, "step": 98925 }, { "epoch": 2.31, "learning_rate": 1.2770470066414776e-05, "loss": 0.007, "step": 98930 }, { "epoch": 2.31, "learning_rate": 1.2769590011968741e-05, "loss": 0.0519, "step": 98935 }, { "epoch": 2.31, "learning_rate": 1.2768709957522706e-05, "loss": 0.0777, "step": 98940 }, { "epoch": 2.31, "learning_rate": 1.276782990307667e-05, "loss": 0.0928, "step": 98945 }, { "epoch": 2.31, "learning_rate": 1.2766949848630636e-05, "loss": 0.0575, "step": 98950 }, { "epoch": 2.31, "learning_rate": 1.27660697941846e-05, "loss": 0.0153, "step": 98955 }, { "epoch": 2.31, "learning_rate": 1.2765189739738566e-05, "loss": 0.0467, "step": 98960 }, { "epoch": 2.31, "learning_rate": 1.276430968529253e-05, "loss": 0.0498, "step": 98965 }, { "epoch": 2.31, "learning_rate": 1.2763429630846497e-05, "loss": 0.2856, "step": 98970 }, { "epoch": 2.31, "learning_rate": 1.276254957640046e-05, "loss": 0.3263, "step": 98975 }, { "epoch": 2.31, "learning_rate": 1.2761669521954427e-05, "loss": 0.0065, "step": 98980 }, { "epoch": 2.31, "learning_rate": 1.276078946750839e-05, "loss": 0.0235, "step": 98985 }, { "epoch": 2.31, "learning_rate": 1.2759909413062355e-05, "loss": 0.0458, "step": 98990 }, { "epoch": 2.31, "learning_rate": 1.275902935861632e-05, "loss": 0.0162, "step": 98995 }, { "epoch": 2.31, "learning_rate": 1.2758149304170285e-05, "loss": 0.081, "step": 99000 }, { "epoch": 2.31, "learning_rate": 1.275726924972425e-05, "loss": 0.1213, "step": 99005 }, { "epoch": 2.31, "learning_rate": 1.2756389195278215e-05, "loss": 0.2933, "step": 99010 }, { "epoch": 2.31, "learning_rate": 1.275550914083218e-05, "loss": 0.2225, "step": 99015 }, { "epoch": 2.31, "learning_rate": 1.2754629086386145e-05, "loss": 0.1339, "step": 99020 }, { "epoch": 2.31, "learning_rate": 1.2753749031940108e-05, "loss": 0.3308, "step": 99025 }, { "epoch": 2.31, "learning_rate": 1.2752868977494075e-05, "loss": 0.0552, "step": 99030 }, { "epoch": 2.31, "learning_rate": 1.2751988923048038e-05, "loss": 0.0412, "step": 99035 }, { "epoch": 2.31, "learning_rate": 1.2751108868602005e-05, "loss": 0.1078, "step": 99040 }, { "epoch": 2.31, "learning_rate": 1.2750228814155968e-05, "loss": 0.0718, "step": 99045 }, { "epoch": 2.31, "learning_rate": 1.2749348759709935e-05, "loss": 0.0487, "step": 99050 }, { "epoch": 2.31, "learning_rate": 1.2748468705263898e-05, "loss": 0.2271, "step": 99055 }, { "epoch": 2.31, "learning_rate": 1.2747588650817864e-05, "loss": 0.1708, "step": 99060 }, { "epoch": 2.31, "learning_rate": 1.274670859637183e-05, "loss": 0.2248, "step": 99065 }, { "epoch": 2.31, "learning_rate": 1.2745828541925794e-05, "loss": 0.1883, "step": 99070 }, { "epoch": 2.31, "learning_rate": 1.274494848747976e-05, "loss": 0.3247, "step": 99075 }, { "epoch": 2.31, "learning_rate": 1.2744068433033724e-05, "loss": 0.0297, "step": 99080 }, { "epoch": 2.31, "learning_rate": 1.274318837858769e-05, "loss": 0.0395, "step": 99085 }, { "epoch": 2.31, "learning_rate": 1.2742308324141654e-05, "loss": 0.0274, "step": 99090 }, { "epoch": 2.31, "learning_rate": 1.2741428269695619e-05, "loss": 0.0787, "step": 99095 }, { "epoch": 2.31, "learning_rate": 1.2740548215249584e-05, "loss": 0.0752, "step": 99100 }, { "epoch": 2.31, "learning_rate": 1.2739668160803549e-05, "loss": 0.0972, "step": 99105 }, { "epoch": 2.31, "learning_rate": 1.2738788106357514e-05, "loss": 0.3151, "step": 99110 }, { "epoch": 2.31, "learning_rate": 1.2737908051911479e-05, "loss": 0.0274, "step": 99115 }, { "epoch": 2.31, "learning_rate": 1.2737027997465444e-05, "loss": 0.271, "step": 99120 }, { "epoch": 2.31, "learning_rate": 1.2736147943019409e-05, "loss": 0.276, "step": 99125 }, { "epoch": 2.31, "learning_rate": 1.2735267888573374e-05, "loss": 0.0368, "step": 99130 }, { "epoch": 2.31, "learning_rate": 1.2734387834127339e-05, "loss": 0.0256, "step": 99135 }, { "epoch": 2.31, "learning_rate": 1.2733507779681302e-05, "loss": 0.1226, "step": 99140 }, { "epoch": 2.31, "learning_rate": 1.2732627725235269e-05, "loss": 0.1177, "step": 99145 }, { "epoch": 2.31, "learning_rate": 1.2731747670789232e-05, "loss": 0.0758, "step": 99150 }, { "epoch": 2.31, "learning_rate": 1.2730867616343199e-05, "loss": 0.0995, "step": 99155 }, { "epoch": 2.31, "learning_rate": 1.2729987561897162e-05, "loss": 0.1443, "step": 99160 }, { "epoch": 2.31, "learning_rate": 1.2729107507451129e-05, "loss": 0.1745, "step": 99165 }, { "epoch": 2.31, "learning_rate": 1.2728227453005092e-05, "loss": 0.346, "step": 99170 }, { "epoch": 2.31, "learning_rate": 1.2727347398559057e-05, "loss": 0.152, "step": 99175 }, { "epoch": 2.31, "learning_rate": 1.2726467344113022e-05, "loss": 0.0641, "step": 99180 }, { "epoch": 2.31, "learning_rate": 1.2725587289666987e-05, "loss": 0.0306, "step": 99185 }, { "epoch": 2.31, "learning_rate": 1.2724707235220952e-05, "loss": 0.0121, "step": 99190 }, { "epoch": 2.31, "learning_rate": 1.2723827180774917e-05, "loss": 0.1516, "step": 99195 }, { "epoch": 2.31, "learning_rate": 1.2722947126328882e-05, "loss": 0.1204, "step": 99200 }, { "epoch": 2.31, "learning_rate": 1.2722067071882847e-05, "loss": 0.0682, "step": 99205 }, { "epoch": 2.31, "learning_rate": 1.2721187017436812e-05, "loss": 0.1393, "step": 99210 }, { "epoch": 2.31, "learning_rate": 1.2720306962990777e-05, "loss": 0.1973, "step": 99215 }, { "epoch": 2.31, "learning_rate": 1.2719426908544742e-05, "loss": 0.2909, "step": 99220 }, { "epoch": 2.32, "learning_rate": 1.2718546854098707e-05, "loss": 0.193, "step": 99225 }, { "epoch": 2.32, "learning_rate": 1.2717666799652673e-05, "loss": 0.0443, "step": 99230 }, { "epoch": 2.32, "learning_rate": 1.2716786745206638e-05, "loss": 0.0995, "step": 99235 }, { "epoch": 2.32, "learning_rate": 1.2715906690760603e-05, "loss": 0.0276, "step": 99240 }, { "epoch": 2.32, "learning_rate": 1.2715026636314566e-05, "loss": 0.0801, "step": 99245 }, { "epoch": 2.32, "learning_rate": 1.2714146581868533e-05, "loss": 0.1192, "step": 99250 }, { "epoch": 2.32, "learning_rate": 1.2713266527422496e-05, "loss": 0.0941, "step": 99255 }, { "epoch": 2.32, "learning_rate": 1.2712386472976463e-05, "loss": 0.0834, "step": 99260 }, { "epoch": 2.32, "learning_rate": 1.2711506418530426e-05, "loss": 0.1286, "step": 99265 }, { "epoch": 2.32, "learning_rate": 1.2710626364084393e-05, "loss": 0.5839, "step": 99270 }, { "epoch": 2.32, "learning_rate": 1.2709746309638356e-05, "loss": 0.3491, "step": 99275 }, { "epoch": 2.32, "learning_rate": 1.2708866255192323e-05, "loss": 0.0101, "step": 99280 }, { "epoch": 2.32, "learning_rate": 1.2707986200746286e-05, "loss": 0.0945, "step": 99285 }, { "epoch": 2.32, "learning_rate": 1.2707106146300251e-05, "loss": 0.0889, "step": 99290 }, { "epoch": 2.32, "learning_rate": 1.2706226091854216e-05, "loss": 0.1171, "step": 99295 }, { "epoch": 2.32, "learning_rate": 1.2705346037408181e-05, "loss": 0.0465, "step": 99300 }, { "epoch": 2.32, "learning_rate": 1.2704465982962146e-05, "loss": 0.0968, "step": 99305 }, { "epoch": 2.32, "learning_rate": 1.2703585928516111e-05, "loss": 0.0618, "step": 99310 }, { "epoch": 2.32, "learning_rate": 1.2702705874070076e-05, "loss": 0.1138, "step": 99315 }, { "epoch": 2.32, "learning_rate": 1.2701825819624041e-05, "loss": 0.2481, "step": 99320 }, { "epoch": 2.32, "learning_rate": 1.2700945765178005e-05, "loss": 0.1964, "step": 99325 }, { "epoch": 2.32, "learning_rate": 1.2700065710731971e-05, "loss": 0.0032, "step": 99330 }, { "epoch": 2.32, "learning_rate": 1.2699185656285936e-05, "loss": 0.0267, "step": 99335 }, { "epoch": 2.32, "learning_rate": 1.2698305601839901e-05, "loss": 0.0613, "step": 99340 }, { "epoch": 2.32, "learning_rate": 1.2697425547393866e-05, "loss": 0.0627, "step": 99345 }, { "epoch": 2.32, "learning_rate": 1.2696545492947831e-05, "loss": 0.0541, "step": 99350 }, { "epoch": 2.32, "learning_rate": 1.2695665438501796e-05, "loss": 0.0568, "step": 99355 }, { "epoch": 2.32, "learning_rate": 1.269478538405576e-05, "loss": 0.1171, "step": 99360 }, { "epoch": 2.32, "learning_rate": 1.2693905329609726e-05, "loss": 0.1823, "step": 99365 }, { "epoch": 2.32, "learning_rate": 1.269302527516369e-05, "loss": 0.0988, "step": 99370 }, { "epoch": 2.32, "learning_rate": 1.2692145220717656e-05, "loss": 0.2937, "step": 99375 }, { "epoch": 2.32, "learning_rate": 1.269126516627162e-05, "loss": 0.0338, "step": 99380 }, { "epoch": 2.32, "learning_rate": 1.2690385111825586e-05, "loss": 0.0006, "step": 99385 }, { "epoch": 2.32, "learning_rate": 1.268950505737955e-05, "loss": 0.0126, "step": 99390 }, { "epoch": 2.32, "learning_rate": 1.2688625002933515e-05, "loss": 0.0495, "step": 99395 }, { "epoch": 2.32, "learning_rate": 1.268774494848748e-05, "loss": 0.061, "step": 99400 }, { "epoch": 2.32, "learning_rate": 1.2686864894041445e-05, "loss": 0.0124, "step": 99405 }, { "epoch": 2.32, "learning_rate": 1.268598483959541e-05, "loss": 0.0946, "step": 99410 }, { "epoch": 2.32, "learning_rate": 1.2685104785149375e-05, "loss": 0.2313, "step": 99415 }, { "epoch": 2.32, "learning_rate": 1.268422473070334e-05, "loss": 0.1526, "step": 99420 }, { "epoch": 2.32, "learning_rate": 1.2683344676257305e-05, "loss": 0.3644, "step": 99425 }, { "epoch": 2.32, "learning_rate": 1.268246462181127e-05, "loss": 0.1029, "step": 99430 }, { "epoch": 2.32, "learning_rate": 1.2681584567365235e-05, "loss": 0.047, "step": 99435 }, { "epoch": 2.32, "learning_rate": 1.2680704512919198e-05, "loss": 0.0443, "step": 99440 }, { "epoch": 2.32, "learning_rate": 1.2679824458473165e-05, "loss": 0.0404, "step": 99445 }, { "epoch": 2.32, "learning_rate": 1.2678944404027128e-05, "loss": 0.0535, "step": 99450 }, { "epoch": 2.32, "learning_rate": 1.2678064349581095e-05, "loss": 0.0765, "step": 99455 }, { "epoch": 2.32, "learning_rate": 1.2677184295135058e-05, "loss": 0.0527, "step": 99460 }, { "epoch": 2.32, "learning_rate": 1.2676304240689025e-05, "loss": 0.1911, "step": 99465 }, { "epoch": 2.32, "learning_rate": 1.2675424186242988e-05, "loss": 0.2771, "step": 99470 }, { "epoch": 2.32, "learning_rate": 1.2674544131796953e-05, "loss": 0.3444, "step": 99475 }, { "epoch": 2.32, "learning_rate": 1.267366407735092e-05, "loss": 0.062, "step": 99480 }, { "epoch": 2.32, "learning_rate": 1.2672784022904883e-05, "loss": 0.0844, "step": 99485 }, { "epoch": 2.32, "learning_rate": 1.267190396845885e-05, "loss": 0.0778, "step": 99490 }, { "epoch": 2.32, "learning_rate": 1.2671023914012814e-05, "loss": 0.1017, "step": 99495 }, { "epoch": 2.32, "learning_rate": 1.267014385956678e-05, "loss": 0.0408, "step": 99500 }, { "epoch": 2.32, "learning_rate": 1.2669263805120744e-05, "loss": 0.1679, "step": 99505 }, { "epoch": 2.32, "learning_rate": 1.2668383750674709e-05, "loss": 0.074, "step": 99510 }, { "epoch": 2.32, "learning_rate": 1.2667503696228674e-05, "loss": 0.2205, "step": 99515 }, { "epoch": 2.32, "learning_rate": 1.2666623641782639e-05, "loss": 0.3528, "step": 99520 }, { "epoch": 2.32, "learning_rate": 1.2665743587336604e-05, "loss": 0.2404, "step": 99525 }, { "epoch": 2.32, "learning_rate": 1.2664863532890569e-05, "loss": 0.0286, "step": 99530 }, { "epoch": 2.32, "learning_rate": 1.2663983478444534e-05, "loss": 0.0223, "step": 99535 }, { "epoch": 2.32, "learning_rate": 1.2663103423998499e-05, "loss": 0.0645, "step": 99540 }, { "epoch": 2.32, "learning_rate": 1.2662223369552462e-05, "loss": 0.0245, "step": 99545 }, { "epoch": 2.32, "learning_rate": 1.2661343315106429e-05, "loss": 0.0514, "step": 99550 }, { "epoch": 2.32, "learning_rate": 1.2660463260660392e-05, "loss": 0.0764, "step": 99555 }, { "epoch": 2.32, "learning_rate": 1.2659583206214359e-05, "loss": 0.094, "step": 99560 }, { "epoch": 2.32, "learning_rate": 1.2658703151768322e-05, "loss": 0.1643, "step": 99565 }, { "epoch": 2.32, "learning_rate": 1.2657823097322289e-05, "loss": 0.1814, "step": 99570 }, { "epoch": 2.32, "learning_rate": 1.2656943042876252e-05, "loss": 0.1993, "step": 99575 }, { "epoch": 2.32, "learning_rate": 1.2656062988430219e-05, "loss": 0.0585, "step": 99580 }, { "epoch": 2.32, "learning_rate": 1.2655182933984182e-05, "loss": 0.0034, "step": 99585 }, { "epoch": 2.32, "learning_rate": 1.2654302879538147e-05, "loss": 0.0233, "step": 99590 }, { "epoch": 2.32, "learning_rate": 1.2653422825092112e-05, "loss": 0.0055, "step": 99595 }, { "epoch": 2.32, "learning_rate": 1.2652542770646077e-05, "loss": 0.1041, "step": 99600 }, { "epoch": 2.32, "learning_rate": 1.2651662716200042e-05, "loss": 0.0661, "step": 99605 }, { "epoch": 2.32, "learning_rate": 1.2650782661754007e-05, "loss": 0.2102, "step": 99610 }, { "epoch": 2.32, "learning_rate": 1.2649902607307974e-05, "loss": 0.1649, "step": 99615 }, { "epoch": 2.32, "learning_rate": 1.2649022552861937e-05, "loss": 0.1249, "step": 99620 }, { "epoch": 2.32, "learning_rate": 1.2648142498415902e-05, "loss": 0.2265, "step": 99625 }, { "epoch": 2.32, "learning_rate": 1.2647262443969867e-05, "loss": 0.0189, "step": 99630 }, { "epoch": 2.32, "learning_rate": 1.2646382389523832e-05, "loss": 0.0014, "step": 99635 }, { "epoch": 2.32, "learning_rate": 1.2645502335077797e-05, "loss": 0.0195, "step": 99640 }, { "epoch": 2.32, "learning_rate": 1.2644622280631762e-05, "loss": 0.1286, "step": 99645 }, { "epoch": 2.32, "learning_rate": 1.2643742226185727e-05, "loss": 0.0294, "step": 99650 }, { "epoch": 2.33, "learning_rate": 1.2642862171739692e-05, "loss": 0.1931, "step": 99655 }, { "epoch": 2.33, "learning_rate": 1.2641982117293656e-05, "loss": 0.1305, "step": 99660 }, { "epoch": 2.33, "learning_rate": 1.2641102062847623e-05, "loss": 0.1582, "step": 99665 }, { "epoch": 2.33, "learning_rate": 1.2640222008401586e-05, "loss": 0.2076, "step": 99670 }, { "epoch": 2.33, "learning_rate": 1.2639341953955553e-05, "loss": 0.2226, "step": 99675 }, { "epoch": 2.33, "learning_rate": 1.2638461899509516e-05, "loss": 0.0094, "step": 99680 }, { "epoch": 2.33, "learning_rate": 1.2637581845063483e-05, "loss": 0.0825, "step": 99685 }, { "epoch": 2.33, "learning_rate": 1.2636701790617446e-05, "loss": 0.0019, "step": 99690 }, { "epoch": 2.33, "learning_rate": 1.2635821736171411e-05, "loss": 0.0735, "step": 99695 }, { "epoch": 2.33, "learning_rate": 1.2634941681725376e-05, "loss": 0.0454, "step": 99700 }, { "epoch": 2.33, "learning_rate": 1.2634061627279341e-05, "loss": 0.2871, "step": 99705 }, { "epoch": 2.33, "learning_rate": 1.2633181572833306e-05, "loss": 0.0303, "step": 99710 }, { "epoch": 2.33, "learning_rate": 1.2632301518387271e-05, "loss": 0.194, "step": 99715 }, { "epoch": 2.33, "learning_rate": 1.2631421463941236e-05, "loss": 0.2019, "step": 99720 }, { "epoch": 2.33, "learning_rate": 1.2630541409495201e-05, "loss": 0.2986, "step": 99725 }, { "epoch": 2.33, "learning_rate": 1.2629661355049166e-05, "loss": 0.0271, "step": 99730 }, { "epoch": 2.33, "learning_rate": 1.2628781300603131e-05, "loss": 0.0137, "step": 99735 }, { "epoch": 2.33, "learning_rate": 1.2627901246157094e-05, "loss": 0.0586, "step": 99740 }, { "epoch": 2.33, "learning_rate": 1.2627021191711061e-05, "loss": 0.0663, "step": 99745 }, { "epoch": 2.33, "learning_rate": 1.2626141137265026e-05, "loss": 0.0058, "step": 99750 }, { "epoch": 2.33, "learning_rate": 1.2625261082818991e-05, "loss": 0.0272, "step": 99755 }, { "epoch": 2.33, "learning_rate": 1.2624381028372956e-05, "loss": 0.0226, "step": 99760 }, { "epoch": 2.33, "learning_rate": 1.2623500973926921e-05, "loss": 0.1002, "step": 99765 }, { "epoch": 2.33, "learning_rate": 1.2622620919480886e-05, "loss": 0.1536, "step": 99770 }, { "epoch": 2.33, "learning_rate": 1.262174086503485e-05, "loss": 0.231, "step": 99775 }, { "epoch": 2.33, "learning_rate": 1.2620860810588816e-05, "loss": 0.034, "step": 99780 }, { "epoch": 2.33, "learning_rate": 1.261998075614278e-05, "loss": 0.0444, "step": 99785 }, { "epoch": 2.33, "learning_rate": 1.2619100701696746e-05, "loss": 0.0121, "step": 99790 }, { "epoch": 2.33, "learning_rate": 1.261822064725071e-05, "loss": 0.0896, "step": 99795 }, { "epoch": 2.33, "learning_rate": 1.2617340592804676e-05, "loss": 0.0069, "step": 99800 }, { "epoch": 2.33, "learning_rate": 1.261646053835864e-05, "loss": 0.1415, "step": 99805 }, { "epoch": 2.33, "learning_rate": 1.2615580483912605e-05, "loss": 0.2881, "step": 99810 }, { "epoch": 2.33, "learning_rate": 1.261470042946657e-05, "loss": 0.0961, "step": 99815 }, { "epoch": 2.33, "learning_rate": 1.2613820375020535e-05, "loss": 0.1133, "step": 99820 }, { "epoch": 2.33, "learning_rate": 1.26129403205745e-05, "loss": 0.2966, "step": 99825 }, { "epoch": 2.33, "learning_rate": 1.2612060266128465e-05, "loss": 0.0825, "step": 99830 }, { "epoch": 2.33, "learning_rate": 1.261118021168243e-05, "loss": 0.0312, "step": 99835 }, { "epoch": 2.33, "learning_rate": 1.2610300157236395e-05, "loss": 0.0333, "step": 99840 }, { "epoch": 2.33, "learning_rate": 1.2609420102790358e-05, "loss": 0.0612, "step": 99845 }, { "epoch": 2.33, "learning_rate": 1.2608540048344325e-05, "loss": 0.0802, "step": 99850 }, { "epoch": 2.33, "learning_rate": 1.2607659993898288e-05, "loss": 0.0148, "step": 99855 }, { "epoch": 2.33, "learning_rate": 1.2606779939452255e-05, "loss": 0.038, "step": 99860 }, { "epoch": 2.33, "learning_rate": 1.2605899885006218e-05, "loss": 0.101, "step": 99865 }, { "epoch": 2.33, "learning_rate": 1.2605019830560185e-05, "loss": 0.3556, "step": 99870 }, { "epoch": 2.33, "learning_rate": 1.2604139776114148e-05, "loss": 0.2597, "step": 99875 }, { "epoch": 2.33, "learning_rate": 1.2603259721668113e-05, "loss": 0.0615, "step": 99880 }, { "epoch": 2.33, "learning_rate": 1.260237966722208e-05, "loss": 0.0922, "step": 99885 }, { "epoch": 2.33, "learning_rate": 1.2601499612776043e-05, "loss": 0.068, "step": 99890 }, { "epoch": 2.33, "learning_rate": 1.260061955833001e-05, "loss": 0.0149, "step": 99895 }, { "epoch": 2.33, "learning_rate": 1.2599739503883973e-05, "loss": 0.0655, "step": 99900 }, { "epoch": 2.33, "learning_rate": 1.259885944943794e-05, "loss": 0.1243, "step": 99905 }, { "epoch": 2.33, "learning_rate": 1.2597979394991903e-05, "loss": 0.0689, "step": 99910 }, { "epoch": 2.33, "learning_rate": 1.259709934054587e-05, "loss": 0.1115, "step": 99915 }, { "epoch": 2.33, "learning_rate": 1.2596219286099833e-05, "loss": 0.3715, "step": 99920 }, { "epoch": 2.33, "learning_rate": 1.2595339231653799e-05, "loss": 0.2249, "step": 99925 }, { "epoch": 2.33, "learning_rate": 1.2594459177207764e-05, "loss": 0.0941, "step": 99930 }, { "epoch": 2.33, "learning_rate": 1.2593579122761729e-05, "loss": 0.0074, "step": 99935 }, { "epoch": 2.33, "learning_rate": 1.2592699068315694e-05, "loss": 0.0848, "step": 99940 }, { "epoch": 2.33, "learning_rate": 1.2591819013869659e-05, "loss": 0.051, "step": 99945 }, { "epoch": 2.33, "learning_rate": 1.2590938959423624e-05, "loss": 0.1468, "step": 99950 }, { "epoch": 2.33, "learning_rate": 1.2590058904977589e-05, "loss": 0.0843, "step": 99955 }, { "epoch": 2.33, "learning_rate": 1.2589178850531552e-05, "loss": 0.1344, "step": 99960 }, { "epoch": 2.33, "learning_rate": 1.2588298796085519e-05, "loss": 0.0963, "step": 99965 }, { "epoch": 2.33, "learning_rate": 1.2587418741639482e-05, "loss": 0.6725, "step": 99970 }, { "epoch": 2.33, "learning_rate": 1.2586538687193449e-05, "loss": 0.5605, "step": 99975 }, { "epoch": 2.33, "learning_rate": 1.2585658632747412e-05, "loss": 0.0211, "step": 99980 }, { "epoch": 2.33, "learning_rate": 1.2584778578301379e-05, "loss": 0.0025, "step": 99985 }, { "epoch": 2.33, "learning_rate": 1.2583898523855342e-05, "loss": 0.0256, "step": 99990 }, { "epoch": 2.33, "learning_rate": 1.2583018469409307e-05, "loss": 0.0325, "step": 99995 }, { "epoch": 2.33, "learning_rate": 1.2582138414963272e-05, "loss": 0.0259, "step": 100000 }, { "epoch": 2.33, "learning_rate": 1.2581258360517237e-05, "loss": 0.064, "step": 100005 }, { "epoch": 2.33, "learning_rate": 1.2580378306071202e-05, "loss": 0.2318, "step": 100010 }, { "epoch": 2.33, "learning_rate": 1.2579498251625167e-05, "loss": 0.2462, "step": 100015 }, { "epoch": 2.33, "learning_rate": 1.2578618197179134e-05, "loss": 0.2745, "step": 100020 }, { "epoch": 2.33, "learning_rate": 1.2577738142733097e-05, "loss": 0.3826, "step": 100025 }, { "epoch": 2.33, "learning_rate": 1.2576858088287062e-05, "loss": 0.0082, "step": 100030 }, { "epoch": 2.33, "learning_rate": 1.2575978033841027e-05, "loss": 0.154, "step": 100035 }, { "epoch": 2.33, "learning_rate": 1.2575097979394992e-05, "loss": 0.0346, "step": 100040 }, { "epoch": 2.33, "learning_rate": 1.2574217924948957e-05, "loss": 0.0756, "step": 100045 }, { "epoch": 2.33, "learning_rate": 1.2573337870502922e-05, "loss": 0.0957, "step": 100050 }, { "epoch": 2.33, "learning_rate": 1.2572457816056887e-05, "loss": 0.0954, "step": 100055 }, { "epoch": 2.33, "learning_rate": 1.2571577761610852e-05, "loss": 0.073, "step": 100060 }, { "epoch": 2.33, "learning_rate": 1.2570697707164817e-05, "loss": 0.0896, "step": 100065 }, { "epoch": 2.33, "learning_rate": 1.2569817652718782e-05, "loss": 0.4154, "step": 100070 }, { "epoch": 2.33, "learning_rate": 1.2568937598272746e-05, "loss": 0.4047, "step": 100075 }, { "epoch": 2.33, "learning_rate": 1.2568057543826712e-05, "loss": 0.071, "step": 100080 }, { "epoch": 2.34, "learning_rate": 1.2567177489380676e-05, "loss": 0.003, "step": 100085 }, { "epoch": 2.34, "learning_rate": 1.2566297434934642e-05, "loss": 0.0425, "step": 100090 }, { "epoch": 2.34, "learning_rate": 1.2565417380488606e-05, "loss": 0.0382, "step": 100095 }, { "epoch": 2.34, "learning_rate": 1.2564537326042573e-05, "loss": 0.032, "step": 100100 }, { "epoch": 2.34, "learning_rate": 1.2563657271596536e-05, "loss": 0.0522, "step": 100105 }, { "epoch": 2.34, "learning_rate": 1.2562777217150501e-05, "loss": 0.1143, "step": 100110 }, { "epoch": 2.34, "learning_rate": 1.2561897162704466e-05, "loss": 0.1467, "step": 100115 }, { "epoch": 2.34, "learning_rate": 1.2561017108258431e-05, "loss": 0.2772, "step": 100120 }, { "epoch": 2.34, "learning_rate": 1.2560137053812396e-05, "loss": 0.4843, "step": 100125 }, { "epoch": 2.34, "learning_rate": 1.2559256999366361e-05, "loss": 0.0388, "step": 100130 }, { "epoch": 2.34, "learning_rate": 1.2558376944920326e-05, "loss": 0.0617, "step": 100135 }, { "epoch": 2.34, "learning_rate": 1.2557496890474291e-05, "loss": 0.0266, "step": 100140 }, { "epoch": 2.34, "learning_rate": 1.2556616836028254e-05, "loss": 0.028, "step": 100145 }, { "epoch": 2.34, "learning_rate": 1.2555736781582221e-05, "loss": 0.0469, "step": 100150 }, { "epoch": 2.34, "learning_rate": 1.2554856727136186e-05, "loss": 0.1846, "step": 100155 }, { "epoch": 2.34, "learning_rate": 1.2553976672690151e-05, "loss": 0.076, "step": 100160 }, { "epoch": 2.34, "learning_rate": 1.2553096618244116e-05, "loss": 0.1784, "step": 100165 }, { "epoch": 2.34, "learning_rate": 1.2552216563798081e-05, "loss": 0.138, "step": 100170 }, { "epoch": 2.34, "learning_rate": 1.2551336509352046e-05, "loss": 0.2306, "step": 100175 }, { "epoch": 2.34, "learning_rate": 1.255045645490601e-05, "loss": 0.0737, "step": 100180 }, { "epoch": 2.34, "learning_rate": 1.2549576400459976e-05, "loss": 0.0702, "step": 100185 }, { "epoch": 2.34, "learning_rate": 1.254869634601394e-05, "loss": 0.015, "step": 100190 }, { "epoch": 2.34, "learning_rate": 1.2547816291567906e-05, "loss": 0.0296, "step": 100195 }, { "epoch": 2.34, "learning_rate": 1.254693623712187e-05, "loss": 0.0474, "step": 100200 }, { "epoch": 2.34, "learning_rate": 1.2546056182675836e-05, "loss": 0.0748, "step": 100205 }, { "epoch": 2.34, "learning_rate": 1.25451761282298e-05, "loss": 0.1458, "step": 100210 }, { "epoch": 2.34, "learning_rate": 1.2544296073783766e-05, "loss": 0.1964, "step": 100215 }, { "epoch": 2.34, "learning_rate": 1.254341601933773e-05, "loss": 0.1306, "step": 100220 }, { "epoch": 2.34, "learning_rate": 1.2542535964891695e-05, "loss": 0.1824, "step": 100225 }, { "epoch": 2.34, "learning_rate": 1.254165591044566e-05, "loss": 0.0379, "step": 100230 }, { "epoch": 2.34, "learning_rate": 1.2540775855999625e-05, "loss": 0.021, "step": 100235 }, { "epoch": 2.34, "learning_rate": 1.253989580155359e-05, "loss": 0.0014, "step": 100240 }, { "epoch": 2.34, "learning_rate": 1.2539015747107555e-05, "loss": 0.1089, "step": 100245 }, { "epoch": 2.34, "learning_rate": 1.253813569266152e-05, "loss": 0.0653, "step": 100250 }, { "epoch": 2.34, "learning_rate": 1.2537255638215485e-05, "loss": 0.1823, "step": 100255 }, { "epoch": 2.34, "learning_rate": 1.2536375583769448e-05, "loss": 0.1888, "step": 100260 }, { "epoch": 2.34, "learning_rate": 1.2535495529323415e-05, "loss": 0.2039, "step": 100265 }, { "epoch": 2.34, "learning_rate": 1.2534615474877378e-05, "loss": 0.3313, "step": 100270 }, { "epoch": 2.34, "learning_rate": 1.2533735420431345e-05, "loss": 0.2297, "step": 100275 }, { "epoch": 2.34, "learning_rate": 1.2532855365985308e-05, "loss": 0.0486, "step": 100280 }, { "epoch": 2.34, "learning_rate": 1.2531975311539275e-05, "loss": 0.0071, "step": 100285 }, { "epoch": 2.34, "learning_rate": 1.253109525709324e-05, "loss": 0.0554, "step": 100290 }, { "epoch": 2.34, "learning_rate": 1.2530215202647203e-05, "loss": 0.1471, "step": 100295 }, { "epoch": 2.34, "learning_rate": 1.252933514820117e-05, "loss": 0.0627, "step": 100300 }, { "epoch": 2.34, "learning_rate": 1.2528455093755133e-05, "loss": 0.0377, "step": 100305 }, { "epoch": 2.34, "learning_rate": 1.25275750393091e-05, "loss": 0.0985, "step": 100310 }, { "epoch": 2.34, "learning_rate": 1.2526694984863063e-05, "loss": 0.1178, "step": 100315 }, { "epoch": 2.34, "learning_rate": 1.252581493041703e-05, "loss": 0.139, "step": 100320 }, { "epoch": 2.34, "learning_rate": 1.2524934875970993e-05, "loss": 0.1996, "step": 100325 }, { "epoch": 2.34, "learning_rate": 1.2524054821524958e-05, "loss": 0.0461, "step": 100330 }, { "epoch": 2.34, "learning_rate": 1.2523174767078923e-05, "loss": 0.0488, "step": 100335 }, { "epoch": 2.34, "learning_rate": 1.2522294712632888e-05, "loss": 0.0163, "step": 100340 }, { "epoch": 2.34, "learning_rate": 1.2521414658186853e-05, "loss": 0.0311, "step": 100345 }, { "epoch": 2.34, "learning_rate": 1.2520534603740818e-05, "loss": 0.0665, "step": 100350 }, { "epoch": 2.34, "learning_rate": 1.2519654549294784e-05, "loss": 0.105, "step": 100355 }, { "epoch": 2.34, "learning_rate": 1.2518774494848749e-05, "loss": 0.0956, "step": 100360 }, { "epoch": 2.34, "learning_rate": 1.2517894440402714e-05, "loss": 0.1567, "step": 100365 }, { "epoch": 2.34, "learning_rate": 1.2517014385956679e-05, "loss": 0.318, "step": 100370 }, { "epoch": 2.34, "learning_rate": 1.2516134331510642e-05, "loss": 0.2459, "step": 100375 }, { "epoch": 2.34, "learning_rate": 1.2515254277064609e-05, "loss": 0.124, "step": 100380 }, { "epoch": 2.34, "learning_rate": 1.2514374222618572e-05, "loss": 0.0281, "step": 100385 }, { "epoch": 2.34, "learning_rate": 1.2513494168172539e-05, "loss": 0.0197, "step": 100390 }, { "epoch": 2.34, "learning_rate": 1.2512614113726502e-05, "loss": 0.0462, "step": 100395 }, { "epoch": 2.34, "learning_rate": 1.2511734059280469e-05, "loss": 0.0429, "step": 100400 }, { "epoch": 2.34, "learning_rate": 1.2510854004834432e-05, "loss": 0.1588, "step": 100405 }, { "epoch": 2.34, "learning_rate": 1.2509973950388397e-05, "loss": 0.075, "step": 100410 }, { "epoch": 2.34, "learning_rate": 1.2509093895942362e-05, "loss": 0.2112, "step": 100415 }, { "epoch": 2.34, "learning_rate": 1.2508213841496327e-05, "loss": 0.2562, "step": 100420 }, { "epoch": 2.34, "learning_rate": 1.2507333787050294e-05, "loss": 0.1556, "step": 100425 }, { "epoch": 2.34, "learning_rate": 1.2506453732604257e-05, "loss": 0.0339, "step": 100430 }, { "epoch": 2.34, "learning_rate": 1.2505573678158224e-05, "loss": 0.0323, "step": 100435 }, { "epoch": 2.34, "learning_rate": 1.2504693623712187e-05, "loss": 0.0492, "step": 100440 }, { "epoch": 2.34, "learning_rate": 1.2503813569266152e-05, "loss": 0.1222, "step": 100445 }, { "epoch": 2.34, "learning_rate": 1.2502933514820117e-05, "loss": 0.1036, "step": 100450 }, { "epoch": 2.34, "learning_rate": 1.2502053460374082e-05, "loss": 0.0938, "step": 100455 }, { "epoch": 2.34, "learning_rate": 1.2501173405928047e-05, "loss": 0.1337, "step": 100460 }, { "epoch": 2.34, "learning_rate": 1.2500293351482012e-05, "loss": 0.2168, "step": 100465 }, { "epoch": 2.34, "learning_rate": 1.2499413297035977e-05, "loss": 0.19, "step": 100470 }, { "epoch": 2.34, "learning_rate": 1.2498533242589942e-05, "loss": 0.4909, "step": 100475 }, { "epoch": 2.34, "learning_rate": 1.2497653188143906e-05, "loss": 0.053, "step": 100480 }, { "epoch": 2.34, "learning_rate": 1.2496773133697872e-05, "loss": 0.0419, "step": 100485 }, { "epoch": 2.34, "learning_rate": 1.2495893079251836e-05, "loss": 0.0198, "step": 100490 }, { "epoch": 2.34, "learning_rate": 1.2495013024805802e-05, "loss": 0.0683, "step": 100495 }, { "epoch": 2.34, "learning_rate": 1.2494132970359766e-05, "loss": 0.0714, "step": 100500 }, { "epoch": 2.34, "learning_rate": 1.2493252915913732e-05, "loss": 0.0683, "step": 100505 }, { "epoch": 2.35, "learning_rate": 1.2492372861467696e-05, "loss": 0.0206, "step": 100510 }, { "epoch": 2.35, "learning_rate": 1.2491492807021662e-05, "loss": 0.148, "step": 100515 }, { "epoch": 2.35, "learning_rate": 1.2490612752575626e-05, "loss": 0.2136, "step": 100520 }, { "epoch": 2.35, "learning_rate": 1.248973269812959e-05, "loss": 0.1833, "step": 100525 }, { "epoch": 2.35, "learning_rate": 1.2488852643683556e-05, "loss": 0.0771, "step": 100530 }, { "epoch": 2.35, "learning_rate": 1.2487972589237521e-05, "loss": 0.0387, "step": 100535 }, { "epoch": 2.35, "learning_rate": 1.2487092534791486e-05, "loss": 0.0434, "step": 100540 }, { "epoch": 2.35, "learning_rate": 1.2486212480345451e-05, "loss": 0.0414, "step": 100545 }, { "epoch": 2.35, "learning_rate": 1.2485332425899416e-05, "loss": 0.0964, "step": 100550 }, { "epoch": 2.35, "learning_rate": 1.2484452371453381e-05, "loss": 0.1329, "step": 100555 }, { "epoch": 2.35, "learning_rate": 1.2483572317007346e-05, "loss": 0.1486, "step": 100560 }, { "epoch": 2.35, "learning_rate": 1.2482692262561311e-05, "loss": 0.3763, "step": 100565 }, { "epoch": 2.35, "learning_rate": 1.2481812208115276e-05, "loss": 0.2819, "step": 100570 }, { "epoch": 2.35, "learning_rate": 1.2480932153669241e-05, "loss": 0.1683, "step": 100575 }, { "epoch": 2.35, "learning_rate": 1.2480052099223206e-05, "loss": 0.024, "step": 100580 }, { "epoch": 2.35, "learning_rate": 1.2479172044777171e-05, "loss": 0.1077, "step": 100585 }, { "epoch": 2.35, "learning_rate": 1.2478291990331136e-05, "loss": 0.0406, "step": 100590 }, { "epoch": 2.35, "learning_rate": 1.24774119358851e-05, "loss": 0.0719, "step": 100595 }, { "epoch": 2.35, "learning_rate": 1.2476531881439066e-05, "loss": 0.048, "step": 100600 }, { "epoch": 2.35, "learning_rate": 1.247565182699303e-05, "loss": 0.0894, "step": 100605 }, { "epoch": 2.35, "learning_rate": 1.2474771772546996e-05, "loss": 0.2448, "step": 100610 }, { "epoch": 2.35, "learning_rate": 1.247389171810096e-05, "loss": 0.218, "step": 100615 }, { "epoch": 2.35, "learning_rate": 1.2473011663654926e-05, "loss": 0.2413, "step": 100620 }, { "epoch": 2.35, "learning_rate": 1.247213160920889e-05, "loss": 0.2865, "step": 100625 }, { "epoch": 2.35, "learning_rate": 1.2471251554762855e-05, "loss": 0.1294, "step": 100630 }, { "epoch": 2.35, "learning_rate": 1.247037150031682e-05, "loss": 0.0441, "step": 100635 }, { "epoch": 2.35, "learning_rate": 1.2469491445870785e-05, "loss": 0.0194, "step": 100640 }, { "epoch": 2.35, "learning_rate": 1.246861139142475e-05, "loss": 0.0538, "step": 100645 }, { "epoch": 2.35, "learning_rate": 1.2467731336978715e-05, "loss": 0.0978, "step": 100650 }, { "epoch": 2.35, "learning_rate": 1.246685128253268e-05, "loss": 0.1239, "step": 100655 }, { "epoch": 2.35, "learning_rate": 1.2465971228086645e-05, "loss": 0.1405, "step": 100660 }, { "epoch": 2.35, "learning_rate": 1.246509117364061e-05, "loss": 0.0761, "step": 100665 }, { "epoch": 2.35, "learning_rate": 1.2464211119194575e-05, "loss": 0.2181, "step": 100670 }, { "epoch": 2.35, "learning_rate": 1.2463331064748538e-05, "loss": 0.3131, "step": 100675 }, { "epoch": 2.35, "learning_rate": 1.2462451010302505e-05, "loss": 0.054, "step": 100680 }, { "epoch": 2.35, "learning_rate": 1.2461570955856468e-05, "loss": 0.0591, "step": 100685 }, { "epoch": 2.35, "learning_rate": 1.2460690901410435e-05, "loss": 0.0284, "step": 100690 }, { "epoch": 2.35, "learning_rate": 1.24598108469644e-05, "loss": 0.0883, "step": 100695 }, { "epoch": 2.35, "learning_rate": 1.2458930792518365e-05, "loss": 0.036, "step": 100700 }, { "epoch": 2.35, "learning_rate": 1.245805073807233e-05, "loss": 0.0074, "step": 100705 }, { "epoch": 2.35, "learning_rate": 1.2457170683626293e-05, "loss": 0.1747, "step": 100710 }, { "epoch": 2.35, "learning_rate": 1.245629062918026e-05, "loss": 0.2058, "step": 100715 }, { "epoch": 2.35, "learning_rate": 1.2455410574734223e-05, "loss": 0.1014, "step": 100720 }, { "epoch": 2.35, "learning_rate": 1.245453052028819e-05, "loss": 0.2558, "step": 100725 }, { "epoch": 2.35, "learning_rate": 1.2453650465842153e-05, "loss": 0.0813, "step": 100730 }, { "epoch": 2.35, "learning_rate": 1.245277041139612e-05, "loss": 0.0342, "step": 100735 }, { "epoch": 2.35, "learning_rate": 1.2451890356950083e-05, "loss": 0.0892, "step": 100740 }, { "epoch": 2.35, "learning_rate": 1.2451010302504048e-05, "loss": 0.0966, "step": 100745 }, { "epoch": 2.35, "learning_rate": 1.2450130248058013e-05, "loss": 0.0308, "step": 100750 }, { "epoch": 2.35, "learning_rate": 1.2449250193611978e-05, "loss": 0.1035, "step": 100755 }, { "epoch": 2.35, "learning_rate": 1.2448370139165943e-05, "loss": 0.1608, "step": 100760 }, { "epoch": 2.35, "learning_rate": 1.2447490084719908e-05, "loss": 0.1588, "step": 100765 }, { "epoch": 2.35, "learning_rate": 1.2446610030273873e-05, "loss": 0.2393, "step": 100770 }, { "epoch": 2.35, "learning_rate": 1.2445729975827838e-05, "loss": 0.2925, "step": 100775 }, { "epoch": 2.35, "learning_rate": 1.2444849921381802e-05, "loss": 0.1282, "step": 100780 }, { "epoch": 2.35, "learning_rate": 1.2443969866935768e-05, "loss": 0.0151, "step": 100785 }, { "epoch": 2.35, "learning_rate": 1.2443089812489732e-05, "loss": 0.0347, "step": 100790 }, { "epoch": 2.35, "learning_rate": 1.2442209758043699e-05, "loss": 0.0489, "step": 100795 }, { "epoch": 2.35, "learning_rate": 1.2441329703597662e-05, "loss": 0.0725, "step": 100800 }, { "epoch": 2.35, "learning_rate": 1.2440449649151629e-05, "loss": 0.0951, "step": 100805 }, { "epoch": 2.35, "learning_rate": 1.2439569594705592e-05, "loss": 0.0461, "step": 100810 }, { "epoch": 2.35, "learning_rate": 1.2438689540259559e-05, "loss": 0.1166, "step": 100815 }, { "epoch": 2.35, "learning_rate": 1.2437809485813522e-05, "loss": 0.2465, "step": 100820 }, { "epoch": 2.35, "learning_rate": 1.2436929431367487e-05, "loss": 0.2686, "step": 100825 }, { "epoch": 2.35, "learning_rate": 1.2436049376921454e-05, "loss": 0.0951, "step": 100830 }, { "epoch": 2.35, "learning_rate": 1.2435169322475417e-05, "loss": 0.0627, "step": 100835 }, { "epoch": 2.35, "learning_rate": 1.2434289268029384e-05, "loss": 0.1103, "step": 100840 }, { "epoch": 2.35, "learning_rate": 1.2433409213583347e-05, "loss": 0.1011, "step": 100845 }, { "epoch": 2.35, "learning_rate": 1.2432529159137314e-05, "loss": 0.1131, "step": 100850 }, { "epoch": 2.35, "learning_rate": 1.2431649104691277e-05, "loss": 0.1484, "step": 100855 }, { "epoch": 2.35, "learning_rate": 1.2430769050245242e-05, "loss": 0.1358, "step": 100860 }, { "epoch": 2.35, "learning_rate": 1.2429888995799207e-05, "loss": 0.1805, "step": 100865 }, { "epoch": 2.35, "learning_rate": 1.2429008941353172e-05, "loss": 0.2239, "step": 100870 }, { "epoch": 2.35, "learning_rate": 1.2428128886907137e-05, "loss": 0.2185, "step": 100875 }, { "epoch": 2.35, "learning_rate": 1.2427248832461102e-05, "loss": 0.0632, "step": 100880 }, { "epoch": 2.35, "learning_rate": 1.2426368778015067e-05, "loss": 0.0697, "step": 100885 }, { "epoch": 2.35, "learning_rate": 1.2425488723569032e-05, "loss": 0.0686, "step": 100890 }, { "epoch": 2.35, "learning_rate": 1.2424608669122996e-05, "loss": 0.0076, "step": 100895 }, { "epoch": 2.35, "learning_rate": 1.2423728614676962e-05, "loss": 0.0321, "step": 100900 }, { "epoch": 2.35, "learning_rate": 1.2422848560230926e-05, "loss": 0.0552, "step": 100905 }, { "epoch": 2.35, "learning_rate": 1.2421968505784892e-05, "loss": 0.0364, "step": 100910 }, { "epoch": 2.35, "learning_rate": 1.2421088451338856e-05, "loss": 0.1764, "step": 100915 }, { "epoch": 2.35, "learning_rate": 1.2420208396892822e-05, "loss": 0.2286, "step": 100920 }, { "epoch": 2.35, "learning_rate": 1.2419328342446786e-05, "loss": 0.1474, "step": 100925 }, { "epoch": 2.35, "learning_rate": 1.241844828800075e-05, "loss": 0.0703, "step": 100930 }, { "epoch": 2.35, "learning_rate": 1.2417568233554716e-05, "loss": 0.0086, "step": 100935 }, { "epoch": 2.36, "learning_rate": 1.241668817910868e-05, "loss": 0.0152, "step": 100940 }, { "epoch": 2.36, "learning_rate": 1.2415808124662646e-05, "loss": 0.0146, "step": 100945 }, { "epoch": 2.36, "learning_rate": 1.241492807021661e-05, "loss": 0.1069, "step": 100950 }, { "epoch": 2.36, "learning_rate": 1.2414048015770576e-05, "loss": 0.0378, "step": 100955 }, { "epoch": 2.36, "learning_rate": 1.241316796132454e-05, "loss": 0.0632, "step": 100960 }, { "epoch": 2.36, "learning_rate": 1.2412287906878506e-05, "loss": 0.0854, "step": 100965 }, { "epoch": 2.36, "learning_rate": 1.2411407852432471e-05, "loss": 0.2351, "step": 100970 }, { "epoch": 2.36, "learning_rate": 1.2410527797986436e-05, "loss": 0.2718, "step": 100975 }, { "epoch": 2.36, "learning_rate": 1.2409647743540401e-05, "loss": 0.0676, "step": 100980 }, { "epoch": 2.36, "learning_rate": 1.2408767689094366e-05, "loss": 0.0063, "step": 100985 }, { "epoch": 2.36, "learning_rate": 1.2407887634648331e-05, "loss": 0.0476, "step": 100990 }, { "epoch": 2.36, "learning_rate": 1.2407007580202296e-05, "loss": 0.1336, "step": 100995 }, { "epoch": 2.36, "learning_rate": 1.2406127525756261e-05, "loss": 0.0369, "step": 101000 }, { "epoch": 2.36, "learning_rate": 1.2405247471310226e-05, "loss": 0.1172, "step": 101005 }, { "epoch": 2.36, "learning_rate": 1.240436741686419e-05, "loss": 0.0753, "step": 101010 }, { "epoch": 2.36, "learning_rate": 1.2403487362418156e-05, "loss": 0.1736, "step": 101015 }, { "epoch": 2.36, "learning_rate": 1.240260730797212e-05, "loss": 0.1493, "step": 101020 }, { "epoch": 2.36, "learning_rate": 1.2401727253526086e-05, "loss": 0.2624, "step": 101025 }, { "epoch": 2.36, "learning_rate": 1.240084719908005e-05, "loss": 0.0056, "step": 101030 }, { "epoch": 2.36, "learning_rate": 1.2399967144634016e-05, "loss": 0.0314, "step": 101035 }, { "epoch": 2.36, "learning_rate": 1.239908709018798e-05, "loss": 0.0613, "step": 101040 }, { "epoch": 2.36, "learning_rate": 1.2398207035741944e-05, "loss": 0.0663, "step": 101045 }, { "epoch": 2.36, "learning_rate": 1.239732698129591e-05, "loss": 0.067, "step": 101050 }, { "epoch": 2.36, "learning_rate": 1.2396446926849875e-05, "loss": 0.0649, "step": 101055 }, { "epoch": 2.36, "learning_rate": 1.239556687240384e-05, "loss": 0.1078, "step": 101060 }, { "epoch": 2.36, "learning_rate": 1.2394686817957805e-05, "loss": 0.1997, "step": 101065 }, { "epoch": 2.36, "learning_rate": 1.239380676351177e-05, "loss": 0.0911, "step": 101070 }, { "epoch": 2.36, "learning_rate": 1.2392926709065735e-05, "loss": 0.1991, "step": 101075 }, { "epoch": 2.36, "learning_rate": 1.2392046654619698e-05, "loss": 0.1086, "step": 101080 }, { "epoch": 2.36, "learning_rate": 1.2391166600173665e-05, "loss": 0.0242, "step": 101085 }, { "epoch": 2.36, "learning_rate": 1.2390286545727628e-05, "loss": 0.0323, "step": 101090 }, { "epoch": 2.36, "learning_rate": 1.2389406491281595e-05, "loss": 0.0288, "step": 101095 }, { "epoch": 2.36, "learning_rate": 1.238852643683556e-05, "loss": 0.1087, "step": 101100 }, { "epoch": 2.36, "learning_rate": 1.2387646382389525e-05, "loss": 0.0695, "step": 101105 }, { "epoch": 2.36, "learning_rate": 1.238676632794349e-05, "loss": 0.248, "step": 101110 }, { "epoch": 2.36, "learning_rate": 1.2385886273497455e-05, "loss": 0.1033, "step": 101115 }, { "epoch": 2.36, "learning_rate": 1.238500621905142e-05, "loss": 0.1891, "step": 101120 }, { "epoch": 2.36, "learning_rate": 1.2384126164605383e-05, "loss": 0.3377, "step": 101125 }, { "epoch": 2.36, "learning_rate": 1.238324611015935e-05, "loss": 0.0188, "step": 101130 }, { "epoch": 2.36, "learning_rate": 1.2382366055713313e-05, "loss": 0.0164, "step": 101135 }, { "epoch": 2.36, "learning_rate": 1.238148600126728e-05, "loss": 0.0424, "step": 101140 }, { "epoch": 2.36, "learning_rate": 1.2380605946821243e-05, "loss": 0.1386, "step": 101145 }, { "epoch": 2.36, "learning_rate": 1.237972589237521e-05, "loss": 0.037, "step": 101150 }, { "epoch": 2.36, "learning_rate": 1.2378845837929173e-05, "loss": 0.0742, "step": 101155 }, { "epoch": 2.36, "learning_rate": 1.2377965783483138e-05, "loss": 0.1399, "step": 101160 }, { "epoch": 2.36, "learning_rate": 1.2377085729037103e-05, "loss": 0.0425, "step": 101165 }, { "epoch": 2.36, "learning_rate": 1.2376205674591068e-05, "loss": 0.1374, "step": 101170 }, { "epoch": 2.36, "learning_rate": 1.2375325620145033e-05, "loss": 0.2543, "step": 101175 }, { "epoch": 2.36, "learning_rate": 1.2374445565698998e-05, "loss": 0.09, "step": 101180 }, { "epoch": 2.36, "learning_rate": 1.2373565511252963e-05, "loss": 0.0816, "step": 101185 }, { "epoch": 2.36, "learning_rate": 1.2372685456806928e-05, "loss": 0.0108, "step": 101190 }, { "epoch": 2.36, "learning_rate": 1.2371805402360892e-05, "loss": 0.0545, "step": 101195 }, { "epoch": 2.36, "learning_rate": 1.2370925347914858e-05, "loss": 0.101, "step": 101200 }, { "epoch": 2.36, "learning_rate": 1.2370045293468822e-05, "loss": 0.1086, "step": 101205 }, { "epoch": 2.36, "learning_rate": 1.2369165239022788e-05, "loss": 0.1509, "step": 101210 }, { "epoch": 2.36, "learning_rate": 1.2368285184576752e-05, "loss": 0.1245, "step": 101215 }, { "epoch": 2.36, "learning_rate": 1.2367405130130718e-05, "loss": 0.1515, "step": 101220 }, { "epoch": 2.36, "learning_rate": 1.2366525075684682e-05, "loss": 0.3035, "step": 101225 }, { "epoch": 2.36, "learning_rate": 1.2365645021238647e-05, "loss": 0.0863, "step": 101230 }, { "epoch": 2.36, "learning_rate": 1.2364764966792612e-05, "loss": 0.0632, "step": 101235 }, { "epoch": 2.36, "learning_rate": 1.2363884912346577e-05, "loss": 0.0313, "step": 101240 }, { "epoch": 2.36, "learning_rate": 1.2363004857900544e-05, "loss": 0.0159, "step": 101245 }, { "epoch": 2.36, "learning_rate": 1.2362124803454507e-05, "loss": 0.0765, "step": 101250 }, { "epoch": 2.36, "learning_rate": 1.2361244749008474e-05, "loss": 0.038, "step": 101255 }, { "epoch": 2.36, "learning_rate": 1.2360364694562437e-05, "loss": 0.1353, "step": 101260 }, { "epoch": 2.36, "learning_rate": 1.2359484640116404e-05, "loss": 0.135, "step": 101265 }, { "epoch": 2.36, "learning_rate": 1.2358604585670367e-05, "loss": 0.1609, "step": 101270 }, { "epoch": 2.36, "learning_rate": 1.2357724531224332e-05, "loss": 0.3967, "step": 101275 }, { "epoch": 2.36, "learning_rate": 1.2356844476778297e-05, "loss": 0.0439, "step": 101280 }, { "epoch": 2.36, "learning_rate": 1.2355964422332262e-05, "loss": 0.0257, "step": 101285 }, { "epoch": 2.36, "learning_rate": 1.2355084367886227e-05, "loss": 0.0021, "step": 101290 }, { "epoch": 2.36, "learning_rate": 1.2354204313440192e-05, "loss": 0.0251, "step": 101295 }, { "epoch": 2.36, "learning_rate": 1.2353324258994157e-05, "loss": 0.0747, "step": 101300 }, { "epoch": 2.36, "learning_rate": 1.2352444204548122e-05, "loss": 0.0577, "step": 101305 }, { "epoch": 2.36, "learning_rate": 1.2351564150102085e-05, "loss": 0.1408, "step": 101310 }, { "epoch": 2.36, "learning_rate": 1.2350684095656052e-05, "loss": 0.1539, "step": 101315 }, { "epoch": 2.36, "learning_rate": 1.2349804041210016e-05, "loss": 0.2462, "step": 101320 }, { "epoch": 2.36, "learning_rate": 1.2348923986763982e-05, "loss": 0.2959, "step": 101325 }, { "epoch": 2.36, "learning_rate": 1.2348043932317946e-05, "loss": 0.0223, "step": 101330 }, { "epoch": 2.36, "learning_rate": 1.2347163877871912e-05, "loss": 0.0031, "step": 101335 }, { "epoch": 2.36, "learning_rate": 1.2346283823425876e-05, "loss": 0.0728, "step": 101340 }, { "epoch": 2.36, "learning_rate": 1.234540376897984e-05, "loss": 0.016, "step": 101345 }, { "epoch": 2.36, "learning_rate": 1.2344523714533806e-05, "loss": 0.0505, "step": 101350 }, { "epoch": 2.36, "learning_rate": 1.234364366008777e-05, "loss": 0.0636, "step": 101355 }, { "epoch": 2.36, "learning_rate": 1.2342763605641736e-05, "loss": 0.1364, "step": 101360 }, { "epoch": 2.36, "learning_rate": 1.23418835511957e-05, "loss": 0.1462, "step": 101365 }, { "epoch": 2.37, "learning_rate": 1.2341003496749666e-05, "loss": 0.1041, "step": 101370 }, { "epoch": 2.37, "learning_rate": 1.234012344230363e-05, "loss": 0.4263, "step": 101375 }, { "epoch": 2.37, "learning_rate": 1.2339243387857596e-05, "loss": 0.0154, "step": 101380 }, { "epoch": 2.37, "learning_rate": 1.233836333341156e-05, "loss": 0.0103, "step": 101385 }, { "epoch": 2.37, "learning_rate": 1.2337483278965526e-05, "loss": 0.0611, "step": 101390 }, { "epoch": 2.37, "learning_rate": 1.233660322451949e-05, "loss": 0.0948, "step": 101395 }, { "epoch": 2.37, "learning_rate": 1.2335723170073456e-05, "loss": 0.043, "step": 101400 }, { "epoch": 2.37, "learning_rate": 1.2334843115627421e-05, "loss": 0.162, "step": 101405 }, { "epoch": 2.37, "learning_rate": 1.2333963061181386e-05, "loss": 0.1158, "step": 101410 }, { "epoch": 2.37, "learning_rate": 1.2333083006735351e-05, "loss": 0.2062, "step": 101415 }, { "epoch": 2.37, "learning_rate": 1.2332202952289316e-05, "loss": 0.1363, "step": 101420 }, { "epoch": 2.37, "learning_rate": 1.233132289784328e-05, "loss": 0.3915, "step": 101425 }, { "epoch": 2.37, "learning_rate": 1.2330442843397246e-05, "loss": 0.0019, "step": 101430 }, { "epoch": 2.37, "learning_rate": 1.232956278895121e-05, "loss": 0.0287, "step": 101435 }, { "epoch": 2.37, "learning_rate": 1.2328682734505176e-05, "loss": 0.0248, "step": 101440 }, { "epoch": 2.37, "learning_rate": 1.232780268005914e-05, "loss": 0.0432, "step": 101445 }, { "epoch": 2.37, "learning_rate": 1.2326922625613106e-05, "loss": 0.0189, "step": 101450 }, { "epoch": 2.37, "learning_rate": 1.232604257116707e-05, "loss": 0.1051, "step": 101455 }, { "epoch": 2.37, "learning_rate": 1.2325162516721034e-05, "loss": 0.1388, "step": 101460 }, { "epoch": 2.37, "learning_rate": 1.2324458473164207e-05, "loss": 0.1825, "step": 101465 }, { "epoch": 2.37, "learning_rate": 1.232357841871817e-05, "loss": 0.1956, "step": 101470 }, { "epoch": 2.37, "learning_rate": 1.2322698364272137e-05, "loss": 0.3224, "step": 101475 }, { "epoch": 2.37, "learning_rate": 1.23218183098261e-05, "loss": 0.0054, "step": 101480 }, { "epoch": 2.37, "learning_rate": 1.2320938255380068e-05, "loss": 0.0744, "step": 101485 }, { "epoch": 2.37, "learning_rate": 1.232005820093403e-05, "loss": 0.0457, "step": 101490 }, { "epoch": 2.37, "learning_rate": 1.2319178146487998e-05, "loss": 0.0476, "step": 101495 }, { "epoch": 2.37, "learning_rate": 1.2318298092041961e-05, "loss": 0.0974, "step": 101500 }, { "epoch": 2.37, "learning_rate": 1.2317418037595926e-05, "loss": 0.1065, "step": 101505 }, { "epoch": 2.37, "learning_rate": 1.2316537983149891e-05, "loss": 0.3389, "step": 101510 }, { "epoch": 2.37, "learning_rate": 1.2315657928703856e-05, "loss": 0.2433, "step": 101515 }, { "epoch": 2.37, "learning_rate": 1.2314777874257821e-05, "loss": 0.2176, "step": 101520 }, { "epoch": 2.37, "learning_rate": 1.2313897819811786e-05, "loss": 0.4625, "step": 101525 }, { "epoch": 2.37, "learning_rate": 1.2313017765365751e-05, "loss": 0.0335, "step": 101530 }, { "epoch": 2.37, "learning_rate": 1.2312137710919716e-05, "loss": 0.0527, "step": 101535 }, { "epoch": 2.37, "learning_rate": 1.231125765647368e-05, "loss": 0.0347, "step": 101540 }, { "epoch": 2.37, "learning_rate": 1.2310377602027646e-05, "loss": 0.0542, "step": 101545 }, { "epoch": 2.37, "learning_rate": 1.230949754758161e-05, "loss": 0.1418, "step": 101550 }, { "epoch": 2.37, "learning_rate": 1.2308617493135576e-05, "loss": 0.075, "step": 101555 }, { "epoch": 2.37, "learning_rate": 1.230773743868954e-05, "loss": 0.0131, "step": 101560 }, { "epoch": 2.37, "learning_rate": 1.2306857384243506e-05, "loss": 0.3019, "step": 101565 }, { "epoch": 2.37, "learning_rate": 1.2305977329797471e-05, "loss": 0.1848, "step": 101570 }, { "epoch": 2.37, "learning_rate": 1.2305097275351434e-05, "loss": 0.2591, "step": 101575 }, { "epoch": 2.37, "learning_rate": 1.2304217220905401e-05, "loss": 0.0432, "step": 101580 }, { "epoch": 2.37, "learning_rate": 1.2303337166459365e-05, "loss": 0.002, "step": 101585 }, { "epoch": 2.37, "learning_rate": 1.2302457112013331e-05, "loss": 0.0141, "step": 101590 }, { "epoch": 2.37, "learning_rate": 1.2301577057567295e-05, "loss": 0.0701, "step": 101595 }, { "epoch": 2.37, "learning_rate": 1.2300697003121261e-05, "loss": 0.0603, "step": 101600 }, { "epoch": 2.37, "learning_rate": 1.2299816948675225e-05, "loss": 0.1127, "step": 101605 }, { "epoch": 2.37, "learning_rate": 1.2298936894229191e-05, "loss": 0.1118, "step": 101610 }, { "epoch": 2.37, "learning_rate": 1.2298056839783155e-05, "loss": 0.1666, "step": 101615 }, { "epoch": 2.37, "learning_rate": 1.229717678533712e-05, "loss": 0.1993, "step": 101620 }, { "epoch": 2.37, "learning_rate": 1.2296296730891085e-05, "loss": 0.3575, "step": 101625 }, { "epoch": 2.37, "learning_rate": 1.229541667644505e-05, "loss": 0.0214, "step": 101630 }, { "epoch": 2.37, "learning_rate": 1.2294536621999015e-05, "loss": 0.0504, "step": 101635 }, { "epoch": 2.37, "learning_rate": 1.229365656755298e-05, "loss": 0.0025, "step": 101640 }, { "epoch": 2.37, "learning_rate": 1.2292776513106945e-05, "loss": 0.0807, "step": 101645 }, { "epoch": 2.37, "learning_rate": 1.229189645866091e-05, "loss": 0.0521, "step": 101650 }, { "epoch": 2.37, "learning_rate": 1.2291016404214873e-05, "loss": 0.2139, "step": 101655 }, { "epoch": 2.37, "learning_rate": 1.229013634976884e-05, "loss": 0.0542, "step": 101660 }, { "epoch": 2.37, "learning_rate": 1.2289256295322803e-05, "loss": 0.0834, "step": 101665 }, { "epoch": 2.37, "learning_rate": 1.228837624087677e-05, "loss": 0.4055, "step": 101670 }, { "epoch": 2.37, "learning_rate": 1.2287496186430733e-05, "loss": 0.201, "step": 101675 }, { "epoch": 2.37, "learning_rate": 1.22866161319847e-05, "loss": 0.0252, "step": 101680 }, { "epoch": 2.37, "learning_rate": 1.2285736077538663e-05, "loss": 0.0806, "step": 101685 }, { "epoch": 2.37, "learning_rate": 1.2284856023092628e-05, "loss": 0.0471, "step": 101690 }, { "epoch": 2.37, "learning_rate": 1.2283975968646593e-05, "loss": 0.1179, "step": 101695 }, { "epoch": 2.37, "learning_rate": 1.2283095914200558e-05, "loss": 0.0381, "step": 101700 }, { "epoch": 2.37, "learning_rate": 1.2282215859754525e-05, "loss": 0.1436, "step": 101705 }, { "epoch": 2.37, "learning_rate": 1.2281335805308488e-05, "loss": 0.1922, "step": 101710 }, { "epoch": 2.37, "learning_rate": 1.2280455750862455e-05, "loss": 0.2683, "step": 101715 }, { "epoch": 2.37, "learning_rate": 1.2279575696416418e-05, "loss": 0.0756, "step": 101720 }, { "epoch": 2.37, "learning_rate": 1.2278695641970383e-05, "loss": 0.1447, "step": 101725 }, { "epoch": 2.37, "learning_rate": 1.2277815587524348e-05, "loss": 0.0239, "step": 101730 }, { "epoch": 2.37, "learning_rate": 1.2276935533078313e-05, "loss": 0.0172, "step": 101735 }, { "epoch": 2.37, "learning_rate": 1.2276055478632278e-05, "loss": 0.0404, "step": 101740 }, { "epoch": 2.37, "learning_rate": 1.2275175424186244e-05, "loss": 0.0606, "step": 101745 }, { "epoch": 2.37, "learning_rate": 1.2274295369740209e-05, "loss": 0.0968, "step": 101750 }, { "epoch": 2.37, "learning_rate": 1.2273415315294174e-05, "loss": 0.0994, "step": 101755 }, { "epoch": 2.37, "learning_rate": 1.2272535260848139e-05, "loss": 0.2048, "step": 101760 }, { "epoch": 2.37, "learning_rate": 1.2271655206402104e-05, "loss": 0.2327, "step": 101765 }, { "epoch": 2.37, "learning_rate": 1.2270775151956067e-05, "loss": 0.5181, "step": 101770 }, { "epoch": 2.37, "learning_rate": 1.2269895097510034e-05, "loss": 0.2584, "step": 101775 }, { "epoch": 2.37, "learning_rate": 1.2269015043063997e-05, "loss": 0.0355, "step": 101780 }, { "epoch": 2.37, "learning_rate": 1.2268134988617964e-05, "loss": 0.0946, "step": 101785 }, { "epoch": 2.37, "learning_rate": 1.2267254934171927e-05, "loss": 0.0719, "step": 101790 }, { "epoch": 2.37, "learning_rate": 1.2266374879725894e-05, "loss": 0.0226, "step": 101795 }, { "epoch": 2.38, "learning_rate": 1.2265494825279857e-05, "loss": 0.0328, "step": 101800 }, { "epoch": 2.38, "learning_rate": 1.2264614770833822e-05, "loss": 0.0495, "step": 101805 }, { "epoch": 2.38, "learning_rate": 1.2263734716387787e-05, "loss": 0.0813, "step": 101810 }, { "epoch": 2.38, "learning_rate": 1.2262854661941752e-05, "loss": 0.0203, "step": 101815 }, { "epoch": 2.38, "learning_rate": 1.2261974607495717e-05, "loss": 0.2045, "step": 101820 }, { "epoch": 2.38, "learning_rate": 1.2261094553049682e-05, "loss": 0.2168, "step": 101825 }, { "epoch": 2.38, "learning_rate": 1.2260214498603647e-05, "loss": 0.0096, "step": 101830 }, { "epoch": 2.38, "learning_rate": 1.2259334444157612e-05, "loss": 0.0036, "step": 101835 }, { "epoch": 2.38, "learning_rate": 1.2258454389711577e-05, "loss": 0.1481, "step": 101840 }, { "epoch": 2.38, "learning_rate": 1.2257574335265542e-05, "loss": 0.1383, "step": 101845 }, { "epoch": 2.38, "learning_rate": 1.2256694280819507e-05, "loss": 0.0671, "step": 101850 }, { "epoch": 2.38, "learning_rate": 1.2255814226373472e-05, "loss": 0.0776, "step": 101855 }, { "epoch": 2.38, "learning_rate": 1.2254934171927437e-05, "loss": 0.1976, "step": 101860 }, { "epoch": 2.38, "learning_rate": 1.2254054117481402e-05, "loss": 0.0716, "step": 101865 }, { "epoch": 2.38, "learning_rate": 1.2253174063035367e-05, "loss": 0.2836, "step": 101870 }, { "epoch": 2.38, "learning_rate": 1.225229400858933e-05, "loss": 0.2853, "step": 101875 }, { "epoch": 2.38, "learning_rate": 1.2251413954143297e-05, "loss": 0.0149, "step": 101880 }, { "epoch": 2.38, "learning_rate": 1.225053389969726e-05, "loss": 0.0101, "step": 101885 }, { "epoch": 2.38, "learning_rate": 1.2249653845251227e-05, "loss": 0.0516, "step": 101890 }, { "epoch": 2.38, "learning_rate": 1.224877379080519e-05, "loss": 0.0965, "step": 101895 }, { "epoch": 2.38, "learning_rate": 1.2247893736359157e-05, "loss": 0.0422, "step": 101900 }, { "epoch": 2.38, "learning_rate": 1.224701368191312e-05, "loss": 0.0546, "step": 101905 }, { "epoch": 2.38, "learning_rate": 1.2246133627467087e-05, "loss": 0.2696, "step": 101910 }, { "epoch": 2.38, "learning_rate": 1.224525357302105e-05, "loss": 0.1835, "step": 101915 }, { "epoch": 2.38, "learning_rate": 1.2244373518575016e-05, "loss": 0.1931, "step": 101920 }, { "epoch": 2.38, "learning_rate": 1.2243493464128981e-05, "loss": 0.2284, "step": 101925 }, { "epoch": 2.38, "learning_rate": 1.2242613409682946e-05, "loss": 0.0517, "step": 101930 }, { "epoch": 2.38, "learning_rate": 1.2241733355236911e-05, "loss": 0.037, "step": 101935 }, { "epoch": 2.38, "learning_rate": 1.2240853300790876e-05, "loss": 0.0358, "step": 101940 }, { "epoch": 2.38, "learning_rate": 1.2239973246344841e-05, "loss": 0.0706, "step": 101945 }, { "epoch": 2.38, "learning_rate": 1.2239093191898806e-05, "loss": 0.026, "step": 101950 }, { "epoch": 2.38, "learning_rate": 1.223821313745277e-05, "loss": 0.0276, "step": 101955 }, { "epoch": 2.38, "learning_rate": 1.2237333083006736e-05, "loss": 0.0858, "step": 101960 }, { "epoch": 2.38, "learning_rate": 1.22364530285607e-05, "loss": 0.1061, "step": 101965 }, { "epoch": 2.38, "learning_rate": 1.2235572974114666e-05, "loss": 0.2989, "step": 101970 }, { "epoch": 2.38, "learning_rate": 1.2234692919668631e-05, "loss": 0.2822, "step": 101975 }, { "epoch": 2.38, "learning_rate": 1.2233812865222596e-05, "loss": 0.0621, "step": 101980 }, { "epoch": 2.38, "learning_rate": 1.2232932810776561e-05, "loss": 0.0281, "step": 101985 }, { "epoch": 2.38, "learning_rate": 1.2232052756330524e-05, "loss": 0.0726, "step": 101990 }, { "epoch": 2.38, "learning_rate": 1.2231172701884491e-05, "loss": 0.3001, "step": 101995 }, { "epoch": 2.38, "learning_rate": 1.2230292647438454e-05, "loss": 0.0497, "step": 102000 }, { "epoch": 2.38, "learning_rate": 1.2229412592992421e-05, "loss": 0.1014, "step": 102005 }, { "epoch": 2.38, "learning_rate": 1.2228532538546385e-05, "loss": 0.0647, "step": 102010 }, { "epoch": 2.38, "learning_rate": 1.2227652484100351e-05, "loss": 0.0646, "step": 102015 }, { "epoch": 2.38, "learning_rate": 1.2226772429654315e-05, "loss": 0.1821, "step": 102020 }, { "epoch": 2.38, "learning_rate": 1.222589237520828e-05, "loss": 0.2648, "step": 102025 }, { "epoch": 2.38, "learning_rate": 1.2225012320762245e-05, "loss": 0.072, "step": 102030 }, { "epoch": 2.38, "learning_rate": 1.222413226631621e-05, "loss": 0.0505, "step": 102035 }, { "epoch": 2.38, "learning_rate": 1.2223252211870175e-05, "loss": 0.0524, "step": 102040 }, { "epoch": 2.38, "learning_rate": 1.222237215742414e-05, "loss": 0.0181, "step": 102045 }, { "epoch": 2.38, "learning_rate": 1.2221492102978105e-05, "loss": 0.0412, "step": 102050 }, { "epoch": 2.38, "learning_rate": 1.222061204853207e-05, "loss": 0.1638, "step": 102055 }, { "epoch": 2.38, "learning_rate": 1.2219731994086035e-05, "loss": 0.1188, "step": 102060 }, { "epoch": 2.38, "learning_rate": 1.221885193964e-05, "loss": 0.0261, "step": 102065 }, { "epoch": 2.38, "learning_rate": 1.2217971885193963e-05, "loss": 0.2581, "step": 102070 }, { "epoch": 2.38, "learning_rate": 1.221709183074793e-05, "loss": 0.2559, "step": 102075 }, { "epoch": 2.38, "learning_rate": 1.2216211776301893e-05, "loss": 0.014, "step": 102080 }, { "epoch": 2.38, "learning_rate": 1.221533172185586e-05, "loss": 0.0165, "step": 102085 }, { "epoch": 2.38, "learning_rate": 1.2214451667409823e-05, "loss": 0.0296, "step": 102090 }, { "epoch": 2.38, "learning_rate": 1.221357161296379e-05, "loss": 0.077, "step": 102095 }, { "epoch": 2.38, "learning_rate": 1.2212691558517753e-05, "loss": 0.0636, "step": 102100 }, { "epoch": 2.38, "learning_rate": 1.2211811504071718e-05, "loss": 0.1157, "step": 102105 }, { "epoch": 2.38, "learning_rate": 1.2210931449625685e-05, "loss": 0.1882, "step": 102110 }, { "epoch": 2.38, "learning_rate": 1.2210051395179648e-05, "loss": 0.1235, "step": 102115 }, { "epoch": 2.38, "learning_rate": 1.2209171340733615e-05, "loss": 0.1186, "step": 102120 }, { "epoch": 2.38, "learning_rate": 1.2208291286287578e-05, "loss": 0.3126, "step": 102125 }, { "epoch": 2.38, "learning_rate": 1.2207411231841545e-05, "loss": 0.0345, "step": 102130 }, { "epoch": 2.38, "learning_rate": 1.2206531177395508e-05, "loss": 0.0424, "step": 102135 }, { "epoch": 2.38, "learning_rate": 1.2205651122949473e-05, "loss": 0.0162, "step": 102140 }, { "epoch": 2.38, "learning_rate": 1.2204771068503438e-05, "loss": 0.0023, "step": 102145 }, { "epoch": 2.38, "learning_rate": 1.2203891014057403e-05, "loss": 0.0757, "step": 102150 }, { "epoch": 2.38, "learning_rate": 1.2203010959611368e-05, "loss": 0.0775, "step": 102155 }, { "epoch": 2.38, "learning_rate": 1.2202130905165333e-05, "loss": 0.0626, "step": 102160 }, { "epoch": 2.38, "learning_rate": 1.2201250850719298e-05, "loss": 0.1494, "step": 102165 }, { "epoch": 2.38, "learning_rate": 1.220054680716247e-05, "loss": 0.1638, "step": 102170 }, { "epoch": 2.38, "learning_rate": 1.2199666752716437e-05, "loss": 0.1887, "step": 102175 }, { "epoch": 2.38, "learning_rate": 1.21987866982704e-05, "loss": 0.0803, "step": 102180 }, { "epoch": 2.38, "learning_rate": 1.2197906643824365e-05, "loss": 0.046, "step": 102185 }, { "epoch": 2.38, "learning_rate": 1.219702658937833e-05, "loss": 0.1172, "step": 102190 }, { "epoch": 2.38, "learning_rate": 1.2196146534932295e-05, "loss": 0.055, "step": 102195 }, { "epoch": 2.38, "learning_rate": 1.219526648048626e-05, "loss": 0.0848, "step": 102200 }, { "epoch": 2.38, "learning_rate": 1.2194386426040225e-05, "loss": 0.0748, "step": 102205 }, { "epoch": 2.38, "learning_rate": 1.219350637159419e-05, "loss": 0.4167, "step": 102210 }, { "epoch": 2.38, "learning_rate": 1.2192626317148155e-05, "loss": 0.0873, "step": 102215 }, { "epoch": 2.38, "learning_rate": 1.2191746262702118e-05, "loss": 0.1584, "step": 102220 }, { "epoch": 2.39, "learning_rate": 1.2190866208256085e-05, "loss": 0.1555, "step": 102225 }, { "epoch": 2.39, "learning_rate": 1.2189986153810048e-05, "loss": 0.0234, "step": 102230 }, { "epoch": 2.39, "learning_rate": 1.2189106099364015e-05, "loss": 0.0545, "step": 102235 }, { "epoch": 2.39, "learning_rate": 1.2188226044917978e-05, "loss": 0.0716, "step": 102240 }, { "epoch": 2.39, "learning_rate": 1.2187345990471945e-05, "loss": 0.1052, "step": 102245 }, { "epoch": 2.39, "learning_rate": 1.2186465936025908e-05, "loss": 0.0716, "step": 102250 }, { "epoch": 2.39, "learning_rate": 1.2185585881579875e-05, "loss": 0.128, "step": 102255 }, { "epoch": 2.39, "learning_rate": 1.2184705827133838e-05, "loss": 0.1642, "step": 102260 }, { "epoch": 2.39, "learning_rate": 1.2183825772687803e-05, "loss": 0.0967, "step": 102265 }, { "epoch": 2.39, "learning_rate": 1.2182945718241769e-05, "loss": 0.1657, "step": 102270 }, { "epoch": 2.39, "learning_rate": 1.2182065663795734e-05, "loss": 0.5405, "step": 102275 }, { "epoch": 2.39, "learning_rate": 1.2181185609349699e-05, "loss": 0.0296, "step": 102280 }, { "epoch": 2.39, "learning_rate": 1.2180305554903664e-05, "loss": 0.0153, "step": 102285 }, { "epoch": 2.39, "learning_rate": 1.2179425500457629e-05, "loss": 0.0439, "step": 102290 }, { "epoch": 2.39, "learning_rate": 1.2178545446011594e-05, "loss": 0.0334, "step": 102295 }, { "epoch": 2.39, "learning_rate": 1.2177665391565559e-05, "loss": 0.0076, "step": 102300 }, { "epoch": 2.39, "learning_rate": 1.2176785337119524e-05, "loss": 0.0673, "step": 102305 }, { "epoch": 2.39, "learning_rate": 1.2175905282673489e-05, "loss": 0.1462, "step": 102310 }, { "epoch": 2.39, "learning_rate": 1.2175025228227454e-05, "loss": 0.1104, "step": 102315 }, { "epoch": 2.39, "learning_rate": 1.2174145173781419e-05, "loss": 0.2746, "step": 102320 }, { "epoch": 2.39, "learning_rate": 1.2173265119335384e-05, "loss": 0.4097, "step": 102325 }, { "epoch": 2.39, "learning_rate": 1.2172385064889349e-05, "loss": 0.0277, "step": 102330 }, { "epoch": 2.39, "learning_rate": 1.2171505010443312e-05, "loss": 0.0573, "step": 102335 }, { "epoch": 2.39, "learning_rate": 1.2170624955997279e-05, "loss": 0.0544, "step": 102340 }, { "epoch": 2.39, "learning_rate": 1.2169744901551242e-05, "loss": 0.0469, "step": 102345 }, { "epoch": 2.39, "learning_rate": 1.2168864847105209e-05, "loss": 0.0534, "step": 102350 }, { "epoch": 2.39, "learning_rate": 1.2167984792659172e-05, "loss": 0.1102, "step": 102355 }, { "epoch": 2.39, "learning_rate": 1.2167104738213139e-05, "loss": 0.0719, "step": 102360 }, { "epoch": 2.39, "learning_rate": 1.2166224683767102e-05, "loss": 0.0197, "step": 102365 }, { "epoch": 2.39, "learning_rate": 1.2165344629321067e-05, "loss": 0.4315, "step": 102370 }, { "epoch": 2.39, "learning_rate": 1.2164464574875032e-05, "loss": 0.3634, "step": 102375 }, { "epoch": 2.39, "learning_rate": 1.2163584520428997e-05, "loss": 0.1221, "step": 102380 }, { "epoch": 2.39, "learning_rate": 1.2162704465982962e-05, "loss": 0.0423, "step": 102385 }, { "epoch": 2.39, "learning_rate": 1.2161824411536927e-05, "loss": 0.0412, "step": 102390 }, { "epoch": 2.39, "learning_rate": 1.2160944357090892e-05, "loss": 0.094, "step": 102395 }, { "epoch": 2.39, "learning_rate": 1.2160064302644857e-05, "loss": 0.0869, "step": 102400 }, { "epoch": 2.39, "learning_rate": 1.2159184248198822e-05, "loss": 0.3172, "step": 102405 }, { "epoch": 2.39, "learning_rate": 1.2158304193752787e-05, "loss": 0.138, "step": 102410 }, { "epoch": 2.39, "learning_rate": 1.215742413930675e-05, "loss": 0.1828, "step": 102415 }, { "epoch": 2.39, "learning_rate": 1.2156544084860717e-05, "loss": 0.2901, "step": 102420 }, { "epoch": 2.39, "learning_rate": 1.215566403041468e-05, "loss": 0.0947, "step": 102425 }, { "epoch": 2.39, "learning_rate": 1.2154783975968647e-05, "loss": 0.038, "step": 102430 }, { "epoch": 2.39, "learning_rate": 1.2153903921522612e-05, "loss": 0.0747, "step": 102435 }, { "epoch": 2.39, "learning_rate": 1.2153023867076578e-05, "loss": 0.0175, "step": 102440 }, { "epoch": 2.39, "learning_rate": 1.2152143812630543e-05, "loss": 0.1165, "step": 102445 }, { "epoch": 2.39, "learning_rate": 1.2151263758184506e-05, "loss": 0.1219, "step": 102450 }, { "epoch": 2.39, "learning_rate": 1.2150383703738473e-05, "loss": 0.0331, "step": 102455 }, { "epoch": 2.39, "learning_rate": 1.2149503649292436e-05, "loss": 0.0682, "step": 102460 }, { "epoch": 2.39, "learning_rate": 1.2148623594846403e-05, "loss": 0.2987, "step": 102465 }, { "epoch": 2.39, "learning_rate": 1.2147743540400366e-05, "loss": 0.1926, "step": 102470 }, { "epoch": 2.39, "learning_rate": 1.2146863485954333e-05, "loss": 0.3918, "step": 102475 }, { "epoch": 2.39, "learning_rate": 1.2145983431508296e-05, "loss": 0.0259, "step": 102480 }, { "epoch": 2.39, "learning_rate": 1.2145103377062261e-05, "loss": 0.0244, "step": 102485 }, { "epoch": 2.39, "learning_rate": 1.2144223322616226e-05, "loss": 0.0123, "step": 102490 }, { "epoch": 2.39, "learning_rate": 1.2143343268170191e-05, "loss": 0.0683, "step": 102495 }, { "epoch": 2.39, "learning_rate": 1.2142463213724156e-05, "loss": 0.1139, "step": 102500 }, { "epoch": 2.39, "learning_rate": 1.2141583159278121e-05, "loss": 0.0829, "step": 102505 }, { "epoch": 2.39, "learning_rate": 1.2140703104832086e-05, "loss": 0.0994, "step": 102510 }, { "epoch": 2.39, "learning_rate": 1.2139823050386051e-05, "loss": 0.0914, "step": 102515 }, { "epoch": 2.39, "learning_rate": 1.2138942995940014e-05, "loss": 0.1778, "step": 102520 }, { "epoch": 2.39, "learning_rate": 1.2138062941493981e-05, "loss": 0.1794, "step": 102525 }, { "epoch": 2.39, "learning_rate": 1.2137182887047944e-05, "loss": 0.0114, "step": 102530 }, { "epoch": 2.39, "learning_rate": 1.2136302832601911e-05, "loss": 0.1141, "step": 102535 }, { "epoch": 2.39, "learning_rate": 1.2135422778155875e-05, "loss": 0.0317, "step": 102540 }, { "epoch": 2.39, "learning_rate": 1.2134542723709841e-05, "loss": 0.0325, "step": 102545 }, { "epoch": 2.39, "learning_rate": 1.2133662669263805e-05, "loss": 0.0361, "step": 102550 }, { "epoch": 2.39, "learning_rate": 1.2132782614817771e-05, "loss": 0.0865, "step": 102555 }, { "epoch": 2.39, "learning_rate": 1.2131902560371735e-05, "loss": 0.1224, "step": 102560 }, { "epoch": 2.39, "learning_rate": 1.21310225059257e-05, "loss": 0.2208, "step": 102565 }, { "epoch": 2.39, "learning_rate": 1.2130142451479665e-05, "loss": 0.3428, "step": 102570 }, { "epoch": 2.39, "learning_rate": 1.212926239703363e-05, "loss": 0.324, "step": 102575 }, { "epoch": 2.39, "learning_rate": 1.2128382342587596e-05, "loss": 0.0016, "step": 102580 }, { "epoch": 2.39, "learning_rate": 1.212750228814156e-05, "loss": 0.0178, "step": 102585 }, { "epoch": 2.39, "learning_rate": 1.2126622233695526e-05, "loss": 0.0522, "step": 102590 }, { "epoch": 2.39, "learning_rate": 1.212574217924949e-05, "loss": 0.0221, "step": 102595 }, { "epoch": 2.39, "learning_rate": 1.2124862124803455e-05, "loss": 0.0273, "step": 102600 }, { "epoch": 2.39, "learning_rate": 1.212398207035742e-05, "loss": 0.031, "step": 102605 }, { "epoch": 2.39, "learning_rate": 1.2123102015911385e-05, "loss": 0.1949, "step": 102610 }, { "epoch": 2.39, "learning_rate": 1.212222196146535e-05, "loss": 0.1411, "step": 102615 }, { "epoch": 2.39, "learning_rate": 1.2121341907019315e-05, "loss": 0.1878, "step": 102620 }, { "epoch": 2.39, "learning_rate": 1.212046185257328e-05, "loss": 0.1183, "step": 102625 }, { "epoch": 2.39, "learning_rate": 1.2119581798127245e-05, "loss": 0.0333, "step": 102630 }, { "epoch": 2.39, "learning_rate": 1.2118701743681208e-05, "loss": 0.0441, "step": 102635 }, { "epoch": 2.39, "learning_rate": 1.2117821689235175e-05, "loss": 0.0637, "step": 102640 }, { "epoch": 2.39, "learning_rate": 1.2116941634789138e-05, "loss": 0.125, "step": 102645 }, { "epoch": 2.39, "learning_rate": 1.2116061580343105e-05, "loss": 0.0334, "step": 102650 }, { "epoch": 2.4, "learning_rate": 1.2115181525897068e-05, "loss": 0.1245, "step": 102655 }, { "epoch": 2.4, "learning_rate": 1.2114301471451035e-05, "loss": 0.1168, "step": 102660 }, { "epoch": 2.4, "learning_rate": 1.2113421417004998e-05, "loss": 0.0284, "step": 102665 }, { "epoch": 2.4, "learning_rate": 1.2112541362558963e-05, "loss": 0.1408, "step": 102670 }, { "epoch": 2.4, "learning_rate": 1.2111661308112928e-05, "loss": 0.1297, "step": 102675 }, { "epoch": 2.4, "learning_rate": 1.2110781253666893e-05, "loss": 0.026, "step": 102680 }, { "epoch": 2.4, "learning_rate": 1.2109901199220858e-05, "loss": 0.0564, "step": 102685 }, { "epoch": 2.4, "learning_rate": 1.2109021144774823e-05, "loss": 0.0469, "step": 102690 }, { "epoch": 2.4, "learning_rate": 1.2108141090328788e-05, "loss": 0.096, "step": 102695 }, { "epoch": 2.4, "learning_rate": 1.2107261035882753e-05, "loss": 0.0738, "step": 102700 }, { "epoch": 2.4, "learning_rate": 1.2106380981436719e-05, "loss": 0.2385, "step": 102705 }, { "epoch": 2.4, "learning_rate": 1.2105500926990684e-05, "loss": 0.0381, "step": 102710 }, { "epoch": 2.4, "learning_rate": 1.2104620872544649e-05, "loss": 0.1604, "step": 102715 }, { "epoch": 2.4, "learning_rate": 1.2103740818098614e-05, "loss": 0.1648, "step": 102720 }, { "epoch": 2.4, "learning_rate": 1.2102860763652579e-05, "loss": 0.242, "step": 102725 }, { "epoch": 2.4, "learning_rate": 1.2101980709206544e-05, "loss": 0.1178, "step": 102730 }, { "epoch": 2.4, "learning_rate": 1.2101100654760509e-05, "loss": 0.0588, "step": 102735 }, { "epoch": 2.4, "learning_rate": 1.2100220600314474e-05, "loss": 0.0244, "step": 102740 }, { "epoch": 2.4, "learning_rate": 1.2099340545868439e-05, "loss": 0.0248, "step": 102745 }, { "epoch": 2.4, "learning_rate": 1.2098460491422402e-05, "loss": 0.1186, "step": 102750 }, { "epoch": 2.4, "learning_rate": 1.2097580436976369e-05, "loss": 0.1226, "step": 102755 }, { "epoch": 2.4, "learning_rate": 1.2096700382530332e-05, "loss": 0.2966, "step": 102760 }, { "epoch": 2.4, "learning_rate": 1.2095820328084299e-05, "loss": 0.0887, "step": 102765 }, { "epoch": 2.4, "learning_rate": 1.2094940273638262e-05, "loss": 0.1365, "step": 102770 }, { "epoch": 2.4, "learning_rate": 1.2094060219192229e-05, "loss": 0.2004, "step": 102775 }, { "epoch": 2.4, "learning_rate": 1.2093180164746192e-05, "loss": 0.1063, "step": 102780 }, { "epoch": 2.4, "learning_rate": 1.2092300110300157e-05, "loss": 0.0263, "step": 102785 }, { "epoch": 2.4, "learning_rate": 1.2091420055854122e-05, "loss": 0.011, "step": 102790 }, { "epoch": 2.4, "learning_rate": 1.2090540001408087e-05, "loss": 0.0038, "step": 102795 }, { "epoch": 2.4, "learning_rate": 1.2089659946962052e-05, "loss": 0.1148, "step": 102800 }, { "epoch": 2.4, "learning_rate": 1.2088779892516017e-05, "loss": 0.0797, "step": 102805 }, { "epoch": 2.4, "learning_rate": 1.2087899838069982e-05, "loss": 0.0319, "step": 102810 }, { "epoch": 2.4, "learning_rate": 1.2087019783623947e-05, "loss": 0.0902, "step": 102815 }, { "epoch": 2.4, "learning_rate": 1.208613972917791e-05, "loss": 0.2807, "step": 102820 }, { "epoch": 2.4, "learning_rate": 1.2085259674731877e-05, "loss": 0.2741, "step": 102825 }, { "epoch": 2.4, "learning_rate": 1.208437962028584e-05, "loss": 0.0801, "step": 102830 }, { "epoch": 2.4, "learning_rate": 1.2083499565839807e-05, "loss": 0.0147, "step": 102835 }, { "epoch": 2.4, "learning_rate": 1.208261951139377e-05, "loss": 0.0224, "step": 102840 }, { "epoch": 2.4, "learning_rate": 1.2081739456947737e-05, "loss": 0.061, "step": 102845 }, { "epoch": 2.4, "learning_rate": 1.2080859402501702e-05, "loss": 0.117, "step": 102850 }, { "epoch": 2.4, "learning_rate": 1.2079979348055667e-05, "loss": 0.1229, "step": 102855 }, { "epoch": 2.4, "learning_rate": 1.2079099293609632e-05, "loss": 0.0497, "step": 102860 }, { "epoch": 2.4, "learning_rate": 1.2078219239163596e-05, "loss": 0.2159, "step": 102865 }, { "epoch": 2.4, "learning_rate": 1.2077339184717562e-05, "loss": 0.3664, "step": 102870 }, { "epoch": 2.4, "learning_rate": 1.2076459130271526e-05, "loss": 0.2723, "step": 102875 }, { "epoch": 2.4, "learning_rate": 1.2075579075825493e-05, "loss": 0.0246, "step": 102880 }, { "epoch": 2.4, "learning_rate": 1.2074699021379456e-05, "loss": 0.0349, "step": 102885 }, { "epoch": 2.4, "learning_rate": 1.2073818966933423e-05, "loss": 0.0211, "step": 102890 }, { "epoch": 2.4, "learning_rate": 1.2072938912487386e-05, "loss": 0.1148, "step": 102895 }, { "epoch": 2.4, "learning_rate": 1.2072058858041351e-05, "loss": 0.0735, "step": 102900 }, { "epoch": 2.4, "learning_rate": 1.2071178803595316e-05, "loss": 0.0354, "step": 102905 }, { "epoch": 2.4, "learning_rate": 1.2070298749149281e-05, "loss": 0.0339, "step": 102910 }, { "epoch": 2.4, "learning_rate": 1.2069418694703246e-05, "loss": 0.141, "step": 102915 }, { "epoch": 2.4, "learning_rate": 1.2068538640257211e-05, "loss": 0.108, "step": 102920 }, { "epoch": 2.4, "learning_rate": 1.2067658585811176e-05, "loss": 0.1631, "step": 102925 }, { "epoch": 2.4, "learning_rate": 1.2066778531365141e-05, "loss": 0.046, "step": 102930 }, { "epoch": 2.4, "learning_rate": 1.2065898476919104e-05, "loss": 0.0292, "step": 102935 }, { "epoch": 2.4, "learning_rate": 1.2065018422473071e-05, "loss": 0.0016, "step": 102940 }, { "epoch": 2.4, "learning_rate": 1.2064138368027034e-05, "loss": 0.0203, "step": 102945 }, { "epoch": 2.4, "learning_rate": 1.2063258313581001e-05, "loss": 0.0728, "step": 102950 }, { "epoch": 2.4, "learning_rate": 1.2062378259134964e-05, "loss": 0.0663, "step": 102955 }, { "epoch": 2.4, "learning_rate": 1.2061498204688931e-05, "loss": 0.1181, "step": 102960 }, { "epoch": 2.4, "learning_rate": 1.2060618150242895e-05, "loss": 0.1837, "step": 102965 }, { "epoch": 2.4, "learning_rate": 1.205973809579686e-05, "loss": 0.1306, "step": 102970 }, { "epoch": 2.4, "learning_rate": 1.2058858041350825e-05, "loss": 0.2213, "step": 102975 }, { "epoch": 2.4, "learning_rate": 1.205797798690479e-05, "loss": 0.0747, "step": 102980 }, { "epoch": 2.4, "learning_rate": 1.2057097932458756e-05, "loss": 0.0798, "step": 102985 }, { "epoch": 2.4, "learning_rate": 1.205621787801272e-05, "loss": 0.0187, "step": 102990 }, { "epoch": 2.4, "learning_rate": 1.2055337823566686e-05, "loss": 0.0694, "step": 102995 }, { "epoch": 2.4, "learning_rate": 1.205445776912065e-05, "loss": 0.0428, "step": 103000 }, { "epoch": 2.4, "learning_rate": 1.2053577714674616e-05, "loss": 0.0155, "step": 103005 }, { "epoch": 2.4, "learning_rate": 1.205269766022858e-05, "loss": 0.0625, "step": 103010 }, { "epoch": 2.4, "learning_rate": 1.2051817605782545e-05, "loss": 0.0815, "step": 103015 }, { "epoch": 2.4, "learning_rate": 1.205093755133651e-05, "loss": 0.1478, "step": 103020 }, { "epoch": 2.4, "learning_rate": 1.2050057496890475e-05, "loss": 0.1264, "step": 103025 }, { "epoch": 2.4, "learning_rate": 1.204917744244444e-05, "loss": 0.0807, "step": 103030 }, { "epoch": 2.4, "learning_rate": 1.2048297387998405e-05, "loss": 0.054, "step": 103035 }, { "epoch": 2.4, "learning_rate": 1.204741733355237e-05, "loss": 0.1053, "step": 103040 }, { "epoch": 2.4, "learning_rate": 1.2046537279106335e-05, "loss": 0.0084, "step": 103045 }, { "epoch": 2.4, "learning_rate": 1.2045657224660298e-05, "loss": 0.0585, "step": 103050 }, { "epoch": 2.4, "learning_rate": 1.2044777170214265e-05, "loss": 0.0515, "step": 103055 }, { "epoch": 2.4, "learning_rate": 1.2043897115768228e-05, "loss": 0.0672, "step": 103060 }, { "epoch": 2.4, "learning_rate": 1.2043017061322195e-05, "loss": 0.107, "step": 103065 }, { "epoch": 2.4, "learning_rate": 1.2042137006876158e-05, "loss": 0.1746, "step": 103070 }, { "epoch": 2.4, "learning_rate": 1.2041256952430125e-05, "loss": 0.3794, "step": 103075 }, { "epoch": 2.4, "learning_rate": 1.2040376897984088e-05, "loss": 0.0888, "step": 103080 }, { "epoch": 2.41, "learning_rate": 1.2039496843538053e-05, "loss": 0.0055, "step": 103085 }, { "epoch": 2.41, "learning_rate": 1.2038616789092018e-05, "loss": 0.0296, "step": 103090 }, { "epoch": 2.41, "learning_rate": 1.2037736734645983e-05, "loss": 0.1344, "step": 103095 }, { "epoch": 2.41, "learning_rate": 1.2036856680199948e-05, "loss": 0.0877, "step": 103100 }, { "epoch": 2.41, "learning_rate": 1.2035976625753913e-05, "loss": 0.0057, "step": 103105 }, { "epoch": 2.41, "learning_rate": 1.2035096571307878e-05, "loss": 0.0986, "step": 103110 }, { "epoch": 2.41, "learning_rate": 1.2034216516861843e-05, "loss": 0.1986, "step": 103115 }, { "epoch": 2.41, "learning_rate": 1.2033336462415808e-05, "loss": 0.0657, "step": 103120 }, { "epoch": 2.41, "learning_rate": 1.2032456407969773e-05, "loss": 0.1929, "step": 103125 }, { "epoch": 2.41, "learning_rate": 1.2031576353523738e-05, "loss": 0.0229, "step": 103130 }, { "epoch": 2.41, "learning_rate": 1.2030696299077704e-05, "loss": 0.0141, "step": 103135 }, { "epoch": 2.41, "learning_rate": 1.2029816244631669e-05, "loss": 0.0959, "step": 103140 }, { "epoch": 2.41, "learning_rate": 1.2028936190185634e-05, "loss": 0.0515, "step": 103145 }, { "epoch": 2.41, "learning_rate": 1.2028056135739599e-05, "loss": 0.1606, "step": 103150 }, { "epoch": 2.41, "learning_rate": 1.2027176081293564e-05, "loss": 0.0432, "step": 103155 }, { "epoch": 2.41, "learning_rate": 1.2026296026847529e-05, "loss": 0.0368, "step": 103160 }, { "epoch": 2.41, "learning_rate": 1.2025415972401492e-05, "loss": 0.1055, "step": 103165 }, { "epoch": 2.41, "learning_rate": 1.2024535917955459e-05, "loss": 0.2093, "step": 103170 }, { "epoch": 2.41, "learning_rate": 1.2023655863509422e-05, "loss": 0.3951, "step": 103175 }, { "epoch": 2.41, "learning_rate": 1.2022775809063389e-05, "loss": 0.1747, "step": 103180 }, { "epoch": 2.41, "learning_rate": 1.2021895754617352e-05, "loss": 0.0014, "step": 103185 }, { "epoch": 2.41, "learning_rate": 1.2021015700171319e-05, "loss": 0.0935, "step": 103190 }, { "epoch": 2.41, "learning_rate": 1.2020135645725282e-05, "loss": 0.0314, "step": 103195 }, { "epoch": 2.41, "learning_rate": 1.2019255591279247e-05, "loss": 0.1576, "step": 103200 }, { "epoch": 2.41, "learning_rate": 1.2018375536833212e-05, "loss": 0.1058, "step": 103205 }, { "epoch": 2.41, "learning_rate": 1.2017495482387177e-05, "loss": 0.1494, "step": 103210 }, { "epoch": 2.41, "learning_rate": 1.2016615427941142e-05, "loss": 0.2722, "step": 103215 }, { "epoch": 2.41, "learning_rate": 1.2015735373495107e-05, "loss": 0.2542, "step": 103220 }, { "epoch": 2.41, "learning_rate": 1.2014855319049072e-05, "loss": 0.1886, "step": 103225 }, { "epoch": 2.41, "learning_rate": 1.2013975264603037e-05, "loss": 0.0509, "step": 103230 }, { "epoch": 2.41, "learning_rate": 1.2013095210157e-05, "loss": 0.0299, "step": 103235 }, { "epoch": 2.41, "learning_rate": 1.2012215155710967e-05, "loss": 0.0312, "step": 103240 }, { "epoch": 2.41, "learning_rate": 1.201133510126493e-05, "loss": 0.0101, "step": 103245 }, { "epoch": 2.41, "learning_rate": 1.2010455046818897e-05, "loss": 0.0338, "step": 103250 }, { "epoch": 2.41, "learning_rate": 1.2009574992372862e-05, "loss": 0.0167, "step": 103255 }, { "epoch": 2.41, "learning_rate": 1.2008694937926827e-05, "loss": 0.1648, "step": 103260 }, { "epoch": 2.41, "learning_rate": 1.2007814883480792e-05, "loss": 0.1195, "step": 103265 }, { "epoch": 2.41, "learning_rate": 1.2006934829034756e-05, "loss": 0.0703, "step": 103270 }, { "epoch": 2.41, "learning_rate": 1.2006054774588722e-05, "loss": 0.2481, "step": 103275 }, { "epoch": 2.41, "learning_rate": 1.2005174720142686e-05, "loss": 0.0588, "step": 103280 }, { "epoch": 2.41, "learning_rate": 1.2004294665696652e-05, "loss": 0.0252, "step": 103285 }, { "epoch": 2.41, "learning_rate": 1.2003414611250616e-05, "loss": 0.0424, "step": 103290 }, { "epoch": 2.41, "learning_rate": 1.2002534556804582e-05, "loss": 0.0805, "step": 103295 }, { "epoch": 2.41, "learning_rate": 1.2001654502358546e-05, "loss": 0.0448, "step": 103300 }, { "epoch": 2.41, "learning_rate": 1.200077444791251e-05, "loss": 0.0831, "step": 103305 }, { "epoch": 2.41, "learning_rate": 1.1999894393466476e-05, "loss": 0.1486, "step": 103310 }, { "epoch": 2.41, "learning_rate": 1.1999014339020441e-05, "loss": 0.0862, "step": 103315 }, { "epoch": 2.41, "learning_rate": 1.1998134284574406e-05, "loss": 0.2956, "step": 103320 }, { "epoch": 2.41, "learning_rate": 1.1997254230128371e-05, "loss": 0.3043, "step": 103325 }, { "epoch": 2.41, "learning_rate": 1.1996374175682336e-05, "loss": 0.035, "step": 103330 }, { "epoch": 2.41, "learning_rate": 1.1995494121236301e-05, "loss": 0.0077, "step": 103335 }, { "epoch": 2.41, "learning_rate": 1.1994614066790266e-05, "loss": 0.059, "step": 103340 }, { "epoch": 2.41, "learning_rate": 1.1993734012344231e-05, "loss": 0.0748, "step": 103345 }, { "epoch": 2.41, "learning_rate": 1.1992853957898194e-05, "loss": 0.0394, "step": 103350 }, { "epoch": 2.41, "learning_rate": 1.1991973903452161e-05, "loss": 0.0648, "step": 103355 }, { "epoch": 2.41, "learning_rate": 1.1991093849006124e-05, "loss": 0.0948, "step": 103360 }, { "epoch": 2.41, "learning_rate": 1.1990213794560091e-05, "loss": 0.2348, "step": 103365 }, { "epoch": 2.41, "learning_rate": 1.1989333740114054e-05, "loss": 0.2701, "step": 103370 }, { "epoch": 2.41, "learning_rate": 1.1988453685668021e-05, "loss": 0.194, "step": 103375 }, { "epoch": 2.41, "learning_rate": 1.1987573631221984e-05, "loss": 0.0258, "step": 103380 }, { "epoch": 2.41, "learning_rate": 1.198669357677595e-05, "loss": 0.0755, "step": 103385 }, { "epoch": 2.41, "learning_rate": 1.1985813522329916e-05, "loss": 0.0247, "step": 103390 }, { "epoch": 2.41, "learning_rate": 1.198493346788388e-05, "loss": 0.1175, "step": 103395 }, { "epoch": 2.41, "learning_rate": 1.1984053413437846e-05, "loss": 0.0858, "step": 103400 }, { "epoch": 2.41, "learning_rate": 1.198317335899181e-05, "loss": 0.1242, "step": 103405 }, { "epoch": 2.41, "learning_rate": 1.1982293304545776e-05, "loss": 0.1362, "step": 103410 }, { "epoch": 2.41, "learning_rate": 1.198141325009974e-05, "loss": 0.1019, "step": 103415 }, { "epoch": 2.41, "learning_rate": 1.1980533195653705e-05, "loss": 0.1544, "step": 103420 }, { "epoch": 2.41, "learning_rate": 1.197965314120767e-05, "loss": 0.3835, "step": 103425 }, { "epoch": 2.41, "learning_rate": 1.1978773086761635e-05, "loss": 0.0674, "step": 103430 }, { "epoch": 2.41, "learning_rate": 1.19778930323156e-05, "loss": 0.001, "step": 103435 }, { "epoch": 2.41, "learning_rate": 1.1977012977869565e-05, "loss": 0.1426, "step": 103440 }, { "epoch": 2.41, "learning_rate": 1.197613292342353e-05, "loss": 0.0544, "step": 103445 }, { "epoch": 2.41, "learning_rate": 1.1975252868977495e-05, "loss": 0.0353, "step": 103450 }, { "epoch": 2.41, "learning_rate": 1.1974372814531458e-05, "loss": 0.1174, "step": 103455 }, { "epoch": 2.41, "learning_rate": 1.1973492760085425e-05, "loss": 0.1311, "step": 103460 }, { "epoch": 2.41, "learning_rate": 1.1972612705639388e-05, "loss": 0.1431, "step": 103465 }, { "epoch": 2.41, "learning_rate": 1.1971732651193355e-05, "loss": 0.22, "step": 103470 }, { "epoch": 2.41, "learning_rate": 1.1970852596747318e-05, "loss": 0.2644, "step": 103475 }, { "epoch": 2.41, "learning_rate": 1.1969972542301285e-05, "loss": 0.0802, "step": 103480 }, { "epoch": 2.41, "learning_rate": 1.1969092487855248e-05, "loss": 0.1206, "step": 103485 }, { "epoch": 2.41, "learning_rate": 1.1968212433409215e-05, "loss": 0.1137, "step": 103490 }, { "epoch": 2.41, "learning_rate": 1.1967332378963178e-05, "loss": 0.0065, "step": 103495 }, { "epoch": 2.41, "learning_rate": 1.1966452324517143e-05, "loss": 0.0099, "step": 103500 }, { "epoch": 2.41, "learning_rate": 1.1965572270071108e-05, "loss": 0.0698, "step": 103505 }, { "epoch": 2.42, "learning_rate": 1.1964692215625073e-05, "loss": 0.141, "step": 103510 }, { "epoch": 2.42, "learning_rate": 1.1963812161179038e-05, "loss": 0.0638, "step": 103515 }, { "epoch": 2.42, "learning_rate": 1.1962932106733003e-05, "loss": 0.1219, "step": 103520 }, { "epoch": 2.42, "learning_rate": 1.196205205228697e-05, "loss": 0.3461, "step": 103525 }, { "epoch": 2.42, "learning_rate": 1.1961171997840933e-05, "loss": 0.0197, "step": 103530 }, { "epoch": 2.42, "learning_rate": 1.1960291943394898e-05, "loss": 0.008, "step": 103535 }, { "epoch": 2.42, "learning_rate": 1.1959411888948863e-05, "loss": 0.1028, "step": 103540 }, { "epoch": 2.42, "learning_rate": 1.1958531834502828e-05, "loss": 0.0337, "step": 103545 }, { "epoch": 2.42, "learning_rate": 1.1957651780056793e-05, "loss": 0.0536, "step": 103550 }, { "epoch": 2.42, "learning_rate": 1.1956771725610758e-05, "loss": 0.0322, "step": 103555 }, { "epoch": 2.42, "learning_rate": 1.1955891671164723e-05, "loss": 0.1517, "step": 103560 }, { "epoch": 2.42, "learning_rate": 1.1955011616718688e-05, "loss": 0.2391, "step": 103565 }, { "epoch": 2.42, "learning_rate": 1.1954131562272652e-05, "loss": 0.1932, "step": 103570 }, { "epoch": 2.42, "learning_rate": 1.1953251507826619e-05, "loss": 0.2839, "step": 103575 }, { "epoch": 2.42, "learning_rate": 1.1952371453380582e-05, "loss": 0.0145, "step": 103580 }, { "epoch": 2.42, "learning_rate": 1.1951491398934549e-05, "loss": 0.0048, "step": 103585 }, { "epoch": 2.42, "learning_rate": 1.1950611344488512e-05, "loss": 0.0436, "step": 103590 }, { "epoch": 2.42, "learning_rate": 1.1949731290042479e-05, "loss": 0.0301, "step": 103595 }, { "epoch": 2.42, "learning_rate": 1.1948851235596442e-05, "loss": 0.085, "step": 103600 }, { "epoch": 2.42, "learning_rate": 1.1947971181150407e-05, "loss": 0.1462, "step": 103605 }, { "epoch": 2.42, "learning_rate": 1.1947091126704372e-05, "loss": 0.1651, "step": 103610 }, { "epoch": 2.42, "learning_rate": 1.1946211072258337e-05, "loss": 0.2737, "step": 103615 }, { "epoch": 2.42, "learning_rate": 1.1945331017812302e-05, "loss": 0.2914, "step": 103620 }, { "epoch": 2.42, "learning_rate": 1.1944450963366267e-05, "loss": 0.3575, "step": 103625 }, { "epoch": 2.42, "learning_rate": 1.1943570908920232e-05, "loss": 0.1482, "step": 103630 }, { "epoch": 2.42, "learning_rate": 1.1942690854474197e-05, "loss": 0.0278, "step": 103635 }, { "epoch": 2.42, "learning_rate": 1.1941810800028162e-05, "loss": 0.0977, "step": 103640 }, { "epoch": 2.42, "learning_rate": 1.1940930745582127e-05, "loss": 0.0137, "step": 103645 }, { "epoch": 2.42, "learning_rate": 1.194005069113609e-05, "loss": 0.0339, "step": 103650 }, { "epoch": 2.42, "learning_rate": 1.1939170636690057e-05, "loss": 0.0751, "step": 103655 }, { "epoch": 2.42, "learning_rate": 1.1938290582244022e-05, "loss": 0.1399, "step": 103660 }, { "epoch": 2.42, "learning_rate": 1.1937410527797987e-05, "loss": 0.2055, "step": 103665 }, { "epoch": 2.42, "learning_rate": 1.1936530473351952e-05, "loss": 0.385, "step": 103670 }, { "epoch": 2.42, "learning_rate": 1.1935650418905917e-05, "loss": 0.4032, "step": 103675 }, { "epoch": 2.42, "learning_rate": 1.1934770364459882e-05, "loss": 0.1141, "step": 103680 }, { "epoch": 2.42, "learning_rate": 1.1933890310013846e-05, "loss": 0.0075, "step": 103685 }, { "epoch": 2.42, "learning_rate": 1.1933010255567812e-05, "loss": 0.018, "step": 103690 }, { "epoch": 2.42, "learning_rate": 1.1932130201121776e-05, "loss": 0.0213, "step": 103695 }, { "epoch": 2.42, "learning_rate": 1.1931250146675742e-05, "loss": 0.0999, "step": 103700 }, { "epoch": 2.42, "learning_rate": 1.1930370092229706e-05, "loss": 0.0187, "step": 103705 }, { "epoch": 2.42, "learning_rate": 1.1929490037783672e-05, "loss": 0.1439, "step": 103710 }, { "epoch": 2.42, "learning_rate": 1.1928609983337636e-05, "loss": 0.222, "step": 103715 }, { "epoch": 2.42, "learning_rate": 1.19277299288916e-05, "loss": 0.1948, "step": 103720 }, { "epoch": 2.42, "learning_rate": 1.1926849874445566e-05, "loss": 0.1456, "step": 103725 }, { "epoch": 2.42, "learning_rate": 1.192596981999953e-05, "loss": 0.0468, "step": 103730 }, { "epoch": 2.42, "learning_rate": 1.1925089765553496e-05, "loss": 0.0089, "step": 103735 }, { "epoch": 2.42, "learning_rate": 1.192420971110746e-05, "loss": 0.0116, "step": 103740 }, { "epoch": 2.42, "learning_rate": 1.1923329656661426e-05, "loss": 0.0258, "step": 103745 }, { "epoch": 2.42, "learning_rate": 1.1922449602215391e-05, "loss": 0.0319, "step": 103750 }, { "epoch": 2.42, "learning_rate": 1.1921569547769354e-05, "loss": 0.0783, "step": 103755 }, { "epoch": 2.42, "learning_rate": 1.1920689493323321e-05, "loss": 0.1024, "step": 103760 }, { "epoch": 2.42, "learning_rate": 1.1919809438877284e-05, "loss": 0.1484, "step": 103765 }, { "epoch": 2.42, "learning_rate": 1.1918929384431251e-05, "loss": 0.2016, "step": 103770 }, { "epoch": 2.42, "learning_rate": 1.1918049329985214e-05, "loss": 0.4914, "step": 103775 }, { "epoch": 2.42, "learning_rate": 1.1917169275539181e-05, "loss": 0.0566, "step": 103780 }, { "epoch": 2.42, "learning_rate": 1.1916289221093144e-05, "loss": 0.0515, "step": 103785 }, { "epoch": 2.42, "learning_rate": 1.1915409166647111e-05, "loss": 0.051, "step": 103790 }, { "epoch": 2.42, "learning_rate": 1.1914529112201076e-05, "loss": 0.0332, "step": 103795 }, { "epoch": 2.42, "learning_rate": 1.191364905775504e-05, "loss": 0.0538, "step": 103800 }, { "epoch": 2.42, "learning_rate": 1.1912769003309006e-05, "loss": 0.098, "step": 103805 }, { "epoch": 2.42, "learning_rate": 1.191188894886297e-05, "loss": 0.1003, "step": 103810 }, { "epoch": 2.42, "learning_rate": 1.1911008894416936e-05, "loss": 0.1338, "step": 103815 }, { "epoch": 2.42, "learning_rate": 1.19101288399709e-05, "loss": 0.1548, "step": 103820 }, { "epoch": 2.42, "learning_rate": 1.1909248785524866e-05, "loss": 0.2236, "step": 103825 }, { "epoch": 2.42, "learning_rate": 1.190836873107883e-05, "loss": 0.1066, "step": 103830 }, { "epoch": 2.42, "learning_rate": 1.1907488676632795e-05, "loss": 0.0037, "step": 103835 }, { "epoch": 2.42, "learning_rate": 1.190660862218676e-05, "loss": 0.0511, "step": 103840 }, { "epoch": 2.42, "learning_rate": 1.1905728567740725e-05, "loss": 0.0382, "step": 103845 }, { "epoch": 2.42, "learning_rate": 1.190484851329469e-05, "loss": 0.0954, "step": 103850 }, { "epoch": 2.42, "learning_rate": 1.1903968458848655e-05, "loss": 0.1374, "step": 103855 }, { "epoch": 2.42, "learning_rate": 1.190308840440262e-05, "loss": 0.1586, "step": 103860 }, { "epoch": 2.42, "learning_rate": 1.1902208349956585e-05, "loss": 0.1717, "step": 103865 }, { "epoch": 2.42, "learning_rate": 1.1901328295510548e-05, "loss": 0.18, "step": 103870 }, { "epoch": 2.42, "learning_rate": 1.1900448241064515e-05, "loss": 0.2281, "step": 103875 }, { "epoch": 2.42, "learning_rate": 1.1899568186618478e-05, "loss": 0.0135, "step": 103880 }, { "epoch": 2.42, "learning_rate": 1.1898688132172445e-05, "loss": 0.1283, "step": 103885 }, { "epoch": 2.42, "learning_rate": 1.1897808077726408e-05, "loss": 0.0011, "step": 103890 }, { "epoch": 2.42, "learning_rate": 1.1896928023280375e-05, "loss": 0.1161, "step": 103895 }, { "epoch": 2.42, "learning_rate": 1.1896047968834338e-05, "loss": 0.0867, "step": 103900 }, { "epoch": 2.42, "learning_rate": 1.1895167914388303e-05, "loss": 0.0199, "step": 103905 }, { "epoch": 2.42, "learning_rate": 1.1894287859942268e-05, "loss": 0.1251, "step": 103910 }, { "epoch": 2.42, "learning_rate": 1.1893407805496233e-05, "loss": 0.189, "step": 103915 }, { "epoch": 2.42, "learning_rate": 1.1892527751050198e-05, "loss": 0.3371, "step": 103920 }, { "epoch": 2.42, "learning_rate": 1.1891647696604163e-05, "loss": 0.1491, "step": 103925 }, { "epoch": 2.42, "learning_rate": 1.1890767642158128e-05, "loss": 0.0327, "step": 103930 }, { "epoch": 2.42, "learning_rate": 1.1889887587712093e-05, "loss": 0.0639, "step": 103935 }, { "epoch": 2.43, "learning_rate": 1.188900753326606e-05, "loss": 0.0517, "step": 103940 }, { "epoch": 2.43, "learning_rate": 1.1888127478820023e-05, "loss": 0.0373, "step": 103945 }, { "epoch": 2.43, "learning_rate": 1.1887247424373988e-05, "loss": 0.0586, "step": 103950 }, { "epoch": 2.43, "learning_rate": 1.1886367369927953e-05, "loss": 0.0817, "step": 103955 }, { "epoch": 2.43, "learning_rate": 1.1885487315481918e-05, "loss": 0.0949, "step": 103960 }, { "epoch": 2.43, "learning_rate": 1.1884607261035883e-05, "loss": 0.1538, "step": 103965 }, { "epoch": 2.43, "learning_rate": 1.1883727206589848e-05, "loss": 0.2122, "step": 103970 }, { "epoch": 2.43, "learning_rate": 1.1882847152143813e-05, "loss": 0.2057, "step": 103975 }, { "epoch": 2.43, "learning_rate": 1.1881967097697778e-05, "loss": 0.0386, "step": 103980 }, { "epoch": 2.43, "learning_rate": 1.1881087043251742e-05, "loss": 0.0787, "step": 103985 }, { "epoch": 2.43, "learning_rate": 1.1880206988805708e-05, "loss": 0.1057, "step": 103990 }, { "epoch": 2.43, "learning_rate": 1.1879326934359672e-05, "loss": 0.0331, "step": 103995 }, { "epoch": 2.43, "learning_rate": 1.1878446879913639e-05, "loss": 0.0463, "step": 104000 }, { "epoch": 2.43, "learning_rate": 1.1877566825467602e-05, "loss": 0.0752, "step": 104005 }, { "epoch": 2.43, "learning_rate": 1.1876686771021569e-05, "loss": 0.0848, "step": 104010 }, { "epoch": 2.43, "learning_rate": 1.1875806716575532e-05, "loss": 0.2385, "step": 104015 }, { "epoch": 2.43, "learning_rate": 1.1874926662129497e-05, "loss": 0.3019, "step": 104020 }, { "epoch": 2.43, "learning_rate": 1.1874046607683462e-05, "loss": 0.2443, "step": 104025 }, { "epoch": 2.43, "learning_rate": 1.1873166553237427e-05, "loss": 0.1976, "step": 104030 }, { "epoch": 2.43, "learning_rate": 1.1872286498791392e-05, "loss": 0.0378, "step": 104035 }, { "epoch": 2.43, "learning_rate": 1.1871406444345357e-05, "loss": 0.0033, "step": 104040 }, { "epoch": 2.43, "learning_rate": 1.1870526389899322e-05, "loss": 0.1032, "step": 104045 }, { "epoch": 2.43, "learning_rate": 1.1869646335453287e-05, "loss": 0.0352, "step": 104050 }, { "epoch": 2.43, "learning_rate": 1.186876628100725e-05, "loss": 0.0158, "step": 104055 }, { "epoch": 2.43, "learning_rate": 1.1867886226561217e-05, "loss": 0.1045, "step": 104060 }, { "epoch": 2.43, "learning_rate": 1.1867006172115182e-05, "loss": 0.1769, "step": 104065 }, { "epoch": 2.43, "learning_rate": 1.1866126117669147e-05, "loss": 0.2246, "step": 104070 }, { "epoch": 2.43, "learning_rate": 1.1865246063223112e-05, "loss": 0.2705, "step": 104075 }, { "epoch": 2.43, "learning_rate": 1.1864366008777077e-05, "loss": 0.0103, "step": 104080 }, { "epoch": 2.43, "learning_rate": 1.1863485954331042e-05, "loss": 0.0021, "step": 104085 }, { "epoch": 2.43, "learning_rate": 1.1862605899885007e-05, "loss": 0.1195, "step": 104090 }, { "epoch": 2.43, "learning_rate": 1.1861725845438972e-05, "loss": 0.0779, "step": 104095 }, { "epoch": 2.43, "learning_rate": 1.1860845790992936e-05, "loss": 0.1099, "step": 104100 }, { "epoch": 2.43, "learning_rate": 1.1859965736546902e-05, "loss": 0.1265, "step": 104105 }, { "epoch": 2.43, "learning_rate": 1.1859085682100866e-05, "loss": 0.1684, "step": 104110 }, { "epoch": 2.43, "learning_rate": 1.1858205627654832e-05, "loss": 0.0476, "step": 104115 }, { "epoch": 2.43, "learning_rate": 1.1857325573208796e-05, "loss": 0.1711, "step": 104120 }, { "epoch": 2.43, "learning_rate": 1.1856445518762762e-05, "loss": 0.309, "step": 104125 }, { "epoch": 2.43, "learning_rate": 1.1855565464316726e-05, "loss": 0.06, "step": 104130 }, { "epoch": 2.43, "learning_rate": 1.185468540987069e-05, "loss": 0.0104, "step": 104135 }, { "epoch": 2.43, "learning_rate": 1.1853805355424656e-05, "loss": 0.0406, "step": 104140 }, { "epoch": 2.43, "learning_rate": 1.185292530097862e-05, "loss": 0.0959, "step": 104145 }, { "epoch": 2.43, "learning_rate": 1.1852045246532586e-05, "loss": 0.0344, "step": 104150 }, { "epoch": 2.43, "learning_rate": 1.185116519208655e-05, "loss": 0.1107, "step": 104155 }, { "epoch": 2.43, "learning_rate": 1.1850285137640516e-05, "loss": 0.1466, "step": 104160 }, { "epoch": 2.43, "learning_rate": 1.184940508319448e-05, "loss": 0.0836, "step": 104165 }, { "epoch": 2.43, "learning_rate": 1.1848525028748444e-05, "loss": 0.1352, "step": 104170 }, { "epoch": 2.43, "learning_rate": 1.184764497430241e-05, "loss": 0.1979, "step": 104175 }, { "epoch": 2.43, "learning_rate": 1.1846764919856374e-05, "loss": 0.0667, "step": 104180 }, { "epoch": 2.43, "learning_rate": 1.1845884865410341e-05, "loss": 0.0152, "step": 104185 }, { "epoch": 2.43, "learning_rate": 1.1845004810964304e-05, "loss": 0.0111, "step": 104190 }, { "epoch": 2.43, "learning_rate": 1.1844124756518271e-05, "loss": 0.0782, "step": 104195 }, { "epoch": 2.43, "learning_rate": 1.1843244702072234e-05, "loss": 0.112, "step": 104200 }, { "epoch": 2.43, "learning_rate": 1.18423646476262e-05, "loss": 0.1989, "step": 104205 }, { "epoch": 2.43, "learning_rate": 1.1841484593180166e-05, "loss": 0.0895, "step": 104210 }, { "epoch": 2.43, "learning_rate": 1.184060453873413e-05, "loss": 0.2018, "step": 104215 }, { "epoch": 2.43, "learning_rate": 1.1839724484288096e-05, "loss": 0.1899, "step": 104220 }, { "epoch": 2.43, "learning_rate": 1.183884442984206e-05, "loss": 0.1324, "step": 104225 }, { "epoch": 2.43, "learning_rate": 1.1837964375396026e-05, "loss": 0.0436, "step": 104230 }, { "epoch": 2.43, "learning_rate": 1.183708432094999e-05, "loss": 0.0614, "step": 104235 }, { "epoch": 2.43, "learning_rate": 1.1836204266503956e-05, "loss": 0.0249, "step": 104240 }, { "epoch": 2.43, "learning_rate": 1.183532421205792e-05, "loss": 0.0344, "step": 104245 }, { "epoch": 2.43, "learning_rate": 1.1834444157611884e-05, "loss": 0.0537, "step": 104250 }, { "epoch": 2.43, "learning_rate": 1.183356410316585e-05, "loss": 0.0178, "step": 104255 }, { "epoch": 2.43, "learning_rate": 1.1832684048719814e-05, "loss": 0.0083, "step": 104260 }, { "epoch": 2.43, "learning_rate": 1.183180399427378e-05, "loss": 0.1552, "step": 104265 }, { "epoch": 2.43, "learning_rate": 1.1830923939827745e-05, "loss": 0.1199, "step": 104270 }, { "epoch": 2.43, "learning_rate": 1.183004388538171e-05, "loss": 0.2728, "step": 104275 }, { "epoch": 2.43, "learning_rate": 1.1829163830935675e-05, "loss": 0.0547, "step": 104280 }, { "epoch": 2.43, "learning_rate": 1.1828283776489638e-05, "loss": 0.0062, "step": 104285 }, { "epoch": 2.43, "learning_rate": 1.1827403722043605e-05, "loss": 0.025, "step": 104290 }, { "epoch": 2.43, "learning_rate": 1.1826523667597568e-05, "loss": 0.0863, "step": 104295 }, { "epoch": 2.43, "learning_rate": 1.1825643613151535e-05, "loss": 0.1677, "step": 104300 }, { "epoch": 2.43, "learning_rate": 1.1824763558705498e-05, "loss": 0.0327, "step": 104305 }, { "epoch": 2.43, "learning_rate": 1.1823883504259465e-05, "loss": 0.0707, "step": 104310 }, { "epoch": 2.43, "learning_rate": 1.1823003449813428e-05, "loss": 0.3189, "step": 104315 }, { "epoch": 2.43, "learning_rate": 1.1822123395367393e-05, "loss": 0.3321, "step": 104320 }, { "epoch": 2.43, "learning_rate": 1.1821243340921358e-05, "loss": 0.1474, "step": 104325 }, { "epoch": 2.43, "learning_rate": 1.1820363286475323e-05, "loss": 0.0157, "step": 104330 }, { "epoch": 2.43, "learning_rate": 1.1819483232029288e-05, "loss": 0.0389, "step": 104335 }, { "epoch": 2.43, "learning_rate": 1.1818603177583253e-05, "loss": 0.0348, "step": 104340 }, { "epoch": 2.43, "learning_rate": 1.181772312313722e-05, "loss": 0.0233, "step": 104345 }, { "epoch": 2.43, "learning_rate": 1.1816843068691183e-05, "loss": 0.0969, "step": 104350 }, { "epoch": 2.43, "learning_rate": 1.1815963014245148e-05, "loss": 0.126, "step": 104355 }, { "epoch": 2.43, "learning_rate": 1.1815082959799113e-05, "loss": 0.1457, "step": 104360 }, { "epoch": 2.43, "learning_rate": 1.1814202905353078e-05, "loss": 0.0696, "step": 104365 }, { "epoch": 2.44, "learning_rate": 1.1813322850907043e-05, "loss": 0.1837, "step": 104370 }, { "epoch": 2.44, "learning_rate": 1.1812442796461008e-05, "loss": 0.155, "step": 104375 }, { "epoch": 2.44, "learning_rate": 1.1811562742014973e-05, "loss": 0.0404, "step": 104380 }, { "epoch": 2.44, "learning_rate": 1.1810682687568938e-05, "loss": 0.0162, "step": 104385 }, { "epoch": 2.44, "learning_rate": 1.1809802633122903e-05, "loss": 0.0377, "step": 104390 }, { "epoch": 2.44, "learning_rate": 1.1808922578676868e-05, "loss": 0.1715, "step": 104395 }, { "epoch": 2.44, "learning_rate": 1.1808042524230832e-05, "loss": 0.0069, "step": 104400 }, { "epoch": 2.44, "learning_rate": 1.1807162469784798e-05, "loss": 0.0493, "step": 104405 }, { "epoch": 2.44, "learning_rate": 1.1806282415338762e-05, "loss": 0.0759, "step": 104410 }, { "epoch": 2.44, "learning_rate": 1.1805402360892728e-05, "loss": 0.2181, "step": 104415 }, { "epoch": 2.44, "learning_rate": 1.1804522306446692e-05, "loss": 0.2796, "step": 104420 }, { "epoch": 2.44, "learning_rate": 1.1803642252000658e-05, "loss": 0.3229, "step": 104425 }, { "epoch": 2.44, "learning_rate": 1.1802762197554622e-05, "loss": 0.0207, "step": 104430 }, { "epoch": 2.44, "learning_rate": 1.1801882143108587e-05, "loss": 0.0151, "step": 104435 }, { "epoch": 2.44, "learning_rate": 1.1801002088662552e-05, "loss": 0.0041, "step": 104440 }, { "epoch": 2.44, "learning_rate": 1.1800122034216517e-05, "loss": 0.0506, "step": 104445 }, { "epoch": 2.44, "learning_rate": 1.1799241979770482e-05, "loss": 0.0303, "step": 104450 }, { "epoch": 2.44, "learning_rate": 1.1798361925324447e-05, "loss": 0.1953, "step": 104455 }, { "epoch": 2.44, "learning_rate": 1.1797481870878412e-05, "loss": 0.0569, "step": 104460 }, { "epoch": 2.44, "learning_rate": 1.1796601816432377e-05, "loss": 0.0644, "step": 104465 }, { "epoch": 2.44, "learning_rate": 1.179572176198634e-05, "loss": 0.2456, "step": 104470 }, { "epoch": 2.44, "learning_rate": 1.1794841707540307e-05, "loss": 0.195, "step": 104475 }, { "epoch": 2.44, "learning_rate": 1.1793961653094272e-05, "loss": 0.1037, "step": 104480 }, { "epoch": 2.44, "learning_rate": 1.1793081598648237e-05, "loss": 0.0557, "step": 104485 }, { "epoch": 2.44, "learning_rate": 1.1792201544202202e-05, "loss": 0.0357, "step": 104490 }, { "epoch": 2.44, "learning_rate": 1.1791321489756167e-05, "loss": 0.0238, "step": 104495 }, { "epoch": 2.44, "learning_rate": 1.1790441435310132e-05, "loss": 0.0594, "step": 104500 }, { "epoch": 2.44, "learning_rate": 1.1789561380864095e-05, "loss": 0.264, "step": 104505 }, { "epoch": 2.44, "learning_rate": 1.1788681326418062e-05, "loss": 0.1058, "step": 104510 }, { "epoch": 2.44, "learning_rate": 1.1787801271972025e-05, "loss": 0.3636, "step": 104515 }, { "epoch": 2.44, "learning_rate": 1.1786921217525992e-05, "loss": 0.1406, "step": 104520 }, { "epoch": 2.44, "learning_rate": 1.1786041163079955e-05, "loss": 0.2324, "step": 104525 }, { "epoch": 2.44, "learning_rate": 1.1785161108633922e-05, "loss": 0.0561, "step": 104530 }, { "epoch": 2.44, "learning_rate": 1.1784281054187886e-05, "loss": 0.0104, "step": 104535 }, { "epoch": 2.44, "learning_rate": 1.1783400999741852e-05, "loss": 0.1104, "step": 104540 }, { "epoch": 2.44, "learning_rate": 1.1782520945295816e-05, "loss": 0.0419, "step": 104545 }, { "epoch": 2.44, "learning_rate": 1.178164089084978e-05, "loss": 0.1351, "step": 104550 }, { "epoch": 2.44, "learning_rate": 1.1780760836403746e-05, "loss": 0.0548, "step": 104555 }, { "epoch": 2.44, "learning_rate": 1.177988078195771e-05, "loss": 0.1101, "step": 104560 }, { "epoch": 2.44, "learning_rate": 1.1779000727511676e-05, "loss": 0.1367, "step": 104565 }, { "epoch": 2.44, "learning_rate": 1.177812067306564e-05, "loss": 0.2416, "step": 104570 }, { "epoch": 2.44, "learning_rate": 1.1777240618619606e-05, "loss": 0.1586, "step": 104575 }, { "epoch": 2.44, "learning_rate": 1.177636056417357e-05, "loss": 0.0633, "step": 104580 }, { "epoch": 2.44, "learning_rate": 1.1775480509727534e-05, "loss": 0.0644, "step": 104585 }, { "epoch": 2.44, "learning_rate": 1.17746004552815e-05, "loss": 0.0443, "step": 104590 }, { "epoch": 2.44, "learning_rate": 1.1773720400835464e-05, "loss": 0.0274, "step": 104595 }, { "epoch": 2.44, "learning_rate": 1.177284034638943e-05, "loss": 0.0343, "step": 104600 }, { "epoch": 2.44, "learning_rate": 1.1771960291943394e-05, "loss": 0.0866, "step": 104605 }, { "epoch": 2.44, "learning_rate": 1.177108023749736e-05, "loss": 0.0273, "step": 104610 }, { "epoch": 2.44, "learning_rate": 1.1770200183051326e-05, "loss": 0.1456, "step": 104615 }, { "epoch": 2.44, "learning_rate": 1.176932012860529e-05, "loss": 0.1707, "step": 104620 }, { "epoch": 2.44, "learning_rate": 1.1768440074159256e-05, "loss": 0.1263, "step": 104625 }, { "epoch": 2.44, "learning_rate": 1.176756001971322e-05, "loss": 0.038, "step": 104630 }, { "epoch": 2.44, "learning_rate": 1.1766679965267186e-05, "loss": 0.011, "step": 104635 }, { "epoch": 2.44, "learning_rate": 1.176579991082115e-05, "loss": 0.0128, "step": 104640 }, { "epoch": 2.44, "learning_rate": 1.1764919856375116e-05, "loss": 0.0342, "step": 104645 }, { "epoch": 2.44, "learning_rate": 1.176403980192908e-05, "loss": 0.0646, "step": 104650 }, { "epoch": 2.44, "learning_rate": 1.1763159747483044e-05, "loss": 0.0436, "step": 104655 }, { "epoch": 2.44, "learning_rate": 1.176227969303701e-05, "loss": 0.1096, "step": 104660 }, { "epoch": 2.44, "learning_rate": 1.1761399638590974e-05, "loss": 0.0881, "step": 104665 }, { "epoch": 2.44, "learning_rate": 1.176051958414494e-05, "loss": 0.1322, "step": 104670 }, { "epoch": 2.44, "learning_rate": 1.1759639529698904e-05, "loss": 0.1337, "step": 104675 }, { "epoch": 2.44, "learning_rate": 1.175875947525287e-05, "loss": 0.1869, "step": 104680 }, { "epoch": 2.44, "learning_rate": 1.1757879420806834e-05, "loss": 0.0366, "step": 104685 }, { "epoch": 2.44, "learning_rate": 1.17569993663608e-05, "loss": 0.0638, "step": 104690 }, { "epoch": 2.44, "learning_rate": 1.1756119311914764e-05, "loss": 0.0335, "step": 104695 }, { "epoch": 2.44, "learning_rate": 1.1755239257468728e-05, "loss": 0.0631, "step": 104700 }, { "epoch": 2.44, "learning_rate": 1.1754359203022695e-05, "loss": 0.2051, "step": 104705 }, { "epoch": 2.44, "learning_rate": 1.1753479148576658e-05, "loss": 0.0503, "step": 104710 }, { "epoch": 2.44, "learning_rate": 1.1752599094130625e-05, "loss": 0.0713, "step": 104715 }, { "epoch": 2.44, "learning_rate": 1.1751719039684588e-05, "loss": 0.3063, "step": 104720 }, { "epoch": 2.44, "learning_rate": 1.1750838985238555e-05, "loss": 0.2868, "step": 104725 }, { "epoch": 2.44, "learning_rate": 1.1749958930792518e-05, "loss": 0.0342, "step": 104730 }, { "epoch": 2.44, "learning_rate": 1.1749078876346483e-05, "loss": 0.0223, "step": 104735 }, { "epoch": 2.44, "learning_rate": 1.1748198821900448e-05, "loss": 0.0284, "step": 104740 }, { "epoch": 2.44, "learning_rate": 1.1747318767454413e-05, "loss": 0.019, "step": 104745 }, { "epoch": 2.44, "learning_rate": 1.174643871300838e-05, "loss": 0.0961, "step": 104750 }, { "epoch": 2.44, "learning_rate": 1.1745558658562343e-05, "loss": 0.0321, "step": 104755 }, { "epoch": 2.44, "learning_rate": 1.174467860411631e-05, "loss": 0.07, "step": 104760 }, { "epoch": 2.44, "learning_rate": 1.1743798549670273e-05, "loss": 0.1295, "step": 104765 }, { "epoch": 2.44, "learning_rate": 1.1742918495224238e-05, "loss": 0.2651, "step": 104770 }, { "epoch": 2.44, "learning_rate": 1.1742038440778203e-05, "loss": 0.2047, "step": 104775 }, { "epoch": 2.44, "learning_rate": 1.1741158386332168e-05, "loss": 0.027, "step": 104780 }, { "epoch": 2.44, "learning_rate": 1.1740278331886133e-05, "loss": 0.1812, "step": 104785 }, { "epoch": 2.44, "learning_rate": 1.1739398277440098e-05, "loss": 0.1513, "step": 104790 }, { "epoch": 2.44, "learning_rate": 1.1738518222994063e-05, "loss": 0.1169, "step": 104795 }, { "epoch": 2.45, "learning_rate": 1.1737638168548028e-05, "loss": 0.0746, "step": 104800 }, { "epoch": 2.45, "learning_rate": 1.1736758114101992e-05, "loss": 0.1029, "step": 104805 }, { "epoch": 2.45, "learning_rate": 1.1735878059655958e-05, "loss": 0.1066, "step": 104810 }, { "epoch": 2.45, "learning_rate": 1.1734998005209922e-05, "loss": 0.2094, "step": 104815 }, { "epoch": 2.45, "learning_rate": 1.1734117950763888e-05, "loss": 0.2722, "step": 104820 }, { "epoch": 2.45, "learning_rate": 1.1733237896317852e-05, "loss": 0.2663, "step": 104825 }, { "epoch": 2.45, "learning_rate": 1.1732357841871818e-05, "loss": 0.0129, "step": 104830 }, { "epoch": 2.45, "learning_rate": 1.1731477787425782e-05, "loss": 0.016, "step": 104835 }, { "epoch": 2.45, "learning_rate": 1.1730597732979748e-05, "loss": 0.0625, "step": 104840 }, { "epoch": 2.45, "learning_rate": 1.1729717678533712e-05, "loss": 0.0342, "step": 104845 }, { "epoch": 2.45, "learning_rate": 1.1728837624087677e-05, "loss": 0.0975, "step": 104850 }, { "epoch": 2.45, "learning_rate": 1.1727957569641642e-05, "loss": 0.091, "step": 104855 }, { "epoch": 2.45, "learning_rate": 1.1727077515195607e-05, "loss": 0.1097, "step": 104860 }, { "epoch": 2.45, "learning_rate": 1.1726197460749572e-05, "loss": 0.1501, "step": 104865 }, { "epoch": 2.45, "learning_rate": 1.1725317406303537e-05, "loss": 0.1712, "step": 104870 }, { "epoch": 2.45, "learning_rate": 1.1724437351857502e-05, "loss": 0.2337, "step": 104875 }, { "epoch": 2.45, "learning_rate": 1.1723557297411467e-05, "loss": 0.0145, "step": 104880 }, { "epoch": 2.45, "learning_rate": 1.1722677242965432e-05, "loss": 0.0076, "step": 104885 }, { "epoch": 2.45, "learning_rate": 1.1721797188519397e-05, "loss": 0.0074, "step": 104890 }, { "epoch": 2.45, "learning_rate": 1.1720917134073362e-05, "loss": 0.1719, "step": 104895 }, { "epoch": 2.45, "learning_rate": 1.1720037079627327e-05, "loss": 0.0508, "step": 104900 }, { "epoch": 2.45, "learning_rate": 1.1719157025181292e-05, "loss": 0.0627, "step": 104905 }, { "epoch": 2.45, "learning_rate": 1.1718276970735257e-05, "loss": 0.0237, "step": 104910 }, { "epoch": 2.45, "learning_rate": 1.1717396916289222e-05, "loss": 0.1591, "step": 104915 }, { "epoch": 2.45, "learning_rate": 1.1716516861843185e-05, "loss": 0.119, "step": 104920 }, { "epoch": 2.45, "learning_rate": 1.1715636807397152e-05, "loss": 0.2756, "step": 104925 }, { "epoch": 2.45, "learning_rate": 1.1714756752951115e-05, "loss": 0.0302, "step": 104930 }, { "epoch": 2.45, "learning_rate": 1.1713876698505082e-05, "loss": 0.0319, "step": 104935 }, { "epoch": 2.45, "learning_rate": 1.1712996644059045e-05, "loss": 0.0971, "step": 104940 }, { "epoch": 2.45, "learning_rate": 1.1712116589613012e-05, "loss": 0.0042, "step": 104945 }, { "epoch": 2.45, "learning_rate": 1.1711236535166975e-05, "loss": 0.1132, "step": 104950 }, { "epoch": 2.45, "learning_rate": 1.171035648072094e-05, "loss": 0.0952, "step": 104955 }, { "epoch": 2.45, "learning_rate": 1.1709476426274906e-05, "loss": 0.1281, "step": 104960 }, { "epoch": 2.45, "learning_rate": 1.170859637182887e-05, "loss": 0.303, "step": 104965 }, { "epoch": 2.45, "learning_rate": 1.1707716317382836e-05, "loss": 0.2999, "step": 104970 }, { "epoch": 2.45, "learning_rate": 1.17068362629368e-05, "loss": 0.3794, "step": 104975 }, { "epoch": 2.45, "learning_rate": 1.1705956208490766e-05, "loss": 0.0491, "step": 104980 }, { "epoch": 2.45, "learning_rate": 1.170507615404473e-05, "loss": 0.0616, "step": 104985 }, { "epoch": 2.45, "learning_rate": 1.1704196099598696e-05, "loss": 0.1042, "step": 104990 }, { "epoch": 2.45, "learning_rate": 1.170331604515266e-05, "loss": 0.0928, "step": 104995 }, { "epoch": 2.45, "learning_rate": 1.1702435990706624e-05, "loss": 0.1287, "step": 105000 }, { "epoch": 2.45, "learning_rate": 1.170155593626059e-05, "loss": 0.0908, "step": 105005 }, { "epoch": 2.45, "learning_rate": 1.1700675881814554e-05, "loss": 0.0561, "step": 105010 }, { "epoch": 2.45, "learning_rate": 1.169979582736852e-05, "loss": 0.2725, "step": 105015 }, { "epoch": 2.45, "learning_rate": 1.1698915772922486e-05, "loss": 0.2244, "step": 105020 }, { "epoch": 2.45, "learning_rate": 1.169803571847645e-05, "loss": 0.2396, "step": 105025 }, { "epoch": 2.45, "learning_rate": 1.1697155664030416e-05, "loss": 0.0324, "step": 105030 }, { "epoch": 2.45, "learning_rate": 1.1696275609584379e-05, "loss": 0.0011, "step": 105035 }, { "epoch": 2.45, "learning_rate": 1.1695395555138346e-05, "loss": 0.1426, "step": 105040 }, { "epoch": 2.45, "learning_rate": 1.169451550069231e-05, "loss": 0.0313, "step": 105045 }, { "epoch": 2.45, "learning_rate": 1.1693635446246276e-05, "loss": 0.1484, "step": 105050 }, { "epoch": 2.45, "learning_rate": 1.169275539180024e-05, "loss": 0.0584, "step": 105055 }, { "epoch": 2.45, "learning_rate": 1.1691875337354206e-05, "loss": 0.2023, "step": 105060 }, { "epoch": 2.45, "learning_rate": 1.169099528290817e-05, "loss": 0.1377, "step": 105065 }, { "epoch": 2.45, "learning_rate": 1.1690115228462134e-05, "loss": 0.1924, "step": 105070 }, { "epoch": 2.45, "learning_rate": 1.16892351740161e-05, "loss": 0.4141, "step": 105075 }, { "epoch": 2.45, "learning_rate": 1.1688355119570064e-05, "loss": 0.1001, "step": 105080 }, { "epoch": 2.45, "learning_rate": 1.168747506512403e-05, "loss": 0.0347, "step": 105085 }, { "epoch": 2.45, "learning_rate": 1.1686595010677994e-05, "loss": 0.0135, "step": 105090 }, { "epoch": 2.45, "learning_rate": 1.168571495623196e-05, "loss": 0.0643, "step": 105095 }, { "epoch": 2.45, "learning_rate": 1.1684834901785924e-05, "loss": 0.0248, "step": 105100 }, { "epoch": 2.45, "learning_rate": 1.1683954847339888e-05, "loss": 0.1069, "step": 105105 }, { "epoch": 2.45, "learning_rate": 1.1683074792893854e-05, "loss": 0.2213, "step": 105110 }, { "epoch": 2.45, "learning_rate": 1.1682194738447818e-05, "loss": 0.1405, "step": 105115 }, { "epoch": 2.45, "learning_rate": 1.1681314684001784e-05, "loss": 0.1293, "step": 105120 }, { "epoch": 2.45, "learning_rate": 1.1680434629555748e-05, "loss": 0.1606, "step": 105125 }, { "epoch": 2.45, "learning_rate": 1.1679554575109715e-05, "loss": 0.0424, "step": 105130 }, { "epoch": 2.45, "learning_rate": 1.1678674520663678e-05, "loss": 0.0116, "step": 105135 }, { "epoch": 2.45, "learning_rate": 1.1677794466217643e-05, "loss": 0.0528, "step": 105140 }, { "epoch": 2.45, "learning_rate": 1.1676914411771608e-05, "loss": 0.0772, "step": 105145 }, { "epoch": 2.45, "learning_rate": 1.1676034357325573e-05, "loss": 0.1423, "step": 105150 }, { "epoch": 2.45, "learning_rate": 1.167515430287954e-05, "loss": 0.2689, "step": 105155 }, { "epoch": 2.45, "learning_rate": 1.1674274248433503e-05, "loss": 0.0618, "step": 105160 }, { "epoch": 2.45, "learning_rate": 1.167339419398747e-05, "loss": 0.1278, "step": 105165 }, { "epoch": 2.45, "learning_rate": 1.1672514139541433e-05, "loss": 0.1806, "step": 105170 }, { "epoch": 2.45, "learning_rate": 1.16716340850954e-05, "loss": 0.2831, "step": 105175 }, { "epoch": 2.45, "learning_rate": 1.1670754030649363e-05, "loss": 0.0057, "step": 105180 }, { "epoch": 2.45, "learning_rate": 1.1669873976203328e-05, "loss": 0.0066, "step": 105185 }, { "epoch": 2.45, "learning_rate": 1.1668993921757293e-05, "loss": 0.1211, "step": 105190 }, { "epoch": 2.45, "learning_rate": 1.1668113867311258e-05, "loss": 0.2311, "step": 105195 }, { "epoch": 2.45, "learning_rate": 1.1667233812865223e-05, "loss": 0.1035, "step": 105200 }, { "epoch": 2.45, "learning_rate": 1.1666353758419188e-05, "loss": 0.0581, "step": 105205 }, { "epoch": 2.45, "learning_rate": 1.1665473703973153e-05, "loss": 0.0634, "step": 105210 }, { "epoch": 2.45, "learning_rate": 1.1664593649527118e-05, "loss": 0.1725, "step": 105215 }, { "epoch": 2.45, "learning_rate": 1.1663713595081081e-05, "loss": 0.1421, "step": 105220 }, { "epoch": 2.46, "learning_rate": 1.1662833540635048e-05, "loss": 0.2929, "step": 105225 }, { "epoch": 2.46, "learning_rate": 1.1661953486189012e-05, "loss": 0.0419, "step": 105230 }, { "epoch": 2.46, "learning_rate": 1.1661073431742978e-05, "loss": 0.0533, "step": 105235 }, { "epoch": 2.46, "learning_rate": 1.1660193377296942e-05, "loss": 0.0089, "step": 105240 }, { "epoch": 2.46, "learning_rate": 1.1659313322850908e-05, "loss": 0.0055, "step": 105245 }, { "epoch": 2.46, "learning_rate": 1.1658433268404872e-05, "loss": 0.0841, "step": 105250 }, { "epoch": 2.46, "learning_rate": 1.1657553213958837e-05, "loss": 0.0889, "step": 105255 }, { "epoch": 2.46, "learning_rate": 1.1656673159512802e-05, "loss": 0.1619, "step": 105260 }, { "epoch": 2.46, "learning_rate": 1.1655793105066767e-05, "loss": 0.0884, "step": 105265 }, { "epoch": 2.46, "learning_rate": 1.1654913050620732e-05, "loss": 0.1575, "step": 105270 }, { "epoch": 2.46, "learning_rate": 1.1654032996174697e-05, "loss": 0.2507, "step": 105275 }, { "epoch": 2.46, "learning_rate": 1.1653152941728662e-05, "loss": 0.069, "step": 105280 }, { "epoch": 2.46, "learning_rate": 1.1652272887282627e-05, "loss": 0.0441, "step": 105285 }, { "epoch": 2.46, "learning_rate": 1.1651392832836592e-05, "loss": 0.0308, "step": 105290 }, { "epoch": 2.46, "learning_rate": 1.1650512778390557e-05, "loss": 0.0857, "step": 105295 }, { "epoch": 2.46, "learning_rate": 1.1649632723944522e-05, "loss": 0.0862, "step": 105300 }, { "epoch": 2.46, "learning_rate": 1.1648752669498487e-05, "loss": 0.0211, "step": 105305 }, { "epoch": 2.46, "learning_rate": 1.1647872615052452e-05, "loss": 0.1008, "step": 105310 }, { "epoch": 2.46, "learning_rate": 1.1646992560606417e-05, "loss": 0.2315, "step": 105315 }, { "epoch": 2.46, "learning_rate": 1.1646112506160382e-05, "loss": 0.2641, "step": 105320 }, { "epoch": 2.46, "learning_rate": 1.1645232451714347e-05, "loss": 0.212, "step": 105325 }, { "epoch": 2.46, "learning_rate": 1.1644352397268312e-05, "loss": 0.0637, "step": 105330 }, { "epoch": 2.46, "learning_rate": 1.1643472342822275e-05, "loss": 0.0492, "step": 105335 }, { "epoch": 2.46, "learning_rate": 1.1642592288376242e-05, "loss": 0.0359, "step": 105340 }, { "epoch": 2.46, "learning_rate": 1.1641712233930205e-05, "loss": 0.0084, "step": 105345 }, { "epoch": 2.46, "learning_rate": 1.1640832179484172e-05, "loss": 0.0438, "step": 105350 }, { "epoch": 2.46, "learning_rate": 1.1639952125038135e-05, "loss": 0.1178, "step": 105355 }, { "epoch": 2.46, "learning_rate": 1.1639072070592102e-05, "loss": 0.0311, "step": 105360 }, { "epoch": 2.46, "learning_rate": 1.1638192016146065e-05, "loss": 0.0492, "step": 105365 }, { "epoch": 2.46, "learning_rate": 1.163731196170003e-05, "loss": 0.1685, "step": 105370 }, { "epoch": 2.46, "learning_rate": 1.1636431907253995e-05, "loss": 0.3071, "step": 105375 }, { "epoch": 2.46, "learning_rate": 1.163555185280796e-05, "loss": 0.046, "step": 105380 }, { "epoch": 2.46, "learning_rate": 1.1634671798361925e-05, "loss": 0.0294, "step": 105385 }, { "epoch": 2.46, "learning_rate": 1.163379174391589e-05, "loss": 0.1173, "step": 105390 }, { "epoch": 2.46, "learning_rate": 1.1632911689469856e-05, "loss": 0.1335, "step": 105395 }, { "epoch": 2.46, "learning_rate": 1.163203163502382e-05, "loss": 0.0866, "step": 105400 }, { "epoch": 2.46, "learning_rate": 1.1631151580577784e-05, "loss": 0.2319, "step": 105405 }, { "epoch": 2.46, "learning_rate": 1.163027152613175e-05, "loss": 0.065, "step": 105410 }, { "epoch": 2.46, "learning_rate": 1.1629391471685714e-05, "loss": 0.1698, "step": 105415 }, { "epoch": 2.46, "learning_rate": 1.162851141723968e-05, "loss": 0.243, "step": 105420 }, { "epoch": 2.46, "learning_rate": 1.1627631362793646e-05, "loss": 0.0734, "step": 105425 }, { "epoch": 2.46, "learning_rate": 1.162675130834761e-05, "loss": 0.0602, "step": 105430 }, { "epoch": 2.46, "learning_rate": 1.1625871253901576e-05, "loss": 0.0013, "step": 105435 }, { "epoch": 2.46, "learning_rate": 1.1624991199455539e-05, "loss": 0.0177, "step": 105440 }, { "epoch": 2.46, "learning_rate": 1.1624111145009506e-05, "loss": 0.0667, "step": 105445 }, { "epoch": 2.46, "learning_rate": 1.1623231090563469e-05, "loss": 0.0405, "step": 105450 }, { "epoch": 2.46, "learning_rate": 1.1622351036117436e-05, "loss": 0.2118, "step": 105455 }, { "epoch": 2.46, "learning_rate": 1.1621470981671399e-05, "loss": 0.0562, "step": 105460 }, { "epoch": 2.46, "learning_rate": 1.1620590927225366e-05, "loss": 0.1101, "step": 105465 }, { "epoch": 2.46, "learning_rate": 1.1619710872779329e-05, "loss": 0.1342, "step": 105470 }, { "epoch": 2.46, "learning_rate": 1.1618830818333296e-05, "loss": 0.1026, "step": 105475 }, { "epoch": 2.46, "learning_rate": 1.161795076388726e-05, "loss": 0.0787, "step": 105480 }, { "epoch": 2.46, "learning_rate": 1.1617070709441224e-05, "loss": 0.0224, "step": 105485 }, { "epoch": 2.46, "learning_rate": 1.161619065499519e-05, "loss": 0.0667, "step": 105490 }, { "epoch": 2.46, "learning_rate": 1.1615310600549154e-05, "loss": 0.0553, "step": 105495 }, { "epoch": 2.46, "learning_rate": 1.161443054610312e-05, "loss": 0.0591, "step": 105500 }, { "epoch": 2.46, "learning_rate": 1.1613550491657084e-05, "loss": 0.029, "step": 105505 }, { "epoch": 2.46, "learning_rate": 1.161267043721105e-05, "loss": 0.0241, "step": 105510 }, { "epoch": 2.46, "learning_rate": 1.1611790382765014e-05, "loss": 0.1195, "step": 105515 }, { "epoch": 2.46, "learning_rate": 1.1610910328318978e-05, "loss": 0.1723, "step": 105520 }, { "epoch": 2.46, "learning_rate": 1.1610030273872944e-05, "loss": 0.419, "step": 105525 }, { "epoch": 2.46, "learning_rate": 1.1609150219426908e-05, "loss": 0.0546, "step": 105530 }, { "epoch": 2.46, "learning_rate": 1.1608270164980874e-05, "loss": 0.0627, "step": 105535 }, { "epoch": 2.46, "learning_rate": 1.1607390110534838e-05, "loss": 0.0517, "step": 105540 }, { "epoch": 2.46, "learning_rate": 1.1606510056088804e-05, "loss": 0.0388, "step": 105545 }, { "epoch": 2.46, "learning_rate": 1.1605630001642768e-05, "loss": 0.082, "step": 105550 }, { "epoch": 2.46, "learning_rate": 1.1604749947196733e-05, "loss": 0.1648, "step": 105555 }, { "epoch": 2.46, "learning_rate": 1.1603869892750698e-05, "loss": 0.2404, "step": 105560 }, { "epoch": 2.46, "learning_rate": 1.1602989838304663e-05, "loss": 0.0493, "step": 105565 }, { "epoch": 2.46, "learning_rate": 1.160210978385863e-05, "loss": 0.2204, "step": 105570 }, { "epoch": 2.46, "learning_rate": 1.1601229729412593e-05, "loss": 0.2337, "step": 105575 }, { "epoch": 2.46, "learning_rate": 1.160034967496656e-05, "loss": 0.0022, "step": 105580 }, { "epoch": 2.46, "learning_rate": 1.1599469620520523e-05, "loss": 0.0924, "step": 105585 }, { "epoch": 2.46, "learning_rate": 1.1598589566074488e-05, "loss": 0.0161, "step": 105590 }, { "epoch": 2.46, "learning_rate": 1.1597709511628453e-05, "loss": 0.0163, "step": 105595 }, { "epoch": 2.46, "learning_rate": 1.1596829457182418e-05, "loss": 0.0385, "step": 105600 }, { "epoch": 2.46, "learning_rate": 1.1595949402736383e-05, "loss": 0.1501, "step": 105605 }, { "epoch": 2.46, "learning_rate": 1.1595069348290348e-05, "loss": 0.0838, "step": 105610 }, { "epoch": 2.46, "learning_rate": 1.1594189293844313e-05, "loss": 0.2975, "step": 105615 }, { "epoch": 2.46, "learning_rate": 1.1593309239398278e-05, "loss": 0.0679, "step": 105620 }, { "epoch": 2.46, "learning_rate": 1.1592429184952243e-05, "loss": 0.1537, "step": 105625 }, { "epoch": 2.46, "learning_rate": 1.1591549130506208e-05, "loss": 0.1395, "step": 105630 }, { "epoch": 2.46, "learning_rate": 1.1590669076060171e-05, "loss": 0.0033, "step": 105635 }, { "epoch": 2.46, "learning_rate": 1.1589789021614138e-05, "loss": 0.0916, "step": 105640 }, { "epoch": 2.46, "learning_rate": 1.1588908967168101e-05, "loss": 0.1184, "step": 105645 }, { "epoch": 2.46, "learning_rate": 1.1588028912722068e-05, "loss": 0.0336, "step": 105650 }, { "epoch": 2.47, "learning_rate": 1.1587148858276031e-05, "loss": 0.1418, "step": 105655 }, { "epoch": 2.47, "learning_rate": 1.1586268803829998e-05, "loss": 0.0798, "step": 105660 }, { "epoch": 2.47, "learning_rate": 1.1585388749383962e-05, "loss": 0.2288, "step": 105665 }, { "epoch": 2.47, "learning_rate": 1.1584508694937927e-05, "loss": 0.2001, "step": 105670 }, { "epoch": 2.47, "learning_rate": 1.1583628640491892e-05, "loss": 0.2999, "step": 105675 }, { "epoch": 2.47, "learning_rate": 1.1582748586045857e-05, "loss": 0.0352, "step": 105680 }, { "epoch": 2.47, "learning_rate": 1.1581868531599822e-05, "loss": 0.1032, "step": 105685 }, { "epoch": 2.47, "learning_rate": 1.1580988477153787e-05, "loss": 0.0173, "step": 105690 }, { "epoch": 2.47, "learning_rate": 1.1580108422707752e-05, "loss": 0.044, "step": 105695 }, { "epoch": 2.47, "learning_rate": 1.1579228368261717e-05, "loss": 0.1041, "step": 105700 }, { "epoch": 2.47, "learning_rate": 1.1578348313815682e-05, "loss": 0.0456, "step": 105705 }, { "epoch": 2.47, "learning_rate": 1.1577468259369647e-05, "loss": 0.3645, "step": 105710 }, { "epoch": 2.47, "learning_rate": 1.1576588204923612e-05, "loss": 0.1097, "step": 105715 }, { "epoch": 2.47, "learning_rate": 1.1575708150477577e-05, "loss": 0.0893, "step": 105720 }, { "epoch": 2.47, "learning_rate": 1.1574828096031542e-05, "loss": 0.5959, "step": 105725 }, { "epoch": 2.47, "learning_rate": 1.1573948041585507e-05, "loss": 0.1356, "step": 105730 }, { "epoch": 2.47, "learning_rate": 1.1573067987139472e-05, "loss": 0.0568, "step": 105735 }, { "epoch": 2.47, "learning_rate": 1.1572187932693435e-05, "loss": 0.0742, "step": 105740 }, { "epoch": 2.47, "learning_rate": 1.1571307878247402e-05, "loss": 0.0445, "step": 105745 }, { "epoch": 2.47, "learning_rate": 1.1570427823801365e-05, "loss": 0.0261, "step": 105750 }, { "epoch": 2.47, "learning_rate": 1.1569547769355332e-05, "loss": 0.058, "step": 105755 }, { "epoch": 2.47, "learning_rate": 1.1568667714909295e-05, "loss": 0.1606, "step": 105760 }, { "epoch": 2.47, "learning_rate": 1.1567787660463262e-05, "loss": 0.1889, "step": 105765 }, { "epoch": 2.47, "learning_rate": 1.1566907606017225e-05, "loss": 0.4577, "step": 105770 }, { "epoch": 2.47, "learning_rate": 1.1566027551571192e-05, "loss": 0.3051, "step": 105775 }, { "epoch": 2.47, "learning_rate": 1.1565147497125155e-05, "loss": 0.077, "step": 105780 }, { "epoch": 2.47, "learning_rate": 1.156426744267912e-05, "loss": 0.0036, "step": 105785 }, { "epoch": 2.47, "learning_rate": 1.1563387388233085e-05, "loss": 0.0457, "step": 105790 }, { "epoch": 2.47, "learning_rate": 1.156250733378705e-05, "loss": 0.0615, "step": 105795 }, { "epoch": 2.47, "learning_rate": 1.1561627279341015e-05, "loss": 0.0506, "step": 105800 }, { "epoch": 2.47, "learning_rate": 1.156074722489498e-05, "loss": 0.0822, "step": 105805 }, { "epoch": 2.47, "learning_rate": 1.1559867170448945e-05, "loss": 0.0527, "step": 105810 }, { "epoch": 2.47, "learning_rate": 1.155898711600291e-05, "loss": 0.0859, "step": 105815 }, { "epoch": 2.47, "learning_rate": 1.1558107061556874e-05, "loss": 0.1898, "step": 105820 }, { "epoch": 2.47, "learning_rate": 1.155722700711084e-05, "loss": 0.2969, "step": 105825 }, { "epoch": 2.47, "learning_rate": 1.1556346952664806e-05, "loss": 0.0594, "step": 105830 }, { "epoch": 2.47, "learning_rate": 1.155546689821877e-05, "loss": 0.041, "step": 105835 }, { "epoch": 2.47, "learning_rate": 1.1554586843772736e-05, "loss": 0.079, "step": 105840 }, { "epoch": 2.47, "learning_rate": 1.15537067893267e-05, "loss": 0.0995, "step": 105845 }, { "epoch": 2.47, "learning_rate": 1.1552826734880666e-05, "loss": 0.1184, "step": 105850 }, { "epoch": 2.47, "learning_rate": 1.1551946680434629e-05, "loss": 0.122, "step": 105855 }, { "epoch": 2.47, "learning_rate": 1.1551066625988596e-05, "loss": 0.0979, "step": 105860 }, { "epoch": 2.47, "learning_rate": 1.1550186571542559e-05, "loss": 0.0589, "step": 105865 }, { "epoch": 2.47, "learning_rate": 1.1549306517096526e-05, "loss": 0.247, "step": 105870 }, { "epoch": 2.47, "learning_rate": 1.1548426462650489e-05, "loss": 0.1854, "step": 105875 }, { "epoch": 2.47, "learning_rate": 1.1547546408204456e-05, "loss": 0.0296, "step": 105880 }, { "epoch": 2.47, "learning_rate": 1.1546666353758419e-05, "loss": 0.0196, "step": 105885 }, { "epoch": 2.47, "learning_rate": 1.1545786299312384e-05, "loss": 0.0692, "step": 105890 }, { "epoch": 2.47, "learning_rate": 1.1544906244866349e-05, "loss": 0.0619, "step": 105895 }, { "epoch": 2.47, "learning_rate": 1.1544026190420314e-05, "loss": 0.0347, "step": 105900 }, { "epoch": 2.47, "learning_rate": 1.1543146135974279e-05, "loss": 0.119, "step": 105905 }, { "epoch": 2.47, "learning_rate": 1.1542266081528244e-05, "loss": 0.0361, "step": 105910 }, { "epoch": 2.47, "learning_rate": 1.154138602708221e-05, "loss": 0.2058, "step": 105915 }, { "epoch": 2.47, "learning_rate": 1.1540505972636174e-05, "loss": 0.3388, "step": 105920 }, { "epoch": 2.47, "learning_rate": 1.153962591819014e-05, "loss": 0.3629, "step": 105925 }, { "epoch": 2.47, "learning_rate": 1.1538745863744104e-05, "loss": 0.0069, "step": 105930 }, { "epoch": 2.47, "learning_rate": 1.1537865809298068e-05, "loss": 0.0685, "step": 105935 }, { "epoch": 2.47, "learning_rate": 1.1536985754852034e-05, "loss": 0.0742, "step": 105940 }, { "epoch": 2.47, "learning_rate": 1.1536105700405998e-05, "loss": 0.049, "step": 105945 }, { "epoch": 2.47, "learning_rate": 1.1535225645959964e-05, "loss": 0.0022, "step": 105950 }, { "epoch": 2.47, "learning_rate": 1.1534345591513928e-05, "loss": 0.0914, "step": 105955 }, { "epoch": 2.47, "learning_rate": 1.1533465537067894e-05, "loss": 0.0852, "step": 105960 }, { "epoch": 2.47, "learning_rate": 1.1532585482621858e-05, "loss": 0.113, "step": 105965 }, { "epoch": 2.47, "learning_rate": 1.1531705428175823e-05, "loss": 0.2158, "step": 105970 }, { "epoch": 2.47, "learning_rate": 1.153082537372979e-05, "loss": 0.1423, "step": 105975 }, { "epoch": 2.47, "learning_rate": 1.1529945319283753e-05, "loss": 0.0064, "step": 105980 }, { "epoch": 2.47, "learning_rate": 1.152906526483772e-05, "loss": 0.0016, "step": 105985 }, { "epoch": 2.47, "learning_rate": 1.1528185210391683e-05, "loss": 0.036, "step": 105990 }, { "epoch": 2.47, "learning_rate": 1.152730515594565e-05, "loss": 0.1096, "step": 105995 }, { "epoch": 2.47, "learning_rate": 1.1526425101499613e-05, "loss": 0.0827, "step": 106000 }, { "epoch": 2.47, "learning_rate": 1.1525545047053578e-05, "loss": 0.1227, "step": 106005 }, { "epoch": 2.47, "learning_rate": 1.1524664992607543e-05, "loss": 0.0956, "step": 106010 }, { "epoch": 2.47, "learning_rate": 1.1523784938161508e-05, "loss": 0.1993, "step": 106015 }, { "epoch": 2.47, "learning_rate": 1.1522904883715473e-05, "loss": 0.3241, "step": 106020 }, { "epoch": 2.47, "learning_rate": 1.1522024829269438e-05, "loss": 0.3615, "step": 106025 }, { "epoch": 2.47, "learning_rate": 1.1521144774823403e-05, "loss": 0.0762, "step": 106030 }, { "epoch": 2.47, "learning_rate": 1.1520264720377368e-05, "loss": 0.0638, "step": 106035 }, { "epoch": 2.47, "learning_rate": 1.1519384665931331e-05, "loss": 0.0256, "step": 106040 }, { "epoch": 2.47, "learning_rate": 1.1518504611485298e-05, "loss": 0.0441, "step": 106045 }, { "epoch": 2.47, "learning_rate": 1.1517624557039261e-05, "loss": 0.0674, "step": 106050 }, { "epoch": 2.47, "learning_rate": 1.1516744502593228e-05, "loss": 0.1298, "step": 106055 }, { "epoch": 2.47, "learning_rate": 1.1515864448147191e-05, "loss": 0.1928, "step": 106060 }, { "epoch": 2.47, "learning_rate": 1.1514984393701158e-05, "loss": 0.1203, "step": 106065 }, { "epoch": 2.47, "learning_rate": 1.1514104339255121e-05, "loss": 0.2072, "step": 106070 }, { "epoch": 2.47, "learning_rate": 1.1513224284809088e-05, "loss": 0.2188, "step": 106075 }, { "epoch": 2.47, "learning_rate": 1.1512344230363051e-05, "loss": 0.0027, "step": 106080 }, { "epoch": 2.48, "learning_rate": 1.1511464175917016e-05, "loss": 0.0106, "step": 106085 }, { "epoch": 2.48, "learning_rate": 1.1510584121470982e-05, "loss": 0.0352, "step": 106090 }, { "epoch": 2.48, "learning_rate": 1.1509704067024947e-05, "loss": 0.0692, "step": 106095 }, { "epoch": 2.48, "learning_rate": 1.1508824012578912e-05, "loss": 0.1146, "step": 106100 }, { "epoch": 2.48, "learning_rate": 1.1507943958132877e-05, "loss": 0.0667, "step": 106105 }, { "epoch": 2.48, "learning_rate": 1.1507063903686843e-05, "loss": 0.0125, "step": 106110 }, { "epoch": 2.48, "learning_rate": 1.1506183849240807e-05, "loss": 0.1421, "step": 106115 }, { "epoch": 2.48, "learning_rate": 1.1505303794794772e-05, "loss": 0.1615, "step": 106120 }, { "epoch": 2.48, "learning_rate": 1.1504423740348737e-05, "loss": 0.458, "step": 106125 }, { "epoch": 2.48, "learning_rate": 1.1503543685902702e-05, "loss": 0.0591, "step": 106130 }, { "epoch": 2.48, "learning_rate": 1.1502663631456667e-05, "loss": 0.0102, "step": 106135 }, { "epoch": 2.48, "learning_rate": 1.1501783577010632e-05, "loss": 0.021, "step": 106140 }, { "epoch": 2.48, "learning_rate": 1.1500903522564597e-05, "loss": 0.0402, "step": 106145 }, { "epoch": 2.48, "learning_rate": 1.1500023468118562e-05, "loss": 0.1119, "step": 106150 }, { "epoch": 2.48, "learning_rate": 1.1499143413672525e-05, "loss": 0.1188, "step": 106155 }, { "epoch": 2.48, "learning_rate": 1.1498263359226492e-05, "loss": 0.1061, "step": 106160 }, { "epoch": 2.48, "learning_rate": 1.1497383304780455e-05, "loss": 0.2884, "step": 106165 }, { "epoch": 2.48, "learning_rate": 1.1496503250334422e-05, "loss": 0.3617, "step": 106170 }, { "epoch": 2.48, "learning_rate": 1.1495623195888385e-05, "loss": 0.2053, "step": 106175 }, { "epoch": 2.48, "learning_rate": 1.1494743141442352e-05, "loss": 0.0054, "step": 106180 }, { "epoch": 2.48, "learning_rate": 1.1493863086996315e-05, "loss": 0.0128, "step": 106185 }, { "epoch": 2.48, "learning_rate": 1.149298303255028e-05, "loss": 0.0647, "step": 106190 }, { "epoch": 2.48, "learning_rate": 1.1492102978104245e-05, "loss": 0.0017, "step": 106195 }, { "epoch": 2.48, "learning_rate": 1.149122292365821e-05, "loss": 0.0608, "step": 106200 }, { "epoch": 2.48, "learning_rate": 1.1490342869212175e-05, "loss": 0.1358, "step": 106205 }, { "epoch": 2.48, "learning_rate": 1.148946281476614e-05, "loss": 0.0348, "step": 106210 }, { "epoch": 2.48, "learning_rate": 1.1488582760320105e-05, "loss": 0.2259, "step": 106215 }, { "epoch": 2.48, "learning_rate": 1.148770270587407e-05, "loss": 0.1307, "step": 106220 }, { "epoch": 2.48, "learning_rate": 1.1486822651428035e-05, "loss": 0.2559, "step": 106225 }, { "epoch": 2.48, "learning_rate": 1.1485942596982e-05, "loss": 0.0229, "step": 106230 }, { "epoch": 2.48, "learning_rate": 1.1485062542535964e-05, "loss": 0.0252, "step": 106235 }, { "epoch": 2.48, "learning_rate": 1.148418248808993e-05, "loss": 0.0037, "step": 106240 }, { "epoch": 2.48, "learning_rate": 1.1483302433643895e-05, "loss": 0.0244, "step": 106245 }, { "epoch": 2.48, "learning_rate": 1.148242237919786e-05, "loss": 0.046, "step": 106250 }, { "epoch": 2.48, "learning_rate": 1.1481542324751825e-05, "loss": 0.1394, "step": 106255 }, { "epoch": 2.48, "learning_rate": 1.148066227030579e-05, "loss": 0.1052, "step": 106260 }, { "epoch": 2.48, "learning_rate": 1.1479782215859756e-05, "loss": 0.2075, "step": 106265 }, { "epoch": 2.48, "learning_rate": 1.1478902161413719e-05, "loss": 0.1176, "step": 106270 }, { "epoch": 2.48, "learning_rate": 1.1478022106967686e-05, "loss": 0.3167, "step": 106275 }, { "epoch": 2.48, "learning_rate": 1.1477142052521649e-05, "loss": 0.0822, "step": 106280 }, { "epoch": 2.48, "learning_rate": 1.1476261998075616e-05, "loss": 0.0468, "step": 106285 }, { "epoch": 2.48, "learning_rate": 1.1475381943629579e-05, "loss": 0.0089, "step": 106290 }, { "epoch": 2.48, "learning_rate": 1.1474501889183546e-05, "loss": 0.0908, "step": 106295 }, { "epoch": 2.48, "learning_rate": 1.1473621834737509e-05, "loss": 0.0108, "step": 106300 }, { "epoch": 2.48, "learning_rate": 1.1472741780291474e-05, "loss": 0.1258, "step": 106305 }, { "epoch": 2.48, "learning_rate": 1.1471861725845439e-05, "loss": 0.0832, "step": 106310 }, { "epoch": 2.48, "learning_rate": 1.1470981671399404e-05, "loss": 0.0619, "step": 106315 }, { "epoch": 2.48, "learning_rate": 1.1470101616953369e-05, "loss": 0.2615, "step": 106320 }, { "epoch": 2.48, "learning_rate": 1.1469221562507334e-05, "loss": 0.203, "step": 106325 }, { "epoch": 2.48, "learning_rate": 1.1468341508061299e-05, "loss": 0.1046, "step": 106330 }, { "epoch": 2.48, "learning_rate": 1.1467461453615264e-05, "loss": 0.0263, "step": 106335 }, { "epoch": 2.48, "learning_rate": 1.1466581399169227e-05, "loss": 0.0156, "step": 106340 }, { "epoch": 2.48, "learning_rate": 1.1465701344723194e-05, "loss": 0.0787, "step": 106345 }, { "epoch": 2.48, "learning_rate": 1.1464821290277157e-05, "loss": 0.1002, "step": 106350 }, { "epoch": 2.48, "learning_rate": 1.1463941235831124e-05, "loss": 0.1543, "step": 106355 }, { "epoch": 2.48, "learning_rate": 1.1463061181385088e-05, "loss": 0.0405, "step": 106360 }, { "epoch": 2.48, "learning_rate": 1.1462181126939054e-05, "loss": 0.0975, "step": 106365 }, { "epoch": 2.48, "learning_rate": 1.1461301072493018e-05, "loss": 0.3022, "step": 106370 }, { "epoch": 2.48, "learning_rate": 1.1460421018046984e-05, "loss": 0.1905, "step": 106375 }, { "epoch": 2.48, "learning_rate": 1.145954096360095e-05, "loss": 0.0261, "step": 106380 }, { "epoch": 2.48, "learning_rate": 1.1458660909154913e-05, "loss": 0.0312, "step": 106385 }, { "epoch": 2.48, "learning_rate": 1.145778085470888e-05, "loss": 0.0524, "step": 106390 }, { "epoch": 2.48, "learning_rate": 1.1456900800262843e-05, "loss": 0.0175, "step": 106395 }, { "epoch": 2.48, "learning_rate": 1.145602074581681e-05, "loss": 0.0561, "step": 106400 }, { "epoch": 2.48, "learning_rate": 1.1455140691370773e-05, "loss": 0.1465, "step": 106405 }, { "epoch": 2.48, "learning_rate": 1.145426063692474e-05, "loss": 0.0578, "step": 106410 }, { "epoch": 2.48, "learning_rate": 1.1453380582478703e-05, "loss": 0.0889, "step": 106415 }, { "epoch": 2.48, "learning_rate": 1.1452500528032668e-05, "loss": 0.1731, "step": 106420 }, { "epoch": 2.48, "learning_rate": 1.1451620473586633e-05, "loss": 0.1927, "step": 106425 }, { "epoch": 2.48, "learning_rate": 1.1450740419140598e-05, "loss": 0.0157, "step": 106430 }, { "epoch": 2.48, "learning_rate": 1.1449860364694563e-05, "loss": 0.0289, "step": 106435 }, { "epoch": 2.48, "learning_rate": 1.1448980310248528e-05, "loss": 0.0729, "step": 106440 }, { "epoch": 2.48, "learning_rate": 1.1448100255802493e-05, "loss": 0.0454, "step": 106445 }, { "epoch": 2.48, "learning_rate": 1.1447220201356458e-05, "loss": 0.0347, "step": 106450 }, { "epoch": 2.48, "learning_rate": 1.1446340146910421e-05, "loss": 0.0666, "step": 106455 }, { "epoch": 2.48, "learning_rate": 1.1445460092464388e-05, "loss": 0.1481, "step": 106460 }, { "epoch": 2.48, "learning_rate": 1.1444580038018351e-05, "loss": 0.1665, "step": 106465 }, { "epoch": 2.48, "learning_rate": 1.1443699983572318e-05, "loss": 0.1663, "step": 106470 }, { "epoch": 2.48, "learning_rate": 1.1442819929126281e-05, "loss": 0.4644, "step": 106475 }, { "epoch": 2.48, "learning_rate": 1.1441939874680248e-05, "loss": 0.0438, "step": 106480 }, { "epoch": 2.48, "learning_rate": 1.1441059820234211e-05, "loss": 0.0376, "step": 106485 }, { "epoch": 2.48, "learning_rate": 1.1440179765788176e-05, "loss": 0.0735, "step": 106490 }, { "epoch": 2.48, "learning_rate": 1.1439299711342141e-05, "loss": 0.0432, "step": 106495 }, { "epoch": 2.48, "learning_rate": 1.1438419656896106e-05, "loss": 0.0987, "step": 106500 }, { "epoch": 2.48, "learning_rate": 1.1437539602450071e-05, "loss": 0.094, "step": 106505 }, { "epoch": 2.49, "learning_rate": 1.1436659548004036e-05, "loss": 0.1852, "step": 106510 }, { "epoch": 2.49, "learning_rate": 1.1435779493558003e-05, "loss": 0.0691, "step": 106515 }, { "epoch": 2.49, "learning_rate": 1.1434899439111966e-05, "loss": 0.3113, "step": 106520 }, { "epoch": 2.49, "learning_rate": 1.1434019384665933e-05, "loss": 0.2411, "step": 106525 }, { "epoch": 2.49, "learning_rate": 1.1433139330219897e-05, "loss": 0.0149, "step": 106530 }, { "epoch": 2.49, "learning_rate": 1.1432259275773862e-05, "loss": 0.0047, "step": 106535 }, { "epoch": 2.49, "learning_rate": 1.1431379221327827e-05, "loss": 0.0182, "step": 106540 }, { "epoch": 2.49, "learning_rate": 1.1430499166881792e-05, "loss": 0.0661, "step": 106545 }, { "epoch": 2.49, "learning_rate": 1.1429619112435757e-05, "loss": 0.0631, "step": 106550 }, { "epoch": 2.49, "learning_rate": 1.1428739057989722e-05, "loss": 0.2248, "step": 106555 }, { "epoch": 2.49, "learning_rate": 1.1427859003543687e-05, "loss": 0.0531, "step": 106560 }, { "epoch": 2.49, "learning_rate": 1.1426978949097652e-05, "loss": 0.0348, "step": 106565 }, { "epoch": 2.49, "learning_rate": 1.1426098894651615e-05, "loss": 0.202, "step": 106570 }, { "epoch": 2.49, "learning_rate": 1.1425218840205582e-05, "loss": 0.2672, "step": 106575 }, { "epoch": 2.49, "learning_rate": 1.1424338785759545e-05, "loss": 0.0234, "step": 106580 }, { "epoch": 2.49, "learning_rate": 1.1423458731313512e-05, "loss": 0.0328, "step": 106585 }, { "epoch": 2.49, "learning_rate": 1.1422578676867475e-05, "loss": 0.0188, "step": 106590 }, { "epoch": 2.49, "learning_rate": 1.1421698622421442e-05, "loss": 0.0667, "step": 106595 }, { "epoch": 2.49, "learning_rate": 1.1420818567975405e-05, "loss": 0.049, "step": 106600 }, { "epoch": 2.49, "learning_rate": 1.141993851352937e-05, "loss": 0.0847, "step": 106605 }, { "epoch": 2.49, "learning_rate": 1.1419058459083335e-05, "loss": 0.0619, "step": 106610 }, { "epoch": 2.49, "learning_rate": 1.14181784046373e-05, "loss": 0.1464, "step": 106615 }, { "epoch": 2.49, "learning_rate": 1.1417298350191265e-05, "loss": 0.1691, "step": 106620 }, { "epoch": 2.49, "learning_rate": 1.141641829574523e-05, "loss": 0.2659, "step": 106625 }, { "epoch": 2.49, "learning_rate": 1.1415538241299195e-05, "loss": 0.0289, "step": 106630 }, { "epoch": 2.49, "learning_rate": 1.141465818685316e-05, "loss": 0.027, "step": 106635 }, { "epoch": 2.49, "learning_rate": 1.1413778132407124e-05, "loss": 0.0829, "step": 106640 }, { "epoch": 2.49, "learning_rate": 1.141289807796109e-05, "loss": 0.0079, "step": 106645 }, { "epoch": 2.49, "learning_rate": 1.1412018023515055e-05, "loss": 0.0483, "step": 106650 }, { "epoch": 2.49, "learning_rate": 1.141113796906902e-05, "loss": 0.0662, "step": 106655 }, { "epoch": 2.49, "learning_rate": 1.1410257914622985e-05, "loss": 0.2006, "step": 106660 }, { "epoch": 2.49, "learning_rate": 1.140937786017695e-05, "loss": 0.1069, "step": 106665 }, { "epoch": 2.49, "learning_rate": 1.1408497805730915e-05, "loss": 0.1378, "step": 106670 }, { "epoch": 2.49, "learning_rate": 1.140761775128488e-05, "loss": 0.2207, "step": 106675 }, { "epoch": 2.49, "learning_rate": 1.1406737696838845e-05, "loss": 0.0343, "step": 106680 }, { "epoch": 2.49, "learning_rate": 1.1405857642392809e-05, "loss": 0.0149, "step": 106685 }, { "epoch": 2.49, "learning_rate": 1.1404977587946776e-05, "loss": 0.0356, "step": 106690 }, { "epoch": 2.49, "learning_rate": 1.1404273544389945e-05, "loss": 0.0998, "step": 106695 }, { "epoch": 2.49, "learning_rate": 1.1403393489943912e-05, "loss": 0.0163, "step": 106700 }, { "epoch": 2.49, "learning_rate": 1.1402513435497877e-05, "loss": 0.1023, "step": 106705 }, { "epoch": 2.49, "learning_rate": 1.1401633381051842e-05, "loss": 0.0156, "step": 106710 }, { "epoch": 2.49, "learning_rate": 1.1400753326605807e-05, "loss": 0.1394, "step": 106715 }, { "epoch": 2.49, "learning_rate": 1.1399873272159772e-05, "loss": 0.1612, "step": 106720 }, { "epoch": 2.49, "learning_rate": 1.1398993217713737e-05, "loss": 0.3175, "step": 106725 }, { "epoch": 2.49, "learning_rate": 1.13981131632677e-05, "loss": 0.034, "step": 106730 }, { "epoch": 2.49, "learning_rate": 1.1397233108821667e-05, "loss": 0.0153, "step": 106735 }, { "epoch": 2.49, "learning_rate": 1.139635305437563e-05, "loss": 0.0466, "step": 106740 }, { "epoch": 2.49, "learning_rate": 1.1395472999929597e-05, "loss": 0.0561, "step": 106745 }, { "epoch": 2.49, "learning_rate": 1.139459294548356e-05, "loss": 0.1648, "step": 106750 }, { "epoch": 2.49, "learning_rate": 1.1393712891037527e-05, "loss": 0.0808, "step": 106755 }, { "epoch": 2.49, "learning_rate": 1.139283283659149e-05, "loss": 0.2291, "step": 106760 }, { "epoch": 2.49, "learning_rate": 1.1391952782145455e-05, "loss": 0.0927, "step": 106765 }, { "epoch": 2.49, "learning_rate": 1.139107272769942e-05, "loss": 0.3224, "step": 106770 }, { "epoch": 2.49, "learning_rate": 1.1390192673253385e-05, "loss": 0.196, "step": 106775 }, { "epoch": 2.49, "learning_rate": 1.138931261880735e-05, "loss": 0.0642, "step": 106780 }, { "epoch": 2.49, "learning_rate": 1.1388432564361316e-05, "loss": 0.0322, "step": 106785 }, { "epoch": 2.49, "learning_rate": 1.138755250991528e-05, "loss": 0.0446, "step": 106790 }, { "epoch": 2.49, "learning_rate": 1.1386672455469246e-05, "loss": 0.032, "step": 106795 }, { "epoch": 2.49, "learning_rate": 1.1385792401023209e-05, "loss": 0.0673, "step": 106800 }, { "epoch": 2.49, "learning_rate": 1.1384912346577176e-05, "loss": 0.1325, "step": 106805 }, { "epoch": 2.49, "learning_rate": 1.1384032292131139e-05, "loss": 0.0921, "step": 106810 }, { "epoch": 2.49, "learning_rate": 1.1383152237685106e-05, "loss": 0.0104, "step": 106815 }, { "epoch": 2.49, "learning_rate": 1.1382272183239069e-05, "loss": 0.1736, "step": 106820 }, { "epoch": 2.49, "learning_rate": 1.1381392128793036e-05, "loss": 0.254, "step": 106825 }, { "epoch": 2.49, "learning_rate": 1.1380512074346999e-05, "loss": 0.0597, "step": 106830 }, { "epoch": 2.49, "learning_rate": 1.1379632019900964e-05, "loss": 0.0418, "step": 106835 }, { "epoch": 2.49, "learning_rate": 1.137875196545493e-05, "loss": 0.0038, "step": 106840 }, { "epoch": 2.49, "learning_rate": 1.1377871911008894e-05, "loss": 0.0329, "step": 106845 }, { "epoch": 2.49, "learning_rate": 1.137699185656286e-05, "loss": 0.0836, "step": 106850 }, { "epoch": 2.49, "learning_rate": 1.1376111802116824e-05, "loss": 0.1325, "step": 106855 }, { "epoch": 2.49, "learning_rate": 1.137523174767079e-05, "loss": 0.1856, "step": 106860 }, { "epoch": 2.49, "learning_rate": 1.1374351693224754e-05, "loss": 0.1604, "step": 106865 }, { "epoch": 2.49, "learning_rate": 1.1373471638778721e-05, "loss": 0.2255, "step": 106870 }, { "epoch": 2.49, "learning_rate": 1.1372591584332684e-05, "loss": 0.274, "step": 106875 }, { "epoch": 2.49, "learning_rate": 1.137171152988665e-05, "loss": 0.0882, "step": 106880 }, { "epoch": 2.49, "learning_rate": 1.1370831475440614e-05, "loss": 0.0229, "step": 106885 }, { "epoch": 2.49, "learning_rate": 1.136995142099458e-05, "loss": 0.1, "step": 106890 }, { "epoch": 2.49, "learning_rate": 1.1369071366548544e-05, "loss": 0.0725, "step": 106895 }, { "epoch": 2.49, "learning_rate": 1.136819131210251e-05, "loss": 0.0502, "step": 106900 }, { "epoch": 2.49, "learning_rate": 1.1367311257656474e-05, "loss": 0.0292, "step": 106905 }, { "epoch": 2.49, "learning_rate": 1.136643120321044e-05, "loss": 0.0661, "step": 106910 }, { "epoch": 2.49, "learning_rate": 1.1365551148764403e-05, "loss": 0.2041, "step": 106915 }, { "epoch": 2.49, "learning_rate": 1.136467109431837e-05, "loss": 0.0859, "step": 106920 }, { "epoch": 2.49, "learning_rate": 1.1363791039872333e-05, "loss": 0.206, "step": 106925 }, { "epoch": 2.49, "learning_rate": 1.13629109854263e-05, "loss": 0.0344, "step": 106930 }, { "epoch": 2.49, "learning_rate": 1.1362030930980263e-05, "loss": 0.0295, "step": 106935 }, { "epoch": 2.5, "learning_rate": 1.136115087653423e-05, "loss": 0.0518, "step": 106940 }, { "epoch": 2.5, "learning_rate": 1.1360270822088193e-05, "loss": 0.0066, "step": 106945 }, { "epoch": 2.5, "learning_rate": 1.1359390767642158e-05, "loss": 0.1456, "step": 106950 }, { "epoch": 2.5, "learning_rate": 1.1358510713196123e-05, "loss": 0.0839, "step": 106955 }, { "epoch": 2.5, "learning_rate": 1.1357630658750088e-05, "loss": 0.0718, "step": 106960 }, { "epoch": 2.5, "learning_rate": 1.1356750604304053e-05, "loss": 0.1223, "step": 106965 }, { "epoch": 2.5, "learning_rate": 1.1355870549858018e-05, "loss": 0.1462, "step": 106970 }, { "epoch": 2.5, "learning_rate": 1.1354990495411983e-05, "loss": 0.1686, "step": 106975 }, { "epoch": 2.5, "learning_rate": 1.1354110440965948e-05, "loss": 0.0056, "step": 106980 }, { "epoch": 2.5, "learning_rate": 1.1353230386519913e-05, "loss": 0.0092, "step": 106985 }, { "epoch": 2.5, "learning_rate": 1.1352350332073878e-05, "loss": 0.0433, "step": 106990 }, { "epoch": 2.5, "learning_rate": 1.1351470277627843e-05, "loss": 0.0118, "step": 106995 }, { "epoch": 2.5, "learning_rate": 1.1350590223181808e-05, "loss": 0.082, "step": 107000 }, { "epoch": 2.5, "learning_rate": 1.1349710168735773e-05, "loss": 0.1213, "step": 107005 }, { "epoch": 2.5, "learning_rate": 1.1348830114289738e-05, "loss": 0.1007, "step": 107010 }, { "epoch": 2.5, "learning_rate": 1.1347950059843703e-05, "loss": 0.1354, "step": 107015 }, { "epoch": 2.5, "learning_rate": 1.1347070005397668e-05, "loss": 0.0977, "step": 107020 }, { "epoch": 2.5, "learning_rate": 1.1346189950951633e-05, "loss": 0.1055, "step": 107025 }, { "epoch": 2.5, "learning_rate": 1.1345309896505596e-05, "loss": 0.0672, "step": 107030 }, { "epoch": 2.5, "learning_rate": 1.1344429842059563e-05, "loss": 0.0265, "step": 107035 }, { "epoch": 2.5, "learning_rate": 1.1343549787613526e-05, "loss": 0.0678, "step": 107040 }, { "epoch": 2.5, "learning_rate": 1.1342669733167493e-05, "loss": 0.0953, "step": 107045 }, { "epoch": 2.5, "learning_rate": 1.1341789678721457e-05, "loss": 0.0102, "step": 107050 }, { "epoch": 2.5, "learning_rate": 1.1340909624275423e-05, "loss": 0.1811, "step": 107055 }, { "epoch": 2.5, "learning_rate": 1.1340029569829387e-05, "loss": 0.1181, "step": 107060 }, { "epoch": 2.5, "learning_rate": 1.1339149515383352e-05, "loss": 0.0625, "step": 107065 }, { "epoch": 2.5, "learning_rate": 1.1338269460937317e-05, "loss": 0.3177, "step": 107070 }, { "epoch": 2.5, "learning_rate": 1.1337389406491282e-05, "loss": 0.15, "step": 107075 }, { "epoch": 2.5, "learning_rate": 1.1336509352045247e-05, "loss": 0.0115, "step": 107080 }, { "epoch": 2.5, "learning_rate": 1.1335629297599212e-05, "loss": 0.042, "step": 107085 }, { "epoch": 2.5, "learning_rate": 1.1334749243153177e-05, "loss": 0.037, "step": 107090 }, { "epoch": 2.5, "learning_rate": 1.1333869188707142e-05, "loss": 0.0468, "step": 107095 }, { "epoch": 2.5, "learning_rate": 1.1332989134261105e-05, "loss": 0.027, "step": 107100 }, { "epoch": 2.5, "learning_rate": 1.1332109079815072e-05, "loss": 0.0682, "step": 107105 }, { "epoch": 2.5, "learning_rate": 1.1331229025369037e-05, "loss": 0.0719, "step": 107110 }, { "epoch": 2.5, "learning_rate": 1.1330348970923002e-05, "loss": 0.1938, "step": 107115 }, { "epoch": 2.5, "learning_rate": 1.1329468916476967e-05, "loss": 0.0932, "step": 107120 }, { "epoch": 2.5, "learning_rate": 1.1328588862030932e-05, "loss": 0.0885, "step": 107125 }, { "epoch": 2.5, "learning_rate": 1.1327708807584897e-05, "loss": 0.0483, "step": 107130 }, { "epoch": 2.5, "learning_rate": 1.132682875313886e-05, "loss": 0.001, "step": 107135 }, { "epoch": 2.5, "learning_rate": 1.1325948698692827e-05, "loss": 0.056, "step": 107140 }, { "epoch": 2.5, "learning_rate": 1.132506864424679e-05, "loss": 0.031, "step": 107145 }, { "epoch": 2.5, "learning_rate": 1.1324188589800757e-05, "loss": 0.0869, "step": 107150 }, { "epoch": 2.5, "learning_rate": 1.132330853535472e-05, "loss": 0.0167, "step": 107155 }, { "epoch": 2.5, "learning_rate": 1.1322428480908687e-05, "loss": 0.2455, "step": 107160 }, { "epoch": 2.5, "learning_rate": 1.132154842646265e-05, "loss": 0.1019, "step": 107165 }, { "epoch": 2.5, "learning_rate": 1.1320668372016617e-05, "loss": 0.2622, "step": 107170 }, { "epoch": 2.5, "learning_rate": 1.131978831757058e-05, "loss": 0.147, "step": 107175 }, { "epoch": 2.5, "learning_rate": 1.1318908263124545e-05, "loss": 0.0566, "step": 107180 }, { "epoch": 2.5, "learning_rate": 1.131802820867851e-05, "loss": 0.0589, "step": 107185 }, { "epoch": 2.5, "learning_rate": 1.1317148154232475e-05, "loss": 0.0063, "step": 107190 }, { "epoch": 2.5, "learning_rate": 1.131626809978644e-05, "loss": 0.0156, "step": 107195 }, { "epoch": 2.5, "learning_rate": 1.1315388045340405e-05, "loss": 0.1531, "step": 107200 }, { "epoch": 2.5, "learning_rate": 1.131450799089437e-05, "loss": 0.0658, "step": 107205 }, { "epoch": 2.5, "learning_rate": 1.1313627936448335e-05, "loss": 0.073, "step": 107210 }, { "epoch": 2.5, "learning_rate": 1.1312747882002299e-05, "loss": 0.1033, "step": 107215 }, { "epoch": 2.5, "learning_rate": 1.1311867827556266e-05, "loss": 0.2589, "step": 107220 }, { "epoch": 2.5, "learning_rate": 1.1310987773110229e-05, "loss": 0.3835, "step": 107225 }, { "epoch": 2.5, "learning_rate": 1.1310107718664196e-05, "loss": 0.0225, "step": 107230 }, { "epoch": 2.5, "learning_rate": 1.1309227664218159e-05, "loss": 0.0293, "step": 107235 }, { "epoch": 2.5, "learning_rate": 1.1308347609772126e-05, "loss": 0.0777, "step": 107240 }, { "epoch": 2.5, "learning_rate": 1.1307467555326089e-05, "loss": 0.0535, "step": 107245 }, { "epoch": 2.5, "learning_rate": 1.1306587500880054e-05, "loss": 0.1096, "step": 107250 }, { "epoch": 2.5, "learning_rate": 1.130570744643402e-05, "loss": 0.0533, "step": 107255 }, { "epoch": 2.5, "learning_rate": 1.1304827391987984e-05, "loss": 0.0397, "step": 107260 }, { "epoch": 2.5, "learning_rate": 1.130394733754195e-05, "loss": 0.1904, "step": 107265 }, { "epoch": 2.5, "learning_rate": 1.1303067283095914e-05, "loss": 0.189, "step": 107270 }, { "epoch": 2.5, "learning_rate": 1.130218722864988e-05, "loss": 0.1846, "step": 107275 }, { "epoch": 2.5, "learning_rate": 1.1301307174203844e-05, "loss": 0.0352, "step": 107280 }, { "epoch": 2.5, "learning_rate": 1.1300427119757809e-05, "loss": 0.0266, "step": 107285 }, { "epoch": 2.5, "learning_rate": 1.1299547065311774e-05, "loss": 0.0648, "step": 107290 }, { "epoch": 2.5, "learning_rate": 1.1298667010865739e-05, "loss": 0.1028, "step": 107295 }, { "epoch": 2.5, "learning_rate": 1.1297786956419704e-05, "loss": 0.0151, "step": 107300 }, { "epoch": 2.5, "learning_rate": 1.129690690197367e-05, "loss": 0.0249, "step": 107305 }, { "epoch": 2.5, "learning_rate": 1.1296026847527634e-05, "loss": 0.0811, "step": 107310 }, { "epoch": 2.5, "learning_rate": 1.12951467930816e-05, "loss": 0.1049, "step": 107315 }, { "epoch": 2.5, "learning_rate": 1.1294266738635564e-05, "loss": 0.2278, "step": 107320 }, { "epoch": 2.5, "learning_rate": 1.129338668418953e-05, "loss": 0.2977, "step": 107325 }, { "epoch": 2.5, "learning_rate": 1.1292506629743493e-05, "loss": 0.0502, "step": 107330 }, { "epoch": 2.5, "learning_rate": 1.129162657529746e-05, "loss": 0.0233, "step": 107335 }, { "epoch": 2.5, "learning_rate": 1.1290746520851423e-05, "loss": 0.0997, "step": 107340 }, { "epoch": 2.5, "learning_rate": 1.128986646640539e-05, "loss": 0.0717, "step": 107345 }, { "epoch": 2.5, "learning_rate": 1.1288986411959353e-05, "loss": 0.1292, "step": 107350 }, { "epoch": 2.5, "learning_rate": 1.128810635751332e-05, "loss": 0.1484, "step": 107355 }, { "epoch": 2.5, "learning_rate": 1.1287226303067283e-05, "loss": 0.0082, "step": 107360 }, { "epoch": 2.5, "learning_rate": 1.1286346248621248e-05, "loss": 0.1304, "step": 107365 }, { "epoch": 2.51, "learning_rate": 1.1285466194175213e-05, "loss": 0.2011, "step": 107370 }, { "epoch": 2.51, "learning_rate": 1.1284586139729178e-05, "loss": 0.2535, "step": 107375 }, { "epoch": 2.51, "learning_rate": 1.1283706085283143e-05, "loss": 0.0519, "step": 107380 }, { "epoch": 2.51, "learning_rate": 1.1282826030837108e-05, "loss": 0.0403, "step": 107385 }, { "epoch": 2.51, "learning_rate": 1.1281945976391075e-05, "loss": 0.0326, "step": 107390 }, { "epoch": 2.51, "learning_rate": 1.1281065921945038e-05, "loss": 0.0184, "step": 107395 }, { "epoch": 2.51, "learning_rate": 1.1280185867499003e-05, "loss": 0.0723, "step": 107400 }, { "epoch": 2.51, "learning_rate": 1.1279305813052968e-05, "loss": 0.0715, "step": 107405 }, { "epoch": 2.51, "learning_rate": 1.1278425758606933e-05, "loss": 0.0604, "step": 107410 }, { "epoch": 2.51, "learning_rate": 1.1277545704160898e-05, "loss": 0.1037, "step": 107415 }, { "epoch": 2.51, "learning_rate": 1.1276665649714863e-05, "loss": 0.2242, "step": 107420 }, { "epoch": 2.51, "learning_rate": 1.1275785595268828e-05, "loss": 0.2016, "step": 107425 }, { "epoch": 2.51, "learning_rate": 1.1274905540822793e-05, "loss": 0.0379, "step": 107430 }, { "epoch": 2.51, "learning_rate": 1.1274025486376756e-05, "loss": 0.0395, "step": 107435 }, { "epoch": 2.51, "learning_rate": 1.1273145431930723e-05, "loss": 0.0646, "step": 107440 }, { "epoch": 2.51, "learning_rate": 1.1272265377484686e-05, "loss": 0.0452, "step": 107445 }, { "epoch": 2.51, "learning_rate": 1.1271385323038653e-05, "loss": 0.068, "step": 107450 }, { "epoch": 2.51, "learning_rate": 1.1270505268592616e-05, "loss": 0.0425, "step": 107455 }, { "epoch": 2.51, "learning_rate": 1.1269625214146583e-05, "loss": 0.0705, "step": 107460 }, { "epoch": 2.51, "learning_rate": 1.1268745159700546e-05, "loss": 0.1552, "step": 107465 }, { "epoch": 2.51, "learning_rate": 1.1267865105254513e-05, "loss": 0.2267, "step": 107470 }, { "epoch": 2.51, "learning_rate": 1.1266985050808476e-05, "loss": 0.1471, "step": 107475 }, { "epoch": 2.51, "learning_rate": 1.1266104996362442e-05, "loss": 0.0449, "step": 107480 }, { "epoch": 2.51, "learning_rate": 1.1265224941916407e-05, "loss": 0.0234, "step": 107485 }, { "epoch": 2.51, "learning_rate": 1.1264344887470372e-05, "loss": 0.0217, "step": 107490 }, { "epoch": 2.51, "learning_rate": 1.1263464833024337e-05, "loss": 0.1088, "step": 107495 }, { "epoch": 2.51, "learning_rate": 1.1262584778578302e-05, "loss": 0.123, "step": 107500 }, { "epoch": 2.51, "learning_rate": 1.1261704724132267e-05, "loss": 0.0337, "step": 107505 }, { "epoch": 2.51, "learning_rate": 1.1260824669686232e-05, "loss": 0.1517, "step": 107510 }, { "epoch": 2.51, "learning_rate": 1.1259944615240195e-05, "loss": 0.1756, "step": 107515 }, { "epoch": 2.51, "learning_rate": 1.1259064560794162e-05, "loss": 0.2906, "step": 107520 }, { "epoch": 2.51, "learning_rate": 1.1258184506348127e-05, "loss": 0.1658, "step": 107525 }, { "epoch": 2.51, "learning_rate": 1.1257304451902092e-05, "loss": 0.0237, "step": 107530 }, { "epoch": 2.51, "learning_rate": 1.1256424397456057e-05, "loss": 0.0434, "step": 107535 }, { "epoch": 2.51, "learning_rate": 1.1255544343010022e-05, "loss": 0.1036, "step": 107540 }, { "epoch": 2.51, "learning_rate": 1.1254664288563987e-05, "loss": 0.0353, "step": 107545 }, { "epoch": 2.51, "learning_rate": 1.125378423411795e-05, "loss": 0.0811, "step": 107550 }, { "epoch": 2.51, "learning_rate": 1.1252904179671917e-05, "loss": 0.0806, "step": 107555 }, { "epoch": 2.51, "learning_rate": 1.125202412522588e-05, "loss": 0.085, "step": 107560 }, { "epoch": 2.51, "learning_rate": 1.1251144070779847e-05, "loss": 0.0695, "step": 107565 }, { "epoch": 2.51, "learning_rate": 1.125026401633381e-05, "loss": 0.1584, "step": 107570 }, { "epoch": 2.51, "learning_rate": 1.1249383961887777e-05, "loss": 0.2032, "step": 107575 }, { "epoch": 2.51, "learning_rate": 1.124850390744174e-05, "loss": 0.0794, "step": 107580 }, { "epoch": 2.51, "learning_rate": 1.1247623852995705e-05, "loss": 0.0024, "step": 107585 }, { "epoch": 2.51, "learning_rate": 1.124674379854967e-05, "loss": 0.0143, "step": 107590 }, { "epoch": 2.51, "learning_rate": 1.1245863744103635e-05, "loss": 0.0146, "step": 107595 }, { "epoch": 2.51, "learning_rate": 1.12449836896576e-05, "loss": 0.0961, "step": 107600 }, { "epoch": 2.51, "learning_rate": 1.1244103635211565e-05, "loss": 0.0927, "step": 107605 }, { "epoch": 2.51, "learning_rate": 1.124322358076553e-05, "loss": 0.0729, "step": 107610 }, { "epoch": 2.51, "learning_rate": 1.1242343526319495e-05, "loss": 0.1291, "step": 107615 }, { "epoch": 2.51, "learning_rate": 1.124146347187346e-05, "loss": 0.075, "step": 107620 }, { "epoch": 2.51, "learning_rate": 1.1240583417427425e-05, "loss": 0.1834, "step": 107625 }, { "epoch": 2.51, "learning_rate": 1.1239703362981389e-05, "loss": 0.0852, "step": 107630 }, { "epoch": 2.51, "learning_rate": 1.1238823308535355e-05, "loss": 0.0547, "step": 107635 }, { "epoch": 2.51, "learning_rate": 1.1237943254089319e-05, "loss": 0.0451, "step": 107640 }, { "epoch": 2.51, "learning_rate": 1.1237063199643285e-05, "loss": 0.0926, "step": 107645 }, { "epoch": 2.51, "learning_rate": 1.1236183145197249e-05, "loss": 0.0717, "step": 107650 }, { "epoch": 2.51, "learning_rate": 1.1235303090751216e-05, "loss": 0.151, "step": 107655 }, { "epoch": 2.51, "learning_rate": 1.123442303630518e-05, "loss": 0.1256, "step": 107660 }, { "epoch": 2.51, "learning_rate": 1.1233542981859144e-05, "loss": 0.0605, "step": 107665 }, { "epoch": 2.51, "learning_rate": 1.123266292741311e-05, "loss": 0.1293, "step": 107670 }, { "epoch": 2.51, "learning_rate": 1.1231782872967074e-05, "loss": 0.2758, "step": 107675 }, { "epoch": 2.51, "learning_rate": 1.123090281852104e-05, "loss": 0.0005, "step": 107680 }, { "epoch": 2.51, "learning_rate": 1.1230022764075004e-05, "loss": 0.0422, "step": 107685 }, { "epoch": 2.51, "learning_rate": 1.122914270962897e-05, "loss": 0.0798, "step": 107690 }, { "epoch": 2.51, "learning_rate": 1.1228262655182934e-05, "loss": 0.0499, "step": 107695 }, { "epoch": 2.51, "learning_rate": 1.1227382600736899e-05, "loss": 0.0278, "step": 107700 }, { "epoch": 2.51, "learning_rate": 1.1226502546290864e-05, "loss": 0.2275, "step": 107705 }, { "epoch": 2.51, "learning_rate": 1.1225622491844829e-05, "loss": 0.0358, "step": 107710 }, { "epoch": 2.51, "learning_rate": 1.1224742437398794e-05, "loss": 0.159, "step": 107715 }, { "epoch": 2.51, "learning_rate": 1.1223862382952759e-05, "loss": 0.1508, "step": 107720 }, { "epoch": 2.51, "learning_rate": 1.1222982328506724e-05, "loss": 0.4075, "step": 107725 }, { "epoch": 2.51, "learning_rate": 1.1222102274060689e-05, "loss": 0.0004, "step": 107730 }, { "epoch": 2.51, "learning_rate": 1.1221222219614652e-05, "loss": 0.0447, "step": 107735 }, { "epoch": 2.51, "learning_rate": 1.122034216516862e-05, "loss": 0.0761, "step": 107740 }, { "epoch": 2.51, "learning_rate": 1.1219462110722583e-05, "loss": 0.0025, "step": 107745 }, { "epoch": 2.51, "learning_rate": 1.121858205627655e-05, "loss": 0.0565, "step": 107750 }, { "epoch": 2.51, "learning_rate": 1.1217702001830513e-05, "loss": 0.0992, "step": 107755 }, { "epoch": 2.51, "learning_rate": 1.121682194738448e-05, "loss": 0.2327, "step": 107760 }, { "epoch": 2.51, "learning_rate": 1.1215941892938443e-05, "loss": 0.0897, "step": 107765 }, { "epoch": 2.51, "learning_rate": 1.1215061838492408e-05, "loss": 0.1767, "step": 107770 }, { "epoch": 2.51, "learning_rate": 1.1214181784046373e-05, "loss": 0.3481, "step": 107775 }, { "epoch": 2.51, "learning_rate": 1.1213301729600338e-05, "loss": 0.0645, "step": 107780 }, { "epoch": 2.51, "learning_rate": 1.1212421675154303e-05, "loss": 0.0793, "step": 107785 }, { "epoch": 2.51, "learning_rate": 1.1211541620708268e-05, "loss": 0.0241, "step": 107790 }, { "epoch": 2.51, "learning_rate": 1.1210661566262234e-05, "loss": 0.1067, "step": 107795 }, { "epoch": 2.52, "learning_rate": 1.1209781511816198e-05, "loss": 0.0511, "step": 107800 }, { "epoch": 2.52, "learning_rate": 1.1208901457370164e-05, "loss": 0.0018, "step": 107805 }, { "epoch": 2.52, "learning_rate": 1.1208021402924128e-05, "loss": 0.0869, "step": 107810 }, { "epoch": 2.52, "learning_rate": 1.1207141348478093e-05, "loss": 0.1662, "step": 107815 }, { "epoch": 2.52, "learning_rate": 1.1206261294032058e-05, "loss": 0.1976, "step": 107820 }, { "epoch": 2.52, "learning_rate": 1.1205381239586023e-05, "loss": 0.1857, "step": 107825 }, { "epoch": 2.52, "learning_rate": 1.1204501185139988e-05, "loss": 0.0595, "step": 107830 }, { "epoch": 2.52, "learning_rate": 1.1203621130693953e-05, "loss": 0.038, "step": 107835 }, { "epoch": 2.52, "learning_rate": 1.1202741076247918e-05, "loss": 0.0295, "step": 107840 }, { "epoch": 2.52, "learning_rate": 1.1201861021801883e-05, "loss": 0.0258, "step": 107845 }, { "epoch": 2.52, "learning_rate": 1.1200980967355846e-05, "loss": 0.1267, "step": 107850 }, { "epoch": 2.52, "learning_rate": 1.1200100912909813e-05, "loss": 0.0653, "step": 107855 }, { "epoch": 2.52, "learning_rate": 1.1199220858463776e-05, "loss": 0.0534, "step": 107860 }, { "epoch": 2.52, "learning_rate": 1.1198340804017743e-05, "loss": 0.1691, "step": 107865 }, { "epoch": 2.52, "learning_rate": 1.1197460749571706e-05, "loss": 0.2226, "step": 107870 }, { "epoch": 2.52, "learning_rate": 1.1196580695125673e-05, "loss": 0.1949, "step": 107875 }, { "epoch": 2.52, "learning_rate": 1.1195700640679636e-05, "loss": 0.0588, "step": 107880 }, { "epoch": 2.52, "learning_rate": 1.1194820586233601e-05, "loss": 0.0199, "step": 107885 }, { "epoch": 2.52, "learning_rate": 1.1193940531787566e-05, "loss": 0.0944, "step": 107890 }, { "epoch": 2.52, "learning_rate": 1.1193060477341531e-05, "loss": 0.0096, "step": 107895 }, { "epoch": 2.52, "learning_rate": 1.1192180422895496e-05, "loss": 0.0376, "step": 107900 }, { "epoch": 2.52, "learning_rate": 1.1191300368449461e-05, "loss": 0.0977, "step": 107905 }, { "epoch": 2.52, "learning_rate": 1.1190420314003426e-05, "loss": 0.0879, "step": 107910 }, { "epoch": 2.52, "learning_rate": 1.1189540259557392e-05, "loss": 0.0324, "step": 107915 }, { "epoch": 2.52, "learning_rate": 1.1188660205111355e-05, "loss": 0.2053, "step": 107920 }, { "epoch": 2.52, "learning_rate": 1.1187780150665322e-05, "loss": 0.1525, "step": 107925 }, { "epoch": 2.52, "learning_rate": 1.1186900096219287e-05, "loss": 0.0793, "step": 107930 }, { "epoch": 2.52, "learning_rate": 1.1186020041773252e-05, "loss": 0.0002, "step": 107935 }, { "epoch": 2.52, "learning_rate": 1.1185139987327217e-05, "loss": 0.072, "step": 107940 }, { "epoch": 2.52, "learning_rate": 1.1184259932881182e-05, "loss": 0.0323, "step": 107945 }, { "epoch": 2.52, "learning_rate": 1.1183379878435147e-05, "loss": 0.0129, "step": 107950 }, { "epoch": 2.52, "learning_rate": 1.1182499823989112e-05, "loss": 0.0396, "step": 107955 }, { "epoch": 2.52, "learning_rate": 1.1181619769543077e-05, "loss": 0.1916, "step": 107960 }, { "epoch": 2.52, "learning_rate": 1.118073971509704e-05, "loss": 0.0792, "step": 107965 }, { "epoch": 2.52, "learning_rate": 1.1179859660651007e-05, "loss": 0.2398, "step": 107970 }, { "epoch": 2.52, "learning_rate": 1.117897960620497e-05, "loss": 0.1566, "step": 107975 }, { "epoch": 2.52, "learning_rate": 1.1178099551758937e-05, "loss": 0.1404, "step": 107980 }, { "epoch": 2.52, "learning_rate": 1.11772194973129e-05, "loss": 0.0237, "step": 107985 }, { "epoch": 2.52, "learning_rate": 1.1176339442866867e-05, "loss": 0.0367, "step": 107990 }, { "epoch": 2.52, "learning_rate": 1.117545938842083e-05, "loss": 0.1193, "step": 107995 }, { "epoch": 2.52, "learning_rate": 1.1174579333974795e-05, "loss": 0.0652, "step": 108000 }, { "epoch": 2.52, "learning_rate": 1.117369927952876e-05, "loss": 0.1511, "step": 108005 }, { "epoch": 2.52, "learning_rate": 1.1172819225082725e-05, "loss": 0.0857, "step": 108010 }, { "epoch": 2.52, "learning_rate": 1.117193917063669e-05, "loss": 0.085, "step": 108015 }, { "epoch": 2.52, "learning_rate": 1.1171059116190655e-05, "loss": 0.154, "step": 108020 }, { "epoch": 2.52, "learning_rate": 1.117017906174462e-05, "loss": 0.2165, "step": 108025 }, { "epoch": 2.52, "learning_rate": 1.1169299007298585e-05, "loss": 0.0653, "step": 108030 }, { "epoch": 2.52, "learning_rate": 1.1168418952852549e-05, "loss": 0.0036, "step": 108035 }, { "epoch": 2.52, "learning_rate": 1.1167538898406515e-05, "loss": 0.0601, "step": 108040 }, { "epoch": 2.52, "learning_rate": 1.1166658843960479e-05, "loss": 0.056, "step": 108045 }, { "epoch": 2.52, "learning_rate": 1.1165778789514445e-05, "loss": 0.0494, "step": 108050 }, { "epoch": 2.52, "learning_rate": 1.1164898735068409e-05, "loss": 0.0417, "step": 108055 }, { "epoch": 2.52, "learning_rate": 1.1164018680622375e-05, "loss": 0.1132, "step": 108060 }, { "epoch": 2.52, "learning_rate": 1.116313862617634e-05, "loss": 0.214, "step": 108065 }, { "epoch": 2.52, "learning_rate": 1.1162258571730304e-05, "loss": 0.2991, "step": 108070 }, { "epoch": 2.52, "learning_rate": 1.116137851728427e-05, "loss": 0.3132, "step": 108075 }, { "epoch": 2.52, "learning_rate": 1.1160498462838234e-05, "loss": 0.0219, "step": 108080 }, { "epoch": 2.52, "learning_rate": 1.11596184083922e-05, "loss": 0.052, "step": 108085 }, { "epoch": 2.52, "learning_rate": 1.1158738353946164e-05, "loss": 0.0869, "step": 108090 }, { "epoch": 2.52, "learning_rate": 1.115785829950013e-05, "loss": 0.1074, "step": 108095 }, { "epoch": 2.52, "learning_rate": 1.1156978245054094e-05, "loss": 0.0383, "step": 108100 }, { "epoch": 2.52, "learning_rate": 1.115609819060806e-05, "loss": 0.1544, "step": 108105 }, { "epoch": 2.52, "learning_rate": 1.1155218136162024e-05, "loss": 0.1306, "step": 108110 }, { "epoch": 2.52, "learning_rate": 1.1154338081715989e-05, "loss": 0.1782, "step": 108115 }, { "epoch": 2.52, "learning_rate": 1.1153458027269954e-05, "loss": 0.0714, "step": 108120 }, { "epoch": 2.52, "learning_rate": 1.1152577972823919e-05, "loss": 0.2641, "step": 108125 }, { "epoch": 2.52, "learning_rate": 1.1151697918377884e-05, "loss": 0.0435, "step": 108130 }, { "epoch": 2.52, "learning_rate": 1.1150817863931849e-05, "loss": 0.1077, "step": 108135 }, { "epoch": 2.52, "learning_rate": 1.1149937809485814e-05, "loss": 0.1665, "step": 108140 }, { "epoch": 2.52, "learning_rate": 1.1149057755039779e-05, "loss": 0.1111, "step": 108145 }, { "epoch": 2.52, "learning_rate": 1.1148177700593742e-05, "loss": 0.1073, "step": 108150 }, { "epoch": 2.52, "learning_rate": 1.1147297646147709e-05, "loss": 0.0479, "step": 108155 }, { "epoch": 2.52, "learning_rate": 1.1146417591701672e-05, "loss": 0.2926, "step": 108160 }, { "epoch": 2.52, "learning_rate": 1.114553753725564e-05, "loss": 0.08, "step": 108165 }, { "epoch": 2.52, "learning_rate": 1.1144657482809602e-05, "loss": 0.1878, "step": 108170 }, { "epoch": 2.52, "learning_rate": 1.114377742836357e-05, "loss": 0.2873, "step": 108175 }, { "epoch": 2.52, "learning_rate": 1.1142897373917533e-05, "loss": 0.0892, "step": 108180 }, { "epoch": 2.52, "learning_rate": 1.1142017319471498e-05, "loss": 0.0101, "step": 108185 }, { "epoch": 2.52, "learning_rate": 1.1141137265025463e-05, "loss": 0.0656, "step": 108190 }, { "epoch": 2.52, "learning_rate": 1.1140257210579428e-05, "loss": 0.0056, "step": 108195 }, { "epoch": 2.52, "learning_rate": 1.1139377156133394e-05, "loss": 0.0506, "step": 108200 }, { "epoch": 2.52, "learning_rate": 1.1138497101687358e-05, "loss": 0.0741, "step": 108205 }, { "epoch": 2.52, "learning_rate": 1.1137617047241324e-05, "loss": 0.1148, "step": 108210 }, { "epoch": 2.52, "learning_rate": 1.1136736992795288e-05, "loss": 0.1056, "step": 108215 }, { "epoch": 2.52, "learning_rate": 1.1135856938349253e-05, "loss": 0.2617, "step": 108220 }, { "epoch": 2.53, "learning_rate": 1.1134976883903218e-05, "loss": 0.1459, "step": 108225 }, { "epoch": 2.53, "learning_rate": 1.1134096829457183e-05, "loss": 0.0237, "step": 108230 }, { "epoch": 2.53, "learning_rate": 1.1133216775011148e-05, "loss": 0.0212, "step": 108235 }, { "epoch": 2.53, "learning_rate": 1.1132336720565113e-05, "loss": 0.004, "step": 108240 }, { "epoch": 2.53, "learning_rate": 1.1131456666119078e-05, "loss": 0.0078, "step": 108245 }, { "epoch": 2.53, "learning_rate": 1.1130576611673043e-05, "loss": 0.043, "step": 108250 }, { "epoch": 2.53, "learning_rate": 1.1129696557227008e-05, "loss": 0.0226, "step": 108255 }, { "epoch": 2.53, "learning_rate": 1.1128816502780973e-05, "loss": 0.1159, "step": 108260 }, { "epoch": 2.53, "learning_rate": 1.1127936448334936e-05, "loss": 0.1066, "step": 108265 }, { "epoch": 2.53, "learning_rate": 1.1127056393888903e-05, "loss": 0.0853, "step": 108270 }, { "epoch": 2.53, "learning_rate": 1.1126176339442866e-05, "loss": 0.2258, "step": 108275 }, { "epoch": 2.53, "learning_rate": 1.1125296284996833e-05, "loss": 0.0218, "step": 108280 }, { "epoch": 2.53, "learning_rate": 1.1124416230550796e-05, "loss": 0.0242, "step": 108285 }, { "epoch": 2.53, "learning_rate": 1.1123536176104763e-05, "loss": 0.0277, "step": 108290 }, { "epoch": 2.53, "learning_rate": 1.1122656121658726e-05, "loss": 0.0681, "step": 108295 }, { "epoch": 2.53, "learning_rate": 1.1121776067212691e-05, "loss": 0.038, "step": 108300 }, { "epoch": 2.53, "learning_rate": 1.1120896012766656e-05, "loss": 0.085, "step": 108305 }, { "epoch": 2.53, "learning_rate": 1.1120015958320621e-05, "loss": 0.1005, "step": 108310 }, { "epoch": 2.53, "learning_rate": 1.1119135903874586e-05, "loss": 0.1822, "step": 108315 }, { "epoch": 2.53, "learning_rate": 1.1118255849428551e-05, "loss": 0.2806, "step": 108320 }, { "epoch": 2.53, "learning_rate": 1.1117375794982516e-05, "loss": 0.2152, "step": 108325 }, { "epoch": 2.53, "learning_rate": 1.1116495740536481e-05, "loss": 0.098, "step": 108330 }, { "epoch": 2.53, "learning_rate": 1.1115615686090446e-05, "loss": 0.0075, "step": 108335 }, { "epoch": 2.53, "learning_rate": 1.1114735631644411e-05, "loss": 0.0856, "step": 108340 }, { "epoch": 2.53, "learning_rate": 1.1113855577198377e-05, "loss": 0.0233, "step": 108345 }, { "epoch": 2.53, "learning_rate": 1.1112975522752342e-05, "loss": 0.1556, "step": 108350 }, { "epoch": 2.53, "learning_rate": 1.1112095468306307e-05, "loss": 0.101, "step": 108355 }, { "epoch": 2.53, "learning_rate": 1.1111215413860272e-05, "loss": 0.063, "step": 108360 }, { "epoch": 2.53, "learning_rate": 1.1110335359414237e-05, "loss": 0.1507, "step": 108365 }, { "epoch": 2.53, "learning_rate": 1.11094553049682e-05, "loss": 0.3329, "step": 108370 }, { "epoch": 2.53, "learning_rate": 1.1108575250522167e-05, "loss": 0.1761, "step": 108375 }, { "epoch": 2.53, "learning_rate": 1.110769519607613e-05, "loss": 0.0389, "step": 108380 }, { "epoch": 2.53, "learning_rate": 1.1106815141630097e-05, "loss": 0.0342, "step": 108385 }, { "epoch": 2.53, "learning_rate": 1.110593508718406e-05, "loss": 0.0098, "step": 108390 }, { "epoch": 2.53, "learning_rate": 1.1105055032738027e-05, "loss": 0.083, "step": 108395 }, { "epoch": 2.53, "learning_rate": 1.110417497829199e-05, "loss": 0.1842, "step": 108400 }, { "epoch": 2.53, "learning_rate": 1.1103294923845957e-05, "loss": 0.1016, "step": 108405 }, { "epoch": 2.53, "learning_rate": 1.110241486939992e-05, "loss": 0.146, "step": 108410 }, { "epoch": 2.53, "learning_rate": 1.1101534814953885e-05, "loss": 0.1536, "step": 108415 }, { "epoch": 2.53, "learning_rate": 1.110065476050785e-05, "loss": 0.124, "step": 108420 }, { "epoch": 2.53, "learning_rate": 1.1099774706061815e-05, "loss": 0.3115, "step": 108425 }, { "epoch": 2.53, "learning_rate": 1.109889465161578e-05, "loss": 0.0008, "step": 108430 }, { "epoch": 2.53, "learning_rate": 1.1098014597169745e-05, "loss": 0.0411, "step": 108435 }, { "epoch": 2.53, "learning_rate": 1.109713454272371e-05, "loss": 0.0399, "step": 108440 }, { "epoch": 2.53, "learning_rate": 1.1096254488277675e-05, "loss": 0.0207, "step": 108445 }, { "epoch": 2.53, "learning_rate": 1.1095374433831639e-05, "loss": 0.1696, "step": 108450 }, { "epoch": 2.53, "learning_rate": 1.1094494379385605e-05, "loss": 0.0492, "step": 108455 }, { "epoch": 2.53, "learning_rate": 1.1093614324939569e-05, "loss": 0.1222, "step": 108460 }, { "epoch": 2.53, "learning_rate": 1.1092734270493535e-05, "loss": 0.1314, "step": 108465 }, { "epoch": 2.53, "learning_rate": 1.10918542160475e-05, "loss": 0.111, "step": 108470 }, { "epoch": 2.53, "learning_rate": 1.1090974161601465e-05, "loss": 0.2193, "step": 108475 }, { "epoch": 2.53, "learning_rate": 1.109009410715543e-05, "loss": 0.0413, "step": 108480 }, { "epoch": 2.53, "learning_rate": 1.1089214052709394e-05, "loss": 0.0278, "step": 108485 }, { "epoch": 2.53, "learning_rate": 1.108833399826336e-05, "loss": 0.0142, "step": 108490 }, { "epoch": 2.53, "learning_rate": 1.1087453943817324e-05, "loss": 0.036, "step": 108495 }, { "epoch": 2.53, "learning_rate": 1.108657388937129e-05, "loss": 0.0976, "step": 108500 }, { "epoch": 2.53, "learning_rate": 1.1085693834925254e-05, "loss": 0.0196, "step": 108505 }, { "epoch": 2.53, "learning_rate": 1.108481378047922e-05, "loss": 0.0325, "step": 108510 }, { "epoch": 2.53, "learning_rate": 1.1083933726033184e-05, "loss": 0.3136, "step": 108515 }, { "epoch": 2.53, "learning_rate": 1.1083053671587149e-05, "loss": 0.2995, "step": 108520 }, { "epoch": 2.53, "learning_rate": 1.1082173617141114e-05, "loss": 0.3974, "step": 108525 }, { "epoch": 2.53, "learning_rate": 1.1081293562695079e-05, "loss": 0.0002, "step": 108530 }, { "epoch": 2.53, "learning_rate": 1.1080413508249044e-05, "loss": 0.0703, "step": 108535 }, { "epoch": 2.53, "learning_rate": 1.1079533453803009e-05, "loss": 0.0326, "step": 108540 }, { "epoch": 2.53, "learning_rate": 1.1078653399356974e-05, "loss": 0.0516, "step": 108545 }, { "epoch": 2.53, "learning_rate": 1.1077773344910939e-05, "loss": 0.0912, "step": 108550 }, { "epoch": 2.53, "learning_rate": 1.1076893290464904e-05, "loss": 0.2029, "step": 108555 }, { "epoch": 2.53, "learning_rate": 1.1076013236018869e-05, "loss": 0.1531, "step": 108560 }, { "epoch": 2.53, "learning_rate": 1.1075133181572832e-05, "loss": 0.047, "step": 108565 }, { "epoch": 2.53, "learning_rate": 1.1074253127126799e-05, "loss": 0.2036, "step": 108570 }, { "epoch": 2.53, "learning_rate": 1.1073373072680762e-05, "loss": 0.2004, "step": 108575 }, { "epoch": 2.53, "learning_rate": 1.1072493018234729e-05, "loss": 0.0418, "step": 108580 }, { "epoch": 2.53, "learning_rate": 1.1071612963788692e-05, "loss": 0.038, "step": 108585 }, { "epoch": 2.53, "learning_rate": 1.1070732909342659e-05, "loss": 0.0113, "step": 108590 }, { "epoch": 2.53, "learning_rate": 1.1069852854896622e-05, "loss": 0.1055, "step": 108595 }, { "epoch": 2.53, "learning_rate": 1.1068972800450587e-05, "loss": 0.0894, "step": 108600 }, { "epoch": 2.53, "learning_rate": 1.1068092746004554e-05, "loss": 0.1907, "step": 108605 }, { "epoch": 2.53, "learning_rate": 1.1067212691558518e-05, "loss": 0.2441, "step": 108610 }, { "epoch": 2.53, "learning_rate": 1.1066332637112484e-05, "loss": 0.0879, "step": 108615 }, { "epoch": 2.53, "learning_rate": 1.1065452582666448e-05, "loss": 0.1011, "step": 108620 }, { "epoch": 2.53, "learning_rate": 1.1064572528220414e-05, "loss": 0.1423, "step": 108625 }, { "epoch": 2.53, "learning_rate": 1.1063692473774378e-05, "loss": 0.0532, "step": 108630 }, { "epoch": 2.53, "learning_rate": 1.1062812419328343e-05, "loss": 0.0253, "step": 108635 }, { "epoch": 2.53, "learning_rate": 1.1061932364882308e-05, "loss": 0.0169, "step": 108640 }, { "epoch": 2.53, "learning_rate": 1.1061052310436273e-05, "loss": 0.0573, "step": 108645 }, { "epoch": 2.53, "learning_rate": 1.1060172255990238e-05, "loss": 0.0073, "step": 108650 }, { "epoch": 2.54, "learning_rate": 1.1059292201544203e-05, "loss": 0.0937, "step": 108655 }, { "epoch": 2.54, "learning_rate": 1.1058412147098168e-05, "loss": 0.072, "step": 108660 }, { "epoch": 2.54, "learning_rate": 1.1057532092652133e-05, "loss": 0.1358, "step": 108665 }, { "epoch": 2.54, "learning_rate": 1.1056652038206096e-05, "loss": 0.2092, "step": 108670 }, { "epoch": 2.54, "learning_rate": 1.1055771983760063e-05, "loss": 0.4501, "step": 108675 }, { "epoch": 2.54, "learning_rate": 1.1054891929314026e-05, "loss": 0.0296, "step": 108680 }, { "epoch": 2.54, "learning_rate": 1.1054011874867993e-05, "loss": 0.0269, "step": 108685 }, { "epoch": 2.54, "learning_rate": 1.1053131820421956e-05, "loss": 0.0506, "step": 108690 }, { "epoch": 2.54, "learning_rate": 1.1052251765975923e-05, "loss": 0.1017, "step": 108695 }, { "epoch": 2.54, "learning_rate": 1.1051371711529886e-05, "loss": 0.0673, "step": 108700 }, { "epoch": 2.54, "learning_rate": 1.1050491657083853e-05, "loss": 0.1008, "step": 108705 }, { "epoch": 2.54, "learning_rate": 1.1049611602637816e-05, "loss": 0.09, "step": 108710 }, { "epoch": 2.54, "learning_rate": 1.1048731548191781e-05, "loss": 0.0677, "step": 108715 }, { "epoch": 2.54, "learning_rate": 1.1047851493745746e-05, "loss": 0.2899, "step": 108720 }, { "epoch": 2.54, "learning_rate": 1.1046971439299711e-05, "loss": 0.135, "step": 108725 }, { "epoch": 2.54, "learning_rate": 1.1046091384853676e-05, "loss": 0.0275, "step": 108730 }, { "epoch": 2.54, "learning_rate": 1.1045211330407641e-05, "loss": 0.1491, "step": 108735 }, { "epoch": 2.54, "learning_rate": 1.1044331275961606e-05, "loss": 0.0108, "step": 108740 }, { "epoch": 2.54, "learning_rate": 1.1043451221515571e-05, "loss": 0.042, "step": 108745 }, { "epoch": 2.54, "learning_rate": 1.1042571167069536e-05, "loss": 0.0923, "step": 108750 }, { "epoch": 2.54, "learning_rate": 1.1041691112623501e-05, "loss": 0.1094, "step": 108755 }, { "epoch": 2.54, "learning_rate": 1.1040811058177466e-05, "loss": 0.1807, "step": 108760 }, { "epoch": 2.54, "learning_rate": 1.1039931003731431e-05, "loss": 0.0788, "step": 108765 }, { "epoch": 2.54, "learning_rate": 1.1039050949285396e-05, "loss": 0.1218, "step": 108770 }, { "epoch": 2.54, "learning_rate": 1.1038170894839361e-05, "loss": 0.4325, "step": 108775 }, { "epoch": 2.54, "learning_rate": 1.1037290840393327e-05, "loss": 0.0265, "step": 108780 }, { "epoch": 2.54, "learning_rate": 1.103641078594729e-05, "loss": 0.0197, "step": 108785 }, { "epoch": 2.54, "learning_rate": 1.1035530731501257e-05, "loss": 0.033, "step": 108790 }, { "epoch": 2.54, "learning_rate": 1.103465067705522e-05, "loss": 0.015, "step": 108795 }, { "epoch": 2.54, "learning_rate": 1.1033770622609187e-05, "loss": 0.0081, "step": 108800 }, { "epoch": 2.54, "learning_rate": 1.103289056816315e-05, "loss": 0.0408, "step": 108805 }, { "epoch": 2.54, "learning_rate": 1.1032010513717117e-05, "loss": 0.0095, "step": 108810 }, { "epoch": 2.54, "learning_rate": 1.103113045927108e-05, "loss": 0.0925, "step": 108815 }, { "epoch": 2.54, "learning_rate": 1.1030250404825045e-05, "loss": 0.1913, "step": 108820 }, { "epoch": 2.54, "learning_rate": 1.102937035037901e-05, "loss": 0.1354, "step": 108825 }, { "epoch": 2.54, "learning_rate": 1.1028490295932975e-05, "loss": 0.0271, "step": 108830 }, { "epoch": 2.54, "learning_rate": 1.102761024148694e-05, "loss": 0.0937, "step": 108835 }, { "epoch": 2.54, "learning_rate": 1.1026730187040905e-05, "loss": 0.0441, "step": 108840 }, { "epoch": 2.54, "learning_rate": 1.102585013259487e-05, "loss": 0.0583, "step": 108845 }, { "epoch": 2.54, "learning_rate": 1.1024970078148835e-05, "loss": 0.044, "step": 108850 }, { "epoch": 2.54, "learning_rate": 1.10240900237028e-05, "loss": 0.0525, "step": 108855 }, { "epoch": 2.54, "learning_rate": 1.1023209969256765e-05, "loss": 0.1799, "step": 108860 }, { "epoch": 2.54, "learning_rate": 1.1022329914810728e-05, "loss": 0.0682, "step": 108865 }, { "epoch": 2.54, "learning_rate": 1.1021449860364695e-05, "loss": 0.2361, "step": 108870 }, { "epoch": 2.54, "learning_rate": 1.102056980591866e-05, "loss": 0.2706, "step": 108875 }, { "epoch": 2.54, "learning_rate": 1.1019689751472625e-05, "loss": 0.0329, "step": 108880 }, { "epoch": 2.54, "learning_rate": 1.101880969702659e-05, "loss": 0.0008, "step": 108885 }, { "epoch": 2.54, "learning_rate": 1.1017929642580555e-05, "loss": 0.058, "step": 108890 }, { "epoch": 2.54, "learning_rate": 1.101704958813452e-05, "loss": 0.029, "step": 108895 }, { "epoch": 2.54, "learning_rate": 1.1016169533688484e-05, "loss": 0.03, "step": 108900 }, { "epoch": 2.54, "learning_rate": 1.101528947924245e-05, "loss": 0.0317, "step": 108905 }, { "epoch": 2.54, "learning_rate": 1.1014409424796414e-05, "loss": 0.1007, "step": 108910 }, { "epoch": 2.54, "learning_rate": 1.101352937035038e-05, "loss": 0.1168, "step": 108915 }, { "epoch": 2.54, "learning_rate": 1.1012649315904344e-05, "loss": 0.0167, "step": 108920 }, { "epoch": 2.54, "learning_rate": 1.101176926145831e-05, "loss": 0.1216, "step": 108925 }, { "epoch": 2.54, "learning_rate": 1.1010889207012274e-05, "loss": 0.0024, "step": 108930 }, { "epoch": 2.54, "learning_rate": 1.1010185163455447e-05, "loss": 0.0396, "step": 108935 }, { "epoch": 2.54, "learning_rate": 1.1009305109009412e-05, "loss": 0.0243, "step": 108940 }, { "epoch": 2.54, "learning_rate": 1.1008425054563375e-05, "loss": 0.0905, "step": 108945 }, { "epoch": 2.54, "learning_rate": 1.1007545000117342e-05, "loss": 0.1194, "step": 108950 }, { "epoch": 2.54, "learning_rate": 1.1006664945671305e-05, "loss": 0.0658, "step": 108955 }, { "epoch": 2.54, "learning_rate": 1.1005784891225272e-05, "loss": 0.2983, "step": 108960 }, { "epoch": 2.54, "learning_rate": 1.1004904836779235e-05, "loss": 0.237, "step": 108965 }, { "epoch": 2.54, "learning_rate": 1.1004024782333202e-05, "loss": 0.1224, "step": 108970 }, { "epoch": 2.54, "learning_rate": 1.1003144727887165e-05, "loss": 0.275, "step": 108975 }, { "epoch": 2.54, "learning_rate": 1.100226467344113e-05, "loss": 0.0261, "step": 108980 }, { "epoch": 2.54, "learning_rate": 1.1001384618995095e-05, "loss": 0.0069, "step": 108985 }, { "epoch": 2.54, "learning_rate": 1.100050456454906e-05, "loss": 0.0102, "step": 108990 }, { "epoch": 2.54, "learning_rate": 1.0999624510103025e-05, "loss": 0.0456, "step": 108995 }, { "epoch": 2.54, "learning_rate": 1.099874445565699e-05, "loss": 0.147, "step": 109000 }, { "epoch": 2.54, "learning_rate": 1.0997864401210955e-05, "loss": 0.0271, "step": 109005 }, { "epoch": 2.54, "learning_rate": 1.099698434676492e-05, "loss": 0.1729, "step": 109010 }, { "epoch": 2.54, "learning_rate": 1.0996104292318884e-05, "loss": 0.0719, "step": 109015 }, { "epoch": 2.54, "learning_rate": 1.099522423787285e-05, "loss": 0.2336, "step": 109020 }, { "epoch": 2.54, "learning_rate": 1.0994344183426814e-05, "loss": 0.2163, "step": 109025 }, { "epoch": 2.54, "learning_rate": 1.099346412898078e-05, "loss": 0.0121, "step": 109030 }, { "epoch": 2.54, "learning_rate": 1.0992584074534744e-05, "loss": 0.0214, "step": 109035 }, { "epoch": 2.54, "learning_rate": 1.099170402008871e-05, "loss": 0.0337, "step": 109040 }, { "epoch": 2.54, "learning_rate": 1.0990823965642674e-05, "loss": 0.04, "step": 109045 }, { "epoch": 2.54, "learning_rate": 1.098994391119664e-05, "loss": 0.0452, "step": 109050 }, { "epoch": 2.54, "learning_rate": 1.0989063856750604e-05, "loss": 0.0413, "step": 109055 }, { "epoch": 2.54, "learning_rate": 1.0988183802304569e-05, "loss": 0.0347, "step": 109060 }, { "epoch": 2.54, "learning_rate": 1.0987303747858534e-05, "loss": 0.1074, "step": 109065 }, { "epoch": 2.54, "learning_rate": 1.0986423693412499e-05, "loss": 0.1635, "step": 109070 }, { "epoch": 2.54, "learning_rate": 1.0985543638966466e-05, "loss": 0.0667, "step": 109075 }, { "epoch": 2.54, "learning_rate": 1.0984663584520429e-05, "loss": 0.0477, "step": 109080 }, { "epoch": 2.55, "learning_rate": 1.0983783530074396e-05, "loss": 0.0435, "step": 109085 }, { "epoch": 2.55, "learning_rate": 1.0982903475628359e-05, "loss": 0.0397, "step": 109090 }, { "epoch": 2.55, "learning_rate": 1.0982023421182324e-05, "loss": 0.0509, "step": 109095 }, { "epoch": 2.55, "learning_rate": 1.0981143366736289e-05, "loss": 0.0016, "step": 109100 }, { "epoch": 2.55, "learning_rate": 1.0980263312290254e-05, "loss": 0.1786, "step": 109105 }, { "epoch": 2.55, "learning_rate": 1.0979383257844219e-05, "loss": 0.1461, "step": 109110 }, { "epoch": 2.55, "learning_rate": 1.0978503203398184e-05, "loss": 0.1227, "step": 109115 }, { "epoch": 2.55, "learning_rate": 1.0977623148952149e-05, "loss": 0.1802, "step": 109120 }, { "epoch": 2.55, "learning_rate": 1.0976743094506114e-05, "loss": 0.2737, "step": 109125 }, { "epoch": 2.55, "learning_rate": 1.0975863040060077e-05, "loss": 0.0931, "step": 109130 }, { "epoch": 2.55, "learning_rate": 1.0974982985614044e-05, "loss": 0.009, "step": 109135 }, { "epoch": 2.55, "learning_rate": 1.0974102931168008e-05, "loss": 0.0664, "step": 109140 }, { "epoch": 2.55, "learning_rate": 1.0973222876721974e-05, "loss": 0.0204, "step": 109145 }, { "epoch": 2.55, "learning_rate": 1.0972342822275938e-05, "loss": 0.1495, "step": 109150 }, { "epoch": 2.55, "learning_rate": 1.0971462767829904e-05, "loss": 0.0425, "step": 109155 }, { "epoch": 2.55, "learning_rate": 1.0970582713383868e-05, "loss": 0.0578, "step": 109160 }, { "epoch": 2.55, "learning_rate": 1.0969702658937833e-05, "loss": 0.2019, "step": 109165 }, { "epoch": 2.55, "learning_rate": 1.0968822604491798e-05, "loss": 0.1515, "step": 109170 }, { "epoch": 2.55, "learning_rate": 1.0967942550045763e-05, "loss": 0.3219, "step": 109175 }, { "epoch": 2.55, "learning_rate": 1.0967062495599728e-05, "loss": 0.0603, "step": 109180 }, { "epoch": 2.55, "learning_rate": 1.0966182441153693e-05, "loss": 0.0097, "step": 109185 }, { "epoch": 2.55, "learning_rate": 1.0965302386707658e-05, "loss": 0.0315, "step": 109190 }, { "epoch": 2.55, "learning_rate": 1.0964422332261623e-05, "loss": 0.073, "step": 109195 }, { "epoch": 2.55, "learning_rate": 1.0963542277815588e-05, "loss": 0.0623, "step": 109200 }, { "epoch": 2.55, "learning_rate": 1.0962662223369553e-05, "loss": 0.0711, "step": 109205 }, { "epoch": 2.55, "learning_rate": 1.0961782168923518e-05, "loss": 0.1551, "step": 109210 }, { "epoch": 2.55, "learning_rate": 1.0960902114477483e-05, "loss": 0.1335, "step": 109215 }, { "epoch": 2.55, "learning_rate": 1.0960022060031448e-05, "loss": 0.2025, "step": 109220 }, { "epoch": 2.55, "learning_rate": 1.0959142005585413e-05, "loss": 0.2296, "step": 109225 }, { "epoch": 2.55, "learning_rate": 1.0958261951139378e-05, "loss": 0.0344, "step": 109230 }, { "epoch": 2.55, "learning_rate": 1.0957381896693343e-05, "loss": 0.0922, "step": 109235 }, { "epoch": 2.55, "learning_rate": 1.0956501842247308e-05, "loss": 0.0348, "step": 109240 }, { "epoch": 2.55, "learning_rate": 1.0955621787801271e-05, "loss": 0.115, "step": 109245 }, { "epoch": 2.55, "learning_rate": 1.0954741733355238e-05, "loss": 0.0139, "step": 109250 }, { "epoch": 2.55, "learning_rate": 1.0953861678909201e-05, "loss": 0.0792, "step": 109255 }, { "epoch": 2.55, "learning_rate": 1.0952981624463168e-05, "loss": 0.1393, "step": 109260 }, { "epoch": 2.55, "learning_rate": 1.0952101570017131e-05, "loss": 0.1246, "step": 109265 }, { "epoch": 2.55, "learning_rate": 1.0951221515571098e-05, "loss": 0.1388, "step": 109270 }, { "epoch": 2.55, "learning_rate": 1.0950341461125061e-05, "loss": 0.3286, "step": 109275 }, { "epoch": 2.55, "learning_rate": 1.0949461406679026e-05, "loss": 0.0595, "step": 109280 }, { "epoch": 2.55, "learning_rate": 1.0948581352232991e-05, "loss": 0.0038, "step": 109285 }, { "epoch": 2.55, "learning_rate": 1.0947701297786956e-05, "loss": 0.083, "step": 109290 }, { "epoch": 2.55, "learning_rate": 1.0946821243340921e-05, "loss": 0.0079, "step": 109295 }, { "epoch": 2.55, "learning_rate": 1.0945941188894887e-05, "loss": 0.0821, "step": 109300 }, { "epoch": 2.55, "learning_rate": 1.0945061134448852e-05, "loss": 0.1051, "step": 109305 }, { "epoch": 2.55, "learning_rate": 1.0944181080002817e-05, "loss": 0.02, "step": 109310 }, { "epoch": 2.55, "learning_rate": 1.094330102555678e-05, "loss": 0.3121, "step": 109315 }, { "epoch": 2.55, "learning_rate": 1.0942420971110747e-05, "loss": 0.1169, "step": 109320 }, { "epoch": 2.55, "learning_rate": 1.094154091666471e-05, "loss": 0.3478, "step": 109325 }, { "epoch": 2.55, "learning_rate": 1.0940660862218677e-05, "loss": 0.0331, "step": 109330 }, { "epoch": 2.55, "learning_rate": 1.093978080777264e-05, "loss": 0.0242, "step": 109335 }, { "epoch": 2.55, "learning_rate": 1.0938900753326607e-05, "loss": 0.2149, "step": 109340 }, { "epoch": 2.55, "learning_rate": 1.0938020698880572e-05, "loss": 0.1658, "step": 109345 }, { "epoch": 2.55, "learning_rate": 1.0937140644434537e-05, "loss": 0.1025, "step": 109350 }, { "epoch": 2.55, "learning_rate": 1.0936260589988502e-05, "loss": 0.0679, "step": 109355 }, { "epoch": 2.55, "learning_rate": 1.0935380535542465e-05, "loss": 0.0958, "step": 109360 }, { "epoch": 2.55, "learning_rate": 1.0934500481096432e-05, "loss": 0.1483, "step": 109365 }, { "epoch": 2.55, "learning_rate": 1.0933620426650395e-05, "loss": 0.1093, "step": 109370 }, { "epoch": 2.55, "learning_rate": 1.0932740372204362e-05, "loss": 0.0765, "step": 109375 }, { "epoch": 2.55, "learning_rate": 1.0931860317758325e-05, "loss": 0.036, "step": 109380 }, { "epoch": 2.55, "learning_rate": 1.0930980263312292e-05, "loss": 0.0356, "step": 109385 }, { "epoch": 2.55, "learning_rate": 1.0930100208866255e-05, "loss": 0.0191, "step": 109390 }, { "epoch": 2.55, "learning_rate": 1.092922015442022e-05, "loss": 0.1179, "step": 109395 }, { "epoch": 2.55, "learning_rate": 1.0928340099974185e-05, "loss": 0.0468, "step": 109400 }, { "epoch": 2.55, "learning_rate": 1.092746004552815e-05, "loss": 0.069, "step": 109405 }, { "epoch": 2.55, "learning_rate": 1.0926579991082115e-05, "loss": 0.1051, "step": 109410 }, { "epoch": 2.55, "learning_rate": 1.092569993663608e-05, "loss": 0.0901, "step": 109415 }, { "epoch": 2.55, "learning_rate": 1.0924819882190045e-05, "loss": 0.1169, "step": 109420 }, { "epoch": 2.55, "learning_rate": 1.092393982774401e-05, "loss": 0.1685, "step": 109425 }, { "epoch": 2.55, "learning_rate": 1.0923059773297974e-05, "loss": 0.1096, "step": 109430 }, { "epoch": 2.55, "learning_rate": 1.092217971885194e-05, "loss": 0.0164, "step": 109435 }, { "epoch": 2.55, "learning_rate": 1.0921299664405904e-05, "loss": 0.0257, "step": 109440 }, { "epoch": 2.55, "learning_rate": 1.092041960995987e-05, "loss": 0.0394, "step": 109445 }, { "epoch": 2.55, "learning_rate": 1.0919539555513834e-05, "loss": 0.0317, "step": 109450 }, { "epoch": 2.55, "learning_rate": 1.09186595010678e-05, "loss": 0.1856, "step": 109455 }, { "epoch": 2.55, "learning_rate": 1.0917779446621764e-05, "loss": 0.1332, "step": 109460 }, { "epoch": 2.55, "learning_rate": 1.0916899392175729e-05, "loss": 0.0577, "step": 109465 }, { "epoch": 2.55, "learning_rate": 1.0916019337729694e-05, "loss": 0.1062, "step": 109470 }, { "epoch": 2.55, "learning_rate": 1.0915139283283659e-05, "loss": 0.1952, "step": 109475 }, { "epoch": 2.55, "learning_rate": 1.0914259228837626e-05, "loss": 0.078, "step": 109480 }, { "epoch": 2.55, "learning_rate": 1.0913379174391589e-05, "loss": 0.0221, "step": 109485 }, { "epoch": 2.55, "learning_rate": 1.0912499119945556e-05, "loss": 0.0726, "step": 109490 }, { "epoch": 2.55, "learning_rate": 1.0911619065499519e-05, "loss": 0.0769, "step": 109495 }, { "epoch": 2.55, "learning_rate": 1.0910739011053486e-05, "loss": 0.1413, "step": 109500 }, { "epoch": 2.55, "learning_rate": 1.0909858956607449e-05, "loss": 0.1093, "step": 109505 }, { "epoch": 2.55, "learning_rate": 1.0908978902161414e-05, "loss": 0.1064, "step": 109510 }, { "epoch": 2.56, "learning_rate": 1.0908098847715379e-05, "loss": 0.1828, "step": 109515 }, { "epoch": 2.56, "learning_rate": 1.0907218793269344e-05, "loss": 0.0533, "step": 109520 }, { "epoch": 2.56, "learning_rate": 1.0906338738823309e-05, "loss": 0.4226, "step": 109525 }, { "epoch": 2.56, "learning_rate": 1.0905458684377274e-05, "loss": 0.0443, "step": 109530 }, { "epoch": 2.56, "learning_rate": 1.0904578629931239e-05, "loss": 0.0596, "step": 109535 }, { "epoch": 2.56, "learning_rate": 1.0903698575485204e-05, "loss": 0.068, "step": 109540 }, { "epoch": 2.56, "learning_rate": 1.0902818521039167e-05, "loss": 0.0031, "step": 109545 }, { "epoch": 2.56, "learning_rate": 1.0901938466593134e-05, "loss": 0.0254, "step": 109550 }, { "epoch": 2.56, "learning_rate": 1.0901058412147097e-05, "loss": 0.0339, "step": 109555 }, { "epoch": 2.56, "learning_rate": 1.0900178357701064e-05, "loss": 0.071, "step": 109560 }, { "epoch": 2.56, "learning_rate": 1.0899298303255028e-05, "loss": 0.0459, "step": 109565 }, { "epoch": 2.56, "learning_rate": 1.0898418248808994e-05, "loss": 0.1871, "step": 109570 }, { "epoch": 2.56, "learning_rate": 1.0897538194362958e-05, "loss": 0.2744, "step": 109575 }, { "epoch": 2.56, "learning_rate": 1.0896658139916923e-05, "loss": 0.0627, "step": 109580 }, { "epoch": 2.56, "learning_rate": 1.0895778085470888e-05, "loss": 0.0623, "step": 109585 }, { "epoch": 2.56, "learning_rate": 1.0894898031024853e-05, "loss": 0.0133, "step": 109590 }, { "epoch": 2.56, "learning_rate": 1.0894017976578818e-05, "loss": 0.2064, "step": 109595 }, { "epoch": 2.56, "learning_rate": 1.0893137922132783e-05, "loss": 0.1008, "step": 109600 }, { "epoch": 2.56, "learning_rate": 1.0892257867686748e-05, "loss": 0.0569, "step": 109605 }, { "epoch": 2.56, "learning_rate": 1.0891377813240713e-05, "loss": 0.055, "step": 109610 }, { "epoch": 2.56, "learning_rate": 1.0890497758794678e-05, "loss": 0.0934, "step": 109615 }, { "epoch": 2.56, "learning_rate": 1.0889617704348643e-05, "loss": 0.1329, "step": 109620 }, { "epoch": 2.56, "learning_rate": 1.0888737649902608e-05, "loss": 0.2281, "step": 109625 }, { "epoch": 2.56, "learning_rate": 1.0887857595456573e-05, "loss": 0.0005, "step": 109630 }, { "epoch": 2.56, "learning_rate": 1.0886977541010538e-05, "loss": 0.0125, "step": 109635 }, { "epoch": 2.56, "learning_rate": 1.0886097486564503e-05, "loss": 0.0017, "step": 109640 }, { "epoch": 2.56, "learning_rate": 1.0885217432118468e-05, "loss": 0.0586, "step": 109645 }, { "epoch": 2.56, "learning_rate": 1.0884337377672433e-05, "loss": 0.03, "step": 109650 }, { "epoch": 2.56, "learning_rate": 1.0883457323226398e-05, "loss": 0.0958, "step": 109655 }, { "epoch": 2.56, "learning_rate": 1.0882577268780361e-05, "loss": 0.1615, "step": 109660 }, { "epoch": 2.56, "learning_rate": 1.0881697214334328e-05, "loss": 0.1148, "step": 109665 }, { "epoch": 2.56, "learning_rate": 1.0880817159888291e-05, "loss": 0.2286, "step": 109670 }, { "epoch": 2.56, "learning_rate": 1.0879937105442258e-05, "loss": 0.4791, "step": 109675 }, { "epoch": 2.56, "learning_rate": 1.0879057050996221e-05, "loss": 0.0044, "step": 109680 }, { "epoch": 2.56, "learning_rate": 1.0878176996550188e-05, "loss": 0.109, "step": 109685 }, { "epoch": 2.56, "learning_rate": 1.0877296942104151e-05, "loss": 0.0658, "step": 109690 }, { "epoch": 2.56, "learning_rate": 1.0876416887658116e-05, "loss": 0.0993, "step": 109695 }, { "epoch": 2.56, "learning_rate": 1.0875536833212081e-05, "loss": 0.1789, "step": 109700 }, { "epoch": 2.56, "learning_rate": 1.0874656778766046e-05, "loss": 0.0189, "step": 109705 }, { "epoch": 2.56, "learning_rate": 1.0873776724320011e-05, "loss": 0.1361, "step": 109710 }, { "epoch": 2.56, "learning_rate": 1.0872896669873976e-05, "loss": 0.189, "step": 109715 }, { "epoch": 2.56, "learning_rate": 1.0872016615427941e-05, "loss": 0.2287, "step": 109720 }, { "epoch": 2.56, "learning_rate": 1.0871136560981906e-05, "loss": 0.2875, "step": 109725 }, { "epoch": 2.56, "learning_rate": 1.087025650653587e-05, "loss": 0.016, "step": 109730 }, { "epoch": 2.56, "learning_rate": 1.0869376452089837e-05, "loss": 0.0019, "step": 109735 }, { "epoch": 2.56, "learning_rate": 1.08684963976438e-05, "loss": 0.0199, "step": 109740 }, { "epoch": 2.56, "learning_rate": 1.0867616343197767e-05, "loss": 0.1153, "step": 109745 }, { "epoch": 2.56, "learning_rate": 1.0866736288751732e-05, "loss": 0.0473, "step": 109750 }, { "epoch": 2.56, "learning_rate": 1.0865856234305697e-05, "loss": 0.0309, "step": 109755 }, { "epoch": 2.56, "learning_rate": 1.0864976179859662e-05, "loss": 0.1111, "step": 109760 }, { "epoch": 2.56, "learning_rate": 1.0864096125413625e-05, "loss": 0.1336, "step": 109765 }, { "epoch": 2.56, "learning_rate": 1.0863216070967592e-05, "loss": 0.3764, "step": 109770 }, { "epoch": 2.56, "learning_rate": 1.0862336016521555e-05, "loss": 0.1219, "step": 109775 }, { "epoch": 2.56, "learning_rate": 1.0861455962075522e-05, "loss": 0.0805, "step": 109780 }, { "epoch": 2.56, "learning_rate": 1.0860575907629485e-05, "loss": 0.0362, "step": 109785 }, { "epoch": 2.56, "learning_rate": 1.0859695853183452e-05, "loss": 0.0478, "step": 109790 }, { "epoch": 2.56, "learning_rate": 1.0858815798737415e-05, "loss": 0.0169, "step": 109795 }, { "epoch": 2.56, "learning_rate": 1.0857935744291382e-05, "loss": 0.1354, "step": 109800 }, { "epoch": 2.56, "learning_rate": 1.0857055689845345e-05, "loss": 0.0485, "step": 109805 }, { "epoch": 2.56, "learning_rate": 1.085617563539931e-05, "loss": 0.0434, "step": 109810 }, { "epoch": 2.56, "learning_rate": 1.0855295580953275e-05, "loss": 0.2699, "step": 109815 }, { "epoch": 2.56, "learning_rate": 1.085441552650724e-05, "loss": 0.2575, "step": 109820 }, { "epoch": 2.56, "learning_rate": 1.0853535472061205e-05, "loss": 0.3598, "step": 109825 }, { "epoch": 2.56, "learning_rate": 1.085265541761517e-05, "loss": 0.0956, "step": 109830 }, { "epoch": 2.56, "learning_rate": 1.0851775363169135e-05, "loss": 0.073, "step": 109835 }, { "epoch": 2.56, "learning_rate": 1.08508953087231e-05, "loss": 0.0335, "step": 109840 }, { "epoch": 2.56, "learning_rate": 1.0850015254277064e-05, "loss": 0.0346, "step": 109845 }, { "epoch": 2.56, "learning_rate": 1.084913519983103e-05, "loss": 0.0173, "step": 109850 }, { "epoch": 2.56, "learning_rate": 1.0848255145384994e-05, "loss": 0.1552, "step": 109855 }, { "epoch": 2.56, "learning_rate": 1.084737509093896e-05, "loss": 0.0474, "step": 109860 }, { "epoch": 2.56, "learning_rate": 1.0846495036492924e-05, "loss": 0.2591, "step": 109865 }, { "epoch": 2.56, "learning_rate": 1.084561498204689e-05, "loss": 0.1635, "step": 109870 }, { "epoch": 2.56, "learning_rate": 1.0844734927600854e-05, "loss": 0.1267, "step": 109875 }, { "epoch": 2.56, "learning_rate": 1.0843854873154819e-05, "loss": 0.0314, "step": 109880 }, { "epoch": 2.56, "learning_rate": 1.0842974818708785e-05, "loss": 0.1222, "step": 109885 }, { "epoch": 2.56, "learning_rate": 1.0842094764262749e-05, "loss": 0.0075, "step": 109890 }, { "epoch": 2.56, "learning_rate": 1.0841214709816715e-05, "loss": 0.003, "step": 109895 }, { "epoch": 2.56, "learning_rate": 1.0840334655370679e-05, "loss": 0.1208, "step": 109900 }, { "epoch": 2.56, "learning_rate": 1.0839454600924646e-05, "loss": 0.112, "step": 109905 }, { "epoch": 2.56, "learning_rate": 1.0838574546478609e-05, "loss": 0.0103, "step": 109910 }, { "epoch": 2.56, "learning_rate": 1.0837694492032574e-05, "loss": 0.2486, "step": 109915 }, { "epoch": 2.56, "learning_rate": 1.0836814437586539e-05, "loss": 0.2134, "step": 109920 }, { "epoch": 2.56, "learning_rate": 1.0835934383140504e-05, "loss": 0.2986, "step": 109925 }, { "epoch": 2.56, "learning_rate": 1.0835054328694469e-05, "loss": 0.1173, "step": 109930 }, { "epoch": 2.56, "learning_rate": 1.0834174274248434e-05, "loss": 0.0142, "step": 109935 }, { "epoch": 2.57, "learning_rate": 1.0833294219802399e-05, "loss": 0.0049, "step": 109940 }, { "epoch": 2.57, "learning_rate": 1.0832414165356364e-05, "loss": 0.0991, "step": 109945 }, { "epoch": 2.57, "learning_rate": 1.0831534110910329e-05, "loss": 0.1544, "step": 109950 }, { "epoch": 2.57, "learning_rate": 1.0830654056464294e-05, "loss": 0.1455, "step": 109955 }, { "epoch": 2.57, "learning_rate": 1.0829774002018257e-05, "loss": 0.4101, "step": 109960 }, { "epoch": 2.57, "learning_rate": 1.0828893947572224e-05, "loss": 0.1603, "step": 109965 }, { "epoch": 2.57, "learning_rate": 1.0828013893126187e-05, "loss": 0.2497, "step": 109970 }, { "epoch": 2.57, "learning_rate": 1.0827133838680154e-05, "loss": 0.1525, "step": 109975 }, { "epoch": 2.57, "learning_rate": 1.0826253784234117e-05, "loss": 0.0729, "step": 109980 }, { "epoch": 2.57, "learning_rate": 1.0825373729788084e-05, "loss": 0.0167, "step": 109985 }, { "epoch": 2.57, "learning_rate": 1.0824493675342047e-05, "loss": 0.0747, "step": 109990 }, { "epoch": 2.57, "learning_rate": 1.0823613620896012e-05, "loss": 0.0595, "step": 109995 }, { "epoch": 2.57, "learning_rate": 1.0822733566449978e-05, "loss": 0.0237, "step": 110000 }, { "epoch": 2.57, "learning_rate": 1.0821853512003943e-05, "loss": 0.1472, "step": 110005 }, { "epoch": 2.57, "learning_rate": 1.0820973457557908e-05, "loss": 0.0925, "step": 110010 }, { "epoch": 2.57, "learning_rate": 1.0820093403111873e-05, "loss": 0.0714, "step": 110015 }, { "epoch": 2.57, "learning_rate": 1.0819213348665838e-05, "loss": 0.1341, "step": 110020 }, { "epoch": 2.57, "learning_rate": 1.0818333294219803e-05, "loss": 0.1389, "step": 110025 }, { "epoch": 2.57, "learning_rate": 1.0817453239773768e-05, "loss": 0.0138, "step": 110030 }, { "epoch": 2.57, "learning_rate": 1.0816573185327733e-05, "loss": 0.0089, "step": 110035 }, { "epoch": 2.57, "learning_rate": 1.0815693130881698e-05, "loss": 0.0061, "step": 110040 }, { "epoch": 2.57, "learning_rate": 1.0814813076435663e-05, "loss": 0.0392, "step": 110045 }, { "epoch": 2.57, "learning_rate": 1.0813933021989628e-05, "loss": 0.2461, "step": 110050 }, { "epoch": 2.57, "learning_rate": 1.0813052967543593e-05, "loss": 0.0454, "step": 110055 }, { "epoch": 2.57, "learning_rate": 1.0812172913097558e-05, "loss": 0.092, "step": 110060 }, { "epoch": 2.57, "learning_rate": 1.0811292858651521e-05, "loss": 0.0951, "step": 110065 }, { "epoch": 2.57, "learning_rate": 1.0810412804205488e-05, "loss": 0.355, "step": 110070 }, { "epoch": 2.57, "learning_rate": 1.0809532749759451e-05, "loss": 0.2092, "step": 110075 }, { "epoch": 2.57, "learning_rate": 1.0808652695313418e-05, "loss": 0.0531, "step": 110080 }, { "epoch": 2.57, "learning_rate": 1.0807772640867381e-05, "loss": 0.1146, "step": 110085 }, { "epoch": 2.57, "learning_rate": 1.0806892586421348e-05, "loss": 0.0801, "step": 110090 }, { "epoch": 2.57, "learning_rate": 1.0806012531975311e-05, "loss": 0.0497, "step": 110095 }, { "epoch": 2.57, "learning_rate": 1.0805132477529278e-05, "loss": 0.0369, "step": 110100 }, { "epoch": 2.57, "learning_rate": 1.0804252423083241e-05, "loss": 0.0661, "step": 110105 }, { "epoch": 2.57, "learning_rate": 1.0803372368637206e-05, "loss": 0.0829, "step": 110110 }, { "epoch": 2.57, "learning_rate": 1.0802492314191171e-05, "loss": 0.2027, "step": 110115 }, { "epoch": 2.57, "learning_rate": 1.0801612259745136e-05, "loss": 0.1938, "step": 110120 }, { "epoch": 2.57, "learning_rate": 1.0800732205299101e-05, "loss": 0.4927, "step": 110125 }, { "epoch": 2.57, "learning_rate": 1.0799852150853066e-05, "loss": 0.0421, "step": 110130 }, { "epoch": 2.57, "learning_rate": 1.0798972096407031e-05, "loss": 0.0216, "step": 110135 }, { "epoch": 2.57, "learning_rate": 1.0798092041960996e-05, "loss": 0.0457, "step": 110140 }, { "epoch": 2.57, "learning_rate": 1.079721198751496e-05, "loss": 0.0026, "step": 110145 }, { "epoch": 2.57, "learning_rate": 1.0796331933068926e-05, "loss": 0.0944, "step": 110150 }, { "epoch": 2.57, "learning_rate": 1.0795451878622891e-05, "loss": 0.0416, "step": 110155 }, { "epoch": 2.57, "learning_rate": 1.0794571824176856e-05, "loss": 0.061, "step": 110160 }, { "epoch": 2.57, "learning_rate": 1.0793691769730822e-05, "loss": 0.0474, "step": 110165 }, { "epoch": 2.57, "learning_rate": 1.0792811715284787e-05, "loss": 0.0754, "step": 110170 }, { "epoch": 2.57, "learning_rate": 1.0791931660838752e-05, "loss": 0.2149, "step": 110175 }, { "epoch": 2.57, "learning_rate": 1.0791051606392715e-05, "loss": 0.0237, "step": 110180 }, { "epoch": 2.57, "learning_rate": 1.0790171551946682e-05, "loss": 0.0116, "step": 110185 }, { "epoch": 2.57, "learning_rate": 1.0789291497500645e-05, "loss": 0.0296, "step": 110190 }, { "epoch": 2.57, "learning_rate": 1.0788411443054612e-05, "loss": 0.0398, "step": 110195 }, { "epoch": 2.57, "learning_rate": 1.0787531388608575e-05, "loss": 0.0312, "step": 110200 }, { "epoch": 2.57, "learning_rate": 1.0786651334162542e-05, "loss": 0.045, "step": 110205 }, { "epoch": 2.57, "learning_rate": 1.0785771279716505e-05, "loss": 0.108, "step": 110210 }, { "epoch": 2.57, "learning_rate": 1.078489122527047e-05, "loss": 0.2181, "step": 110215 }, { "epoch": 2.57, "learning_rate": 1.0784011170824435e-05, "loss": 0.1477, "step": 110220 }, { "epoch": 2.57, "learning_rate": 1.07831311163784e-05, "loss": 0.1041, "step": 110225 }, { "epoch": 2.57, "learning_rate": 1.0782251061932365e-05, "loss": 0.012, "step": 110230 }, { "epoch": 2.57, "learning_rate": 1.078137100748633e-05, "loss": 0.0019, "step": 110235 }, { "epoch": 2.57, "learning_rate": 1.0780490953040295e-05, "loss": 0.075, "step": 110240 }, { "epoch": 2.57, "learning_rate": 1.077961089859426e-05, "loss": 0.0007, "step": 110245 }, { "epoch": 2.57, "learning_rate": 1.0778730844148225e-05, "loss": 0.0768, "step": 110250 }, { "epoch": 2.57, "learning_rate": 1.077785078970219e-05, "loss": 0.1336, "step": 110255 }, { "epoch": 2.57, "learning_rate": 1.0776970735256154e-05, "loss": 0.1306, "step": 110260 }, { "epoch": 2.57, "learning_rate": 1.077609068081012e-05, "loss": 0.0478, "step": 110265 }, { "epoch": 2.57, "learning_rate": 1.0775210626364084e-05, "loss": 0.1976, "step": 110270 }, { "epoch": 2.57, "learning_rate": 1.077433057191805e-05, "loss": 0.1901, "step": 110275 }, { "epoch": 2.57, "learning_rate": 1.0773450517472014e-05, "loss": 0.021, "step": 110280 }, { "epoch": 2.57, "learning_rate": 1.077257046302598e-05, "loss": 0.0304, "step": 110285 }, { "epoch": 2.57, "learning_rate": 1.0771690408579944e-05, "loss": 0.1211, "step": 110290 }, { "epoch": 2.57, "learning_rate": 1.0770810354133909e-05, "loss": 0.0993, "step": 110295 }, { "epoch": 2.57, "learning_rate": 1.0769930299687875e-05, "loss": 0.0986, "step": 110300 }, { "epoch": 2.57, "learning_rate": 1.0769050245241839e-05, "loss": 0.1229, "step": 110305 }, { "epoch": 2.57, "learning_rate": 1.0768170190795805e-05, "loss": 0.0821, "step": 110310 }, { "epoch": 2.57, "learning_rate": 1.0767290136349769e-05, "loss": 0.1358, "step": 110315 }, { "epoch": 2.57, "learning_rate": 1.0766410081903735e-05, "loss": 0.243, "step": 110320 }, { "epoch": 2.57, "learning_rate": 1.0765530027457699e-05, "loss": 0.2298, "step": 110325 }, { "epoch": 2.57, "learning_rate": 1.0764649973011664e-05, "loss": 0.0586, "step": 110330 }, { "epoch": 2.57, "learning_rate": 1.0763769918565629e-05, "loss": 0.0135, "step": 110335 }, { "epoch": 2.57, "learning_rate": 1.0762889864119594e-05, "loss": 0.0631, "step": 110340 }, { "epoch": 2.57, "learning_rate": 1.0762009809673559e-05, "loss": 0.0262, "step": 110345 }, { "epoch": 2.57, "learning_rate": 1.0761129755227524e-05, "loss": 0.1038, "step": 110350 }, { "epoch": 2.57, "learning_rate": 1.0760249700781489e-05, "loss": 0.0421, "step": 110355 }, { "epoch": 2.57, "learning_rate": 1.0759369646335454e-05, "loss": 0.1934, "step": 110360 }, { "epoch": 2.57, "learning_rate": 1.0758489591889417e-05, "loss": 0.1586, "step": 110365 }, { "epoch": 2.58, "learning_rate": 1.0757609537443384e-05, "loss": 0.1319, "step": 110370 }, { "epoch": 2.58, "learning_rate": 1.0756729482997347e-05, "loss": 0.2596, "step": 110375 }, { "epoch": 2.58, "learning_rate": 1.0755849428551314e-05, "loss": 0.0692, "step": 110380 }, { "epoch": 2.58, "learning_rate": 1.0754969374105277e-05, "loss": 0.0227, "step": 110385 }, { "epoch": 2.58, "learning_rate": 1.0754089319659244e-05, "loss": 0.0114, "step": 110390 }, { "epoch": 2.58, "learning_rate": 1.0753209265213207e-05, "loss": 0.0099, "step": 110395 }, { "epoch": 2.58, "learning_rate": 1.0752329210767174e-05, "loss": 0.0159, "step": 110400 }, { "epoch": 2.58, "learning_rate": 1.0751449156321137e-05, "loss": 0.0262, "step": 110405 }, { "epoch": 2.58, "learning_rate": 1.0750569101875102e-05, "loss": 0.0688, "step": 110410 }, { "epoch": 2.58, "learning_rate": 1.0749689047429067e-05, "loss": 0.188, "step": 110415 }, { "epoch": 2.58, "learning_rate": 1.0748808992983032e-05, "loss": 0.0792, "step": 110420 }, { "epoch": 2.58, "learning_rate": 1.0747928938536997e-05, "loss": 0.2543, "step": 110425 }, { "epoch": 2.58, "learning_rate": 1.0747048884090963e-05, "loss": 0.0506, "step": 110430 }, { "epoch": 2.58, "learning_rate": 1.074616882964493e-05, "loss": 0.0933, "step": 110435 }, { "epoch": 2.58, "learning_rate": 1.0745288775198893e-05, "loss": 0.1063, "step": 110440 }, { "epoch": 2.58, "learning_rate": 1.0744408720752858e-05, "loss": 0.1813, "step": 110445 }, { "epoch": 2.58, "learning_rate": 1.0743528666306823e-05, "loss": 0.0766, "step": 110450 }, { "epoch": 2.58, "learning_rate": 1.0742648611860788e-05, "loss": 0.0841, "step": 110455 }, { "epoch": 2.58, "learning_rate": 1.0741768557414753e-05, "loss": 0.0447, "step": 110460 }, { "epoch": 2.58, "learning_rate": 1.0740888502968718e-05, "loss": 0.2007, "step": 110465 }, { "epoch": 2.58, "learning_rate": 1.0740008448522683e-05, "loss": 0.0922, "step": 110470 }, { "epoch": 2.58, "learning_rate": 1.0739304404965854e-05, "loss": 0.3966, "step": 110475 }, { "epoch": 2.58, "learning_rate": 1.0738424350519819e-05, "loss": 0.0006, "step": 110480 }, { "epoch": 2.58, "learning_rate": 1.0737544296073784e-05, "loss": 0.0172, "step": 110485 }, { "epoch": 2.58, "learning_rate": 1.0736664241627749e-05, "loss": 0.0506, "step": 110490 }, { "epoch": 2.58, "learning_rate": 1.0735784187181714e-05, "loss": 0.037, "step": 110495 }, { "epoch": 2.58, "learning_rate": 1.0734904132735679e-05, "loss": 0.045, "step": 110500 }, { "epoch": 2.58, "learning_rate": 1.0734024078289644e-05, "loss": 0.1054, "step": 110505 }, { "epoch": 2.58, "learning_rate": 1.0733144023843609e-05, "loss": 0.0345, "step": 110510 }, { "epoch": 2.58, "learning_rate": 1.0732263969397574e-05, "loss": 0.2223, "step": 110515 }, { "epoch": 2.58, "learning_rate": 1.073138391495154e-05, "loss": 0.1256, "step": 110520 }, { "epoch": 2.58, "learning_rate": 1.0730503860505503e-05, "loss": 0.1002, "step": 110525 }, { "epoch": 2.58, "learning_rate": 1.072962380605947e-05, "loss": 0.0559, "step": 110530 }, { "epoch": 2.58, "learning_rate": 1.0728743751613433e-05, "loss": 0.0519, "step": 110535 }, { "epoch": 2.58, "learning_rate": 1.07278636971674e-05, "loss": 0.007, "step": 110540 }, { "epoch": 2.58, "learning_rate": 1.0726983642721363e-05, "loss": 0.0363, "step": 110545 }, { "epoch": 2.58, "learning_rate": 1.072610358827533e-05, "loss": 0.0534, "step": 110550 }, { "epoch": 2.58, "learning_rate": 1.0725223533829293e-05, "loss": 0.129, "step": 110555 }, { "epoch": 2.58, "learning_rate": 1.0724343479383258e-05, "loss": 0.1625, "step": 110560 }, { "epoch": 2.58, "learning_rate": 1.0723463424937223e-05, "loss": 0.1288, "step": 110565 }, { "epoch": 2.58, "learning_rate": 1.0722583370491188e-05, "loss": 0.2514, "step": 110570 }, { "epoch": 2.58, "learning_rate": 1.0721703316045153e-05, "loss": 0.4816, "step": 110575 }, { "epoch": 2.58, "learning_rate": 1.0720823261599118e-05, "loss": 0.0111, "step": 110580 }, { "epoch": 2.58, "learning_rate": 1.0719943207153083e-05, "loss": 0.0154, "step": 110585 }, { "epoch": 2.58, "learning_rate": 1.0719063152707048e-05, "loss": 0.0918, "step": 110590 }, { "epoch": 2.58, "learning_rate": 1.0718183098261013e-05, "loss": 0.1241, "step": 110595 }, { "epoch": 2.58, "learning_rate": 1.0717303043814978e-05, "loss": 0.0631, "step": 110600 }, { "epoch": 2.58, "learning_rate": 1.0716422989368941e-05, "loss": 0.0475, "step": 110605 }, { "epoch": 2.58, "learning_rate": 1.0715542934922908e-05, "loss": 0.1135, "step": 110610 }, { "epoch": 2.58, "learning_rate": 1.0714662880476871e-05, "loss": 0.1167, "step": 110615 }, { "epoch": 2.58, "learning_rate": 1.0713782826030838e-05, "loss": 0.2517, "step": 110620 }, { "epoch": 2.58, "learning_rate": 1.0712902771584803e-05, "loss": 0.2098, "step": 110625 }, { "epoch": 2.58, "learning_rate": 1.0712022717138768e-05, "loss": 0.0182, "step": 110630 }, { "epoch": 2.58, "learning_rate": 1.0711142662692733e-05, "loss": 0.0486, "step": 110635 }, { "epoch": 2.58, "learning_rate": 1.0710262608246696e-05, "loss": 0.0927, "step": 110640 }, { "epoch": 2.58, "learning_rate": 1.0709382553800663e-05, "loss": 0.0755, "step": 110645 }, { "epoch": 2.58, "learning_rate": 1.0708502499354626e-05, "loss": 0.0451, "step": 110650 }, { "epoch": 2.58, "learning_rate": 1.0707622444908593e-05, "loss": 0.0651, "step": 110655 }, { "epoch": 2.58, "learning_rate": 1.0706742390462556e-05, "loss": 0.1283, "step": 110660 }, { "epoch": 2.58, "learning_rate": 1.0705862336016523e-05, "loss": 0.0396, "step": 110665 }, { "epoch": 2.58, "learning_rate": 1.0704982281570486e-05, "loss": 0.3042, "step": 110670 }, { "epoch": 2.58, "learning_rate": 1.0704102227124451e-05, "loss": 0.427, "step": 110675 }, { "epoch": 2.58, "learning_rate": 1.0703222172678416e-05, "loss": 0.0187, "step": 110680 }, { "epoch": 2.58, "learning_rate": 1.0702342118232381e-05, "loss": 0.0271, "step": 110685 }, { "epoch": 2.58, "learning_rate": 1.0701462063786347e-05, "loss": 0.0642, "step": 110690 }, { "epoch": 2.58, "learning_rate": 1.0700582009340312e-05, "loss": 0.0697, "step": 110695 }, { "epoch": 2.58, "learning_rate": 1.0699701954894277e-05, "loss": 0.0613, "step": 110700 }, { "epoch": 2.58, "learning_rate": 1.0698821900448242e-05, "loss": 0.0796, "step": 110705 }, { "epoch": 2.58, "learning_rate": 1.0697941846002205e-05, "loss": 0.1004, "step": 110710 }, { "epoch": 2.58, "learning_rate": 1.0697061791556172e-05, "loss": 0.1705, "step": 110715 }, { "epoch": 2.58, "learning_rate": 1.0696181737110135e-05, "loss": 0.1428, "step": 110720 }, { "epoch": 2.58, "learning_rate": 1.0695301682664102e-05, "loss": 0.2057, "step": 110725 }, { "epoch": 2.58, "learning_rate": 1.0694421628218065e-05, "loss": 0.0356, "step": 110730 }, { "epoch": 2.58, "learning_rate": 1.0693541573772032e-05, "loss": 0.0098, "step": 110735 }, { "epoch": 2.58, "learning_rate": 1.0692661519325995e-05, "loss": 0.0446, "step": 110740 }, { "epoch": 2.58, "learning_rate": 1.0691781464879962e-05, "loss": 0.0314, "step": 110745 }, { "epoch": 2.58, "learning_rate": 1.0690901410433925e-05, "loss": 0.0515, "step": 110750 }, { "epoch": 2.58, "learning_rate": 1.069002135598789e-05, "loss": 0.0337, "step": 110755 }, { "epoch": 2.58, "learning_rate": 1.0689141301541857e-05, "loss": 0.1651, "step": 110760 }, { "epoch": 2.58, "learning_rate": 1.068826124709582e-05, "loss": 0.0796, "step": 110765 }, { "epoch": 2.58, "learning_rate": 1.0687381192649787e-05, "loss": 0.196, "step": 110770 }, { "epoch": 2.58, "learning_rate": 1.068650113820375e-05, "loss": 0.1987, "step": 110775 }, { "epoch": 2.58, "learning_rate": 1.0685621083757717e-05, "loss": 0.0576, "step": 110780 }, { "epoch": 2.58, "learning_rate": 1.068474102931168e-05, "loss": 0.057, "step": 110785 }, { "epoch": 2.58, "learning_rate": 1.0683860974865645e-05, "loss": 0.1914, "step": 110790 }, { "epoch": 2.58, "learning_rate": 1.068298092041961e-05, "loss": 0.0547, "step": 110795 }, { "epoch": 2.59, "learning_rate": 1.0682100865973575e-05, "loss": 0.0258, "step": 110800 }, { "epoch": 2.59, "learning_rate": 1.068122081152754e-05, "loss": 0.069, "step": 110805 }, { "epoch": 2.59, "learning_rate": 1.0680340757081505e-05, "loss": 0.1158, "step": 110810 }, { "epoch": 2.59, "learning_rate": 1.067946070263547e-05, "loss": 0.1349, "step": 110815 }, { "epoch": 2.59, "learning_rate": 1.0678580648189435e-05, "loss": 0.1331, "step": 110820 }, { "epoch": 2.59, "learning_rate": 1.0677700593743399e-05, "loss": 0.3571, "step": 110825 }, { "epoch": 2.59, "learning_rate": 1.0676820539297365e-05, "loss": 0.0611, "step": 110830 }, { "epoch": 2.59, "learning_rate": 1.0675940484851329e-05, "loss": 0.0021, "step": 110835 }, { "epoch": 2.59, "learning_rate": 1.0675060430405295e-05, "loss": 0.0425, "step": 110840 }, { "epoch": 2.59, "learning_rate": 1.0674180375959259e-05, "loss": 0.0367, "step": 110845 }, { "epoch": 2.59, "learning_rate": 1.0673300321513225e-05, "loss": 0.0756, "step": 110850 }, { "epoch": 2.59, "learning_rate": 1.0672420267067189e-05, "loss": 0.1536, "step": 110855 }, { "epoch": 2.59, "learning_rate": 1.0671540212621154e-05, "loss": 0.2347, "step": 110860 }, { "epoch": 2.59, "learning_rate": 1.0670660158175119e-05, "loss": 0.0905, "step": 110865 }, { "epoch": 2.59, "learning_rate": 1.0669780103729084e-05, "loss": 0.207, "step": 110870 }, { "epoch": 2.59, "learning_rate": 1.0668900049283049e-05, "loss": 0.17, "step": 110875 }, { "epoch": 2.59, "learning_rate": 1.0668019994837014e-05, "loss": 0.0658, "step": 110880 }, { "epoch": 2.59, "learning_rate": 1.0667139940390979e-05, "loss": 0.0628, "step": 110885 }, { "epoch": 2.59, "learning_rate": 1.0666259885944944e-05, "loss": 0.0371, "step": 110890 }, { "epoch": 2.59, "learning_rate": 1.066537983149891e-05, "loss": 0.0204, "step": 110895 }, { "epoch": 2.59, "learning_rate": 1.0664499777052874e-05, "loss": 0.0999, "step": 110900 }, { "epoch": 2.59, "learning_rate": 1.0663619722606839e-05, "loss": 0.0766, "step": 110905 }, { "epoch": 2.59, "learning_rate": 1.0662739668160804e-05, "loss": 0.1005, "step": 110910 }, { "epoch": 2.59, "learning_rate": 1.0661859613714769e-05, "loss": 0.1019, "step": 110915 }, { "epoch": 2.59, "learning_rate": 1.0660979559268734e-05, "loss": 0.1607, "step": 110920 }, { "epoch": 2.59, "learning_rate": 1.0660099504822699e-05, "loss": 0.1463, "step": 110925 }, { "epoch": 2.59, "learning_rate": 1.0659219450376664e-05, "loss": 0.1187, "step": 110930 }, { "epoch": 2.59, "learning_rate": 1.0658339395930629e-05, "loss": 0.004, "step": 110935 }, { "epoch": 2.59, "learning_rate": 1.0657459341484592e-05, "loss": 0.0466, "step": 110940 }, { "epoch": 2.59, "learning_rate": 1.065657928703856e-05, "loss": 0.0576, "step": 110945 }, { "epoch": 2.59, "learning_rate": 1.0655699232592522e-05, "loss": 0.0599, "step": 110950 }, { "epoch": 2.59, "learning_rate": 1.065481917814649e-05, "loss": 0.0362, "step": 110955 }, { "epoch": 2.59, "learning_rate": 1.0653939123700453e-05, "loss": 0.1523, "step": 110960 }, { "epoch": 2.59, "learning_rate": 1.065305906925442e-05, "loss": 0.1687, "step": 110965 }, { "epoch": 2.59, "learning_rate": 1.0652179014808383e-05, "loss": 0.2414, "step": 110970 }, { "epoch": 2.59, "learning_rate": 1.0651298960362348e-05, "loss": 0.2114, "step": 110975 }, { "epoch": 2.59, "learning_rate": 1.0650418905916313e-05, "loss": 0.0368, "step": 110980 }, { "epoch": 2.59, "learning_rate": 1.0649538851470278e-05, "loss": 0.0662, "step": 110985 }, { "epoch": 2.59, "learning_rate": 1.0648658797024243e-05, "loss": 0.0246, "step": 110990 }, { "epoch": 2.59, "learning_rate": 1.0647778742578208e-05, "loss": 0.0212, "step": 110995 }, { "epoch": 2.59, "learning_rate": 1.0646898688132173e-05, "loss": 0.0711, "step": 111000 }, { "epoch": 2.59, "learning_rate": 1.0646018633686138e-05, "loss": 0.1765, "step": 111005 }, { "epoch": 2.59, "learning_rate": 1.0645138579240101e-05, "loss": 0.2044, "step": 111010 }, { "epoch": 2.59, "learning_rate": 1.0644258524794068e-05, "loss": 0.1714, "step": 111015 }, { "epoch": 2.59, "learning_rate": 1.0643378470348031e-05, "loss": 0.4245, "step": 111020 }, { "epoch": 2.59, "learning_rate": 1.0642498415901998e-05, "loss": 0.1726, "step": 111025 }, { "epoch": 2.59, "learning_rate": 1.0641618361455963e-05, "loss": 0.0179, "step": 111030 }, { "epoch": 2.59, "learning_rate": 1.0640738307009928e-05, "loss": 0.0023, "step": 111035 }, { "epoch": 2.59, "learning_rate": 1.0639858252563893e-05, "loss": 0.0293, "step": 111040 }, { "epoch": 2.59, "learning_rate": 1.0638978198117858e-05, "loss": 0.0329, "step": 111045 }, { "epoch": 2.59, "learning_rate": 1.0638098143671823e-05, "loss": 0.0013, "step": 111050 }, { "epoch": 2.59, "learning_rate": 1.0637218089225786e-05, "loss": 0.0203, "step": 111055 }, { "epoch": 2.59, "learning_rate": 1.0636338034779753e-05, "loss": 0.0713, "step": 111060 }, { "epoch": 2.59, "learning_rate": 1.0635457980333716e-05, "loss": 0.0665, "step": 111065 }, { "epoch": 2.59, "learning_rate": 1.0634577925887683e-05, "loss": 0.1417, "step": 111070 }, { "epoch": 2.59, "learning_rate": 1.0633697871441646e-05, "loss": 0.1503, "step": 111075 }, { "epoch": 2.59, "learning_rate": 1.0632817816995613e-05, "loss": 0.0384, "step": 111080 }, { "epoch": 2.59, "learning_rate": 1.0631937762549576e-05, "loss": 0.2463, "step": 111085 }, { "epoch": 2.59, "learning_rate": 1.0631057708103541e-05, "loss": 0.0588, "step": 111090 }, { "epoch": 2.59, "learning_rate": 1.0630177653657506e-05, "loss": 0.0551, "step": 111095 }, { "epoch": 2.59, "learning_rate": 1.0629297599211471e-05, "loss": 0.0305, "step": 111100 }, { "epoch": 2.59, "learning_rate": 1.0628417544765436e-05, "loss": 0.0426, "step": 111105 }, { "epoch": 2.59, "learning_rate": 1.0627537490319401e-05, "loss": 0.0858, "step": 111110 }, { "epoch": 2.59, "learning_rate": 1.0626657435873366e-05, "loss": 0.068, "step": 111115 }, { "epoch": 2.59, "learning_rate": 1.0625777381427331e-05, "loss": 0.4151, "step": 111120 }, { "epoch": 2.59, "learning_rate": 1.0624897326981295e-05, "loss": 0.3339, "step": 111125 }, { "epoch": 2.59, "learning_rate": 1.0624017272535262e-05, "loss": 0.0247, "step": 111130 }, { "epoch": 2.59, "learning_rate": 1.0623137218089225e-05, "loss": 0.025, "step": 111135 }, { "epoch": 2.59, "learning_rate": 1.0622257163643192e-05, "loss": 0.0932, "step": 111140 }, { "epoch": 2.59, "learning_rate": 1.0621377109197155e-05, "loss": 0.0231, "step": 111145 }, { "epoch": 2.59, "learning_rate": 1.0620497054751122e-05, "loss": 0.063, "step": 111150 }, { "epoch": 2.59, "learning_rate": 1.0619617000305085e-05, "loss": 0.2242, "step": 111155 }, { "epoch": 2.59, "learning_rate": 1.061873694585905e-05, "loss": 0.068, "step": 111160 }, { "epoch": 2.59, "learning_rate": 1.0617856891413017e-05, "loss": 0.1971, "step": 111165 }, { "epoch": 2.59, "learning_rate": 1.061697683696698e-05, "loss": 0.186, "step": 111170 }, { "epoch": 2.59, "learning_rate": 1.0616096782520947e-05, "loss": 0.1684, "step": 111175 }, { "epoch": 2.59, "learning_rate": 1.061521672807491e-05, "loss": 0.0625, "step": 111180 }, { "epoch": 2.59, "learning_rate": 1.0614336673628877e-05, "loss": 0.0376, "step": 111185 }, { "epoch": 2.59, "learning_rate": 1.061345661918284e-05, "loss": 0.0403, "step": 111190 }, { "epoch": 2.59, "learning_rate": 1.0612576564736805e-05, "loss": 0.0179, "step": 111195 }, { "epoch": 2.59, "learning_rate": 1.061169651029077e-05, "loss": 0.0218, "step": 111200 }, { "epoch": 2.59, "learning_rate": 1.0610816455844735e-05, "loss": 0.0504, "step": 111205 }, { "epoch": 2.59, "learning_rate": 1.06099364013987e-05, "loss": 0.0528, "step": 111210 }, { "epoch": 2.59, "learning_rate": 1.0609056346952665e-05, "loss": 0.0691, "step": 111215 }, { "epoch": 2.59, "learning_rate": 1.060817629250663e-05, "loss": 0.3393, "step": 111220 }, { "epoch": 2.6, "learning_rate": 1.0607296238060595e-05, "loss": 0.1555, "step": 111225 }, { "epoch": 2.6, "learning_rate": 1.060641618361456e-05, "loss": 0.0212, "step": 111230 }, { "epoch": 2.6, "learning_rate": 1.0605536129168525e-05, "loss": 0.0675, "step": 111235 }, { "epoch": 2.6, "learning_rate": 1.0604656074722489e-05, "loss": 0.0233, "step": 111240 }, { "epoch": 2.6, "learning_rate": 1.0603776020276455e-05, "loss": 0.1655, "step": 111245 }, { "epoch": 2.6, "learning_rate": 1.0602895965830419e-05, "loss": 0.1009, "step": 111250 }, { "epoch": 2.6, "learning_rate": 1.0602015911384385e-05, "loss": 0.0367, "step": 111255 }, { "epoch": 2.6, "learning_rate": 1.0601135856938349e-05, "loss": 0.0822, "step": 111260 }, { "epoch": 2.6, "learning_rate": 1.0600255802492315e-05, "loss": 0.0915, "step": 111265 }, { "epoch": 2.6, "learning_rate": 1.0599375748046279e-05, "loss": 0.1894, "step": 111270 }, { "epoch": 2.6, "learning_rate": 1.0598495693600244e-05, "loss": 0.2826, "step": 111275 }, { "epoch": 2.6, "learning_rate": 1.0597615639154209e-05, "loss": 0.0277, "step": 111280 }, { "epoch": 2.6, "learning_rate": 1.0596735584708174e-05, "loss": 0.0154, "step": 111285 }, { "epoch": 2.6, "learning_rate": 1.0595855530262139e-05, "loss": 0.0857, "step": 111290 }, { "epoch": 2.6, "learning_rate": 1.0594975475816104e-05, "loss": 0.0629, "step": 111295 }, { "epoch": 2.6, "learning_rate": 1.059409542137007e-05, "loss": 0.0246, "step": 111300 }, { "epoch": 2.6, "learning_rate": 1.0593215366924034e-05, "loss": 0.1767, "step": 111305 }, { "epoch": 2.6, "learning_rate": 1.0592335312477999e-05, "loss": 0.159, "step": 111310 }, { "epoch": 2.6, "learning_rate": 1.0591455258031964e-05, "loss": 0.1874, "step": 111315 }, { "epoch": 2.6, "learning_rate": 1.0590575203585929e-05, "loss": 0.2146, "step": 111320 }, { "epoch": 2.6, "learning_rate": 1.0589695149139894e-05, "loss": 0.2501, "step": 111325 }, { "epoch": 2.6, "learning_rate": 1.0588815094693859e-05, "loss": 0.008, "step": 111330 }, { "epoch": 2.6, "learning_rate": 1.0587935040247824e-05, "loss": 0.0284, "step": 111335 }, { "epoch": 2.6, "learning_rate": 1.0587054985801789e-05, "loss": 0.0262, "step": 111340 }, { "epoch": 2.6, "learning_rate": 1.0586174931355752e-05, "loss": 0.1188, "step": 111345 }, { "epoch": 2.6, "learning_rate": 1.0585294876909719e-05, "loss": 0.099, "step": 111350 }, { "epoch": 2.6, "learning_rate": 1.0584414822463682e-05, "loss": 0.0612, "step": 111355 }, { "epoch": 2.6, "learning_rate": 1.0583534768017649e-05, "loss": 0.0628, "step": 111360 }, { "epoch": 2.6, "learning_rate": 1.0582654713571612e-05, "loss": 0.1229, "step": 111365 }, { "epoch": 2.6, "learning_rate": 1.0581774659125579e-05, "loss": 0.1397, "step": 111370 }, { "epoch": 2.6, "learning_rate": 1.0580894604679542e-05, "loss": 0.2142, "step": 111375 }, { "epoch": 2.6, "learning_rate": 1.058001455023351e-05, "loss": 0.0264, "step": 111380 }, { "epoch": 2.6, "learning_rate": 1.0579134495787472e-05, "loss": 0.0605, "step": 111385 }, { "epoch": 2.6, "learning_rate": 1.0578254441341438e-05, "loss": 0.0935, "step": 111390 }, { "epoch": 2.6, "learning_rate": 1.0577374386895403e-05, "loss": 0.0233, "step": 111395 }, { "epoch": 2.6, "learning_rate": 1.0576494332449368e-05, "loss": 0.2151, "step": 111400 }, { "epoch": 2.6, "learning_rate": 1.0575614278003333e-05, "loss": 0.06, "step": 111405 }, { "epoch": 2.6, "learning_rate": 1.0574734223557298e-05, "loss": 0.0068, "step": 111410 }, { "epoch": 2.6, "learning_rate": 1.0573854169111263e-05, "loss": 0.1188, "step": 111415 }, { "epoch": 2.6, "learning_rate": 1.0572974114665228e-05, "loss": 0.1471, "step": 111420 }, { "epoch": 2.6, "learning_rate": 1.0572094060219191e-05, "loss": 0.1749, "step": 111425 }, { "epoch": 2.6, "learning_rate": 1.0571214005773158e-05, "loss": 0.0358, "step": 111430 }, { "epoch": 2.6, "learning_rate": 1.0570333951327123e-05, "loss": 0.0917, "step": 111435 }, { "epoch": 2.6, "learning_rate": 1.0569453896881088e-05, "loss": 0.0062, "step": 111440 }, { "epoch": 2.6, "learning_rate": 1.0568573842435053e-05, "loss": 0.0226, "step": 111445 }, { "epoch": 2.6, "learning_rate": 1.0567693787989018e-05, "loss": 0.078, "step": 111450 }, { "epoch": 2.6, "learning_rate": 1.0566813733542983e-05, "loss": 0.2579, "step": 111455 }, { "epoch": 2.6, "learning_rate": 1.0565933679096946e-05, "loss": 0.0551, "step": 111460 }, { "epoch": 2.6, "learning_rate": 1.0565053624650913e-05, "loss": 0.2174, "step": 111465 }, { "epoch": 2.6, "learning_rate": 1.0564173570204876e-05, "loss": 0.3294, "step": 111470 }, { "epoch": 2.6, "learning_rate": 1.0563293515758843e-05, "loss": 0.1024, "step": 111475 }, { "epoch": 2.6, "learning_rate": 1.0562413461312806e-05, "loss": 0.053, "step": 111480 }, { "epoch": 2.6, "learning_rate": 1.0561533406866773e-05, "loss": 0.0105, "step": 111485 }, { "epoch": 2.6, "learning_rate": 1.0560653352420736e-05, "loss": 0.0628, "step": 111490 }, { "epoch": 2.6, "learning_rate": 1.0559773297974701e-05, "loss": 0.0265, "step": 111495 }, { "epoch": 2.6, "learning_rate": 1.0558893243528666e-05, "loss": 0.0166, "step": 111500 }, { "epoch": 2.6, "learning_rate": 1.0558013189082631e-05, "loss": 0.081, "step": 111505 }, { "epoch": 2.6, "learning_rate": 1.0557133134636596e-05, "loss": 0.053, "step": 111510 }, { "epoch": 2.6, "learning_rate": 1.0556253080190561e-05, "loss": 0.1219, "step": 111515 }, { "epoch": 2.6, "learning_rate": 1.0555373025744526e-05, "loss": 0.394, "step": 111520 }, { "epoch": 2.6, "learning_rate": 1.0554492971298491e-05, "loss": 0.1232, "step": 111525 }, { "epoch": 2.6, "learning_rate": 1.0553612916852456e-05, "loss": 0.1107, "step": 111530 }, { "epoch": 2.6, "learning_rate": 1.0552732862406421e-05, "loss": 0.0126, "step": 111535 }, { "epoch": 2.6, "learning_rate": 1.0551852807960385e-05, "loss": 0.0954, "step": 111540 }, { "epoch": 2.6, "learning_rate": 1.0550972753514351e-05, "loss": 0.0939, "step": 111545 }, { "epoch": 2.6, "learning_rate": 1.0550092699068315e-05, "loss": 0.0979, "step": 111550 }, { "epoch": 2.6, "learning_rate": 1.0549212644622282e-05, "loss": 0.116, "step": 111555 }, { "epoch": 2.6, "learning_rate": 1.0548332590176245e-05, "loss": 0.0874, "step": 111560 }, { "epoch": 2.6, "learning_rate": 1.0547452535730212e-05, "loss": 0.1647, "step": 111565 }, { "epoch": 2.6, "learning_rate": 1.0546572481284177e-05, "loss": 0.2002, "step": 111570 }, { "epoch": 2.6, "learning_rate": 1.054569242683814e-05, "loss": 0.0484, "step": 111575 }, { "epoch": 2.6, "learning_rate": 1.0544812372392107e-05, "loss": 0.0223, "step": 111580 }, { "epoch": 2.6, "learning_rate": 1.054393231794607e-05, "loss": 0.0004, "step": 111585 }, { "epoch": 2.6, "learning_rate": 1.0543052263500037e-05, "loss": 0.0639, "step": 111590 }, { "epoch": 2.6, "learning_rate": 1.0542172209054e-05, "loss": 0.0456, "step": 111595 }, { "epoch": 2.6, "learning_rate": 1.0541292154607967e-05, "loss": 0.0268, "step": 111600 }, { "epoch": 2.6, "learning_rate": 1.054041210016193e-05, "loss": 0.0252, "step": 111605 }, { "epoch": 2.6, "learning_rate": 1.0539532045715895e-05, "loss": 0.0867, "step": 111610 }, { "epoch": 2.6, "learning_rate": 1.053865199126986e-05, "loss": 0.0312, "step": 111615 }, { "epoch": 2.6, "learning_rate": 1.0537771936823825e-05, "loss": 0.0183, "step": 111620 }, { "epoch": 2.6, "learning_rate": 1.053689188237779e-05, "loss": 0.2456, "step": 111625 }, { "epoch": 2.6, "learning_rate": 1.0536011827931755e-05, "loss": 0.1511, "step": 111630 }, { "epoch": 2.6, "learning_rate": 1.053513177348572e-05, "loss": 0.0102, "step": 111635 }, { "epoch": 2.6, "learning_rate": 1.0534251719039685e-05, "loss": 0.0591, "step": 111640 }, { "epoch": 2.6, "learning_rate": 1.0533371664593648e-05, "loss": 0.1013, "step": 111645 }, { "epoch": 2.6, "learning_rate": 1.0532491610147615e-05, "loss": 0.0147, "step": 111650 }, { "epoch": 2.61, "learning_rate": 1.0531611555701579e-05, "loss": 0.0632, "step": 111655 }, { "epoch": 2.61, "learning_rate": 1.0530731501255545e-05, "loss": 0.0793, "step": 111660 }, { "epoch": 2.61, "learning_rate": 1.0529851446809509e-05, "loss": 0.0731, "step": 111665 }, { "epoch": 2.61, "learning_rate": 1.0528971392363475e-05, "loss": 0.1632, "step": 111670 }, { "epoch": 2.61, "learning_rate": 1.0528091337917439e-05, "loss": 0.2144, "step": 111675 }, { "epoch": 2.61, "learning_rate": 1.0527211283471405e-05, "loss": 0.0205, "step": 111680 }, { "epoch": 2.61, "learning_rate": 1.0526331229025369e-05, "loss": 0.0532, "step": 111685 }, { "epoch": 2.61, "learning_rate": 1.0525451174579334e-05, "loss": 0.0515, "step": 111690 }, { "epoch": 2.61, "learning_rate": 1.0524571120133299e-05, "loss": 0.1189, "step": 111695 }, { "epoch": 2.61, "learning_rate": 1.0523691065687264e-05, "loss": 0.1529, "step": 111700 }, { "epoch": 2.61, "learning_rate": 1.0522811011241229e-05, "loss": 0.0976, "step": 111705 }, { "epoch": 2.61, "learning_rate": 1.0521930956795194e-05, "loss": 0.225, "step": 111710 }, { "epoch": 2.61, "learning_rate": 1.052105090234916e-05, "loss": 0.1031, "step": 111715 }, { "epoch": 2.61, "learning_rate": 1.0520170847903124e-05, "loss": 0.2021, "step": 111720 }, { "epoch": 2.61, "learning_rate": 1.0519290793457089e-05, "loss": 0.1128, "step": 111725 }, { "epoch": 2.61, "learning_rate": 1.0518410739011054e-05, "loss": 0.0916, "step": 111730 }, { "epoch": 2.61, "learning_rate": 1.0517530684565019e-05, "loss": 0.0443, "step": 111735 }, { "epoch": 2.61, "learning_rate": 1.0516650630118984e-05, "loss": 0.1181, "step": 111740 }, { "epoch": 2.61, "learning_rate": 1.0515770575672949e-05, "loss": 0.0612, "step": 111745 }, { "epoch": 2.61, "learning_rate": 1.0514890521226914e-05, "loss": 0.0703, "step": 111750 }, { "epoch": 2.61, "learning_rate": 1.0514010466780879e-05, "loss": 0.0413, "step": 111755 }, { "epoch": 2.61, "learning_rate": 1.0513130412334842e-05, "loss": 0.2172, "step": 111760 }, { "epoch": 2.61, "learning_rate": 1.0512250357888809e-05, "loss": 0.0826, "step": 111765 }, { "epoch": 2.61, "learning_rate": 1.0511370303442772e-05, "loss": 0.3065, "step": 111770 }, { "epoch": 2.61, "learning_rate": 1.0510490248996739e-05, "loss": 0.233, "step": 111775 }, { "epoch": 2.61, "learning_rate": 1.0509610194550702e-05, "loss": 0.0015, "step": 111780 }, { "epoch": 2.61, "learning_rate": 1.0508730140104669e-05, "loss": 0.0915, "step": 111785 }, { "epoch": 2.61, "learning_rate": 1.0507850085658632e-05, "loss": 0.0812, "step": 111790 }, { "epoch": 2.61, "learning_rate": 1.0506970031212597e-05, "loss": 0.0199, "step": 111795 }, { "epoch": 2.61, "learning_rate": 1.0506089976766562e-05, "loss": 0.0991, "step": 111800 }, { "epoch": 2.61, "learning_rate": 1.0505209922320527e-05, "loss": 0.0807, "step": 111805 }, { "epoch": 2.61, "learning_rate": 1.0504329867874492e-05, "loss": 0.059, "step": 111810 }, { "epoch": 2.61, "learning_rate": 1.0503449813428457e-05, "loss": 0.1642, "step": 111815 }, { "epoch": 2.61, "learning_rate": 1.0502569758982423e-05, "loss": 0.2378, "step": 111820 }, { "epoch": 2.61, "learning_rate": 1.0501689704536388e-05, "loss": 0.2218, "step": 111825 }, { "epoch": 2.61, "learning_rate": 1.0500809650090353e-05, "loss": 0.0745, "step": 111830 }, { "epoch": 2.61, "learning_rate": 1.0499929595644318e-05, "loss": 0.0004, "step": 111835 }, { "epoch": 2.61, "learning_rate": 1.0499049541198283e-05, "loss": 0.0044, "step": 111840 }, { "epoch": 2.61, "learning_rate": 1.0498169486752248e-05, "loss": 0.0379, "step": 111845 }, { "epoch": 2.61, "learning_rate": 1.0497289432306213e-05, "loss": 0.0176, "step": 111850 }, { "epoch": 2.61, "learning_rate": 1.0496409377860178e-05, "loss": 0.0953, "step": 111855 }, { "epoch": 2.61, "learning_rate": 1.0495529323414143e-05, "loss": 0.1006, "step": 111860 }, { "epoch": 2.61, "learning_rate": 1.0494649268968108e-05, "loss": 0.1139, "step": 111865 }, { "epoch": 2.61, "learning_rate": 1.0493769214522073e-05, "loss": 0.1741, "step": 111870 }, { "epoch": 2.61, "learning_rate": 1.0492889160076036e-05, "loss": 0.1889, "step": 111875 }, { "epoch": 2.61, "learning_rate": 1.0492009105630003e-05, "loss": 0.1611, "step": 111880 }, { "epoch": 2.61, "learning_rate": 1.0491129051183966e-05, "loss": 0.0057, "step": 111885 }, { "epoch": 2.61, "learning_rate": 1.0490248996737933e-05, "loss": 0.009, "step": 111890 }, { "epoch": 2.61, "learning_rate": 1.0489368942291896e-05, "loss": 0.1187, "step": 111895 }, { "epoch": 2.61, "learning_rate": 1.0488488887845863e-05, "loss": 0.0595, "step": 111900 }, { "epoch": 2.61, "learning_rate": 1.0487608833399826e-05, "loss": 0.0605, "step": 111905 }, { "epoch": 2.61, "learning_rate": 1.0486728778953791e-05, "loss": 0.0701, "step": 111910 }, { "epoch": 2.61, "learning_rate": 1.0485848724507756e-05, "loss": 0.0931, "step": 111915 }, { "epoch": 2.61, "learning_rate": 1.0484968670061721e-05, "loss": 0.1001, "step": 111920 }, { "epoch": 2.61, "learning_rate": 1.0484088615615686e-05, "loss": 0.1925, "step": 111925 }, { "epoch": 2.61, "learning_rate": 1.0483208561169651e-05, "loss": 0.0731, "step": 111930 }, { "epoch": 2.61, "learning_rate": 1.0482328506723616e-05, "loss": 0.0557, "step": 111935 }, { "epoch": 2.61, "learning_rate": 1.0481448452277581e-05, "loss": 0.1053, "step": 111940 }, { "epoch": 2.61, "learning_rate": 1.0480568397831545e-05, "loss": 0.0004, "step": 111945 }, { "epoch": 2.61, "learning_rate": 1.0479688343385511e-05, "loss": 0.1275, "step": 111950 }, { "epoch": 2.61, "learning_rate": 1.0478808288939475e-05, "loss": 0.1149, "step": 111955 }, { "epoch": 2.61, "learning_rate": 1.0477928234493441e-05, "loss": 0.091, "step": 111960 }, { "epoch": 2.61, "learning_rate": 1.0477048180047405e-05, "loss": 0.089, "step": 111965 }, { "epoch": 2.61, "learning_rate": 1.0476168125601371e-05, "loss": 0.0422, "step": 111970 }, { "epoch": 2.61, "learning_rate": 1.0475288071155335e-05, "loss": 0.2393, "step": 111975 }, { "epoch": 2.61, "learning_rate": 1.0474408016709301e-05, "loss": 0.0295, "step": 111980 }, { "epoch": 2.61, "learning_rate": 1.0473527962263266e-05, "loss": 0.0427, "step": 111985 }, { "epoch": 2.61, "learning_rate": 1.047264790781723e-05, "loss": 0.0401, "step": 111990 }, { "epoch": 2.61, "learning_rate": 1.0471767853371197e-05, "loss": 0.1152, "step": 111995 }, { "epoch": 2.61, "learning_rate": 1.047088779892516e-05, "loss": 0.0974, "step": 112000 }, { "epoch": 2.61, "learning_rate": 1.0470007744479127e-05, "loss": 0.1892, "step": 112005 }, { "epoch": 2.61, "learning_rate": 1.046912769003309e-05, "loss": 0.0516, "step": 112010 }, { "epoch": 2.61, "learning_rate": 1.0468247635587057e-05, "loss": 0.1284, "step": 112015 }, { "epoch": 2.61, "learning_rate": 1.046736758114102e-05, "loss": 0.3643, "step": 112020 }, { "epoch": 2.61, "learning_rate": 1.0466487526694985e-05, "loss": 0.5213, "step": 112025 }, { "epoch": 2.61, "learning_rate": 1.046560747224895e-05, "loss": 0.0618, "step": 112030 }, { "epoch": 2.61, "learning_rate": 1.0464727417802915e-05, "loss": 0.0747, "step": 112035 }, { "epoch": 2.61, "learning_rate": 1.046384736335688e-05, "loss": 0.0662, "step": 112040 }, { "epoch": 2.61, "learning_rate": 1.0462967308910845e-05, "loss": 0.0557, "step": 112045 }, { "epoch": 2.61, "learning_rate": 1.046208725446481e-05, "loss": 0.0835, "step": 112050 }, { "epoch": 2.61, "learning_rate": 1.0461207200018775e-05, "loss": 0.1088, "step": 112055 }, { "epoch": 2.61, "learning_rate": 1.0460327145572738e-05, "loss": 0.0335, "step": 112060 }, { "epoch": 2.61, "learning_rate": 1.0459447091126705e-05, "loss": 0.2438, "step": 112065 }, { "epoch": 2.61, "learning_rate": 1.0458567036680668e-05, "loss": 0.1414, "step": 112070 }, { "epoch": 2.61, "learning_rate": 1.0457686982234635e-05, "loss": 0.0936, "step": 112075 }, { "epoch": 2.61, "learning_rate": 1.0456806927788598e-05, "loss": 0.0561, "step": 112080 }, { "epoch": 2.62, "learning_rate": 1.0455926873342565e-05, "loss": 0.0008, "step": 112085 }, { "epoch": 2.62, "learning_rate": 1.0455046818896529e-05, "loss": 0.0462, "step": 112090 }, { "epoch": 2.62, "learning_rate": 1.0454166764450494e-05, "loss": 0.0265, "step": 112095 }, { "epoch": 2.62, "learning_rate": 1.0453286710004459e-05, "loss": 0.1343, "step": 112100 }, { "epoch": 2.62, "learning_rate": 1.0452406655558424e-05, "loss": 0.1262, "step": 112105 }, { "epoch": 2.62, "learning_rate": 1.0451526601112389e-05, "loss": 0.156, "step": 112110 }, { "epoch": 2.62, "learning_rate": 1.0450646546666354e-05, "loss": 0.1617, "step": 112115 }, { "epoch": 2.62, "learning_rate": 1.044976649222032e-05, "loss": 0.1404, "step": 112120 }, { "epoch": 2.62, "learning_rate": 1.0448886437774284e-05, "loss": 0.3577, "step": 112125 }, { "epoch": 2.62, "learning_rate": 1.044800638332825e-05, "loss": 0.0275, "step": 112130 }, { "epoch": 2.62, "learning_rate": 1.0447126328882214e-05, "loss": 0.077, "step": 112135 }, { "epoch": 2.62, "learning_rate": 1.0446246274436179e-05, "loss": 0.0235, "step": 112140 }, { "epoch": 2.62, "learning_rate": 1.0445366219990144e-05, "loss": 0.0628, "step": 112145 }, { "epoch": 2.62, "learning_rate": 1.0444486165544109e-05, "loss": 0.0159, "step": 112150 }, { "epoch": 2.62, "learning_rate": 1.0443606111098074e-05, "loss": 0.052, "step": 112155 }, { "epoch": 2.62, "learning_rate": 1.0442726056652039e-05, "loss": 0.1161, "step": 112160 }, { "epoch": 2.62, "learning_rate": 1.0441846002206004e-05, "loss": 0.1793, "step": 112165 }, { "epoch": 2.62, "learning_rate": 1.0440965947759969e-05, "loss": 0.151, "step": 112170 }, { "epoch": 2.62, "learning_rate": 1.0440085893313932e-05, "loss": 0.1196, "step": 112175 }, { "epoch": 2.62, "learning_rate": 1.0439205838867899e-05, "loss": 0.1233, "step": 112180 }, { "epoch": 2.62, "learning_rate": 1.0438325784421862e-05, "loss": 0.0113, "step": 112185 }, { "epoch": 2.62, "learning_rate": 1.0437445729975829e-05, "loss": 0.0171, "step": 112190 }, { "epoch": 2.62, "learning_rate": 1.0436565675529792e-05, "loss": 0.0525, "step": 112195 }, { "epoch": 2.62, "learning_rate": 1.0435685621083759e-05, "loss": 0.1313, "step": 112200 }, { "epoch": 2.62, "learning_rate": 1.0434805566637722e-05, "loss": 0.0613, "step": 112205 }, { "epoch": 2.62, "learning_rate": 1.0433925512191687e-05, "loss": 0.1089, "step": 112210 }, { "epoch": 2.62, "learning_rate": 1.0433045457745652e-05, "loss": 0.1828, "step": 112215 }, { "epoch": 2.62, "learning_rate": 1.0432165403299617e-05, "loss": 0.1602, "step": 112220 }, { "epoch": 2.62, "learning_rate": 1.0431285348853582e-05, "loss": 0.2612, "step": 112225 }, { "epoch": 2.62, "learning_rate": 1.0430405294407547e-05, "loss": 0.0587, "step": 112230 }, { "epoch": 2.62, "learning_rate": 1.0429525239961512e-05, "loss": 0.0319, "step": 112235 }, { "epoch": 2.62, "learning_rate": 1.0428645185515477e-05, "loss": 0.1765, "step": 112240 }, { "epoch": 2.62, "learning_rate": 1.042776513106944e-05, "loss": 0.088, "step": 112245 }, { "epoch": 2.62, "learning_rate": 1.0426885076623407e-05, "loss": 0.0234, "step": 112250 }, { "epoch": 2.62, "learning_rate": 1.0426005022177373e-05, "loss": 0.0839, "step": 112255 }, { "epoch": 2.62, "learning_rate": 1.0425124967731338e-05, "loss": 0.0655, "step": 112260 }, { "epoch": 2.62, "learning_rate": 1.0424244913285303e-05, "loss": 0.0184, "step": 112265 }, { "epoch": 2.62, "learning_rate": 1.0423364858839268e-05, "loss": 0.1471, "step": 112270 }, { "epoch": 2.62, "learning_rate": 1.0422484804393233e-05, "loss": 0.2158, "step": 112275 }, { "epoch": 2.62, "learning_rate": 1.0421604749947198e-05, "loss": 0.041, "step": 112280 }, { "epoch": 2.62, "learning_rate": 1.0420724695501163e-05, "loss": 0.004, "step": 112285 }, { "epoch": 2.62, "learning_rate": 1.0419844641055126e-05, "loss": 0.0305, "step": 112290 }, { "epoch": 2.62, "learning_rate": 1.0418964586609093e-05, "loss": 0.0397, "step": 112295 }, { "epoch": 2.62, "learning_rate": 1.0418084532163056e-05, "loss": 0.0264, "step": 112300 }, { "epoch": 2.62, "learning_rate": 1.0417204477717023e-05, "loss": 0.1008, "step": 112305 }, { "epoch": 2.62, "learning_rate": 1.0416324423270986e-05, "loss": 0.1256, "step": 112310 }, { "epoch": 2.62, "learning_rate": 1.0415444368824953e-05, "loss": 0.1404, "step": 112315 }, { "epoch": 2.62, "learning_rate": 1.0414564314378916e-05, "loss": 0.2998, "step": 112320 }, { "epoch": 2.62, "learning_rate": 1.0413684259932881e-05, "loss": 0.2557, "step": 112325 }, { "epoch": 2.62, "learning_rate": 1.0412804205486846e-05, "loss": 0.1598, "step": 112330 }, { "epoch": 2.62, "learning_rate": 1.0411924151040811e-05, "loss": 0.016, "step": 112335 }, { "epoch": 2.62, "learning_rate": 1.0411044096594776e-05, "loss": 0.0096, "step": 112340 }, { "epoch": 2.62, "learning_rate": 1.0410164042148741e-05, "loss": 0.0243, "step": 112345 }, { "epoch": 2.62, "learning_rate": 1.0409283987702706e-05, "loss": 0.0681, "step": 112350 }, { "epoch": 2.62, "learning_rate": 1.0408403933256671e-05, "loss": 0.0463, "step": 112355 }, { "epoch": 2.62, "learning_rate": 1.0407523878810635e-05, "loss": 0.0654, "step": 112360 }, { "epoch": 2.62, "learning_rate": 1.0406643824364601e-05, "loss": 0.0362, "step": 112365 }, { "epoch": 2.62, "learning_rate": 1.0405763769918565e-05, "loss": 0.0936, "step": 112370 }, { "epoch": 2.62, "learning_rate": 1.0404883715472531e-05, "loss": 0.1439, "step": 112375 }, { "epoch": 2.62, "learning_rate": 1.0404003661026495e-05, "loss": 0.0102, "step": 112380 }, { "epoch": 2.62, "learning_rate": 1.0403123606580461e-05, "loss": 0.0302, "step": 112385 }, { "epoch": 2.62, "learning_rate": 1.0402243552134426e-05, "loss": 0.0193, "step": 112390 }, { "epoch": 2.62, "learning_rate": 1.040136349768839e-05, "loss": 0.1381, "step": 112395 }, { "epoch": 2.62, "learning_rate": 1.0400483443242356e-05, "loss": 0.0051, "step": 112400 }, { "epoch": 2.62, "learning_rate": 1.039960338879632e-05, "loss": 0.0897, "step": 112405 }, { "epoch": 2.62, "learning_rate": 1.0398723334350286e-05, "loss": 0.2777, "step": 112410 }, { "epoch": 2.62, "learning_rate": 1.039784327990425e-05, "loss": 0.2699, "step": 112415 }, { "epoch": 2.62, "learning_rate": 1.0396963225458217e-05, "loss": 0.1825, "step": 112420 }, { "epoch": 2.62, "learning_rate": 1.039608317101218e-05, "loss": 0.2178, "step": 112425 }, { "epoch": 2.62, "learning_rate": 1.0395203116566147e-05, "loss": 0.022, "step": 112430 }, { "epoch": 2.62, "learning_rate": 1.039432306212011e-05, "loss": 0.072, "step": 112435 }, { "epoch": 2.62, "learning_rate": 1.0393443007674075e-05, "loss": 0.0727, "step": 112440 }, { "epoch": 2.62, "learning_rate": 1.039256295322804e-05, "loss": 0.0426, "step": 112445 }, { "epoch": 2.62, "learning_rate": 1.0391682898782005e-05, "loss": 0.0171, "step": 112450 }, { "epoch": 2.62, "learning_rate": 1.039080284433597e-05, "loss": 0.062, "step": 112455 }, { "epoch": 2.62, "learning_rate": 1.0389922789889935e-05, "loss": 0.2462, "step": 112460 }, { "epoch": 2.62, "learning_rate": 1.03890427354439e-05, "loss": 0.3592, "step": 112465 }, { "epoch": 2.62, "learning_rate": 1.0388162680997865e-05, "loss": 0.0891, "step": 112470 }, { "epoch": 2.62, "learning_rate": 1.0387282626551828e-05, "loss": 0.2005, "step": 112475 }, { "epoch": 2.62, "learning_rate": 1.0386402572105795e-05, "loss": 0.0623, "step": 112480 }, { "epoch": 2.62, "learning_rate": 1.0385522517659758e-05, "loss": 0.0001, "step": 112485 }, { "epoch": 2.62, "learning_rate": 1.0384642463213725e-05, "loss": 0.0201, "step": 112490 }, { "epoch": 2.62, "learning_rate": 1.0383762408767688e-05, "loss": 0.0525, "step": 112495 }, { "epoch": 2.62, "learning_rate": 1.0382882354321655e-05, "loss": 0.0012, "step": 112500 }, { "epoch": 2.62, "learning_rate": 1.0382002299875618e-05, "loss": 0.0025, "step": 112505 }, { "epoch": 2.62, "learning_rate": 1.0381122245429583e-05, "loss": 0.1424, "step": 112510 }, { "epoch": 2.63, "learning_rate": 1.0380242190983549e-05, "loss": 0.0819, "step": 112515 }, { "epoch": 2.63, "learning_rate": 1.0379362136537514e-05, "loss": 0.183, "step": 112520 }, { "epoch": 2.63, "learning_rate": 1.037848208209148e-05, "loss": 0.0973, "step": 112525 }, { "epoch": 2.63, "learning_rate": 1.0377602027645444e-05, "loss": 0.0043, "step": 112530 }, { "epoch": 2.63, "learning_rate": 1.037672197319941e-05, "loss": 0.0466, "step": 112535 }, { "epoch": 2.63, "learning_rate": 1.0375841918753374e-05, "loss": 0.0329, "step": 112540 }, { "epoch": 2.63, "learning_rate": 1.0374961864307339e-05, "loss": 0.0371, "step": 112545 }, { "epoch": 2.63, "learning_rate": 1.0374081809861304e-05, "loss": 0.0454, "step": 112550 }, { "epoch": 2.63, "learning_rate": 1.0373201755415269e-05, "loss": 0.1146, "step": 112555 }, { "epoch": 2.63, "learning_rate": 1.0372321700969234e-05, "loss": 0.0434, "step": 112560 }, { "epoch": 2.63, "learning_rate": 1.0371441646523199e-05, "loss": 0.0827, "step": 112565 }, { "epoch": 2.63, "learning_rate": 1.0370561592077164e-05, "loss": 0.3113, "step": 112570 }, { "epoch": 2.63, "learning_rate": 1.0369681537631129e-05, "loss": 0.2372, "step": 112575 }, { "epoch": 2.63, "learning_rate": 1.0368801483185094e-05, "loss": 0.0072, "step": 112580 }, { "epoch": 2.63, "learning_rate": 1.0367921428739059e-05, "loss": 0.024, "step": 112585 }, { "epoch": 2.63, "learning_rate": 1.0367041374293022e-05, "loss": 0.0303, "step": 112590 }, { "epoch": 2.63, "learning_rate": 1.0366161319846989e-05, "loss": 0.004, "step": 112595 }, { "epoch": 2.63, "learning_rate": 1.0365281265400952e-05, "loss": 0.0164, "step": 112600 }, { "epoch": 2.63, "learning_rate": 1.0364401210954919e-05, "loss": 0.0098, "step": 112605 }, { "epoch": 2.63, "learning_rate": 1.0363521156508882e-05, "loss": 0.0745, "step": 112610 }, { "epoch": 2.63, "learning_rate": 1.0362641102062849e-05, "loss": 0.1421, "step": 112615 }, { "epoch": 2.63, "learning_rate": 1.0361761047616812e-05, "loss": 0.2634, "step": 112620 }, { "epoch": 2.63, "learning_rate": 1.0360880993170777e-05, "loss": 0.1807, "step": 112625 }, { "epoch": 2.63, "learning_rate": 1.0360000938724742e-05, "loss": 0.0702, "step": 112630 }, { "epoch": 2.63, "learning_rate": 1.0359120884278707e-05, "loss": 0.0655, "step": 112635 }, { "epoch": 2.63, "learning_rate": 1.0358240829832672e-05, "loss": 0.003, "step": 112640 }, { "epoch": 2.63, "learning_rate": 1.0357360775386637e-05, "loss": 0.0399, "step": 112645 }, { "epoch": 2.63, "learning_rate": 1.0356480720940602e-05, "loss": 0.0521, "step": 112650 }, { "epoch": 2.63, "learning_rate": 1.0355600666494567e-05, "loss": 0.0532, "step": 112655 }, { "epoch": 2.63, "learning_rate": 1.0354720612048532e-05, "loss": 0.09, "step": 112660 }, { "epoch": 2.63, "learning_rate": 1.0353840557602497e-05, "loss": 0.1255, "step": 112665 }, { "epoch": 2.63, "learning_rate": 1.0352960503156462e-05, "loss": 0.1914, "step": 112670 }, { "epoch": 2.63, "learning_rate": 1.0352080448710427e-05, "loss": 0.1588, "step": 112675 }, { "epoch": 2.63, "learning_rate": 1.0351200394264392e-05, "loss": 0.01, "step": 112680 }, { "epoch": 2.63, "learning_rate": 1.0350320339818358e-05, "loss": 0.0375, "step": 112685 }, { "epoch": 2.63, "learning_rate": 1.0349440285372323e-05, "loss": 0.0968, "step": 112690 }, { "epoch": 2.63, "learning_rate": 1.0348560230926286e-05, "loss": 0.0851, "step": 112695 }, { "epoch": 2.63, "learning_rate": 1.0347680176480253e-05, "loss": 0.0465, "step": 112700 }, { "epoch": 2.63, "learning_rate": 1.0346800122034216e-05, "loss": 0.0906, "step": 112705 }, { "epoch": 2.63, "learning_rate": 1.0345920067588183e-05, "loss": 0.0329, "step": 112710 }, { "epoch": 2.63, "learning_rate": 1.0345040013142146e-05, "loss": 0.0465, "step": 112715 }, { "epoch": 2.63, "learning_rate": 1.0344159958696113e-05, "loss": 0.253, "step": 112720 }, { "epoch": 2.63, "learning_rate": 1.0343279904250076e-05, "loss": 0.3837, "step": 112725 }, { "epoch": 2.63, "learning_rate": 1.0342399849804043e-05, "loss": 0.0163, "step": 112730 }, { "epoch": 2.63, "learning_rate": 1.0341519795358006e-05, "loss": 0.0334, "step": 112735 }, { "epoch": 2.63, "learning_rate": 1.0340639740911971e-05, "loss": 0.1032, "step": 112740 }, { "epoch": 2.63, "learning_rate": 1.0339759686465936e-05, "loss": 0.0469, "step": 112745 }, { "epoch": 2.63, "learning_rate": 1.0338879632019901e-05, "loss": 0.0392, "step": 112750 }, { "epoch": 2.63, "learning_rate": 1.0337999577573866e-05, "loss": 0.0899, "step": 112755 }, { "epoch": 2.63, "learning_rate": 1.0337119523127831e-05, "loss": 0.197, "step": 112760 }, { "epoch": 2.63, "learning_rate": 1.0336239468681796e-05, "loss": 0.0951, "step": 112765 }, { "epoch": 2.63, "learning_rate": 1.0335359414235761e-05, "loss": 0.0842, "step": 112770 }, { "epoch": 2.63, "learning_rate": 1.0334479359789724e-05, "loss": 0.195, "step": 112775 }, { "epoch": 2.63, "learning_rate": 1.0333599305343691e-05, "loss": 0.0013, "step": 112780 }, { "epoch": 2.63, "learning_rate": 1.0332719250897655e-05, "loss": 0.0004, "step": 112785 }, { "epoch": 2.63, "learning_rate": 1.0331839196451621e-05, "loss": 0.0742, "step": 112790 }, { "epoch": 2.63, "learning_rate": 1.0330959142005586e-05, "loss": 0.0373, "step": 112795 }, { "epoch": 2.63, "learning_rate": 1.0330079087559551e-05, "loss": 0.0669, "step": 112800 }, { "epoch": 2.63, "learning_rate": 1.0329199033113516e-05, "loss": 0.1646, "step": 112805 }, { "epoch": 2.63, "learning_rate": 1.032831897866748e-05, "loss": 0.0294, "step": 112810 }, { "epoch": 2.63, "learning_rate": 1.0327438924221446e-05, "loss": 0.0933, "step": 112815 }, { "epoch": 2.63, "learning_rate": 1.032655886977541e-05, "loss": 0.1202, "step": 112820 }, { "epoch": 2.63, "learning_rate": 1.0325678815329376e-05, "loss": 0.245, "step": 112825 }, { "epoch": 2.63, "learning_rate": 1.032479876088334e-05, "loss": 0.0421, "step": 112830 }, { "epoch": 2.63, "learning_rate": 1.0323918706437306e-05, "loss": 0.0636, "step": 112835 }, { "epoch": 2.63, "learning_rate": 1.032303865199127e-05, "loss": 0.018, "step": 112840 }, { "epoch": 2.63, "learning_rate": 1.0322158597545235e-05, "loss": 0.0665, "step": 112845 }, { "epoch": 2.63, "learning_rate": 1.03212785430992e-05, "loss": 0.0313, "step": 112850 }, { "epoch": 2.63, "learning_rate": 1.0320398488653165e-05, "loss": 0.032, "step": 112855 }, { "epoch": 2.63, "learning_rate": 1.031951843420713e-05, "loss": 0.0615, "step": 112860 }, { "epoch": 2.63, "learning_rate": 1.0318638379761095e-05, "loss": 0.0669, "step": 112865 }, { "epoch": 2.63, "learning_rate": 1.031775832531506e-05, "loss": 0.1306, "step": 112870 }, { "epoch": 2.63, "learning_rate": 1.0316878270869025e-05, "loss": 0.1856, "step": 112875 }, { "epoch": 2.63, "learning_rate": 1.031599821642299e-05, "loss": 0.0963, "step": 112880 }, { "epoch": 2.63, "learning_rate": 1.0315118161976955e-05, "loss": 0.0221, "step": 112885 }, { "epoch": 2.63, "learning_rate": 1.0314238107530918e-05, "loss": 0.1275, "step": 112890 }, { "epoch": 2.63, "learning_rate": 1.0313358053084885e-05, "loss": 0.0125, "step": 112895 }, { "epoch": 2.63, "learning_rate": 1.0312477998638848e-05, "loss": 0.0542, "step": 112900 }, { "epoch": 2.63, "learning_rate": 1.0311597944192815e-05, "loss": 0.2123, "step": 112905 }, { "epoch": 2.63, "learning_rate": 1.0310717889746778e-05, "loss": 0.104, "step": 112910 }, { "epoch": 2.63, "learning_rate": 1.0309837835300745e-05, "loss": 0.0702, "step": 112915 }, { "epoch": 2.63, "learning_rate": 1.0308957780854708e-05, "loss": 0.301, "step": 112920 }, { "epoch": 2.63, "learning_rate": 1.0308077726408673e-05, "loss": 0.1517, "step": 112925 }, { "epoch": 2.63, "learning_rate": 1.030719767196264e-05, "loss": 0.1247, "step": 112930 }, { "epoch": 2.63, "learning_rate": 1.0306317617516603e-05, "loss": 0.0179, "step": 112935 }, { "epoch": 2.64, "learning_rate": 1.030543756307057e-05, "loss": 0.0222, "step": 112940 }, { "epoch": 2.64, "learning_rate": 1.0304557508624533e-05, "loss": 0.0501, "step": 112945 }, { "epoch": 2.64, "learning_rate": 1.03036774541785e-05, "loss": 0.1011, "step": 112950 }, { "epoch": 2.64, "learning_rate": 1.0302797399732464e-05, "loss": 0.0149, "step": 112955 }, { "epoch": 2.64, "learning_rate": 1.0301917345286429e-05, "loss": 0.0657, "step": 112960 }, { "epoch": 2.64, "learning_rate": 1.0301037290840394e-05, "loss": 0.1144, "step": 112965 }, { "epoch": 2.64, "learning_rate": 1.0300157236394359e-05, "loss": 0.1325, "step": 112970 }, { "epoch": 2.64, "learning_rate": 1.0299277181948324e-05, "loss": 0.2219, "step": 112975 }, { "epoch": 2.64, "learning_rate": 1.0298397127502289e-05, "loss": 0.0334, "step": 112980 }, { "epoch": 2.64, "learning_rate": 1.0297517073056254e-05, "loss": 0.0634, "step": 112985 }, { "epoch": 2.64, "learning_rate": 1.0296637018610219e-05, "loss": 0.0465, "step": 112990 }, { "epoch": 2.64, "learning_rate": 1.0295756964164182e-05, "loss": 0.0629, "step": 112995 }, { "epoch": 2.64, "learning_rate": 1.0294876909718149e-05, "loss": 0.085, "step": 113000 }, { "epoch": 2.64, "learning_rate": 1.0293996855272112e-05, "loss": 0.1141, "step": 113005 }, { "epoch": 2.64, "learning_rate": 1.0293116800826079e-05, "loss": 0.2028, "step": 113010 }, { "epoch": 2.64, "learning_rate": 1.0292236746380042e-05, "loss": 0.2361, "step": 113015 }, { "epoch": 2.64, "learning_rate": 1.0291356691934009e-05, "loss": 0.2976, "step": 113020 }, { "epoch": 2.64, "learning_rate": 1.0290476637487972e-05, "loss": 0.2871, "step": 113025 }, { "epoch": 2.64, "learning_rate": 1.0289596583041939e-05, "loss": 0.0094, "step": 113030 }, { "epoch": 2.64, "learning_rate": 1.0288716528595902e-05, "loss": 0.0068, "step": 113035 }, { "epoch": 2.64, "learning_rate": 1.0287836474149867e-05, "loss": 0.0518, "step": 113040 }, { "epoch": 2.64, "learning_rate": 1.0286956419703832e-05, "loss": 0.0176, "step": 113045 }, { "epoch": 2.64, "learning_rate": 1.0286076365257797e-05, "loss": 0.0528, "step": 113050 }, { "epoch": 2.64, "learning_rate": 1.0285196310811762e-05, "loss": 0.1677, "step": 113055 }, { "epoch": 2.64, "learning_rate": 1.0284316256365727e-05, "loss": 0.1092, "step": 113060 }, { "epoch": 2.64, "learning_rate": 1.0283436201919692e-05, "loss": 0.1094, "step": 113065 }, { "epoch": 2.64, "learning_rate": 1.0282556147473657e-05, "loss": 0.2272, "step": 113070 }, { "epoch": 2.64, "learning_rate": 1.0281676093027622e-05, "loss": 0.2621, "step": 113075 }, { "epoch": 2.64, "learning_rate": 1.0280796038581587e-05, "loss": 0.027, "step": 113080 }, { "epoch": 2.64, "learning_rate": 1.0279915984135552e-05, "loss": 0.0211, "step": 113085 }, { "epoch": 2.64, "learning_rate": 1.0279035929689517e-05, "loss": 0.0511, "step": 113090 }, { "epoch": 2.64, "learning_rate": 1.0278155875243482e-05, "loss": 0.0027, "step": 113095 }, { "epoch": 2.64, "learning_rate": 1.0277275820797447e-05, "loss": 0.0759, "step": 113100 }, { "epoch": 2.64, "learning_rate": 1.0276395766351412e-05, "loss": 0.0528, "step": 113105 }, { "epoch": 2.64, "learning_rate": 1.0275515711905376e-05, "loss": 0.6676, "step": 113110 }, { "epoch": 2.64, "learning_rate": 1.0274635657459342e-05, "loss": 0.046, "step": 113115 }, { "epoch": 2.64, "learning_rate": 1.0273755603013306e-05, "loss": 0.1554, "step": 113120 }, { "epoch": 2.64, "learning_rate": 1.0272875548567273e-05, "loss": 0.213, "step": 113125 }, { "epoch": 2.64, "learning_rate": 1.0271995494121236e-05, "loss": 0.0932, "step": 113130 }, { "epoch": 2.64, "learning_rate": 1.0271115439675203e-05, "loss": 0.0155, "step": 113135 }, { "epoch": 2.64, "learning_rate": 1.0270235385229166e-05, "loss": 0.0347, "step": 113140 }, { "epoch": 2.64, "learning_rate": 1.0269355330783131e-05, "loss": 0.1262, "step": 113145 }, { "epoch": 2.64, "learning_rate": 1.0268475276337096e-05, "loss": 0.111, "step": 113150 }, { "epoch": 2.64, "learning_rate": 1.0267595221891061e-05, "loss": 0.0972, "step": 113155 }, { "epoch": 2.64, "learning_rate": 1.0266715167445026e-05, "loss": 0.06, "step": 113160 }, { "epoch": 2.64, "learning_rate": 1.0265835112998991e-05, "loss": 0.208, "step": 113165 }, { "epoch": 2.64, "learning_rate": 1.0264955058552956e-05, "loss": 0.2155, "step": 113170 }, { "epoch": 2.64, "learning_rate": 1.0264075004106921e-05, "loss": 0.2556, "step": 113175 }, { "epoch": 2.64, "learning_rate": 1.0263194949660884e-05, "loss": 0.0081, "step": 113180 }, { "epoch": 2.64, "learning_rate": 1.0262314895214851e-05, "loss": 0.089, "step": 113185 }, { "epoch": 2.64, "learning_rate": 1.0261434840768814e-05, "loss": 0.0217, "step": 113190 }, { "epoch": 2.64, "learning_rate": 1.0260554786322781e-05, "loss": 0.0241, "step": 113195 }, { "epoch": 2.64, "learning_rate": 1.0259674731876746e-05, "loss": 0.1495, "step": 113200 }, { "epoch": 2.64, "learning_rate": 1.0258794677430711e-05, "loss": 0.0923, "step": 113205 }, { "epoch": 2.64, "learning_rate": 1.0257914622984676e-05, "loss": 0.0708, "step": 113210 }, { "epoch": 2.64, "learning_rate": 1.0257034568538641e-05, "loss": 0.1026, "step": 113215 }, { "epoch": 2.64, "learning_rate": 1.0256154514092606e-05, "loss": 0.1876, "step": 113220 }, { "epoch": 2.64, "learning_rate": 1.025527445964657e-05, "loss": 0.1356, "step": 113225 }, { "epoch": 2.64, "learning_rate": 1.0254394405200536e-05, "loss": 0.0496, "step": 113230 }, { "epoch": 2.64, "learning_rate": 1.02535143507545e-05, "loss": 0.0496, "step": 113235 }, { "epoch": 2.64, "learning_rate": 1.0252634296308466e-05, "loss": 0.0574, "step": 113240 }, { "epoch": 2.64, "learning_rate": 1.025175424186243e-05, "loss": 0.1532, "step": 113245 }, { "epoch": 2.64, "learning_rate": 1.0250874187416396e-05, "loss": 0.1187, "step": 113250 }, { "epoch": 2.64, "learning_rate": 1.024999413297036e-05, "loss": 0.0641, "step": 113255 }, { "epoch": 2.64, "learning_rate": 1.0249114078524325e-05, "loss": 0.0605, "step": 113260 }, { "epoch": 2.64, "learning_rate": 1.024823402407829e-05, "loss": 0.3079, "step": 113265 }, { "epoch": 2.64, "learning_rate": 1.0247353969632255e-05, "loss": 0.0951, "step": 113270 }, { "epoch": 2.64, "learning_rate": 1.024647391518622e-05, "loss": 0.7453, "step": 113275 }, { "epoch": 2.64, "learning_rate": 1.0245593860740185e-05, "loss": 0.066, "step": 113280 }, { "epoch": 2.64, "learning_rate": 1.024471380629415e-05, "loss": 0.0091, "step": 113285 }, { "epoch": 2.64, "learning_rate": 1.0243833751848115e-05, "loss": 0.0465, "step": 113290 }, { "epoch": 2.64, "learning_rate": 1.0242953697402078e-05, "loss": 0.0604, "step": 113295 }, { "epoch": 2.64, "learning_rate": 1.0242073642956045e-05, "loss": 0.0041, "step": 113300 }, { "epoch": 2.64, "learning_rate": 1.0241193588510008e-05, "loss": 0.0635, "step": 113305 }, { "epoch": 2.64, "learning_rate": 1.0240313534063975e-05, "loss": 0.1366, "step": 113310 }, { "epoch": 2.64, "learning_rate": 1.0239433479617938e-05, "loss": 0.1332, "step": 113315 }, { "epoch": 2.64, "learning_rate": 1.0238553425171905e-05, "loss": 0.3135, "step": 113320 }, { "epoch": 2.64, "learning_rate": 1.0237673370725868e-05, "loss": 0.2806, "step": 113325 }, { "epoch": 2.64, "learning_rate": 1.0236793316279833e-05, "loss": 0.0115, "step": 113330 }, { "epoch": 2.64, "learning_rate": 1.0235913261833798e-05, "loss": 0.0279, "step": 113335 }, { "epoch": 2.64, "learning_rate": 1.0235033207387763e-05, "loss": 0.0497, "step": 113340 }, { "epoch": 2.64, "learning_rate": 1.023415315294173e-05, "loss": 0.0403, "step": 113345 }, { "epoch": 2.64, "learning_rate": 1.0233273098495693e-05, "loss": 0.0239, "step": 113350 }, { "epoch": 2.64, "learning_rate": 1.023239304404966e-05, "loss": 0.2004, "step": 113355 }, { "epoch": 2.64, "learning_rate": 1.0231512989603623e-05, "loss": 0.1466, "step": 113360 }, { "epoch": 2.64, "learning_rate": 1.023063293515759e-05, "loss": 0.1935, "step": 113365 }, { "epoch": 2.65, "learning_rate": 1.0229752880711553e-05, "loss": 0.2513, "step": 113370 }, { "epoch": 2.65, "learning_rate": 1.0228872826265518e-05, "loss": 0.189, "step": 113375 }, { "epoch": 2.65, "learning_rate": 1.0227992771819484e-05, "loss": 0.0198, "step": 113380 }, { "epoch": 2.65, "learning_rate": 1.0227112717373449e-05, "loss": 0.001, "step": 113385 }, { "epoch": 2.65, "learning_rate": 1.0226232662927414e-05, "loss": 0.0307, "step": 113390 }, { "epoch": 2.65, "learning_rate": 1.0225352608481379e-05, "loss": 0.062, "step": 113395 }, { "epoch": 2.65, "learning_rate": 1.0224472554035344e-05, "loss": 0.1059, "step": 113400 }, { "epoch": 2.65, "learning_rate": 1.0223592499589309e-05, "loss": 0.0303, "step": 113405 }, { "epoch": 2.65, "learning_rate": 1.0222712445143272e-05, "loss": 0.0236, "step": 113410 }, { "epoch": 2.65, "learning_rate": 1.0221832390697239e-05, "loss": 0.1359, "step": 113415 }, { "epoch": 2.65, "learning_rate": 1.0220952336251202e-05, "loss": 0.2147, "step": 113420 }, { "epoch": 2.65, "learning_rate": 1.0220072281805169e-05, "loss": 0.3334, "step": 113425 }, { "epoch": 2.65, "learning_rate": 1.0219192227359132e-05, "loss": 0.0313, "step": 113430 }, { "epoch": 2.65, "learning_rate": 1.0218312172913099e-05, "loss": 0.0292, "step": 113435 }, { "epoch": 2.65, "learning_rate": 1.0217432118467062e-05, "loss": 0.0823, "step": 113440 }, { "epoch": 2.65, "learning_rate": 1.0216552064021027e-05, "loss": 0.0769, "step": 113445 }, { "epoch": 2.65, "learning_rate": 1.0215672009574992e-05, "loss": 0.0551, "step": 113450 }, { "epoch": 2.65, "learning_rate": 1.0214791955128957e-05, "loss": 0.1696, "step": 113455 }, { "epoch": 2.65, "learning_rate": 1.0213911900682922e-05, "loss": 0.1524, "step": 113460 }, { "epoch": 2.65, "learning_rate": 1.0213031846236887e-05, "loss": 0.0909, "step": 113465 }, { "epoch": 2.65, "learning_rate": 1.0212151791790852e-05, "loss": 0.2594, "step": 113470 }, { "epoch": 2.65, "learning_rate": 1.0211271737344817e-05, "loss": 0.1718, "step": 113475 }, { "epoch": 2.65, "learning_rate": 1.0210391682898782e-05, "loss": 0.1161, "step": 113480 }, { "epoch": 2.65, "learning_rate": 1.0209511628452747e-05, "loss": 0.0146, "step": 113485 }, { "epoch": 2.65, "learning_rate": 1.0208631574006712e-05, "loss": 0.0234, "step": 113490 }, { "epoch": 2.65, "learning_rate": 1.0207751519560677e-05, "loss": 0.0328, "step": 113495 }, { "epoch": 2.65, "learning_rate": 1.0206871465114642e-05, "loss": 0.0859, "step": 113500 }, { "epoch": 2.65, "learning_rate": 1.0205991410668607e-05, "loss": 0.1274, "step": 113505 }, { "epoch": 2.65, "learning_rate": 1.0205111356222572e-05, "loss": 0.1326, "step": 113510 }, { "epoch": 2.65, "learning_rate": 1.0204231301776537e-05, "loss": 0.1476, "step": 113515 }, { "epoch": 2.65, "learning_rate": 1.0203351247330502e-05, "loss": 0.1933, "step": 113520 }, { "epoch": 2.65, "learning_rate": 1.0202471192884466e-05, "loss": 0.314, "step": 113525 }, { "epoch": 2.65, "learning_rate": 1.0201591138438432e-05, "loss": 0.0014, "step": 113530 }, { "epoch": 2.65, "learning_rate": 1.0200711083992396e-05, "loss": 0.0018, "step": 113535 }, { "epoch": 2.65, "learning_rate": 1.0199831029546362e-05, "loss": 0.0129, "step": 113540 }, { "epoch": 2.65, "learning_rate": 1.0198950975100326e-05, "loss": 0.0748, "step": 113545 }, { "epoch": 2.65, "learning_rate": 1.0198070920654293e-05, "loss": 0.0029, "step": 113550 }, { "epoch": 2.65, "learning_rate": 1.0197190866208256e-05, "loss": 0.0994, "step": 113555 }, { "epoch": 2.65, "learning_rate": 1.0196310811762221e-05, "loss": 0.1216, "step": 113560 }, { "epoch": 2.65, "learning_rate": 1.0195430757316186e-05, "loss": 0.2097, "step": 113565 }, { "epoch": 2.65, "learning_rate": 1.0194550702870151e-05, "loss": 0.1199, "step": 113570 }, { "epoch": 2.65, "learning_rate": 1.0193670648424116e-05, "loss": 0.1201, "step": 113575 }, { "epoch": 2.65, "learning_rate": 1.0192790593978081e-05, "loss": 0.0712, "step": 113580 }, { "epoch": 2.65, "learning_rate": 1.0191910539532046e-05, "loss": 0.0582, "step": 113585 }, { "epoch": 2.65, "learning_rate": 1.0191030485086011e-05, "loss": 0.0326, "step": 113590 }, { "epoch": 2.65, "learning_rate": 1.0190150430639974e-05, "loss": 0.0047, "step": 113595 }, { "epoch": 2.65, "learning_rate": 1.0189270376193941e-05, "loss": 0.0815, "step": 113600 }, { "epoch": 2.65, "learning_rate": 1.0188390321747906e-05, "loss": 0.0108, "step": 113605 }, { "epoch": 2.65, "learning_rate": 1.0187510267301871e-05, "loss": 0.0263, "step": 113610 }, { "epoch": 2.65, "learning_rate": 1.0186630212855836e-05, "loss": 0.1389, "step": 113615 }, { "epoch": 2.65, "learning_rate": 1.0185750158409801e-05, "loss": 0.1763, "step": 113620 }, { "epoch": 2.65, "learning_rate": 1.0184870103963766e-05, "loss": 0.38, "step": 113625 }, { "epoch": 2.65, "learning_rate": 1.018399004951773e-05, "loss": 0.0002, "step": 113630 }, { "epoch": 2.65, "learning_rate": 1.0183109995071696e-05, "loss": 0.0018, "step": 113635 }, { "epoch": 2.65, "learning_rate": 1.018222994062566e-05, "loss": 0.0068, "step": 113640 }, { "epoch": 2.65, "learning_rate": 1.0181349886179626e-05, "loss": 0.0137, "step": 113645 }, { "epoch": 2.65, "learning_rate": 1.018046983173359e-05, "loss": 0.0554, "step": 113650 }, { "epoch": 2.65, "learning_rate": 1.0179589777287556e-05, "loss": 0.0679, "step": 113655 }, { "epoch": 2.65, "learning_rate": 1.017870972284152e-05, "loss": 0.1238, "step": 113660 }, { "epoch": 2.65, "learning_rate": 1.0177829668395486e-05, "loss": 0.1914, "step": 113665 }, { "epoch": 2.65, "learning_rate": 1.017694961394945e-05, "loss": 0.1599, "step": 113670 }, { "epoch": 2.65, "learning_rate": 1.0176069559503415e-05, "loss": 0.2527, "step": 113675 }, { "epoch": 2.65, "learning_rate": 1.017518950505738e-05, "loss": 0.0382, "step": 113680 }, { "epoch": 2.65, "learning_rate": 1.0174309450611345e-05, "loss": 0.0365, "step": 113685 }, { "epoch": 2.65, "learning_rate": 1.017342939616531e-05, "loss": 0.0195, "step": 113690 }, { "epoch": 2.65, "learning_rate": 1.0172549341719275e-05, "loss": 0.2422, "step": 113695 }, { "epoch": 2.65, "learning_rate": 1.017166928727324e-05, "loss": 0.0082, "step": 113700 }, { "epoch": 2.65, "learning_rate": 1.0170789232827205e-05, "loss": 0.1099, "step": 113705 }, { "epoch": 2.65, "learning_rate": 1.0169909178381168e-05, "loss": 0.0911, "step": 113710 }, { "epoch": 2.65, "learning_rate": 1.0169029123935135e-05, "loss": 0.0453, "step": 113715 }, { "epoch": 2.65, "learning_rate": 1.0168149069489098e-05, "loss": 0.5169, "step": 113720 }, { "epoch": 2.65, "learning_rate": 1.0167269015043065e-05, "loss": 0.3096, "step": 113725 }, { "epoch": 2.65, "learning_rate": 1.0166388960597028e-05, "loss": 0.046, "step": 113730 }, { "epoch": 2.65, "learning_rate": 1.0165508906150995e-05, "loss": 0.0318, "step": 113735 }, { "epoch": 2.65, "learning_rate": 1.0164628851704958e-05, "loss": 0.0307, "step": 113740 }, { "epoch": 2.65, "learning_rate": 1.0163748797258923e-05, "loss": 0.0085, "step": 113745 }, { "epoch": 2.65, "learning_rate": 1.016286874281289e-05, "loss": 0.1028, "step": 113750 }, { "epoch": 2.65, "learning_rate": 1.0161988688366853e-05, "loss": 0.0557, "step": 113755 }, { "epoch": 2.65, "learning_rate": 1.016110863392082e-05, "loss": 0.186, "step": 113760 }, { "epoch": 2.65, "learning_rate": 1.0160228579474783e-05, "loss": 0.1356, "step": 113765 }, { "epoch": 2.65, "learning_rate": 1.015934852502875e-05, "loss": 0.165, "step": 113770 }, { "epoch": 2.65, "learning_rate": 1.0158468470582713e-05, "loss": 0.1384, "step": 113775 }, { "epoch": 2.65, "learning_rate": 1.0157588416136678e-05, "loss": 0.1259, "step": 113780 }, { "epoch": 2.65, "learning_rate": 1.0156708361690643e-05, "loss": 0.0143, "step": 113785 }, { "epoch": 2.65, "learning_rate": 1.0155828307244608e-05, "loss": 0.0878, "step": 113790 }, { "epoch": 2.65, "learning_rate": 1.0154948252798573e-05, "loss": 0.0342, "step": 113795 }, { "epoch": 2.66, "learning_rate": 1.0154068198352538e-05, "loss": 0.0855, "step": 113800 }, { "epoch": 2.66, "learning_rate": 1.0153188143906503e-05, "loss": 0.0982, "step": 113805 }, { "epoch": 2.66, "learning_rate": 1.0152308089460468e-05, "loss": 0.0228, "step": 113810 }, { "epoch": 2.66, "learning_rate": 1.0151428035014434e-05, "loss": 0.1426, "step": 113815 }, { "epoch": 2.66, "learning_rate": 1.0150547980568399e-05, "loss": 0.1603, "step": 113820 }, { "epoch": 2.66, "learning_rate": 1.0149667926122362e-05, "loss": 0.2119, "step": 113825 }, { "epoch": 2.66, "learning_rate": 1.0148787871676329e-05, "loss": 0.0585, "step": 113830 }, { "epoch": 2.66, "learning_rate": 1.0147907817230292e-05, "loss": 0.0006, "step": 113835 }, { "epoch": 2.66, "learning_rate": 1.0147027762784259e-05, "loss": 0.0085, "step": 113840 }, { "epoch": 2.66, "learning_rate": 1.0146147708338222e-05, "loss": 0.0702, "step": 113845 }, { "epoch": 2.66, "learning_rate": 1.0145267653892189e-05, "loss": 0.1131, "step": 113850 }, { "epoch": 2.66, "learning_rate": 1.0144387599446152e-05, "loss": 0.1605, "step": 113855 }, { "epoch": 2.66, "learning_rate": 1.0143507545000117e-05, "loss": 0.1411, "step": 113860 }, { "epoch": 2.66, "learning_rate": 1.0142627490554082e-05, "loss": 0.1543, "step": 113865 }, { "epoch": 2.66, "learning_rate": 1.0141747436108047e-05, "loss": 0.4197, "step": 113870 }, { "epoch": 2.66, "learning_rate": 1.0140867381662012e-05, "loss": 0.2484, "step": 113875 }, { "epoch": 2.66, "learning_rate": 1.0139987327215977e-05, "loss": 0.0292, "step": 113880 }, { "epoch": 2.66, "learning_rate": 1.0139107272769944e-05, "loss": 0.0208, "step": 113885 }, { "epoch": 2.66, "learning_rate": 1.0138227218323907e-05, "loss": 0.0609, "step": 113890 }, { "epoch": 2.66, "learning_rate": 1.0137347163877872e-05, "loss": 0.0344, "step": 113895 }, { "epoch": 2.66, "learning_rate": 1.0136467109431837e-05, "loss": 0.087, "step": 113900 }, { "epoch": 2.66, "learning_rate": 1.0135587054985802e-05, "loss": 0.0663, "step": 113905 }, { "epoch": 2.66, "learning_rate": 1.0134707000539767e-05, "loss": 0.0974, "step": 113910 }, { "epoch": 2.66, "learning_rate": 1.0133826946093732e-05, "loss": 0.091, "step": 113915 }, { "epoch": 2.66, "learning_rate": 1.0132946891647697e-05, "loss": 0.1781, "step": 113920 }, { "epoch": 2.66, "learning_rate": 1.0132066837201662e-05, "loss": 0.3641, "step": 113925 }, { "epoch": 2.66, "learning_rate": 1.0131186782755626e-05, "loss": 0.02, "step": 113930 }, { "epoch": 2.66, "learning_rate": 1.0130306728309592e-05, "loss": 0.0217, "step": 113935 }, { "epoch": 2.66, "learning_rate": 1.0129426673863556e-05, "loss": 0.0028, "step": 113940 }, { "epoch": 2.66, "learning_rate": 1.0128546619417522e-05, "loss": 0.0482, "step": 113945 }, { "epoch": 2.66, "learning_rate": 1.0127666564971486e-05, "loss": 0.0011, "step": 113950 }, { "epoch": 2.66, "learning_rate": 1.0126786510525452e-05, "loss": 0.1013, "step": 113955 }, { "epoch": 2.66, "learning_rate": 1.0125906456079416e-05, "loss": 0.1406, "step": 113960 }, { "epoch": 2.66, "learning_rate": 1.0125026401633382e-05, "loss": 0.1283, "step": 113965 }, { "epoch": 2.66, "learning_rate": 1.0124146347187346e-05, "loss": 0.3883, "step": 113970 }, { "epoch": 2.66, "learning_rate": 1.012326629274131e-05, "loss": 0.1877, "step": 113975 }, { "epoch": 2.66, "learning_rate": 1.0122386238295276e-05, "loss": 0.0014, "step": 113980 }, { "epoch": 2.66, "learning_rate": 1.012150618384924e-05, "loss": 0.0311, "step": 113985 }, { "epoch": 2.66, "learning_rate": 1.0120626129403206e-05, "loss": 0.0414, "step": 113990 }, { "epoch": 2.66, "learning_rate": 1.0119746074957171e-05, "loss": 0.0591, "step": 113995 }, { "epoch": 2.66, "learning_rate": 1.0118866020511136e-05, "loss": 0.0169, "step": 114000 }, { "epoch": 2.66, "learning_rate": 1.0117985966065101e-05, "loss": 0.0541, "step": 114005 }, { "epoch": 2.66, "learning_rate": 1.0117105911619064e-05, "loss": 0.0474, "step": 114010 }, { "epoch": 2.66, "learning_rate": 1.0116225857173031e-05, "loss": 0.1073, "step": 114015 }, { "epoch": 2.66, "learning_rate": 1.0115345802726996e-05, "loss": 0.1562, "step": 114020 }, { "epoch": 2.66, "learning_rate": 1.0114465748280961e-05, "loss": 0.3442, "step": 114025 }, { "epoch": 2.66, "learning_rate": 1.0113585693834926e-05, "loss": 0.0146, "step": 114030 }, { "epoch": 2.66, "learning_rate": 1.0112705639388891e-05, "loss": 0.035, "step": 114035 }, { "epoch": 2.66, "learning_rate": 1.0111825584942856e-05, "loss": 0.0386, "step": 114040 }, { "epoch": 2.66, "learning_rate": 1.011094553049682e-05, "loss": 0.04, "step": 114045 }, { "epoch": 2.66, "learning_rate": 1.0110065476050786e-05, "loss": 0.1151, "step": 114050 }, { "epoch": 2.66, "learning_rate": 1.010918542160475e-05, "loss": 0.0378, "step": 114055 }, { "epoch": 2.66, "learning_rate": 1.0108305367158716e-05, "loss": 0.1195, "step": 114060 }, { "epoch": 2.66, "learning_rate": 1.010742531271268e-05, "loss": 0.1865, "step": 114065 }, { "epoch": 2.66, "learning_rate": 1.0106545258266646e-05, "loss": 0.1176, "step": 114070 }, { "epoch": 2.66, "learning_rate": 1.010566520382061e-05, "loss": 0.1474, "step": 114075 }, { "epoch": 2.66, "learning_rate": 1.0104785149374575e-05, "loss": 0.0268, "step": 114080 }, { "epoch": 2.66, "learning_rate": 1.010390509492854e-05, "loss": 0.1012, "step": 114085 }, { "epoch": 2.66, "learning_rate": 1.0103025040482505e-05, "loss": 0.0324, "step": 114090 }, { "epoch": 2.66, "learning_rate": 1.010214498603647e-05, "loss": 0.0493, "step": 114095 }, { "epoch": 2.66, "learning_rate": 1.0101264931590435e-05, "loss": 0.0335, "step": 114100 }, { "epoch": 2.66, "learning_rate": 1.01003848771444e-05, "loss": 0.0569, "step": 114105 }, { "epoch": 2.66, "learning_rate": 1.0099504822698365e-05, "loss": 0.1005, "step": 114110 }, { "epoch": 2.66, "learning_rate": 1.009862476825233e-05, "loss": 0.103, "step": 114115 }, { "epoch": 2.66, "learning_rate": 1.0097744713806295e-05, "loss": 0.1912, "step": 114120 }, { "epoch": 2.66, "learning_rate": 1.0096864659360258e-05, "loss": 0.2122, "step": 114125 }, { "epoch": 2.66, "learning_rate": 1.0095984604914225e-05, "loss": 0.1274, "step": 114130 }, { "epoch": 2.66, "learning_rate": 1.0095104550468188e-05, "loss": 0.133, "step": 114135 }, { "epoch": 2.66, "learning_rate": 1.0094224496022155e-05, "loss": 0.0569, "step": 114140 }, { "epoch": 2.66, "learning_rate": 1.0093344441576118e-05, "loss": 0.0793, "step": 114145 }, { "epoch": 2.66, "learning_rate": 1.0092464387130085e-05, "loss": 0.0232, "step": 114150 }, { "epoch": 2.66, "learning_rate": 1.009158433268405e-05, "loss": 0.0639, "step": 114155 }, { "epoch": 2.66, "learning_rate": 1.0090704278238013e-05, "loss": 0.2163, "step": 114160 }, { "epoch": 2.66, "learning_rate": 1.008982422379198e-05, "loss": 0.0343, "step": 114165 }, { "epoch": 2.66, "learning_rate": 1.0088944169345943e-05, "loss": 0.2682, "step": 114170 }, { "epoch": 2.66, "learning_rate": 1.008806411489991e-05, "loss": 0.3564, "step": 114175 }, { "epoch": 2.66, "learning_rate": 1.0087184060453873e-05, "loss": 0.0372, "step": 114180 }, { "epoch": 2.66, "learning_rate": 1.008630400600784e-05, "loss": 0.0063, "step": 114185 }, { "epoch": 2.66, "learning_rate": 1.0085423951561803e-05, "loss": 0.0572, "step": 114190 }, { "epoch": 2.66, "learning_rate": 1.0084543897115768e-05, "loss": 0.0362, "step": 114195 }, { "epoch": 2.66, "learning_rate": 1.0083663842669733e-05, "loss": 0.0195, "step": 114200 }, { "epoch": 2.66, "learning_rate": 1.0082783788223698e-05, "loss": 0.0295, "step": 114205 }, { "epoch": 2.66, "learning_rate": 1.0081903733777663e-05, "loss": 0.0743, "step": 114210 }, { "epoch": 2.66, "learning_rate": 1.0081023679331628e-05, "loss": 0.0851, "step": 114215 }, { "epoch": 2.66, "learning_rate": 1.0080143624885593e-05, "loss": 0.2067, "step": 114220 }, { "epoch": 2.67, "learning_rate": 1.0079263570439558e-05, "loss": 0.3076, "step": 114225 }, { "epoch": 2.67, "learning_rate": 1.0078383515993522e-05, "loss": 0.0269, "step": 114230 }, { "epoch": 2.67, "learning_rate": 1.0077503461547488e-05, "loss": 0.0126, "step": 114235 }, { "epoch": 2.67, "learning_rate": 1.0076623407101452e-05, "loss": 0.0513, "step": 114240 }, { "epoch": 2.67, "learning_rate": 1.0075743352655418e-05, "loss": 0.0692, "step": 114245 }, { "epoch": 2.67, "learning_rate": 1.0074863298209382e-05, "loss": 0.0678, "step": 114250 }, { "epoch": 2.67, "learning_rate": 1.0073983243763349e-05, "loss": 0.2256, "step": 114255 }, { "epoch": 2.67, "learning_rate": 1.0073103189317312e-05, "loss": 0.0845, "step": 114260 }, { "epoch": 2.67, "learning_rate": 1.0072223134871279e-05, "loss": 0.263, "step": 114265 }, { "epoch": 2.67, "learning_rate": 1.0071343080425242e-05, "loss": 0.0848, "step": 114270 }, { "epoch": 2.67, "learning_rate": 1.0070463025979207e-05, "loss": 0.4222, "step": 114275 }, { "epoch": 2.67, "learning_rate": 1.0069582971533172e-05, "loss": 0.0395, "step": 114280 }, { "epoch": 2.67, "learning_rate": 1.0068702917087137e-05, "loss": 0.0028, "step": 114285 }, { "epoch": 2.67, "learning_rate": 1.0067822862641104e-05, "loss": 0.0516, "step": 114290 }, { "epoch": 2.67, "learning_rate": 1.0066942808195067e-05, "loss": 0.0139, "step": 114295 }, { "epoch": 2.67, "learning_rate": 1.0066062753749034e-05, "loss": 0.0054, "step": 114300 }, { "epoch": 2.67, "learning_rate": 1.0065182699302997e-05, "loss": 0.1327, "step": 114305 }, { "epoch": 2.67, "learning_rate": 1.0064302644856962e-05, "loss": 0.1912, "step": 114310 }, { "epoch": 2.67, "learning_rate": 1.0063422590410927e-05, "loss": 0.0794, "step": 114315 }, { "epoch": 2.67, "learning_rate": 1.0062542535964892e-05, "loss": 0.1281, "step": 114320 }, { "epoch": 2.67, "learning_rate": 1.0061662481518857e-05, "loss": 0.1391, "step": 114325 }, { "epoch": 2.67, "learning_rate": 1.0060782427072822e-05, "loss": 0.0395, "step": 114330 }, { "epoch": 2.67, "learning_rate": 1.0059902372626787e-05, "loss": 0.0276, "step": 114335 }, { "epoch": 2.67, "learning_rate": 1.0059022318180752e-05, "loss": 0.0025, "step": 114340 }, { "epoch": 2.67, "learning_rate": 1.0058142263734716e-05, "loss": 0.1202, "step": 114345 }, { "epoch": 2.67, "learning_rate": 1.0057262209288682e-05, "loss": 0.0624, "step": 114350 }, { "epoch": 2.67, "learning_rate": 1.0056382154842646e-05, "loss": 0.0495, "step": 114355 }, { "epoch": 2.67, "learning_rate": 1.0055502100396612e-05, "loss": 0.0489, "step": 114360 }, { "epoch": 2.67, "learning_rate": 1.0054622045950576e-05, "loss": 0.2189, "step": 114365 }, { "epoch": 2.67, "learning_rate": 1.0053741991504542e-05, "loss": 0.1774, "step": 114370 }, { "epoch": 2.67, "learning_rate": 1.0052861937058506e-05, "loss": 0.2155, "step": 114375 }, { "epoch": 2.67, "learning_rate": 1.005198188261247e-05, "loss": 0.0371, "step": 114380 }, { "epoch": 2.67, "learning_rate": 1.0051101828166436e-05, "loss": 0.1562, "step": 114385 }, { "epoch": 2.67, "learning_rate": 1.00502217737204e-05, "loss": 0.0404, "step": 114390 }, { "epoch": 2.67, "learning_rate": 1.0049341719274366e-05, "loss": 0.0059, "step": 114395 }, { "epoch": 2.67, "learning_rate": 1.004846166482833e-05, "loss": 0.0501, "step": 114400 }, { "epoch": 2.67, "learning_rate": 1.0047581610382296e-05, "loss": 0.0694, "step": 114405 }, { "epoch": 2.67, "learning_rate": 1.004670155593626e-05, "loss": 0.3182, "step": 114410 }, { "epoch": 2.67, "learning_rate": 1.0045821501490226e-05, "loss": 0.1239, "step": 114415 }, { "epoch": 2.67, "learning_rate": 1.004494144704419e-05, "loss": 0.2217, "step": 114420 }, { "epoch": 2.67, "learning_rate": 1.0044061392598156e-05, "loss": 0.2233, "step": 114425 }, { "epoch": 2.67, "learning_rate": 1.0043181338152121e-05, "loss": 0.0576, "step": 114430 }, { "epoch": 2.67, "learning_rate": 1.0042301283706086e-05, "loss": 0.0019, "step": 114435 }, { "epoch": 2.67, "learning_rate": 1.0041421229260051e-05, "loss": 0.1058, "step": 114440 }, { "epoch": 2.67, "learning_rate": 1.0040541174814016e-05, "loss": 0.0184, "step": 114445 }, { "epoch": 2.67, "learning_rate": 1.0039661120367981e-05, "loss": 0.0357, "step": 114450 }, { "epoch": 2.67, "learning_rate": 1.0038781065921946e-05, "loss": 0.1033, "step": 114455 }, { "epoch": 2.67, "learning_rate": 1.003790101147591e-05, "loss": 0.0828, "step": 114460 }, { "epoch": 2.67, "learning_rate": 1.0037020957029876e-05, "loss": 0.0948, "step": 114465 }, { "epoch": 2.67, "learning_rate": 1.003614090258384e-05, "loss": 0.1595, "step": 114470 }, { "epoch": 2.67, "learning_rate": 1.0035260848137806e-05, "loss": 0.2325, "step": 114475 }, { "epoch": 2.67, "learning_rate": 1.003438079369177e-05, "loss": 0.0011, "step": 114480 }, { "epoch": 2.67, "learning_rate": 1.0033500739245736e-05, "loss": 0.0224, "step": 114485 }, { "epoch": 2.67, "learning_rate": 1.00326206847997e-05, "loss": 0.1151, "step": 114490 }, { "epoch": 2.67, "learning_rate": 1.0031740630353664e-05, "loss": 0.0153, "step": 114495 }, { "epoch": 2.67, "learning_rate": 1.003086057590763e-05, "loss": 0.1015, "step": 114500 }, { "epoch": 2.67, "learning_rate": 1.0029980521461594e-05, "loss": 0.0461, "step": 114505 }, { "epoch": 2.67, "learning_rate": 1.002910046701556e-05, "loss": 0.0715, "step": 114510 }, { "epoch": 2.67, "learning_rate": 1.0028220412569525e-05, "loss": 0.2052, "step": 114515 }, { "epoch": 2.67, "learning_rate": 1.002734035812349e-05, "loss": 0.1258, "step": 114520 }, { "epoch": 2.67, "learning_rate": 1.0026460303677455e-05, "loss": 0.1936, "step": 114525 }, { "epoch": 2.67, "learning_rate": 1.0025580249231418e-05, "loss": 0.1024, "step": 114530 }, { "epoch": 2.67, "learning_rate": 1.0024700194785385e-05, "loss": 0.0233, "step": 114535 }, { "epoch": 2.67, "learning_rate": 1.0023820140339348e-05, "loss": 0.0059, "step": 114540 }, { "epoch": 2.67, "learning_rate": 1.0022940085893315e-05, "loss": 0.0725, "step": 114545 }, { "epoch": 2.67, "learning_rate": 1.0022060031447278e-05, "loss": 0.0689, "step": 114550 }, { "epoch": 2.67, "learning_rate": 1.0021179977001245e-05, "loss": 0.1212, "step": 114555 }, { "epoch": 2.67, "learning_rate": 1.002029992255521e-05, "loss": 0.1205, "step": 114560 }, { "epoch": 2.67, "learning_rate": 1.0019419868109175e-05, "loss": 0.2221, "step": 114565 }, { "epoch": 2.67, "learning_rate": 1.001853981366314e-05, "loss": 0.2216, "step": 114570 }, { "epoch": 2.67, "learning_rate": 1.0017659759217103e-05, "loss": 0.4853, "step": 114575 }, { "epoch": 2.67, "learning_rate": 1.001677970477107e-05, "loss": 0.0379, "step": 114580 }, { "epoch": 2.67, "learning_rate": 1.0015899650325033e-05, "loss": 0.0058, "step": 114585 }, { "epoch": 2.67, "learning_rate": 1.0015019595879e-05, "loss": 0.0653, "step": 114590 }, { "epoch": 2.67, "learning_rate": 1.0014139541432963e-05, "loss": 0.0119, "step": 114595 }, { "epoch": 2.67, "learning_rate": 1.001325948698693e-05, "loss": 0.0022, "step": 114600 }, { "epoch": 2.67, "learning_rate": 1.0012379432540893e-05, "loss": 0.1055, "step": 114605 }, { "epoch": 2.67, "learning_rate": 1.0011499378094858e-05, "loss": 0.0427, "step": 114610 }, { "epoch": 2.67, "learning_rate": 1.0010619323648823e-05, "loss": 0.1601, "step": 114615 }, { "epoch": 2.67, "learning_rate": 1.0009739269202788e-05, "loss": 0.2267, "step": 114620 }, { "epoch": 2.67, "learning_rate": 1.0008859214756753e-05, "loss": 0.1285, "step": 114625 }, { "epoch": 2.67, "learning_rate": 1.0007979160310718e-05, "loss": 0.0356, "step": 114630 }, { "epoch": 2.67, "learning_rate": 1.0007099105864683e-05, "loss": 0.021, "step": 114635 }, { "epoch": 2.67, "learning_rate": 1.0006219051418648e-05, "loss": 0.0168, "step": 114640 }, { "epoch": 2.67, "learning_rate": 1.0005338996972612e-05, "loss": 0.1323, "step": 114645 }, { "epoch": 2.67, "learning_rate": 1.0004458942526578e-05, "loss": 0.114, "step": 114650 }, { "epoch": 2.68, "learning_rate": 1.0003578888080542e-05, "loss": 0.1153, "step": 114655 }, { "epoch": 2.68, "learning_rate": 1.0002874844523715e-05, "loss": 0.0775, "step": 114660 }, { "epoch": 2.68, "learning_rate": 1.000199479007768e-05, "loss": 0.1994, "step": 114665 }, { "epoch": 2.68, "learning_rate": 1.0001114735631645e-05, "loss": 0.26, "step": 114670 }, { "epoch": 2.68, "learning_rate": 1.000023468118561e-05, "loss": 0.2919, "step": 114675 }, { "epoch": 2.68, "learning_rate": 9.999354626739575e-06, "loss": 0.0479, "step": 114680 }, { "epoch": 2.68, "learning_rate": 9.99847457229354e-06, "loss": 0.053, "step": 114685 }, { "epoch": 2.68, "learning_rate": 9.997594517847503e-06, "loss": 0.0116, "step": 114690 }, { "epoch": 2.68, "learning_rate": 9.99671446340147e-06, "loss": 0.0223, "step": 114695 }, { "epoch": 2.68, "learning_rate": 9.995834408955433e-06, "loss": 0.0981, "step": 114700 }, { "epoch": 2.68, "learning_rate": 9.9949543545094e-06, "loss": 0.067, "step": 114705 }, { "epoch": 2.68, "learning_rate": 9.994074300063363e-06, "loss": 0.037, "step": 114710 }, { "epoch": 2.68, "learning_rate": 9.99319424561733e-06, "loss": 0.1224, "step": 114715 }, { "epoch": 2.68, "learning_rate": 9.992314191171293e-06, "loss": 0.1562, "step": 114720 }, { "epoch": 2.68, "learning_rate": 9.991434136725258e-06, "loss": 0.1997, "step": 114725 }, { "epoch": 2.68, "learning_rate": 9.990554082279223e-06, "loss": 0.0284, "step": 114730 }, { "epoch": 2.68, "learning_rate": 9.989674027833188e-06, "loss": 0.049, "step": 114735 }, { "epoch": 2.68, "learning_rate": 9.988793973387153e-06, "loss": 0.0082, "step": 114740 }, { "epoch": 2.68, "learning_rate": 9.987913918941118e-06, "loss": 0.0684, "step": 114745 }, { "epoch": 2.68, "learning_rate": 9.987033864495083e-06, "loss": 0.0168, "step": 114750 }, { "epoch": 2.68, "learning_rate": 9.986153810049048e-06, "loss": 0.0647, "step": 114755 }, { "epoch": 2.68, "learning_rate": 9.985273755603015e-06, "loss": 0.1285, "step": 114760 }, { "epoch": 2.68, "learning_rate": 9.984393701156978e-06, "loss": 0.1261, "step": 114765 }, { "epoch": 2.68, "learning_rate": 9.983513646710944e-06, "loss": 0.181, "step": 114770 }, { "epoch": 2.68, "learning_rate": 9.982633592264909e-06, "loss": 0.2071, "step": 114775 }, { "epoch": 2.68, "learning_rate": 9.981753537818874e-06, "loss": 0.0013, "step": 114780 }, { "epoch": 2.68, "learning_rate": 9.980873483372839e-06, "loss": 0.0014, "step": 114785 }, { "epoch": 2.68, "learning_rate": 9.979993428926804e-06, "loss": 0.0082, "step": 114790 }, { "epoch": 2.68, "learning_rate": 9.979113374480769e-06, "loss": 0.1233, "step": 114795 }, { "epoch": 2.68, "learning_rate": 9.978233320034734e-06, "loss": 0.0584, "step": 114800 }, { "epoch": 2.68, "learning_rate": 9.977353265588697e-06, "loss": 0.0549, "step": 114805 }, { "epoch": 2.68, "learning_rate": 9.976473211142664e-06, "loss": 0.146, "step": 114810 }, { "epoch": 2.68, "learning_rate": 9.975593156696627e-06, "loss": 0.2273, "step": 114815 }, { "epoch": 2.68, "learning_rate": 9.974713102250594e-06, "loss": 0.1966, "step": 114820 }, { "epoch": 2.68, "learning_rate": 9.973833047804557e-06, "loss": 0.1529, "step": 114825 }, { "epoch": 2.68, "learning_rate": 9.972952993358524e-06, "loss": 0.1063, "step": 114830 }, { "epoch": 2.68, "learning_rate": 9.972072938912487e-06, "loss": 0.0794, "step": 114835 }, { "epoch": 2.68, "learning_rate": 9.971192884466452e-06, "loss": 0.0099, "step": 114840 }, { "epoch": 2.68, "learning_rate": 9.970312830020417e-06, "loss": 0.1067, "step": 114845 }, { "epoch": 2.68, "learning_rate": 9.969432775574382e-06, "loss": 0.1923, "step": 114850 }, { "epoch": 2.68, "learning_rate": 9.968552721128347e-06, "loss": 0.0786, "step": 114855 }, { "epoch": 2.68, "learning_rate": 9.967672666682312e-06, "loss": 0.1992, "step": 114860 }, { "epoch": 2.68, "learning_rate": 9.966792612236277e-06, "loss": 0.0717, "step": 114865 }, { "epoch": 2.68, "learning_rate": 9.965912557790242e-06, "loss": 0.1825, "step": 114870 }, { "epoch": 2.68, "learning_rate": 9.965032503344206e-06, "loss": 0.1139, "step": 114875 }, { "epoch": 2.68, "learning_rate": 9.964152448898172e-06, "loss": 0.1173, "step": 114880 }, { "epoch": 2.68, "learning_rate": 9.963272394452137e-06, "loss": 0.0339, "step": 114885 }, { "epoch": 2.68, "learning_rate": 9.962392340006102e-06, "loss": 0.0036, "step": 114890 }, { "epoch": 2.68, "learning_rate": 9.961512285560067e-06, "loss": 0.0672, "step": 114895 }, { "epoch": 2.68, "learning_rate": 9.960632231114032e-06, "loss": 0.0275, "step": 114900 }, { "epoch": 2.68, "learning_rate": 9.959752176667997e-06, "loss": 0.0845, "step": 114905 }, { "epoch": 2.68, "learning_rate": 9.958872122221962e-06, "loss": 0.2213, "step": 114910 }, { "epoch": 2.68, "learning_rate": 9.957992067775927e-06, "loss": 0.0871, "step": 114915 }, { "epoch": 2.68, "learning_rate": 9.95711201332989e-06, "loss": 0.1575, "step": 114920 }, { "epoch": 2.68, "learning_rate": 9.956231958883857e-06, "loss": 0.1876, "step": 114925 }, { "epoch": 2.68, "learning_rate": 9.95535190443782e-06, "loss": 0.0768, "step": 114930 }, { "epoch": 2.68, "learning_rate": 9.954471849991787e-06, "loss": 0.0167, "step": 114935 }, { "epoch": 2.68, "learning_rate": 9.95359179554575e-06, "loss": 0.0273, "step": 114940 }, { "epoch": 2.68, "learning_rate": 9.952711741099718e-06, "loss": 0.0621, "step": 114945 }, { "epoch": 2.68, "learning_rate": 9.951831686653681e-06, "loss": 0.1037, "step": 114950 }, { "epoch": 2.68, "learning_rate": 9.950951632207646e-06, "loss": 0.1328, "step": 114955 }, { "epoch": 2.68, "learning_rate": 9.950071577761611e-06, "loss": 0.037, "step": 114960 }, { "epoch": 2.68, "learning_rate": 9.949191523315576e-06, "loss": 0.0897, "step": 114965 }, { "epoch": 2.68, "learning_rate": 9.948311468869541e-06, "loss": 0.2504, "step": 114970 }, { "epoch": 2.68, "learning_rate": 9.947431414423506e-06, "loss": 0.1974, "step": 114975 }, { "epoch": 2.68, "learning_rate": 9.946551359977471e-06, "loss": 0.1277, "step": 114980 }, { "epoch": 2.68, "learning_rate": 9.945671305531436e-06, "loss": 0.0055, "step": 114985 }, { "epoch": 2.68, "learning_rate": 9.9447912510854e-06, "loss": 0.0112, "step": 114990 }, { "epoch": 2.68, "learning_rate": 9.943911196639366e-06, "loss": 0.1183, "step": 114995 }, { "epoch": 2.68, "learning_rate": 9.94303114219333e-06, "loss": 0.0959, "step": 115000 }, { "epoch": 2.68, "learning_rate": 9.942151087747296e-06, "loss": 0.0392, "step": 115005 }, { "epoch": 2.68, "learning_rate": 9.94127103330126e-06, "loss": 0.1378, "step": 115010 }, { "epoch": 2.68, "learning_rate": 9.940390978855226e-06, "loss": 0.2127, "step": 115015 }, { "epoch": 2.68, "learning_rate": 9.93951092440919e-06, "loss": 0.1695, "step": 115020 }, { "epoch": 2.68, "learning_rate": 9.938630869963154e-06, "loss": 0.1655, "step": 115025 }, { "epoch": 2.68, "learning_rate": 9.937750815517121e-06, "loss": 0.048, "step": 115030 }, { "epoch": 2.68, "learning_rate": 9.936870761071085e-06, "loss": 0.0764, "step": 115035 }, { "epoch": 2.68, "learning_rate": 9.935990706625051e-06, "loss": 0.0019, "step": 115040 }, { "epoch": 2.68, "learning_rate": 9.935110652179015e-06, "loss": 0.0901, "step": 115045 }, { "epoch": 2.68, "learning_rate": 9.934230597732981e-06, "loss": 0.0714, "step": 115050 }, { "epoch": 2.68, "learning_rate": 9.933350543286945e-06, "loss": 0.0911, "step": 115055 }, { "epoch": 2.68, "learning_rate": 9.932646499730118e-06, "loss": 0.0592, "step": 115060 }, { "epoch": 2.68, "learning_rate": 9.931766445284081e-06, "loss": 0.1138, "step": 115065 }, { "epoch": 2.68, "learning_rate": 9.930886390838046e-06, "loss": 0.3155, "step": 115070 }, { "epoch": 2.68, "learning_rate": 9.930006336392011e-06, "loss": 0.2235, "step": 115075 }, { "epoch": 2.68, "learning_rate": 9.929126281945976e-06, "loss": 0.0892, "step": 115080 }, { "epoch": 2.69, "learning_rate": 9.928246227499943e-06, "loss": 0.0087, "step": 115085 }, { "epoch": 2.69, "learning_rate": 9.927366173053906e-06, "loss": 0.0101, "step": 115090 }, { "epoch": 2.69, "learning_rate": 9.926486118607873e-06, "loss": 0.0014, "step": 115095 }, { "epoch": 2.69, "learning_rate": 9.925606064161836e-06, "loss": 0.0839, "step": 115100 }, { "epoch": 2.69, "learning_rate": 9.924726009715803e-06, "loss": 0.0367, "step": 115105 }, { "epoch": 2.69, "learning_rate": 9.923845955269766e-06, "loss": 0.043, "step": 115110 }, { "epoch": 2.69, "learning_rate": 9.922965900823731e-06, "loss": 0.013, "step": 115115 }, { "epoch": 2.69, "learning_rate": 9.922085846377696e-06, "loss": 0.1584, "step": 115120 }, { "epoch": 2.69, "learning_rate": 9.921205791931661e-06, "loss": 0.1264, "step": 115125 }, { "epoch": 2.69, "learning_rate": 9.920325737485626e-06, "loss": 0.0811, "step": 115130 }, { "epoch": 2.69, "learning_rate": 9.919445683039591e-06, "loss": 0.0316, "step": 115135 }, { "epoch": 2.69, "learning_rate": 9.918565628593556e-06, "loss": 0.0621, "step": 115140 }, { "epoch": 2.69, "learning_rate": 9.917685574147521e-06, "loss": 0.0057, "step": 115145 }, { "epoch": 2.69, "learning_rate": 9.916805519701485e-06, "loss": 0.0695, "step": 115150 }, { "epoch": 2.69, "learning_rate": 9.915925465255451e-06, "loss": 0.0799, "step": 115155 }, { "epoch": 2.69, "learning_rate": 9.915045410809415e-06, "loss": 0.1012, "step": 115160 }, { "epoch": 2.69, "learning_rate": 9.914165356363381e-06, "loss": 0.0516, "step": 115165 }, { "epoch": 2.69, "learning_rate": 9.913285301917345e-06, "loss": 0.1469, "step": 115170 }, { "epoch": 2.69, "learning_rate": 9.912405247471311e-06, "loss": 0.3266, "step": 115175 }, { "epoch": 2.69, "learning_rate": 9.911525193025275e-06, "loss": 0.0108, "step": 115180 }, { "epoch": 2.69, "learning_rate": 9.91064513857924e-06, "loss": 0.0284, "step": 115185 }, { "epoch": 2.69, "learning_rate": 9.909765084133205e-06, "loss": 0.055, "step": 115190 }, { "epoch": 2.69, "learning_rate": 9.90888502968717e-06, "loss": 0.0014, "step": 115195 }, { "epoch": 2.69, "learning_rate": 9.908004975241135e-06, "loss": 0.0987, "step": 115200 }, { "epoch": 2.69, "learning_rate": 9.9071249207951e-06, "loss": 0.0734, "step": 115205 }, { "epoch": 2.69, "learning_rate": 9.906244866349065e-06, "loss": 0.0615, "step": 115210 }, { "epoch": 2.69, "learning_rate": 9.90536481190303e-06, "loss": 0.0763, "step": 115215 }, { "epoch": 2.69, "learning_rate": 9.904484757456995e-06, "loss": 0.1441, "step": 115220 }, { "epoch": 2.69, "learning_rate": 9.90360470301096e-06, "loss": 0.227, "step": 115225 }, { "epoch": 2.69, "learning_rate": 9.902724648564925e-06, "loss": 0.0463, "step": 115230 }, { "epoch": 2.69, "learning_rate": 9.90184459411889e-06, "loss": 0.0305, "step": 115235 }, { "epoch": 2.69, "learning_rate": 9.900964539672855e-06, "loss": 0.0311, "step": 115240 }, { "epoch": 2.69, "learning_rate": 9.90008448522682e-06, "loss": 0.0681, "step": 115245 }, { "epoch": 2.69, "learning_rate": 9.899204430780785e-06, "loss": 0.0892, "step": 115250 }, { "epoch": 2.69, "learning_rate": 9.89832437633475e-06, "loss": 0.0764, "step": 115255 }, { "epoch": 2.69, "learning_rate": 9.897444321888715e-06, "loss": 0.1085, "step": 115260 }, { "epoch": 2.69, "learning_rate": 9.896564267442678e-06, "loss": 0.1645, "step": 115265 }, { "epoch": 2.69, "learning_rate": 9.895684212996645e-06, "loss": 0.3135, "step": 115270 }, { "epoch": 2.69, "learning_rate": 9.894804158550608e-06, "loss": 0.3653, "step": 115275 }, { "epoch": 2.69, "learning_rate": 9.893924104104575e-06, "loss": 0.0096, "step": 115280 }, { "epoch": 2.69, "learning_rate": 9.893044049658538e-06, "loss": 0.0362, "step": 115285 }, { "epoch": 2.69, "learning_rate": 9.892163995212505e-06, "loss": 0.0721, "step": 115290 }, { "epoch": 2.69, "learning_rate": 9.891283940766469e-06, "loss": 0.0485, "step": 115295 }, { "epoch": 2.69, "learning_rate": 9.890403886320434e-06, "loss": 0.0565, "step": 115300 }, { "epoch": 2.69, "learning_rate": 9.889523831874399e-06, "loss": 0.021, "step": 115305 }, { "epoch": 2.69, "learning_rate": 9.888643777428364e-06, "loss": 0.2204, "step": 115310 }, { "epoch": 2.69, "learning_rate": 9.887763722982329e-06, "loss": 0.1815, "step": 115315 }, { "epoch": 2.69, "learning_rate": 9.886883668536294e-06, "loss": 0.2258, "step": 115320 }, { "epoch": 2.69, "learning_rate": 9.886003614090259e-06, "loss": 0.1306, "step": 115325 }, { "epoch": 2.69, "learning_rate": 9.885123559644224e-06, "loss": 0.0003, "step": 115330 }, { "epoch": 2.69, "learning_rate": 9.884243505198187e-06, "loss": 0.0632, "step": 115335 }, { "epoch": 2.69, "learning_rate": 9.883363450752154e-06, "loss": 0.0003, "step": 115340 }, { "epoch": 2.69, "learning_rate": 9.882483396306117e-06, "loss": 0.0558, "step": 115345 }, { "epoch": 2.69, "learning_rate": 9.881603341860084e-06, "loss": 0.0608, "step": 115350 }, { "epoch": 2.69, "learning_rate": 9.880723287414049e-06, "loss": 0.0506, "step": 115355 }, { "epoch": 2.69, "learning_rate": 9.879843232968014e-06, "loss": 0.1566, "step": 115360 }, { "epoch": 2.69, "learning_rate": 9.878963178521979e-06, "loss": 0.1471, "step": 115365 }, { "epoch": 2.69, "learning_rate": 9.878083124075942e-06, "loss": 0.2449, "step": 115370 }, { "epoch": 2.69, "learning_rate": 9.877203069629909e-06, "loss": 0.1785, "step": 115375 }, { "epoch": 2.69, "learning_rate": 9.876323015183872e-06, "loss": 0.0994, "step": 115380 }, { "epoch": 2.69, "learning_rate": 9.875442960737839e-06, "loss": 0.0379, "step": 115385 }, { "epoch": 2.69, "learning_rate": 9.874562906291802e-06, "loss": 0.0181, "step": 115390 }, { "epoch": 2.69, "learning_rate": 9.873682851845769e-06, "loss": 0.0503, "step": 115395 }, { "epoch": 2.69, "learning_rate": 9.872802797399732e-06, "loss": 0.129, "step": 115400 }, { "epoch": 2.69, "learning_rate": 9.871922742953699e-06, "loss": 0.1041, "step": 115405 }, { "epoch": 2.69, "learning_rate": 9.871042688507662e-06, "loss": 0.1414, "step": 115410 }, { "epoch": 2.69, "learning_rate": 9.870162634061627e-06, "loss": 0.0835, "step": 115415 }, { "epoch": 2.69, "learning_rate": 9.869282579615592e-06, "loss": 0.2873, "step": 115420 }, { "epoch": 2.69, "learning_rate": 9.868402525169557e-06, "loss": 0.2337, "step": 115425 }, { "epoch": 2.69, "learning_rate": 9.867522470723522e-06, "loss": 0.0361, "step": 115430 }, { "epoch": 2.69, "learning_rate": 9.866642416277487e-06, "loss": 0.0213, "step": 115435 }, { "epoch": 2.69, "learning_rate": 9.865762361831452e-06, "loss": 0.0794, "step": 115440 }, { "epoch": 2.69, "learning_rate": 9.864882307385417e-06, "loss": 0.0032, "step": 115445 }, { "epoch": 2.69, "learning_rate": 9.86400225293938e-06, "loss": 0.1114, "step": 115450 }, { "epoch": 2.69, "learning_rate": 9.863122198493347e-06, "loss": 0.1159, "step": 115455 }, { "epoch": 2.69, "learning_rate": 9.86224214404731e-06, "loss": 0.028, "step": 115460 }, { "epoch": 2.69, "learning_rate": 9.861362089601278e-06, "loss": 0.131, "step": 115465 }, { "epoch": 2.69, "learning_rate": 9.86048203515524e-06, "loss": 0.0663, "step": 115470 }, { "epoch": 2.69, "learning_rate": 9.859601980709208e-06, "loss": 0.1795, "step": 115475 }, { "epoch": 2.69, "learning_rate": 9.858721926263171e-06, "loss": 0.0495, "step": 115480 }, { "epoch": 2.69, "learning_rate": 9.857841871817136e-06, "loss": 0.0026, "step": 115485 }, { "epoch": 2.69, "learning_rate": 9.856961817371103e-06, "loss": 0.0664, "step": 115490 }, { "epoch": 2.69, "learning_rate": 9.856081762925066e-06, "loss": 0.0436, "step": 115495 }, { "epoch": 2.69, "learning_rate": 9.855201708479033e-06, "loss": 0.0116, "step": 115500 }, { "epoch": 2.69, "learning_rate": 9.854321654032996e-06, "loss": 0.1568, "step": 115505 }, { "epoch": 2.69, "learning_rate": 9.853441599586963e-06, "loss": 0.0486, "step": 115510 }, { "epoch": 2.7, "learning_rate": 9.852561545140926e-06, "loss": 0.1555, "step": 115515 }, { "epoch": 2.7, "learning_rate": 9.851681490694891e-06, "loss": 0.1597, "step": 115520 }, { "epoch": 2.7, "learning_rate": 9.850801436248856e-06, "loss": 0.2531, "step": 115525 }, { "epoch": 2.7, "learning_rate": 9.849921381802821e-06, "loss": 0.0089, "step": 115530 }, { "epoch": 2.7, "learning_rate": 9.849041327356786e-06, "loss": 0.0986, "step": 115535 }, { "epoch": 2.7, "learning_rate": 9.848161272910751e-06, "loss": 0.0033, "step": 115540 }, { "epoch": 2.7, "learning_rate": 9.847281218464716e-06, "loss": 0.0279, "step": 115545 }, { "epoch": 2.7, "learning_rate": 9.846401164018681e-06, "loss": 0.0016, "step": 115550 }, { "epoch": 2.7, "learning_rate": 9.845521109572646e-06, "loss": 0.0264, "step": 115555 }, { "epoch": 2.7, "learning_rate": 9.844641055126611e-06, "loss": 0.0552, "step": 115560 }, { "epoch": 2.7, "learning_rate": 9.843761000680575e-06, "loss": 0.0658, "step": 115565 }, { "epoch": 2.7, "learning_rate": 9.842880946234541e-06, "loss": 0.2192, "step": 115570 }, { "epoch": 2.7, "learning_rate": 9.842000891788505e-06, "loss": 0.3526, "step": 115575 }, { "epoch": 2.7, "learning_rate": 9.841120837342471e-06, "loss": 0.0161, "step": 115580 }, { "epoch": 2.7, "learning_rate": 9.840240782896435e-06, "loss": 0.0476, "step": 115585 }, { "epoch": 2.7, "learning_rate": 9.839360728450401e-06, "loss": 0.0294, "step": 115590 }, { "epoch": 2.7, "learning_rate": 9.838480674004365e-06, "loss": 0.0203, "step": 115595 }, { "epoch": 2.7, "learning_rate": 9.83760061955833e-06, "loss": 0.0643, "step": 115600 }, { "epoch": 2.7, "learning_rate": 9.836720565112295e-06, "loss": 0.0555, "step": 115605 }, { "epoch": 2.7, "learning_rate": 9.83584051066626e-06, "loss": 0.0769, "step": 115610 }, { "epoch": 2.7, "learning_rate": 9.834960456220225e-06, "loss": 0.1225, "step": 115615 }, { "epoch": 2.7, "learning_rate": 9.83408040177419e-06, "loss": 0.1814, "step": 115620 }, { "epoch": 2.7, "learning_rate": 9.833200347328156e-06, "loss": 0.2653, "step": 115625 }, { "epoch": 2.7, "learning_rate": 9.83232029288212e-06, "loss": 0.0551, "step": 115630 }, { "epoch": 2.7, "learning_rate": 9.831440238436085e-06, "loss": 0.0031, "step": 115635 }, { "epoch": 2.7, "learning_rate": 9.83056018399005e-06, "loss": 0.0589, "step": 115640 }, { "epoch": 2.7, "learning_rate": 9.829680129544015e-06, "loss": 0.0635, "step": 115645 }, { "epoch": 2.7, "learning_rate": 9.82880007509798e-06, "loss": 0.0306, "step": 115650 }, { "epoch": 2.7, "learning_rate": 9.827920020651945e-06, "loss": 0.1057, "step": 115655 }, { "epoch": 2.7, "learning_rate": 9.82703996620591e-06, "loss": 0.1433, "step": 115660 }, { "epoch": 2.7, "learning_rate": 9.826159911759875e-06, "loss": 0.1259, "step": 115665 }, { "epoch": 2.7, "learning_rate": 9.825279857313838e-06, "loss": 0.3838, "step": 115670 }, { "epoch": 2.7, "learning_rate": 9.824399802867805e-06, "loss": 0.1574, "step": 115675 }, { "epoch": 2.7, "learning_rate": 9.823519748421768e-06, "loss": 0.0228, "step": 115680 }, { "epoch": 2.7, "learning_rate": 9.822639693975735e-06, "loss": 0.0144, "step": 115685 }, { "epoch": 2.7, "learning_rate": 9.821759639529698e-06, "loss": 0.0548, "step": 115690 }, { "epoch": 2.7, "learning_rate": 9.820879585083665e-06, "loss": 0.1125, "step": 115695 }, { "epoch": 2.7, "learning_rate": 9.819999530637628e-06, "loss": 0.0157, "step": 115700 }, { "epoch": 2.7, "learning_rate": 9.819119476191595e-06, "loss": 0.0115, "step": 115705 }, { "epoch": 2.7, "learning_rate": 9.818239421745558e-06, "loss": 0.1989, "step": 115710 }, { "epoch": 2.7, "learning_rate": 9.817359367299523e-06, "loss": 0.0246, "step": 115715 }, { "epoch": 2.7, "learning_rate": 9.816479312853488e-06, "loss": 0.2733, "step": 115720 }, { "epoch": 2.7, "learning_rate": 9.815599258407453e-06, "loss": 0.1822, "step": 115725 }, { "epoch": 2.7, "learning_rate": 9.814719203961419e-06, "loss": 0.0436, "step": 115730 }, { "epoch": 2.7, "learning_rate": 9.813839149515384e-06, "loss": 0.0243, "step": 115735 }, { "epoch": 2.7, "learning_rate": 9.812959095069349e-06, "loss": 0.0324, "step": 115740 }, { "epoch": 2.7, "learning_rate": 9.812079040623314e-06, "loss": 0.0434, "step": 115745 }, { "epoch": 2.7, "learning_rate": 9.811198986177277e-06, "loss": 0.0444, "step": 115750 }, { "epoch": 2.7, "learning_rate": 9.810318931731244e-06, "loss": 0.0929, "step": 115755 }, { "epoch": 2.7, "learning_rate": 9.809438877285209e-06, "loss": 0.0382, "step": 115760 }, { "epoch": 2.7, "learning_rate": 9.808558822839174e-06, "loss": 0.0662, "step": 115765 }, { "epoch": 2.7, "learning_rate": 9.807678768393139e-06, "loss": 0.3817, "step": 115770 }, { "epoch": 2.7, "learning_rate": 9.806798713947104e-06, "loss": 0.1992, "step": 115775 }, { "epoch": 2.7, "learning_rate": 9.805918659501069e-06, "loss": 0.0495, "step": 115780 }, { "epoch": 2.7, "learning_rate": 9.805038605055032e-06, "loss": 0.0101, "step": 115785 }, { "epoch": 2.7, "learning_rate": 9.804158550608999e-06, "loss": 0.0578, "step": 115790 }, { "epoch": 2.7, "learning_rate": 9.803278496162962e-06, "loss": 0.0374, "step": 115795 }, { "epoch": 2.7, "learning_rate": 9.802398441716929e-06, "loss": 0.0595, "step": 115800 }, { "epoch": 2.7, "learning_rate": 9.801518387270892e-06, "loss": 0.058, "step": 115805 }, { "epoch": 2.7, "learning_rate": 9.800638332824859e-06, "loss": 0.0962, "step": 115810 }, { "epoch": 2.7, "learning_rate": 9.799758278378822e-06, "loss": 0.2003, "step": 115815 }, { "epoch": 2.7, "learning_rate": 9.798878223932787e-06, "loss": 0.0897, "step": 115820 }, { "epoch": 2.7, "learning_rate": 9.797998169486752e-06, "loss": 0.3912, "step": 115825 }, { "epoch": 2.7, "learning_rate": 9.797118115040717e-06, "loss": 0.022, "step": 115830 }, { "epoch": 2.7, "learning_rate": 9.796238060594682e-06, "loss": 0.0824, "step": 115835 }, { "epoch": 2.7, "learning_rate": 9.795358006148647e-06, "loss": 0.0188, "step": 115840 }, { "epoch": 2.7, "learning_rate": 9.794477951702612e-06, "loss": 0.0191, "step": 115845 }, { "epoch": 2.7, "learning_rate": 9.793597897256577e-06, "loss": 0.0557, "step": 115850 }, { "epoch": 2.7, "learning_rate": 9.792717842810542e-06, "loss": 0.1216, "step": 115855 }, { "epoch": 2.7, "learning_rate": 9.791837788364507e-06, "loss": 0.1788, "step": 115860 }, { "epoch": 2.7, "learning_rate": 9.79095773391847e-06, "loss": 0.1503, "step": 115865 }, { "epoch": 2.7, "learning_rate": 9.790077679472437e-06, "loss": 0.2282, "step": 115870 }, { "epoch": 2.7, "learning_rate": 9.7891976250264e-06, "loss": 0.2068, "step": 115875 }, { "epoch": 2.7, "learning_rate": 9.788317570580367e-06, "loss": 0.1064, "step": 115880 }, { "epoch": 2.7, "learning_rate": 9.78743751613433e-06, "loss": 0.0031, "step": 115885 }, { "epoch": 2.7, "learning_rate": 9.786557461688297e-06, "loss": 0.055, "step": 115890 }, { "epoch": 2.7, "learning_rate": 9.785677407242262e-06, "loss": 0.1161, "step": 115895 }, { "epoch": 2.7, "learning_rate": 9.784797352796226e-06, "loss": 0.0266, "step": 115900 }, { "epoch": 2.7, "learning_rate": 9.783917298350193e-06, "loss": 0.0006, "step": 115905 }, { "epoch": 2.7, "learning_rate": 9.783037243904156e-06, "loss": 0.1372, "step": 115910 }, { "epoch": 2.7, "learning_rate": 9.782157189458123e-06, "loss": 0.0783, "step": 115915 }, { "epoch": 2.7, "learning_rate": 9.781277135012086e-06, "loss": 0.1472, "step": 115920 }, { "epoch": 2.7, "learning_rate": 9.780397080566053e-06, "loss": 0.2001, "step": 115925 }, { "epoch": 2.7, "learning_rate": 9.779517026120016e-06, "loss": 0.026, "step": 115930 }, { "epoch": 2.7, "learning_rate": 9.778636971673981e-06, "loss": 0.0297, "step": 115935 }, { "epoch": 2.71, "learning_rate": 9.777756917227946e-06, "loss": 0.0027, "step": 115940 }, { "epoch": 2.71, "learning_rate": 9.776876862781911e-06, "loss": 0.0167, "step": 115945 }, { "epoch": 2.71, "learning_rate": 9.775996808335876e-06, "loss": 0.053, "step": 115950 }, { "epoch": 2.71, "learning_rate": 9.775116753889841e-06, "loss": 0.1321, "step": 115955 }, { "epoch": 2.71, "learning_rate": 9.774236699443806e-06, "loss": 0.2413, "step": 115960 }, { "epoch": 2.71, "learning_rate": 9.773356644997771e-06, "loss": 0.1613, "step": 115965 }, { "epoch": 2.71, "learning_rate": 9.772476590551734e-06, "loss": 0.1855, "step": 115970 }, { "epoch": 2.71, "learning_rate": 9.771596536105701e-06, "loss": 0.412, "step": 115975 }, { "epoch": 2.71, "learning_rate": 9.770716481659664e-06, "loss": 0.0232, "step": 115980 }, { "epoch": 2.71, "learning_rate": 9.769836427213631e-06, "loss": 0.0175, "step": 115985 }, { "epoch": 2.71, "learning_rate": 9.768956372767595e-06, "loss": 0.0245, "step": 115990 }, { "epoch": 2.71, "learning_rate": 9.768076318321561e-06, "loss": 0.0915, "step": 115995 }, { "epoch": 2.71, "learning_rate": 9.767196263875525e-06, "loss": 0.097, "step": 116000 }, { "epoch": 2.71, "learning_rate": 9.766316209429491e-06, "loss": 0.0615, "step": 116005 }, { "epoch": 2.71, "learning_rate": 9.765436154983455e-06, "loss": 0.0557, "step": 116010 }, { "epoch": 2.71, "learning_rate": 9.76455610053742e-06, "loss": 0.2688, "step": 116015 }, { "epoch": 2.71, "learning_rate": 9.763676046091385e-06, "loss": 0.1742, "step": 116020 }, { "epoch": 2.71, "learning_rate": 9.76279599164535e-06, "loss": 0.1447, "step": 116025 }, { "epoch": 2.71, "learning_rate": 9.761915937199316e-06, "loss": 0.0632, "step": 116030 }, { "epoch": 2.71, "learning_rate": 9.76103588275328e-06, "loss": 0.0177, "step": 116035 }, { "epoch": 2.71, "learning_rate": 9.760155828307246e-06, "loss": 0.0178, "step": 116040 }, { "epoch": 2.71, "learning_rate": 9.75927577386121e-06, "loss": 0.0445, "step": 116045 }, { "epoch": 2.71, "learning_rate": 9.758395719415175e-06, "loss": 0.0287, "step": 116050 }, { "epoch": 2.71, "learning_rate": 9.75751566496914e-06, "loss": 0.1892, "step": 116055 }, { "epoch": 2.71, "learning_rate": 9.756635610523105e-06, "loss": 0.0421, "step": 116060 }, { "epoch": 2.71, "learning_rate": 9.75575555607707e-06, "loss": 0.128, "step": 116065 }, { "epoch": 2.71, "learning_rate": 9.754875501631035e-06, "loss": 0.2488, "step": 116070 }, { "epoch": 2.71, "learning_rate": 9.753995447185e-06, "loss": 0.0917, "step": 116075 }, { "epoch": 2.71, "learning_rate": 9.753115392738965e-06, "loss": 0.0384, "step": 116080 }, { "epoch": 2.71, "learning_rate": 9.752235338292928e-06, "loss": 0.0836, "step": 116085 }, { "epoch": 2.71, "learning_rate": 9.751355283846895e-06, "loss": 0.0475, "step": 116090 }, { "epoch": 2.71, "learning_rate": 9.750475229400858e-06, "loss": 0.0849, "step": 116095 }, { "epoch": 2.71, "learning_rate": 9.749595174954825e-06, "loss": 0.0083, "step": 116100 }, { "epoch": 2.71, "learning_rate": 9.748715120508788e-06, "loss": 0.1055, "step": 116105 }, { "epoch": 2.71, "learning_rate": 9.747835066062755e-06, "loss": 0.013, "step": 116110 }, { "epoch": 2.71, "learning_rate": 9.746955011616718e-06, "loss": 0.1541, "step": 116115 }, { "epoch": 2.71, "learning_rate": 9.746074957170683e-06, "loss": 0.1954, "step": 116120 }, { "epoch": 2.71, "learning_rate": 9.745194902724648e-06, "loss": 0.0832, "step": 116125 }, { "epoch": 2.71, "learning_rate": 9.744314848278613e-06, "loss": 0.0557, "step": 116130 }, { "epoch": 2.71, "learning_rate": 9.743434793832578e-06, "loss": 0.0526, "step": 116135 }, { "epoch": 2.71, "learning_rate": 9.742554739386543e-06, "loss": 0.1207, "step": 116140 }, { "epoch": 2.71, "learning_rate": 9.741674684940508e-06, "loss": 0.0325, "step": 116145 }, { "epoch": 2.71, "learning_rate": 9.740794630494473e-06, "loss": 0.1007, "step": 116150 }, { "epoch": 2.71, "learning_rate": 9.739914576048438e-06, "loss": 0.1465, "step": 116155 }, { "epoch": 2.71, "learning_rate": 9.739034521602404e-06, "loss": 0.1824, "step": 116160 }, { "epoch": 2.71, "learning_rate": 9.738154467156369e-06, "loss": 0.1281, "step": 116165 }, { "epoch": 2.71, "learning_rate": 9.737274412710334e-06, "loss": 0.1398, "step": 116170 }, { "epoch": 2.71, "learning_rate": 9.736394358264299e-06, "loss": 0.3423, "step": 116175 }, { "epoch": 2.71, "learning_rate": 9.735514303818264e-06, "loss": 0.053, "step": 116180 }, { "epoch": 2.71, "learning_rate": 9.734634249372229e-06, "loss": 0.0184, "step": 116185 }, { "epoch": 2.71, "learning_rate": 9.733754194926194e-06, "loss": 0.0272, "step": 116190 }, { "epoch": 2.71, "learning_rate": 9.732874140480159e-06, "loss": 0.0305, "step": 116195 }, { "epoch": 2.71, "learning_rate": 9.731994086034122e-06, "loss": 0.0106, "step": 116200 }, { "epoch": 2.71, "learning_rate": 9.731114031588089e-06, "loss": 0.0625, "step": 116205 }, { "epoch": 2.71, "learning_rate": 9.730233977142052e-06, "loss": 0.0306, "step": 116210 }, { "epoch": 2.71, "learning_rate": 9.729353922696019e-06, "loss": 0.1387, "step": 116215 }, { "epoch": 2.71, "learning_rate": 9.728473868249982e-06, "loss": 0.191, "step": 116220 }, { "epoch": 2.71, "learning_rate": 9.727593813803949e-06, "loss": 0.2324, "step": 116225 }, { "epoch": 2.71, "learning_rate": 9.726713759357912e-06, "loss": 0.0762, "step": 116230 }, { "epoch": 2.71, "learning_rate": 9.725833704911877e-06, "loss": 0.0622, "step": 116235 }, { "epoch": 2.71, "learning_rate": 9.724953650465842e-06, "loss": 0.0178, "step": 116240 }, { "epoch": 2.71, "learning_rate": 9.724073596019807e-06, "loss": 0.0271, "step": 116245 }, { "epoch": 2.71, "learning_rate": 9.723193541573772e-06, "loss": 0.0518, "step": 116250 }, { "epoch": 2.71, "learning_rate": 9.722313487127737e-06, "loss": 0.2, "step": 116255 }, { "epoch": 2.71, "learning_rate": 9.721433432681702e-06, "loss": 0.1539, "step": 116260 }, { "epoch": 2.71, "learning_rate": 9.720553378235667e-06, "loss": 0.0872, "step": 116265 }, { "epoch": 2.71, "learning_rate": 9.71967332378963e-06, "loss": 0.1964, "step": 116270 }, { "epoch": 2.71, "learning_rate": 9.718793269343597e-06, "loss": 0.1702, "step": 116275 }, { "epoch": 2.71, "learning_rate": 9.71791321489756e-06, "loss": 0.0459, "step": 116280 }, { "epoch": 2.71, "learning_rate": 9.717033160451527e-06, "loss": 0.0433, "step": 116285 }, { "epoch": 2.71, "learning_rate": 9.71615310600549e-06, "loss": 0.0356, "step": 116290 }, { "epoch": 2.71, "learning_rate": 9.715273051559457e-06, "loss": 0.0454, "step": 116295 }, { "epoch": 2.71, "learning_rate": 9.714392997113422e-06, "loss": 0.1244, "step": 116300 }, { "epoch": 2.71, "learning_rate": 9.713512942667387e-06, "loss": 0.0146, "step": 116305 }, { "epoch": 2.71, "learning_rate": 9.712632888221352e-06, "loss": 0.1446, "step": 116310 }, { "epoch": 2.71, "learning_rate": 9.711752833775316e-06, "loss": 0.1749, "step": 116315 }, { "epoch": 2.71, "learning_rate": 9.710872779329282e-06, "loss": 0.1111, "step": 116320 }, { "epoch": 2.71, "learning_rate": 9.709992724883246e-06, "loss": 0.4396, "step": 116325 }, { "epoch": 2.71, "learning_rate": 9.709112670437213e-06, "loss": 0.0693, "step": 116330 }, { "epoch": 2.71, "learning_rate": 9.708232615991176e-06, "loss": 0.0052, "step": 116335 }, { "epoch": 2.71, "learning_rate": 9.707352561545143e-06, "loss": 0.0825, "step": 116340 }, { "epoch": 2.71, "learning_rate": 9.706472507099106e-06, "loss": 0.0328, "step": 116345 }, { "epoch": 2.71, "learning_rate": 9.705592452653071e-06, "loss": 0.0115, "step": 116350 }, { "epoch": 2.71, "learning_rate": 9.704712398207036e-06, "loss": 0.0219, "step": 116355 }, { "epoch": 2.71, "learning_rate": 9.703832343761001e-06, "loss": 0.0603, "step": 116360 }, { "epoch": 2.71, "learning_rate": 9.702952289314966e-06, "loss": 0.1235, "step": 116365 }, { "epoch": 2.72, "learning_rate": 9.702072234868931e-06, "loss": 0.0867, "step": 116370 }, { "epoch": 2.72, "learning_rate": 9.701192180422896e-06, "loss": 0.2357, "step": 116375 }, { "epoch": 2.72, "learning_rate": 9.700312125976861e-06, "loss": 0.0854, "step": 116380 }, { "epoch": 2.72, "learning_rate": 9.699432071530824e-06, "loss": 0.001, "step": 116385 }, { "epoch": 2.72, "learning_rate": 9.698552017084791e-06, "loss": 0.0153, "step": 116390 }, { "epoch": 2.72, "learning_rate": 9.697671962638754e-06, "loss": 0.0564, "step": 116395 }, { "epoch": 2.72, "learning_rate": 9.696791908192721e-06, "loss": 0.0522, "step": 116400 }, { "epoch": 2.72, "learning_rate": 9.695911853746684e-06, "loss": 0.0652, "step": 116405 }, { "epoch": 2.72, "learning_rate": 9.695031799300651e-06, "loss": 0.1323, "step": 116410 }, { "epoch": 2.72, "learning_rate": 9.694151744854614e-06, "loss": 0.0839, "step": 116415 }, { "epoch": 2.72, "learning_rate": 9.69327169040858e-06, "loss": 0.0746, "step": 116420 }, { "epoch": 2.72, "learning_rate": 9.692391635962545e-06, "loss": 0.1445, "step": 116425 }, { "epoch": 2.72, "learning_rate": 9.69151158151651e-06, "loss": 0.0264, "step": 116430 }, { "epoch": 2.72, "learning_rate": 9.690631527070475e-06, "loss": 0.034, "step": 116435 }, { "epoch": 2.72, "learning_rate": 9.68975147262444e-06, "loss": 0.0131, "step": 116440 }, { "epoch": 2.72, "learning_rate": 9.688871418178406e-06, "loss": 0.2041, "step": 116445 }, { "epoch": 2.72, "learning_rate": 9.68799136373237e-06, "loss": 0.025, "step": 116450 }, { "epoch": 2.72, "learning_rate": 9.687111309286335e-06, "loss": 0.0951, "step": 116455 }, { "epoch": 2.72, "learning_rate": 9.6862312548403e-06, "loss": 0.1168, "step": 116460 }, { "epoch": 2.72, "learning_rate": 9.685351200394265e-06, "loss": 0.0816, "step": 116465 }, { "epoch": 2.72, "learning_rate": 9.68447114594823e-06, "loss": 0.1351, "step": 116470 }, { "epoch": 2.72, "learning_rate": 9.683591091502195e-06, "loss": 0.2245, "step": 116475 }, { "epoch": 2.72, "learning_rate": 9.68271103705616e-06, "loss": 0.0709, "step": 116480 }, { "epoch": 2.72, "learning_rate": 9.681830982610125e-06, "loss": 0.0033, "step": 116485 }, { "epoch": 2.72, "learning_rate": 9.68095092816409e-06, "loss": 0.0026, "step": 116490 }, { "epoch": 2.72, "learning_rate": 9.680070873718055e-06, "loss": 0.0527, "step": 116495 }, { "epoch": 2.72, "learning_rate": 9.679190819272018e-06, "loss": 0.088, "step": 116500 }, { "epoch": 2.72, "learning_rate": 9.678310764825985e-06, "loss": 0.0989, "step": 116505 }, { "epoch": 2.72, "learning_rate": 9.677430710379948e-06, "loss": 0.1456, "step": 116510 }, { "epoch": 2.72, "learning_rate": 9.676550655933915e-06, "loss": 0.0356, "step": 116515 }, { "epoch": 2.72, "learning_rate": 9.675670601487878e-06, "loss": 0.1256, "step": 116520 }, { "epoch": 2.72, "learning_rate": 9.674790547041845e-06, "loss": 0.154, "step": 116525 }, { "epoch": 2.72, "learning_rate": 9.673910492595808e-06, "loss": 0.0473, "step": 116530 }, { "epoch": 2.72, "learning_rate": 9.673030438149773e-06, "loss": 0.0828, "step": 116535 }, { "epoch": 2.72, "learning_rate": 9.672150383703738e-06, "loss": 0.0449, "step": 116540 }, { "epoch": 2.72, "learning_rate": 9.671270329257703e-06, "loss": 0.0036, "step": 116545 }, { "epoch": 2.72, "learning_rate": 9.670390274811668e-06, "loss": 0.098, "step": 116550 }, { "epoch": 2.72, "learning_rate": 9.669510220365633e-06, "loss": 0.1012, "step": 116555 }, { "epoch": 2.72, "learning_rate": 9.668630165919598e-06, "loss": 0.0918, "step": 116560 }, { "epoch": 2.72, "learning_rate": 9.667750111473563e-06, "loss": 0.2549, "step": 116565 }, { "epoch": 2.72, "learning_rate": 9.666870057027528e-06, "loss": 0.0942, "step": 116570 }, { "epoch": 2.72, "learning_rate": 9.665990002581493e-06, "loss": 0.1283, "step": 116575 }, { "epoch": 2.72, "learning_rate": 9.665109948135458e-06, "loss": 0.0058, "step": 116580 }, { "epoch": 2.72, "learning_rate": 9.664229893689423e-06, "loss": 0.0084, "step": 116585 }, { "epoch": 2.72, "learning_rate": 9.663349839243388e-06, "loss": 0.0285, "step": 116590 }, { "epoch": 2.72, "learning_rate": 9.662469784797354e-06, "loss": 0.0992, "step": 116595 }, { "epoch": 2.72, "learning_rate": 9.661589730351319e-06, "loss": 0.0613, "step": 116600 }, { "epoch": 2.72, "learning_rate": 9.660709675905282e-06, "loss": 0.1301, "step": 116605 }, { "epoch": 2.72, "learning_rate": 9.659829621459249e-06, "loss": 0.0896, "step": 116610 }, { "epoch": 2.72, "learning_rate": 9.658949567013212e-06, "loss": 0.1477, "step": 116615 }, { "epoch": 2.72, "learning_rate": 9.658069512567179e-06, "loss": 0.2157, "step": 116620 }, { "epoch": 2.72, "learning_rate": 9.657189458121142e-06, "loss": 0.3403, "step": 116625 }, { "epoch": 2.72, "learning_rate": 9.656309403675109e-06, "loss": 0.0502, "step": 116630 }, { "epoch": 2.72, "learning_rate": 9.655429349229072e-06, "loss": 0.0079, "step": 116635 }, { "epoch": 2.72, "learning_rate": 9.654549294783039e-06, "loss": 0.0037, "step": 116640 }, { "epoch": 2.72, "learning_rate": 9.653669240337002e-06, "loss": 0.0172, "step": 116645 }, { "epoch": 2.72, "learning_rate": 9.652789185890967e-06, "loss": 0.0456, "step": 116650 }, { "epoch": 2.72, "learning_rate": 9.651909131444932e-06, "loss": 0.0861, "step": 116655 }, { "epoch": 2.72, "learning_rate": 9.651029076998897e-06, "loss": 0.108, "step": 116660 }, { "epoch": 2.72, "learning_rate": 9.650149022552862e-06, "loss": 0.1545, "step": 116665 }, { "epoch": 2.72, "learning_rate": 9.649268968106827e-06, "loss": 0.1158, "step": 116670 }, { "epoch": 2.72, "learning_rate": 9.648388913660792e-06, "loss": 0.2417, "step": 116675 }, { "epoch": 2.72, "learning_rate": 9.647508859214757e-06, "loss": 0.008, "step": 116680 }, { "epoch": 2.72, "learning_rate": 9.64662880476872e-06, "loss": 0.0044, "step": 116685 }, { "epoch": 2.72, "learning_rate": 9.645748750322687e-06, "loss": 0.031, "step": 116690 }, { "epoch": 2.72, "learning_rate": 9.64486869587665e-06, "loss": 0.0099, "step": 116695 }, { "epoch": 2.72, "learning_rate": 9.643988641430617e-06, "loss": 0.0292, "step": 116700 }, { "epoch": 2.72, "learning_rate": 9.64310858698458e-06, "loss": 0.048, "step": 116705 }, { "epoch": 2.72, "learning_rate": 9.642228532538547e-06, "loss": 0.0471, "step": 116710 }, { "epoch": 2.72, "learning_rate": 9.641348478092512e-06, "loss": 0.236, "step": 116715 }, { "epoch": 2.72, "learning_rate": 9.640468423646476e-06, "loss": 0.0609, "step": 116720 }, { "epoch": 2.72, "learning_rate": 9.639588369200442e-06, "loss": 0.1728, "step": 116725 }, { "epoch": 2.72, "learning_rate": 9.638708314754406e-06, "loss": 0.03, "step": 116730 }, { "epoch": 2.72, "learning_rate": 9.637828260308372e-06, "loss": 0.0065, "step": 116735 }, { "epoch": 2.72, "learning_rate": 9.636948205862336e-06, "loss": 0.0228, "step": 116740 }, { "epoch": 2.72, "learning_rate": 9.636068151416302e-06, "loss": 0.0051, "step": 116745 }, { "epoch": 2.72, "learning_rate": 9.635188096970266e-06, "loss": 0.0573, "step": 116750 }, { "epoch": 2.72, "learning_rate": 9.63430804252423e-06, "loss": 0.1058, "step": 116755 }, { "epoch": 2.72, "learning_rate": 9.633427988078196e-06, "loss": 0.1671, "step": 116760 }, { "epoch": 2.72, "learning_rate": 9.63254793363216e-06, "loss": 0.1287, "step": 116765 }, { "epoch": 2.72, "learning_rate": 9.631667879186126e-06, "loss": 0.2896, "step": 116770 }, { "epoch": 2.72, "learning_rate": 9.630787824740091e-06, "loss": 0.2154, "step": 116775 }, { "epoch": 2.72, "learning_rate": 9.629907770294056e-06, "loss": 0.0747, "step": 116780 }, { "epoch": 2.72, "learning_rate": 9.629027715848021e-06, "loss": 0.0561, "step": 116785 }, { "epoch": 2.72, "learning_rate": 9.628147661401986e-06, "loss": 0.0156, "step": 116790 }, { "epoch": 2.72, "learning_rate": 9.627267606955951e-06, "loss": 0.0547, "step": 116795 }, { "epoch": 2.73, "learning_rate": 9.626387552509914e-06, "loss": 0.0053, "step": 116800 }, { "epoch": 2.73, "learning_rate": 9.625507498063881e-06, "loss": 0.0458, "step": 116805 }, { "epoch": 2.73, "learning_rate": 9.624627443617844e-06, "loss": 0.1914, "step": 116810 }, { "epoch": 2.73, "learning_rate": 9.623747389171811e-06, "loss": 0.2996, "step": 116815 }, { "epoch": 2.73, "learning_rate": 9.622867334725774e-06, "loss": 0.3684, "step": 116820 }, { "epoch": 2.73, "learning_rate": 9.621987280279741e-06, "loss": 0.4474, "step": 116825 }, { "epoch": 2.73, "learning_rate": 9.621107225833704e-06, "loss": 0.051, "step": 116830 }, { "epoch": 2.73, "learning_rate": 9.62022717138767e-06, "loss": 0.0471, "step": 116835 }, { "epoch": 2.73, "learning_rate": 9.619347116941634e-06, "loss": 0.01, "step": 116840 }, { "epoch": 2.73, "learning_rate": 9.6184670624956e-06, "loss": 0.0054, "step": 116845 }, { "epoch": 2.73, "learning_rate": 9.617587008049566e-06, "loss": 0.0915, "step": 116850 }, { "epoch": 2.73, "learning_rate": 9.61670695360353e-06, "loss": 0.014, "step": 116855 }, { "epoch": 2.73, "learning_rate": 9.615826899157496e-06, "loss": 0.0999, "step": 116860 }, { "epoch": 2.73, "learning_rate": 9.61494684471146e-06, "loss": 0.1108, "step": 116865 }, { "epoch": 2.73, "learning_rate": 9.614066790265425e-06, "loss": 0.1695, "step": 116870 }, { "epoch": 2.73, "learning_rate": 9.61318673581939e-06, "loss": 0.1416, "step": 116875 }, { "epoch": 2.73, "learning_rate": 9.612306681373355e-06, "loss": 0.0847, "step": 116880 }, { "epoch": 2.73, "learning_rate": 9.61142662692732e-06, "loss": 0.0771, "step": 116885 }, { "epoch": 2.73, "learning_rate": 9.610546572481285e-06, "loss": 0.0006, "step": 116890 }, { "epoch": 2.73, "learning_rate": 9.60966651803525e-06, "loss": 0.0383, "step": 116895 }, { "epoch": 2.73, "learning_rate": 9.608786463589215e-06, "loss": 0.0189, "step": 116900 }, { "epoch": 2.73, "learning_rate": 9.607906409143178e-06, "loss": 0.0468, "step": 116905 }, { "epoch": 2.73, "learning_rate": 9.607026354697145e-06, "loss": 0.0196, "step": 116910 }, { "epoch": 2.73, "learning_rate": 9.606146300251108e-06, "loss": 0.1653, "step": 116915 }, { "epoch": 2.73, "learning_rate": 9.605266245805075e-06, "loss": 0.3802, "step": 116920 }, { "epoch": 2.73, "learning_rate": 9.604386191359038e-06, "loss": 0.382, "step": 116925 }, { "epoch": 2.73, "learning_rate": 9.603506136913005e-06, "loss": 0.0169, "step": 116930 }, { "epoch": 2.73, "learning_rate": 9.602626082466968e-06, "loss": 0.0202, "step": 116935 }, { "epoch": 2.73, "learning_rate": 9.601746028020935e-06, "loss": 0.0309, "step": 116940 }, { "epoch": 2.73, "learning_rate": 9.600865973574898e-06, "loss": 0.0629, "step": 116945 }, { "epoch": 2.73, "learning_rate": 9.599985919128863e-06, "loss": 0.0289, "step": 116950 }, { "epoch": 2.73, "learning_rate": 9.599105864682828e-06, "loss": 0.0018, "step": 116955 }, { "epoch": 2.73, "learning_rate": 9.598225810236793e-06, "loss": 0.0662, "step": 116960 }, { "epoch": 2.73, "learning_rate": 9.597345755790758e-06, "loss": 0.183, "step": 116965 }, { "epoch": 2.73, "learning_rate": 9.596465701344723e-06, "loss": 0.2273, "step": 116970 }, { "epoch": 2.73, "learning_rate": 9.595585646898688e-06, "loss": 0.286, "step": 116975 }, { "epoch": 2.73, "learning_rate": 9.594705592452653e-06, "loss": 0.0109, "step": 116980 }, { "epoch": 2.73, "learning_rate": 9.593825538006618e-06, "loss": 0.0107, "step": 116985 }, { "epoch": 2.73, "learning_rate": 9.592945483560583e-06, "loss": 0.1323, "step": 116990 }, { "epoch": 2.73, "learning_rate": 9.592065429114548e-06, "loss": 0.0378, "step": 116995 }, { "epoch": 2.73, "learning_rate": 9.591185374668513e-06, "loss": 0.0346, "step": 117000 }, { "epoch": 2.73, "learning_rate": 9.590305320222478e-06, "loss": 0.0419, "step": 117005 }, { "epoch": 2.73, "learning_rate": 9.589425265776443e-06, "loss": 0.0987, "step": 117010 }, { "epoch": 2.73, "learning_rate": 9.588545211330408e-06, "loss": 0.204, "step": 117015 }, { "epoch": 2.73, "learning_rate": 9.587665156884372e-06, "loss": 0.292, "step": 117020 }, { "epoch": 2.73, "learning_rate": 9.586785102438339e-06, "loss": 0.2539, "step": 117025 }, { "epoch": 2.73, "learning_rate": 9.585905047992302e-06, "loss": 0.0553, "step": 117030 }, { "epoch": 2.73, "learning_rate": 9.585024993546269e-06, "loss": 0.0031, "step": 117035 }, { "epoch": 2.73, "learning_rate": 9.584144939100232e-06, "loss": 0.0114, "step": 117040 }, { "epoch": 2.73, "learning_rate": 9.583264884654199e-06, "loss": 0.0169, "step": 117045 }, { "epoch": 2.73, "learning_rate": 9.582384830208162e-06, "loss": 0.0286, "step": 117050 }, { "epoch": 2.73, "learning_rate": 9.581504775762127e-06, "loss": 0.0255, "step": 117055 }, { "epoch": 2.73, "learning_rate": 9.580624721316092e-06, "loss": 0.1917, "step": 117060 }, { "epoch": 2.73, "learning_rate": 9.579744666870057e-06, "loss": 0.1825, "step": 117065 }, { "epoch": 2.73, "learning_rate": 9.578864612424022e-06, "loss": 0.2239, "step": 117070 }, { "epoch": 2.73, "learning_rate": 9.577984557977987e-06, "loss": 0.2297, "step": 117075 }, { "epoch": 2.73, "learning_rate": 9.577104503531952e-06, "loss": 0.0174, "step": 117080 }, { "epoch": 2.73, "learning_rate": 9.576224449085917e-06, "loss": 0.0506, "step": 117085 }, { "epoch": 2.73, "learning_rate": 9.575344394639882e-06, "loss": 0.0318, "step": 117090 }, { "epoch": 2.73, "learning_rate": 9.574464340193847e-06, "loss": 0.0695, "step": 117095 }, { "epoch": 2.73, "learning_rate": 9.57358428574781e-06, "loss": 0.0758, "step": 117100 }, { "epoch": 2.73, "learning_rate": 9.572704231301777e-06, "loss": 0.1299, "step": 117105 }, { "epoch": 2.73, "learning_rate": 9.57182417685574e-06, "loss": 0.2268, "step": 117110 }, { "epoch": 2.73, "learning_rate": 9.570944122409707e-06, "loss": 0.1633, "step": 117115 }, { "epoch": 2.73, "learning_rate": 9.570064067963672e-06, "loss": 0.2304, "step": 117120 }, { "epoch": 2.73, "learning_rate": 9.569184013517637e-06, "loss": 0.3553, "step": 117125 }, { "epoch": 2.73, "learning_rate": 9.568303959071602e-06, "loss": 0.0406, "step": 117130 }, { "epoch": 2.73, "learning_rate": 9.567423904625566e-06, "loss": 0.0403, "step": 117135 }, { "epoch": 2.73, "learning_rate": 9.566543850179532e-06, "loss": 0.0405, "step": 117140 }, { "epoch": 2.73, "learning_rate": 9.565663795733496e-06, "loss": 0.0928, "step": 117145 }, { "epoch": 2.73, "learning_rate": 9.564783741287462e-06, "loss": 0.0166, "step": 117150 }, { "epoch": 2.73, "learning_rate": 9.563903686841426e-06, "loss": 0.0473, "step": 117155 }, { "epoch": 2.73, "learning_rate": 9.563023632395392e-06, "loss": 0.0834, "step": 117160 }, { "epoch": 2.73, "learning_rate": 9.562143577949356e-06, "loss": 0.0926, "step": 117165 }, { "epoch": 2.73, "learning_rate": 9.56126352350332e-06, "loss": 0.1315, "step": 117170 }, { "epoch": 2.73, "learning_rate": 9.560383469057286e-06, "loss": 0.0976, "step": 117175 }, { "epoch": 2.73, "learning_rate": 9.55950341461125e-06, "loss": 0.0858, "step": 117180 }, { "epoch": 2.73, "learning_rate": 9.558623360165216e-06, "loss": 0.057, "step": 117185 }, { "epoch": 2.73, "learning_rate": 9.55774330571918e-06, "loss": 0.0285, "step": 117190 }, { "epoch": 2.73, "learning_rate": 9.556863251273146e-06, "loss": 0.0903, "step": 117195 }, { "epoch": 2.73, "learning_rate": 9.55598319682711e-06, "loss": 0.0491, "step": 117200 }, { "epoch": 2.73, "learning_rate": 9.555103142381074e-06, "loss": 0.0798, "step": 117205 }, { "epoch": 2.73, "learning_rate": 9.554223087935041e-06, "loss": 0.0535, "step": 117210 }, { "epoch": 2.73, "learning_rate": 9.553343033489004e-06, "loss": 0.1824, "step": 117215 }, { "epoch": 2.73, "learning_rate": 9.552462979042971e-06, "loss": 0.22, "step": 117220 }, { "epoch": 2.73, "learning_rate": 9.551582924596934e-06, "loss": 0.0939, "step": 117225 }, { "epoch": 2.74, "learning_rate": 9.550702870150901e-06, "loss": 0.1345, "step": 117230 }, { "epoch": 2.74, "learning_rate": 9.549822815704864e-06, "loss": 0.0853, "step": 117235 }, { "epoch": 2.74, "learning_rate": 9.548942761258831e-06, "loss": 0.0223, "step": 117240 }, { "epoch": 2.74, "learning_rate": 9.548062706812794e-06, "loss": 0.0141, "step": 117245 }, { "epoch": 2.74, "learning_rate": 9.54718265236676e-06, "loss": 0.0032, "step": 117250 }, { "epoch": 2.74, "learning_rate": 9.546302597920726e-06, "loss": 0.109, "step": 117255 }, { "epoch": 2.74, "learning_rate": 9.54542254347469e-06, "loss": 0.0419, "step": 117260 }, { "epoch": 2.74, "learning_rate": 9.544542489028656e-06, "loss": 0.1593, "step": 117265 }, { "epoch": 2.74, "learning_rate": 9.54366243458262e-06, "loss": 0.0868, "step": 117270 }, { "epoch": 2.74, "learning_rate": 9.542782380136586e-06, "loss": 0.2669, "step": 117275 }, { "epoch": 2.74, "learning_rate": 9.54190232569055e-06, "loss": 0.0353, "step": 117280 }, { "epoch": 2.74, "learning_rate": 9.541022271244514e-06, "loss": 0.1419, "step": 117285 }, { "epoch": 2.74, "learning_rate": 9.54014221679848e-06, "loss": 0.0553, "step": 117290 }, { "epoch": 2.74, "learning_rate": 9.539262162352445e-06, "loss": 0.1043, "step": 117295 }, { "epoch": 2.74, "learning_rate": 9.53838210790641e-06, "loss": 0.0173, "step": 117300 }, { "epoch": 2.74, "learning_rate": 9.537502053460375e-06, "loss": 0.0567, "step": 117305 }, { "epoch": 2.74, "learning_rate": 9.53662199901434e-06, "loss": 0.0576, "step": 117310 }, { "epoch": 2.74, "learning_rate": 9.535741944568305e-06, "loss": 0.1245, "step": 117315 }, { "epoch": 2.74, "learning_rate": 9.534861890122268e-06, "loss": 0.1325, "step": 117320 }, { "epoch": 2.74, "learning_rate": 9.533981835676235e-06, "loss": 0.1589, "step": 117325 }, { "epoch": 2.74, "learning_rate": 9.533101781230198e-06, "loss": 0.0192, "step": 117330 }, { "epoch": 2.74, "learning_rate": 9.532221726784165e-06, "loss": 0.0815, "step": 117335 }, { "epoch": 2.74, "learning_rate": 9.531341672338128e-06, "loss": 0.0225, "step": 117340 }, { "epoch": 2.74, "learning_rate": 9.530461617892095e-06, "loss": 0.031, "step": 117345 }, { "epoch": 2.74, "learning_rate": 9.529581563446058e-06, "loss": 0.0692, "step": 117350 }, { "epoch": 2.74, "learning_rate": 9.528701509000023e-06, "loss": 0.1146, "step": 117355 }, { "epoch": 2.74, "learning_rate": 9.527821454553988e-06, "loss": 0.1384, "step": 117360 }, { "epoch": 2.74, "learning_rate": 9.526941400107953e-06, "loss": 0.104, "step": 117365 }, { "epoch": 2.74, "learning_rate": 9.526061345661918e-06, "loss": 0.3767, "step": 117370 }, { "epoch": 2.74, "learning_rate": 9.525181291215883e-06, "loss": 0.2201, "step": 117375 }, { "epoch": 2.74, "learning_rate": 9.524301236769848e-06, "loss": 0.015, "step": 117380 }, { "epoch": 2.74, "learning_rate": 9.523421182323813e-06, "loss": 0.0371, "step": 117385 }, { "epoch": 2.74, "learning_rate": 9.52254112787778e-06, "loss": 0.0708, "step": 117390 }, { "epoch": 2.74, "learning_rate": 9.521661073431743e-06, "loss": 0.0167, "step": 117395 }, { "epoch": 2.74, "learning_rate": 9.520781018985708e-06, "loss": 0.086, "step": 117400 }, { "epoch": 2.74, "learning_rate": 9.519900964539673e-06, "loss": 0.1091, "step": 117405 }, { "epoch": 2.74, "learning_rate": 9.519020910093638e-06, "loss": 0.0149, "step": 117410 }, { "epoch": 2.74, "learning_rate": 9.518140855647603e-06, "loss": 0.1411, "step": 117415 }, { "epoch": 2.74, "learning_rate": 9.517260801201568e-06, "loss": 0.2773, "step": 117420 }, { "epoch": 2.74, "learning_rate": 9.516380746755533e-06, "loss": 0.3288, "step": 117425 }, { "epoch": 2.74, "learning_rate": 9.515500692309498e-06, "loss": 0.1262, "step": 117430 }, { "epoch": 2.74, "learning_rate": 9.514620637863462e-06, "loss": 0.0043, "step": 117435 }, { "epoch": 2.74, "learning_rate": 9.513740583417428e-06, "loss": 0.0408, "step": 117440 }, { "epoch": 2.74, "learning_rate": 9.512860528971392e-06, "loss": 0.0344, "step": 117445 }, { "epoch": 2.74, "learning_rate": 9.511980474525358e-06, "loss": 0.0925, "step": 117450 }, { "epoch": 2.74, "learning_rate": 9.511100420079322e-06, "loss": 0.0377, "step": 117455 }, { "epoch": 2.74, "learning_rate": 9.510220365633289e-06, "loss": 0.0571, "step": 117460 }, { "epoch": 2.74, "learning_rate": 9.509340311187252e-06, "loss": 0.2132, "step": 117465 }, { "epoch": 2.74, "learning_rate": 9.508460256741217e-06, "loss": 0.1833, "step": 117470 }, { "epoch": 2.74, "learning_rate": 9.507580202295182e-06, "loss": 0.1499, "step": 117475 }, { "epoch": 2.74, "learning_rate": 9.506700147849147e-06, "loss": 0.084, "step": 117480 }, { "epoch": 2.74, "learning_rate": 9.505820093403112e-06, "loss": 0.0071, "step": 117485 }, { "epoch": 2.74, "learning_rate": 9.504940038957077e-06, "loss": 0.0513, "step": 117490 }, { "epoch": 2.74, "learning_rate": 9.504059984511042e-06, "loss": 0.0219, "step": 117495 }, { "epoch": 2.74, "learning_rate": 9.503179930065007e-06, "loss": 0.0354, "step": 117500 }, { "epoch": 2.74, "learning_rate": 9.50229987561897e-06, "loss": 0.0739, "step": 117505 }, { "epoch": 2.74, "learning_rate": 9.501419821172937e-06, "loss": 0.1313, "step": 117510 }, { "epoch": 2.74, "learning_rate": 9.5005397667269e-06, "loss": 0.1331, "step": 117515 }, { "epoch": 2.74, "learning_rate": 9.499659712280867e-06, "loss": 0.0931, "step": 117520 }, { "epoch": 2.74, "learning_rate": 9.498779657834832e-06, "loss": 0.2331, "step": 117525 }, { "epoch": 2.74, "learning_rate": 9.497899603388797e-06, "loss": 0.0132, "step": 117530 }, { "epoch": 2.74, "learning_rate": 9.497019548942762e-06, "loss": 0.0243, "step": 117535 }, { "epoch": 2.74, "learning_rate": 9.496139494496727e-06, "loss": 0.0358, "step": 117540 }, { "epoch": 2.74, "learning_rate": 9.495259440050692e-06, "loss": 0.0158, "step": 117545 }, { "epoch": 2.74, "learning_rate": 9.494379385604655e-06, "loss": 0.0538, "step": 117550 }, { "epoch": 2.74, "learning_rate": 9.493499331158622e-06, "loss": 0.0674, "step": 117555 }, { "epoch": 2.74, "learning_rate": 9.492619276712586e-06, "loss": 0.2015, "step": 117560 }, { "epoch": 2.74, "learning_rate": 9.491739222266552e-06, "loss": 0.1791, "step": 117565 }, { "epoch": 2.74, "learning_rate": 9.490859167820516e-06, "loss": 0.2516, "step": 117570 }, { "epoch": 2.74, "learning_rate": 9.489979113374482e-06, "loss": 0.2643, "step": 117575 }, { "epoch": 2.74, "learning_rate": 9.489099058928446e-06, "loss": 0.0572, "step": 117580 }, { "epoch": 2.74, "learning_rate": 9.48821900448241e-06, "loss": 0.0154, "step": 117585 }, { "epoch": 2.74, "learning_rate": 9.487338950036376e-06, "loss": 0.087, "step": 117590 }, { "epoch": 2.74, "learning_rate": 9.48645889559034e-06, "loss": 0.1889, "step": 117595 }, { "epoch": 2.74, "learning_rate": 9.485578841144306e-06, "loss": 0.0633, "step": 117600 }, { "epoch": 2.74, "learning_rate": 9.48469878669827e-06, "loss": 0.068, "step": 117605 }, { "epoch": 2.74, "learning_rate": 9.483818732252236e-06, "loss": 0.0261, "step": 117610 }, { "epoch": 2.74, "learning_rate": 9.4829386778062e-06, "loss": 0.1249, "step": 117615 }, { "epoch": 2.74, "learning_rate": 9.482058623360164e-06, "loss": 0.266, "step": 117620 }, { "epoch": 2.74, "learning_rate": 9.48117856891413e-06, "loss": 0.1723, "step": 117625 }, { "epoch": 2.74, "learning_rate": 9.480298514468094e-06, "loss": 0.0267, "step": 117630 }, { "epoch": 2.74, "learning_rate": 9.47941846002206e-06, "loss": 0.0029, "step": 117635 }, { "epoch": 2.74, "learning_rate": 9.478538405576024e-06, "loss": 0.1707, "step": 117640 }, { "epoch": 2.74, "learning_rate": 9.477658351129991e-06, "loss": 0.0015, "step": 117645 }, { "epoch": 2.74, "learning_rate": 9.476778296683954e-06, "loss": 0.2121, "step": 117650 }, { "epoch": 2.75, "learning_rate": 9.47589824223792e-06, "loss": 0.0627, "step": 117655 }, { "epoch": 2.75, "learning_rate": 9.475018187791886e-06, "loss": 0.0798, "step": 117660 }, { "epoch": 2.75, "learning_rate": 9.47413813334585e-06, "loss": 0.0754, "step": 117665 }, { "epoch": 2.75, "learning_rate": 9.473258078899816e-06, "loss": 0.0753, "step": 117670 }, { "epoch": 2.75, "learning_rate": 9.47237802445378e-06, "loss": 0.2453, "step": 117675 }, { "epoch": 2.75, "learning_rate": 9.471497970007746e-06, "loss": 0.0321, "step": 117680 }, { "epoch": 2.75, "learning_rate": 9.47061791556171e-06, "loss": 0.058, "step": 117685 }, { "epoch": 2.75, "learning_rate": 9.469737861115676e-06, "loss": 0.016, "step": 117690 }, { "epoch": 2.75, "learning_rate": 9.46885780666964e-06, "loss": 0.0121, "step": 117695 }, { "epoch": 2.75, "learning_rate": 9.467977752223604e-06, "loss": 0.11, "step": 117700 }, { "epoch": 2.75, "learning_rate": 9.46709769777757e-06, "loss": 0.0624, "step": 117705 }, { "epoch": 2.75, "learning_rate": 9.466217643331534e-06, "loss": 0.1603, "step": 117710 }, { "epoch": 2.75, "learning_rate": 9.4653375888855e-06, "loss": 0.1469, "step": 117715 }, { "epoch": 2.75, "learning_rate": 9.464457534439464e-06, "loss": 0.1321, "step": 117720 }, { "epoch": 2.75, "learning_rate": 9.46357747999343e-06, "loss": 0.3355, "step": 117725 }, { "epoch": 2.75, "learning_rate": 9.462697425547395e-06, "loss": 0.0755, "step": 117730 }, { "epoch": 2.75, "learning_rate": 9.461817371101358e-06, "loss": 0.0178, "step": 117735 }, { "epoch": 2.75, "learning_rate": 9.460937316655325e-06, "loss": 0.0423, "step": 117740 }, { "epoch": 2.75, "learning_rate": 9.460057262209288e-06, "loss": 0.1179, "step": 117745 }, { "epoch": 2.75, "learning_rate": 9.459177207763255e-06, "loss": 0.0049, "step": 117750 }, { "epoch": 2.75, "learning_rate": 9.458297153317218e-06, "loss": 0.082, "step": 117755 }, { "epoch": 2.75, "learning_rate": 9.457417098871185e-06, "loss": 0.1499, "step": 117760 }, { "epoch": 2.75, "learning_rate": 9.456537044425148e-06, "loss": 0.1713, "step": 117765 }, { "epoch": 2.75, "learning_rate": 9.455656989979113e-06, "loss": 0.1082, "step": 117770 }, { "epoch": 2.75, "learning_rate": 9.454776935533078e-06, "loss": 0.1726, "step": 117775 }, { "epoch": 2.75, "learning_rate": 9.453896881087043e-06, "loss": 0.0271, "step": 117780 }, { "epoch": 2.75, "learning_rate": 9.453016826641008e-06, "loss": 0.0413, "step": 117785 }, { "epoch": 2.75, "learning_rate": 9.452136772194973e-06, "loss": 0.1539, "step": 117790 }, { "epoch": 2.75, "learning_rate": 9.451256717748938e-06, "loss": 0.0242, "step": 117795 }, { "epoch": 2.75, "learning_rate": 9.450376663302903e-06, "loss": 0.0167, "step": 117800 }, { "epoch": 2.75, "learning_rate": 9.449496608856868e-06, "loss": 0.0805, "step": 117805 }, { "epoch": 2.75, "learning_rate": 9.448616554410833e-06, "loss": 0.1601, "step": 117810 }, { "epoch": 2.75, "learning_rate": 9.447736499964798e-06, "loss": 0.0329, "step": 117815 }, { "epoch": 2.75, "learning_rate": 9.446856445518763e-06, "loss": 0.1594, "step": 117820 }, { "epoch": 2.75, "learning_rate": 9.445976391072728e-06, "loss": 0.1548, "step": 117825 }, { "epoch": 2.75, "learning_rate": 9.445096336626693e-06, "loss": 0.048, "step": 117830 }, { "epoch": 2.75, "learning_rate": 9.444216282180658e-06, "loss": 0.2001, "step": 117835 }, { "epoch": 2.75, "learning_rate": 9.443336227734623e-06, "loss": 0.0249, "step": 117840 }, { "epoch": 2.75, "learning_rate": 9.442456173288588e-06, "loss": 0.0669, "step": 117845 }, { "epoch": 2.75, "learning_rate": 9.441576118842552e-06, "loss": 0.1047, "step": 117850 }, { "epoch": 2.75, "learning_rate": 9.440696064396518e-06, "loss": 0.1379, "step": 117855 }, { "epoch": 2.75, "learning_rate": 9.439816009950482e-06, "loss": 0.1146, "step": 117860 }, { "epoch": 2.75, "learning_rate": 9.438935955504448e-06, "loss": 0.0987, "step": 117865 }, { "epoch": 2.75, "learning_rate": 9.438055901058412e-06, "loss": 0.2472, "step": 117870 }, { "epoch": 2.75, "learning_rate": 9.437175846612378e-06, "loss": 0.2157, "step": 117875 }, { "epoch": 2.75, "learning_rate": 9.436295792166342e-06, "loss": 0.01, "step": 117880 }, { "epoch": 2.75, "learning_rate": 9.435415737720307e-06, "loss": 0.0005, "step": 117885 }, { "epoch": 2.75, "learning_rate": 9.434535683274272e-06, "loss": 0.0206, "step": 117890 }, { "epoch": 2.75, "learning_rate": 9.433655628828237e-06, "loss": 0.0205, "step": 117895 }, { "epoch": 2.75, "learning_rate": 9.432775574382202e-06, "loss": 0.0666, "step": 117900 }, { "epoch": 2.75, "learning_rate": 9.431895519936167e-06, "loss": 0.0872, "step": 117905 }, { "epoch": 2.75, "learning_rate": 9.431015465490132e-06, "loss": 0.0297, "step": 117910 }, { "epoch": 2.75, "learning_rate": 9.430135411044097e-06, "loss": 0.0585, "step": 117915 }, { "epoch": 2.75, "learning_rate": 9.42925535659806e-06, "loss": 0.1684, "step": 117920 }, { "epoch": 2.75, "learning_rate": 9.428375302152027e-06, "loss": 0.189, "step": 117925 }, { "epoch": 2.75, "learning_rate": 9.427495247705992e-06, "loss": 0.063, "step": 117930 }, { "epoch": 2.75, "learning_rate": 9.426615193259957e-06, "loss": 0.0477, "step": 117935 }, { "epoch": 2.75, "learning_rate": 9.425735138813922e-06, "loss": 0.0529, "step": 117940 }, { "epoch": 2.75, "learning_rate": 9.424855084367887e-06, "loss": 0.1522, "step": 117945 }, { "epoch": 2.75, "learning_rate": 9.423975029921852e-06, "loss": 0.0575, "step": 117950 }, { "epoch": 2.75, "learning_rate": 9.423094975475815e-06, "loss": 0.0885, "step": 117955 }, { "epoch": 2.75, "learning_rate": 9.422214921029782e-06, "loss": 0.0701, "step": 117960 }, { "epoch": 2.75, "learning_rate": 9.421334866583745e-06, "loss": 0.0154, "step": 117965 }, { "epoch": 2.75, "learning_rate": 9.420454812137712e-06, "loss": 0.1818, "step": 117970 }, { "epoch": 2.75, "learning_rate": 9.419574757691675e-06, "loss": 0.159, "step": 117975 }, { "epoch": 2.75, "learning_rate": 9.418694703245642e-06, "loss": 0.0008, "step": 117980 }, { "epoch": 2.75, "learning_rate": 9.417814648799606e-06, "loss": 0.0459, "step": 117985 }, { "epoch": 2.75, "learning_rate": 9.416934594353572e-06, "loss": 0.1154, "step": 117990 }, { "epoch": 2.75, "learning_rate": 9.416054539907536e-06, "loss": 0.0825, "step": 117995 }, { "epoch": 2.75, "learning_rate": 9.4151744854615e-06, "loss": 0.0247, "step": 118000 }, { "epoch": 2.75, "learning_rate": 9.414294431015466e-06, "loss": 0.0185, "step": 118005 }, { "epoch": 2.75, "learning_rate": 9.41341437656943e-06, "loss": 0.2849, "step": 118010 }, { "epoch": 2.75, "learning_rate": 9.412534322123396e-06, "loss": 0.1888, "step": 118015 }, { "epoch": 2.75, "learning_rate": 9.41165426767736e-06, "loss": 0.1286, "step": 118020 }, { "epoch": 2.75, "learning_rate": 9.410774213231326e-06, "loss": 0.1982, "step": 118025 }, { "epoch": 2.75, "learning_rate": 9.40989415878529e-06, "loss": 0.053, "step": 118030 }, { "epoch": 2.75, "learning_rate": 9.409014104339254e-06, "loss": 0.0515, "step": 118035 }, { "epoch": 2.75, "learning_rate": 9.40813404989322e-06, "loss": 0.0332, "step": 118040 }, { "epoch": 2.75, "learning_rate": 9.407253995447184e-06, "loss": 0.0581, "step": 118045 }, { "epoch": 2.75, "learning_rate": 9.40637394100115e-06, "loss": 0.1074, "step": 118050 }, { "epoch": 2.75, "learning_rate": 9.405493886555114e-06, "loss": 0.1925, "step": 118055 }, { "epoch": 2.75, "learning_rate": 9.40461383210908e-06, "loss": 0.0264, "step": 118060 }, { "epoch": 2.75, "learning_rate": 9.403733777663044e-06, "loss": 0.2978, "step": 118065 }, { "epoch": 2.75, "learning_rate": 9.40285372321701e-06, "loss": 0.143, "step": 118070 }, { "epoch": 2.75, "learning_rate": 9.401973668770976e-06, "loss": 0.1818, "step": 118075 }, { "epoch": 2.75, "learning_rate": 9.40109361432494e-06, "loss": 0.0683, "step": 118080 }, { "epoch": 2.76, "learning_rate": 9.400213559878906e-06, "loss": 0.0081, "step": 118085 }, { "epoch": 2.76, "learning_rate": 9.39933350543287e-06, "loss": 0.0588, "step": 118090 }, { "epoch": 2.76, "learning_rate": 9.398453450986836e-06, "loss": 0.0919, "step": 118095 }, { "epoch": 2.76, "learning_rate": 9.3975733965408e-06, "loss": 0.0527, "step": 118100 }, { "epoch": 2.76, "learning_rate": 9.396693342094764e-06, "loss": 0.055, "step": 118105 }, { "epoch": 2.76, "learning_rate": 9.39581328764873e-06, "loss": 0.0918, "step": 118110 }, { "epoch": 2.76, "learning_rate": 9.394933233202694e-06, "loss": 0.0589, "step": 118115 }, { "epoch": 2.76, "learning_rate": 9.39405317875666e-06, "loss": 0.1874, "step": 118120 }, { "epoch": 2.76, "learning_rate": 9.393173124310624e-06, "loss": 0.2493, "step": 118125 }, { "epoch": 2.76, "learning_rate": 9.39229306986459e-06, "loss": 0.02, "step": 118130 }, { "epoch": 2.76, "learning_rate": 9.391413015418554e-06, "loss": 0.0114, "step": 118135 }, { "epoch": 2.76, "learning_rate": 9.39053296097252e-06, "loss": 0.0189, "step": 118140 }, { "epoch": 2.76, "learning_rate": 9.389652906526484e-06, "loss": 0.0168, "step": 118145 }, { "epoch": 2.76, "learning_rate": 9.388772852080448e-06, "loss": 0.0726, "step": 118150 }, { "epoch": 2.76, "learning_rate": 9.387892797634415e-06, "loss": 0.0076, "step": 118155 }, { "epoch": 2.76, "learning_rate": 9.387012743188378e-06, "loss": 0.2121, "step": 118160 }, { "epoch": 2.76, "learning_rate": 9.386132688742345e-06, "loss": 0.1639, "step": 118165 }, { "epoch": 2.76, "learning_rate": 9.385252634296308e-06, "loss": 0.2019, "step": 118170 }, { "epoch": 2.76, "learning_rate": 9.384372579850275e-06, "loss": 0.3474, "step": 118175 }, { "epoch": 2.76, "learning_rate": 9.383492525404238e-06, "loss": 0.0284, "step": 118180 }, { "epoch": 2.76, "learning_rate": 9.382612470958203e-06, "loss": 0.0442, "step": 118185 }, { "epoch": 2.76, "learning_rate": 9.381732416512168e-06, "loss": 0.0895, "step": 118190 }, { "epoch": 2.76, "learning_rate": 9.380852362066133e-06, "loss": 0.1326, "step": 118195 }, { "epoch": 2.76, "learning_rate": 9.379972307620098e-06, "loss": 0.0795, "step": 118200 }, { "epoch": 2.76, "learning_rate": 9.379092253174063e-06, "loss": 0.0615, "step": 118205 }, { "epoch": 2.76, "learning_rate": 9.37821219872803e-06, "loss": 0.1587, "step": 118210 }, { "epoch": 2.76, "learning_rate": 9.377332144281993e-06, "loss": 0.0505, "step": 118215 }, { "epoch": 2.76, "learning_rate": 9.376452089835958e-06, "loss": 0.1735, "step": 118220 }, { "epoch": 2.76, "learning_rate": 9.375572035389923e-06, "loss": 0.4239, "step": 118225 }, { "epoch": 2.76, "learning_rate": 9.374691980943888e-06, "loss": 0.0303, "step": 118230 }, { "epoch": 2.76, "learning_rate": 9.373811926497853e-06, "loss": 0.0724, "step": 118235 }, { "epoch": 2.76, "learning_rate": 9.372931872051818e-06, "loss": 0.0074, "step": 118240 }, { "epoch": 2.76, "learning_rate": 9.372051817605783e-06, "loss": 0.033, "step": 118245 }, { "epoch": 2.76, "learning_rate": 9.371171763159748e-06, "loss": 0.0136, "step": 118250 }, { "epoch": 2.76, "learning_rate": 9.370291708713712e-06, "loss": 0.0703, "step": 118255 }, { "epoch": 2.76, "learning_rate": 9.369411654267678e-06, "loss": 0.2375, "step": 118260 }, { "epoch": 2.76, "learning_rate": 9.368531599821642e-06, "loss": 0.08, "step": 118265 }, { "epoch": 2.76, "learning_rate": 9.367651545375608e-06, "loss": 0.2896, "step": 118270 }, { "epoch": 2.76, "learning_rate": 9.366771490929572e-06, "loss": 0.1618, "step": 118275 }, { "epoch": 2.76, "learning_rate": 9.365891436483538e-06, "loss": 0.069, "step": 118280 }, { "epoch": 2.76, "learning_rate": 9.365011382037502e-06, "loss": 0.0014, "step": 118285 }, { "epoch": 2.76, "learning_rate": 9.364131327591468e-06, "loss": 0.0291, "step": 118290 }, { "epoch": 2.76, "learning_rate": 9.363251273145432e-06, "loss": 0.053, "step": 118295 }, { "epoch": 2.76, "learning_rate": 9.362371218699397e-06, "loss": 0.0136, "step": 118300 }, { "epoch": 2.76, "learning_rate": 9.361491164253362e-06, "loss": 0.0412, "step": 118305 }, { "epoch": 2.76, "learning_rate": 9.360611109807327e-06, "loss": 0.0158, "step": 118310 }, { "epoch": 2.76, "learning_rate": 9.359731055361292e-06, "loss": 0.2163, "step": 118315 }, { "epoch": 2.76, "learning_rate": 9.358851000915257e-06, "loss": 0.3897, "step": 118320 }, { "epoch": 2.76, "learning_rate": 9.357970946469222e-06, "loss": 0.0987, "step": 118325 }, { "epoch": 2.76, "learning_rate": 9.357090892023187e-06, "loss": 0.0224, "step": 118330 }, { "epoch": 2.76, "learning_rate": 9.35621083757715e-06, "loss": 0.0367, "step": 118335 }, { "epoch": 2.76, "learning_rate": 9.355330783131117e-06, "loss": 0.0575, "step": 118340 }, { "epoch": 2.76, "learning_rate": 9.354450728685082e-06, "loss": 0.0584, "step": 118345 }, { "epoch": 2.76, "learning_rate": 9.353570674239047e-06, "loss": 0.0373, "step": 118350 }, { "epoch": 2.76, "learning_rate": 9.352690619793012e-06, "loss": 0.0597, "step": 118355 }, { "epoch": 2.76, "learning_rate": 9.351810565346977e-06, "loss": 0.0225, "step": 118360 }, { "epoch": 2.76, "learning_rate": 9.350930510900942e-06, "loss": 0.2802, "step": 118365 }, { "epoch": 2.76, "learning_rate": 9.350050456454905e-06, "loss": 0.3093, "step": 118370 }, { "epoch": 2.76, "learning_rate": 9.349170402008872e-06, "loss": 0.1409, "step": 118375 }, { "epoch": 2.76, "learning_rate": 9.348290347562835e-06, "loss": 0.1245, "step": 118380 }, { "epoch": 2.76, "learning_rate": 9.347410293116802e-06, "loss": 0.0089, "step": 118385 }, { "epoch": 2.76, "learning_rate": 9.346530238670765e-06, "loss": 0.0468, "step": 118390 }, { "epoch": 2.76, "learning_rate": 9.345650184224732e-06, "loss": 0.0261, "step": 118395 }, { "epoch": 2.76, "learning_rate": 9.344770129778695e-06, "loss": 0.0762, "step": 118400 }, { "epoch": 2.76, "learning_rate": 9.34389007533266e-06, "loss": 0.0987, "step": 118405 }, { "epoch": 2.76, "learning_rate": 9.343010020886625e-06, "loss": 0.05, "step": 118410 }, { "epoch": 2.76, "learning_rate": 9.34212996644059e-06, "loss": 0.183, "step": 118415 }, { "epoch": 2.76, "learning_rate": 9.341249911994556e-06, "loss": 0.3803, "step": 118420 }, { "epoch": 2.76, "learning_rate": 9.34036985754852e-06, "loss": 0.1977, "step": 118425 }, { "epoch": 2.76, "learning_rate": 9.339489803102486e-06, "loss": 0.0172, "step": 118430 }, { "epoch": 2.76, "learning_rate": 9.33860974865645e-06, "loss": 0.013, "step": 118435 }, { "epoch": 2.76, "learning_rate": 9.337729694210414e-06, "loss": 0.0088, "step": 118440 }, { "epoch": 2.76, "learning_rate": 9.33684963976438e-06, "loss": 0.0455, "step": 118445 }, { "epoch": 2.76, "learning_rate": 9.335969585318344e-06, "loss": 0.0912, "step": 118450 }, { "epoch": 2.76, "learning_rate": 9.33508953087231e-06, "loss": 0.0915, "step": 118455 }, { "epoch": 2.76, "learning_rate": 9.334209476426274e-06, "loss": 0.049, "step": 118460 }, { "epoch": 2.76, "learning_rate": 9.33332942198024e-06, "loss": 0.1539, "step": 118465 }, { "epoch": 2.76, "learning_rate": 9.332449367534204e-06, "loss": 0.0504, "step": 118470 }, { "epoch": 2.76, "learning_rate": 9.33156931308817e-06, "loss": 0.2817, "step": 118475 }, { "epoch": 2.76, "learning_rate": 9.330689258642136e-06, "loss": 0.0191, "step": 118480 }, { "epoch": 2.76, "learning_rate": 9.329809204196099e-06, "loss": 0.0103, "step": 118485 }, { "epoch": 2.76, "learning_rate": 9.328929149750066e-06, "loss": 0.0867, "step": 118490 }, { "epoch": 2.76, "learning_rate": 9.328049095304029e-06, "loss": 0.0412, "step": 118495 }, { "epoch": 2.76, "learning_rate": 9.327169040857996e-06, "loss": 0.0515, "step": 118500 }, { "epoch": 2.76, "learning_rate": 9.32628898641196e-06, "loss": 0.0294, "step": 118505 }, { "epoch": 2.76, "learning_rate": 9.325408931965926e-06, "loss": 0.1426, "step": 118510 }, { "epoch": 2.77, "learning_rate": 9.32452887751989e-06, "loss": 0.0819, "step": 118515 }, { "epoch": 2.77, "learning_rate": 9.323648823073854e-06, "loss": 0.2447, "step": 118520 }, { "epoch": 2.77, "learning_rate": 9.32276876862782e-06, "loss": 0.2026, "step": 118525 }, { "epoch": 2.77, "learning_rate": 9.321888714181784e-06, "loss": 0.0439, "step": 118530 }, { "epoch": 2.77, "learning_rate": 9.32100865973575e-06, "loss": 0.0033, "step": 118535 }, { "epoch": 2.77, "learning_rate": 9.320128605289714e-06, "loss": 0.0144, "step": 118540 }, { "epoch": 2.77, "learning_rate": 9.31924855084368e-06, "loss": 0.0334, "step": 118545 }, { "epoch": 2.77, "learning_rate": 9.318368496397644e-06, "loss": 0.0677, "step": 118550 }, { "epoch": 2.77, "learning_rate": 9.317488441951608e-06, "loss": 0.1185, "step": 118555 }, { "epoch": 2.77, "learning_rate": 9.316608387505574e-06, "loss": 0.2133, "step": 118560 }, { "epoch": 2.77, "learning_rate": 9.315728333059538e-06, "loss": 0.0465, "step": 118565 }, { "epoch": 2.77, "learning_rate": 9.314848278613504e-06, "loss": 0.1685, "step": 118570 }, { "epoch": 2.77, "learning_rate": 9.313968224167468e-06, "loss": 0.2094, "step": 118575 }, { "epoch": 2.77, "learning_rate": 9.313088169721434e-06, "loss": 0.0793, "step": 118580 }, { "epoch": 2.77, "learning_rate": 9.312208115275398e-06, "loss": 0.0094, "step": 118585 }, { "epoch": 2.77, "learning_rate": 9.311328060829363e-06, "loss": 0.0167, "step": 118590 }, { "epoch": 2.77, "learning_rate": 9.310448006383328e-06, "loss": 0.0589, "step": 118595 }, { "epoch": 2.77, "learning_rate": 9.309567951937293e-06, "loss": 0.0936, "step": 118600 }, { "epoch": 2.77, "learning_rate": 9.308687897491258e-06, "loss": 0.0073, "step": 118605 }, { "epoch": 2.77, "learning_rate": 9.307807843045223e-06, "loss": 0.207, "step": 118610 }, { "epoch": 2.77, "learning_rate": 9.30692778859919e-06, "loss": 0.0478, "step": 118615 }, { "epoch": 2.77, "learning_rate": 9.306047734153153e-06, "loss": 0.2222, "step": 118620 }, { "epoch": 2.77, "learning_rate": 9.30516767970712e-06, "loss": 0.2512, "step": 118625 }, { "epoch": 2.77, "learning_rate": 9.304287625261083e-06, "loss": 0.0205, "step": 118630 }, { "epoch": 2.77, "learning_rate": 9.303407570815048e-06, "loss": 0.0023, "step": 118635 }, { "epoch": 2.77, "learning_rate": 9.302527516369013e-06, "loss": 0.025, "step": 118640 }, { "epoch": 2.77, "learning_rate": 9.301647461922978e-06, "loss": 0.011, "step": 118645 }, { "epoch": 2.77, "learning_rate": 9.300767407476943e-06, "loss": 0.0001, "step": 118650 }, { "epoch": 2.77, "learning_rate": 9.299887353030908e-06, "loss": 0.0664, "step": 118655 }, { "epoch": 2.77, "learning_rate": 9.299007298584873e-06, "loss": 0.1678, "step": 118660 }, { "epoch": 2.77, "learning_rate": 9.298127244138838e-06, "loss": 0.1621, "step": 118665 }, { "epoch": 2.77, "learning_rate": 9.297247189692801e-06, "loss": 0.3069, "step": 118670 }, { "epoch": 2.77, "learning_rate": 9.296367135246768e-06, "loss": 0.3323, "step": 118675 }, { "epoch": 2.77, "learning_rate": 9.295487080800731e-06, "loss": 0.0462, "step": 118680 }, { "epoch": 2.77, "learning_rate": 9.294607026354698e-06, "loss": 0.0162, "step": 118685 }, { "epoch": 2.77, "learning_rate": 9.293726971908662e-06, "loss": 0.0117, "step": 118690 }, { "epoch": 2.77, "learning_rate": 9.292846917462628e-06, "loss": 0.0846, "step": 118695 }, { "epoch": 2.77, "learning_rate": 9.291966863016592e-06, "loss": 0.07, "step": 118700 }, { "epoch": 2.77, "learning_rate": 9.291086808570557e-06, "loss": 0.0492, "step": 118705 }, { "epoch": 2.77, "learning_rate": 9.290206754124522e-06, "loss": 0.1657, "step": 118710 }, { "epoch": 2.77, "learning_rate": 9.289326699678487e-06, "loss": 0.0588, "step": 118715 }, { "epoch": 2.77, "learning_rate": 9.288446645232452e-06, "loss": 0.1799, "step": 118720 }, { "epoch": 2.77, "learning_rate": 9.287566590786417e-06, "loss": 0.2343, "step": 118725 }, { "epoch": 2.77, "learning_rate": 9.286686536340382e-06, "loss": 0.0119, "step": 118730 }, { "epoch": 2.77, "learning_rate": 9.285806481894347e-06, "loss": 0.001, "step": 118735 }, { "epoch": 2.77, "learning_rate": 9.28492642744831e-06, "loss": 0.1953, "step": 118740 }, { "epoch": 2.77, "learning_rate": 9.284046373002277e-06, "loss": 0.0673, "step": 118745 }, { "epoch": 2.77, "learning_rate": 9.283166318556242e-06, "loss": 0.0848, "step": 118750 }, { "epoch": 2.77, "learning_rate": 9.282286264110207e-06, "loss": 0.0536, "step": 118755 }, { "epoch": 2.77, "learning_rate": 9.281406209664172e-06, "loss": 0.2279, "step": 118760 }, { "epoch": 2.77, "learning_rate": 9.280526155218137e-06, "loss": 0.2518, "step": 118765 }, { "epoch": 2.77, "learning_rate": 9.279646100772102e-06, "loss": 0.1811, "step": 118770 }, { "epoch": 2.77, "learning_rate": 9.278766046326067e-06, "loss": 0.4361, "step": 118775 }, { "epoch": 2.77, "learning_rate": 9.277885991880032e-06, "loss": 0.0647, "step": 118780 }, { "epoch": 2.77, "learning_rate": 9.277005937433995e-06, "loss": 0.0942, "step": 118785 }, { "epoch": 2.77, "learning_rate": 9.276125882987962e-06, "loss": 0.0026, "step": 118790 }, { "epoch": 2.77, "learning_rate": 9.275245828541925e-06, "loss": 0.0232, "step": 118795 }, { "epoch": 2.77, "learning_rate": 9.274365774095892e-06, "loss": 0.0919, "step": 118800 }, { "epoch": 2.77, "learning_rate": 9.273485719649855e-06, "loss": 0.0789, "step": 118805 }, { "epoch": 2.77, "learning_rate": 9.272605665203822e-06, "loss": 0.0399, "step": 118810 }, { "epoch": 2.77, "learning_rate": 9.271725610757785e-06, "loss": 0.0646, "step": 118815 }, { "epoch": 2.77, "learning_rate": 9.27084555631175e-06, "loss": 0.2334, "step": 118820 }, { "epoch": 2.77, "learning_rate": 9.269965501865715e-06, "loss": 0.2787, "step": 118825 }, { "epoch": 2.77, "learning_rate": 9.26908544741968e-06, "loss": 0.0023, "step": 118830 }, { "epoch": 2.77, "learning_rate": 9.268205392973645e-06, "loss": 0.1473, "step": 118835 }, { "epoch": 2.77, "learning_rate": 9.26732533852761e-06, "loss": 0.0891, "step": 118840 }, { "epoch": 2.77, "learning_rate": 9.266445284081575e-06, "loss": 0.0349, "step": 118845 }, { "epoch": 2.77, "learning_rate": 9.26556522963554e-06, "loss": 0.0481, "step": 118850 }, { "epoch": 2.77, "learning_rate": 9.264685175189504e-06, "loss": 0.0929, "step": 118855 }, { "epoch": 2.77, "learning_rate": 9.26380512074347e-06, "loss": 0.1076, "step": 118860 }, { "epoch": 2.77, "learning_rate": 9.262925066297434e-06, "loss": 0.1286, "step": 118865 }, { "epoch": 2.77, "learning_rate": 9.2620450118514e-06, "loss": 0.1511, "step": 118870 }, { "epoch": 2.77, "learning_rate": 9.261164957405364e-06, "loss": 0.1536, "step": 118875 }, { "epoch": 2.77, "learning_rate": 9.26028490295933e-06, "loss": 0.0434, "step": 118880 }, { "epoch": 2.77, "learning_rate": 9.259404848513296e-06, "loss": 0.002, "step": 118885 }, { "epoch": 2.77, "learning_rate": 9.258524794067259e-06, "loss": 0.0561, "step": 118890 }, { "epoch": 2.77, "learning_rate": 9.257644739621226e-06, "loss": 0.1215, "step": 118895 }, { "epoch": 2.77, "learning_rate": 9.256764685175189e-06, "loss": 0.0903, "step": 118900 }, { "epoch": 2.77, "learning_rate": 9.255884630729156e-06, "loss": 0.0714, "step": 118905 }, { "epoch": 2.77, "learning_rate": 9.255004576283119e-06, "loss": 0.18, "step": 118910 }, { "epoch": 2.77, "learning_rate": 9.254124521837086e-06, "loss": 0.1556, "step": 118915 }, { "epoch": 2.77, "learning_rate": 9.253244467391049e-06, "loss": 0.1998, "step": 118920 }, { "epoch": 2.77, "learning_rate": 9.252364412945016e-06, "loss": 0.3845, "step": 118925 }, { "epoch": 2.77, "learning_rate": 9.251484358498979e-06, "loss": 0.0648, "step": 118930 }, { "epoch": 2.77, "learning_rate": 9.250604304052944e-06, "loss": 0.0784, "step": 118935 }, { "epoch": 2.78, "learning_rate": 9.24972424960691e-06, "loss": 0.0784, "step": 118940 }, { "epoch": 2.78, "learning_rate": 9.248844195160874e-06, "loss": 0.079, "step": 118945 }, { "epoch": 2.78, "learning_rate": 9.24796414071484e-06, "loss": 0.0603, "step": 118950 }, { "epoch": 2.78, "learning_rate": 9.247084086268804e-06, "loss": 0.1194, "step": 118955 }, { "epoch": 2.78, "learning_rate": 9.24620403182277e-06, "loss": 0.0606, "step": 118960 }, { "epoch": 2.78, "learning_rate": 9.245323977376734e-06, "loss": 0.2374, "step": 118965 }, { "epoch": 2.78, "learning_rate": 9.244443922930698e-06, "loss": 0.1356, "step": 118970 }, { "epoch": 2.78, "learning_rate": 9.243563868484664e-06, "loss": 0.2239, "step": 118975 }, { "epoch": 2.78, "learning_rate": 9.242683814038628e-06, "loss": 0.0465, "step": 118980 }, { "epoch": 2.78, "learning_rate": 9.241803759592594e-06, "loss": 0.1186, "step": 118985 }, { "epoch": 2.78, "learning_rate": 9.240923705146558e-06, "loss": 0.0763, "step": 118990 }, { "epoch": 2.78, "learning_rate": 9.240043650700524e-06, "loss": 0.054, "step": 118995 }, { "epoch": 2.78, "learning_rate": 9.239163596254488e-06, "loss": 0.0552, "step": 119000 }, { "epoch": 2.78, "learning_rate": 9.238283541808453e-06, "loss": 0.0719, "step": 119005 }, { "epoch": 2.78, "learning_rate": 9.237403487362418e-06, "loss": 0.0466, "step": 119010 }, { "epoch": 2.78, "learning_rate": 9.236523432916383e-06, "loss": 0.1176, "step": 119015 }, { "epoch": 2.78, "learning_rate": 9.23564337847035e-06, "loss": 0.2619, "step": 119020 }, { "epoch": 2.78, "learning_rate": 9.234763324024313e-06, "loss": 0.1072, "step": 119025 }, { "epoch": 2.78, "learning_rate": 9.23388326957828e-06, "loss": 0.0044, "step": 119030 }, { "epoch": 2.78, "learning_rate": 9.233003215132243e-06, "loss": 0.0048, "step": 119035 }, { "epoch": 2.78, "learning_rate": 9.232123160686208e-06, "loss": 0.0012, "step": 119040 }, { "epoch": 2.78, "learning_rate": 9.231243106240173e-06, "loss": 0.0483, "step": 119045 }, { "epoch": 2.78, "learning_rate": 9.230363051794138e-06, "loss": 0.0759, "step": 119050 }, { "epoch": 2.78, "learning_rate": 9.229482997348103e-06, "loss": 0.1517, "step": 119055 }, { "epoch": 2.78, "learning_rate": 9.228602942902068e-06, "loss": 0.0915, "step": 119060 }, { "epoch": 2.78, "learning_rate": 9.227722888456033e-06, "loss": 0.1457, "step": 119065 }, { "epoch": 2.78, "learning_rate": 9.226842834009998e-06, "loss": 0.2767, "step": 119070 }, { "epoch": 2.78, "learning_rate": 9.225962779563963e-06, "loss": 0.2089, "step": 119075 }, { "epoch": 2.78, "learning_rate": 9.225082725117928e-06, "loss": 0.0503, "step": 119080 }, { "epoch": 2.78, "learning_rate": 9.224202670671891e-06, "loss": 0.0017, "step": 119085 }, { "epoch": 2.78, "learning_rate": 9.223322616225858e-06, "loss": 0.0487, "step": 119090 }, { "epoch": 2.78, "learning_rate": 9.222442561779821e-06, "loss": 0.0229, "step": 119095 }, { "epoch": 2.78, "learning_rate": 9.221562507333788e-06, "loss": 0.003, "step": 119100 }, { "epoch": 2.78, "learning_rate": 9.220682452887751e-06, "loss": 0.0722, "step": 119105 }, { "epoch": 2.78, "learning_rate": 9.219802398441718e-06, "loss": 0.0846, "step": 119110 }, { "epoch": 2.78, "learning_rate": 9.218922343995682e-06, "loss": 0.1248, "step": 119115 }, { "epoch": 2.78, "learning_rate": 9.218042289549647e-06, "loss": 0.2023, "step": 119120 }, { "epoch": 2.78, "learning_rate": 9.217162235103612e-06, "loss": 0.1712, "step": 119125 }, { "epoch": 2.78, "learning_rate": 9.216282180657577e-06, "loss": 0.0181, "step": 119130 }, { "epoch": 2.78, "learning_rate": 9.215402126211542e-06, "loss": 0.0065, "step": 119135 }, { "epoch": 2.78, "learning_rate": 9.214522071765507e-06, "loss": 0.002, "step": 119140 }, { "epoch": 2.78, "learning_rate": 9.213642017319472e-06, "loss": 0.1816, "step": 119145 }, { "epoch": 2.78, "learning_rate": 9.212761962873437e-06, "loss": 0.021, "step": 119150 }, { "epoch": 2.78, "learning_rate": 9.211881908427402e-06, "loss": 0.0023, "step": 119155 }, { "epoch": 2.78, "learning_rate": 9.211001853981367e-06, "loss": 0.2781, "step": 119160 }, { "epoch": 2.78, "learning_rate": 9.210121799535332e-06, "loss": 0.0967, "step": 119165 }, { "epoch": 2.78, "learning_rate": 9.209241745089297e-06, "loss": 0.1512, "step": 119170 }, { "epoch": 2.78, "learning_rate": 9.208361690643262e-06, "loss": 0.2064, "step": 119175 }, { "epoch": 2.78, "learning_rate": 9.207481636197227e-06, "loss": 0.0489, "step": 119180 }, { "epoch": 2.78, "learning_rate": 9.206601581751192e-06, "loss": 0.0317, "step": 119185 }, { "epoch": 2.78, "learning_rate": 9.205721527305155e-06, "loss": 0.0697, "step": 119190 }, { "epoch": 2.78, "learning_rate": 9.204841472859122e-06, "loss": 0.0103, "step": 119195 }, { "epoch": 2.78, "learning_rate": 9.203961418413085e-06, "loss": 0.0712, "step": 119200 }, { "epoch": 2.78, "learning_rate": 9.203081363967052e-06, "loss": 0.0335, "step": 119205 }, { "epoch": 2.78, "learning_rate": 9.202201309521015e-06, "loss": 0.0265, "step": 119210 }, { "epoch": 2.78, "learning_rate": 9.201321255074982e-06, "loss": 0.1595, "step": 119215 }, { "epoch": 2.78, "learning_rate": 9.200441200628945e-06, "loss": 0.2288, "step": 119220 }, { "epoch": 2.78, "learning_rate": 9.199561146182912e-06, "loss": 0.1775, "step": 119225 }, { "epoch": 2.78, "learning_rate": 9.198681091736875e-06, "loss": 0.0648, "step": 119230 }, { "epoch": 2.78, "learning_rate": 9.19780103729084e-06, "loss": 0.0008, "step": 119235 }, { "epoch": 2.78, "learning_rate": 9.196920982844805e-06, "loss": 0.0456, "step": 119240 }, { "epoch": 2.78, "learning_rate": 9.19604092839877e-06, "loss": 0.0669, "step": 119245 }, { "epoch": 2.78, "learning_rate": 9.195160873952735e-06, "loss": 0.0489, "step": 119250 }, { "epoch": 2.78, "learning_rate": 9.1942808195067e-06, "loss": 0.0595, "step": 119255 }, { "epoch": 2.78, "learning_rate": 9.193400765060665e-06, "loss": 0.0901, "step": 119260 }, { "epoch": 2.78, "learning_rate": 9.19252071061463e-06, "loss": 0.107, "step": 119265 }, { "epoch": 2.78, "learning_rate": 9.191640656168594e-06, "loss": 0.0857, "step": 119270 }, { "epoch": 2.78, "learning_rate": 9.19076060172256e-06, "loss": 0.1969, "step": 119275 }, { "epoch": 2.78, "learning_rate": 9.189880547276524e-06, "loss": 0.0487, "step": 119280 }, { "epoch": 2.78, "learning_rate": 9.18900049283049e-06, "loss": 0.1489, "step": 119285 }, { "epoch": 2.78, "learning_rate": 9.188120438384456e-06, "loss": 0.043, "step": 119290 }, { "epoch": 2.78, "learning_rate": 9.18724038393842e-06, "loss": 0.0909, "step": 119295 }, { "epoch": 2.78, "learning_rate": 9.186360329492386e-06, "loss": 0.0637, "step": 119300 }, { "epoch": 2.78, "learning_rate": 9.185480275046349e-06, "loss": 0.0944, "step": 119305 }, { "epoch": 2.78, "learning_rate": 9.184600220600316e-06, "loss": 0.1744, "step": 119310 }, { "epoch": 2.78, "learning_rate": 9.183720166154279e-06, "loss": 0.1111, "step": 119315 }, { "epoch": 2.78, "learning_rate": 9.182840111708246e-06, "loss": 0.117, "step": 119320 }, { "epoch": 2.78, "learning_rate": 9.181960057262209e-06, "loss": 0.2422, "step": 119325 }, { "epoch": 2.78, "learning_rate": 9.181080002816176e-06, "loss": 0.0355, "step": 119330 }, { "epoch": 2.78, "learning_rate": 9.180199948370139e-06, "loss": 0.004, "step": 119335 }, { "epoch": 2.78, "learning_rate": 9.179319893924104e-06, "loss": 0.0473, "step": 119340 }, { "epoch": 2.78, "learning_rate": 9.178439839478069e-06, "loss": 0.0025, "step": 119345 }, { "epoch": 2.78, "learning_rate": 9.177559785032034e-06, "loss": 0.0242, "step": 119350 }, { "epoch": 2.78, "learning_rate": 9.176679730585999e-06, "loss": 0.1348, "step": 119355 }, { "epoch": 2.78, "learning_rate": 9.175799676139964e-06, "loss": 0.1521, "step": 119360 }, { "epoch": 2.78, "learning_rate": 9.174919621693929e-06, "loss": 0.0756, "step": 119365 }, { "epoch": 2.79, "learning_rate": 9.174039567247894e-06, "loss": 0.1611, "step": 119370 }, { "epoch": 2.79, "learning_rate": 9.17315951280186e-06, "loss": 0.1917, "step": 119375 }, { "epoch": 2.79, "learning_rate": 9.172279458355824e-06, "loss": 0.0227, "step": 119380 }, { "epoch": 2.79, "learning_rate": 9.171399403909788e-06, "loss": 0.0207, "step": 119385 }, { "epoch": 2.79, "learning_rate": 9.170519349463754e-06, "loss": 0.0245, "step": 119390 }, { "epoch": 2.79, "learning_rate": 9.169639295017718e-06, "loss": 0.0849, "step": 119395 }, { "epoch": 2.79, "learning_rate": 9.168759240571684e-06, "loss": 0.1687, "step": 119400 }, { "epoch": 2.79, "learning_rate": 9.167879186125648e-06, "loss": 0.0953, "step": 119405 }, { "epoch": 2.79, "learning_rate": 9.166999131679614e-06, "loss": 0.0825, "step": 119410 }, { "epoch": 2.79, "learning_rate": 9.166119077233578e-06, "loss": 0.0719, "step": 119415 }, { "epoch": 2.79, "learning_rate": 9.165239022787543e-06, "loss": 0.1154, "step": 119420 }, { "epoch": 2.79, "learning_rate": 9.16435896834151e-06, "loss": 0.1728, "step": 119425 }, { "epoch": 2.79, "learning_rate": 9.163478913895473e-06, "loss": 0.0955, "step": 119430 }, { "epoch": 2.79, "learning_rate": 9.16259885944944e-06, "loss": 0.0267, "step": 119435 }, { "epoch": 2.79, "learning_rate": 9.161718805003403e-06, "loss": 0.0098, "step": 119440 }, { "epoch": 2.79, "learning_rate": 9.16083875055737e-06, "loss": 0.0137, "step": 119445 }, { "epoch": 2.79, "learning_rate": 9.159958696111333e-06, "loss": 0.1063, "step": 119450 }, { "epoch": 2.79, "learning_rate": 9.159078641665298e-06, "loss": 0.0042, "step": 119455 }, { "epoch": 2.79, "learning_rate": 9.158198587219263e-06, "loss": 0.0734, "step": 119460 }, { "epoch": 2.79, "learning_rate": 9.157318532773228e-06, "loss": 0.1228, "step": 119465 }, { "epoch": 2.79, "learning_rate": 9.156438478327193e-06, "loss": 0.1216, "step": 119470 }, { "epoch": 2.79, "learning_rate": 9.155558423881158e-06, "loss": 0.29, "step": 119475 }, { "epoch": 2.79, "learning_rate": 9.154678369435123e-06, "loss": 0.0713, "step": 119480 }, { "epoch": 2.79, "learning_rate": 9.153798314989088e-06, "loss": 0.0333, "step": 119485 }, { "epoch": 2.79, "learning_rate": 9.152918260543051e-06, "loss": 0.0806, "step": 119490 }, { "epoch": 2.79, "learning_rate": 9.152038206097018e-06, "loss": 0.0395, "step": 119495 }, { "epoch": 2.79, "learning_rate": 9.151158151650981e-06, "loss": 0.3831, "step": 119500 }, { "epoch": 2.79, "learning_rate": 9.150278097204948e-06, "loss": 0.058, "step": 119505 }, { "epoch": 2.79, "learning_rate": 9.149398042758911e-06, "loss": 0.0645, "step": 119510 }, { "epoch": 2.79, "learning_rate": 9.148517988312878e-06, "loss": 0.128, "step": 119515 }, { "epoch": 2.79, "learning_rate": 9.147637933866841e-06, "loss": 0.2842, "step": 119520 }, { "epoch": 2.79, "learning_rate": 9.146757879420808e-06, "loss": 0.2302, "step": 119525 }, { "epoch": 2.79, "learning_rate": 9.145877824974771e-06, "loss": 0.0799, "step": 119530 }, { "epoch": 2.79, "learning_rate": 9.144997770528736e-06, "loss": 0.0217, "step": 119535 }, { "epoch": 2.79, "learning_rate": 9.144117716082701e-06, "loss": 0.0222, "step": 119540 }, { "epoch": 2.79, "learning_rate": 9.143237661636666e-06, "loss": 0.0502, "step": 119545 }, { "epoch": 2.79, "learning_rate": 9.142357607190632e-06, "loss": 0.0049, "step": 119550 }, { "epoch": 2.79, "learning_rate": 9.141477552744597e-06, "loss": 0.0396, "step": 119555 }, { "epoch": 2.79, "learning_rate": 9.140597498298562e-06, "loss": 0.0671, "step": 119560 }, { "epoch": 2.79, "learning_rate": 9.139717443852527e-06, "loss": 0.1283, "step": 119565 }, { "epoch": 2.79, "learning_rate": 9.138837389406492e-06, "loss": 0.1763, "step": 119570 }, { "epoch": 2.79, "learning_rate": 9.137957334960457e-06, "loss": 0.2823, "step": 119575 }, { "epoch": 2.79, "learning_rate": 9.137077280514422e-06, "loss": 0.0643, "step": 119580 }, { "epoch": 2.79, "learning_rate": 9.136197226068387e-06, "loss": 0.0314, "step": 119585 }, { "epoch": 2.79, "learning_rate": 9.135317171622352e-06, "loss": 0.0433, "step": 119590 }, { "epoch": 2.79, "learning_rate": 9.134437117176317e-06, "loss": 0.0433, "step": 119595 }, { "epoch": 2.79, "learning_rate": 9.133557062730282e-06, "loss": 0.0932, "step": 119600 }, { "epoch": 2.79, "learning_rate": 9.132677008284245e-06, "loss": 0.0574, "step": 119605 }, { "epoch": 2.79, "learning_rate": 9.131796953838212e-06, "loss": 0.1487, "step": 119610 }, { "epoch": 2.79, "learning_rate": 9.130916899392175e-06, "loss": 0.2267, "step": 119615 }, { "epoch": 2.79, "learning_rate": 9.130036844946142e-06, "loss": 0.35, "step": 119620 }, { "epoch": 2.79, "learning_rate": 9.129156790500105e-06, "loss": 0.3753, "step": 119625 }, { "epoch": 2.79, "learning_rate": 9.128276736054072e-06, "loss": 0.004, "step": 119630 }, { "epoch": 2.79, "learning_rate": 9.127396681608035e-06, "loss": 0.0179, "step": 119635 }, { "epoch": 2.79, "learning_rate": 9.126516627162e-06, "loss": 0.0866, "step": 119640 }, { "epoch": 2.79, "learning_rate": 9.125636572715965e-06, "loss": 0.0224, "step": 119645 }, { "epoch": 2.79, "learning_rate": 9.12475651826993e-06, "loss": 0.0757, "step": 119650 }, { "epoch": 2.79, "learning_rate": 9.123876463823895e-06, "loss": 0.0341, "step": 119655 }, { "epoch": 2.79, "learning_rate": 9.12299640937786e-06, "loss": 0.1391, "step": 119660 }, { "epoch": 2.79, "learning_rate": 9.122116354931825e-06, "loss": 0.1272, "step": 119665 }, { "epoch": 2.79, "learning_rate": 9.12123630048579e-06, "loss": 0.1574, "step": 119670 }, { "epoch": 2.79, "learning_rate": 9.120356246039755e-06, "loss": 0.0535, "step": 119675 }, { "epoch": 2.79, "learning_rate": 9.11947619159372e-06, "loss": 0.0395, "step": 119680 }, { "epoch": 2.79, "learning_rate": 9.118596137147684e-06, "loss": 0.0098, "step": 119685 }, { "epoch": 2.79, "learning_rate": 9.11771608270165e-06, "loss": 0.0051, "step": 119690 }, { "epoch": 2.79, "learning_rate": 9.116836028255615e-06, "loss": 0.0108, "step": 119695 }, { "epoch": 2.79, "learning_rate": 9.11595597380958e-06, "loss": 0.0507, "step": 119700 }, { "epoch": 2.79, "learning_rate": 9.115075919363545e-06, "loss": 0.1518, "step": 119705 }, { "epoch": 2.79, "learning_rate": 9.11419586491751e-06, "loss": 0.0825, "step": 119710 }, { "epoch": 2.79, "learning_rate": 9.113315810471476e-06, "loss": 0.163, "step": 119715 }, { "epoch": 2.79, "learning_rate": 9.112435756025439e-06, "loss": 0.3258, "step": 119720 }, { "epoch": 2.79, "learning_rate": 9.111555701579406e-06, "loss": 0.0958, "step": 119725 }, { "epoch": 2.79, "learning_rate": 9.110675647133369e-06, "loss": 0.0013, "step": 119730 }, { "epoch": 2.79, "learning_rate": 9.109795592687336e-06, "loss": 0.0097, "step": 119735 }, { "epoch": 2.79, "learning_rate": 9.108915538241299e-06, "loss": 0.0166, "step": 119740 }, { "epoch": 2.79, "learning_rate": 9.108035483795266e-06, "loss": 0.0319, "step": 119745 }, { "epoch": 2.79, "learning_rate": 9.107155429349229e-06, "loss": 0.0645, "step": 119750 }, { "epoch": 2.79, "learning_rate": 9.106275374903194e-06, "loss": 0.0301, "step": 119755 }, { "epoch": 2.79, "learning_rate": 9.105395320457159e-06, "loss": 0.24, "step": 119760 }, { "epoch": 2.79, "learning_rate": 9.104515266011124e-06, "loss": 0.0599, "step": 119765 }, { "epoch": 2.79, "learning_rate": 9.103635211565089e-06, "loss": 0.1331, "step": 119770 }, { "epoch": 2.79, "learning_rate": 9.102755157119054e-06, "loss": 0.2293, "step": 119775 }, { "epoch": 2.79, "learning_rate": 9.101875102673019e-06, "loss": 0.0062, "step": 119780 }, { "epoch": 2.79, "learning_rate": 9.100995048226984e-06, "loss": 0.0242, "step": 119785 }, { "epoch": 2.79, "learning_rate": 9.100114993780947e-06, "loss": 0.0444, "step": 119790 }, { "epoch": 2.79, "learning_rate": 9.099234939334914e-06, "loss": 0.003, "step": 119795 }, { "epoch": 2.8, "learning_rate": 9.098354884888877e-06, "loss": 0.0083, "step": 119800 }, { "epoch": 2.8, "learning_rate": 9.097474830442844e-06, "loss": 0.0462, "step": 119805 }, { "epoch": 2.8, "learning_rate": 9.096594775996808e-06, "loss": 0.1009, "step": 119810 }, { "epoch": 2.8, "learning_rate": 9.095714721550774e-06, "loss": 0.2717, "step": 119815 }, { "epoch": 2.8, "learning_rate": 9.094834667104738e-06, "loss": 0.2799, "step": 119820 }, { "epoch": 2.8, "learning_rate": 9.093954612658704e-06, "loss": 0.1269, "step": 119825 }, { "epoch": 2.8, "learning_rate": 9.093074558212668e-06, "loss": 0.0254, "step": 119830 }, { "epoch": 2.8, "learning_rate": 9.092194503766633e-06, "loss": 0.0222, "step": 119835 }, { "epoch": 2.8, "learning_rate": 9.0913144493206e-06, "loss": 0.0449, "step": 119840 }, { "epoch": 2.8, "learning_rate": 9.090434394874563e-06, "loss": 0.0604, "step": 119845 }, { "epoch": 2.8, "learning_rate": 9.08955434042853e-06, "loss": 0.0439, "step": 119850 }, { "epoch": 2.8, "learning_rate": 9.088674285982493e-06, "loss": 0.0685, "step": 119855 }, { "epoch": 2.8, "learning_rate": 9.08779423153646e-06, "loss": 0.0837, "step": 119860 }, { "epoch": 2.8, "learning_rate": 9.086914177090423e-06, "loss": 0.08, "step": 119865 }, { "epoch": 2.8, "learning_rate": 9.086034122644388e-06, "loss": 0.2377, "step": 119870 }, { "epoch": 2.8, "learning_rate": 9.085154068198353e-06, "loss": 0.1669, "step": 119875 }, { "epoch": 2.8, "learning_rate": 9.084274013752318e-06, "loss": 0.0021, "step": 119880 }, { "epoch": 2.8, "learning_rate": 9.083393959306283e-06, "loss": 0.0251, "step": 119885 }, { "epoch": 2.8, "learning_rate": 9.082513904860248e-06, "loss": 0.004, "step": 119890 }, { "epoch": 2.8, "learning_rate": 9.081633850414213e-06, "loss": 0.1246, "step": 119895 }, { "epoch": 2.8, "learning_rate": 9.080753795968178e-06, "loss": 0.0246, "step": 119900 }, { "epoch": 2.8, "learning_rate": 9.079873741522141e-06, "loss": 0.0279, "step": 119905 }, { "epoch": 2.8, "learning_rate": 9.078993687076108e-06, "loss": 0.031, "step": 119910 }, { "epoch": 2.8, "learning_rate": 9.078113632630071e-06, "loss": 0.101, "step": 119915 }, { "epoch": 2.8, "learning_rate": 9.077233578184038e-06, "loss": 0.1643, "step": 119920 }, { "epoch": 2.8, "learning_rate": 9.076353523738001e-06, "loss": 0.3182, "step": 119925 }, { "epoch": 2.8, "learning_rate": 9.075473469291968e-06, "loss": 0.04, "step": 119930 }, { "epoch": 2.8, "learning_rate": 9.074593414845931e-06, "loss": 0.0389, "step": 119935 }, { "epoch": 2.8, "learning_rate": 9.073713360399896e-06, "loss": 0.02, "step": 119940 }, { "epoch": 2.8, "learning_rate": 9.072833305953861e-06, "loss": 0.0114, "step": 119945 }, { "epoch": 2.8, "learning_rate": 9.071953251507826e-06, "loss": 0.042, "step": 119950 }, { "epoch": 2.8, "learning_rate": 9.071073197061791e-06, "loss": 0.0045, "step": 119955 }, { "epoch": 2.8, "learning_rate": 9.070193142615756e-06, "loss": 0.0316, "step": 119960 }, { "epoch": 2.8, "learning_rate": 9.069313088169721e-06, "loss": 0.118, "step": 119965 }, { "epoch": 2.8, "learning_rate": 9.068433033723686e-06, "loss": 0.1387, "step": 119970 }, { "epoch": 2.8, "learning_rate": 9.067552979277653e-06, "loss": 0.3026, "step": 119975 }, { "epoch": 2.8, "learning_rate": 9.066672924831617e-06, "loss": 0.2665, "step": 119980 }, { "epoch": 2.8, "learning_rate": 9.065792870385582e-06, "loss": 0.0788, "step": 119985 }, { "epoch": 2.8, "learning_rate": 9.064912815939547e-06, "loss": 0.0339, "step": 119990 }, { "epoch": 2.8, "learning_rate": 9.064032761493512e-06, "loss": 0.0842, "step": 119995 }, { "epoch": 2.8, "learning_rate": 9.063152707047477e-06, "loss": 0.0439, "step": 120000 }, { "epoch": 2.8, "learning_rate": 9.062272652601442e-06, "loss": 0.0403, "step": 120005 }, { "epoch": 2.8, "learning_rate": 9.061392598155407e-06, "loss": 0.0985, "step": 120010 }, { "epoch": 2.8, "learning_rate": 9.060512543709372e-06, "loss": 0.1311, "step": 120015 }, { "epoch": 2.8, "learning_rate": 9.059632489263335e-06, "loss": 0.0752, "step": 120020 }, { "epoch": 2.8, "learning_rate": 9.058752434817302e-06, "loss": 0.1984, "step": 120025 }, { "epoch": 2.8, "learning_rate": 9.057872380371265e-06, "loss": 0.1212, "step": 120030 }, { "epoch": 2.8, "learning_rate": 9.056992325925232e-06, "loss": 0.009, "step": 120035 }, { "epoch": 2.8, "learning_rate": 9.056112271479195e-06, "loss": 0.0022, "step": 120040 }, { "epoch": 2.8, "learning_rate": 9.055232217033162e-06, "loss": 0.0823, "step": 120045 }, { "epoch": 2.8, "learning_rate": 9.054352162587125e-06, "loss": 0.0246, "step": 120050 }, { "epoch": 2.8, "learning_rate": 9.05347210814109e-06, "loss": 0.0036, "step": 120055 }, { "epoch": 2.8, "learning_rate": 9.052592053695055e-06, "loss": 0.1694, "step": 120060 }, { "epoch": 2.8, "learning_rate": 9.05171199924902e-06, "loss": 0.0466, "step": 120065 }, { "epoch": 2.8, "learning_rate": 9.050831944802985e-06, "loss": 0.1829, "step": 120070 }, { "epoch": 2.8, "learning_rate": 9.04995189035695e-06, "loss": 0.1503, "step": 120075 }, { "epoch": 2.8, "learning_rate": 9.049071835910915e-06, "loss": 0.0792, "step": 120080 }, { "epoch": 2.8, "learning_rate": 9.04819178146488e-06, "loss": 0.0165, "step": 120085 }, { "epoch": 2.8, "learning_rate": 9.047311727018844e-06, "loss": 0.064, "step": 120090 }, { "epoch": 2.8, "learning_rate": 9.04643167257281e-06, "loss": 0.1062, "step": 120095 }, { "epoch": 2.8, "learning_rate": 9.045551618126774e-06, "loss": 0.0046, "step": 120100 }, { "epoch": 2.8, "learning_rate": 9.04467156368074e-06, "loss": 0.0256, "step": 120105 }, { "epoch": 2.8, "learning_rate": 9.043791509234705e-06, "loss": 0.1255, "step": 120110 }, { "epoch": 2.8, "learning_rate": 9.04291145478867e-06, "loss": 0.1784, "step": 120115 }, { "epoch": 2.8, "learning_rate": 9.042031400342635e-06, "loss": 0.2463, "step": 120120 }, { "epoch": 2.8, "learning_rate": 9.0411513458966e-06, "loss": 0.4342, "step": 120125 }, { "epoch": 2.8, "learning_rate": 9.040271291450565e-06, "loss": 0.0193, "step": 120130 }, { "epoch": 2.8, "learning_rate": 9.039391237004529e-06, "loss": 0.0211, "step": 120135 }, { "epoch": 2.8, "learning_rate": 9.038511182558495e-06, "loss": 0.0858, "step": 120140 }, { "epoch": 2.8, "learning_rate": 9.037631128112459e-06, "loss": 0.0064, "step": 120145 }, { "epoch": 2.8, "learning_rate": 9.036751073666426e-06, "loss": 0.0342, "step": 120150 }, { "epoch": 2.8, "learning_rate": 9.035871019220389e-06, "loss": 0.0593, "step": 120155 }, { "epoch": 2.8, "learning_rate": 9.034990964774356e-06, "loss": 0.0763, "step": 120160 }, { "epoch": 2.8, "learning_rate": 9.034110910328319e-06, "loss": 0.0641, "step": 120165 }, { "epoch": 2.8, "learning_rate": 9.033230855882284e-06, "loss": 0.1218, "step": 120170 }, { "epoch": 2.8, "learning_rate": 9.032350801436249e-06, "loss": 0.3572, "step": 120175 }, { "epoch": 2.8, "learning_rate": 9.031470746990214e-06, "loss": 0.0061, "step": 120180 }, { "epoch": 2.8, "learning_rate": 9.030590692544179e-06, "loss": 0.0005, "step": 120185 }, { "epoch": 2.8, "learning_rate": 9.029710638098144e-06, "loss": 0.106, "step": 120190 }, { "epoch": 2.8, "learning_rate": 9.028830583652109e-06, "loss": 0.0688, "step": 120195 }, { "epoch": 2.8, "learning_rate": 9.027950529206074e-06, "loss": 0.1128, "step": 120200 }, { "epoch": 2.8, "learning_rate": 9.027070474760037e-06, "loss": 0.0647, "step": 120205 }, { "epoch": 2.8, "learning_rate": 9.026190420314004e-06, "loss": 0.1209, "step": 120210 }, { "epoch": 2.8, "learning_rate": 9.025310365867967e-06, "loss": 0.1928, "step": 120215 }, { "epoch": 2.8, "learning_rate": 9.024430311421934e-06, "loss": 0.277, "step": 120220 }, { "epoch": 2.8, "learning_rate": 9.023550256975897e-06, "loss": 0.1439, "step": 120225 }, { "epoch": 2.81, "learning_rate": 9.022670202529864e-06, "loss": 0.0422, "step": 120230 }, { "epoch": 2.81, "learning_rate": 9.021790148083827e-06, "loss": 0.0307, "step": 120235 }, { "epoch": 2.81, "learning_rate": 9.020910093637792e-06, "loss": 0.0477, "step": 120240 }, { "epoch": 2.81, "learning_rate": 9.02003003919176e-06, "loss": 0.1344, "step": 120245 }, { "epoch": 2.81, "learning_rate": 9.019149984745723e-06, "loss": 0.073, "step": 120250 }, { "epoch": 2.81, "learning_rate": 9.01826993029969e-06, "loss": 0.031, "step": 120255 }, { "epoch": 2.81, "learning_rate": 9.017389875853653e-06, "loss": 0.0658, "step": 120260 }, { "epoch": 2.81, "learning_rate": 9.01650982140762e-06, "loss": 0.2412, "step": 120265 }, { "epoch": 2.81, "learning_rate": 9.015629766961583e-06, "loss": 0.3973, "step": 120270 }, { "epoch": 2.81, "learning_rate": 9.014749712515548e-06, "loss": 0.2426, "step": 120275 }, { "epoch": 2.81, "learning_rate": 9.013869658069513e-06, "loss": 0.0343, "step": 120280 }, { "epoch": 2.81, "learning_rate": 9.012989603623478e-06, "loss": 0.0214, "step": 120285 }, { "epoch": 2.81, "learning_rate": 9.012109549177443e-06, "loss": 0.0185, "step": 120290 }, { "epoch": 2.81, "learning_rate": 9.011229494731408e-06, "loss": 0.1694, "step": 120295 }, { "epoch": 2.81, "learning_rate": 9.010349440285373e-06, "loss": 0.0157, "step": 120300 }, { "epoch": 2.81, "learning_rate": 9.009469385839338e-06, "loss": 0.0658, "step": 120305 }, { "epoch": 2.81, "learning_rate": 9.008589331393303e-06, "loss": 0.1439, "step": 120310 }, { "epoch": 2.81, "learning_rate": 9.007709276947268e-06, "loss": 0.0723, "step": 120315 }, { "epoch": 2.81, "learning_rate": 9.006829222501231e-06, "loss": 0.1072, "step": 120320 }, { "epoch": 2.81, "learning_rate": 9.005949168055198e-06, "loss": 0.2446, "step": 120325 }, { "epoch": 2.81, "learning_rate": 9.005069113609161e-06, "loss": 0.0389, "step": 120330 }, { "epoch": 2.81, "learning_rate": 9.004189059163128e-06, "loss": 0.0297, "step": 120335 }, { "epoch": 2.81, "learning_rate": 9.003309004717091e-06, "loss": 0.0237, "step": 120340 }, { "epoch": 2.81, "learning_rate": 9.002428950271058e-06, "loss": 0.0457, "step": 120345 }, { "epoch": 2.81, "learning_rate": 9.001548895825021e-06, "loss": 0.163, "step": 120350 }, { "epoch": 2.81, "learning_rate": 9.000668841378986e-06, "loss": 0.0376, "step": 120355 }, { "epoch": 2.81, "learning_rate": 8.999788786932951e-06, "loss": 0.0312, "step": 120360 }, { "epoch": 2.81, "learning_rate": 8.998908732486916e-06, "loss": 0.1177, "step": 120365 }, { "epoch": 2.81, "learning_rate": 8.998028678040881e-06, "loss": 0.0548, "step": 120370 }, { "epoch": 2.81, "learning_rate": 8.997148623594846e-06, "loss": 0.1247, "step": 120375 }, { "epoch": 2.81, "learning_rate": 8.996268569148813e-06, "loss": 0.0076, "step": 120380 }, { "epoch": 2.81, "learning_rate": 8.995388514702776e-06, "loss": 0.0392, "step": 120385 }, { "epoch": 2.81, "learning_rate": 8.994508460256741e-06, "loss": 0.0227, "step": 120390 }, { "epoch": 2.81, "learning_rate": 8.993628405810706e-06, "loss": 0.0143, "step": 120395 }, { "epoch": 2.81, "learning_rate": 8.992748351364671e-06, "loss": 0.1728, "step": 120400 }, { "epoch": 2.81, "learning_rate": 8.991868296918636e-06, "loss": 0.0659, "step": 120405 }, { "epoch": 2.81, "learning_rate": 8.990988242472601e-06, "loss": 0.1915, "step": 120410 }, { "epoch": 2.81, "learning_rate": 8.990108188026567e-06, "loss": 0.0679, "step": 120415 }, { "epoch": 2.81, "learning_rate": 8.989228133580532e-06, "loss": 0.3837, "step": 120420 }, { "epoch": 2.81, "learning_rate": 8.988348079134495e-06, "loss": 0.3545, "step": 120425 }, { "epoch": 2.81, "learning_rate": 8.987468024688462e-06, "loss": 0.087, "step": 120430 }, { "epoch": 2.81, "learning_rate": 8.986587970242425e-06, "loss": 0.0253, "step": 120435 }, { "epoch": 2.81, "learning_rate": 8.985707915796392e-06, "loss": 0.0077, "step": 120440 }, { "epoch": 2.81, "learning_rate": 8.984827861350355e-06, "loss": 0.0127, "step": 120445 }, { "epoch": 2.81, "learning_rate": 8.983947806904322e-06, "loss": 0.0466, "step": 120450 }, { "epoch": 2.81, "learning_rate": 8.983067752458285e-06, "loss": 0.0229, "step": 120455 }, { "epoch": 2.81, "learning_rate": 8.982187698012252e-06, "loss": 0.09, "step": 120460 }, { "epoch": 2.81, "learning_rate": 8.981307643566215e-06, "loss": 0.1052, "step": 120465 }, { "epoch": 2.81, "learning_rate": 8.98042758912018e-06, "loss": 0.0459, "step": 120470 }, { "epoch": 2.81, "learning_rate": 8.979547534674145e-06, "loss": 0.1514, "step": 120475 }, { "epoch": 2.81, "learning_rate": 8.97866748022811e-06, "loss": 0.0139, "step": 120480 }, { "epoch": 2.81, "learning_rate": 8.977787425782075e-06, "loss": 0.0005, "step": 120485 }, { "epoch": 2.81, "learning_rate": 8.97690737133604e-06, "loss": 0.0104, "step": 120490 }, { "epoch": 2.81, "learning_rate": 8.976027316890005e-06, "loss": 0.0041, "step": 120495 }, { "epoch": 2.81, "learning_rate": 8.97514726244397e-06, "loss": 0.1136, "step": 120500 }, { "epoch": 2.81, "learning_rate": 8.974267207997933e-06, "loss": 0.0546, "step": 120505 }, { "epoch": 2.81, "learning_rate": 8.9733871535519e-06, "loss": 0.1204, "step": 120510 }, { "epoch": 2.81, "learning_rate": 8.972507099105865e-06, "loss": 0.1076, "step": 120515 }, { "epoch": 2.81, "learning_rate": 8.97162704465983e-06, "loss": 0.2202, "step": 120520 }, { "epoch": 2.81, "learning_rate": 8.970746990213795e-06, "loss": 0.1676, "step": 120525 }, { "epoch": 2.81, "learning_rate": 8.96986693576776e-06, "loss": 0.0582, "step": 120530 }, { "epoch": 2.81, "learning_rate": 8.968986881321725e-06, "loss": 0.0484, "step": 120535 }, { "epoch": 2.81, "learning_rate": 8.968106826875689e-06, "loss": 0.0575, "step": 120540 }, { "epoch": 2.81, "learning_rate": 8.967226772429655e-06, "loss": 0.0559, "step": 120545 }, { "epoch": 2.81, "learning_rate": 8.966346717983619e-06, "loss": 0.1104, "step": 120550 }, { "epoch": 2.81, "learning_rate": 8.965466663537585e-06, "loss": 0.1159, "step": 120555 }, { "epoch": 2.81, "learning_rate": 8.964586609091549e-06, "loss": 0.1901, "step": 120560 }, { "epoch": 2.81, "learning_rate": 8.963706554645515e-06, "loss": 0.1477, "step": 120565 }, { "epoch": 2.81, "learning_rate": 8.962826500199479e-06, "loss": 0.1617, "step": 120570 }, { "epoch": 2.81, "learning_rate": 8.961946445753444e-06, "loss": 0.2499, "step": 120575 }, { "epoch": 2.81, "learning_rate": 8.961066391307409e-06, "loss": 0.0297, "step": 120580 }, { "epoch": 2.81, "learning_rate": 8.960186336861374e-06, "loss": 0.0047, "step": 120585 }, { "epoch": 2.81, "learning_rate": 8.959306282415339e-06, "loss": 0.1384, "step": 120590 }, { "epoch": 2.81, "learning_rate": 8.958426227969304e-06, "loss": 0.024, "step": 120595 }, { "epoch": 2.81, "learning_rate": 8.957546173523269e-06, "loss": 0.2126, "step": 120600 }, { "epoch": 2.81, "learning_rate": 8.956666119077234e-06, "loss": 0.0543, "step": 120605 }, { "epoch": 2.81, "learning_rate": 8.955786064631199e-06, "loss": 0.0177, "step": 120610 }, { "epoch": 2.81, "learning_rate": 8.954906010185164e-06, "loss": 0.1114, "step": 120615 }, { "epoch": 2.81, "learning_rate": 8.954025955739127e-06, "loss": 0.0648, "step": 120620 }, { "epoch": 2.81, "learning_rate": 8.953145901293094e-06, "loss": 0.128, "step": 120625 }, { "epoch": 2.81, "learning_rate": 8.952265846847057e-06, "loss": 0.0685, "step": 120630 }, { "epoch": 2.81, "learning_rate": 8.951385792401024e-06, "loss": 0.0551, "step": 120635 }, { "epoch": 2.81, "learning_rate": 8.950505737954987e-06, "loss": 0.001, "step": 120640 }, { "epoch": 2.81, "learning_rate": 8.949625683508954e-06, "loss": 0.0024, "step": 120645 }, { "epoch": 2.81, "learning_rate": 8.948745629062919e-06, "loss": 0.1113, "step": 120650 }, { "epoch": 2.82, "learning_rate": 8.947865574616882e-06, "loss": 0.0645, "step": 120655 }, { "epoch": 2.82, "learning_rate": 8.946985520170849e-06, "loss": 0.0255, "step": 120660 }, { "epoch": 2.82, "learning_rate": 8.946105465724812e-06, "loss": 0.0253, "step": 120665 }, { "epoch": 2.82, "learning_rate": 8.94522541127878e-06, "loss": 0.3234, "step": 120670 }, { "epoch": 2.82, "learning_rate": 8.944345356832743e-06, "loss": 0.1666, "step": 120675 }, { "epoch": 2.82, "learning_rate": 8.94346530238671e-06, "loss": 0.1612, "step": 120680 }, { "epoch": 2.82, "learning_rate": 8.942585247940673e-06, "loss": 0.0424, "step": 120685 }, { "epoch": 2.82, "learning_rate": 8.941705193494638e-06, "loss": 0.0375, "step": 120690 }, { "epoch": 2.82, "learning_rate": 8.940825139048603e-06, "loss": 0.0036, "step": 120695 }, { "epoch": 2.82, "learning_rate": 8.939945084602568e-06, "loss": 0.029, "step": 120700 }, { "epoch": 2.82, "learning_rate": 8.939065030156533e-06, "loss": 0.0775, "step": 120705 }, { "epoch": 2.82, "learning_rate": 8.938184975710498e-06, "loss": 0.2275, "step": 120710 }, { "epoch": 2.82, "learning_rate": 8.937304921264463e-06, "loss": 0.1836, "step": 120715 }, { "epoch": 2.82, "learning_rate": 8.936424866818428e-06, "loss": 0.1517, "step": 120720 }, { "epoch": 2.82, "learning_rate": 8.935544812372391e-06, "loss": 0.2422, "step": 120725 }, { "epoch": 2.82, "learning_rate": 8.934664757926358e-06, "loss": 0.0152, "step": 120730 }, { "epoch": 2.82, "learning_rate": 8.933784703480321e-06, "loss": 0.0426, "step": 120735 }, { "epoch": 2.82, "learning_rate": 8.932904649034288e-06, "loss": 0.0108, "step": 120740 }, { "epoch": 2.82, "learning_rate": 8.932024594588251e-06, "loss": 0.1685, "step": 120745 }, { "epoch": 2.82, "learning_rate": 8.931144540142218e-06, "loss": 0.0053, "step": 120750 }, { "epoch": 2.82, "learning_rate": 8.930264485696181e-06, "loss": 0.2044, "step": 120755 }, { "epoch": 2.82, "learning_rate": 8.929384431250148e-06, "loss": 0.0578, "step": 120760 }, { "epoch": 2.82, "learning_rate": 8.928504376804111e-06, "loss": 0.0542, "step": 120765 }, { "epoch": 2.82, "learning_rate": 8.927624322358076e-06, "loss": 0.0999, "step": 120770 }, { "epoch": 2.82, "learning_rate": 8.926744267912041e-06, "loss": 0.1537, "step": 120775 }, { "epoch": 2.82, "learning_rate": 8.925864213466006e-06, "loss": 0.0835, "step": 120780 }, { "epoch": 2.82, "learning_rate": 8.924984159019973e-06, "loss": 0.03, "step": 120785 }, { "epoch": 2.82, "learning_rate": 8.924104104573936e-06, "loss": 0.112, "step": 120790 }, { "epoch": 2.82, "learning_rate": 8.923224050127903e-06, "loss": 0.0291, "step": 120795 }, { "epoch": 2.82, "learning_rate": 8.922343995681866e-06, "loss": 0.082, "step": 120800 }, { "epoch": 2.82, "learning_rate": 8.921463941235831e-06, "loss": 0.1285, "step": 120805 }, { "epoch": 2.82, "learning_rate": 8.920583886789796e-06, "loss": 0.1502, "step": 120810 }, { "epoch": 2.82, "learning_rate": 8.919703832343761e-06, "loss": 0.2945, "step": 120815 }, { "epoch": 2.82, "learning_rate": 8.918823777897726e-06, "loss": 0.1655, "step": 120820 }, { "epoch": 2.82, "learning_rate": 8.917943723451691e-06, "loss": 0.1411, "step": 120825 }, { "epoch": 2.82, "learning_rate": 8.917063669005656e-06, "loss": 0.0242, "step": 120830 }, { "epoch": 2.82, "learning_rate": 8.916183614559621e-06, "loss": 0.083, "step": 120835 }, { "epoch": 2.82, "learning_rate": 8.915303560113585e-06, "loss": 0.0227, "step": 120840 }, { "epoch": 2.82, "learning_rate": 8.914423505667552e-06, "loss": 0.1137, "step": 120845 }, { "epoch": 2.82, "learning_rate": 8.913543451221515e-06, "loss": 0.1022, "step": 120850 }, { "epoch": 2.82, "learning_rate": 8.912663396775482e-06, "loss": 0.0571, "step": 120855 }, { "epoch": 2.82, "learning_rate": 8.911783342329445e-06, "loss": 0.0944, "step": 120860 }, { "epoch": 2.82, "learning_rate": 8.910903287883412e-06, "loss": 0.223, "step": 120865 }, { "epoch": 2.82, "learning_rate": 8.910023233437375e-06, "loss": 0.1991, "step": 120870 }, { "epoch": 2.82, "learning_rate": 8.90914317899134e-06, "loss": 0.1954, "step": 120875 }, { "epoch": 2.82, "learning_rate": 8.908263124545305e-06, "loss": 0.0686, "step": 120880 }, { "epoch": 2.82, "learning_rate": 8.90738307009927e-06, "loss": 0.0107, "step": 120885 }, { "epoch": 2.82, "learning_rate": 8.906503015653235e-06, "loss": 0.1183, "step": 120890 }, { "epoch": 2.82, "learning_rate": 8.9056229612072e-06, "loss": 0.0373, "step": 120895 }, { "epoch": 2.82, "learning_rate": 8.904742906761165e-06, "loss": 0.1176, "step": 120900 }, { "epoch": 2.82, "learning_rate": 8.90386285231513e-06, "loss": 0.1983, "step": 120905 }, { "epoch": 2.82, "learning_rate": 8.902982797869095e-06, "loss": 0.0295, "step": 120910 }, { "epoch": 2.82, "learning_rate": 8.90210274342306e-06, "loss": 0.1617, "step": 120915 }, { "epoch": 2.82, "learning_rate": 8.901222688977025e-06, "loss": 0.3221, "step": 120920 }, { "epoch": 2.82, "learning_rate": 8.90034263453099e-06, "loss": 0.2433, "step": 120925 }, { "epoch": 2.82, "learning_rate": 8.899462580084955e-06, "loss": 0.0107, "step": 120930 }, { "epoch": 2.82, "learning_rate": 8.89858252563892e-06, "loss": 0.0141, "step": 120935 }, { "epoch": 2.82, "learning_rate": 8.897702471192885e-06, "loss": 0.0426, "step": 120940 }, { "epoch": 2.82, "learning_rate": 8.89682241674685e-06, "loss": 0.0276, "step": 120945 }, { "epoch": 2.82, "learning_rate": 8.895942362300815e-06, "loss": 0.0743, "step": 120950 }, { "epoch": 2.82, "learning_rate": 8.895062307854779e-06, "loss": 0.0604, "step": 120955 }, { "epoch": 2.82, "learning_rate": 8.894182253408745e-06, "loss": 0.0522, "step": 120960 }, { "epoch": 2.82, "learning_rate": 8.893302198962709e-06, "loss": 0.2484, "step": 120965 }, { "epoch": 2.82, "learning_rate": 8.892422144516675e-06, "loss": 0.1346, "step": 120970 }, { "epoch": 2.82, "learning_rate": 8.891542090070639e-06, "loss": 0.1314, "step": 120975 }, { "epoch": 2.82, "learning_rate": 8.890662035624605e-06, "loss": 0.0517, "step": 120980 }, { "epoch": 2.82, "learning_rate": 8.889781981178569e-06, "loss": 0.0249, "step": 120985 }, { "epoch": 2.82, "learning_rate": 8.888901926732534e-06, "loss": 0.0106, "step": 120990 }, { "epoch": 2.82, "learning_rate": 8.888021872286499e-06, "loss": 0.0645, "step": 120995 }, { "epoch": 2.82, "learning_rate": 8.887141817840464e-06, "loss": 0.0383, "step": 121000 }, { "epoch": 2.82, "learning_rate": 8.886261763394429e-06, "loss": 0.0516, "step": 121005 }, { "epoch": 2.82, "learning_rate": 8.885381708948394e-06, "loss": 0.1535, "step": 121010 }, { "epoch": 2.82, "learning_rate": 8.884501654502359e-06, "loss": 0.0356, "step": 121015 }, { "epoch": 2.82, "learning_rate": 8.883621600056324e-06, "loss": 0.1734, "step": 121020 }, { "epoch": 2.82, "learning_rate": 8.882741545610287e-06, "loss": 0.2326, "step": 121025 }, { "epoch": 2.82, "learning_rate": 8.881861491164254e-06, "loss": 0.01, "step": 121030 }, { "epoch": 2.82, "learning_rate": 8.880981436718217e-06, "loss": 0.0001, "step": 121035 }, { "epoch": 2.82, "learning_rate": 8.880101382272184e-06, "loss": 0.028, "step": 121040 }, { "epoch": 2.82, "learning_rate": 8.879221327826147e-06, "loss": 0.1185, "step": 121045 }, { "epoch": 2.82, "learning_rate": 8.878341273380114e-06, "loss": 0.0116, "step": 121050 }, { "epoch": 2.82, "learning_rate": 8.877461218934079e-06, "loss": 0.0057, "step": 121055 }, { "epoch": 2.82, "learning_rate": 8.876581164488044e-06, "loss": 0.0869, "step": 121060 }, { "epoch": 2.82, "learning_rate": 8.875701110042009e-06, "loss": 0.1468, "step": 121065 }, { "epoch": 2.82, "learning_rate": 8.874821055595972e-06, "loss": 0.2061, "step": 121070 }, { "epoch": 2.82, "learning_rate": 8.873941001149939e-06, "loss": 0.4409, "step": 121075 }, { "epoch": 2.82, "learning_rate": 8.873060946703902e-06, "loss": 0.0867, "step": 121080 }, { "epoch": 2.83, "learning_rate": 8.872180892257869e-06, "loss": 0.0318, "step": 121085 }, { "epoch": 2.83, "learning_rate": 8.871300837811832e-06, "loss": 0.1019, "step": 121090 }, { "epoch": 2.83, "learning_rate": 8.870420783365799e-06, "loss": 0.1072, "step": 121095 }, { "epoch": 2.83, "learning_rate": 8.869540728919762e-06, "loss": 0.0205, "step": 121100 }, { "epoch": 2.83, "learning_rate": 8.868660674473727e-06, "loss": 0.0594, "step": 121105 }, { "epoch": 2.83, "learning_rate": 8.867780620027693e-06, "loss": 0.0703, "step": 121110 }, { "epoch": 2.83, "learning_rate": 8.866900565581658e-06, "loss": 0.1279, "step": 121115 }, { "epoch": 2.83, "learning_rate": 8.866020511135623e-06, "loss": 0.1752, "step": 121120 }, { "epoch": 2.83, "learning_rate": 8.865140456689588e-06, "loss": 0.3695, "step": 121125 }, { "epoch": 2.83, "learning_rate": 8.864260402243553e-06, "loss": 0.0627, "step": 121130 }, { "epoch": 2.83, "learning_rate": 8.863380347797518e-06, "loss": 0.0017, "step": 121135 }, { "epoch": 2.83, "learning_rate": 8.862500293351481e-06, "loss": 0.0795, "step": 121140 }, { "epoch": 2.83, "learning_rate": 8.861620238905448e-06, "loss": 0.05, "step": 121145 }, { "epoch": 2.83, "learning_rate": 8.860740184459411e-06, "loss": 0.0601, "step": 121150 }, { "epoch": 2.83, "learning_rate": 8.859860130013378e-06, "loss": 0.1879, "step": 121155 }, { "epoch": 2.83, "learning_rate": 8.858980075567341e-06, "loss": 0.0927, "step": 121160 }, { "epoch": 2.83, "learning_rate": 8.858100021121308e-06, "loss": 0.2991, "step": 121165 }, { "epoch": 2.83, "learning_rate": 8.857219966675271e-06, "loss": 0.1715, "step": 121170 }, { "epoch": 2.83, "learning_rate": 8.856339912229236e-06, "loss": 0.2841, "step": 121175 }, { "epoch": 2.83, "learning_rate": 8.855459857783201e-06, "loss": 0.1043, "step": 121180 }, { "epoch": 2.83, "learning_rate": 8.854579803337166e-06, "loss": 0.0461, "step": 121185 }, { "epoch": 2.83, "learning_rate": 8.853699748891131e-06, "loss": 0.001, "step": 121190 }, { "epoch": 2.83, "learning_rate": 8.852819694445096e-06, "loss": 0.0544, "step": 121195 }, { "epoch": 2.83, "learning_rate": 8.851939639999063e-06, "loss": 0.0729, "step": 121200 }, { "epoch": 2.83, "learning_rate": 8.851059585553026e-06, "loss": 0.1749, "step": 121205 }, { "epoch": 2.83, "learning_rate": 8.850179531106993e-06, "loss": 0.1644, "step": 121210 }, { "epoch": 2.83, "learning_rate": 8.849299476660956e-06, "loss": 0.1094, "step": 121215 }, { "epoch": 2.83, "learning_rate": 8.848595433104128e-06, "loss": 0.2287, "step": 121220 }, { "epoch": 2.83, "learning_rate": 8.847715378658093e-06, "loss": 0.193, "step": 121225 }, { "epoch": 2.83, "learning_rate": 8.846835324212058e-06, "loss": 0.0054, "step": 121230 }, { "epoch": 2.83, "learning_rate": 8.845955269766023e-06, "loss": 0.0166, "step": 121235 }, { "epoch": 2.83, "learning_rate": 8.845075215319988e-06, "loss": 0.0095, "step": 121240 }, { "epoch": 2.83, "learning_rate": 8.844195160873953e-06, "loss": 0.0023, "step": 121245 }, { "epoch": 2.83, "learning_rate": 8.843315106427918e-06, "loss": 0.0414, "step": 121250 }, { "epoch": 2.83, "learning_rate": 8.842435051981884e-06, "loss": 0.0515, "step": 121255 }, { "epoch": 2.83, "learning_rate": 8.841554997535848e-06, "loss": 0.0466, "step": 121260 }, { "epoch": 2.83, "learning_rate": 8.840674943089813e-06, "loss": 0.1382, "step": 121265 }, { "epoch": 2.83, "learning_rate": 8.839794888643778e-06, "loss": 0.1964, "step": 121270 }, { "epoch": 2.83, "learning_rate": 8.838914834197743e-06, "loss": 0.1385, "step": 121275 }, { "epoch": 2.83, "learning_rate": 8.838034779751708e-06, "loss": 0.0398, "step": 121280 }, { "epoch": 2.83, "learning_rate": 8.837154725305673e-06, "loss": 0.066, "step": 121285 }, { "epoch": 2.83, "learning_rate": 8.836274670859638e-06, "loss": 0.0539, "step": 121290 }, { "epoch": 2.83, "learning_rate": 8.835394616413603e-06, "loss": 0.1055, "step": 121295 }, { "epoch": 2.83, "learning_rate": 8.834514561967566e-06, "loss": 0.0007, "step": 121300 }, { "epoch": 2.83, "learning_rate": 8.833634507521533e-06, "loss": 0.1483, "step": 121305 }, { "epoch": 2.83, "learning_rate": 8.832754453075496e-06, "loss": 0.2143, "step": 121310 }, { "epoch": 2.83, "learning_rate": 8.831874398629463e-06, "loss": 0.23, "step": 121315 }, { "epoch": 2.83, "learning_rate": 8.830994344183426e-06, "loss": 0.1411, "step": 121320 }, { "epoch": 2.83, "learning_rate": 8.830114289737393e-06, "loss": 0.1224, "step": 121325 }, { "epoch": 2.83, "learning_rate": 8.829234235291356e-06, "loss": 0.0364, "step": 121330 }, { "epoch": 2.83, "learning_rate": 8.828354180845321e-06, "loss": 0.0302, "step": 121335 }, { "epoch": 2.83, "learning_rate": 8.827474126399286e-06, "loss": 0.0002, "step": 121340 }, { "epoch": 2.83, "learning_rate": 8.826594071953251e-06, "loss": 0.1133, "step": 121345 }, { "epoch": 2.83, "learning_rate": 8.825714017507216e-06, "loss": 0.0836, "step": 121350 }, { "epoch": 2.83, "learning_rate": 8.824833963061181e-06, "loss": 0.0565, "step": 121355 }, { "epoch": 2.83, "learning_rate": 8.823953908615146e-06, "loss": 0.0822, "step": 121360 }, { "epoch": 2.83, "learning_rate": 8.823073854169111e-06, "loss": 0.0561, "step": 121365 }, { "epoch": 2.83, "learning_rate": 8.822193799723075e-06, "loss": 0.1383, "step": 121370 }, { "epoch": 2.83, "learning_rate": 8.821313745277042e-06, "loss": 0.1077, "step": 121375 }, { "epoch": 2.83, "learning_rate": 8.820433690831007e-06, "loss": 0.0142, "step": 121380 }, { "epoch": 2.83, "learning_rate": 8.819553636384972e-06, "loss": 0.0337, "step": 121385 }, { "epoch": 2.83, "learning_rate": 8.818673581938937e-06, "loss": 0.049, "step": 121390 }, { "epoch": 2.83, "learning_rate": 8.817793527492902e-06, "loss": 0.0096, "step": 121395 }, { "epoch": 2.83, "learning_rate": 8.816913473046867e-06, "loss": 0.0856, "step": 121400 }, { "epoch": 2.83, "learning_rate": 8.816033418600832e-06, "loss": 0.1131, "step": 121405 }, { "epoch": 2.83, "learning_rate": 8.815153364154797e-06, "loss": 0.0811, "step": 121410 }, { "epoch": 2.83, "learning_rate": 8.81427330970876e-06, "loss": 0.0671, "step": 121415 }, { "epoch": 2.83, "learning_rate": 8.813393255262727e-06, "loss": 0.2519, "step": 121420 }, { "epoch": 2.83, "learning_rate": 8.81251320081669e-06, "loss": 0.1323, "step": 121425 }, { "epoch": 2.83, "learning_rate": 8.811633146370657e-06, "loss": 0.0936, "step": 121430 }, { "epoch": 2.83, "learning_rate": 8.81075309192462e-06, "loss": 0.0611, "step": 121435 }, { "epoch": 2.83, "learning_rate": 8.809873037478587e-06, "loss": 0.0202, "step": 121440 }, { "epoch": 2.83, "learning_rate": 8.80899298303255e-06, "loss": 0.0989, "step": 121445 }, { "epoch": 2.83, "learning_rate": 8.808112928586515e-06, "loss": 0.0628, "step": 121450 }, { "epoch": 2.83, "learning_rate": 8.80723287414048e-06, "loss": 0.0197, "step": 121455 }, { "epoch": 2.83, "learning_rate": 8.806352819694445e-06, "loss": 0.172, "step": 121460 }, { "epoch": 2.83, "learning_rate": 8.80547276524841e-06, "loss": 0.1097, "step": 121465 }, { "epoch": 2.83, "learning_rate": 8.804592710802375e-06, "loss": 0.1487, "step": 121470 }, { "epoch": 2.83, "learning_rate": 8.80371265635634e-06, "loss": 0.11, "step": 121475 }, { "epoch": 2.83, "learning_rate": 8.802832601910305e-06, "loss": 0.0648, "step": 121480 }, { "epoch": 2.83, "learning_rate": 8.801952547464269e-06, "loss": 0.057, "step": 121485 }, { "epoch": 2.83, "learning_rate": 8.801072493018235e-06, "loss": 0.0088, "step": 121490 }, { "epoch": 2.83, "learning_rate": 8.800192438572199e-06, "loss": 0.022, "step": 121495 }, { "epoch": 2.83, "learning_rate": 8.799312384126165e-06, "loss": 0.044, "step": 121500 }, { "epoch": 2.83, "learning_rate": 8.798432329680129e-06, "loss": 0.1869, "step": 121505 }, { "epoch": 2.83, "learning_rate": 8.797552275234095e-06, "loss": 0.0654, "step": 121510 }, { "epoch": 2.84, "learning_rate": 8.796672220788059e-06, "loss": 0.0953, "step": 121515 }, { "epoch": 2.84, "learning_rate": 8.795792166342024e-06, "loss": 0.0909, "step": 121520 }, { "epoch": 2.84, "learning_rate": 8.79491211189599e-06, "loss": 0.2157, "step": 121525 }, { "epoch": 2.84, "learning_rate": 8.794032057449954e-06, "loss": 0.0585, "step": 121530 }, { "epoch": 2.84, "learning_rate": 8.79315200300392e-06, "loss": 0.0026, "step": 121535 }, { "epoch": 2.84, "learning_rate": 8.792271948557884e-06, "loss": 0.1015, "step": 121540 }, { "epoch": 2.84, "learning_rate": 8.79139189411185e-06, "loss": 0.1086, "step": 121545 }, { "epoch": 2.84, "learning_rate": 8.790511839665814e-06, "loss": 0.0252, "step": 121550 }, { "epoch": 2.84, "learning_rate": 8.78963178521978e-06, "loss": 0.1944, "step": 121555 }, { "epoch": 2.84, "learning_rate": 8.788751730773744e-06, "loss": 0.0935, "step": 121560 }, { "epoch": 2.84, "learning_rate": 8.787871676327709e-06, "loss": 0.1864, "step": 121565 }, { "epoch": 2.84, "learning_rate": 8.786991621881674e-06, "loss": 0.156, "step": 121570 }, { "epoch": 2.84, "learning_rate": 8.786111567435639e-06, "loss": 0.1997, "step": 121575 }, { "epoch": 2.84, "learning_rate": 8.785231512989604e-06, "loss": 0.0026, "step": 121580 }, { "epoch": 2.84, "learning_rate": 8.784351458543569e-06, "loss": 0.0062, "step": 121585 }, { "epoch": 2.84, "learning_rate": 8.783471404097534e-06, "loss": 0.0314, "step": 121590 }, { "epoch": 2.84, "learning_rate": 8.782591349651499e-06, "loss": 0.0172, "step": 121595 }, { "epoch": 2.84, "learning_rate": 8.781711295205462e-06, "loss": 0.0746, "step": 121600 }, { "epoch": 2.84, "learning_rate": 8.780831240759429e-06, "loss": 0.1069, "step": 121605 }, { "epoch": 2.84, "learning_rate": 8.779951186313392e-06, "loss": 0.0356, "step": 121610 }, { "epoch": 2.84, "learning_rate": 8.779071131867359e-06, "loss": 0.2639, "step": 121615 }, { "epoch": 2.84, "learning_rate": 8.778191077421322e-06, "loss": 0.2645, "step": 121620 }, { "epoch": 2.84, "learning_rate": 8.77731102297529e-06, "loss": 0.0905, "step": 121625 }, { "epoch": 2.84, "learning_rate": 8.776430968529252e-06, "loss": 0.0319, "step": 121630 }, { "epoch": 2.84, "learning_rate": 8.775550914083218e-06, "loss": 0.0008, "step": 121635 }, { "epoch": 2.84, "learning_rate": 8.774670859637183e-06, "loss": 0.0356, "step": 121640 }, { "epoch": 2.84, "learning_rate": 8.773790805191148e-06, "loss": 0.0159, "step": 121645 }, { "epoch": 2.84, "learning_rate": 8.772910750745113e-06, "loss": 0.0185, "step": 121650 }, { "epoch": 2.84, "learning_rate": 8.772030696299078e-06, "loss": 0.1867, "step": 121655 }, { "epoch": 2.84, "learning_rate": 8.771150641853044e-06, "loss": 0.0322, "step": 121660 }, { "epoch": 2.84, "learning_rate": 8.770270587407008e-06, "loss": 0.1828, "step": 121665 }, { "epoch": 2.84, "learning_rate": 8.769390532960973e-06, "loss": 0.2198, "step": 121670 }, { "epoch": 2.84, "learning_rate": 8.768510478514938e-06, "loss": 0.2245, "step": 121675 }, { "epoch": 2.84, "learning_rate": 8.767630424068903e-06, "loss": 0.075, "step": 121680 }, { "epoch": 2.84, "learning_rate": 8.766750369622868e-06, "loss": 0.0541, "step": 121685 }, { "epoch": 2.84, "learning_rate": 8.765870315176833e-06, "loss": 0.0055, "step": 121690 }, { "epoch": 2.84, "learning_rate": 8.764990260730798e-06, "loss": 0.0539, "step": 121695 }, { "epoch": 2.84, "learning_rate": 8.764110206284763e-06, "loss": 0.0621, "step": 121700 }, { "epoch": 2.84, "learning_rate": 8.763230151838728e-06, "loss": 0.0532, "step": 121705 }, { "epoch": 2.84, "learning_rate": 8.762350097392693e-06, "loss": 0.1625, "step": 121710 }, { "epoch": 2.84, "learning_rate": 8.761470042946656e-06, "loss": 0.0717, "step": 121715 }, { "epoch": 2.84, "learning_rate": 8.760589988500623e-06, "loss": 0.3219, "step": 121720 }, { "epoch": 2.84, "learning_rate": 8.759709934054586e-06, "loss": 0.3973, "step": 121725 }, { "epoch": 2.84, "learning_rate": 8.758829879608553e-06, "loss": 0.0214, "step": 121730 }, { "epoch": 2.84, "learning_rate": 8.757949825162516e-06, "loss": 0.0728, "step": 121735 }, { "epoch": 2.84, "learning_rate": 8.757069770716483e-06, "loss": 0.0326, "step": 121740 }, { "epoch": 2.84, "learning_rate": 8.756189716270446e-06, "loss": 0.0104, "step": 121745 }, { "epoch": 2.84, "learning_rate": 8.755309661824411e-06, "loss": 0.0056, "step": 121750 }, { "epoch": 2.84, "learning_rate": 8.754429607378376e-06, "loss": 0.1203, "step": 121755 }, { "epoch": 2.84, "learning_rate": 8.753549552932341e-06, "loss": 0.0344, "step": 121760 }, { "epoch": 2.84, "learning_rate": 8.752669498486306e-06, "loss": 0.0914, "step": 121765 }, { "epoch": 2.84, "learning_rate": 8.751789444040271e-06, "loss": 0.5579, "step": 121770 }, { "epoch": 2.84, "learning_rate": 8.750909389594236e-06, "loss": 0.3042, "step": 121775 }, { "epoch": 2.84, "learning_rate": 8.750029335148201e-06, "loss": 0.0119, "step": 121780 }, { "epoch": 2.84, "learning_rate": 8.749149280702165e-06, "loss": 0.065, "step": 121785 }, { "epoch": 2.84, "learning_rate": 8.748269226256131e-06, "loss": 0.1014, "step": 121790 }, { "epoch": 2.84, "learning_rate": 8.747389171810096e-06, "loss": 0.03, "step": 121795 }, { "epoch": 2.84, "learning_rate": 8.746509117364061e-06, "loss": 0.0092, "step": 121800 }, { "epoch": 2.84, "learning_rate": 8.745629062918027e-06, "loss": 0.0813, "step": 121805 }, { "epoch": 2.84, "learning_rate": 8.744749008471992e-06, "loss": 0.1562, "step": 121810 }, { "epoch": 2.84, "learning_rate": 8.743868954025957e-06, "loss": 0.0999, "step": 121815 }, { "epoch": 2.84, "learning_rate": 8.74298889957992e-06, "loss": 0.1153, "step": 121820 }, { "epoch": 2.84, "learning_rate": 8.742108845133887e-06, "loss": 0.1761, "step": 121825 }, { "epoch": 2.84, "learning_rate": 8.74122879068785e-06, "loss": 0.0463, "step": 121830 }, { "epoch": 2.84, "learning_rate": 8.740348736241817e-06, "loss": 0.0179, "step": 121835 }, { "epoch": 2.84, "learning_rate": 8.73946868179578e-06, "loss": 0.0146, "step": 121840 }, { "epoch": 2.84, "learning_rate": 8.738588627349747e-06, "loss": 0.0119, "step": 121845 }, { "epoch": 2.84, "learning_rate": 8.73770857290371e-06, "loss": 0.0948, "step": 121850 }, { "epoch": 2.84, "learning_rate": 8.736828518457677e-06, "loss": 0.0791, "step": 121855 }, { "epoch": 2.84, "learning_rate": 8.73594846401164e-06, "loss": 0.107, "step": 121860 }, { "epoch": 2.84, "learning_rate": 8.735068409565605e-06, "loss": 0.1284, "step": 121865 }, { "epoch": 2.84, "learning_rate": 8.73418835511957e-06, "loss": 0.287, "step": 121870 }, { "epoch": 2.84, "learning_rate": 8.733308300673535e-06, "loss": 0.4916, "step": 121875 }, { "epoch": 2.84, "learning_rate": 8.7324282462275e-06, "loss": 0.0232, "step": 121880 }, { "epoch": 2.84, "learning_rate": 8.731548191781465e-06, "loss": 0.0629, "step": 121885 }, { "epoch": 2.84, "learning_rate": 8.73066813733543e-06, "loss": 0.0146, "step": 121890 }, { "epoch": 2.84, "learning_rate": 8.729788082889395e-06, "loss": 0.0091, "step": 121895 }, { "epoch": 2.84, "learning_rate": 8.728908028443359e-06, "loss": 0.1289, "step": 121900 }, { "epoch": 2.84, "learning_rate": 8.728027973997325e-06, "loss": 0.0634, "step": 121905 }, { "epoch": 2.84, "learning_rate": 8.727147919551289e-06, "loss": 0.1406, "step": 121910 }, { "epoch": 2.84, "learning_rate": 8.726267865105255e-06, "loss": 0.1703, "step": 121915 }, { "epoch": 2.84, "learning_rate": 8.725387810659219e-06, "loss": 0.2738, "step": 121920 }, { "epoch": 2.84, "learning_rate": 8.724507756213185e-06, "loss": 0.2161, "step": 121925 }, { "epoch": 2.84, "learning_rate": 8.72362770176715e-06, "loss": 0.0699, "step": 121930 }, { "epoch": 2.84, "learning_rate": 8.722747647321114e-06, "loss": 0.0003, "step": 121935 }, { "epoch": 2.85, "learning_rate": 8.72186759287508e-06, "loss": 0.0093, "step": 121940 }, { "epoch": 2.85, "learning_rate": 8.720987538429044e-06, "loss": 0.0431, "step": 121945 }, { "epoch": 2.85, "learning_rate": 8.72010748398301e-06, "loss": 0.0602, "step": 121950 }, { "epoch": 2.85, "learning_rate": 8.719227429536974e-06, "loss": 0.0994, "step": 121955 }, { "epoch": 2.85, "learning_rate": 8.71834737509094e-06, "loss": 0.113, "step": 121960 }, { "epoch": 2.85, "learning_rate": 8.717467320644904e-06, "loss": 0.0509, "step": 121965 }, { "epoch": 2.85, "learning_rate": 8.716587266198869e-06, "loss": 0.2887, "step": 121970 }, { "epoch": 2.85, "learning_rate": 8.715707211752834e-06, "loss": 0.089, "step": 121975 }, { "epoch": 2.85, "learning_rate": 8.714827157306799e-06, "loss": 0.0672, "step": 121980 }, { "epoch": 2.85, "learning_rate": 8.713947102860764e-06, "loss": 0.0049, "step": 121985 }, { "epoch": 2.85, "learning_rate": 8.713067048414729e-06, "loss": 0.0001, "step": 121990 }, { "epoch": 2.85, "learning_rate": 8.712186993968694e-06, "loss": 0.062, "step": 121995 }, { "epoch": 2.85, "learning_rate": 8.711306939522659e-06, "loss": 0.0378, "step": 122000 }, { "epoch": 2.85, "learning_rate": 8.710426885076624e-06, "loss": 0.0885, "step": 122005 }, { "epoch": 2.85, "learning_rate": 8.709546830630589e-06, "loss": 0.0254, "step": 122010 }, { "epoch": 2.85, "learning_rate": 8.708666776184552e-06, "loss": 0.1415, "step": 122015 }, { "epoch": 2.85, "learning_rate": 8.707786721738519e-06, "loss": 0.0977, "step": 122020 }, { "epoch": 2.85, "learning_rate": 8.706906667292482e-06, "loss": 0.1563, "step": 122025 }, { "epoch": 2.85, "learning_rate": 8.706026612846449e-06, "loss": 0.0265, "step": 122030 }, { "epoch": 2.85, "learning_rate": 8.705146558400412e-06, "loss": 0.0553, "step": 122035 }, { "epoch": 2.85, "learning_rate": 8.704266503954379e-06, "loss": 0.01, "step": 122040 }, { "epoch": 2.85, "learning_rate": 8.703386449508342e-06, "loss": 0.194, "step": 122045 }, { "epoch": 2.85, "learning_rate": 8.702506395062307e-06, "loss": 0.1604, "step": 122050 }, { "epoch": 2.85, "learning_rate": 8.701626340616272e-06, "loss": 0.0013, "step": 122055 }, { "epoch": 2.85, "learning_rate": 8.700746286170237e-06, "loss": 0.2352, "step": 122060 }, { "epoch": 2.85, "learning_rate": 8.699866231724204e-06, "loss": 0.1219, "step": 122065 }, { "epoch": 2.85, "learning_rate": 8.698986177278168e-06, "loss": 0.1864, "step": 122070 }, { "epoch": 2.85, "learning_rate": 8.698106122832134e-06, "loss": 0.2566, "step": 122075 }, { "epoch": 2.85, "learning_rate": 8.697226068386098e-06, "loss": 0.0309, "step": 122080 }, { "epoch": 2.85, "learning_rate": 8.696346013940063e-06, "loss": 0.0472, "step": 122085 }, { "epoch": 2.85, "learning_rate": 8.695465959494028e-06, "loss": 0.0223, "step": 122090 }, { "epoch": 2.85, "learning_rate": 8.694585905047993e-06, "loss": 0.0146, "step": 122095 }, { "epoch": 2.85, "learning_rate": 8.693705850601958e-06, "loss": 0.0704, "step": 122100 }, { "epoch": 2.85, "learning_rate": 8.692825796155923e-06, "loss": 0.0177, "step": 122105 }, { "epoch": 2.85, "learning_rate": 8.691945741709888e-06, "loss": 0.2102, "step": 122110 }, { "epoch": 2.85, "learning_rate": 8.691065687263853e-06, "loss": 0.1542, "step": 122115 }, { "epoch": 2.85, "learning_rate": 8.690185632817816e-06, "loss": 0.2081, "step": 122120 }, { "epoch": 2.85, "learning_rate": 8.689305578371783e-06, "loss": 0.1366, "step": 122125 }, { "epoch": 2.85, "learning_rate": 8.688425523925746e-06, "loss": 0.0122, "step": 122130 }, { "epoch": 2.85, "learning_rate": 8.687545469479713e-06, "loss": 0.096, "step": 122135 }, { "epoch": 2.85, "learning_rate": 8.686665415033676e-06, "loss": 0.0135, "step": 122140 }, { "epoch": 2.85, "learning_rate": 8.685785360587643e-06, "loss": 0.0514, "step": 122145 }, { "epoch": 2.85, "learning_rate": 8.684905306141606e-06, "loss": 0.0984, "step": 122150 }, { "epoch": 2.85, "learning_rate": 8.684025251695573e-06, "loss": 0.0451, "step": 122155 }, { "epoch": 2.85, "learning_rate": 8.683145197249536e-06, "loss": 0.1603, "step": 122160 }, { "epoch": 2.85, "learning_rate": 8.682265142803501e-06, "loss": 0.1573, "step": 122165 }, { "epoch": 2.85, "learning_rate": 8.681385088357466e-06, "loss": 0.274, "step": 122170 }, { "epoch": 2.85, "learning_rate": 8.680505033911431e-06, "loss": 0.2012, "step": 122175 }, { "epoch": 2.85, "learning_rate": 8.679624979465396e-06, "loss": 0.0286, "step": 122180 }, { "epoch": 2.85, "learning_rate": 8.678744925019361e-06, "loss": 0.043, "step": 122185 }, { "epoch": 2.85, "learning_rate": 8.677864870573326e-06, "loss": 0.0212, "step": 122190 }, { "epoch": 2.85, "learning_rate": 8.676984816127291e-06, "loss": 0.0091, "step": 122195 }, { "epoch": 2.85, "learning_rate": 8.676104761681256e-06, "loss": 0.0803, "step": 122200 }, { "epoch": 2.85, "learning_rate": 8.675224707235221e-06, "loss": 0.0271, "step": 122205 }, { "epoch": 2.85, "learning_rate": 8.674344652789186e-06, "loss": 0.0081, "step": 122210 }, { "epoch": 2.85, "learning_rate": 8.673464598343151e-06, "loss": 0.1325, "step": 122215 }, { "epoch": 2.85, "learning_rate": 8.672584543897116e-06, "loss": 0.1246, "step": 122220 }, { "epoch": 2.85, "learning_rate": 8.671704489451081e-06, "loss": 0.1229, "step": 122225 }, { "epoch": 2.85, "learning_rate": 8.670824435005046e-06, "loss": 0.131, "step": 122230 }, { "epoch": 2.85, "learning_rate": 8.66994438055901e-06, "loss": 0.0554, "step": 122235 }, { "epoch": 2.85, "learning_rate": 8.669064326112977e-06, "loss": 0.0427, "step": 122240 }, { "epoch": 2.85, "learning_rate": 8.66818427166694e-06, "loss": 0.0535, "step": 122245 }, { "epoch": 2.85, "learning_rate": 8.667304217220907e-06, "loss": 0.0128, "step": 122250 }, { "epoch": 2.85, "learning_rate": 8.66642416277487e-06, "loss": 0.1533, "step": 122255 }, { "epoch": 2.85, "learning_rate": 8.665544108328837e-06, "loss": 0.1987, "step": 122260 }, { "epoch": 2.85, "learning_rate": 8.6646640538828e-06, "loss": 0.1447, "step": 122265 }, { "epoch": 2.85, "learning_rate": 8.663783999436765e-06, "loss": 0.2823, "step": 122270 }, { "epoch": 2.85, "learning_rate": 8.66290394499073e-06, "loss": 0.0986, "step": 122275 }, { "epoch": 2.85, "learning_rate": 8.662023890544695e-06, "loss": 0.055, "step": 122280 }, { "epoch": 2.85, "learning_rate": 8.66114383609866e-06, "loss": 0.0019, "step": 122285 }, { "epoch": 2.85, "learning_rate": 8.660263781652625e-06, "loss": 0.0068, "step": 122290 }, { "epoch": 2.85, "learning_rate": 8.65938372720659e-06, "loss": 0.0534, "step": 122295 }, { "epoch": 2.85, "learning_rate": 8.658503672760555e-06, "loss": 0.1354, "step": 122300 }, { "epoch": 2.85, "learning_rate": 8.65762361831452e-06, "loss": 0.0481, "step": 122305 }, { "epoch": 2.85, "learning_rate": 8.656743563868485e-06, "loss": 0.1069, "step": 122310 }, { "epoch": 2.85, "learning_rate": 8.655863509422448e-06, "loss": 0.1553, "step": 122315 }, { "epoch": 2.85, "learning_rate": 8.654983454976415e-06, "loss": 0.2001, "step": 122320 }, { "epoch": 2.85, "learning_rate": 8.654103400530378e-06, "loss": 0.2431, "step": 122325 }, { "epoch": 2.85, "learning_rate": 8.653223346084345e-06, "loss": 0.0193, "step": 122330 }, { "epoch": 2.85, "learning_rate": 8.65234329163831e-06, "loss": 0.0021, "step": 122335 }, { "epoch": 2.85, "learning_rate": 8.651463237192275e-06, "loss": 0.0276, "step": 122340 }, { "epoch": 2.85, "learning_rate": 8.65058318274624e-06, "loss": 0.0651, "step": 122345 }, { "epoch": 2.85, "learning_rate": 8.649703128300204e-06, "loss": 0.0653, "step": 122350 }, { "epoch": 2.85, "learning_rate": 8.64882307385417e-06, "loss": 0.0369, "step": 122355 }, { "epoch": 2.85, "learning_rate": 8.647943019408134e-06, "loss": 0.1748, "step": 122360 }, { "epoch": 2.85, "learning_rate": 8.6470629649621e-06, "loss": 0.0774, "step": 122365 }, { "epoch": 2.86, "learning_rate": 8.646182910516064e-06, "loss": 0.1076, "step": 122370 }, { "epoch": 2.86, "learning_rate": 8.64530285607003e-06, "loss": 0.2505, "step": 122375 }, { "epoch": 2.86, "learning_rate": 8.644422801623994e-06, "loss": 0.0053, "step": 122380 }, { "epoch": 2.86, "learning_rate": 8.643542747177959e-06, "loss": 0.022, "step": 122385 }, { "epoch": 2.86, "learning_rate": 8.642662692731924e-06, "loss": 0.0521, "step": 122390 }, { "epoch": 2.86, "learning_rate": 8.641782638285889e-06, "loss": 0.0509, "step": 122395 }, { "epoch": 2.86, "learning_rate": 8.640902583839854e-06, "loss": 0.0405, "step": 122400 }, { "epoch": 2.86, "learning_rate": 8.640022529393819e-06, "loss": 0.0301, "step": 122405 }, { "epoch": 2.86, "learning_rate": 8.639142474947784e-06, "loss": 0.0207, "step": 122410 }, { "epoch": 2.86, "learning_rate": 8.638262420501749e-06, "loss": 0.1583, "step": 122415 }, { "epoch": 2.86, "learning_rate": 8.637382366055712e-06, "loss": 0.1399, "step": 122420 }, { "epoch": 2.86, "learning_rate": 8.636502311609679e-06, "loss": 0.1491, "step": 122425 }, { "epoch": 2.86, "learning_rate": 8.635622257163642e-06, "loss": 0.032, "step": 122430 }, { "epoch": 2.86, "learning_rate": 8.634742202717609e-06, "loss": 0.0644, "step": 122435 }, { "epoch": 2.86, "learning_rate": 8.633862148271572e-06, "loss": 0.0802, "step": 122440 }, { "epoch": 2.86, "learning_rate": 8.632982093825539e-06, "loss": 0.0552, "step": 122445 }, { "epoch": 2.86, "learning_rate": 8.632102039379502e-06, "loss": 0.0642, "step": 122450 }, { "epoch": 2.86, "learning_rate": 8.631221984933469e-06, "loss": 0.0654, "step": 122455 }, { "epoch": 2.86, "learning_rate": 8.630341930487432e-06, "loss": 0.1593, "step": 122460 }, { "epoch": 2.86, "learning_rate": 8.629461876041397e-06, "loss": 0.1992, "step": 122465 }, { "epoch": 2.86, "learning_rate": 8.628581821595364e-06, "loss": 0.2802, "step": 122470 }, { "epoch": 2.86, "learning_rate": 8.627701767149327e-06, "loss": 0.3319, "step": 122475 }, { "epoch": 2.86, "learning_rate": 8.626821712703294e-06, "loss": 0.0144, "step": 122480 }, { "epoch": 2.86, "learning_rate": 8.625941658257257e-06, "loss": 0.023, "step": 122485 }, { "epoch": 2.86, "learning_rate": 8.625061603811224e-06, "loss": 0.0491, "step": 122490 }, { "epoch": 2.86, "learning_rate": 8.624181549365187e-06, "loss": 0.0386, "step": 122495 }, { "epoch": 2.86, "learning_rate": 8.623301494919153e-06, "loss": 0.0669, "step": 122500 }, { "epoch": 2.86, "learning_rate": 8.622421440473118e-06, "loss": 0.0563, "step": 122505 }, { "epoch": 2.86, "learning_rate": 8.621541386027083e-06, "loss": 0.1372, "step": 122510 }, { "epoch": 2.86, "learning_rate": 8.620661331581048e-06, "loss": 0.0432, "step": 122515 }, { "epoch": 2.86, "learning_rate": 8.619781277135013e-06, "loss": 0.2055, "step": 122520 }, { "epoch": 2.86, "learning_rate": 8.618901222688978e-06, "loss": 0.322, "step": 122525 }, { "epoch": 2.86, "learning_rate": 8.618021168242943e-06, "loss": 0.0097, "step": 122530 }, { "epoch": 2.86, "learning_rate": 8.617141113796906e-06, "loss": 0.0129, "step": 122535 }, { "epoch": 2.86, "learning_rate": 8.616261059350873e-06, "loss": 0.0288, "step": 122540 }, { "epoch": 2.86, "learning_rate": 8.615381004904836e-06, "loss": 0.06, "step": 122545 }, { "epoch": 2.86, "learning_rate": 8.614500950458803e-06, "loss": 0.0581, "step": 122550 }, { "epoch": 2.86, "learning_rate": 8.613620896012766e-06, "loss": 0.1267, "step": 122555 }, { "epoch": 2.86, "learning_rate": 8.612740841566733e-06, "loss": 0.2041, "step": 122560 }, { "epoch": 2.86, "learning_rate": 8.611860787120696e-06, "loss": 0.1833, "step": 122565 }, { "epoch": 2.86, "learning_rate": 8.610980732674661e-06, "loss": 0.2163, "step": 122570 }, { "epoch": 2.86, "learning_rate": 8.610100678228626e-06, "loss": 0.2207, "step": 122575 }, { "epoch": 2.86, "learning_rate": 8.609220623782591e-06, "loss": 0.0391, "step": 122580 }, { "epoch": 2.86, "learning_rate": 8.608340569336556e-06, "loss": 0.054, "step": 122585 }, { "epoch": 2.86, "learning_rate": 8.607460514890521e-06, "loss": 0.005, "step": 122590 }, { "epoch": 2.86, "learning_rate": 8.606580460444486e-06, "loss": 0.071, "step": 122595 }, { "epoch": 2.86, "learning_rate": 8.605700405998451e-06, "loss": 0.0778, "step": 122600 }, { "epoch": 2.86, "learning_rate": 8.604820351552416e-06, "loss": 0.129, "step": 122605 }, { "epoch": 2.86, "learning_rate": 8.603940297106381e-06, "loss": 0.0391, "step": 122610 }, { "epoch": 2.86, "learning_rate": 8.603060242660346e-06, "loss": 0.14, "step": 122615 }, { "epoch": 2.86, "learning_rate": 8.602180188214311e-06, "loss": 0.2702, "step": 122620 }, { "epoch": 2.86, "learning_rate": 8.601300133768276e-06, "loss": 0.2691, "step": 122625 }, { "epoch": 2.86, "learning_rate": 8.600420079322241e-06, "loss": 0.0343, "step": 122630 }, { "epoch": 2.86, "learning_rate": 8.599540024876206e-06, "loss": 0.0194, "step": 122635 }, { "epoch": 2.86, "learning_rate": 8.598659970430171e-06, "loss": 0.0537, "step": 122640 }, { "epoch": 2.86, "learning_rate": 8.597779915984136e-06, "loss": 0.0618, "step": 122645 }, { "epoch": 2.86, "learning_rate": 8.5968998615381e-06, "loss": 0.1225, "step": 122650 }, { "epoch": 2.86, "learning_rate": 8.596019807092066e-06, "loss": 0.1514, "step": 122655 }, { "epoch": 2.86, "learning_rate": 8.59513975264603e-06, "loss": 0.1057, "step": 122660 }, { "epoch": 2.86, "learning_rate": 8.594259698199996e-06, "loss": 0.0971, "step": 122665 }, { "epoch": 2.86, "learning_rate": 8.59337964375396e-06, "loss": 0.1713, "step": 122670 }, { "epoch": 2.86, "learning_rate": 8.592499589307927e-06, "loss": 0.2206, "step": 122675 }, { "epoch": 2.86, "learning_rate": 8.59161953486189e-06, "loss": 0.0402, "step": 122680 }, { "epoch": 2.86, "learning_rate": 8.590739480415855e-06, "loss": 0.0349, "step": 122685 }, { "epoch": 2.86, "learning_rate": 8.58985942596982e-06, "loss": 0.0497, "step": 122690 }, { "epoch": 2.86, "learning_rate": 8.588979371523785e-06, "loss": 0.0027, "step": 122695 }, { "epoch": 2.86, "learning_rate": 8.58809931707775e-06, "loss": 0.0374, "step": 122700 }, { "epoch": 2.86, "learning_rate": 8.587219262631715e-06, "loss": 0.0505, "step": 122705 }, { "epoch": 2.86, "learning_rate": 8.58633920818568e-06, "loss": 0.0706, "step": 122710 }, { "epoch": 2.86, "learning_rate": 8.585459153739645e-06, "loss": 0.2144, "step": 122715 }, { "epoch": 2.86, "learning_rate": 8.584579099293608e-06, "loss": 0.1656, "step": 122720 }, { "epoch": 2.86, "learning_rate": 8.583699044847575e-06, "loss": 0.3025, "step": 122725 }, { "epoch": 2.86, "learning_rate": 8.582818990401538e-06, "loss": 0.051, "step": 122730 }, { "epoch": 2.86, "learning_rate": 8.581938935955505e-06, "loss": 0.0074, "step": 122735 }, { "epoch": 2.86, "learning_rate": 8.58105888150947e-06, "loss": 0.0294, "step": 122740 }, { "epoch": 2.86, "learning_rate": 8.580178827063435e-06, "loss": 0.0259, "step": 122745 }, { "epoch": 2.86, "learning_rate": 8.5792987726174e-06, "loss": 0.0508, "step": 122750 }, { "epoch": 2.86, "learning_rate": 8.578418718171365e-06, "loss": 0.0735, "step": 122755 }, { "epoch": 2.86, "learning_rate": 8.57753866372533e-06, "loss": 0.0215, "step": 122760 }, { "epoch": 2.86, "learning_rate": 8.576658609279294e-06, "loss": 0.1653, "step": 122765 }, { "epoch": 2.86, "learning_rate": 8.57577855483326e-06, "loss": 0.1211, "step": 122770 }, { "epoch": 2.86, "learning_rate": 8.574898500387224e-06, "loss": 0.282, "step": 122775 }, { "epoch": 2.86, "learning_rate": 8.57401844594119e-06, "loss": 0.0145, "step": 122780 }, { "epoch": 2.86, "learning_rate": 8.573138391495154e-06, "loss": 0.0232, "step": 122785 }, { "epoch": 2.86, "learning_rate": 8.57225833704912e-06, "loss": 0.0346, "step": 122790 }, { "epoch": 2.86, "learning_rate": 8.571378282603084e-06, "loss": 0.0364, "step": 122795 }, { "epoch": 2.87, "learning_rate": 8.570498228157049e-06, "loss": 0.0705, "step": 122800 }, { "epoch": 2.87, "learning_rate": 8.569618173711014e-06, "loss": 0.1, "step": 122805 }, { "epoch": 2.87, "learning_rate": 8.568738119264979e-06, "loss": 0.0691, "step": 122810 }, { "epoch": 2.87, "learning_rate": 8.567858064818944e-06, "loss": 0.098, "step": 122815 }, { "epoch": 2.87, "learning_rate": 8.566978010372909e-06, "loss": 0.1035, "step": 122820 }, { "epoch": 2.87, "learning_rate": 8.566097955926874e-06, "loss": 0.2167, "step": 122825 }, { "epoch": 2.87, "learning_rate": 8.565217901480839e-06, "loss": 0.0227, "step": 122830 }, { "epoch": 2.87, "learning_rate": 8.564337847034802e-06, "loss": 0.0267, "step": 122835 }, { "epoch": 2.87, "learning_rate": 8.563457792588769e-06, "loss": 0.0305, "step": 122840 }, { "epoch": 2.87, "learning_rate": 8.562577738142732e-06, "loss": 0.0428, "step": 122845 }, { "epoch": 2.87, "learning_rate": 8.561697683696699e-06, "loss": 0.032, "step": 122850 }, { "epoch": 2.87, "learning_rate": 8.560817629250662e-06, "loss": 0.1436, "step": 122855 }, { "epoch": 2.87, "learning_rate": 8.559937574804629e-06, "loss": 0.1066, "step": 122860 }, { "epoch": 2.87, "learning_rate": 8.559057520358592e-06, "loss": 0.0407, "step": 122865 }, { "epoch": 2.87, "learning_rate": 8.558177465912557e-06, "loss": 0.1566, "step": 122870 }, { "epoch": 2.87, "learning_rate": 8.557297411466522e-06, "loss": 0.2358, "step": 122875 }, { "epoch": 2.87, "learning_rate": 8.556417357020487e-06, "loss": 0.0208, "step": 122880 }, { "epoch": 2.87, "learning_rate": 8.555537302574454e-06, "loss": 0.0545, "step": 122885 }, { "epoch": 2.87, "learning_rate": 8.554657248128417e-06, "loss": 0.0317, "step": 122890 }, { "epoch": 2.87, "learning_rate": 8.553777193682384e-06, "loss": 0.0289, "step": 122895 }, { "epoch": 2.87, "learning_rate": 8.552897139236347e-06, "loss": 0.0814, "step": 122900 }, { "epoch": 2.87, "learning_rate": 8.552017084790312e-06, "loss": 0.1787, "step": 122905 }, { "epoch": 2.87, "learning_rate": 8.551137030344277e-06, "loss": 0.1147, "step": 122910 }, { "epoch": 2.87, "learning_rate": 8.550256975898242e-06, "loss": 0.0974, "step": 122915 }, { "epoch": 2.87, "learning_rate": 8.549376921452207e-06, "loss": 0.3378, "step": 122920 }, { "epoch": 2.87, "learning_rate": 8.548496867006172e-06, "loss": 0.2738, "step": 122925 }, { "epoch": 2.87, "learning_rate": 8.547616812560138e-06, "loss": 0.0439, "step": 122930 }, { "epoch": 2.87, "learning_rate": 8.546736758114103e-06, "loss": 0.0042, "step": 122935 }, { "epoch": 2.87, "learning_rate": 8.545856703668068e-06, "loss": 0.0151, "step": 122940 }, { "epoch": 2.87, "learning_rate": 8.544976649222033e-06, "loss": 0.0324, "step": 122945 }, { "epoch": 2.87, "learning_rate": 8.544096594775996e-06, "loss": 0.0394, "step": 122950 }, { "epoch": 2.87, "learning_rate": 8.543216540329963e-06, "loss": 0.1185, "step": 122955 }, { "epoch": 2.87, "learning_rate": 8.542336485883926e-06, "loss": 0.0608, "step": 122960 }, { "epoch": 2.87, "learning_rate": 8.541456431437893e-06, "loss": 0.1535, "step": 122965 }, { "epoch": 2.87, "learning_rate": 8.540576376991856e-06, "loss": 0.0832, "step": 122970 }, { "epoch": 2.87, "learning_rate": 8.539696322545823e-06, "loss": 0.3872, "step": 122975 }, { "epoch": 2.87, "learning_rate": 8.538816268099786e-06, "loss": 0.0191, "step": 122980 }, { "epoch": 2.87, "learning_rate": 8.537936213653751e-06, "loss": 0.0144, "step": 122985 }, { "epoch": 2.87, "learning_rate": 8.537056159207716e-06, "loss": 0.0571, "step": 122990 }, { "epoch": 2.87, "learning_rate": 8.536176104761681e-06, "loss": 0.0812, "step": 122995 }, { "epoch": 2.87, "learning_rate": 8.535296050315646e-06, "loss": 0.0209, "step": 123000 }, { "epoch": 2.87, "learning_rate": 8.534415995869611e-06, "loss": 0.0207, "step": 123005 }, { "epoch": 2.87, "learning_rate": 8.533535941423576e-06, "loss": 0.0871, "step": 123010 }, { "epoch": 2.87, "learning_rate": 8.532655886977541e-06, "loss": 0.1477, "step": 123015 }, { "epoch": 2.87, "learning_rate": 8.531775832531506e-06, "loss": 0.2506, "step": 123020 }, { "epoch": 2.87, "learning_rate": 8.530895778085471e-06, "loss": 0.0925, "step": 123025 }, { "epoch": 2.87, "learning_rate": 8.530015723639436e-06, "loss": 0.0285, "step": 123030 }, { "epoch": 2.87, "learning_rate": 8.529135669193401e-06, "loss": 0.0286, "step": 123035 }, { "epoch": 2.87, "learning_rate": 8.528255614747366e-06, "loss": 0.0175, "step": 123040 }, { "epoch": 2.87, "learning_rate": 8.527375560301331e-06, "loss": 0.0557, "step": 123045 }, { "epoch": 2.87, "learning_rate": 8.526495505855296e-06, "loss": 0.0784, "step": 123050 }, { "epoch": 2.87, "learning_rate": 8.52561545140926e-06, "loss": 0.2296, "step": 123055 }, { "epoch": 2.87, "learning_rate": 8.524735396963226e-06, "loss": 0.0789, "step": 123060 }, { "epoch": 2.87, "learning_rate": 8.52385534251719e-06, "loss": 0.129, "step": 123065 }, { "epoch": 2.87, "learning_rate": 8.522975288071156e-06, "loss": 0.1319, "step": 123070 }, { "epoch": 2.87, "learning_rate": 8.52209523362512e-06, "loss": 0.2854, "step": 123075 }, { "epoch": 2.87, "learning_rate": 8.521215179179086e-06, "loss": 0.0204, "step": 123080 }, { "epoch": 2.87, "learning_rate": 8.52033512473305e-06, "loss": 0.0025, "step": 123085 }, { "epoch": 2.87, "learning_rate": 8.519455070287016e-06, "loss": 0.0466, "step": 123090 }, { "epoch": 2.87, "learning_rate": 8.51857501584098e-06, "loss": 0.0617, "step": 123095 }, { "epoch": 2.87, "learning_rate": 8.517694961394945e-06, "loss": 0.07, "step": 123100 }, { "epoch": 2.87, "learning_rate": 8.51681490694891e-06, "loss": 0.158, "step": 123105 }, { "epoch": 2.87, "learning_rate": 8.515934852502875e-06, "loss": 0.1341, "step": 123110 }, { "epoch": 2.87, "learning_rate": 8.51505479805684e-06, "loss": 0.0158, "step": 123115 }, { "epoch": 2.87, "learning_rate": 8.514174743610805e-06, "loss": 0.3311, "step": 123120 }, { "epoch": 2.87, "learning_rate": 8.51329468916477e-06, "loss": 0.4746, "step": 123125 }, { "epoch": 2.87, "learning_rate": 8.512414634718735e-06, "loss": 0.0195, "step": 123130 }, { "epoch": 2.87, "learning_rate": 8.511534580272698e-06, "loss": 0.04, "step": 123135 }, { "epoch": 2.87, "learning_rate": 8.510654525826665e-06, "loss": 0.0441, "step": 123140 }, { "epoch": 2.87, "learning_rate": 8.509774471380628e-06, "loss": 0.0344, "step": 123145 }, { "epoch": 2.87, "learning_rate": 8.508894416934595e-06, "loss": 0.084, "step": 123150 }, { "epoch": 2.87, "learning_rate": 8.50801436248856e-06, "loss": 0.0839, "step": 123155 }, { "epoch": 2.87, "learning_rate": 8.507134308042525e-06, "loss": 0.3213, "step": 123160 }, { "epoch": 2.87, "learning_rate": 8.50625425359649e-06, "loss": 0.1896, "step": 123165 }, { "epoch": 2.87, "learning_rate": 8.505374199150453e-06, "loss": 0.1797, "step": 123170 }, { "epoch": 2.87, "learning_rate": 8.50449414470442e-06, "loss": 0.2447, "step": 123175 }, { "epoch": 2.87, "learning_rate": 8.503614090258383e-06, "loss": 0.089, "step": 123180 }, { "epoch": 2.87, "learning_rate": 8.50273403581235e-06, "loss": 0.0098, "step": 123185 }, { "epoch": 2.87, "learning_rate": 8.501853981366313e-06, "loss": 0.0682, "step": 123190 }, { "epoch": 2.87, "learning_rate": 8.50097392692028e-06, "loss": 0.0555, "step": 123195 }, { "epoch": 2.87, "learning_rate": 8.500093872474244e-06, "loss": 0.1353, "step": 123200 }, { "epoch": 2.87, "learning_rate": 8.499213818028209e-06, "loss": 0.0375, "step": 123205 }, { "epoch": 2.87, "learning_rate": 8.498333763582174e-06, "loss": 0.0518, "step": 123210 }, { "epoch": 2.87, "learning_rate": 8.497453709136139e-06, "loss": 0.0156, "step": 123215 }, { "epoch": 2.87, "learning_rate": 8.496573654690104e-06, "loss": 0.1708, "step": 123220 }, { "epoch": 2.87, "learning_rate": 8.495693600244069e-06, "loss": 0.2694, "step": 123225 }, { "epoch": 2.88, "learning_rate": 8.494813545798034e-06, "loss": 0.034, "step": 123230 }, { "epoch": 2.88, "learning_rate": 8.493933491351999e-06, "loss": 0.0386, "step": 123235 }, { "epoch": 2.88, "learning_rate": 8.493053436905964e-06, "loss": 0.0525, "step": 123240 }, { "epoch": 2.88, "learning_rate": 8.492173382459929e-06, "loss": 0.0662, "step": 123245 }, { "epoch": 2.88, "learning_rate": 8.491293328013892e-06, "loss": 0.1123, "step": 123250 }, { "epoch": 2.88, "learning_rate": 8.490413273567859e-06, "loss": 0.0656, "step": 123255 }, { "epoch": 2.88, "learning_rate": 8.489533219121822e-06, "loss": 0.0784, "step": 123260 }, { "epoch": 2.88, "learning_rate": 8.488653164675789e-06, "loss": 0.1024, "step": 123265 }, { "epoch": 2.88, "learning_rate": 8.487773110229752e-06, "loss": 0.2729, "step": 123270 }, { "epoch": 2.88, "learning_rate": 8.486893055783719e-06, "loss": 0.1445, "step": 123275 }, { "epoch": 2.88, "learning_rate": 8.486013001337682e-06, "loss": 0.0271, "step": 123280 }, { "epoch": 2.88, "learning_rate": 8.485132946891647e-06, "loss": 0.0028, "step": 123285 }, { "epoch": 2.88, "learning_rate": 8.484252892445614e-06, "loss": 0.0315, "step": 123290 }, { "epoch": 2.88, "learning_rate": 8.483372837999577e-06, "loss": 0.0619, "step": 123295 }, { "epoch": 2.88, "learning_rate": 8.482492783553544e-06, "loss": 0.0703, "step": 123300 }, { "epoch": 2.88, "learning_rate": 8.481612729107507e-06, "loss": 0.1329, "step": 123305 }, { "epoch": 2.88, "learning_rate": 8.480732674661474e-06, "loss": 0.0693, "step": 123310 }, { "epoch": 2.88, "learning_rate": 8.479852620215437e-06, "loss": 0.1185, "step": 123315 }, { "epoch": 2.88, "learning_rate": 8.478972565769402e-06, "loss": 0.0817, "step": 123320 }, { "epoch": 2.88, "learning_rate": 8.478092511323367e-06, "loss": 0.1939, "step": 123325 }, { "epoch": 2.88, "learning_rate": 8.477212456877332e-06, "loss": 0.0149, "step": 123330 }, { "epoch": 2.88, "learning_rate": 8.476332402431297e-06, "loss": 0.0163, "step": 123335 }, { "epoch": 2.88, "learning_rate": 8.475452347985262e-06, "loss": 0.0018, "step": 123340 }, { "epoch": 2.88, "learning_rate": 8.474572293539227e-06, "loss": 0.1257, "step": 123345 }, { "epoch": 2.88, "learning_rate": 8.473692239093192e-06, "loss": 0.0268, "step": 123350 }, { "epoch": 2.88, "learning_rate": 8.472812184647156e-06, "loss": 0.171, "step": 123355 }, { "epoch": 2.88, "learning_rate": 8.471932130201122e-06, "loss": 0.0885, "step": 123360 }, { "epoch": 2.88, "learning_rate": 8.471052075755086e-06, "loss": 0.0864, "step": 123365 }, { "epoch": 2.88, "learning_rate": 8.470172021309053e-06, "loss": 0.1941, "step": 123370 }, { "epoch": 2.88, "learning_rate": 8.469291966863016e-06, "loss": 0.3871, "step": 123375 }, { "epoch": 2.88, "learning_rate": 8.468411912416983e-06, "loss": 0.1189, "step": 123380 }, { "epoch": 2.88, "learning_rate": 8.467531857970946e-06, "loss": 0.0006, "step": 123385 }, { "epoch": 2.88, "learning_rate": 8.466651803524913e-06, "loss": 0.0536, "step": 123390 }, { "epoch": 2.88, "learning_rate": 8.465771749078876e-06, "loss": 0.0983, "step": 123395 }, { "epoch": 2.88, "learning_rate": 8.464891694632841e-06, "loss": 0.0263, "step": 123400 }, { "epoch": 2.88, "learning_rate": 8.464011640186806e-06, "loss": 0.0437, "step": 123405 }, { "epoch": 2.88, "learning_rate": 8.463131585740771e-06, "loss": 0.1332, "step": 123410 }, { "epoch": 2.88, "learning_rate": 8.462251531294736e-06, "loss": 0.1898, "step": 123415 }, { "epoch": 2.88, "learning_rate": 8.461371476848701e-06, "loss": 0.1671, "step": 123420 }, { "epoch": 2.88, "learning_rate": 8.460491422402668e-06, "loss": 0.2865, "step": 123425 }, { "epoch": 2.88, "learning_rate": 8.459611367956631e-06, "loss": 0.0002, "step": 123430 }, { "epoch": 2.88, "learning_rate": 8.458731313510596e-06, "loss": 0.0122, "step": 123435 }, { "epoch": 2.88, "learning_rate": 8.457851259064561e-06, "loss": 0.0195, "step": 123440 }, { "epoch": 2.88, "learning_rate": 8.456971204618526e-06, "loss": 0.0097, "step": 123445 }, { "epoch": 2.88, "learning_rate": 8.456091150172491e-06, "loss": 0.1137, "step": 123450 }, { "epoch": 2.88, "learning_rate": 8.455211095726456e-06, "loss": 0.0649, "step": 123455 }, { "epoch": 2.88, "learning_rate": 8.454331041280421e-06, "loss": 0.0914, "step": 123460 }, { "epoch": 2.88, "learning_rate": 8.453450986834386e-06, "loss": 0.1365, "step": 123465 }, { "epoch": 2.88, "learning_rate": 8.45257093238835e-06, "loss": 0.0996, "step": 123470 }, { "epoch": 2.88, "learning_rate": 8.451690877942316e-06, "loss": 0.316, "step": 123475 }, { "epoch": 2.88, "learning_rate": 8.45081082349628e-06, "loss": 0.0276, "step": 123480 }, { "epoch": 2.88, "learning_rate": 8.449930769050246e-06, "loss": 0.0757, "step": 123485 }, { "epoch": 2.88, "learning_rate": 8.44905071460421e-06, "loss": 0.0518, "step": 123490 }, { "epoch": 2.88, "learning_rate": 8.448170660158176e-06, "loss": 0.0007, "step": 123495 }, { "epoch": 2.88, "learning_rate": 8.44729060571214e-06, "loss": 0.1937, "step": 123500 }, { "epoch": 2.88, "learning_rate": 8.446410551266105e-06, "loss": 0.0097, "step": 123505 }, { "epoch": 2.88, "learning_rate": 8.44553049682007e-06, "loss": 0.2073, "step": 123510 }, { "epoch": 2.88, "learning_rate": 8.444650442374035e-06, "loss": 0.0974, "step": 123515 }, { "epoch": 2.88, "learning_rate": 8.443770387928e-06, "loss": 0.0956, "step": 123520 }, { "epoch": 2.88, "learning_rate": 8.442890333481965e-06, "loss": 0.3742, "step": 123525 }, { "epoch": 2.88, "learning_rate": 8.44201027903593e-06, "loss": 0.0381, "step": 123530 }, { "epoch": 2.88, "learning_rate": 8.441130224589895e-06, "loss": 0.0123, "step": 123535 }, { "epoch": 2.88, "learning_rate": 8.44025017014386e-06, "loss": 0.0003, "step": 123540 }, { "epoch": 2.88, "learning_rate": 8.439370115697825e-06, "loss": 0.0004, "step": 123545 }, { "epoch": 2.88, "learning_rate": 8.438490061251788e-06, "loss": 0.0722, "step": 123550 }, { "epoch": 2.88, "learning_rate": 8.437610006805755e-06, "loss": 0.0774, "step": 123555 }, { "epoch": 2.88, "learning_rate": 8.43672995235972e-06, "loss": 0.1345, "step": 123560 }, { "epoch": 2.88, "learning_rate": 8.436025908802891e-06, "loss": 0.1541, "step": 123565 }, { "epoch": 2.88, "learning_rate": 8.435145854356856e-06, "loss": 0.1272, "step": 123570 }, { "epoch": 2.88, "learning_rate": 8.434265799910821e-06, "loss": 0.3033, "step": 123575 }, { "epoch": 2.88, "learning_rate": 8.433385745464786e-06, "loss": 0.0631, "step": 123580 }, { "epoch": 2.88, "learning_rate": 8.432505691018751e-06, "loss": 0.0016, "step": 123585 }, { "epoch": 2.88, "learning_rate": 8.431625636572716e-06, "loss": 0.0222, "step": 123590 }, { "epoch": 2.88, "learning_rate": 8.43074558212668e-06, "loss": 0.0991, "step": 123595 }, { "epoch": 2.88, "learning_rate": 8.429865527680646e-06, "loss": 0.0362, "step": 123600 }, { "epoch": 2.88, "learning_rate": 8.42898547323461e-06, "loss": 0.0301, "step": 123605 }, { "epoch": 2.88, "learning_rate": 8.428105418788576e-06, "loss": 0.0276, "step": 123610 }, { "epoch": 2.88, "learning_rate": 8.427225364342541e-06, "loss": 0.0795, "step": 123615 }, { "epoch": 2.88, "learning_rate": 8.426345309896506e-06, "loss": 0.1815, "step": 123620 }, { "epoch": 2.88, "learning_rate": 8.425465255450472e-06, "loss": 0.1201, "step": 123625 }, { "epoch": 2.88, "learning_rate": 8.424585201004435e-06, "loss": 0.0439, "step": 123630 }, { "epoch": 2.88, "learning_rate": 8.423705146558402e-06, "loss": 0.0204, "step": 123635 }, { "epoch": 2.88, "learning_rate": 8.422825092112365e-06, "loss": 0.0239, "step": 123640 }, { "epoch": 2.88, "learning_rate": 8.421945037666332e-06, "loss": 0.0807, "step": 123645 }, { "epoch": 2.88, "learning_rate": 8.421064983220295e-06, "loss": 0.0702, "step": 123650 }, { "epoch": 2.89, "learning_rate": 8.420184928774262e-06, "loss": 0.0537, "step": 123655 }, { "epoch": 2.89, "learning_rate": 8.419304874328225e-06, "loss": 0.0842, "step": 123660 }, { "epoch": 2.89, "learning_rate": 8.41842481988219e-06, "loss": 0.0588, "step": 123665 }, { "epoch": 2.89, "learning_rate": 8.417544765436155e-06, "loss": 0.244, "step": 123670 }, { "epoch": 2.89, "learning_rate": 8.41666471099012e-06, "loss": 0.2847, "step": 123675 }, { "epoch": 2.89, "learning_rate": 8.415784656544085e-06, "loss": 0.0589, "step": 123680 }, { "epoch": 2.89, "learning_rate": 8.41490460209805e-06, "loss": 0.0972, "step": 123685 }, { "epoch": 2.89, "learning_rate": 8.414024547652015e-06, "loss": 0.0215, "step": 123690 }, { "epoch": 2.89, "learning_rate": 8.41314449320598e-06, "loss": 0.0165, "step": 123695 }, { "epoch": 2.89, "learning_rate": 8.412264438759943e-06, "loss": 0.0226, "step": 123700 }, { "epoch": 2.89, "learning_rate": 8.41138438431391e-06, "loss": 0.0705, "step": 123705 }, { "epoch": 2.89, "learning_rate": 8.410504329867873e-06, "loss": 0.1252, "step": 123710 }, { "epoch": 2.89, "learning_rate": 8.40962427542184e-06, "loss": 0.2115, "step": 123715 }, { "epoch": 2.89, "learning_rate": 8.408744220975804e-06, "loss": 0.2954, "step": 123720 }, { "epoch": 2.89, "learning_rate": 8.40786416652977e-06, "loss": 0.1762, "step": 123725 }, { "epoch": 2.89, "learning_rate": 8.406984112083734e-06, "loss": 0.0316, "step": 123730 }, { "epoch": 2.89, "learning_rate": 8.4061040576377e-06, "loss": 0.0353, "step": 123735 }, { "epoch": 2.89, "learning_rate": 8.405224003191664e-06, "loss": 0.0386, "step": 123740 }, { "epoch": 2.89, "learning_rate": 8.404343948745629e-06, "loss": 0.0107, "step": 123745 }, { "epoch": 2.89, "learning_rate": 8.403463894299595e-06, "loss": 0.2076, "step": 123750 }, { "epoch": 2.89, "learning_rate": 8.402583839853559e-06, "loss": 0.1339, "step": 123755 }, { "epoch": 2.89, "learning_rate": 8.401703785407525e-06, "loss": 0.0732, "step": 123760 }, { "epoch": 2.89, "learning_rate": 8.400823730961489e-06, "loss": 0.0992, "step": 123765 }, { "epoch": 2.89, "learning_rate": 8.399943676515455e-06, "loss": 0.0614, "step": 123770 }, { "epoch": 2.89, "learning_rate": 8.399063622069419e-06, "loss": 0.1932, "step": 123775 }, { "epoch": 2.89, "learning_rate": 8.398183567623384e-06, "loss": 0.0707, "step": 123780 }, { "epoch": 2.89, "learning_rate": 8.397303513177349e-06, "loss": 0.0513, "step": 123785 }, { "epoch": 2.89, "learning_rate": 8.396423458731314e-06, "loss": 0.0365, "step": 123790 }, { "epoch": 2.89, "learning_rate": 8.395543404285279e-06, "loss": 0.0003, "step": 123795 }, { "epoch": 2.89, "learning_rate": 8.394663349839244e-06, "loss": 0.0239, "step": 123800 }, { "epoch": 2.89, "learning_rate": 8.393783295393209e-06, "loss": 0.0738, "step": 123805 }, { "epoch": 2.89, "learning_rate": 8.392903240947174e-06, "loss": 0.0614, "step": 123810 }, { "epoch": 2.89, "learning_rate": 8.392023186501137e-06, "loss": 0.1719, "step": 123815 }, { "epoch": 2.89, "learning_rate": 8.391143132055104e-06, "loss": 0.1769, "step": 123820 }, { "epoch": 2.89, "learning_rate": 8.390263077609067e-06, "loss": 0.4086, "step": 123825 }, { "epoch": 2.89, "learning_rate": 8.389383023163034e-06, "loss": 0.0077, "step": 123830 }, { "epoch": 2.89, "learning_rate": 8.388502968716997e-06, "loss": 0.0004, "step": 123835 }, { "epoch": 2.89, "learning_rate": 8.387622914270964e-06, "loss": 0.0076, "step": 123840 }, { "epoch": 2.89, "learning_rate": 8.386742859824927e-06, "loss": 0.0861, "step": 123845 }, { "epoch": 2.89, "learning_rate": 8.385862805378892e-06, "loss": 0.1156, "step": 123850 }, { "epoch": 2.89, "learning_rate": 8.384982750932857e-06, "loss": 0.1031, "step": 123855 }, { "epoch": 2.89, "learning_rate": 8.384102696486822e-06, "loss": 0.0075, "step": 123860 }, { "epoch": 2.89, "learning_rate": 8.383222642040787e-06, "loss": 0.2433, "step": 123865 }, { "epoch": 2.89, "learning_rate": 8.382342587594752e-06, "loss": 0.1425, "step": 123870 }, { "epoch": 2.89, "learning_rate": 8.381462533148717e-06, "loss": 0.1227, "step": 123875 }, { "epoch": 2.89, "learning_rate": 8.380582478702682e-06, "loss": 0.006, "step": 123880 }, { "epoch": 2.89, "learning_rate": 8.37970242425665e-06, "loss": 0.0437, "step": 123885 }, { "epoch": 2.89, "learning_rate": 8.378822369810613e-06, "loss": 0.0235, "step": 123890 }, { "epoch": 2.89, "learning_rate": 8.377942315364578e-06, "loss": 0.0825, "step": 123895 }, { "epoch": 2.89, "learning_rate": 8.377062260918543e-06, "loss": 0.066, "step": 123900 }, { "epoch": 2.89, "learning_rate": 8.376182206472508e-06, "loss": 0.0561, "step": 123905 }, { "epoch": 2.89, "learning_rate": 8.375302152026473e-06, "loss": 0.1974, "step": 123910 }, { "epoch": 2.89, "learning_rate": 8.374422097580438e-06, "loss": 0.0789, "step": 123915 }, { "epoch": 2.89, "learning_rate": 8.373542043134403e-06, "loss": 0.2354, "step": 123920 }, { "epoch": 2.89, "learning_rate": 8.372661988688368e-06, "loss": 0.2939, "step": 123925 }, { "epoch": 2.89, "learning_rate": 8.371781934242331e-06, "loss": 0.0747, "step": 123930 }, { "epoch": 2.89, "learning_rate": 8.370901879796298e-06, "loss": 0.0428, "step": 123935 }, { "epoch": 2.89, "learning_rate": 8.370021825350261e-06, "loss": 0.0052, "step": 123940 }, { "epoch": 2.89, "learning_rate": 8.369141770904228e-06, "loss": 0.0123, "step": 123945 }, { "epoch": 2.89, "learning_rate": 8.368261716458191e-06, "loss": 0.0093, "step": 123950 }, { "epoch": 2.89, "learning_rate": 8.367381662012158e-06, "loss": 0.038, "step": 123955 }, { "epoch": 2.89, "learning_rate": 8.366501607566121e-06, "loss": 0.1706, "step": 123960 }, { "epoch": 2.89, "learning_rate": 8.365621553120086e-06, "loss": 0.2301, "step": 123965 }, { "epoch": 2.89, "learning_rate": 8.364741498674051e-06, "loss": 0.0616, "step": 123970 }, { "epoch": 2.89, "learning_rate": 8.363861444228016e-06, "loss": 0.1398, "step": 123975 }, { "epoch": 2.89, "learning_rate": 8.362981389781981e-06, "loss": 0.0205, "step": 123980 }, { "epoch": 2.89, "learning_rate": 8.362101335335946e-06, "loss": 0.0762, "step": 123985 }, { "epoch": 2.89, "learning_rate": 8.361221280889911e-06, "loss": 0.0139, "step": 123990 }, { "epoch": 2.89, "learning_rate": 8.360341226443876e-06, "loss": 0.0545, "step": 123995 }, { "epoch": 2.89, "learning_rate": 8.35946117199784e-06, "loss": 0.0806, "step": 124000 }, { "epoch": 2.89, "learning_rate": 8.358581117551806e-06, "loss": 0.2102, "step": 124005 }, { "epoch": 2.89, "learning_rate": 8.35770106310577e-06, "loss": 0.0015, "step": 124010 }, { "epoch": 2.89, "learning_rate": 8.356821008659736e-06, "loss": 0.0892, "step": 124015 }, { "epoch": 2.89, "learning_rate": 8.355940954213701e-06, "loss": 0.0757, "step": 124020 }, { "epoch": 2.89, "learning_rate": 8.355060899767666e-06, "loss": 0.2264, "step": 124025 }, { "epoch": 2.89, "learning_rate": 8.354180845321631e-06, "loss": 0.0046, "step": 124030 }, { "epoch": 2.89, "learning_rate": 8.353300790875596e-06, "loss": 0.0136, "step": 124035 }, { "epoch": 2.89, "learning_rate": 8.352420736429561e-06, "loss": 0.0379, "step": 124040 }, { "epoch": 2.89, "learning_rate": 8.351540681983525e-06, "loss": 0.0107, "step": 124045 }, { "epoch": 2.89, "learning_rate": 8.350660627537491e-06, "loss": 0.0127, "step": 124050 }, { "epoch": 2.89, "learning_rate": 8.349780573091455e-06, "loss": 0.0572, "step": 124055 }, { "epoch": 2.89, "learning_rate": 8.348900518645422e-06, "loss": 0.2102, "step": 124060 }, { "epoch": 2.89, "learning_rate": 8.348020464199385e-06, "loss": 0.1964, "step": 124065 }, { "epoch": 2.89, "learning_rate": 8.347140409753352e-06, "loss": 0.1443, "step": 124070 }, { "epoch": 2.89, "learning_rate": 8.346260355307315e-06, "loss": 0.1134, "step": 124075 }, { "epoch": 2.89, "learning_rate": 8.34538030086128e-06, "loss": 0.0411, "step": 124080 }, { "epoch": 2.9, "learning_rate": 8.344500246415245e-06, "loss": 0.1121, "step": 124085 }, { "epoch": 2.9, "learning_rate": 8.34362019196921e-06, "loss": 0.0378, "step": 124090 }, { "epoch": 2.9, "learning_rate": 8.342740137523175e-06, "loss": 0.0883, "step": 124095 }, { "epoch": 2.9, "learning_rate": 8.34186008307714e-06, "loss": 0.0542, "step": 124100 }, { "epoch": 2.9, "learning_rate": 8.340980028631105e-06, "loss": 0.091, "step": 124105 }, { "epoch": 2.9, "learning_rate": 8.34009997418507e-06, "loss": 0.1458, "step": 124110 }, { "epoch": 2.9, "learning_rate": 8.339219919739033e-06, "loss": 0.1088, "step": 124115 }, { "epoch": 2.9, "learning_rate": 8.338339865293e-06, "loss": 0.3889, "step": 124120 }, { "epoch": 2.9, "learning_rate": 8.337459810846963e-06, "loss": 0.1181, "step": 124125 }, { "epoch": 2.9, "learning_rate": 8.33657975640093e-06, "loss": 0.0501, "step": 124130 }, { "epoch": 2.9, "learning_rate": 8.335699701954893e-06, "loss": 0.0067, "step": 124135 }, { "epoch": 2.9, "learning_rate": 8.33481964750886e-06, "loss": 0.0019, "step": 124140 }, { "epoch": 2.9, "learning_rate": 8.333939593062823e-06, "loss": 0.0135, "step": 124145 }, { "epoch": 2.9, "learning_rate": 8.333059538616789e-06, "loss": 0.0179, "step": 124150 }, { "epoch": 2.9, "learning_rate": 8.332179484170755e-06, "loss": 0.1084, "step": 124155 }, { "epoch": 2.9, "learning_rate": 8.331299429724719e-06, "loss": 0.2149, "step": 124160 }, { "epoch": 2.9, "learning_rate": 8.330419375278685e-06, "loss": 0.2678, "step": 124165 }, { "epoch": 2.9, "learning_rate": 8.329539320832649e-06, "loss": 0.2307, "step": 124170 }, { "epoch": 2.9, "learning_rate": 8.328659266386615e-06, "loss": 0.1958, "step": 124175 }, { "epoch": 2.9, "learning_rate": 8.327779211940579e-06, "loss": 0.127, "step": 124180 }, { "epoch": 2.9, "learning_rate": 8.326899157494545e-06, "loss": 0.0434, "step": 124185 }, { "epoch": 2.9, "learning_rate": 8.326019103048509e-06, "loss": 0.0301, "step": 124190 }, { "epoch": 2.9, "learning_rate": 8.325139048602474e-06, "loss": 0.054, "step": 124195 }, { "epoch": 2.9, "learning_rate": 8.324258994156439e-06, "loss": 0.0237, "step": 124200 }, { "epoch": 2.9, "learning_rate": 8.323378939710404e-06, "loss": 0.0217, "step": 124205 }, { "epoch": 2.9, "learning_rate": 8.322498885264369e-06, "loss": 0.0537, "step": 124210 }, { "epoch": 2.9, "learning_rate": 8.321618830818334e-06, "loss": 0.0781, "step": 124215 }, { "epoch": 2.9, "learning_rate": 8.320738776372299e-06, "loss": 0.0832, "step": 124220 }, { "epoch": 2.9, "learning_rate": 8.319858721926264e-06, "loss": 0.218, "step": 124225 }, { "epoch": 2.9, "learning_rate": 8.318978667480227e-06, "loss": 0.0618, "step": 124230 }, { "epoch": 2.9, "learning_rate": 8.318098613034194e-06, "loss": 0.0046, "step": 124235 }, { "epoch": 2.9, "learning_rate": 8.317218558588157e-06, "loss": 0.0766, "step": 124240 }, { "epoch": 2.9, "learning_rate": 8.316338504142124e-06, "loss": 0.0984, "step": 124245 }, { "epoch": 2.9, "learning_rate": 8.315458449696087e-06, "loss": 0.1784, "step": 124250 }, { "epoch": 2.9, "learning_rate": 8.314578395250054e-06, "loss": 0.1591, "step": 124255 }, { "epoch": 2.9, "learning_rate": 8.313698340804017e-06, "loss": 0.31, "step": 124260 }, { "epoch": 2.9, "learning_rate": 8.312818286357982e-06, "loss": 0.0481, "step": 124265 }, { "epoch": 2.9, "learning_rate": 8.311938231911947e-06, "loss": 0.248, "step": 124270 }, { "epoch": 2.9, "learning_rate": 8.311058177465912e-06, "loss": 0.1576, "step": 124275 }, { "epoch": 2.9, "learning_rate": 8.310178123019877e-06, "loss": 0.0691, "step": 124280 }, { "epoch": 2.9, "learning_rate": 8.309298068573842e-06, "loss": 0.0169, "step": 124285 }, { "epoch": 2.9, "learning_rate": 8.308418014127807e-06, "loss": 0.0149, "step": 124290 }, { "epoch": 2.9, "learning_rate": 8.307537959681772e-06, "loss": 0.068, "step": 124295 }, { "epoch": 2.9, "learning_rate": 8.306657905235737e-06, "loss": 0.081, "step": 124300 }, { "epoch": 2.9, "learning_rate": 8.305777850789702e-06, "loss": 0.0304, "step": 124305 }, { "epoch": 2.9, "learning_rate": 8.304897796343667e-06, "loss": 0.0151, "step": 124310 }, { "epoch": 2.9, "learning_rate": 8.304017741897632e-06, "loss": 0.1094, "step": 124315 }, { "epoch": 2.9, "learning_rate": 8.303137687451598e-06, "loss": 0.2207, "step": 124320 }, { "epoch": 2.9, "learning_rate": 8.302257633005563e-06, "loss": 0.1661, "step": 124325 }, { "epoch": 2.9, "learning_rate": 8.301377578559528e-06, "loss": 0.0254, "step": 124330 }, { "epoch": 2.9, "learning_rate": 8.300497524113493e-06, "loss": 0.0094, "step": 124335 }, { "epoch": 2.9, "learning_rate": 8.299617469667458e-06, "loss": 0.0753, "step": 124340 }, { "epoch": 2.9, "learning_rate": 8.298737415221421e-06, "loss": 0.0014, "step": 124345 }, { "epoch": 2.9, "learning_rate": 8.297857360775388e-06, "loss": 0.012, "step": 124350 }, { "epoch": 2.9, "learning_rate": 8.296977306329351e-06, "loss": 0.0262, "step": 124355 }, { "epoch": 2.9, "learning_rate": 8.296097251883318e-06, "loss": 0.108, "step": 124360 }, { "epoch": 2.9, "learning_rate": 8.295217197437281e-06, "loss": 0.2478, "step": 124365 }, { "epoch": 2.9, "learning_rate": 8.294337142991248e-06, "loss": 0.1473, "step": 124370 }, { "epoch": 2.9, "learning_rate": 8.293457088545211e-06, "loss": 0.3131, "step": 124375 }, { "epoch": 2.9, "learning_rate": 8.292577034099176e-06, "loss": 0.0256, "step": 124380 }, { "epoch": 2.9, "learning_rate": 8.291696979653141e-06, "loss": 0.0171, "step": 124385 }, { "epoch": 2.9, "learning_rate": 8.290816925207106e-06, "loss": 0.009, "step": 124390 }, { "epoch": 2.9, "learning_rate": 8.289936870761071e-06, "loss": 0.027, "step": 124395 }, { "epoch": 2.9, "learning_rate": 8.289056816315036e-06, "loss": 0.0726, "step": 124400 }, { "epoch": 2.9, "learning_rate": 8.288176761869001e-06, "loss": 0.0942, "step": 124405 }, { "epoch": 2.9, "learning_rate": 8.287296707422966e-06, "loss": 0.1209, "step": 124410 }, { "epoch": 2.9, "learning_rate": 8.28641665297693e-06, "loss": 0.0768, "step": 124415 }, { "epoch": 2.9, "learning_rate": 8.285536598530896e-06, "loss": 0.2339, "step": 124420 }, { "epoch": 2.9, "learning_rate": 8.284656544084861e-06, "loss": 0.1391, "step": 124425 }, { "epoch": 2.9, "learning_rate": 8.283776489638826e-06, "loss": 0.0674, "step": 124430 }, { "epoch": 2.9, "learning_rate": 8.282896435192791e-06, "loss": 0.0483, "step": 124435 }, { "epoch": 2.9, "learning_rate": 8.282016380746756e-06, "loss": 0.006, "step": 124440 }, { "epoch": 2.9, "learning_rate": 8.281136326300721e-06, "loss": 0.0185, "step": 124445 }, { "epoch": 2.9, "learning_rate": 8.280256271854685e-06, "loss": 0.021, "step": 124450 }, { "epoch": 2.9, "learning_rate": 8.279376217408651e-06, "loss": 0.1007, "step": 124455 }, { "epoch": 2.9, "learning_rate": 8.278496162962615e-06, "loss": 0.0559, "step": 124460 }, { "epoch": 2.9, "learning_rate": 8.277616108516581e-06, "loss": 0.0633, "step": 124465 }, { "epoch": 2.9, "learning_rate": 8.276736054070545e-06, "loss": 0.2727, "step": 124470 }, { "epoch": 2.9, "learning_rate": 8.275855999624511e-06, "loss": 0.2169, "step": 124475 }, { "epoch": 2.9, "learning_rate": 8.274975945178475e-06, "loss": 0.0182, "step": 124480 }, { "epoch": 2.9, "learning_rate": 8.274095890732441e-06, "loss": 0.007, "step": 124485 }, { "epoch": 2.9, "learning_rate": 8.273215836286405e-06, "loss": 0.0262, "step": 124490 }, { "epoch": 2.9, "learning_rate": 8.27233578184037e-06, "loss": 0.0376, "step": 124495 }, { "epoch": 2.9, "learning_rate": 8.271455727394335e-06, "loss": 0.0697, "step": 124500 }, { "epoch": 2.9, "learning_rate": 8.2705756729483e-06, "loss": 0.0513, "step": 124505 }, { "epoch": 2.9, "learning_rate": 8.269695618502265e-06, "loss": 0.0881, "step": 124510 }, { "epoch": 2.91, "learning_rate": 8.26881556405623e-06, "loss": 0.1483, "step": 124515 }, { "epoch": 2.91, "learning_rate": 8.267935509610195e-06, "loss": 0.1657, "step": 124520 }, { "epoch": 2.91, "learning_rate": 8.26705545516416e-06, "loss": 0.2754, "step": 124525 }, { "epoch": 2.91, "learning_rate": 8.266175400718123e-06, "loss": 0.0622, "step": 124530 }, { "epoch": 2.91, "learning_rate": 8.26529534627209e-06, "loss": 0.0433, "step": 124535 }, { "epoch": 2.91, "learning_rate": 8.264415291826053e-06, "loss": 0.1001, "step": 124540 }, { "epoch": 2.91, "learning_rate": 8.26353523738002e-06, "loss": 0.0461, "step": 124545 }, { "epoch": 2.91, "learning_rate": 8.262655182933983e-06, "loss": 0.1392, "step": 124550 }, { "epoch": 2.91, "learning_rate": 8.26177512848795e-06, "loss": 0.1324, "step": 124555 }, { "epoch": 2.91, "learning_rate": 8.260895074041913e-06, "loss": 0.0188, "step": 124560 }, { "epoch": 2.91, "learning_rate": 8.260015019595878e-06, "loss": 0.1373, "step": 124565 }, { "epoch": 2.91, "learning_rate": 8.259134965149845e-06, "loss": 0.2071, "step": 124570 }, { "epoch": 2.91, "learning_rate": 8.258254910703808e-06, "loss": 0.1621, "step": 124575 }, { "epoch": 2.91, "learning_rate": 8.257374856257775e-06, "loss": 0.0491, "step": 124580 }, { "epoch": 2.91, "learning_rate": 8.256494801811739e-06, "loss": 0.0038, "step": 124585 }, { "epoch": 2.91, "learning_rate": 8.255614747365705e-06, "loss": 0.0435, "step": 124590 }, { "epoch": 2.91, "learning_rate": 8.254734692919669e-06, "loss": 0.0024, "step": 124595 }, { "epoch": 2.91, "learning_rate": 8.253854638473634e-06, "loss": 0.0443, "step": 124600 }, { "epoch": 2.91, "learning_rate": 8.252974584027599e-06, "loss": 0.1387, "step": 124605 }, { "epoch": 2.91, "learning_rate": 8.252094529581564e-06, "loss": 0.1358, "step": 124610 }, { "epoch": 2.91, "learning_rate": 8.251214475135529e-06, "loss": 0.044, "step": 124615 }, { "epoch": 2.91, "learning_rate": 8.250334420689494e-06, "loss": 0.3013, "step": 124620 }, { "epoch": 2.91, "learning_rate": 8.249454366243459e-06, "loss": 0.248, "step": 124625 }, { "epoch": 2.91, "learning_rate": 8.248574311797424e-06, "loss": 0.0573, "step": 124630 }, { "epoch": 2.91, "learning_rate": 8.247694257351389e-06, "loss": 0.0039, "step": 124635 }, { "epoch": 2.91, "learning_rate": 8.246814202905354e-06, "loss": 0.002, "step": 124640 }, { "epoch": 2.91, "learning_rate": 8.245934148459317e-06, "loss": 0.0791, "step": 124645 }, { "epoch": 2.91, "learning_rate": 8.245054094013284e-06, "loss": 0.0197, "step": 124650 }, { "epoch": 2.91, "learning_rate": 8.244174039567247e-06, "loss": 0.1216, "step": 124655 }, { "epoch": 2.91, "learning_rate": 8.243293985121214e-06, "loss": 0.1283, "step": 124660 }, { "epoch": 2.91, "learning_rate": 8.242413930675177e-06, "loss": 0.106, "step": 124665 }, { "epoch": 2.91, "learning_rate": 8.241533876229144e-06, "loss": 0.1719, "step": 124670 }, { "epoch": 2.91, "learning_rate": 8.240653821783107e-06, "loss": 0.2203, "step": 124675 }, { "epoch": 2.91, "learning_rate": 8.239773767337072e-06, "loss": 0.0306, "step": 124680 }, { "epoch": 2.91, "learning_rate": 8.238893712891037e-06, "loss": 0.0238, "step": 124685 }, { "epoch": 2.91, "learning_rate": 8.238013658445002e-06, "loss": 0.0365, "step": 124690 }, { "epoch": 2.91, "learning_rate": 8.237133603998967e-06, "loss": 0.0464, "step": 124695 }, { "epoch": 2.91, "learning_rate": 8.236253549552932e-06, "loss": 0.1008, "step": 124700 }, { "epoch": 2.91, "learning_rate": 8.235373495106899e-06, "loss": 0.0622, "step": 124705 }, { "epoch": 2.91, "learning_rate": 8.234493440660862e-06, "loss": 0.1145, "step": 124710 }, { "epoch": 2.91, "learning_rate": 8.233613386214827e-06, "loss": 0.0951, "step": 124715 }, { "epoch": 2.91, "learning_rate": 8.232733331768792e-06, "loss": 0.1473, "step": 124720 }, { "epoch": 2.91, "learning_rate": 8.231853277322757e-06, "loss": 0.3509, "step": 124725 }, { "epoch": 2.91, "learning_rate": 8.230973222876722e-06, "loss": 0.0111, "step": 124730 }, { "epoch": 2.91, "learning_rate": 8.230093168430687e-06, "loss": 0.003, "step": 124735 }, { "epoch": 2.91, "learning_rate": 8.229213113984652e-06, "loss": 0.0272, "step": 124740 }, { "epoch": 2.91, "learning_rate": 8.228333059538617e-06, "loss": 0.0849, "step": 124745 }, { "epoch": 2.91, "learning_rate": 8.22745300509258e-06, "loss": 0.0254, "step": 124750 }, { "epoch": 2.91, "learning_rate": 8.226572950646548e-06, "loss": 0.2186, "step": 124755 }, { "epoch": 2.91, "learning_rate": 8.22569289620051e-06, "loss": 0.2296, "step": 124760 }, { "epoch": 2.91, "learning_rate": 8.224812841754478e-06, "loss": 0.088, "step": 124765 }, { "epoch": 2.91, "learning_rate": 8.223932787308441e-06, "loss": 0.1193, "step": 124770 }, { "epoch": 2.91, "learning_rate": 8.223052732862408e-06, "loss": 0.1149, "step": 124775 }, { "epoch": 2.91, "learning_rate": 8.222172678416371e-06, "loss": 0.0606, "step": 124780 }, { "epoch": 2.91, "learning_rate": 8.221292623970338e-06, "loss": 0.0464, "step": 124785 }, { "epoch": 2.91, "learning_rate": 8.220412569524301e-06, "loss": 0.002, "step": 124790 }, { "epoch": 2.91, "learning_rate": 8.219532515078266e-06, "loss": 0.0424, "step": 124795 }, { "epoch": 2.91, "learning_rate": 8.218652460632231e-06, "loss": 0.0307, "step": 124800 }, { "epoch": 2.91, "learning_rate": 8.217772406186196e-06, "loss": 0.1668, "step": 124805 }, { "epoch": 2.91, "learning_rate": 8.216892351740161e-06, "loss": 0.0091, "step": 124810 }, { "epoch": 2.91, "learning_rate": 8.216012297294126e-06, "loss": 0.1511, "step": 124815 }, { "epoch": 2.91, "learning_rate": 8.215132242848091e-06, "loss": 0.1204, "step": 124820 }, { "epoch": 2.91, "learning_rate": 8.214252188402056e-06, "loss": 0.341, "step": 124825 }, { "epoch": 2.91, "learning_rate": 8.21337213395602e-06, "loss": 0.0794, "step": 124830 }, { "epoch": 2.91, "learning_rate": 8.212492079509986e-06, "loss": 0.046, "step": 124835 }, { "epoch": 2.91, "learning_rate": 8.211612025063951e-06, "loss": 0.0414, "step": 124840 }, { "epoch": 2.91, "learning_rate": 8.210731970617916e-06, "loss": 0.0651, "step": 124845 }, { "epoch": 2.91, "learning_rate": 8.209851916171881e-06, "loss": 0.074, "step": 124850 }, { "epoch": 2.91, "learning_rate": 8.208971861725846e-06, "loss": 0.0404, "step": 124855 }, { "epoch": 2.91, "learning_rate": 8.208091807279811e-06, "loss": 0.0274, "step": 124860 }, { "epoch": 2.91, "learning_rate": 8.207211752833775e-06, "loss": 0.067, "step": 124865 }, { "epoch": 2.91, "learning_rate": 8.206331698387741e-06, "loss": 0.1394, "step": 124870 }, { "epoch": 2.91, "learning_rate": 8.205451643941705e-06, "loss": 0.1221, "step": 124875 }, { "epoch": 2.91, "learning_rate": 8.204571589495671e-06, "loss": 0.0152, "step": 124880 }, { "epoch": 2.91, "learning_rate": 8.203691535049635e-06, "loss": 0.0187, "step": 124885 }, { "epoch": 2.91, "learning_rate": 8.202811480603601e-06, "loss": 0.0145, "step": 124890 }, { "epoch": 2.91, "learning_rate": 8.201931426157565e-06, "loss": 0.0004, "step": 124895 }, { "epoch": 2.91, "learning_rate": 8.20105137171153e-06, "loss": 0.031, "step": 124900 }, { "epoch": 2.91, "learning_rate": 8.200171317265495e-06, "loss": 0.0738, "step": 124905 }, { "epoch": 2.91, "learning_rate": 8.19929126281946e-06, "loss": 0.0649, "step": 124910 }, { "epoch": 2.91, "learning_rate": 8.198411208373425e-06, "loss": 0.2384, "step": 124915 }, { "epoch": 2.91, "learning_rate": 8.19753115392739e-06, "loss": 0.2202, "step": 124920 }, { "epoch": 2.91, "learning_rate": 8.196651099481355e-06, "loss": 0.1697, "step": 124925 }, { "epoch": 2.91, "learning_rate": 8.19577104503532e-06, "loss": 0.0055, "step": 124930 }, { "epoch": 2.91, "learning_rate": 8.194890990589285e-06, "loss": 0.0021, "step": 124935 }, { "epoch": 2.91, "learning_rate": 8.19401093614325e-06, "loss": 0.0673, "step": 124940 }, { "epoch": 2.92, "learning_rate": 8.193130881697213e-06, "loss": 0.0345, "step": 124945 }, { "epoch": 2.92, "learning_rate": 8.19225082725118e-06, "loss": 0.0762, "step": 124950 }, { "epoch": 2.92, "learning_rate": 8.191370772805143e-06, "loss": 0.0244, "step": 124955 }, { "epoch": 2.92, "learning_rate": 8.19049071835911e-06, "loss": 0.0574, "step": 124960 }, { "epoch": 2.92, "learning_rate": 8.189610663913073e-06, "loss": 0.0899, "step": 124965 }, { "epoch": 2.92, "learning_rate": 8.188906620356246e-06, "loss": 0.1813, "step": 124970 }, { "epoch": 2.92, "learning_rate": 8.188026565910211e-06, "loss": 0.2736, "step": 124975 }, { "epoch": 2.92, "learning_rate": 8.187146511464176e-06, "loss": 0.1228, "step": 124980 }, { "epoch": 2.92, "learning_rate": 8.186266457018141e-06, "loss": 0.0389, "step": 124985 }, { "epoch": 2.92, "learning_rate": 8.185386402572105e-06, "loss": 0.0056, "step": 124990 }, { "epoch": 2.92, "learning_rate": 8.184506348126071e-06, "loss": 0.0748, "step": 124995 }, { "epoch": 2.92, "learning_rate": 8.183626293680035e-06, "loss": 0.026, "step": 125000 }, { "epoch": 2.92, "learning_rate": 8.182746239234001e-06, "loss": 0.1024, "step": 125005 }, { "epoch": 2.92, "learning_rate": 8.181866184787965e-06, "loss": 0.1859, "step": 125010 }, { "epoch": 2.92, "learning_rate": 8.180986130341932e-06, "loss": 0.03, "step": 125015 }, { "epoch": 2.92, "learning_rate": 8.180106075895895e-06, "loss": 0.0537, "step": 125020 }, { "epoch": 2.92, "learning_rate": 8.17922602144986e-06, "loss": 0.2611, "step": 125025 }, { "epoch": 2.92, "learning_rate": 8.178345967003827e-06, "loss": 0.0734, "step": 125030 }, { "epoch": 2.92, "learning_rate": 8.17746591255779e-06, "loss": 0.0047, "step": 125035 }, { "epoch": 2.92, "learning_rate": 8.176585858111757e-06, "loss": 0.0095, "step": 125040 }, { "epoch": 2.92, "learning_rate": 8.17570580366572e-06, "loss": 0.0005, "step": 125045 }, { "epoch": 2.92, "learning_rate": 8.174825749219687e-06, "loss": 0.0295, "step": 125050 }, { "epoch": 2.92, "learning_rate": 8.17394569477365e-06, "loss": 0.0367, "step": 125055 }, { "epoch": 2.92, "learning_rate": 8.173065640327615e-06, "loss": 0.1516, "step": 125060 }, { "epoch": 2.92, "learning_rate": 8.17218558588158e-06, "loss": 0.092, "step": 125065 }, { "epoch": 2.92, "learning_rate": 8.171305531435545e-06, "loss": 0.1401, "step": 125070 }, { "epoch": 2.92, "learning_rate": 8.17042547698951e-06, "loss": 0.2013, "step": 125075 }, { "epoch": 2.92, "learning_rate": 8.169545422543475e-06, "loss": 0.0383, "step": 125080 }, { "epoch": 2.92, "learning_rate": 8.16866536809744e-06, "loss": 0.0019, "step": 125085 }, { "epoch": 2.92, "learning_rate": 8.167785313651405e-06, "loss": 0.0394, "step": 125090 }, { "epoch": 2.92, "learning_rate": 8.166905259205368e-06, "loss": 0.1007, "step": 125095 }, { "epoch": 2.92, "learning_rate": 8.166025204759335e-06, "loss": 0.0873, "step": 125100 }, { "epoch": 2.92, "learning_rate": 8.165145150313298e-06, "loss": 0.1503, "step": 125105 }, { "epoch": 2.92, "learning_rate": 8.164265095867265e-06, "loss": 0.2112, "step": 125110 }, { "epoch": 2.92, "learning_rate": 8.163385041421229e-06, "loss": 0.0432, "step": 125115 }, { "epoch": 2.92, "learning_rate": 8.162504986975195e-06, "loss": 0.3373, "step": 125120 }, { "epoch": 2.92, "learning_rate": 8.161624932529159e-06, "loss": 0.1602, "step": 125125 }, { "epoch": 2.92, "learning_rate": 8.160744878083125e-06, "loss": 0.0989, "step": 125130 }, { "epoch": 2.92, "learning_rate": 8.159864823637089e-06, "loss": 0.0107, "step": 125135 }, { "epoch": 2.92, "learning_rate": 8.158984769191054e-06, "loss": 0.0214, "step": 125140 }, { "epoch": 2.92, "learning_rate": 8.158104714745019e-06, "loss": 0.0651, "step": 125145 }, { "epoch": 2.92, "learning_rate": 8.157224660298984e-06, "loss": 0.0145, "step": 125150 }, { "epoch": 2.92, "learning_rate": 8.156344605852949e-06, "loss": 0.0255, "step": 125155 }, { "epoch": 2.92, "learning_rate": 8.155464551406914e-06, "loss": 0.0448, "step": 125160 }, { "epoch": 2.92, "learning_rate": 8.15458449696088e-06, "loss": 0.161, "step": 125165 }, { "epoch": 2.92, "learning_rate": 8.153704442514844e-06, "loss": 0.1045, "step": 125170 }, { "epoch": 2.92, "learning_rate": 8.152824388068809e-06, "loss": 0.1536, "step": 125175 }, { "epoch": 2.92, "learning_rate": 8.151944333622774e-06, "loss": 0.0521, "step": 125180 }, { "epoch": 2.92, "learning_rate": 8.151064279176739e-06, "loss": 0.0314, "step": 125185 }, { "epoch": 2.92, "learning_rate": 8.150184224730704e-06, "loss": 0.0267, "step": 125190 }, { "epoch": 2.92, "learning_rate": 8.149304170284669e-06, "loss": 0.0448, "step": 125195 }, { "epoch": 2.92, "learning_rate": 8.148424115838634e-06, "loss": 0.0442, "step": 125200 }, { "epoch": 2.92, "learning_rate": 8.147544061392599e-06, "loss": 0.0541, "step": 125205 }, { "epoch": 2.92, "learning_rate": 8.146664006946562e-06, "loss": 0.0245, "step": 125210 }, { "epoch": 2.92, "learning_rate": 8.145783952500529e-06, "loss": 0.2241, "step": 125215 }, { "epoch": 2.92, "learning_rate": 8.144903898054492e-06, "loss": 0.1964, "step": 125220 }, { "epoch": 2.92, "learning_rate": 8.144199854497665e-06, "loss": 0.2381, "step": 125225 }, { "epoch": 2.92, "learning_rate": 8.14331980005163e-06, "loss": 0.0277, "step": 125230 }, { "epoch": 2.92, "learning_rate": 8.142439745605595e-06, "loss": 0.0073, "step": 125235 }, { "epoch": 2.92, "learning_rate": 8.14155969115956e-06, "loss": 0.0381, "step": 125240 }, { "epoch": 2.92, "learning_rate": 8.140679636713525e-06, "loss": 0.0148, "step": 125245 }, { "epoch": 2.92, "learning_rate": 8.13979958226749e-06, "loss": 0.02, "step": 125250 }, { "epoch": 2.92, "learning_rate": 8.138919527821454e-06, "loss": 0.0618, "step": 125255 }, { "epoch": 2.92, "learning_rate": 8.13803947337542e-06, "loss": 0.1223, "step": 125260 }, { "epoch": 2.92, "learning_rate": 8.137159418929384e-06, "loss": 0.0561, "step": 125265 }, { "epoch": 2.92, "learning_rate": 8.13627936448335e-06, "loss": 0.1821, "step": 125270 }, { "epoch": 2.92, "learning_rate": 8.135399310037314e-06, "loss": 0.1272, "step": 125275 }, { "epoch": 2.92, "learning_rate": 8.13451925559128e-06, "loss": 0.0191, "step": 125280 }, { "epoch": 2.92, "learning_rate": 8.133639201145244e-06, "loss": 0.1429, "step": 125285 }, { "epoch": 2.92, "learning_rate": 8.132759146699209e-06, "loss": 0.157, "step": 125290 }, { "epoch": 2.92, "learning_rate": 8.131879092253174e-06, "loss": 0.0368, "step": 125295 }, { "epoch": 2.92, "learning_rate": 8.130999037807139e-06, "loss": 0.0233, "step": 125300 }, { "epoch": 2.92, "learning_rate": 8.130118983361104e-06, "loss": 0.0189, "step": 125305 }, { "epoch": 2.92, "learning_rate": 8.129238928915069e-06, "loss": 0.062, "step": 125310 }, { "epoch": 2.92, "learning_rate": 8.128358874469034e-06, "loss": 0.0143, "step": 125315 }, { "epoch": 2.92, "learning_rate": 8.127478820022999e-06, "loss": 0.1527, "step": 125320 }, { "epoch": 2.92, "learning_rate": 8.126598765576964e-06, "loss": 0.1431, "step": 125325 }, { "epoch": 2.92, "learning_rate": 8.125718711130929e-06, "loss": 0.0025, "step": 125330 }, { "epoch": 2.92, "learning_rate": 8.124838656684892e-06, "loss": 0.05, "step": 125335 }, { "epoch": 2.92, "learning_rate": 8.123958602238859e-06, "loss": 0.0196, "step": 125340 }, { "epoch": 2.92, "learning_rate": 8.123078547792822e-06, "loss": 0.0047, "step": 125345 }, { "epoch": 2.92, "learning_rate": 8.122198493346789e-06, "loss": 0.0403, "step": 125350 }, { "epoch": 2.92, "learning_rate": 8.121318438900754e-06, "loss": 0.0644, "step": 125355 }, { "epoch": 2.92, "learning_rate": 8.12043838445472e-06, "loss": 0.2829, "step": 125360 }, { "epoch": 2.92, "learning_rate": 8.119558330008684e-06, "loss": 0.1497, "step": 125365 }, { "epoch": 2.93, "learning_rate": 8.118678275562648e-06, "loss": 0.1569, "step": 125370 }, { "epoch": 2.93, "learning_rate": 8.117798221116614e-06, "loss": 0.1962, "step": 125375 }, { "epoch": 2.93, "learning_rate": 8.116918166670578e-06, "loss": 0.1363, "step": 125380 }, { "epoch": 2.93, "learning_rate": 8.116038112224544e-06, "loss": 0.0848, "step": 125385 }, { "epoch": 2.93, "learning_rate": 8.115158057778508e-06, "loss": 0.0325, "step": 125390 }, { "epoch": 2.93, "learning_rate": 8.114278003332474e-06, "loss": 0.0293, "step": 125395 }, { "epoch": 2.93, "learning_rate": 8.113397948886438e-06, "loss": 0.0157, "step": 125400 }, { "epoch": 2.93, "learning_rate": 8.112517894440403e-06, "loss": 0.1171, "step": 125405 }, { "epoch": 2.93, "learning_rate": 8.111637839994368e-06, "loss": 0.117, "step": 125410 }, { "epoch": 2.93, "learning_rate": 8.110757785548333e-06, "loss": 0.203, "step": 125415 }, { "epoch": 2.93, "learning_rate": 8.109877731102298e-06, "loss": 0.2006, "step": 125420 }, { "epoch": 2.93, "learning_rate": 8.108997676656263e-06, "loss": 0.3551, "step": 125425 }, { "epoch": 2.93, "learning_rate": 8.108117622210228e-06, "loss": 0.0026, "step": 125430 }, { "epoch": 2.93, "learning_rate": 8.107237567764193e-06, "loss": 0.0492, "step": 125435 }, { "epoch": 2.93, "learning_rate": 8.106357513318156e-06, "loss": 0.0434, "step": 125440 }, { "epoch": 2.93, "learning_rate": 8.105477458872123e-06, "loss": 0.1529, "step": 125445 }, { "epoch": 2.93, "learning_rate": 8.104597404426086e-06, "loss": 0.0839, "step": 125450 }, { "epoch": 2.93, "learning_rate": 8.103717349980053e-06, "loss": 0.0706, "step": 125455 }, { "epoch": 2.93, "learning_rate": 8.102837295534016e-06, "loss": 0.0814, "step": 125460 }, { "epoch": 2.93, "learning_rate": 8.101957241087983e-06, "loss": 0.0308, "step": 125465 }, { "epoch": 2.93, "learning_rate": 8.101077186641946e-06, "loss": 0.1054, "step": 125470 }, { "epoch": 2.93, "learning_rate": 8.100197132195913e-06, "loss": 0.0898, "step": 125475 }, { "epoch": 2.93, "learning_rate": 8.099317077749876e-06, "loss": 0.0264, "step": 125480 }, { "epoch": 2.93, "learning_rate": 8.098437023303841e-06, "loss": 0.0494, "step": 125485 }, { "epoch": 2.93, "learning_rate": 8.097556968857808e-06, "loss": 0.0401, "step": 125490 }, { "epoch": 2.93, "learning_rate": 8.096676914411771e-06, "loss": 0.0293, "step": 125495 }, { "epoch": 2.93, "learning_rate": 8.095796859965738e-06, "loss": 0.0583, "step": 125500 }, { "epoch": 2.93, "learning_rate": 8.094916805519701e-06, "loss": 0.1319, "step": 125505 }, { "epoch": 2.93, "learning_rate": 8.094036751073668e-06, "loss": 0.1777, "step": 125510 }, { "epoch": 2.93, "learning_rate": 8.093156696627631e-06, "loss": 0.0568, "step": 125515 }, { "epoch": 2.93, "learning_rate": 8.092276642181596e-06, "loss": 0.1699, "step": 125520 }, { "epoch": 2.93, "learning_rate": 8.091396587735561e-06, "loss": 0.1504, "step": 125525 }, { "epoch": 2.93, "learning_rate": 8.090516533289526e-06, "loss": 0.0059, "step": 125530 }, { "epoch": 2.93, "learning_rate": 8.089636478843491e-06, "loss": 0.0397, "step": 125535 }, { "epoch": 2.93, "learning_rate": 8.088756424397457e-06, "loss": 0.012, "step": 125540 }, { "epoch": 2.93, "learning_rate": 8.087876369951422e-06, "loss": 0.0391, "step": 125545 }, { "epoch": 2.93, "learning_rate": 8.086996315505387e-06, "loss": 0.019, "step": 125550 }, { "epoch": 2.93, "learning_rate": 8.08611626105935e-06, "loss": 0.0177, "step": 125555 }, { "epoch": 2.93, "learning_rate": 8.085236206613317e-06, "loss": 0.0462, "step": 125560 }, { "epoch": 2.93, "learning_rate": 8.08435615216728e-06, "loss": 0.1446, "step": 125565 }, { "epoch": 2.93, "learning_rate": 8.083476097721247e-06, "loss": 0.1436, "step": 125570 }, { "epoch": 2.93, "learning_rate": 8.08259604327521e-06, "loss": 0.1375, "step": 125575 }, { "epoch": 2.93, "learning_rate": 8.081715988829177e-06, "loss": 0.0291, "step": 125580 }, { "epoch": 2.93, "learning_rate": 8.08083593438314e-06, "loss": 0.0103, "step": 125585 }, { "epoch": 2.93, "learning_rate": 8.079955879937105e-06, "loss": 0.0342, "step": 125590 }, { "epoch": 2.93, "learning_rate": 8.07907582549107e-06, "loss": 0.0221, "step": 125595 }, { "epoch": 2.93, "learning_rate": 8.078195771045035e-06, "loss": 0.159, "step": 125600 }, { "epoch": 2.93, "learning_rate": 8.077315716599e-06, "loss": 0.067, "step": 125605 }, { "epoch": 2.93, "learning_rate": 8.076435662152965e-06, "loss": 0.1198, "step": 125610 }, { "epoch": 2.93, "learning_rate": 8.07555560770693e-06, "loss": 0.2235, "step": 125615 }, { "epoch": 2.93, "learning_rate": 8.074675553260895e-06, "loss": 0.1667, "step": 125620 }, { "epoch": 2.93, "learning_rate": 8.07379549881486e-06, "loss": 0.1328, "step": 125625 }, { "epoch": 2.93, "learning_rate": 8.072915444368825e-06, "loss": 0.021, "step": 125630 }, { "epoch": 2.93, "learning_rate": 8.07203538992279e-06, "loss": 0.0129, "step": 125635 }, { "epoch": 2.93, "learning_rate": 8.071155335476755e-06, "loss": 0.0706, "step": 125640 }, { "epoch": 2.93, "learning_rate": 8.07027528103072e-06, "loss": 0.1003, "step": 125645 }, { "epoch": 2.93, "learning_rate": 8.069395226584685e-06, "loss": 0.0161, "step": 125650 }, { "epoch": 2.93, "learning_rate": 8.06851517213865e-06, "loss": 0.0246, "step": 125655 }, { "epoch": 2.93, "learning_rate": 8.067635117692615e-06, "loss": 0.0843, "step": 125660 }, { "epoch": 2.93, "learning_rate": 8.06675506324658e-06, "loss": 0.0257, "step": 125665 }, { "epoch": 2.93, "learning_rate": 8.065875008800544e-06, "loss": 0.0732, "step": 125670 }, { "epoch": 2.93, "learning_rate": 8.06499495435451e-06, "loss": 0.1836, "step": 125675 }, { "epoch": 2.93, "learning_rate": 8.064114899908474e-06, "loss": 0.1169, "step": 125680 }, { "epoch": 2.93, "learning_rate": 8.06323484546244e-06, "loss": 0.1476, "step": 125685 }, { "epoch": 2.93, "learning_rate": 8.062354791016404e-06, "loss": 0.029, "step": 125690 }, { "epoch": 2.93, "learning_rate": 8.06147473657037e-06, "loss": 0.0069, "step": 125695 }, { "epoch": 2.93, "learning_rate": 8.060594682124334e-06, "loss": 0.1049, "step": 125700 }, { "epoch": 2.93, "learning_rate": 8.059714627678299e-06, "loss": 0.1748, "step": 125705 }, { "epoch": 2.93, "learning_rate": 8.058834573232264e-06, "loss": 0.0962, "step": 125710 }, { "epoch": 2.93, "learning_rate": 8.057954518786229e-06, "loss": 0.0684, "step": 125715 }, { "epoch": 2.93, "learning_rate": 8.057074464340194e-06, "loss": 0.0974, "step": 125720 }, { "epoch": 2.93, "learning_rate": 8.056194409894159e-06, "loss": 0.1557, "step": 125725 }, { "epoch": 2.93, "learning_rate": 8.055314355448124e-06, "loss": 0.0303, "step": 125730 }, { "epoch": 2.93, "learning_rate": 8.054434301002089e-06, "loss": 0.009, "step": 125735 }, { "epoch": 2.93, "learning_rate": 8.053554246556052e-06, "loss": 0.0045, "step": 125740 }, { "epoch": 2.93, "learning_rate": 8.052674192110019e-06, "loss": 0.079, "step": 125745 }, { "epoch": 2.93, "learning_rate": 8.051794137663982e-06, "loss": 0.0931, "step": 125750 }, { "epoch": 2.93, "learning_rate": 8.050914083217949e-06, "loss": 0.0518, "step": 125755 }, { "epoch": 2.93, "learning_rate": 8.050034028771914e-06, "loss": 0.1246, "step": 125760 }, { "epoch": 2.93, "learning_rate": 8.049153974325879e-06, "loss": 0.0804, "step": 125765 }, { "epoch": 2.93, "learning_rate": 8.048273919879844e-06, "loss": 0.2476, "step": 125770 }, { "epoch": 2.93, "learning_rate": 8.047393865433809e-06, "loss": 0.2425, "step": 125775 }, { "epoch": 2.93, "learning_rate": 8.046513810987774e-06, "loss": 0.0963, "step": 125780 }, { "epoch": 2.93, "learning_rate": 8.045633756541737e-06, "loss": 0.0003, "step": 125785 }, { "epoch": 2.93, "learning_rate": 8.044753702095704e-06, "loss": 0.0617, "step": 125790 }, { "epoch": 2.93, "learning_rate": 8.043873647649667e-06, "loss": 0.0454, "step": 125795 }, { "epoch": 2.94, "learning_rate": 8.042993593203634e-06, "loss": 0.0216, "step": 125800 }, { "epoch": 2.94, "learning_rate": 8.042113538757598e-06, "loss": 0.1056, "step": 125805 }, { "epoch": 2.94, "learning_rate": 8.041233484311564e-06, "loss": 0.0127, "step": 125810 }, { "epoch": 2.94, "learning_rate": 8.040353429865528e-06, "loss": 0.0869, "step": 125815 }, { "epoch": 2.94, "learning_rate": 8.039473375419493e-06, "loss": 0.0399, "step": 125820 }, { "epoch": 2.94, "learning_rate": 8.038593320973458e-06, "loss": 0.2229, "step": 125825 }, { "epoch": 2.94, "learning_rate": 8.037713266527423e-06, "loss": 0.0266, "step": 125830 }, { "epoch": 2.94, "learning_rate": 8.036833212081388e-06, "loss": 0.0003, "step": 125835 }, { "epoch": 2.94, "learning_rate": 8.035953157635353e-06, "loss": 0.0382, "step": 125840 }, { "epoch": 2.94, "learning_rate": 8.035073103189318e-06, "loss": 0.0442, "step": 125845 }, { "epoch": 2.94, "learning_rate": 8.034193048743283e-06, "loss": 0.1188, "step": 125850 }, { "epoch": 2.94, "learning_rate": 8.033312994297246e-06, "loss": 0.0874, "step": 125855 }, { "epoch": 2.94, "learning_rate": 8.032432939851213e-06, "loss": 0.0737, "step": 125860 }, { "epoch": 2.94, "learning_rate": 8.031552885405176e-06, "loss": 0.1153, "step": 125865 }, { "epoch": 2.94, "learning_rate": 8.030672830959143e-06, "loss": 0.1837, "step": 125870 }, { "epoch": 2.94, "learning_rate": 8.029792776513106e-06, "loss": 0.2423, "step": 125875 }, { "epoch": 2.94, "learning_rate": 8.028912722067073e-06, "loss": 0.0038, "step": 125880 }, { "epoch": 2.94, "learning_rate": 8.028032667621036e-06, "loss": 0.0551, "step": 125885 }, { "epoch": 2.94, "learning_rate": 8.027152613175001e-06, "loss": 0.1279, "step": 125890 }, { "epoch": 2.94, "learning_rate": 8.026272558728966e-06, "loss": 0.077, "step": 125895 }, { "epoch": 2.94, "learning_rate": 8.025392504282931e-06, "loss": 0.0342, "step": 125900 }, { "epoch": 2.94, "learning_rate": 8.024512449836898e-06, "loss": 0.1039, "step": 125905 }, { "epoch": 2.94, "learning_rate": 8.023632395390861e-06, "loss": 0.1191, "step": 125910 }, { "epoch": 2.94, "learning_rate": 8.022752340944828e-06, "loss": 0.2942, "step": 125915 }, { "epoch": 2.94, "learning_rate": 8.021872286498791e-06, "loss": 0.1125, "step": 125920 }, { "epoch": 2.94, "learning_rate": 8.020992232052758e-06, "loss": 0.2062, "step": 125925 }, { "epoch": 2.94, "learning_rate": 8.020112177606721e-06, "loss": 0.0234, "step": 125930 }, { "epoch": 2.94, "learning_rate": 8.019232123160686e-06, "loss": 0.0518, "step": 125935 }, { "epoch": 2.94, "learning_rate": 8.018352068714651e-06, "loss": 0.0662, "step": 125940 }, { "epoch": 2.94, "learning_rate": 8.017472014268616e-06, "loss": 0.0375, "step": 125945 }, { "epoch": 2.94, "learning_rate": 8.016591959822581e-06, "loss": 0.0162, "step": 125950 }, { "epoch": 2.94, "learning_rate": 8.015711905376546e-06, "loss": 0.019, "step": 125955 }, { "epoch": 2.94, "learning_rate": 8.014831850930511e-06, "loss": 0.0238, "step": 125960 }, { "epoch": 2.94, "learning_rate": 8.014127807373683e-06, "loss": 0.2233, "step": 125965 }, { "epoch": 2.94, "learning_rate": 8.01324775292765e-06, "loss": 0.1346, "step": 125970 }, { "epoch": 2.94, "learning_rate": 8.012367698481613e-06, "loss": 0.2099, "step": 125975 }, { "epoch": 2.94, "learning_rate": 8.011487644035578e-06, "loss": 0.0659, "step": 125980 }, { "epoch": 2.94, "learning_rate": 8.010607589589543e-06, "loss": 0.0292, "step": 125985 }, { "epoch": 2.94, "learning_rate": 8.009727535143508e-06, "loss": 0.0102, "step": 125990 }, { "epoch": 2.94, "learning_rate": 8.008847480697473e-06, "loss": 0.0309, "step": 125995 }, { "epoch": 2.94, "learning_rate": 8.007967426251438e-06, "loss": 0.0257, "step": 126000 }, { "epoch": 2.94, "learning_rate": 8.007087371805403e-06, "loss": 0.0563, "step": 126005 }, { "epoch": 2.94, "learning_rate": 8.006207317359368e-06, "loss": 0.1768, "step": 126010 }, { "epoch": 2.94, "learning_rate": 8.005327262913331e-06, "loss": 0.1638, "step": 126015 }, { "epoch": 2.94, "learning_rate": 8.004447208467298e-06, "loss": 0.1121, "step": 126020 }, { "epoch": 2.94, "learning_rate": 8.003567154021261e-06, "loss": 0.1902, "step": 126025 }, { "epoch": 2.94, "learning_rate": 8.002687099575228e-06, "loss": 0.1985, "step": 126030 }, { "epoch": 2.94, "learning_rate": 8.001807045129191e-06, "loss": 0.012, "step": 126035 }, { "epoch": 2.94, "learning_rate": 8.000926990683158e-06, "loss": 0.0293, "step": 126040 }, { "epoch": 2.94, "learning_rate": 8.000046936237121e-06, "loss": 0.0151, "step": 126045 }, { "epoch": 2.94, "learning_rate": 7.999166881791086e-06, "loss": 0.0862, "step": 126050 }, { "epoch": 2.94, "learning_rate": 7.998286827345051e-06, "loss": 0.0178, "step": 126055 }, { "epoch": 2.94, "learning_rate": 7.997406772899016e-06, "loss": 0.048, "step": 126060 }, { "epoch": 2.94, "learning_rate": 7.996526718452982e-06, "loss": 0.0921, "step": 126065 }, { "epoch": 2.94, "learning_rate": 7.995646664006947e-06, "loss": 0.2597, "step": 126070 }, { "epoch": 2.94, "learning_rate": 7.994766609560912e-06, "loss": 0.2626, "step": 126075 }, { "epoch": 2.94, "learning_rate": 7.993886555114877e-06, "loss": 0.0136, "step": 126080 }, { "epoch": 2.94, "learning_rate": 7.993006500668842e-06, "loss": 0.0017, "step": 126085 }, { "epoch": 2.94, "learning_rate": 7.992126446222807e-06, "loss": 0.044, "step": 126090 }, { "epoch": 2.94, "learning_rate": 7.991246391776772e-06, "loss": 0.1561, "step": 126095 }, { "epoch": 2.94, "learning_rate": 7.990366337330737e-06, "loss": 0.0448, "step": 126100 }, { "epoch": 2.94, "learning_rate": 7.989486282884702e-06, "loss": 0.0456, "step": 126105 }, { "epoch": 2.94, "learning_rate": 7.988606228438667e-06, "loss": 0.0285, "step": 126110 }, { "epoch": 2.94, "learning_rate": 7.987726173992632e-06, "loss": 0.1413, "step": 126115 }, { "epoch": 2.94, "learning_rate": 7.986846119546597e-06, "loss": 0.1567, "step": 126120 }, { "epoch": 2.94, "learning_rate": 7.985966065100562e-06, "loss": 0.3227, "step": 126125 }, { "epoch": 2.94, "learning_rate": 7.985086010654525e-06, "loss": 0.0603, "step": 126130 }, { "epoch": 2.94, "learning_rate": 7.984205956208492e-06, "loss": 0.0014, "step": 126135 }, { "epoch": 2.94, "learning_rate": 7.983325901762455e-06, "loss": 0.0864, "step": 126140 }, { "epoch": 2.94, "learning_rate": 7.982445847316422e-06, "loss": 0.0302, "step": 126145 }, { "epoch": 2.94, "learning_rate": 7.981565792870385e-06, "loss": 0.1859, "step": 126150 }, { "epoch": 2.94, "learning_rate": 7.980685738424352e-06, "loss": 0.0294, "step": 126155 }, { "epoch": 2.94, "learning_rate": 7.979805683978315e-06, "loss": 0.0282, "step": 126160 }, { "epoch": 2.94, "learning_rate": 7.97892562953228e-06, "loss": 0.0631, "step": 126165 }, { "epoch": 2.94, "learning_rate": 7.978045575086245e-06, "loss": 0.18, "step": 126170 }, { "epoch": 2.94, "learning_rate": 7.97716552064021e-06, "loss": 0.1893, "step": 126175 }, { "epoch": 2.94, "learning_rate": 7.976285466194175e-06, "loss": 0.0324, "step": 126180 }, { "epoch": 2.94, "learning_rate": 7.97540541174814e-06, "loss": 0.0835, "step": 126185 }, { "epoch": 2.94, "learning_rate": 7.974525357302105e-06, "loss": 0.0471, "step": 126190 }, { "epoch": 2.94, "learning_rate": 7.97364530285607e-06, "loss": 0.0773, "step": 126195 }, { "epoch": 2.94, "learning_rate": 7.972765248410034e-06, "loss": 0.0117, "step": 126200 }, { "epoch": 2.94, "learning_rate": 7.971885193964e-06, "loss": 0.0238, "step": 126205 }, { "epoch": 2.94, "learning_rate": 7.971005139517964e-06, "loss": 0.0898, "step": 126210 }, { "epoch": 2.94, "learning_rate": 7.97012508507193e-06, "loss": 0.1125, "step": 126215 }, { "epoch": 2.94, "learning_rate": 7.969245030625894e-06, "loss": 0.2083, "step": 126220 }, { "epoch": 2.94, "learning_rate": 7.96836497617986e-06, "loss": 0.1338, "step": 126225 }, { "epoch": 2.95, "learning_rate": 7.967484921733826e-06, "loss": 0.0005, "step": 126230 }, { "epoch": 2.95, "learning_rate": 7.966604867287789e-06, "loss": 0.0193, "step": 126235 }, { "epoch": 2.95, "learning_rate": 7.965724812841756e-06, "loss": 0.0212, "step": 126240 }, { "epoch": 2.95, "learning_rate": 7.964844758395719e-06, "loss": 0.1249, "step": 126245 }, { "epoch": 2.95, "learning_rate": 7.963964703949686e-06, "loss": 0.1178, "step": 126250 }, { "epoch": 2.95, "learning_rate": 7.963084649503649e-06, "loss": 0.0286, "step": 126255 }, { "epoch": 2.95, "learning_rate": 7.962204595057616e-06, "loss": 0.1915, "step": 126260 }, { "epoch": 2.95, "learning_rate": 7.961324540611579e-06, "loss": 0.0726, "step": 126265 }, { "epoch": 2.95, "learning_rate": 7.960444486165546e-06, "loss": 0.1877, "step": 126270 }, { "epoch": 2.95, "learning_rate": 7.959564431719509e-06, "loss": 0.2005, "step": 126275 }, { "epoch": 2.95, "learning_rate": 7.958684377273474e-06, "loss": 0.0207, "step": 126280 }, { "epoch": 2.95, "learning_rate": 7.957804322827439e-06, "loss": 0.0442, "step": 126285 }, { "epoch": 2.95, "learning_rate": 7.956924268381404e-06, "loss": 0.0315, "step": 126290 }, { "epoch": 2.95, "learning_rate": 7.956044213935369e-06, "loss": 0.0272, "step": 126295 }, { "epoch": 2.95, "learning_rate": 7.955164159489334e-06, "loss": 0.0167, "step": 126300 }, { "epoch": 2.95, "learning_rate": 7.954284105043299e-06, "loss": 0.0468, "step": 126305 }, { "epoch": 2.95, "learning_rate": 7.953404050597264e-06, "loss": 0.111, "step": 126310 }, { "epoch": 2.95, "learning_rate": 7.952523996151227e-06, "loss": 0.1026, "step": 126315 }, { "epoch": 2.95, "learning_rate": 7.951643941705194e-06, "loss": 0.1331, "step": 126320 }, { "epoch": 2.95, "learning_rate": 7.950763887259158e-06, "loss": 0.1784, "step": 126325 }, { "epoch": 2.95, "learning_rate": 7.949883832813124e-06, "loss": 0.0254, "step": 126330 }, { "epoch": 2.95, "learning_rate": 7.949003778367088e-06, "loss": 0.0636, "step": 126335 }, { "epoch": 2.95, "learning_rate": 7.948123723921054e-06, "loss": 0.0669, "step": 126340 }, { "epoch": 2.95, "learning_rate": 7.947243669475018e-06, "loss": 0.0808, "step": 126345 }, { "epoch": 2.95, "learning_rate": 7.946363615028983e-06, "loss": 0.0325, "step": 126350 }, { "epoch": 2.95, "learning_rate": 7.945483560582948e-06, "loss": 0.1102, "step": 126355 }, { "epoch": 2.95, "learning_rate": 7.944603506136913e-06, "loss": 0.0294, "step": 126360 }, { "epoch": 2.95, "learning_rate": 7.94372345169088e-06, "loss": 0.1551, "step": 126365 }, { "epoch": 2.95, "learning_rate": 7.942843397244843e-06, "loss": 0.165, "step": 126370 }, { "epoch": 2.95, "learning_rate": 7.94196334279881e-06, "loss": 0.4898, "step": 126375 }, { "epoch": 2.95, "learning_rate": 7.941083288352773e-06, "loss": 0.0608, "step": 126380 }, { "epoch": 2.95, "learning_rate": 7.940203233906738e-06, "loss": 0.0041, "step": 126385 }, { "epoch": 2.95, "learning_rate": 7.939323179460703e-06, "loss": 0.0064, "step": 126390 }, { "epoch": 2.95, "learning_rate": 7.938443125014668e-06, "loss": 0.0492, "step": 126395 }, { "epoch": 2.95, "learning_rate": 7.937563070568633e-06, "loss": 0.097, "step": 126400 }, { "epoch": 2.95, "learning_rate": 7.936683016122598e-06, "loss": 0.1267, "step": 126405 }, { "epoch": 2.95, "learning_rate": 7.935802961676563e-06, "loss": 0.0681, "step": 126410 }, { "epoch": 2.95, "learning_rate": 7.934922907230528e-06, "loss": 0.0866, "step": 126415 }, { "epoch": 2.95, "learning_rate": 7.934042852784493e-06, "loss": 0.1092, "step": 126420 }, { "epoch": 2.95, "learning_rate": 7.933162798338458e-06, "loss": 0.2229, "step": 126425 }, { "epoch": 2.95, "learning_rate": 7.932282743892421e-06, "loss": 0.0509, "step": 126430 }, { "epoch": 2.95, "learning_rate": 7.931402689446388e-06, "loss": 0.0163, "step": 126435 }, { "epoch": 2.95, "learning_rate": 7.930522635000351e-06, "loss": 0.0056, "step": 126440 }, { "epoch": 2.95, "learning_rate": 7.929642580554318e-06, "loss": 0.0053, "step": 126445 }, { "epoch": 2.95, "learning_rate": 7.928762526108281e-06, "loss": 0.078, "step": 126450 }, { "epoch": 2.95, "learning_rate": 7.927882471662248e-06, "loss": 0.0953, "step": 126455 }, { "epoch": 2.95, "learning_rate": 7.927002417216211e-06, "loss": 0.1364, "step": 126460 }, { "epoch": 2.95, "learning_rate": 7.926122362770176e-06, "loss": 0.1427, "step": 126465 }, { "epoch": 2.95, "learning_rate": 7.925242308324141e-06, "loss": 0.2715, "step": 126470 }, { "epoch": 2.95, "learning_rate": 7.924362253878106e-06, "loss": 0.2043, "step": 126475 }, { "epoch": 2.95, "learning_rate": 7.923482199432071e-06, "loss": 0.0867, "step": 126480 }, { "epoch": 2.95, "learning_rate": 7.922602144986036e-06, "loss": 0.0008, "step": 126485 }, { "epoch": 2.95, "learning_rate": 7.921722090540001e-06, "loss": 0.0145, "step": 126490 }, { "epoch": 2.95, "learning_rate": 7.920842036093967e-06, "loss": 0.0475, "step": 126495 }, { "epoch": 2.95, "learning_rate": 7.919961981647932e-06, "loss": 0.0241, "step": 126500 }, { "epoch": 2.95, "learning_rate": 7.919081927201897e-06, "loss": 0.05, "step": 126505 }, { "epoch": 2.95, "learning_rate": 7.918201872755862e-06, "loss": 0.0458, "step": 126510 }, { "epoch": 2.95, "learning_rate": 7.917321818309827e-06, "loss": 0.1011, "step": 126515 }, { "epoch": 2.95, "learning_rate": 7.916441763863792e-06, "loss": 0.2477, "step": 126520 }, { "epoch": 2.95, "learning_rate": 7.915561709417757e-06, "loss": 0.1319, "step": 126525 }, { "epoch": 2.95, "learning_rate": 7.914681654971722e-06, "loss": 0.0065, "step": 126530 }, { "epoch": 2.95, "learning_rate": 7.913801600525685e-06, "loss": 0.0006, "step": 126535 }, { "epoch": 2.95, "learning_rate": 7.912921546079652e-06, "loss": 0.0098, "step": 126540 }, { "epoch": 2.95, "learning_rate": 7.912041491633615e-06, "loss": 0.021, "step": 126545 }, { "epoch": 2.95, "learning_rate": 7.911161437187582e-06, "loss": 0.0518, "step": 126550 }, { "epoch": 2.95, "learning_rate": 7.910281382741545e-06, "loss": 0.096, "step": 126555 }, { "epoch": 2.95, "learning_rate": 7.909401328295512e-06, "loss": 0.1148, "step": 126560 }, { "epoch": 2.95, "learning_rate": 7.908521273849475e-06, "loss": 0.0429, "step": 126565 }, { "epoch": 2.95, "learning_rate": 7.907641219403442e-06, "loss": 0.4182, "step": 126570 }, { "epoch": 2.95, "learning_rate": 7.906761164957405e-06, "loss": 0.1687, "step": 126575 }, { "epoch": 2.95, "learning_rate": 7.90588111051137e-06, "loss": 0.0459, "step": 126580 }, { "epoch": 2.95, "learning_rate": 7.905001056065335e-06, "loss": 0.043, "step": 126585 }, { "epoch": 2.95, "learning_rate": 7.9041210016193e-06, "loss": 0.0022, "step": 126590 }, { "epoch": 2.95, "learning_rate": 7.903240947173265e-06, "loss": 0.0607, "step": 126595 }, { "epoch": 2.95, "learning_rate": 7.90236089272723e-06, "loss": 0.0985, "step": 126600 }, { "epoch": 2.95, "learning_rate": 7.901480838281195e-06, "loss": 0.0799, "step": 126605 }, { "epoch": 2.95, "learning_rate": 7.90060078383516e-06, "loss": 0.0986, "step": 126610 }, { "epoch": 2.95, "learning_rate": 7.899720729389124e-06, "loss": 0.1269, "step": 126615 }, { "epoch": 2.95, "learning_rate": 7.89884067494309e-06, "loss": 0.1749, "step": 126620 }, { "epoch": 2.95, "learning_rate": 7.897960620497054e-06, "loss": 0.1941, "step": 126625 }, { "epoch": 2.95, "learning_rate": 7.89708056605102e-06, "loss": 0.032, "step": 126630 }, { "epoch": 2.95, "learning_rate": 7.896200511604985e-06, "loss": 0.0185, "step": 126635 }, { "epoch": 2.95, "learning_rate": 7.89532045715895e-06, "loss": 0.0496, "step": 126640 }, { "epoch": 2.95, "learning_rate": 7.894440402712915e-06, "loss": 0.0245, "step": 126645 }, { "epoch": 2.95, "learning_rate": 7.893560348266879e-06, "loss": 0.1028, "step": 126650 }, { "epoch": 2.96, "learning_rate": 7.892680293820845e-06, "loss": 0.0752, "step": 126655 }, { "epoch": 2.96, "learning_rate": 7.891800239374809e-06, "loss": 0.108, "step": 126660 }, { "epoch": 2.96, "learning_rate": 7.890920184928776e-06, "loss": 0.0379, "step": 126665 }, { "epoch": 2.96, "learning_rate": 7.890040130482739e-06, "loss": 0.1915, "step": 126670 }, { "epoch": 2.96, "learning_rate": 7.889160076036706e-06, "loss": 0.1662, "step": 126675 }, { "epoch": 2.96, "learning_rate": 7.888280021590669e-06, "loss": 0.0121, "step": 126680 }, { "epoch": 2.96, "learning_rate": 7.887399967144634e-06, "loss": 0.0445, "step": 126685 }, { "epoch": 2.96, "learning_rate": 7.886519912698599e-06, "loss": 0.0404, "step": 126690 }, { "epoch": 2.96, "learning_rate": 7.885639858252564e-06, "loss": 0.0522, "step": 126695 }, { "epoch": 2.96, "learning_rate": 7.884759803806529e-06, "loss": 0.0776, "step": 126700 }, { "epoch": 2.96, "learning_rate": 7.883879749360494e-06, "loss": 0.1338, "step": 126705 }, { "epoch": 2.96, "learning_rate": 7.882999694914459e-06, "loss": 0.0357, "step": 126710 }, { "epoch": 2.96, "learning_rate": 7.882119640468424e-06, "loss": 0.1031, "step": 126715 }, { "epoch": 2.96, "learning_rate": 7.881239586022389e-06, "loss": 0.2405, "step": 126720 }, { "epoch": 2.96, "learning_rate": 7.880359531576354e-06, "loss": 0.2028, "step": 126725 }, { "epoch": 2.96, "learning_rate": 7.879479477130317e-06, "loss": 0.0066, "step": 126730 }, { "epoch": 2.96, "learning_rate": 7.878599422684284e-06, "loss": 0.0627, "step": 126735 }, { "epoch": 2.96, "learning_rate": 7.877719368238247e-06, "loss": 0.0067, "step": 126740 }, { "epoch": 2.96, "learning_rate": 7.876839313792214e-06, "loss": 0.0352, "step": 126745 }, { "epoch": 2.96, "learning_rate": 7.875959259346177e-06, "loss": 0.0877, "step": 126750 }, { "epoch": 2.96, "learning_rate": 7.875079204900144e-06, "loss": 0.188, "step": 126755 }, { "epoch": 2.96, "learning_rate": 7.874199150454108e-06, "loss": 0.1361, "step": 126760 }, { "epoch": 2.96, "learning_rate": 7.873319096008073e-06, "loss": 0.0528, "step": 126765 }, { "epoch": 2.96, "learning_rate": 7.87243904156204e-06, "loss": 0.1893, "step": 126770 }, { "epoch": 2.96, "learning_rate": 7.871558987116003e-06, "loss": 0.1882, "step": 126775 }, { "epoch": 2.96, "learning_rate": 7.87067893266997e-06, "loss": 0.0374, "step": 126780 }, { "epoch": 2.96, "learning_rate": 7.869798878223933e-06, "loss": 0.0274, "step": 126785 }, { "epoch": 2.96, "learning_rate": 7.8689188237779e-06, "loss": 0.0998, "step": 126790 }, { "epoch": 2.96, "learning_rate": 7.868038769331863e-06, "loss": 0.0494, "step": 126795 }, { "epoch": 2.96, "learning_rate": 7.867158714885828e-06, "loss": 0.0066, "step": 126800 }, { "epoch": 2.96, "learning_rate": 7.866278660439793e-06, "loss": 0.0181, "step": 126805 }, { "epoch": 2.96, "learning_rate": 7.865398605993758e-06, "loss": 0.0564, "step": 126810 }, { "epoch": 2.96, "learning_rate": 7.864518551547723e-06, "loss": 0.0841, "step": 126815 }, { "epoch": 2.96, "learning_rate": 7.863638497101688e-06, "loss": 0.3539, "step": 126820 }, { "epoch": 2.96, "learning_rate": 7.862758442655653e-06, "loss": 0.2467, "step": 126825 }, { "epoch": 2.96, "learning_rate": 7.861878388209618e-06, "loss": 0.0147, "step": 126830 }, { "epoch": 2.96, "learning_rate": 7.860998333763581e-06, "loss": 0.0287, "step": 126835 }, { "epoch": 2.96, "learning_rate": 7.860118279317548e-06, "loss": 0.0139, "step": 126840 }, { "epoch": 2.96, "learning_rate": 7.859238224871511e-06, "loss": 0.06, "step": 126845 }, { "epoch": 2.96, "learning_rate": 7.858358170425478e-06, "loss": 0.1652, "step": 126850 }, { "epoch": 2.96, "learning_rate": 7.857478115979441e-06, "loss": 0.0034, "step": 126855 }, { "epoch": 2.96, "learning_rate": 7.856598061533408e-06, "loss": 0.091, "step": 126860 }, { "epoch": 2.96, "learning_rate": 7.855718007087371e-06, "loss": 0.1223, "step": 126865 }, { "epoch": 2.96, "learning_rate": 7.854837952641338e-06, "loss": 0.0913, "step": 126870 }, { "epoch": 2.96, "learning_rate": 7.853957898195301e-06, "loss": 0.1852, "step": 126875 }, { "epoch": 2.96, "learning_rate": 7.853077843749266e-06, "loss": 0.0752, "step": 126880 }, { "epoch": 2.96, "learning_rate": 7.852197789303231e-06, "loss": 0.0266, "step": 126885 }, { "epoch": 2.96, "learning_rate": 7.851317734857196e-06, "loss": 0.0327, "step": 126890 }, { "epoch": 2.96, "learning_rate": 7.850437680411161e-06, "loss": 0.0511, "step": 126895 }, { "epoch": 2.96, "learning_rate": 7.849557625965126e-06, "loss": 0.0555, "step": 126900 }, { "epoch": 2.96, "learning_rate": 7.848677571519093e-06, "loss": 0.0191, "step": 126905 }, { "epoch": 2.96, "learning_rate": 7.847797517073056e-06, "loss": 0.0302, "step": 126910 }, { "epoch": 2.96, "learning_rate": 7.846917462627021e-06, "loss": 0.0836, "step": 126915 }, { "epoch": 2.96, "learning_rate": 7.846037408180986e-06, "loss": 0.1432, "step": 126920 }, { "epoch": 2.96, "learning_rate": 7.845157353734951e-06, "loss": 0.1563, "step": 126925 }, { "epoch": 2.96, "learning_rate": 7.844277299288917e-06, "loss": 0.0389, "step": 126930 }, { "epoch": 2.96, "learning_rate": 7.843397244842882e-06, "loss": 0.0589, "step": 126935 }, { "epoch": 2.96, "learning_rate": 7.842517190396847e-06, "loss": 0.0256, "step": 126940 }, { "epoch": 2.96, "learning_rate": 7.841637135950812e-06, "loss": 0.0272, "step": 126945 }, { "epoch": 2.96, "learning_rate": 7.840757081504775e-06, "loss": 0.0253, "step": 126950 }, { "epoch": 2.96, "learning_rate": 7.839877027058742e-06, "loss": 0.0929, "step": 126955 }, { "epoch": 2.96, "learning_rate": 7.838996972612705e-06, "loss": 0.0805, "step": 126960 }, { "epoch": 2.96, "learning_rate": 7.838116918166672e-06, "loss": 0.0958, "step": 126965 }, { "epoch": 2.96, "learning_rate": 7.837236863720635e-06, "loss": 0.2384, "step": 126970 }, { "epoch": 2.96, "learning_rate": 7.836356809274602e-06, "loss": 0.3902, "step": 126975 }, { "epoch": 2.96, "learning_rate": 7.835476754828565e-06, "loss": 0.0599, "step": 126980 }, { "epoch": 2.96, "learning_rate": 7.83459670038253e-06, "loss": 0.0127, "step": 126985 }, { "epoch": 2.96, "learning_rate": 7.833716645936495e-06, "loss": 0.1203, "step": 126990 }, { "epoch": 2.96, "learning_rate": 7.83283659149046e-06, "loss": 0.0304, "step": 126995 }, { "epoch": 2.96, "learning_rate": 7.831956537044425e-06, "loss": 0.0012, "step": 127000 }, { "epoch": 2.96, "learning_rate": 7.83107648259839e-06, "loss": 0.0324, "step": 127005 }, { "epoch": 2.96, "learning_rate": 7.830196428152355e-06, "loss": 0.1393, "step": 127010 }, { "epoch": 2.96, "learning_rate": 7.82931637370632e-06, "loss": 0.1175, "step": 127015 }, { "epoch": 2.96, "learning_rate": 7.828436319260285e-06, "loss": 0.1642, "step": 127020 }, { "epoch": 2.96, "learning_rate": 7.82755626481425e-06, "loss": 0.2522, "step": 127025 }, { "epoch": 2.96, "learning_rate": 7.826676210368214e-06, "loss": 0.121, "step": 127030 }, { "epoch": 2.96, "learning_rate": 7.82579615592218e-06, "loss": 0.0196, "step": 127035 }, { "epoch": 2.96, "learning_rate": 7.824916101476145e-06, "loss": 0.0234, "step": 127040 }, { "epoch": 2.96, "learning_rate": 7.82403604703011e-06, "loss": 0.0036, "step": 127045 }, { "epoch": 2.96, "learning_rate": 7.823155992584075e-06, "loss": 0.0142, "step": 127050 }, { "epoch": 2.96, "learning_rate": 7.82227593813804e-06, "loss": 0.0357, "step": 127055 }, { "epoch": 2.96, "learning_rate": 7.821395883692005e-06, "loss": 0.063, "step": 127060 }, { "epoch": 2.96, "learning_rate": 7.820515829245969e-06, "loss": 0.0193, "step": 127065 }, { "epoch": 2.96, "learning_rate": 7.819635774799935e-06, "loss": 0.1897, "step": 127070 }, { "epoch": 2.96, "learning_rate": 7.818755720353899e-06, "loss": 0.3175, "step": 127075 }, { "epoch": 2.96, "learning_rate": 7.817875665907865e-06, "loss": 0.0556, "step": 127080 }, { "epoch": 2.97, "learning_rate": 7.816995611461829e-06, "loss": 0.0324, "step": 127085 }, { "epoch": 2.97, "learning_rate": 7.816115557015795e-06, "loss": 0.0946, "step": 127090 }, { "epoch": 2.97, "learning_rate": 7.815235502569759e-06, "loss": 0.0482, "step": 127095 }, { "epoch": 2.97, "learning_rate": 7.814355448123724e-06, "loss": 0.0482, "step": 127100 }, { "epoch": 2.97, "learning_rate": 7.813475393677689e-06, "loss": 0.0876, "step": 127105 }, { "epoch": 2.97, "learning_rate": 7.812595339231654e-06, "loss": 0.033, "step": 127110 }, { "epoch": 2.97, "learning_rate": 7.811715284785619e-06, "loss": 0.1166, "step": 127115 }, { "epoch": 2.97, "learning_rate": 7.810835230339584e-06, "loss": 0.0605, "step": 127120 }, { "epoch": 2.97, "learning_rate": 7.809955175893549e-06, "loss": 0.5048, "step": 127125 }, { "epoch": 2.97, "learning_rate": 7.809075121447514e-06, "loss": 0.0249, "step": 127130 }, { "epoch": 2.97, "learning_rate": 7.808195067001477e-06, "loss": 0.0292, "step": 127135 }, { "epoch": 2.97, "learning_rate": 7.807315012555444e-06, "loss": 0.0609, "step": 127140 }, { "epoch": 2.97, "learning_rate": 7.806434958109407e-06, "loss": 0.0371, "step": 127145 }, { "epoch": 2.97, "learning_rate": 7.805554903663374e-06, "loss": 0.0421, "step": 127150 }, { "epoch": 2.97, "learning_rate": 7.804674849217337e-06, "loss": 0.0008, "step": 127155 }, { "epoch": 2.97, "learning_rate": 7.803794794771304e-06, "loss": 0.0832, "step": 127160 }, { "epoch": 2.97, "learning_rate": 7.802914740325267e-06, "loss": 0.1289, "step": 127165 }, { "epoch": 2.97, "learning_rate": 7.802034685879234e-06, "loss": 0.337, "step": 127170 }, { "epoch": 2.97, "learning_rate": 7.801154631433199e-06, "loss": 0.2619, "step": 127175 }, { "epoch": 2.97, "learning_rate": 7.800274576987162e-06, "loss": 0.0442, "step": 127180 }, { "epoch": 2.97, "learning_rate": 7.79939452254113e-06, "loss": 0.0185, "step": 127185 }, { "epoch": 2.97, "learning_rate": 7.798514468095093e-06, "loss": 0.0274, "step": 127190 }, { "epoch": 2.97, "learning_rate": 7.79763441364906e-06, "loss": 0.0812, "step": 127195 }, { "epoch": 2.97, "learning_rate": 7.796754359203023e-06, "loss": 0.1087, "step": 127200 }, { "epoch": 2.97, "learning_rate": 7.79587430475699e-06, "loss": 0.1239, "step": 127205 }, { "epoch": 2.97, "learning_rate": 7.794994250310953e-06, "loss": 0.0777, "step": 127210 }, { "epoch": 2.97, "learning_rate": 7.794114195864918e-06, "loss": 0.115, "step": 127215 }, { "epoch": 2.97, "learning_rate": 7.793234141418883e-06, "loss": 0.0495, "step": 127220 }, { "epoch": 2.97, "learning_rate": 7.792354086972848e-06, "loss": 0.223, "step": 127225 }, { "epoch": 2.97, "learning_rate": 7.791474032526813e-06, "loss": 0.0034, "step": 127230 }, { "epoch": 2.97, "learning_rate": 7.790593978080778e-06, "loss": 0.0281, "step": 127235 }, { "epoch": 2.97, "learning_rate": 7.789713923634743e-06, "loss": 0.0924, "step": 127240 }, { "epoch": 2.97, "learning_rate": 7.788833869188708e-06, "loss": 0.0093, "step": 127245 }, { "epoch": 2.97, "learning_rate": 7.787953814742671e-06, "loss": 0.0503, "step": 127250 }, { "epoch": 2.97, "learning_rate": 7.787073760296638e-06, "loss": 0.0662, "step": 127255 }, { "epoch": 2.97, "learning_rate": 7.786193705850601e-06, "loss": 0.0975, "step": 127260 }, { "epoch": 2.97, "learning_rate": 7.785313651404568e-06, "loss": 0.2771, "step": 127265 }, { "epoch": 2.97, "learning_rate": 7.784433596958531e-06, "loss": 0.1043, "step": 127270 }, { "epoch": 2.97, "learning_rate": 7.783553542512498e-06, "loss": 0.2677, "step": 127275 }, { "epoch": 2.97, "learning_rate": 7.782673488066461e-06, "loss": 0.0011, "step": 127280 }, { "epoch": 2.97, "learning_rate": 7.781793433620426e-06, "loss": 0.0109, "step": 127285 }, { "epoch": 2.97, "learning_rate": 7.780913379174391e-06, "loss": 0.014, "step": 127290 }, { "epoch": 2.97, "learning_rate": 7.780033324728356e-06, "loss": 0.0402, "step": 127295 }, { "epoch": 2.97, "learning_rate": 7.779153270282321e-06, "loss": 0.079, "step": 127300 }, { "epoch": 2.97, "learning_rate": 7.778273215836286e-06, "loss": 0.0205, "step": 127305 }, { "epoch": 2.97, "learning_rate": 7.777393161390251e-06, "loss": 0.0388, "step": 127310 }, { "epoch": 2.97, "learning_rate": 7.776513106944216e-06, "loss": 0.2385, "step": 127315 }, { "epoch": 2.97, "learning_rate": 7.775633052498183e-06, "loss": 0.1209, "step": 127320 }, { "epoch": 2.97, "learning_rate": 7.774752998052146e-06, "loss": 0.0966, "step": 127325 }, { "epoch": 2.97, "learning_rate": 7.773872943606111e-06, "loss": 0.0398, "step": 127330 }, { "epoch": 2.97, "learning_rate": 7.772992889160076e-06, "loss": 0.0134, "step": 127335 }, { "epoch": 2.97, "learning_rate": 7.772112834714041e-06, "loss": 0.0171, "step": 127340 }, { "epoch": 2.97, "learning_rate": 7.771232780268006e-06, "loss": 0.0003, "step": 127345 }, { "epoch": 2.97, "learning_rate": 7.770352725821971e-06, "loss": 0.0977, "step": 127350 }, { "epoch": 2.97, "learning_rate": 7.769472671375936e-06, "loss": 0.0249, "step": 127355 }, { "epoch": 2.97, "learning_rate": 7.768592616929902e-06, "loss": 0.0172, "step": 127360 }, { "epoch": 2.97, "learning_rate": 7.767712562483865e-06, "loss": 0.1341, "step": 127365 }, { "epoch": 2.97, "learning_rate": 7.766832508037832e-06, "loss": 0.0742, "step": 127370 }, { "epoch": 2.97, "learning_rate": 7.765952453591795e-06, "loss": 0.3973, "step": 127375 }, { "epoch": 2.97, "learning_rate": 7.765072399145762e-06, "loss": 0.0396, "step": 127380 }, { "epoch": 2.97, "learning_rate": 7.764192344699725e-06, "loss": 0.0411, "step": 127385 }, { "epoch": 2.97, "learning_rate": 7.763312290253692e-06, "loss": 0.0155, "step": 127390 }, { "epoch": 2.97, "learning_rate": 7.762432235807655e-06, "loss": 0.0422, "step": 127395 }, { "epoch": 2.97, "learning_rate": 7.76155218136162e-06, "loss": 0.0208, "step": 127400 }, { "epoch": 2.97, "learning_rate": 7.760672126915585e-06, "loss": 0.0515, "step": 127405 }, { "epoch": 2.97, "learning_rate": 7.75979207246955e-06, "loss": 0.0353, "step": 127410 }, { "epoch": 2.97, "learning_rate": 7.758912018023515e-06, "loss": 0.0933, "step": 127415 }, { "epoch": 2.97, "learning_rate": 7.75803196357748e-06, "loss": 0.2466, "step": 127420 }, { "epoch": 2.97, "learning_rate": 7.757151909131445e-06, "loss": 0.1583, "step": 127425 }, { "epoch": 2.97, "learning_rate": 7.75627185468541e-06, "loss": 0.0509, "step": 127430 }, { "epoch": 2.97, "learning_rate": 7.755391800239373e-06, "loss": 0.0004, "step": 127435 }, { "epoch": 2.97, "learning_rate": 7.75451174579334e-06, "loss": 0.0007, "step": 127440 }, { "epoch": 2.97, "learning_rate": 7.753631691347305e-06, "loss": 0.1198, "step": 127445 }, { "epoch": 2.97, "learning_rate": 7.75275163690127e-06, "loss": 0.03, "step": 127450 }, { "epoch": 2.97, "learning_rate": 7.751871582455235e-06, "loss": 0.0444, "step": 127455 }, { "epoch": 2.97, "learning_rate": 7.7509915280092e-06, "loss": 0.1006, "step": 127460 }, { "epoch": 2.97, "learning_rate": 7.750111473563165e-06, "loss": 0.0484, "step": 127465 }, { "epoch": 2.97, "learning_rate": 7.74923141911713e-06, "loss": 0.2406, "step": 127470 }, { "epoch": 2.97, "learning_rate": 7.748351364671095e-06, "loss": 0.2322, "step": 127475 }, { "epoch": 2.97, "learning_rate": 7.747471310225059e-06, "loss": 0.0248, "step": 127480 }, { "epoch": 2.97, "learning_rate": 7.746591255779025e-06, "loss": 0.0579, "step": 127485 }, { "epoch": 2.97, "learning_rate": 7.745711201332989e-06, "loss": 0.0575, "step": 127490 }, { "epoch": 2.97, "learning_rate": 7.744831146886955e-06, "loss": 0.0152, "step": 127495 }, { "epoch": 2.97, "learning_rate": 7.743951092440919e-06, "loss": 0.0793, "step": 127500 }, { "epoch": 2.97, "learning_rate": 7.743071037994885e-06, "loss": 0.0392, "step": 127505 }, { "epoch": 2.97, "learning_rate": 7.742190983548849e-06, "loss": 0.0091, "step": 127510 }, { "epoch": 2.98, "learning_rate": 7.741310929102814e-06, "loss": 0.0709, "step": 127515 }, { "epoch": 2.98, "learning_rate": 7.740430874656779e-06, "loss": 0.1927, "step": 127520 }, { "epoch": 2.98, "learning_rate": 7.739550820210744e-06, "loss": 0.1752, "step": 127525 }, { "epoch": 2.98, "learning_rate": 7.738670765764709e-06, "loss": 0.0535, "step": 127530 }, { "epoch": 2.98, "learning_rate": 7.737790711318674e-06, "loss": 0.0148, "step": 127535 }, { "epoch": 2.98, "learning_rate": 7.736910656872639e-06, "loss": 0.0305, "step": 127540 }, { "epoch": 2.98, "learning_rate": 7.736030602426604e-06, "loss": 0.0114, "step": 127545 }, { "epoch": 2.98, "learning_rate": 7.735150547980567e-06, "loss": 0.0523, "step": 127550 }, { "epoch": 2.98, "learning_rate": 7.734270493534534e-06, "loss": 0.0393, "step": 127555 }, { "epoch": 2.98, "learning_rate": 7.733390439088497e-06, "loss": 0.0926, "step": 127560 }, { "epoch": 2.98, "learning_rate": 7.732510384642464e-06, "loss": 0.1629, "step": 127565 }, { "epoch": 2.98, "learning_rate": 7.731630330196427e-06, "loss": 0.3026, "step": 127570 }, { "epoch": 2.98, "learning_rate": 7.730750275750394e-06, "loss": 0.1247, "step": 127575 }, { "epoch": 2.98, "learning_rate": 7.729870221304357e-06, "loss": 0.0104, "step": 127580 }, { "epoch": 2.98, "learning_rate": 7.728990166858322e-06, "loss": 0.0362, "step": 127585 }, { "epoch": 2.98, "learning_rate": 7.728110112412289e-06, "loss": 0.017, "step": 127590 }, { "epoch": 2.98, "learning_rate": 7.727230057966252e-06, "loss": 0.088, "step": 127595 }, { "epoch": 2.98, "learning_rate": 7.726350003520219e-06, "loss": 0.0659, "step": 127600 }, { "epoch": 2.98, "learning_rate": 7.725469949074182e-06, "loss": 0.0544, "step": 127605 }, { "epoch": 2.98, "learning_rate": 7.724589894628149e-06, "loss": 0.1307, "step": 127610 }, { "epoch": 2.98, "learning_rate": 7.723709840182112e-06, "loss": 0.1017, "step": 127615 }, { "epoch": 2.98, "learning_rate": 7.72282978573608e-06, "loss": 0.1368, "step": 127620 }, { "epoch": 2.98, "learning_rate": 7.721949731290043e-06, "loss": 0.5501, "step": 127625 }, { "epoch": 2.98, "learning_rate": 7.721069676844008e-06, "loss": 0.0275, "step": 127630 }, { "epoch": 2.98, "learning_rate": 7.720189622397973e-06, "loss": 0.0019, "step": 127635 }, { "epoch": 2.98, "learning_rate": 7.719309567951938e-06, "loss": 0.0288, "step": 127640 }, { "epoch": 2.98, "learning_rate": 7.718429513505903e-06, "loss": 0.0342, "step": 127645 }, { "epoch": 2.98, "learning_rate": 7.717549459059868e-06, "loss": 0.0267, "step": 127650 }, { "epoch": 2.98, "learning_rate": 7.716669404613833e-06, "loss": 0.009, "step": 127655 }, { "epoch": 2.98, "learning_rate": 7.715789350167798e-06, "loss": 0.172, "step": 127660 }, { "epoch": 2.98, "learning_rate": 7.714909295721761e-06, "loss": 0.0702, "step": 127665 }, { "epoch": 2.98, "learning_rate": 7.714029241275728e-06, "loss": 0.3438, "step": 127670 }, { "epoch": 2.98, "learning_rate": 7.713149186829691e-06, "loss": 0.2168, "step": 127675 }, { "epoch": 2.98, "learning_rate": 7.712269132383658e-06, "loss": 0.066, "step": 127680 }, { "epoch": 2.98, "learning_rate": 7.711389077937621e-06, "loss": 0.057, "step": 127685 }, { "epoch": 2.98, "learning_rate": 7.710509023491588e-06, "loss": 0.1015, "step": 127690 }, { "epoch": 2.98, "learning_rate": 7.709628969045551e-06, "loss": 0.0822, "step": 127695 }, { "epoch": 2.98, "learning_rate": 7.708748914599516e-06, "loss": 0.1866, "step": 127700 }, { "epoch": 2.98, "learning_rate": 7.707868860153481e-06, "loss": 0.1094, "step": 127705 }, { "epoch": 2.98, "learning_rate": 7.706988805707446e-06, "loss": 0.0316, "step": 127710 }, { "epoch": 2.98, "learning_rate": 7.706108751261411e-06, "loss": 0.1104, "step": 127715 }, { "epoch": 2.98, "learning_rate": 7.705228696815376e-06, "loss": 0.0868, "step": 127720 }, { "epoch": 2.98, "learning_rate": 7.704348642369343e-06, "loss": 0.3125, "step": 127725 }, { "epoch": 2.98, "learning_rate": 7.703468587923306e-06, "loss": 0.0004, "step": 127730 }, { "epoch": 2.98, "learning_rate": 7.702588533477271e-06, "loss": 0.1336, "step": 127735 }, { "epoch": 2.98, "learning_rate": 7.701708479031236e-06, "loss": 0.0644, "step": 127740 }, { "epoch": 2.98, "learning_rate": 7.700828424585201e-06, "loss": 0.0186, "step": 127745 }, { "epoch": 2.98, "learning_rate": 7.699948370139166e-06, "loss": 0.1628, "step": 127750 }, { "epoch": 2.98, "learning_rate": 7.699068315693131e-06, "loss": 0.113, "step": 127755 }, { "epoch": 2.98, "learning_rate": 7.698188261247096e-06, "loss": 0.1192, "step": 127760 }, { "epoch": 2.98, "learning_rate": 7.697308206801061e-06, "loss": 0.1094, "step": 127765 }, { "epoch": 2.98, "learning_rate": 7.696428152355025e-06, "loss": 0.2374, "step": 127770 }, { "epoch": 2.98, "learning_rate": 7.695548097908991e-06, "loss": 0.2724, "step": 127775 }, { "epoch": 2.98, "learning_rate": 7.694668043462955e-06, "loss": 0.0293, "step": 127780 }, { "epoch": 2.98, "learning_rate": 7.693787989016921e-06, "loss": 0.112, "step": 127785 }, { "epoch": 2.98, "learning_rate": 7.692907934570885e-06, "loss": 0.0164, "step": 127790 }, { "epoch": 2.98, "learning_rate": 7.692027880124852e-06, "loss": 0.0469, "step": 127795 }, { "epoch": 2.98, "learning_rate": 7.691147825678815e-06, "loss": 0.0558, "step": 127800 }, { "epoch": 2.98, "learning_rate": 7.690267771232782e-06, "loss": 0.0534, "step": 127805 }, { "epoch": 2.98, "learning_rate": 7.689387716786745e-06, "loss": 0.0771, "step": 127810 }, { "epoch": 2.98, "learning_rate": 7.68850766234071e-06, "loss": 0.1075, "step": 127815 }, { "epoch": 2.98, "learning_rate": 7.687627607894675e-06, "loss": 0.5176, "step": 127820 }, { "epoch": 2.98, "learning_rate": 7.68674755344864e-06, "loss": 0.1868, "step": 127825 }, { "epoch": 2.98, "learning_rate": 7.685867499002605e-06, "loss": 0.0517, "step": 127830 }, { "epoch": 2.98, "learning_rate": 7.68498744455657e-06, "loss": 0.0019, "step": 127835 }, { "epoch": 2.98, "learning_rate": 7.684107390110535e-06, "loss": 0.0059, "step": 127840 }, { "epoch": 2.98, "learning_rate": 7.6832273356645e-06, "loss": 0.0031, "step": 127845 }, { "epoch": 2.98, "learning_rate": 7.682347281218463e-06, "loss": 0.035, "step": 127850 }, { "epoch": 2.98, "learning_rate": 7.68146722677243e-06, "loss": 0.0337, "step": 127855 }, { "epoch": 2.98, "learning_rate": 7.680587172326395e-06, "loss": 0.0377, "step": 127860 }, { "epoch": 2.98, "learning_rate": 7.67970711788036e-06, "loss": 0.0581, "step": 127865 }, { "epoch": 2.98, "learning_rate": 7.678827063434325e-06, "loss": 0.2087, "step": 127870 }, { "epoch": 2.98, "learning_rate": 7.67794700898829e-06, "loss": 0.2159, "step": 127875 }, { "epoch": 2.98, "learning_rate": 7.677066954542255e-06, "loss": 0.0643, "step": 127880 }, { "epoch": 2.98, "learning_rate": 7.676186900096218e-06, "loss": 0.0372, "step": 127885 }, { "epoch": 2.98, "learning_rate": 7.675306845650185e-06, "loss": 0.038, "step": 127890 }, { "epoch": 2.98, "learning_rate": 7.674426791204149e-06, "loss": 0.0232, "step": 127895 }, { "epoch": 2.98, "learning_rate": 7.673546736758115e-06, "loss": 0.0867, "step": 127900 }, { "epoch": 2.98, "learning_rate": 7.672666682312079e-06, "loss": 0.0085, "step": 127905 }, { "epoch": 2.98, "learning_rate": 7.671786627866045e-06, "loss": 0.1326, "step": 127910 }, { "epoch": 2.98, "learning_rate": 7.670906573420009e-06, "loss": 0.0452, "step": 127915 }, { "epoch": 2.98, "learning_rate": 7.670026518973974e-06, "loss": 0.2316, "step": 127920 }, { "epoch": 2.98, "learning_rate": 7.669146464527939e-06, "loss": 0.1844, "step": 127925 }, { "epoch": 2.98, "learning_rate": 7.668266410081904e-06, "loss": 0.1397, "step": 127930 }, { "epoch": 2.98, "learning_rate": 7.667386355635869e-06, "loss": 0.0005, "step": 127935 }, { "epoch": 2.98, "learning_rate": 7.666506301189834e-06, "loss": 0.0362, "step": 127940 }, { "epoch": 2.99, "learning_rate": 7.665626246743799e-06, "loss": 0.0619, "step": 127945 }, { "epoch": 2.99, "learning_rate": 7.664746192297764e-06, "loss": 0.0822, "step": 127950 }, { "epoch": 2.99, "learning_rate": 7.663866137851729e-06, "loss": 0.1014, "step": 127955 }, { "epoch": 2.99, "learning_rate": 7.662986083405694e-06, "loss": 0.0494, "step": 127960 }, { "epoch": 2.99, "learning_rate": 7.662106028959657e-06, "loss": 0.3472, "step": 127965 }, { "epoch": 2.99, "learning_rate": 7.661225974513624e-06, "loss": 0.1413, "step": 127970 }, { "epoch": 2.99, "learning_rate": 7.660345920067587e-06, "loss": 0.2928, "step": 127975 }, { "epoch": 2.99, "learning_rate": 7.659465865621554e-06, "loss": 0.0668, "step": 127980 }, { "epoch": 2.99, "learning_rate": 7.658585811175517e-06, "loss": 0.0817, "step": 127985 }, { "epoch": 2.99, "learning_rate": 7.657705756729484e-06, "loss": 0.0251, "step": 127990 }, { "epoch": 2.99, "learning_rate": 7.656825702283449e-06, "loss": 0.117, "step": 127995 }, { "epoch": 2.99, "learning_rate": 7.655945647837412e-06, "loss": 0.1196, "step": 128000 }, { "epoch": 2.99, "learning_rate": 7.655065593391379e-06, "loss": 0.019, "step": 128005 }, { "epoch": 2.99, "learning_rate": 7.654185538945342e-06, "loss": 0.1118, "step": 128010 }, { "epoch": 2.99, "learning_rate": 7.653305484499309e-06, "loss": 0.1144, "step": 128015 }, { "epoch": 2.99, "learning_rate": 7.652425430053272e-06, "loss": 0.1668, "step": 128020 }, { "epoch": 2.99, "learning_rate": 7.651545375607239e-06, "loss": 0.1874, "step": 128025 }, { "epoch": 2.99, "learning_rate": 7.650665321161202e-06, "loss": 0.0362, "step": 128030 }, { "epoch": 2.99, "learning_rate": 7.649785266715167e-06, "loss": 0.0018, "step": 128035 }, { "epoch": 2.99, "learning_rate": 7.648905212269132e-06, "loss": 0.0479, "step": 128040 }, { "epoch": 2.99, "learning_rate": 7.648025157823097e-06, "loss": 0.1272, "step": 128045 }, { "epoch": 2.99, "learning_rate": 7.647145103377062e-06, "loss": 0.0012, "step": 128050 }, { "epoch": 2.99, "learning_rate": 7.646265048931028e-06, "loss": 0.1259, "step": 128055 }, { "epoch": 2.99, "learning_rate": 7.645384994484993e-06, "loss": 0.0233, "step": 128060 }, { "epoch": 2.99, "learning_rate": 7.644504940038958e-06, "loss": 0.1245, "step": 128065 }, { "epoch": 2.99, "learning_rate": 7.643624885592921e-06, "loss": 0.0809, "step": 128070 }, { "epoch": 2.99, "learning_rate": 7.642744831146888e-06, "loss": 0.3228, "step": 128075 }, { "epoch": 2.99, "learning_rate": 7.641864776700851e-06, "loss": 0.0552, "step": 128080 }, { "epoch": 2.99, "learning_rate": 7.640984722254818e-06, "loss": 0.0137, "step": 128085 }, { "epoch": 2.99, "learning_rate": 7.640104667808781e-06, "loss": 0.0009, "step": 128090 }, { "epoch": 2.99, "learning_rate": 7.639224613362748e-06, "loss": 0.074, "step": 128095 }, { "epoch": 2.99, "learning_rate": 7.638344558916711e-06, "loss": 0.0165, "step": 128100 }, { "epoch": 2.99, "learning_rate": 7.637464504470678e-06, "loss": 0.1257, "step": 128105 }, { "epoch": 2.99, "learning_rate": 7.636584450024641e-06, "loss": 0.0289, "step": 128110 }, { "epoch": 2.99, "learning_rate": 7.635704395578606e-06, "loss": 0.031, "step": 128115 }, { "epoch": 2.99, "learning_rate": 7.634824341132571e-06, "loss": 0.1102, "step": 128120 }, { "epoch": 2.99, "learning_rate": 7.633944286686536e-06, "loss": 0.1302, "step": 128125 }, { "epoch": 2.99, "learning_rate": 7.633064232240503e-06, "loss": 0.0289, "step": 128130 }, { "epoch": 2.99, "learning_rate": 7.632184177794466e-06, "loss": 0.0298, "step": 128135 }, { "epoch": 2.99, "learning_rate": 7.631304123348433e-06, "loss": 0.0088, "step": 128140 }, { "epoch": 2.99, "learning_rate": 7.630424068902396e-06, "loss": 0.0226, "step": 128145 }, { "epoch": 2.99, "learning_rate": 7.629544014456361e-06, "loss": 0.0581, "step": 128150 }, { "epoch": 2.99, "learning_rate": 7.628663960010326e-06, "loss": 0.0213, "step": 128155 }, { "epoch": 2.99, "learning_rate": 7.62778390556429e-06, "loss": 0.174, "step": 128160 }, { "epoch": 2.99, "learning_rate": 7.626903851118256e-06, "loss": 0.0449, "step": 128165 }, { "epoch": 2.99, "learning_rate": 7.6260237966722204e-06, "loss": 0.1814, "step": 128170 }, { "epoch": 2.99, "learning_rate": 7.625143742226186e-06, "loss": 0.258, "step": 128175 }, { "epoch": 2.99, "learning_rate": 7.6242636877801505e-06, "loss": 0.0449, "step": 128180 }, { "epoch": 2.99, "learning_rate": 7.623383633334115e-06, "loss": 0.0059, "step": 128185 }, { "epoch": 2.99, "learning_rate": 7.6225035788880805e-06, "loss": 0.0402, "step": 128190 }, { "epoch": 2.99, "learning_rate": 7.6216235244420455e-06, "loss": 0.0002, "step": 128195 }, { "epoch": 2.99, "learning_rate": 7.620743469996011e-06, "loss": 0.0369, "step": 128200 }, { "epoch": 2.99, "learning_rate": 7.6198634155499756e-06, "loss": 0.0794, "step": 128205 }, { "epoch": 2.99, "learning_rate": 7.6189833611039414e-06, "loss": 0.0823, "step": 128210 }, { "epoch": 2.99, "learning_rate": 7.618103306657906e-06, "loss": 0.0198, "step": 128215 }, { "epoch": 2.99, "learning_rate": 7.61722325221187e-06, "loss": 0.1231, "step": 128220 }, { "epoch": 2.99, "learning_rate": 7.616343197765836e-06, "loss": 0.2078, "step": 128225 }, { "epoch": 2.99, "learning_rate": 7.6154631433198e-06, "loss": 0.064, "step": 128230 }, { "epoch": 2.99, "learning_rate": 7.614583088873766e-06, "loss": 0.0067, "step": 128235 }, { "epoch": 2.99, "learning_rate": 7.61370303442773e-06, "loss": 0.0016, "step": 128240 }, { "epoch": 2.99, "learning_rate": 7.612822979981696e-06, "loss": 0.0681, "step": 128245 }, { "epoch": 2.99, "learning_rate": 7.61194292553566e-06, "loss": 0.0492, "step": 128250 }, { "epoch": 2.99, "learning_rate": 7.611062871089626e-06, "loss": 0.0758, "step": 128255 }, { "epoch": 2.99, "learning_rate": 7.61018281664359e-06, "loss": 0.0293, "step": 128260 }, { "epoch": 2.99, "learning_rate": 7.609302762197554e-06, "loss": 0.3261, "step": 128265 }, { "epoch": 2.99, "learning_rate": 7.60842270775152e-06, "loss": 0.1991, "step": 128270 }, { "epoch": 2.99, "learning_rate": 7.607542653305484e-06, "loss": 0.2059, "step": 128275 }, { "epoch": 2.99, "learning_rate": 7.60666259885945e-06, "loss": 0.048, "step": 128280 }, { "epoch": 2.99, "learning_rate": 7.605782544413414e-06, "loss": 0.0136, "step": 128285 }, { "epoch": 2.99, "learning_rate": 7.60490248996738e-06, "loss": 0.155, "step": 128290 }, { "epoch": 2.99, "learning_rate": 7.604022435521344e-06, "loss": 0.0437, "step": 128295 }, { "epoch": 2.99, "learning_rate": 7.6031423810753084e-06, "loss": 0.0285, "step": 128300 }, { "epoch": 2.99, "learning_rate": 7.602262326629274e-06, "loss": 0.0179, "step": 128305 }, { "epoch": 2.99, "learning_rate": 7.6013822721832385e-06, "loss": 0.2294, "step": 128310 }, { "epoch": 2.99, "learning_rate": 7.600502217737204e-06, "loss": 0.081, "step": 128315 }, { "epoch": 2.99, "learning_rate": 7.5996221632911685e-06, "loss": 0.1729, "step": 128320 }, { "epoch": 2.99, "learning_rate": 7.598742108845134e-06, "loss": 0.3138, "step": 128325 }, { "epoch": 2.99, "learning_rate": 7.597862054399099e-06, "loss": 0.0343, "step": 128330 }, { "epoch": 2.99, "learning_rate": 7.5969819999530636e-06, "loss": 0.0041, "step": 128335 }, { "epoch": 2.99, "learning_rate": 7.5961019455070294e-06, "loss": 0.0987, "step": 128340 }, { "epoch": 2.99, "learning_rate": 7.595221891060994e-06, "loss": 0.0087, "step": 128345 }, { "epoch": 2.99, "learning_rate": 7.5943418366149595e-06, "loss": 0.1486, "step": 128350 }, { "epoch": 2.99, "learning_rate": 7.593461782168924e-06, "loss": 0.155, "step": 128355 }, { "epoch": 2.99, "learning_rate": 7.5925817277228895e-06, "loss": 0.2256, "step": 128360 }, { "epoch": 2.99, "learning_rate": 7.591701673276854e-06, "loss": 0.0418, "step": 128365 }, { "epoch": 3.0, "learning_rate": 7.590821618830818e-06, "loss": 0.1633, "step": 128370 }, { "epoch": 3.0, "learning_rate": 7.589941564384784e-06, "loss": 0.0606, "step": 128375 }, { "epoch": 3.0, "learning_rate": 7.589061509938748e-06, "loss": 0.0103, "step": 128380 }, { "epoch": 3.0, "learning_rate": 7.588181455492714e-06, "loss": 0.023, "step": 128385 }, { "epoch": 3.0, "learning_rate": 7.587301401046678e-06, "loss": 0.0751, "step": 128390 }, { "epoch": 3.0, "learning_rate": 7.586421346600644e-06, "loss": 0.0544, "step": 128395 }, { "epoch": 3.0, "learning_rate": 7.585541292154608e-06, "loss": 0.0767, "step": 128400 }, { "epoch": 3.0, "learning_rate": 7.584661237708574e-06, "loss": 0.0432, "step": 128405 }, { "epoch": 3.0, "learning_rate": 7.583781183262538e-06, "loss": 0.083, "step": 128410 }, { "epoch": 3.0, "learning_rate": 7.582901128816502e-06, "loss": 0.1642, "step": 128415 }, { "epoch": 3.0, "learning_rate": 7.582021074370468e-06, "loss": 0.2308, "step": 128420 }, { "epoch": 3.0, "learning_rate": 7.581141019924432e-06, "loss": 0.3236, "step": 128425 }, { "epoch": 3.0, "learning_rate": 7.580260965478398e-06, "loss": 0.0343, "step": 128430 }, { "epoch": 3.0, "learning_rate": 7.579380911032362e-06, "loss": 0.0011, "step": 128435 }, { "epoch": 3.0, "learning_rate": 7.578500856586328e-06, "loss": 0.075, "step": 128440 }, { "epoch": 3.0, "learning_rate": 7.577620802140292e-06, "loss": 0.0695, "step": 128445 }, { "epoch": 3.0, "learning_rate": 7.5767407476942565e-06, "loss": 0.0743, "step": 128450 }, { "epoch": 3.0, "learning_rate": 7.575860693248222e-06, "loss": 0.0911, "step": 128455 }, { "epoch": 3.0, "learning_rate": 7.5749806388021865e-06, "loss": 0.0993, "step": 128460 }, { "epoch": 3.0, "learning_rate": 7.574100584356152e-06, "loss": 0.0465, "step": 128465 }, { "epoch": 3.0, "learning_rate": 7.5732205299101174e-06, "loss": 0.1676, "step": 128470 }, { "epoch": 3.0, "learning_rate": 7.572340475464083e-06, "loss": 0.3342, "step": 128475 }, { "epoch": 3.0, "learning_rate": 7.5714604210180475e-06, "loss": 0.0292, "step": 128480 }, { "epoch": 3.0, "learning_rate": 7.570580366572012e-06, "loss": 0.0127, "step": 128485 }, { "epoch": 3.0, "learning_rate": 7.5697003121259775e-06, "loss": 0.0594, "step": 128490 }, { "epoch": 3.0, "learning_rate": 7.568820257679942e-06, "loss": 0.0652, "step": 128495 }, { "epoch": 3.0, "learning_rate": 7.5679402032339076e-06, "loss": 0.0372, "step": 128500 }, { "epoch": 3.0, "learning_rate": 7.567060148787872e-06, "loss": 0.1078, "step": 128505 }, { "epoch": 3.0, "learning_rate": 7.566180094341838e-06, "loss": 0.0663, "step": 128510 }, { "epoch": 3.0, "learning_rate": 7.565300039895802e-06, "loss": 0.1317, "step": 128515 }, { "epoch": 3.0, "learning_rate": 7.564419985449766e-06, "loss": 0.2873, "step": 128520 }, { "epoch": 3.0, "learning_rate": 7.563539931003732e-06, "loss": 0.2519, "step": 128525 }, { "epoch": 3.0, "learning_rate": 7.562659876557696e-06, "loss": 0.0413, "step": 128530 }, { "epoch": 3.0, "learning_rate": 7.561779822111662e-06, "loss": 0.0587, "step": 128535 }, { "epoch": 3.0, "learning_rate": 7.560899767665626e-06, "loss": 0.0078, "step": 128540 }, { "epoch": 3.0, "learning_rate": 7.560019713219592e-06, "loss": 0.0536, "step": 128545 }, { "epoch": 3.0, "learning_rate": 7.559139658773556e-06, "loss": 0.0718, "step": 128550 }, { "epoch": 3.0, "learning_rate": 7.558259604327522e-06, "loss": 0.0847, "step": 128555 }, { "epoch": 3.0, "learning_rate": 7.557379549881486e-06, "loss": 0.0619, "step": 128560 }, { "epoch": 3.0, "learning_rate": 7.55649949543545e-06, "loss": 0.1693, "step": 128565 }, { "epoch": 3.0, "learning_rate": 7.555619440989416e-06, "loss": 0.2141, "step": 128570 }, { "epoch": 3.0, "learning_rate": 7.55473938654338e-06, "loss": 0.1441, "step": 128575 }, { "epoch": 3.0, "learning_rate": 7.553859332097346e-06, "loss": 0.0031, "step": 128580 }, { "epoch": 3.0, "eval_cer": 0.010451914362615392, "eval_loss": 0.026185082271695137, "eval_runtime": 690.0155, "eval_samples_per_second": 27.608, "eval_steps_per_second": 6.903, "eval_wer": 0.08524861048520355, "step": 128583 } ], "max_steps": 171444, "num_train_epochs": 4, "total_flos": 2.5584430284220125e+19, "trial_name": null, "trial_params": null }