{ "best_metric": 0.024219496175646782, "best_model_checkpoint": "wav2vec2-base-pem123-32-960h-la/checkpoint-64293", "epoch": 3.0, "global_step": 64293, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 1.5000000000000002e-07, "loss": 2.7078, "step": 5 }, { "epoch": 0.0, "learning_rate": 3.0000000000000004e-07, "loss": 3.2934, "step": 10 }, { "epoch": 0.0, "learning_rate": 4.5e-07, "loss": 3.5202, "step": 15 }, { "epoch": 0.0, "learning_rate": 6.000000000000001e-07, "loss": 3.0796, "step": 20 }, { "epoch": 0.0, "learning_rate": 7.5e-07, "loss": 3.4226, "step": 25 }, { "epoch": 0.0, "learning_rate": 9e-07, "loss": 3.3138, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.0500000000000001e-06, "loss": 3.4498, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.2000000000000002e-06, "loss": 3.3835, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.35e-06, "loss": 3.3571, "step": 45 }, { "epoch": 0.0, "learning_rate": 1.5e-06, "loss": 3.8856, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.65e-06, "loss": 2.0857, "step": 55 }, { "epoch": 0.0, "learning_rate": 1.8e-06, "loss": 2.6621, "step": 60 }, { "epoch": 0.0, "learning_rate": 1.95e-06, "loss": 2.2829, "step": 65 }, { "epoch": 0.0, "learning_rate": 2.1000000000000002e-06, "loss": 2.2943, "step": 70 }, { "epoch": 0.0, "learning_rate": 2.25e-06, "loss": 2.3434, "step": 75 }, { "epoch": 0.0, "learning_rate": 2.4000000000000003e-06, "loss": 2.4087, "step": 80 }, { "epoch": 0.0, "learning_rate": 2.55e-06, "loss": 2.1228, "step": 85 }, { "epoch": 0.0, "learning_rate": 2.7e-06, "loss": 2.2721, "step": 90 }, { "epoch": 0.0, "learning_rate": 2.8500000000000002e-06, "loss": 2.4505, "step": 95 }, { "epoch": 0.0, "learning_rate": 3e-06, "loss": 2.6655, "step": 100 }, { "epoch": 0.0, "learning_rate": 3.15e-06, "loss": 1.4593, "step": 105 }, { "epoch": 0.01, "learning_rate": 3.3e-06, "loss": 1.4055, "step": 110 }, { "epoch": 0.01, "learning_rate": 3.4500000000000004e-06, "loss": 1.485, "step": 115 }, { "epoch": 0.01, "learning_rate": 3.6e-06, "loss": 1.5149, "step": 120 }, { "epoch": 0.01, "learning_rate": 3.75e-06, "loss": 1.3764, "step": 125 }, { "epoch": 0.01, "learning_rate": 3.9e-06, "loss": 1.6748, "step": 130 }, { "epoch": 0.01, "learning_rate": 4.05e-06, "loss": 1.5351, "step": 135 }, { "epoch": 0.01, "learning_rate": 4.2000000000000004e-06, "loss": 1.6763, "step": 140 }, { "epoch": 0.01, "learning_rate": 4.35e-06, "loss": 1.7393, "step": 145 }, { "epoch": 0.01, "learning_rate": 4.5e-06, "loss": 1.6164, "step": 150 }, { "epoch": 0.01, "learning_rate": 4.65e-06, "loss": 1.3382, "step": 155 }, { "epoch": 0.01, "learning_rate": 4.800000000000001e-06, "loss": 1.028, "step": 160 }, { "epoch": 0.01, "learning_rate": 4.95e-06, "loss": 1.1416, "step": 165 }, { "epoch": 0.01, "learning_rate": 5.1e-06, "loss": 0.978, "step": 170 }, { "epoch": 0.01, "learning_rate": 5.25e-06, "loss": 1.169, "step": 175 }, { "epoch": 0.01, "learning_rate": 5.4e-06, "loss": 1.2869, "step": 180 }, { "epoch": 0.01, "learning_rate": 5.55e-06, "loss": 1.5859, "step": 185 }, { "epoch": 0.01, "learning_rate": 5.7000000000000005e-06, "loss": 1.2329, "step": 190 }, { "epoch": 0.01, "learning_rate": 5.850000000000001e-06, "loss": 1.4542, "step": 195 }, { "epoch": 0.01, "learning_rate": 6e-06, "loss": 1.8286, "step": 200 }, { "epoch": 0.01, "learning_rate": 6.1499999999999996e-06, "loss": 1.0484, "step": 205 }, { "epoch": 0.01, "learning_rate": 6.3e-06, "loss": 0.8858, "step": 210 }, { "epoch": 0.01, "learning_rate": 6.45e-06, "loss": 0.9786, "step": 215 }, { "epoch": 0.01, "learning_rate": 6.6e-06, "loss": 1.153, "step": 220 }, { "epoch": 0.01, "learning_rate": 6.750000000000001e-06, "loss": 1.0792, "step": 225 }, { "epoch": 0.01, "learning_rate": 6.900000000000001e-06, "loss": 1.2376, "step": 230 }, { "epoch": 0.01, "learning_rate": 7.049999999999999e-06, "loss": 1.2282, "step": 235 }, { "epoch": 0.01, "learning_rate": 7.2e-06, "loss": 1.211, "step": 240 }, { "epoch": 0.01, "learning_rate": 7.35e-06, "loss": 1.3426, "step": 245 }, { "epoch": 0.01, "learning_rate": 7.5e-06, "loss": 1.6511, "step": 250 }, { "epoch": 0.01, "learning_rate": 7.65e-06, "loss": 1.1288, "step": 255 }, { "epoch": 0.01, "learning_rate": 7.8e-06, "loss": 0.8139, "step": 260 }, { "epoch": 0.01, "learning_rate": 7.95e-06, "loss": 0.985, "step": 265 }, { "epoch": 0.01, "learning_rate": 8.1e-06, "loss": 0.8362, "step": 270 }, { "epoch": 0.01, "learning_rate": 8.25e-06, "loss": 0.9912, "step": 275 }, { "epoch": 0.01, "learning_rate": 8.400000000000001e-06, "loss": 1.2811, "step": 280 }, { "epoch": 0.01, "learning_rate": 8.55e-06, "loss": 1.1056, "step": 285 }, { "epoch": 0.01, "learning_rate": 8.7e-06, "loss": 1.1996, "step": 290 }, { "epoch": 0.01, "learning_rate": 8.85e-06, "loss": 1.4412, "step": 295 }, { "epoch": 0.01, "learning_rate": 9e-06, "loss": 1.9595, "step": 300 }, { "epoch": 0.01, "learning_rate": 9.15e-06, "loss": 0.9321, "step": 305 }, { "epoch": 0.01, "learning_rate": 9.3e-06, "loss": 0.682, "step": 310 }, { "epoch": 0.01, "learning_rate": 9.450000000000001e-06, "loss": 0.7705, "step": 315 }, { "epoch": 0.01, "learning_rate": 9.600000000000001e-06, "loss": 0.8194, "step": 320 }, { "epoch": 0.02, "learning_rate": 9.75e-06, "loss": 0.9736, "step": 325 }, { "epoch": 0.02, "learning_rate": 9.9e-06, "loss": 1.0385, "step": 330 }, { "epoch": 0.02, "learning_rate": 1.005e-05, "loss": 1.0263, "step": 335 }, { "epoch": 0.02, "learning_rate": 1.02e-05, "loss": 1.1658, "step": 340 }, { "epoch": 0.02, "learning_rate": 1.035e-05, "loss": 1.4046, "step": 345 }, { "epoch": 0.02, "learning_rate": 1.05e-05, "loss": 1.7697, "step": 350 }, { "epoch": 0.02, "learning_rate": 1.065e-05, "loss": 0.8677, "step": 355 }, { "epoch": 0.02, "learning_rate": 1.08e-05, "loss": 0.7698, "step": 360 }, { "epoch": 0.02, "learning_rate": 1.095e-05, "loss": 0.7114, "step": 365 }, { "epoch": 0.02, "learning_rate": 1.11e-05, "loss": 0.8821, "step": 370 }, { "epoch": 0.02, "learning_rate": 1.125e-05, "loss": 0.8893, "step": 375 }, { "epoch": 0.02, "learning_rate": 1.1400000000000001e-05, "loss": 0.8909, "step": 380 }, { "epoch": 0.02, "learning_rate": 1.1550000000000001e-05, "loss": 1.0615, "step": 385 }, { "epoch": 0.02, "learning_rate": 1.1700000000000001e-05, "loss": 1.026, "step": 390 }, { "epoch": 0.02, "learning_rate": 1.185e-05, "loss": 1.1341, "step": 395 }, { "epoch": 0.02, "learning_rate": 1.2e-05, "loss": 1.7984, "step": 400 }, { "epoch": 0.02, "learning_rate": 1.215e-05, "loss": 0.946, "step": 405 }, { "epoch": 0.02, "learning_rate": 1.2299999999999999e-05, "loss": 0.5768, "step": 410 }, { "epoch": 0.02, "learning_rate": 1.245e-05, "loss": 0.6476, "step": 415 }, { "epoch": 0.02, "learning_rate": 1.26e-05, "loss": 0.7286, "step": 420 }, { "epoch": 0.02, "learning_rate": 1.275e-05, "loss": 0.7953, "step": 425 }, { "epoch": 0.02, "learning_rate": 1.29e-05, "loss": 1.0026, "step": 430 }, { "epoch": 0.02, "learning_rate": 1.305e-05, "loss": 0.8454, "step": 435 }, { "epoch": 0.02, "learning_rate": 1.32e-05, "loss": 1.0045, "step": 440 }, { "epoch": 0.02, "learning_rate": 1.3350000000000001e-05, "loss": 1.0515, "step": 445 }, { "epoch": 0.02, "learning_rate": 1.3500000000000001e-05, "loss": 1.4769, "step": 450 }, { "epoch": 0.02, "learning_rate": 1.3650000000000001e-05, "loss": 0.8398, "step": 455 }, { "epoch": 0.02, "learning_rate": 1.3800000000000002e-05, "loss": 0.6949, "step": 460 }, { "epoch": 0.02, "learning_rate": 1.395e-05, "loss": 0.6305, "step": 465 }, { "epoch": 0.02, "learning_rate": 1.4099999999999999e-05, "loss": 0.6594, "step": 470 }, { "epoch": 0.02, "learning_rate": 1.4249999999999999e-05, "loss": 0.7755, "step": 475 }, { "epoch": 0.02, "learning_rate": 1.44e-05, "loss": 0.7899, "step": 480 }, { "epoch": 0.02, "learning_rate": 1.455e-05, "loss": 0.9505, "step": 485 }, { "epoch": 0.02, "learning_rate": 1.47e-05, "loss": 0.9007, "step": 490 }, { "epoch": 0.02, "learning_rate": 1.485e-05, "loss": 0.9164, "step": 495 }, { "epoch": 0.02, "learning_rate": 1.5e-05, "loss": 1.4495, "step": 500 }, { "epoch": 0.02, "learning_rate": 1.515e-05, "loss": 0.8424, "step": 505 }, { "epoch": 0.02, "learning_rate": 1.53e-05, "loss": 0.518, "step": 510 }, { "epoch": 0.02, "learning_rate": 1.545e-05, "loss": 0.5492, "step": 515 }, { "epoch": 0.02, "learning_rate": 1.56e-05, "loss": 0.7048, "step": 520 }, { "epoch": 0.02, "learning_rate": 1.575e-05, "loss": 0.7184, "step": 525 }, { "epoch": 0.02, "learning_rate": 1.59e-05, "loss": 0.8037, "step": 530 }, { "epoch": 0.02, "learning_rate": 1.605e-05, "loss": 0.8041, "step": 535 }, { "epoch": 0.03, "learning_rate": 1.62e-05, "loss": 0.9346, "step": 540 }, { "epoch": 0.03, "learning_rate": 1.635e-05, "loss": 1.0695, "step": 545 }, { "epoch": 0.03, "learning_rate": 1.65e-05, "loss": 1.0438, "step": 550 }, { "epoch": 0.03, "learning_rate": 1.665e-05, "loss": 0.7821, "step": 555 }, { "epoch": 0.03, "learning_rate": 1.6800000000000002e-05, "loss": 0.6464, "step": 560 }, { "epoch": 0.03, "learning_rate": 1.695e-05, "loss": 0.5224, "step": 565 }, { "epoch": 0.03, "learning_rate": 1.71e-05, "loss": 0.5888, "step": 570 }, { "epoch": 0.03, "learning_rate": 1.725e-05, "loss": 0.6815, "step": 575 }, { "epoch": 0.03, "learning_rate": 1.74e-05, "loss": 0.7166, "step": 580 }, { "epoch": 0.03, "learning_rate": 1.755e-05, "loss": 0.7879, "step": 585 }, { "epoch": 0.03, "learning_rate": 1.77e-05, "loss": 0.9349, "step": 590 }, { "epoch": 0.03, "learning_rate": 1.785e-05, "loss": 1.1031, "step": 595 }, { "epoch": 0.03, "learning_rate": 1.8e-05, "loss": 1.5274, "step": 600 }, { "epoch": 0.03, "learning_rate": 1.815e-05, "loss": 0.8439, "step": 605 }, { "epoch": 0.03, "learning_rate": 1.83e-05, "loss": 0.51, "step": 610 }, { "epoch": 0.03, "learning_rate": 1.845e-05, "loss": 0.6592, "step": 615 }, { "epoch": 0.03, "learning_rate": 1.86e-05, "loss": 0.5804, "step": 620 }, { "epoch": 0.03, "learning_rate": 1.8750000000000002e-05, "loss": 0.648, "step": 625 }, { "epoch": 0.03, "learning_rate": 1.8900000000000002e-05, "loss": 0.7396, "step": 630 }, { "epoch": 0.03, "learning_rate": 1.9050000000000002e-05, "loss": 0.9051, "step": 635 }, { "epoch": 0.03, "learning_rate": 1.9200000000000003e-05, "loss": 0.9073, "step": 640 }, { "epoch": 0.03, "learning_rate": 1.935e-05, "loss": 1.0043, "step": 645 }, { "epoch": 0.03, "learning_rate": 1.95e-05, "loss": 1.1916, "step": 650 }, { "epoch": 0.03, "learning_rate": 1.965e-05, "loss": 0.6345, "step": 655 }, { "epoch": 0.03, "learning_rate": 1.98e-05, "loss": 0.5134, "step": 660 }, { "epoch": 0.03, "learning_rate": 1.995e-05, "loss": 0.4761, "step": 665 }, { "epoch": 0.03, "learning_rate": 2.01e-05, "loss": 0.6114, "step": 670 }, { "epoch": 0.03, "learning_rate": 2.025e-05, "loss": 0.5128, "step": 675 }, { "epoch": 0.03, "learning_rate": 2.04e-05, "loss": 0.7642, "step": 680 }, { "epoch": 0.03, "learning_rate": 2.055e-05, "loss": 0.7798, "step": 685 }, { "epoch": 0.03, "learning_rate": 2.07e-05, "loss": 0.8385, "step": 690 }, { "epoch": 0.03, "learning_rate": 2.085e-05, "loss": 0.8393, "step": 695 }, { "epoch": 0.03, "learning_rate": 2.1e-05, "loss": 1.2992, "step": 700 }, { "epoch": 0.03, "learning_rate": 2.115e-05, "loss": 0.7748, "step": 705 }, { "epoch": 0.03, "learning_rate": 2.13e-05, "loss": 0.4805, "step": 710 }, { "epoch": 0.03, "learning_rate": 2.145e-05, "loss": 0.4676, "step": 715 }, { "epoch": 0.03, "learning_rate": 2.16e-05, "loss": 0.4895, "step": 720 }, { "epoch": 0.03, "learning_rate": 2.175e-05, "loss": 0.6178, "step": 725 }, { "epoch": 0.03, "learning_rate": 2.19e-05, "loss": 0.6627, "step": 730 }, { "epoch": 0.03, "learning_rate": 2.205e-05, "loss": 0.7188, "step": 735 }, { "epoch": 0.03, "learning_rate": 2.22e-05, "loss": 0.7827, "step": 740 }, { "epoch": 0.03, "learning_rate": 2.235e-05, "loss": 0.9153, "step": 745 }, { "epoch": 0.03, "learning_rate": 2.25e-05, "loss": 1.4324, "step": 750 }, { "epoch": 0.04, "learning_rate": 2.265e-05, "loss": 0.5785, "step": 755 }, { "epoch": 0.04, "learning_rate": 2.2800000000000002e-05, "loss": 0.4344, "step": 760 }, { "epoch": 0.04, "learning_rate": 2.2950000000000002e-05, "loss": 0.4868, "step": 765 }, { "epoch": 0.04, "learning_rate": 2.3100000000000002e-05, "loss": 0.4944, "step": 770 }, { "epoch": 0.04, "learning_rate": 2.3250000000000003e-05, "loss": 0.5598, "step": 775 }, { "epoch": 0.04, "learning_rate": 2.3400000000000003e-05, "loss": 0.5761, "step": 780 }, { "epoch": 0.04, "learning_rate": 2.3550000000000003e-05, "loss": 0.6754, "step": 785 }, { "epoch": 0.04, "learning_rate": 2.37e-05, "loss": 0.9604, "step": 790 }, { "epoch": 0.04, "learning_rate": 2.385e-05, "loss": 1.0913, "step": 795 }, { "epoch": 0.04, "learning_rate": 2.4e-05, "loss": 1.3051, "step": 800 }, { "epoch": 0.04, "learning_rate": 2.415e-05, "loss": 0.6032, "step": 805 }, { "epoch": 0.04, "learning_rate": 2.43e-05, "loss": 0.403, "step": 810 }, { "epoch": 0.04, "learning_rate": 2.4449999999999998e-05, "loss": 0.4457, "step": 815 }, { "epoch": 0.04, "learning_rate": 2.4599999999999998e-05, "loss": 0.4929, "step": 820 }, { "epoch": 0.04, "learning_rate": 2.475e-05, "loss": 0.6268, "step": 825 }, { "epoch": 0.04, "learning_rate": 2.49e-05, "loss": 0.5645, "step": 830 }, { "epoch": 0.04, "learning_rate": 2.505e-05, "loss": 0.6993, "step": 835 }, { "epoch": 0.04, "learning_rate": 2.52e-05, "loss": 0.7833, "step": 840 }, { "epoch": 0.04, "learning_rate": 2.535e-05, "loss": 0.816, "step": 845 }, { "epoch": 0.04, "learning_rate": 2.55e-05, "loss": 1.0046, "step": 850 }, { "epoch": 0.04, "learning_rate": 2.565e-05, "loss": 0.7003, "step": 855 }, { "epoch": 0.04, "learning_rate": 2.58e-05, "loss": 0.3647, "step": 860 }, { "epoch": 0.04, "learning_rate": 2.595e-05, "loss": 0.441, "step": 865 }, { "epoch": 0.04, "learning_rate": 2.61e-05, "loss": 0.5558, "step": 870 }, { "epoch": 0.04, "learning_rate": 2.625e-05, "loss": 0.526, "step": 875 }, { "epoch": 0.04, "learning_rate": 2.64e-05, "loss": 0.7098, "step": 880 }, { "epoch": 0.04, "learning_rate": 2.655e-05, "loss": 0.5895, "step": 885 }, { "epoch": 0.04, "learning_rate": 2.6700000000000002e-05, "loss": 0.7329, "step": 890 }, { "epoch": 0.04, "learning_rate": 2.6850000000000002e-05, "loss": 1.0144, "step": 895 }, { "epoch": 0.04, "learning_rate": 2.7000000000000002e-05, "loss": 1.1521, "step": 900 }, { "epoch": 0.04, "learning_rate": 2.7150000000000003e-05, "loss": 0.6231, "step": 905 }, { "epoch": 0.04, "learning_rate": 2.7300000000000003e-05, "loss": 0.361, "step": 910 }, { "epoch": 0.04, "learning_rate": 2.7450000000000003e-05, "loss": 0.5285, "step": 915 }, { "epoch": 0.04, "learning_rate": 2.7600000000000003e-05, "loss": 0.563, "step": 920 }, { "epoch": 0.04, "learning_rate": 2.7750000000000004e-05, "loss": 0.5804, "step": 925 }, { "epoch": 0.04, "learning_rate": 2.79e-05, "loss": 0.5276, "step": 930 }, { "epoch": 0.04, "learning_rate": 2.805e-05, "loss": 0.5842, "step": 935 }, { "epoch": 0.04, "learning_rate": 2.8199999999999998e-05, "loss": 0.7148, "step": 940 }, { "epoch": 0.04, "learning_rate": 2.8349999999999998e-05, "loss": 0.9909, "step": 945 }, { "epoch": 0.04, "learning_rate": 2.8499999999999998e-05, "loss": 1.0674, "step": 950 }, { "epoch": 0.04, "learning_rate": 2.865e-05, "loss": 0.6094, "step": 955 }, { "epoch": 0.04, "learning_rate": 2.88e-05, "loss": 0.3548, "step": 960 }, { "epoch": 0.05, "learning_rate": 2.895e-05, "loss": 0.4452, "step": 965 }, { "epoch": 0.05, "learning_rate": 2.91e-05, "loss": 0.5272, "step": 970 }, { "epoch": 0.05, "learning_rate": 2.925e-05, "loss": 0.4915, "step": 975 }, { "epoch": 0.05, "learning_rate": 2.94e-05, "loss": 0.5938, "step": 980 }, { "epoch": 0.05, "learning_rate": 2.955e-05, "loss": 0.6842, "step": 985 }, { "epoch": 0.05, "learning_rate": 2.97e-05, "loss": 0.7153, "step": 990 }, { "epoch": 0.05, "learning_rate": 2.985e-05, "loss": 0.8872, "step": 995 }, { "epoch": 0.05, "learning_rate": 3e-05, "loss": 1.0897, "step": 1000 }, { "epoch": 0.05, "learning_rate": 2.999953193016398e-05, "loss": 0.6239, "step": 1005 }, { "epoch": 0.05, "learning_rate": 2.9999063860327964e-05, "loss": 0.3869, "step": 1010 }, { "epoch": 0.05, "learning_rate": 2.9998595790491944e-05, "loss": 0.4034, "step": 1015 }, { "epoch": 0.05, "learning_rate": 2.9998127720655923e-05, "loss": 0.5883, "step": 1020 }, { "epoch": 0.05, "learning_rate": 2.9997659650819903e-05, "loss": 0.5247, "step": 1025 }, { "epoch": 0.05, "learning_rate": 2.9997191580983886e-05, "loss": 0.5544, "step": 1030 }, { "epoch": 0.05, "learning_rate": 2.9996723511147863e-05, "loss": 0.666, "step": 1035 }, { "epoch": 0.05, "learning_rate": 2.9996255441311843e-05, "loss": 0.6601, "step": 1040 }, { "epoch": 0.05, "learning_rate": 2.9995787371475823e-05, "loss": 0.7471, "step": 1045 }, { "epoch": 0.05, "learning_rate": 2.9995319301639806e-05, "loss": 1.0622, "step": 1050 }, { "epoch": 0.05, "learning_rate": 2.9994851231803786e-05, "loss": 0.5234, "step": 1055 }, { "epoch": 0.05, "learning_rate": 2.9994383161967765e-05, "loss": 0.3457, "step": 1060 }, { "epoch": 0.05, "learning_rate": 2.999391509213175e-05, "loss": 0.4613, "step": 1065 }, { "epoch": 0.05, "learning_rate": 2.999344702229573e-05, "loss": 0.5267, "step": 1070 }, { "epoch": 0.05, "learning_rate": 2.9992978952459708e-05, "loss": 0.4031, "step": 1075 }, { "epoch": 0.05, "learning_rate": 2.9992510882623688e-05, "loss": 0.657, "step": 1080 }, { "epoch": 0.05, "learning_rate": 2.999204281278767e-05, "loss": 0.5817, "step": 1085 }, { "epoch": 0.05, "learning_rate": 2.999157474295165e-05, "loss": 0.5876, "step": 1090 }, { "epoch": 0.05, "learning_rate": 2.9991106673115627e-05, "loss": 0.8619, "step": 1095 }, { "epoch": 0.05, "learning_rate": 2.9990638603279607e-05, "loss": 0.9578, "step": 1100 }, { "epoch": 0.05, "learning_rate": 2.999017053344359e-05, "loss": 0.6637, "step": 1105 }, { "epoch": 0.05, "learning_rate": 2.998970246360757e-05, "loss": 0.4149, "step": 1110 }, { "epoch": 0.05, "learning_rate": 2.998923439377155e-05, "loss": 0.2971, "step": 1115 }, { "epoch": 0.05, "learning_rate": 2.998876632393553e-05, "loss": 0.3353, "step": 1120 }, { "epoch": 0.05, "learning_rate": 2.9988298254099513e-05, "loss": 0.4795, "step": 1125 }, { "epoch": 0.05, "learning_rate": 2.9987830184263493e-05, "loss": 0.5698, "step": 1130 }, { "epoch": 0.05, "learning_rate": 2.9987362114427473e-05, "loss": 0.5586, "step": 1135 }, { "epoch": 0.05, "learning_rate": 2.9986894044591456e-05, "loss": 0.7123, "step": 1140 }, { "epoch": 0.05, "learning_rate": 2.9986425974755436e-05, "loss": 0.7498, "step": 1145 }, { "epoch": 0.05, "learning_rate": 2.9985957904919416e-05, "loss": 1.1885, "step": 1150 }, { "epoch": 0.05, "learning_rate": 2.9985489835083395e-05, "loss": 0.5377, "step": 1155 }, { "epoch": 0.05, "learning_rate": 2.9985021765247375e-05, "loss": 0.3324, "step": 1160 }, { "epoch": 0.05, "learning_rate": 2.9984553695411355e-05, "loss": 0.4179, "step": 1165 }, { "epoch": 0.05, "learning_rate": 2.9984085625575335e-05, "loss": 0.4199, "step": 1170 }, { "epoch": 0.05, "learning_rate": 2.9983617555739315e-05, "loss": 0.5361, "step": 1175 }, { "epoch": 0.06, "learning_rate": 2.9983149485903298e-05, "loss": 0.5003, "step": 1180 }, { "epoch": 0.06, "learning_rate": 2.9982681416067278e-05, "loss": 0.7324, "step": 1185 }, { "epoch": 0.06, "learning_rate": 2.9982213346231258e-05, "loss": 0.7964, "step": 1190 }, { "epoch": 0.06, "learning_rate": 2.998174527639524e-05, "loss": 0.7179, "step": 1195 }, { "epoch": 0.06, "learning_rate": 2.998127720655922e-05, "loss": 0.9888, "step": 1200 }, { "epoch": 0.06, "learning_rate": 2.99808091367232e-05, "loss": 0.545, "step": 1205 }, { "epoch": 0.06, "learning_rate": 2.998034106688718e-05, "loss": 0.265, "step": 1210 }, { "epoch": 0.06, "learning_rate": 2.9979872997051163e-05, "loss": 0.4571, "step": 1215 }, { "epoch": 0.06, "learning_rate": 2.9979404927215143e-05, "loss": 0.4603, "step": 1220 }, { "epoch": 0.06, "learning_rate": 2.997893685737912e-05, "loss": 0.4434, "step": 1225 }, { "epoch": 0.06, "learning_rate": 2.99784687875431e-05, "loss": 0.5181, "step": 1230 }, { "epoch": 0.06, "learning_rate": 2.9978000717707083e-05, "loss": 0.6107, "step": 1235 }, { "epoch": 0.06, "learning_rate": 2.9977532647871063e-05, "loss": 0.6827, "step": 1240 }, { "epoch": 0.06, "learning_rate": 2.9977064578035042e-05, "loss": 0.8626, "step": 1245 }, { "epoch": 0.06, "learning_rate": 2.9976596508199026e-05, "loss": 1.0847, "step": 1250 }, { "epoch": 0.06, "learning_rate": 2.9976128438363005e-05, "loss": 0.5537, "step": 1255 }, { "epoch": 0.06, "learning_rate": 2.9975660368526985e-05, "loss": 0.3262, "step": 1260 }, { "epoch": 0.06, "learning_rate": 2.9975192298690965e-05, "loss": 0.4132, "step": 1265 }, { "epoch": 0.06, "learning_rate": 2.9974724228854948e-05, "loss": 0.4272, "step": 1270 }, { "epoch": 0.06, "learning_rate": 2.9974256159018928e-05, "loss": 0.3927, "step": 1275 }, { "epoch": 0.06, "learning_rate": 2.9973788089182908e-05, "loss": 0.4433, "step": 1280 }, { "epoch": 0.06, "learning_rate": 2.9973320019346884e-05, "loss": 0.5682, "step": 1285 }, { "epoch": 0.06, "learning_rate": 2.9972851949510867e-05, "loss": 0.6579, "step": 1290 }, { "epoch": 0.06, "learning_rate": 2.9972383879674847e-05, "loss": 0.9216, "step": 1295 }, { "epoch": 0.06, "learning_rate": 2.9971915809838827e-05, "loss": 1.0738, "step": 1300 }, { "epoch": 0.06, "learning_rate": 2.9971447740002807e-05, "loss": 0.5654, "step": 1305 }, { "epoch": 0.06, "learning_rate": 2.997097967016679e-05, "loss": 0.3394, "step": 1310 }, { "epoch": 0.06, "learning_rate": 2.997051160033077e-05, "loss": 0.3269, "step": 1315 }, { "epoch": 0.06, "learning_rate": 2.997004353049475e-05, "loss": 0.3893, "step": 1320 }, { "epoch": 0.06, "learning_rate": 2.9969575460658733e-05, "loss": 0.4748, "step": 1325 }, { "epoch": 0.06, "learning_rate": 2.9969107390822713e-05, "loss": 0.4428, "step": 1330 }, { "epoch": 0.06, "learning_rate": 2.9968639320986693e-05, "loss": 0.5365, "step": 1335 }, { "epoch": 0.06, "learning_rate": 2.9968171251150672e-05, "loss": 0.6774, "step": 1340 }, { "epoch": 0.06, "learning_rate": 2.9967703181314656e-05, "loss": 0.6508, "step": 1345 }, { "epoch": 0.06, "learning_rate": 2.9967235111478632e-05, "loss": 1.013, "step": 1350 }, { "epoch": 0.06, "learning_rate": 2.9966767041642612e-05, "loss": 0.5972, "step": 1355 }, { "epoch": 0.06, "learning_rate": 2.996629897180659e-05, "loss": 0.3359, "step": 1360 }, { "epoch": 0.06, "learning_rate": 2.9965830901970575e-05, "loss": 0.3574, "step": 1365 }, { "epoch": 0.06, "learning_rate": 2.9965362832134555e-05, "loss": 0.3918, "step": 1370 }, { "epoch": 0.06, "learning_rate": 2.9964894762298535e-05, "loss": 0.4044, "step": 1375 }, { "epoch": 0.06, "learning_rate": 2.9964426692462518e-05, "loss": 0.523, "step": 1380 }, { "epoch": 0.06, "learning_rate": 2.9963958622626498e-05, "loss": 0.5226, "step": 1385 }, { "epoch": 0.06, "learning_rate": 2.9963490552790477e-05, "loss": 0.5051, "step": 1390 }, { "epoch": 0.07, "learning_rate": 2.9963022482954457e-05, "loss": 0.8244, "step": 1395 }, { "epoch": 0.07, "learning_rate": 2.996255441311844e-05, "loss": 1.2422, "step": 1400 }, { "epoch": 0.07, "learning_rate": 2.996208634328242e-05, "loss": 0.5015, "step": 1405 }, { "epoch": 0.07, "learning_rate": 2.99616182734464e-05, "loss": 0.2722, "step": 1410 }, { "epoch": 0.07, "learning_rate": 2.9961150203610376e-05, "loss": 0.3147, "step": 1415 }, { "epoch": 0.07, "learning_rate": 2.996068213377436e-05, "loss": 0.389, "step": 1420 }, { "epoch": 0.07, "learning_rate": 2.996021406393834e-05, "loss": 0.3931, "step": 1425 }, { "epoch": 0.07, "learning_rate": 2.995974599410232e-05, "loss": 0.5088, "step": 1430 }, { "epoch": 0.07, "learning_rate": 2.9959277924266302e-05, "loss": 0.6008, "step": 1435 }, { "epoch": 0.07, "learning_rate": 2.9958809854430282e-05, "loss": 0.6104, "step": 1440 }, { "epoch": 0.07, "learning_rate": 2.9958341784594262e-05, "loss": 0.6142, "step": 1445 }, { "epoch": 0.07, "learning_rate": 2.9957873714758242e-05, "loss": 1.0904, "step": 1450 }, { "epoch": 0.07, "learning_rate": 2.9957405644922225e-05, "loss": 0.5021, "step": 1455 }, { "epoch": 0.07, "learning_rate": 2.9956937575086205e-05, "loss": 0.4151, "step": 1460 }, { "epoch": 0.07, "learning_rate": 2.9956469505250185e-05, "loss": 0.3302, "step": 1465 }, { "epoch": 0.07, "learning_rate": 2.9956001435414165e-05, "loss": 0.3917, "step": 1470 }, { "epoch": 0.07, "learning_rate": 2.9955533365578144e-05, "loss": 0.3972, "step": 1475 }, { "epoch": 0.07, "learning_rate": 2.9955065295742124e-05, "loss": 0.428, "step": 1480 }, { "epoch": 0.07, "learning_rate": 2.9954597225906104e-05, "loss": 0.5828, "step": 1485 }, { "epoch": 0.07, "learning_rate": 2.9954129156070087e-05, "loss": 0.6936, "step": 1490 }, { "epoch": 0.07, "learning_rate": 2.9953661086234067e-05, "loss": 0.7506, "step": 1495 }, { "epoch": 0.07, "learning_rate": 2.9953193016398047e-05, "loss": 1.1826, "step": 1500 }, { "epoch": 0.07, "learning_rate": 2.9952724946562027e-05, "loss": 0.4892, "step": 1505 }, { "epoch": 0.07, "learning_rate": 2.995225687672601e-05, "loss": 0.353, "step": 1510 }, { "epoch": 0.07, "learning_rate": 2.995178880688999e-05, "loss": 0.381, "step": 1515 }, { "epoch": 0.07, "learning_rate": 2.995132073705397e-05, "loss": 0.2913, "step": 1520 }, { "epoch": 0.07, "learning_rate": 2.995085266721795e-05, "loss": 0.3876, "step": 1525 }, { "epoch": 0.07, "learning_rate": 2.9950384597381933e-05, "loss": 0.4861, "step": 1530 }, { "epoch": 0.07, "learning_rate": 2.9949916527545912e-05, "loss": 0.5714, "step": 1535 }, { "epoch": 0.07, "learning_rate": 2.994944845770989e-05, "loss": 0.3839, "step": 1540 }, { "epoch": 0.07, "learning_rate": 2.994898038787387e-05, "loss": 0.688, "step": 1545 }, { "epoch": 0.07, "learning_rate": 2.9948512318037852e-05, "loss": 0.9213, "step": 1550 }, { "epoch": 0.07, "learning_rate": 2.994804424820183e-05, "loss": 0.5367, "step": 1555 }, { "epoch": 0.07, "learning_rate": 2.994757617836581e-05, "loss": 0.2404, "step": 1560 }, { "epoch": 0.07, "learning_rate": 2.9947108108529795e-05, "loss": 0.3007, "step": 1565 }, { "epoch": 0.07, "learning_rate": 2.9946640038693775e-05, "loss": 0.3554, "step": 1570 }, { "epoch": 0.07, "learning_rate": 2.9946171968857754e-05, "loss": 0.3435, "step": 1575 }, { "epoch": 0.07, "learning_rate": 2.9945703899021734e-05, "loss": 0.4196, "step": 1580 }, { "epoch": 0.07, "learning_rate": 2.9945235829185717e-05, "loss": 0.5177, "step": 1585 }, { "epoch": 0.07, "learning_rate": 2.9944767759349697e-05, "loss": 0.6126, "step": 1590 }, { "epoch": 0.07, "learning_rate": 2.9944299689513677e-05, "loss": 0.6459, "step": 1595 }, { "epoch": 0.07, "learning_rate": 2.9943831619677657e-05, "loss": 1.0063, "step": 1600 }, { "epoch": 0.07, "learning_rate": 2.9943363549841637e-05, "loss": 0.4703, "step": 1605 }, { "epoch": 0.08, "learning_rate": 2.9942895480005616e-05, "loss": 0.2602, "step": 1610 }, { "epoch": 0.08, "learning_rate": 2.9942427410169596e-05, "loss": 0.3381, "step": 1615 }, { "epoch": 0.08, "learning_rate": 2.994195934033358e-05, "loss": 0.3428, "step": 1620 }, { "epoch": 0.08, "learning_rate": 2.994149127049756e-05, "loss": 0.5091, "step": 1625 }, { "epoch": 0.08, "learning_rate": 2.994102320066154e-05, "loss": 0.4356, "step": 1630 }, { "epoch": 0.08, "learning_rate": 2.994055513082552e-05, "loss": 0.4724, "step": 1635 }, { "epoch": 0.08, "learning_rate": 2.9940087060989502e-05, "loss": 0.6446, "step": 1640 }, { "epoch": 0.08, "learning_rate": 2.9939618991153482e-05, "loss": 0.8065, "step": 1645 }, { "epoch": 0.08, "learning_rate": 2.9939150921317462e-05, "loss": 1.2063, "step": 1650 }, { "epoch": 0.08, "learning_rate": 2.993868285148144e-05, "loss": 0.4322, "step": 1655 }, { "epoch": 0.08, "learning_rate": 2.9938214781645425e-05, "loss": 0.2961, "step": 1660 }, { "epoch": 0.08, "learning_rate": 2.99377467118094e-05, "loss": 0.3671, "step": 1665 }, { "epoch": 0.08, "learning_rate": 2.993727864197338e-05, "loss": 0.2797, "step": 1670 }, { "epoch": 0.08, "learning_rate": 2.9936810572137364e-05, "loss": 0.3971, "step": 1675 }, { "epoch": 0.08, "learning_rate": 2.9936342502301344e-05, "loss": 0.3879, "step": 1680 }, { "epoch": 0.08, "learning_rate": 2.9935874432465324e-05, "loss": 0.4365, "step": 1685 }, { "epoch": 0.08, "learning_rate": 2.9935406362629304e-05, "loss": 0.7451, "step": 1690 }, { "epoch": 0.08, "learning_rate": 2.9934938292793287e-05, "loss": 0.5118, "step": 1695 }, { "epoch": 0.08, "learning_rate": 2.9934470222957267e-05, "loss": 0.9061, "step": 1700 }, { "epoch": 0.08, "learning_rate": 2.9934002153121247e-05, "loss": 0.4253, "step": 1705 }, { "epoch": 0.08, "learning_rate": 2.9933534083285226e-05, "loss": 0.2794, "step": 1710 }, { "epoch": 0.08, "learning_rate": 2.993306601344921e-05, "loss": 0.2437, "step": 1715 }, { "epoch": 0.08, "learning_rate": 2.993259794361319e-05, "loss": 0.3895, "step": 1720 }, { "epoch": 0.08, "learning_rate": 2.993212987377717e-05, "loss": 0.4005, "step": 1725 }, { "epoch": 0.08, "learning_rate": 2.9931661803941146e-05, "loss": 0.3341, "step": 1730 }, { "epoch": 0.08, "learning_rate": 2.993119373410513e-05, "loss": 0.5051, "step": 1735 }, { "epoch": 0.08, "learning_rate": 2.993072566426911e-05, "loss": 0.6066, "step": 1740 }, { "epoch": 0.08, "learning_rate": 2.993025759443309e-05, "loss": 0.8869, "step": 1745 }, { "epoch": 0.08, "learning_rate": 2.992978952459707e-05, "loss": 1.0657, "step": 1750 }, { "epoch": 0.08, "learning_rate": 2.992932145476105e-05, "loss": 0.3844, "step": 1755 }, { "epoch": 0.08, "learning_rate": 2.992885338492503e-05, "loss": 0.3359, "step": 1760 }, { "epoch": 0.08, "learning_rate": 2.992838531508901e-05, "loss": 0.3245, "step": 1765 }, { "epoch": 0.08, "learning_rate": 2.9927917245252994e-05, "loss": 0.4634, "step": 1770 }, { "epoch": 0.08, "learning_rate": 2.9927449175416974e-05, "loss": 0.3902, "step": 1775 }, { "epoch": 0.08, "learning_rate": 2.9926981105580954e-05, "loss": 0.4418, "step": 1780 }, { "epoch": 0.08, "learning_rate": 2.9926513035744934e-05, "loss": 0.4923, "step": 1785 }, { "epoch": 0.08, "learning_rate": 2.9926044965908914e-05, "loss": 0.5632, "step": 1790 }, { "epoch": 0.08, "learning_rate": 2.9925576896072893e-05, "loss": 0.6368, "step": 1795 }, { "epoch": 0.08, "learning_rate": 2.9925108826236873e-05, "loss": 0.8357, "step": 1800 }, { "epoch": 0.08, "learning_rate": 2.9924640756400856e-05, "loss": 0.5503, "step": 1805 }, { "epoch": 0.08, "learning_rate": 2.9924172686564836e-05, "loss": 0.273, "step": 1810 }, { "epoch": 0.08, "learning_rate": 2.9923704616728816e-05, "loss": 0.2966, "step": 1815 }, { "epoch": 0.08, "learning_rate": 2.9923236546892796e-05, "loss": 0.3135, "step": 1820 }, { "epoch": 0.09, "learning_rate": 2.992276847705678e-05, "loss": 0.4249, "step": 1825 }, { "epoch": 0.09, "learning_rate": 2.992230040722076e-05, "loss": 0.4162, "step": 1830 }, { "epoch": 0.09, "learning_rate": 2.992183233738474e-05, "loss": 0.606, "step": 1835 }, { "epoch": 0.09, "learning_rate": 2.992136426754872e-05, "loss": 0.5282, "step": 1840 }, { "epoch": 0.09, "learning_rate": 2.9920896197712702e-05, "loss": 0.6356, "step": 1845 }, { "epoch": 0.09, "learning_rate": 2.992042812787668e-05, "loss": 0.8718, "step": 1850 }, { "epoch": 0.09, "learning_rate": 2.9919960058040658e-05, "loss": 0.6441, "step": 1855 }, { "epoch": 0.09, "learning_rate": 2.991949198820464e-05, "loss": 0.2587, "step": 1860 }, { "epoch": 0.09, "learning_rate": 2.991902391836862e-05, "loss": 0.394, "step": 1865 }, { "epoch": 0.09, "learning_rate": 2.99185558485326e-05, "loss": 0.4004, "step": 1870 }, { "epoch": 0.09, "learning_rate": 2.991808777869658e-05, "loss": 0.3494, "step": 1875 }, { "epoch": 0.09, "learning_rate": 2.9917619708860564e-05, "loss": 0.4785, "step": 1880 }, { "epoch": 0.09, "learning_rate": 2.9917151639024544e-05, "loss": 0.4616, "step": 1885 }, { "epoch": 0.09, "learning_rate": 2.9916683569188524e-05, "loss": 0.5584, "step": 1890 }, { "epoch": 0.09, "learning_rate": 2.9916215499352503e-05, "loss": 0.6289, "step": 1895 }, { "epoch": 0.09, "learning_rate": 2.9915747429516487e-05, "loss": 0.9955, "step": 1900 }, { "epoch": 0.09, "learning_rate": 2.9915279359680466e-05, "loss": 0.4722, "step": 1905 }, { "epoch": 0.09, "learning_rate": 2.9914811289844446e-05, "loss": 0.3275, "step": 1910 }, { "epoch": 0.09, "learning_rate": 2.9914343220008426e-05, "loss": 0.266, "step": 1915 }, { "epoch": 0.09, "learning_rate": 2.9913875150172406e-05, "loss": 0.3133, "step": 1920 }, { "epoch": 0.09, "learning_rate": 2.9913407080336386e-05, "loss": 0.4386, "step": 1925 }, { "epoch": 0.09, "learning_rate": 2.9912939010500365e-05, "loss": 0.5633, "step": 1930 }, { "epoch": 0.09, "learning_rate": 2.991247094066435e-05, "loss": 0.5064, "step": 1935 }, { "epoch": 0.09, "learning_rate": 2.991200287082833e-05, "loss": 0.6127, "step": 1940 }, { "epoch": 0.09, "learning_rate": 2.9911534800992308e-05, "loss": 0.5771, "step": 1945 }, { "epoch": 0.09, "learning_rate": 2.9911066731156288e-05, "loss": 0.8355, "step": 1950 }, { "epoch": 0.09, "learning_rate": 2.991059866132027e-05, "loss": 0.6079, "step": 1955 }, { "epoch": 0.09, "learning_rate": 2.991013059148425e-05, "loss": 0.2757, "step": 1960 }, { "epoch": 0.09, "learning_rate": 2.990966252164823e-05, "loss": 0.2852, "step": 1965 }, { "epoch": 0.09, "learning_rate": 2.990919445181221e-05, "loss": 0.3789, "step": 1970 }, { "epoch": 0.09, "learning_rate": 2.9908726381976194e-05, "loss": 0.3595, "step": 1975 }, { "epoch": 0.09, "learning_rate": 2.990825831214017e-05, "loss": 0.4526, "step": 1980 }, { "epoch": 0.09, "learning_rate": 2.990779024230415e-05, "loss": 0.469, "step": 1985 }, { "epoch": 0.09, "learning_rate": 2.9907322172468133e-05, "loss": 0.5819, "step": 1990 }, { "epoch": 0.09, "learning_rate": 2.9906854102632113e-05, "loss": 0.5658, "step": 1995 }, { "epoch": 0.09, "learning_rate": 2.9906386032796093e-05, "loss": 0.9492, "step": 2000 }, { "epoch": 0.09, "learning_rate": 2.9905917962960073e-05, "loss": 0.4452, "step": 2005 }, { "epoch": 0.09, "learning_rate": 2.9905449893124056e-05, "loss": 0.3381, "step": 2010 }, { "epoch": 0.09, "learning_rate": 2.9904981823288036e-05, "loss": 0.2989, "step": 2015 }, { "epoch": 0.09, "learning_rate": 2.9904513753452016e-05, "loss": 0.3576, "step": 2020 }, { "epoch": 0.09, "learning_rate": 2.9904045683615996e-05, "loss": 0.3346, "step": 2025 }, { "epoch": 0.09, "learning_rate": 2.990357761377998e-05, "loss": 0.4455, "step": 2030 }, { "epoch": 0.09, "learning_rate": 2.990310954394396e-05, "loss": 0.4583, "step": 2035 }, { "epoch": 0.1, "learning_rate": 2.990264147410794e-05, "loss": 0.5651, "step": 2040 }, { "epoch": 0.1, "learning_rate": 2.9902173404271918e-05, "loss": 0.7309, "step": 2045 }, { "epoch": 0.1, "learning_rate": 2.9901705334435898e-05, "loss": 1.0832, "step": 2050 }, { "epoch": 0.1, "learning_rate": 2.9901237264599878e-05, "loss": 0.4419, "step": 2055 }, { "epoch": 0.1, "learning_rate": 2.9900769194763858e-05, "loss": 0.2238, "step": 2060 }, { "epoch": 0.1, "learning_rate": 2.990030112492784e-05, "loss": 0.3316, "step": 2065 }, { "epoch": 0.1, "learning_rate": 2.989983305509182e-05, "loss": 0.3515, "step": 2070 }, { "epoch": 0.1, "learning_rate": 2.98993649852558e-05, "loss": 0.3687, "step": 2075 }, { "epoch": 0.1, "learning_rate": 2.989889691541978e-05, "loss": 0.3471, "step": 2080 }, { "epoch": 0.1, "learning_rate": 2.9898428845583763e-05, "loss": 0.4767, "step": 2085 }, { "epoch": 0.1, "learning_rate": 2.9897960775747743e-05, "loss": 0.4956, "step": 2090 }, { "epoch": 0.1, "learning_rate": 2.9897492705911723e-05, "loss": 0.5889, "step": 2095 }, { "epoch": 0.1, "learning_rate": 2.9897024636075706e-05, "loss": 0.896, "step": 2100 }, { "epoch": 0.1, "learning_rate": 2.9896556566239686e-05, "loss": 0.5322, "step": 2105 }, { "epoch": 0.1, "learning_rate": 2.9896088496403663e-05, "loss": 0.2182, "step": 2110 }, { "epoch": 0.1, "learning_rate": 2.9895620426567642e-05, "loss": 0.2965, "step": 2115 }, { "epoch": 0.1, "learning_rate": 2.9895152356731626e-05, "loss": 0.3261, "step": 2120 }, { "epoch": 0.1, "learning_rate": 2.9894684286895605e-05, "loss": 0.3001, "step": 2125 }, { "epoch": 0.1, "learning_rate": 2.9894216217059585e-05, "loss": 0.357, "step": 2130 }, { "epoch": 0.1, "learning_rate": 2.9893748147223565e-05, "loss": 0.4138, "step": 2135 }, { "epoch": 0.1, "learning_rate": 2.9893280077387548e-05, "loss": 0.4194, "step": 2140 }, { "epoch": 0.1, "learning_rate": 2.9892812007551528e-05, "loss": 0.6184, "step": 2145 }, { "epoch": 0.1, "learning_rate": 2.9892343937715508e-05, "loss": 0.9583, "step": 2150 }, { "epoch": 0.1, "learning_rate": 2.9891875867879488e-05, "loss": 0.6117, "step": 2155 }, { "epoch": 0.1, "learning_rate": 2.989140779804347e-05, "loss": 0.2569, "step": 2160 }, { "epoch": 0.1, "learning_rate": 2.989093972820745e-05, "loss": 0.3615, "step": 2165 }, { "epoch": 0.1, "learning_rate": 2.9890471658371427e-05, "loss": 0.3633, "step": 2170 }, { "epoch": 0.1, "learning_rate": 2.989000358853541e-05, "loss": 0.3625, "step": 2175 }, { "epoch": 0.1, "learning_rate": 2.988953551869939e-05, "loss": 0.4699, "step": 2180 }, { "epoch": 0.1, "learning_rate": 2.988906744886337e-05, "loss": 0.4034, "step": 2185 }, { "epoch": 0.1, "learning_rate": 2.988859937902735e-05, "loss": 0.5138, "step": 2190 }, { "epoch": 0.1, "learning_rate": 2.9888131309191333e-05, "loss": 0.5981, "step": 2195 }, { "epoch": 0.1, "learning_rate": 2.9887663239355313e-05, "loss": 0.8212, "step": 2200 }, { "epoch": 0.1, "learning_rate": 2.9887195169519293e-05, "loss": 0.4884, "step": 2205 }, { "epoch": 0.1, "learning_rate": 2.9886727099683272e-05, "loss": 0.2304, "step": 2210 }, { "epoch": 0.1, "learning_rate": 2.9886259029847256e-05, "loss": 0.2807, "step": 2215 }, { "epoch": 0.1, "learning_rate": 2.9885790960011236e-05, "loss": 0.2353, "step": 2220 }, { "epoch": 0.1, "learning_rate": 2.9885322890175215e-05, "loss": 0.3482, "step": 2225 }, { "epoch": 0.1, "learning_rate": 2.98848548203392e-05, "loss": 0.3445, "step": 2230 }, { "epoch": 0.1, "learning_rate": 2.9884386750503175e-05, "loss": 0.5496, "step": 2235 }, { "epoch": 0.1, "learning_rate": 2.9883918680667155e-05, "loss": 0.5218, "step": 2240 }, { "epoch": 0.1, "learning_rate": 2.9883450610831135e-05, "loss": 0.4963, "step": 2245 }, { "epoch": 0.1, "learning_rate": 2.9882982540995118e-05, "loss": 0.7415, "step": 2250 }, { "epoch": 0.11, "learning_rate": 2.9882514471159098e-05, "loss": 0.5019, "step": 2255 }, { "epoch": 0.11, "learning_rate": 2.9882046401323077e-05, "loss": 0.2427, "step": 2260 }, { "epoch": 0.11, "learning_rate": 2.9881578331487057e-05, "loss": 0.3198, "step": 2265 }, { "epoch": 0.11, "learning_rate": 2.988111026165104e-05, "loss": 0.3634, "step": 2270 }, { "epoch": 0.11, "learning_rate": 2.988064219181502e-05, "loss": 0.341, "step": 2275 }, { "epoch": 0.11, "learning_rate": 2.9880174121979e-05, "loss": 0.4483, "step": 2280 }, { "epoch": 0.11, "learning_rate": 2.9879706052142983e-05, "loss": 0.3871, "step": 2285 }, { "epoch": 0.11, "learning_rate": 2.9879237982306963e-05, "loss": 0.4825, "step": 2290 }, { "epoch": 0.11, "learning_rate": 2.9878769912470943e-05, "loss": 0.5595, "step": 2295 }, { "epoch": 0.11, "learning_rate": 2.987830184263492e-05, "loss": 0.8071, "step": 2300 }, { "epoch": 0.11, "learning_rate": 2.9877833772798903e-05, "loss": 0.384, "step": 2305 }, { "epoch": 0.11, "learning_rate": 2.9877365702962882e-05, "loss": 0.1561, "step": 2310 }, { "epoch": 0.11, "learning_rate": 2.9876897633126862e-05, "loss": 0.2556, "step": 2315 }, { "epoch": 0.11, "learning_rate": 2.9876429563290842e-05, "loss": 0.3062, "step": 2320 }, { "epoch": 0.11, "learning_rate": 2.9875961493454825e-05, "loss": 0.3907, "step": 2325 }, { "epoch": 0.11, "learning_rate": 2.9875493423618805e-05, "loss": 0.2969, "step": 2330 }, { "epoch": 0.11, "learning_rate": 2.9875025353782785e-05, "loss": 0.4659, "step": 2335 }, { "epoch": 0.11, "learning_rate": 2.9874557283946765e-05, "loss": 0.5984, "step": 2340 }, { "epoch": 0.11, "learning_rate": 2.9874089214110748e-05, "loss": 0.6873, "step": 2345 }, { "epoch": 0.11, "learning_rate": 2.9873621144274728e-05, "loss": 0.9153, "step": 2350 }, { "epoch": 0.11, "learning_rate": 2.9873153074438708e-05, "loss": 0.4355, "step": 2355 }, { "epoch": 0.11, "learning_rate": 2.9872685004602687e-05, "loss": 0.2925, "step": 2360 }, { "epoch": 0.11, "learning_rate": 2.9872216934766667e-05, "loss": 0.3126, "step": 2365 }, { "epoch": 0.11, "learning_rate": 2.9871748864930647e-05, "loss": 0.3362, "step": 2370 }, { "epoch": 0.11, "learning_rate": 2.9871280795094627e-05, "loss": 0.2827, "step": 2375 }, { "epoch": 0.11, "learning_rate": 2.987081272525861e-05, "loss": 0.4077, "step": 2380 }, { "epoch": 0.11, "learning_rate": 2.987034465542259e-05, "loss": 0.4149, "step": 2385 }, { "epoch": 0.11, "learning_rate": 2.986987658558657e-05, "loss": 0.3678, "step": 2390 }, { "epoch": 0.11, "learning_rate": 2.986940851575055e-05, "loss": 0.5219, "step": 2395 }, { "epoch": 0.11, "learning_rate": 2.9868940445914533e-05, "loss": 0.8143, "step": 2400 }, { "epoch": 0.11, "learning_rate": 2.9868472376078512e-05, "loss": 0.4312, "step": 2405 }, { "epoch": 0.11, "learning_rate": 2.9868004306242492e-05, "loss": 0.1985, "step": 2410 }, { "epoch": 0.11, "learning_rate": 2.9867536236406476e-05, "loss": 0.2725, "step": 2415 }, { "epoch": 0.11, "learning_rate": 2.9867068166570455e-05, "loss": 0.301, "step": 2420 }, { "epoch": 0.11, "learning_rate": 2.9866600096734432e-05, "loss": 0.3201, "step": 2425 }, { "epoch": 0.11, "learning_rate": 2.986613202689841e-05, "loss": 0.2769, "step": 2430 }, { "epoch": 0.11, "learning_rate": 2.9865663957062395e-05, "loss": 0.4863, "step": 2435 }, { "epoch": 0.11, "learning_rate": 2.9865195887226375e-05, "loss": 0.4339, "step": 2440 }, { "epoch": 0.11, "learning_rate": 2.9864727817390354e-05, "loss": 0.6175, "step": 2445 }, { "epoch": 0.11, "learning_rate": 2.9864259747554334e-05, "loss": 0.7348, "step": 2450 }, { "epoch": 0.11, "learning_rate": 2.9863791677718317e-05, "loss": 0.3938, "step": 2455 }, { "epoch": 0.11, "learning_rate": 2.9863323607882297e-05, "loss": 0.3108, "step": 2460 }, { "epoch": 0.12, "learning_rate": 2.9862855538046277e-05, "loss": 0.3003, "step": 2465 }, { "epoch": 0.12, "learning_rate": 2.986238746821026e-05, "loss": 0.4002, "step": 2470 }, { "epoch": 0.12, "learning_rate": 2.986191939837424e-05, "loss": 0.3234, "step": 2475 }, { "epoch": 0.12, "learning_rate": 2.986145132853822e-05, "loss": 0.4503, "step": 2480 }, { "epoch": 0.12, "learning_rate": 2.9860983258702196e-05, "loss": 0.5289, "step": 2485 }, { "epoch": 0.12, "learning_rate": 2.986051518886618e-05, "loss": 0.4845, "step": 2490 }, { "epoch": 0.12, "learning_rate": 2.986004711903016e-05, "loss": 0.6735, "step": 2495 }, { "epoch": 0.12, "learning_rate": 2.985957904919414e-05, "loss": 1.0605, "step": 2500 }, { "epoch": 0.12, "learning_rate": 2.985911097935812e-05, "loss": 0.4427, "step": 2505 }, { "epoch": 0.12, "learning_rate": 2.9858642909522102e-05, "loss": 0.1878, "step": 2510 }, { "epoch": 0.12, "learning_rate": 2.9858174839686082e-05, "loss": 0.2638, "step": 2515 }, { "epoch": 0.12, "learning_rate": 2.9857706769850062e-05, "loss": 0.32, "step": 2520 }, { "epoch": 0.12, "learning_rate": 2.985723870001404e-05, "loss": 0.4158, "step": 2525 }, { "epoch": 0.12, "learning_rate": 2.9856770630178025e-05, "loss": 0.4044, "step": 2530 }, { "epoch": 0.12, "learning_rate": 2.9856302560342005e-05, "loss": 0.5286, "step": 2535 }, { "epoch": 0.12, "learning_rate": 2.9855834490505985e-05, "loss": 0.5233, "step": 2540 }, { "epoch": 0.12, "learning_rate": 2.9855366420669968e-05, "loss": 0.4839, "step": 2545 }, { "epoch": 0.12, "learning_rate": 2.9854898350833944e-05, "loss": 1.015, "step": 2550 }, { "epoch": 0.12, "learning_rate": 2.9854430280997924e-05, "loss": 0.4324, "step": 2555 }, { "epoch": 0.12, "learning_rate": 2.9853962211161904e-05, "loss": 0.1776, "step": 2560 }, { "epoch": 0.12, "learning_rate": 2.9853494141325887e-05, "loss": 0.2091, "step": 2565 }, { "epoch": 0.12, "learning_rate": 2.9853026071489867e-05, "loss": 0.3604, "step": 2570 }, { "epoch": 0.12, "learning_rate": 2.9852558001653847e-05, "loss": 0.3578, "step": 2575 }, { "epoch": 0.12, "learning_rate": 2.9852089931817826e-05, "loss": 0.3927, "step": 2580 }, { "epoch": 0.12, "learning_rate": 2.985162186198181e-05, "loss": 0.4487, "step": 2585 }, { "epoch": 0.12, "learning_rate": 2.985115379214579e-05, "loss": 0.4263, "step": 2590 }, { "epoch": 0.12, "learning_rate": 2.985068572230977e-05, "loss": 0.653, "step": 2595 }, { "epoch": 0.12, "learning_rate": 2.9850217652473752e-05, "loss": 0.8854, "step": 2600 }, { "epoch": 0.12, "learning_rate": 2.9849749582637732e-05, "loss": 0.464, "step": 2605 }, { "epoch": 0.12, "learning_rate": 2.9849281512801712e-05, "loss": 0.2695, "step": 2610 }, { "epoch": 0.12, "learning_rate": 2.984881344296569e-05, "loss": 0.2283, "step": 2615 }, { "epoch": 0.12, "learning_rate": 2.9848345373129672e-05, "loss": 0.3573, "step": 2620 }, { "epoch": 0.12, "learning_rate": 2.984787730329365e-05, "loss": 0.2801, "step": 2625 }, { "epoch": 0.12, "learning_rate": 2.984740923345763e-05, "loss": 0.4285, "step": 2630 }, { "epoch": 0.12, "learning_rate": 2.984694116362161e-05, "loss": 0.4369, "step": 2635 }, { "epoch": 0.12, "learning_rate": 2.9846473093785594e-05, "loss": 0.4876, "step": 2640 }, { "epoch": 0.12, "learning_rate": 2.9846005023949574e-05, "loss": 0.6387, "step": 2645 }, { "epoch": 0.12, "learning_rate": 2.9845536954113554e-05, "loss": 0.9479, "step": 2650 }, { "epoch": 0.12, "learning_rate": 2.9845068884277537e-05, "loss": 0.4644, "step": 2655 }, { "epoch": 0.12, "learning_rate": 2.9844600814441517e-05, "loss": 0.1859, "step": 2660 }, { "epoch": 0.12, "learning_rate": 2.9844132744605497e-05, "loss": 0.2329, "step": 2665 }, { "epoch": 0.12, "learning_rate": 2.9843664674769477e-05, "loss": 0.324, "step": 2670 }, { "epoch": 0.12, "learning_rate": 2.9843196604933457e-05, "loss": 0.2328, "step": 2675 }, { "epoch": 0.13, "learning_rate": 2.9842728535097436e-05, "loss": 0.403, "step": 2680 }, { "epoch": 0.13, "learning_rate": 2.9842260465261416e-05, "loss": 0.3824, "step": 2685 }, { "epoch": 0.13, "learning_rate": 2.9841792395425396e-05, "loss": 0.547, "step": 2690 }, { "epoch": 0.13, "learning_rate": 2.984132432558938e-05, "loss": 0.5246, "step": 2695 }, { "epoch": 0.13, "learning_rate": 2.984085625575336e-05, "loss": 0.7502, "step": 2700 }, { "epoch": 0.13, "learning_rate": 2.984038818591734e-05, "loss": 0.4464, "step": 2705 }, { "epoch": 0.13, "learning_rate": 2.983992011608132e-05, "loss": 0.1972, "step": 2710 }, { "epoch": 0.13, "learning_rate": 2.9839452046245302e-05, "loss": 0.3288, "step": 2715 }, { "epoch": 0.13, "learning_rate": 2.983898397640928e-05, "loss": 0.3094, "step": 2720 }, { "epoch": 0.13, "learning_rate": 2.983851590657326e-05, "loss": 0.3497, "step": 2725 }, { "epoch": 0.13, "learning_rate": 2.9838047836737245e-05, "loss": 0.35, "step": 2730 }, { "epoch": 0.13, "learning_rate": 2.9837579766901224e-05, "loss": 0.4482, "step": 2735 }, { "epoch": 0.13, "learning_rate": 2.98371116970652e-05, "loss": 0.4094, "step": 2740 }, { "epoch": 0.13, "learning_rate": 2.983664362722918e-05, "loss": 0.5456, "step": 2745 }, { "epoch": 0.13, "learning_rate": 2.9836175557393164e-05, "loss": 0.6946, "step": 2750 }, { "epoch": 0.13, "learning_rate": 2.9835707487557144e-05, "loss": 0.3944, "step": 2755 }, { "epoch": 0.13, "learning_rate": 2.9835239417721124e-05, "loss": 0.1891, "step": 2760 }, { "epoch": 0.13, "learning_rate": 2.9834771347885103e-05, "loss": 0.1876, "step": 2765 }, { "epoch": 0.13, "learning_rate": 2.9834303278049087e-05, "loss": 0.3824, "step": 2770 }, { "epoch": 0.13, "learning_rate": 2.9833835208213066e-05, "loss": 0.3722, "step": 2775 }, { "epoch": 0.13, "learning_rate": 2.9833367138377046e-05, "loss": 0.4569, "step": 2780 }, { "epoch": 0.13, "learning_rate": 2.983289906854103e-05, "loss": 0.5069, "step": 2785 }, { "epoch": 0.13, "learning_rate": 2.983243099870501e-05, "loss": 0.5151, "step": 2790 }, { "epoch": 0.13, "learning_rate": 2.983196292886899e-05, "loss": 0.5431, "step": 2795 }, { "epoch": 0.13, "learning_rate": 2.983149485903297e-05, "loss": 1.0826, "step": 2800 }, { "epoch": 0.13, "learning_rate": 2.983102678919695e-05, "loss": 0.3695, "step": 2805 }, { "epoch": 0.13, "learning_rate": 2.983055871936093e-05, "loss": 0.2276, "step": 2810 }, { "epoch": 0.13, "learning_rate": 2.983009064952491e-05, "loss": 0.3133, "step": 2815 }, { "epoch": 0.13, "learning_rate": 2.9829622579688888e-05, "loss": 0.3186, "step": 2820 }, { "epoch": 0.13, "learning_rate": 2.982915450985287e-05, "loss": 0.3351, "step": 2825 }, { "epoch": 0.13, "learning_rate": 2.982868644001685e-05, "loss": 0.3789, "step": 2830 }, { "epoch": 0.13, "learning_rate": 2.982821837018083e-05, "loss": 0.3645, "step": 2835 }, { "epoch": 0.13, "learning_rate": 2.9827750300344814e-05, "loss": 0.4059, "step": 2840 }, { "epoch": 0.13, "learning_rate": 2.9827282230508794e-05, "loss": 0.3996, "step": 2845 }, { "epoch": 0.13, "learning_rate": 2.9826814160672774e-05, "loss": 0.6644, "step": 2850 }, { "epoch": 0.13, "learning_rate": 2.9826346090836754e-05, "loss": 0.4051, "step": 2855 }, { "epoch": 0.13, "learning_rate": 2.9825878021000737e-05, "loss": 0.0858, "step": 2860 }, { "epoch": 0.13, "learning_rate": 2.9825409951164713e-05, "loss": 0.2712, "step": 2865 }, { "epoch": 0.13, "learning_rate": 2.9824941881328693e-05, "loss": 0.3066, "step": 2870 }, { "epoch": 0.13, "learning_rate": 2.9824473811492673e-05, "loss": 0.2838, "step": 2875 }, { "epoch": 0.13, "learning_rate": 2.9824005741656656e-05, "loss": 0.3694, "step": 2880 }, { "epoch": 0.13, "learning_rate": 2.9823537671820636e-05, "loss": 0.5235, "step": 2885 }, { "epoch": 0.13, "learning_rate": 2.9823069601984616e-05, "loss": 0.63, "step": 2890 }, { "epoch": 0.14, "learning_rate": 2.98226015321486e-05, "loss": 0.5719, "step": 2895 }, { "epoch": 0.14, "learning_rate": 2.982213346231258e-05, "loss": 0.7585, "step": 2900 }, { "epoch": 0.14, "learning_rate": 2.982166539247656e-05, "loss": 0.3946, "step": 2905 }, { "epoch": 0.14, "learning_rate": 2.982119732264054e-05, "loss": 0.2283, "step": 2910 }, { "epoch": 0.14, "learning_rate": 2.982072925280452e-05, "loss": 0.2433, "step": 2915 }, { "epoch": 0.14, "learning_rate": 2.98202611829685e-05, "loss": 0.2719, "step": 2920 }, { "epoch": 0.14, "learning_rate": 2.981979311313248e-05, "loss": 0.3201, "step": 2925 }, { "epoch": 0.14, "learning_rate": 2.9819325043296458e-05, "loss": 0.34, "step": 2930 }, { "epoch": 0.14, "learning_rate": 2.981885697346044e-05, "loss": 0.4069, "step": 2935 }, { "epoch": 0.14, "learning_rate": 2.981838890362442e-05, "loss": 0.4682, "step": 2940 }, { "epoch": 0.14, "learning_rate": 2.98179208337884e-05, "loss": 0.5898, "step": 2945 }, { "epoch": 0.14, "learning_rate": 2.981745276395238e-05, "loss": 0.6208, "step": 2950 }, { "epoch": 0.14, "learning_rate": 2.9816984694116364e-05, "loss": 0.5141, "step": 2955 }, { "epoch": 0.14, "learning_rate": 2.9816516624280343e-05, "loss": 0.2397, "step": 2960 }, { "epoch": 0.14, "learning_rate": 2.9816048554444323e-05, "loss": 0.211, "step": 2965 }, { "epoch": 0.14, "learning_rate": 2.9815580484608306e-05, "loss": 0.2892, "step": 2970 }, { "epoch": 0.14, "learning_rate": 2.9815112414772286e-05, "loss": 0.4122, "step": 2975 }, { "epoch": 0.14, "learning_rate": 2.9814644344936266e-05, "loss": 0.3687, "step": 2980 }, { "epoch": 0.14, "learning_rate": 2.9814176275100246e-05, "loss": 0.2965, "step": 2985 }, { "epoch": 0.14, "learning_rate": 2.9813708205264226e-05, "loss": 0.5304, "step": 2990 }, { "epoch": 0.14, "learning_rate": 2.9813240135428206e-05, "loss": 0.5774, "step": 2995 }, { "epoch": 0.14, "learning_rate": 2.9812772065592185e-05, "loss": 0.774, "step": 3000 }, { "epoch": 0.14, "learning_rate": 2.9812303995756165e-05, "loss": 0.4824, "step": 3005 }, { "epoch": 0.14, "learning_rate": 2.981183592592015e-05, "loss": 0.186, "step": 3010 }, { "epoch": 0.14, "learning_rate": 2.9811367856084128e-05, "loss": 0.2796, "step": 3015 }, { "epoch": 0.14, "learning_rate": 2.9810899786248108e-05, "loss": 0.2142, "step": 3020 }, { "epoch": 0.14, "learning_rate": 2.981043171641209e-05, "loss": 0.2992, "step": 3025 }, { "epoch": 0.14, "learning_rate": 2.980996364657607e-05, "loss": 0.3176, "step": 3030 }, { "epoch": 0.14, "learning_rate": 2.980949557674005e-05, "loss": 0.5671, "step": 3035 }, { "epoch": 0.14, "learning_rate": 2.980902750690403e-05, "loss": 0.4956, "step": 3040 }, { "epoch": 0.14, "learning_rate": 2.9808559437068014e-05, "loss": 0.5212, "step": 3045 }, { "epoch": 0.14, "learning_rate": 2.9808091367231994e-05, "loss": 0.7228, "step": 3050 }, { "epoch": 0.14, "learning_rate": 2.980762329739597e-05, "loss": 0.3316, "step": 3055 }, { "epoch": 0.14, "learning_rate": 2.980715522755995e-05, "loss": 0.1685, "step": 3060 }, { "epoch": 0.14, "learning_rate": 2.9806687157723933e-05, "loss": 0.2233, "step": 3065 }, { "epoch": 0.14, "learning_rate": 2.9806219087887913e-05, "loss": 0.3173, "step": 3070 }, { "epoch": 0.14, "learning_rate": 2.9805751018051893e-05, "loss": 0.3252, "step": 3075 }, { "epoch": 0.14, "learning_rate": 2.9805282948215876e-05, "loss": 0.326, "step": 3080 }, { "epoch": 0.14, "learning_rate": 2.9804814878379856e-05, "loss": 0.3709, "step": 3085 }, { "epoch": 0.14, "learning_rate": 2.9804346808543836e-05, "loss": 0.5064, "step": 3090 }, { "epoch": 0.14, "learning_rate": 2.9803878738707815e-05, "loss": 0.5457, "step": 3095 }, { "epoch": 0.14, "learning_rate": 2.98034106688718e-05, "loss": 1.0209, "step": 3100 }, { "epoch": 0.14, "learning_rate": 2.980294259903578e-05, "loss": 0.3009, "step": 3105 }, { "epoch": 0.15, "learning_rate": 2.9802474529199758e-05, "loss": 0.2032, "step": 3110 }, { "epoch": 0.15, "learning_rate": 2.9802006459363738e-05, "loss": 0.3715, "step": 3115 }, { "epoch": 0.15, "learning_rate": 2.9801538389527718e-05, "loss": 0.3311, "step": 3120 }, { "epoch": 0.15, "learning_rate": 2.9801070319691698e-05, "loss": 0.3749, "step": 3125 }, { "epoch": 0.15, "learning_rate": 2.9800602249855678e-05, "loss": 0.3707, "step": 3130 }, { "epoch": 0.15, "learning_rate": 2.9800134180019657e-05, "loss": 0.3479, "step": 3135 }, { "epoch": 0.15, "learning_rate": 2.979966611018364e-05, "loss": 0.5007, "step": 3140 }, { "epoch": 0.15, "learning_rate": 2.979919804034762e-05, "loss": 0.555, "step": 3145 }, { "epoch": 0.15, "learning_rate": 2.97987299705116e-05, "loss": 0.6763, "step": 3150 }, { "epoch": 0.15, "learning_rate": 2.9798261900675583e-05, "loss": 0.3482, "step": 3155 }, { "epoch": 0.15, "learning_rate": 2.9797793830839563e-05, "loss": 0.1475, "step": 3160 }, { "epoch": 0.15, "learning_rate": 2.9797325761003543e-05, "loss": 0.1497, "step": 3165 }, { "epoch": 0.15, "learning_rate": 2.9796857691167523e-05, "loss": 0.2737, "step": 3170 }, { "epoch": 0.15, "learning_rate": 2.9796389621331506e-05, "loss": 0.3525, "step": 3175 }, { "epoch": 0.15, "learning_rate": 2.9795921551495482e-05, "loss": 0.3627, "step": 3180 }, { "epoch": 0.15, "learning_rate": 2.9795453481659462e-05, "loss": 0.5222, "step": 3185 }, { "epoch": 0.15, "learning_rate": 2.9794985411823442e-05, "loss": 0.5057, "step": 3190 }, { "epoch": 0.15, "learning_rate": 2.9794517341987425e-05, "loss": 0.4937, "step": 3195 }, { "epoch": 0.15, "learning_rate": 2.9794049272151405e-05, "loss": 0.696, "step": 3200 }, { "epoch": 0.15, "learning_rate": 2.9793581202315385e-05, "loss": 0.5451, "step": 3205 }, { "epoch": 0.15, "learning_rate": 2.9793113132479368e-05, "loss": 0.1988, "step": 3210 }, { "epoch": 0.15, "learning_rate": 2.9792645062643348e-05, "loss": 0.1992, "step": 3215 }, { "epoch": 0.15, "learning_rate": 2.9792176992807328e-05, "loss": 0.263, "step": 3220 }, { "epoch": 0.15, "learning_rate": 2.9791708922971308e-05, "loss": 0.325, "step": 3225 }, { "epoch": 0.15, "learning_rate": 2.979124085313529e-05, "loss": 0.3584, "step": 3230 }, { "epoch": 0.15, "learning_rate": 2.979077278329927e-05, "loss": 0.5023, "step": 3235 }, { "epoch": 0.15, "learning_rate": 2.979030471346325e-05, "loss": 0.6313, "step": 3240 }, { "epoch": 0.15, "learning_rate": 2.9789836643627227e-05, "loss": 0.762, "step": 3245 }, { "epoch": 0.15, "learning_rate": 2.978936857379121e-05, "loss": 0.7333, "step": 3250 }, { "epoch": 0.15, "learning_rate": 2.978890050395519e-05, "loss": 0.3004, "step": 3255 }, { "epoch": 0.15, "learning_rate": 2.978843243411917e-05, "loss": 0.1905, "step": 3260 }, { "epoch": 0.15, "learning_rate": 2.9787964364283153e-05, "loss": 0.2138, "step": 3265 }, { "epoch": 0.15, "learning_rate": 2.9787496294447133e-05, "loss": 0.3551, "step": 3270 }, { "epoch": 0.15, "learning_rate": 2.9787028224611113e-05, "loss": 0.2628, "step": 3275 }, { "epoch": 0.15, "learning_rate": 2.9786560154775092e-05, "loss": 0.3818, "step": 3280 }, { "epoch": 0.15, "learning_rate": 2.9786092084939076e-05, "loss": 0.4099, "step": 3285 }, { "epoch": 0.15, "learning_rate": 2.9785624015103055e-05, "loss": 0.4, "step": 3290 }, { "epoch": 0.15, "learning_rate": 2.9785155945267035e-05, "loss": 0.5284, "step": 3295 }, { "epoch": 0.15, "learning_rate": 2.9784687875431015e-05, "loss": 0.8593, "step": 3300 }, { "epoch": 0.15, "learning_rate": 2.9784219805594998e-05, "loss": 0.4611, "step": 3305 }, { "epoch": 0.15, "learning_rate": 2.9783751735758975e-05, "loss": 0.1586, "step": 3310 }, { "epoch": 0.15, "learning_rate": 2.9783283665922954e-05, "loss": 0.2415, "step": 3315 }, { "epoch": 0.15, "learning_rate": 2.9782815596086934e-05, "loss": 0.2548, "step": 3320 }, { "epoch": 0.16, "learning_rate": 2.9782347526250918e-05, "loss": 0.3687, "step": 3325 }, { "epoch": 0.16, "learning_rate": 2.9781879456414897e-05, "loss": 0.4, "step": 3330 }, { "epoch": 0.16, "learning_rate": 2.9781411386578877e-05, "loss": 0.4192, "step": 3335 }, { "epoch": 0.16, "learning_rate": 2.978094331674286e-05, "loss": 0.3799, "step": 3340 }, { "epoch": 0.16, "learning_rate": 2.978047524690684e-05, "loss": 0.6527, "step": 3345 }, { "epoch": 0.16, "learning_rate": 2.978000717707082e-05, "loss": 0.8422, "step": 3350 }, { "epoch": 0.16, "learning_rate": 2.97795391072348e-05, "loss": 0.4539, "step": 3355 }, { "epoch": 0.16, "learning_rate": 2.9779071037398783e-05, "loss": 0.2319, "step": 3360 }, { "epoch": 0.16, "learning_rate": 2.9778602967562763e-05, "loss": 0.1937, "step": 3365 }, { "epoch": 0.16, "learning_rate": 2.977813489772674e-05, "loss": 0.2484, "step": 3370 }, { "epoch": 0.16, "learning_rate": 2.977766682789072e-05, "loss": 0.2864, "step": 3375 }, { "epoch": 0.16, "learning_rate": 2.9777198758054702e-05, "loss": 0.3665, "step": 3380 }, { "epoch": 0.16, "learning_rate": 2.9776730688218682e-05, "loss": 0.4467, "step": 3385 }, { "epoch": 0.16, "learning_rate": 2.9776262618382662e-05, "loss": 0.4694, "step": 3390 }, { "epoch": 0.16, "learning_rate": 2.9775794548546645e-05, "loss": 0.3766, "step": 3395 }, { "epoch": 0.16, "learning_rate": 2.9775326478710625e-05, "loss": 0.77, "step": 3400 }, { "epoch": 0.16, "learning_rate": 2.9774858408874605e-05, "loss": 0.3687, "step": 3405 }, { "epoch": 0.16, "learning_rate": 2.9774390339038585e-05, "loss": 0.2109, "step": 3410 }, { "epoch": 0.16, "learning_rate": 2.9773922269202568e-05, "loss": 0.255, "step": 3415 }, { "epoch": 0.16, "learning_rate": 2.9773454199366548e-05, "loss": 0.2809, "step": 3420 }, { "epoch": 0.16, "learning_rate": 2.9772986129530527e-05, "loss": 0.3345, "step": 3425 }, { "epoch": 0.16, "learning_rate": 2.9772518059694507e-05, "loss": 0.3559, "step": 3430 }, { "epoch": 0.16, "learning_rate": 2.9772049989858487e-05, "loss": 0.3222, "step": 3435 }, { "epoch": 0.16, "learning_rate": 2.9771581920022467e-05, "loss": 0.4967, "step": 3440 }, { "epoch": 0.16, "learning_rate": 2.9771113850186447e-05, "loss": 0.5792, "step": 3445 }, { "epoch": 0.16, "learning_rate": 2.977064578035043e-05, "loss": 0.889, "step": 3450 }, { "epoch": 0.16, "learning_rate": 2.977017771051441e-05, "loss": 0.352, "step": 3455 }, { "epoch": 0.16, "learning_rate": 2.976970964067839e-05, "loss": 0.1901, "step": 3460 }, { "epoch": 0.16, "learning_rate": 2.976924157084237e-05, "loss": 0.2705, "step": 3465 }, { "epoch": 0.16, "learning_rate": 2.9768773501006353e-05, "loss": 0.2532, "step": 3470 }, { "epoch": 0.16, "learning_rate": 2.9768305431170332e-05, "loss": 0.3798, "step": 3475 }, { "epoch": 0.16, "learning_rate": 2.9767837361334312e-05, "loss": 0.2622, "step": 3480 }, { "epoch": 0.16, "learning_rate": 2.9767369291498292e-05, "loss": 0.3909, "step": 3485 }, { "epoch": 0.16, "learning_rate": 2.9766901221662275e-05, "loss": 0.457, "step": 3490 }, { "epoch": 0.16, "learning_rate": 2.9766433151826255e-05, "loss": 0.4318, "step": 3495 }, { "epoch": 0.16, "learning_rate": 2.976596508199023e-05, "loss": 0.8586, "step": 3500 }, { "epoch": 0.16, "learning_rate": 2.9765497012154215e-05, "loss": 0.3323, "step": 3505 }, { "epoch": 0.16, "learning_rate": 2.9765028942318194e-05, "loss": 0.1509, "step": 3510 }, { "epoch": 0.16, "learning_rate": 2.9764560872482174e-05, "loss": 0.2893, "step": 3515 }, { "epoch": 0.16, "learning_rate": 2.9764092802646154e-05, "loss": 0.2656, "step": 3520 }, { "epoch": 0.16, "learning_rate": 2.9763624732810137e-05, "loss": 0.3075, "step": 3525 }, { "epoch": 0.16, "learning_rate": 2.9763156662974117e-05, "loss": 0.3944, "step": 3530 }, { "epoch": 0.16, "learning_rate": 2.9762688593138097e-05, "loss": 0.3305, "step": 3535 }, { "epoch": 0.17, "learning_rate": 2.9762220523302077e-05, "loss": 0.3977, "step": 3540 }, { "epoch": 0.17, "learning_rate": 2.976175245346606e-05, "loss": 0.5208, "step": 3545 }, { "epoch": 0.17, "learning_rate": 2.976128438363004e-05, "loss": 0.8924, "step": 3550 }, { "epoch": 0.17, "learning_rate": 2.976081631379402e-05, "loss": 0.3387, "step": 3555 }, { "epoch": 0.17, "learning_rate": 2.9760348243957996e-05, "loss": 0.159, "step": 3560 }, { "epoch": 0.17, "learning_rate": 2.975988017412198e-05, "loss": 0.2022, "step": 3565 }, { "epoch": 0.17, "learning_rate": 2.975941210428596e-05, "loss": 0.2047, "step": 3570 }, { "epoch": 0.17, "learning_rate": 2.975894403444994e-05, "loss": 0.3374, "step": 3575 }, { "epoch": 0.17, "learning_rate": 2.9758475964613922e-05, "loss": 0.3447, "step": 3580 }, { "epoch": 0.17, "learning_rate": 2.9758007894777902e-05, "loss": 0.4563, "step": 3585 }, { "epoch": 0.17, "learning_rate": 2.9757539824941882e-05, "loss": 0.3952, "step": 3590 }, { "epoch": 0.17, "learning_rate": 2.975707175510586e-05, "loss": 0.4245, "step": 3595 }, { "epoch": 0.17, "learning_rate": 2.9756603685269845e-05, "loss": 0.5382, "step": 3600 }, { "epoch": 0.17, "learning_rate": 2.9756135615433825e-05, "loss": 0.3683, "step": 3605 }, { "epoch": 0.17, "learning_rate": 2.9755667545597804e-05, "loss": 0.1625, "step": 3610 }, { "epoch": 0.17, "learning_rate": 2.9755199475761784e-05, "loss": 0.2182, "step": 3615 }, { "epoch": 0.17, "learning_rate": 2.9754731405925767e-05, "loss": 0.2643, "step": 3620 }, { "epoch": 0.17, "learning_rate": 2.9754263336089744e-05, "loss": 0.2336, "step": 3625 }, { "epoch": 0.17, "learning_rate": 2.9753795266253724e-05, "loss": 0.3078, "step": 3630 }, { "epoch": 0.17, "learning_rate": 2.9753327196417707e-05, "loss": 0.3142, "step": 3635 }, { "epoch": 0.17, "learning_rate": 2.9752859126581687e-05, "loss": 0.3788, "step": 3640 }, { "epoch": 0.17, "learning_rate": 2.9752391056745667e-05, "loss": 0.4912, "step": 3645 }, { "epoch": 0.17, "learning_rate": 2.9751922986909646e-05, "loss": 0.9646, "step": 3650 }, { "epoch": 0.17, "learning_rate": 2.975145491707363e-05, "loss": 0.3359, "step": 3655 }, { "epoch": 0.17, "learning_rate": 2.975098684723761e-05, "loss": 0.1838, "step": 3660 }, { "epoch": 0.17, "learning_rate": 2.975051877740159e-05, "loss": 0.1841, "step": 3665 }, { "epoch": 0.17, "learning_rate": 2.975005070756557e-05, "loss": 0.2018, "step": 3670 }, { "epoch": 0.17, "learning_rate": 2.9749582637729552e-05, "loss": 0.2357, "step": 3675 }, { "epoch": 0.17, "learning_rate": 2.9749114567893532e-05, "loss": 0.2938, "step": 3680 }, { "epoch": 0.17, "learning_rate": 2.9748646498057512e-05, "loss": 0.2645, "step": 3685 }, { "epoch": 0.17, "learning_rate": 2.974817842822149e-05, "loss": 0.3835, "step": 3690 }, { "epoch": 0.17, "learning_rate": 2.974771035838547e-05, "loss": 0.6618, "step": 3695 }, { "epoch": 0.17, "learning_rate": 2.974724228854945e-05, "loss": 0.9241, "step": 3700 }, { "epoch": 0.17, "learning_rate": 2.974677421871343e-05, "loss": 0.4766, "step": 3705 }, { "epoch": 0.17, "learning_rate": 2.9746306148877414e-05, "loss": 0.1703, "step": 3710 }, { "epoch": 0.17, "learning_rate": 2.9745838079041394e-05, "loss": 0.3016, "step": 3715 }, { "epoch": 0.17, "learning_rate": 2.9745370009205374e-05, "loss": 0.2396, "step": 3720 }, { "epoch": 0.17, "learning_rate": 2.9744901939369354e-05, "loss": 0.2578, "step": 3725 }, { "epoch": 0.17, "learning_rate": 2.9744433869533337e-05, "loss": 0.3347, "step": 3730 }, { "epoch": 0.17, "learning_rate": 2.9743965799697317e-05, "loss": 0.3658, "step": 3735 }, { "epoch": 0.17, "learning_rate": 2.9743497729861297e-05, "loss": 0.3817, "step": 3740 }, { "epoch": 0.17, "learning_rate": 2.9743029660025276e-05, "loss": 0.4796, "step": 3745 }, { "epoch": 0.17, "learning_rate": 2.9742561590189256e-05, "loss": 0.7048, "step": 3750 }, { "epoch": 0.18, "learning_rate": 2.9742093520353236e-05, "loss": 0.3976, "step": 3755 }, { "epoch": 0.18, "learning_rate": 2.9741625450517216e-05, "loss": 0.1921, "step": 3760 }, { "epoch": 0.18, "learning_rate": 2.97411573806812e-05, "loss": 0.2238, "step": 3765 }, { "epoch": 0.18, "learning_rate": 2.974068931084518e-05, "loss": 0.2518, "step": 3770 }, { "epoch": 0.18, "learning_rate": 2.974022124100916e-05, "loss": 0.2936, "step": 3775 }, { "epoch": 0.18, "learning_rate": 2.973975317117314e-05, "loss": 0.2794, "step": 3780 }, { "epoch": 0.18, "learning_rate": 2.9739285101337122e-05, "loss": 0.3757, "step": 3785 }, { "epoch": 0.18, "learning_rate": 2.97388170315011e-05, "loss": 0.2954, "step": 3790 }, { "epoch": 0.18, "learning_rate": 2.973834896166508e-05, "loss": 0.5374, "step": 3795 }, { "epoch": 0.18, "learning_rate": 2.973788089182906e-05, "loss": 0.8038, "step": 3800 }, { "epoch": 0.18, "learning_rate": 2.9737412821993044e-05, "loss": 0.3343, "step": 3805 }, { "epoch": 0.18, "learning_rate": 2.9736944752157024e-05, "loss": 0.2778, "step": 3810 }, { "epoch": 0.18, "learning_rate": 2.9736476682321e-05, "loss": 0.2653, "step": 3815 }, { "epoch": 0.18, "learning_rate": 2.9736008612484984e-05, "loss": 0.2064, "step": 3820 }, { "epoch": 0.18, "learning_rate": 2.9735540542648964e-05, "loss": 0.2617, "step": 3825 }, { "epoch": 0.18, "learning_rate": 2.9735072472812943e-05, "loss": 0.3189, "step": 3830 }, { "epoch": 0.18, "learning_rate": 2.9734604402976923e-05, "loss": 0.448, "step": 3835 }, { "epoch": 0.18, "learning_rate": 2.9734136333140907e-05, "loss": 0.2858, "step": 3840 }, { "epoch": 0.18, "learning_rate": 2.9733668263304886e-05, "loss": 0.3845, "step": 3845 }, { "epoch": 0.18, "learning_rate": 2.9733200193468866e-05, "loss": 0.5596, "step": 3850 }, { "epoch": 0.18, "learning_rate": 2.9732732123632846e-05, "loss": 0.4199, "step": 3855 }, { "epoch": 0.18, "learning_rate": 2.973226405379683e-05, "loss": 0.1667, "step": 3860 }, { "epoch": 0.18, "learning_rate": 2.973179598396081e-05, "loss": 0.2107, "step": 3865 }, { "epoch": 0.18, "learning_rate": 2.973132791412479e-05, "loss": 0.2686, "step": 3870 }, { "epoch": 0.18, "learning_rate": 2.973085984428877e-05, "loss": 0.2633, "step": 3875 }, { "epoch": 0.18, "learning_rate": 2.973039177445275e-05, "loss": 0.2741, "step": 3880 }, { "epoch": 0.18, "learning_rate": 2.9729923704616728e-05, "loss": 0.3799, "step": 3885 }, { "epoch": 0.18, "learning_rate": 2.9729455634780708e-05, "loss": 0.4158, "step": 3890 }, { "epoch": 0.18, "learning_rate": 2.972898756494469e-05, "loss": 0.3757, "step": 3895 }, { "epoch": 0.18, "learning_rate": 2.972851949510867e-05, "loss": 0.8281, "step": 3900 }, { "epoch": 0.18, "learning_rate": 2.972805142527265e-05, "loss": 0.4027, "step": 3905 }, { "epoch": 0.18, "learning_rate": 2.972758335543663e-05, "loss": 0.18, "step": 3910 }, { "epoch": 0.18, "learning_rate": 2.9727115285600614e-05, "loss": 0.2759, "step": 3915 }, { "epoch": 0.18, "learning_rate": 2.9726647215764594e-05, "loss": 0.2724, "step": 3920 }, { "epoch": 0.18, "learning_rate": 2.9726179145928574e-05, "loss": 0.2124, "step": 3925 }, { "epoch": 0.18, "learning_rate": 2.9725711076092553e-05, "loss": 0.2936, "step": 3930 }, { "epoch": 0.18, "learning_rate": 2.9725243006256537e-05, "loss": 0.3913, "step": 3935 }, { "epoch": 0.18, "learning_rate": 2.9724774936420513e-05, "loss": 0.3398, "step": 3940 }, { "epoch": 0.18, "learning_rate": 2.9724306866584493e-05, "loss": 0.4383, "step": 3945 }, { "epoch": 0.18, "learning_rate": 2.9723838796748476e-05, "loss": 0.757, "step": 3950 }, { "epoch": 0.18, "learning_rate": 2.9723370726912456e-05, "loss": 0.3838, "step": 3955 }, { "epoch": 0.18, "learning_rate": 2.9722902657076436e-05, "loss": 0.2418, "step": 3960 }, { "epoch": 0.19, "learning_rate": 2.9722434587240415e-05, "loss": 0.2861, "step": 3965 }, { "epoch": 0.19, "learning_rate": 2.97219665174044e-05, "loss": 0.2855, "step": 3970 }, { "epoch": 0.19, "learning_rate": 2.972149844756838e-05, "loss": 0.26, "step": 3975 }, { "epoch": 0.19, "learning_rate": 2.972103037773236e-05, "loss": 0.2921, "step": 3980 }, { "epoch": 0.19, "learning_rate": 2.9720562307896338e-05, "loss": 0.3622, "step": 3985 }, { "epoch": 0.19, "learning_rate": 2.972009423806032e-05, "loss": 0.4226, "step": 3990 }, { "epoch": 0.19, "learning_rate": 2.97196261682243e-05, "loss": 0.4777, "step": 3995 }, { "epoch": 0.19, "learning_rate": 2.971915809838828e-05, "loss": 0.6818, "step": 4000 }, { "epoch": 0.19, "learning_rate": 2.971869002855226e-05, "loss": 0.3419, "step": 4005 }, { "epoch": 0.19, "learning_rate": 2.971822195871624e-05, "loss": 0.1553, "step": 4010 }, { "epoch": 0.19, "learning_rate": 2.971775388888022e-05, "loss": 0.2399, "step": 4015 }, { "epoch": 0.19, "learning_rate": 2.97172858190442e-05, "loss": 0.2925, "step": 4020 }, { "epoch": 0.19, "learning_rate": 2.9716817749208183e-05, "loss": 0.3096, "step": 4025 }, { "epoch": 0.19, "learning_rate": 2.9716349679372163e-05, "loss": 0.3317, "step": 4030 }, { "epoch": 0.19, "learning_rate": 2.9715881609536143e-05, "loss": 0.3364, "step": 4035 }, { "epoch": 0.19, "learning_rate": 2.9715413539700123e-05, "loss": 0.3538, "step": 4040 }, { "epoch": 0.19, "learning_rate": 2.9714945469864106e-05, "loss": 0.5839, "step": 4045 }, { "epoch": 0.19, "learning_rate": 2.9714477400028086e-05, "loss": 0.8043, "step": 4050 }, { "epoch": 0.19, "learning_rate": 2.9714009330192066e-05, "loss": 0.372, "step": 4055 }, { "epoch": 0.19, "learning_rate": 2.971354126035605e-05, "loss": 0.1749, "step": 4060 }, { "epoch": 0.19, "learning_rate": 2.9713073190520025e-05, "loss": 0.2517, "step": 4065 }, { "epoch": 0.19, "learning_rate": 2.9712605120684005e-05, "loss": 0.2707, "step": 4070 }, { "epoch": 0.19, "learning_rate": 2.9712137050847985e-05, "loss": 0.2532, "step": 4075 }, { "epoch": 0.19, "learning_rate": 2.9711668981011968e-05, "loss": 0.3381, "step": 4080 }, { "epoch": 0.19, "learning_rate": 2.9711200911175948e-05, "loss": 0.4232, "step": 4085 }, { "epoch": 0.19, "learning_rate": 2.9710732841339928e-05, "loss": 0.6174, "step": 4090 }, { "epoch": 0.19, "learning_rate": 2.9710264771503908e-05, "loss": 0.5029, "step": 4095 }, { "epoch": 0.19, "learning_rate": 2.970979670166789e-05, "loss": 0.6542, "step": 4100 }, { "epoch": 0.19, "learning_rate": 2.970932863183187e-05, "loss": 0.3444, "step": 4105 }, { "epoch": 0.19, "learning_rate": 2.970886056199585e-05, "loss": 0.2102, "step": 4110 }, { "epoch": 0.19, "learning_rate": 2.9708392492159834e-05, "loss": 0.19, "step": 4115 }, { "epoch": 0.19, "learning_rate": 2.9707924422323814e-05, "loss": 0.2002, "step": 4120 }, { "epoch": 0.19, "learning_rate": 2.9707456352487793e-05, "loss": 0.2521, "step": 4125 }, { "epoch": 0.19, "learning_rate": 2.970698828265177e-05, "loss": 0.3259, "step": 4130 }, { "epoch": 0.19, "learning_rate": 2.9706520212815753e-05, "loss": 0.3844, "step": 4135 }, { "epoch": 0.19, "learning_rate": 2.9706052142979733e-05, "loss": 0.3614, "step": 4140 }, { "epoch": 0.19, "learning_rate": 2.9705584073143713e-05, "loss": 0.3912, "step": 4145 }, { "epoch": 0.19, "learning_rate": 2.9705116003307692e-05, "loss": 0.7987, "step": 4150 }, { "epoch": 0.19, "learning_rate": 2.9704647933471676e-05, "loss": 0.4521, "step": 4155 }, { "epoch": 0.19, "learning_rate": 2.9704179863635655e-05, "loss": 0.2179, "step": 4160 }, { "epoch": 0.19, "learning_rate": 2.9703711793799635e-05, "loss": 0.1927, "step": 4165 }, { "epoch": 0.19, "learning_rate": 2.9703243723963615e-05, "loss": 0.2198, "step": 4170 }, { "epoch": 0.19, "learning_rate": 2.97027756541276e-05, "loss": 0.1865, "step": 4175 }, { "epoch": 0.2, "learning_rate": 2.9702307584291578e-05, "loss": 0.2242, "step": 4180 }, { "epoch": 0.2, "learning_rate": 2.9701839514455558e-05, "loss": 0.4013, "step": 4185 }, { "epoch": 0.2, "learning_rate": 2.9701371444619538e-05, "loss": 0.4213, "step": 4190 }, { "epoch": 0.2, "learning_rate": 2.9700903374783518e-05, "loss": 0.4226, "step": 4195 }, { "epoch": 0.2, "learning_rate": 2.9700435304947497e-05, "loss": 0.808, "step": 4200 }, { "epoch": 0.2, "learning_rate": 2.9699967235111477e-05, "loss": 0.3286, "step": 4205 }, { "epoch": 0.2, "learning_rate": 2.969949916527546e-05, "loss": 0.1521, "step": 4210 }, { "epoch": 0.2, "learning_rate": 2.969903109543944e-05, "loss": 0.1065, "step": 4215 }, { "epoch": 0.2, "learning_rate": 2.969856302560342e-05, "loss": 0.223, "step": 4220 }, { "epoch": 0.2, "learning_rate": 2.96980949557674e-05, "loss": 0.2412, "step": 4225 }, { "epoch": 0.2, "learning_rate": 2.9697626885931383e-05, "loss": 0.3471, "step": 4230 }, { "epoch": 0.2, "learning_rate": 2.9697158816095363e-05, "loss": 0.3856, "step": 4235 }, { "epoch": 0.2, "learning_rate": 2.9696690746259343e-05, "loss": 0.4513, "step": 4240 }, { "epoch": 0.2, "learning_rate": 2.9696222676423326e-05, "loss": 0.4989, "step": 4245 }, { "epoch": 0.2, "learning_rate": 2.9695754606587306e-05, "loss": 0.7143, "step": 4250 }, { "epoch": 0.2, "learning_rate": 2.9695286536751282e-05, "loss": 0.3467, "step": 4255 }, { "epoch": 0.2, "learning_rate": 2.9694818466915262e-05, "loss": 0.1694, "step": 4260 }, { "epoch": 0.2, "learning_rate": 2.9694350397079245e-05, "loss": 0.1576, "step": 4265 }, { "epoch": 0.2, "learning_rate": 2.9693882327243225e-05, "loss": 0.2404, "step": 4270 }, { "epoch": 0.2, "learning_rate": 2.9693414257407205e-05, "loss": 0.2594, "step": 4275 }, { "epoch": 0.2, "learning_rate": 2.9692946187571185e-05, "loss": 0.3529, "step": 4280 }, { "epoch": 0.2, "learning_rate": 2.9692478117735168e-05, "loss": 0.262, "step": 4285 }, { "epoch": 0.2, "learning_rate": 2.9692010047899148e-05, "loss": 0.4635, "step": 4290 }, { "epoch": 0.2, "learning_rate": 2.9691541978063128e-05, "loss": 0.5061, "step": 4295 }, { "epoch": 0.2, "learning_rate": 2.969107390822711e-05, "loss": 0.8731, "step": 4300 }, { "epoch": 0.2, "learning_rate": 2.969060583839109e-05, "loss": 0.4071, "step": 4305 }, { "epoch": 0.2, "learning_rate": 2.969013776855507e-05, "loss": 0.1383, "step": 4310 }, { "epoch": 0.2, "learning_rate": 2.968966969871905e-05, "loss": 0.3105, "step": 4315 }, { "epoch": 0.2, "learning_rate": 2.968920162888303e-05, "loss": 0.2413, "step": 4320 }, { "epoch": 0.2, "learning_rate": 2.968873355904701e-05, "loss": 0.2718, "step": 4325 }, { "epoch": 0.2, "learning_rate": 2.968826548921099e-05, "loss": 0.2647, "step": 4330 }, { "epoch": 0.2, "learning_rate": 2.968779741937497e-05, "loss": 0.4238, "step": 4335 }, { "epoch": 0.2, "learning_rate": 2.9687329349538953e-05, "loss": 0.3297, "step": 4340 }, { "epoch": 0.2, "learning_rate": 2.9686861279702932e-05, "loss": 0.6178, "step": 4345 }, { "epoch": 0.2, "learning_rate": 2.9686393209866912e-05, "loss": 0.5912, "step": 4350 }, { "epoch": 0.2, "learning_rate": 2.9685925140030892e-05, "loss": 0.3473, "step": 4355 }, { "epoch": 0.2, "learning_rate": 2.9685457070194875e-05, "loss": 0.2283, "step": 4360 }, { "epoch": 0.2, "learning_rate": 2.9684989000358855e-05, "loss": 0.243, "step": 4365 }, { "epoch": 0.2, "learning_rate": 2.9684520930522835e-05, "loss": 0.1972, "step": 4370 }, { "epoch": 0.2, "learning_rate": 2.9684052860686818e-05, "loss": 0.2412, "step": 4375 }, { "epoch": 0.2, "learning_rate": 2.9683584790850795e-05, "loss": 0.3386, "step": 4380 }, { "epoch": 0.2, "learning_rate": 2.9683116721014774e-05, "loss": 0.3353, "step": 4385 }, { "epoch": 0.2, "learning_rate": 2.9682648651178754e-05, "loss": 0.395, "step": 4390 }, { "epoch": 0.21, "learning_rate": 2.9682180581342737e-05, "loss": 0.6366, "step": 4395 }, { "epoch": 0.21, "learning_rate": 2.9681712511506717e-05, "loss": 0.74, "step": 4400 }, { "epoch": 0.21, "learning_rate": 2.9681244441670697e-05, "loss": 0.3046, "step": 4405 }, { "epoch": 0.21, "learning_rate": 2.9680776371834677e-05, "loss": 0.1366, "step": 4410 }, { "epoch": 0.21, "learning_rate": 2.968030830199866e-05, "loss": 0.1812, "step": 4415 }, { "epoch": 0.21, "learning_rate": 2.967984023216264e-05, "loss": 0.1934, "step": 4420 }, { "epoch": 0.21, "learning_rate": 2.967937216232662e-05, "loss": 0.373, "step": 4425 }, { "epoch": 0.21, "learning_rate": 2.9678904092490603e-05, "loss": 0.2788, "step": 4430 }, { "epoch": 0.21, "learning_rate": 2.9678436022654583e-05, "loss": 0.4025, "step": 4435 }, { "epoch": 0.21, "learning_rate": 2.9677967952818563e-05, "loss": 0.319, "step": 4440 }, { "epoch": 0.21, "learning_rate": 2.967749988298254e-05, "loss": 0.4115, "step": 4445 }, { "epoch": 0.21, "learning_rate": 2.9677031813146522e-05, "loss": 0.7493, "step": 4450 }, { "epoch": 0.21, "learning_rate": 2.9676563743310502e-05, "loss": 0.3839, "step": 4455 }, { "epoch": 0.21, "learning_rate": 2.9676095673474482e-05, "loss": 0.1987, "step": 4460 }, { "epoch": 0.21, "learning_rate": 2.967562760363846e-05, "loss": 0.2261, "step": 4465 }, { "epoch": 0.21, "learning_rate": 2.9675159533802445e-05, "loss": 0.2989, "step": 4470 }, { "epoch": 0.21, "learning_rate": 2.9674691463966425e-05, "loss": 0.1901, "step": 4475 }, { "epoch": 0.21, "learning_rate": 2.9674223394130404e-05, "loss": 0.3234, "step": 4480 }, { "epoch": 0.21, "learning_rate": 2.9673755324294388e-05, "loss": 0.2385, "step": 4485 }, { "epoch": 0.21, "learning_rate": 2.9673287254458368e-05, "loss": 0.3662, "step": 4490 }, { "epoch": 0.21, "learning_rate": 2.9672819184622347e-05, "loss": 0.4362, "step": 4495 }, { "epoch": 0.21, "learning_rate": 2.9672351114786327e-05, "loss": 0.6486, "step": 4500 }, { "epoch": 0.21, "learning_rate": 2.967188304495031e-05, "loss": 0.3722, "step": 4505 }, { "epoch": 0.21, "learning_rate": 2.9671414975114287e-05, "loss": 0.1274, "step": 4510 }, { "epoch": 0.21, "learning_rate": 2.9670946905278267e-05, "loss": 0.15, "step": 4515 }, { "epoch": 0.21, "learning_rate": 2.9670478835442246e-05, "loss": 0.3022, "step": 4520 }, { "epoch": 0.21, "learning_rate": 2.967001076560623e-05, "loss": 0.2737, "step": 4525 }, { "epoch": 0.21, "learning_rate": 2.966954269577021e-05, "loss": 0.4052, "step": 4530 }, { "epoch": 0.21, "learning_rate": 2.966907462593419e-05, "loss": 0.3599, "step": 4535 }, { "epoch": 0.21, "learning_rate": 2.966860655609817e-05, "loss": 0.4664, "step": 4540 }, { "epoch": 0.21, "learning_rate": 2.9668138486262152e-05, "loss": 0.5121, "step": 4545 }, { "epoch": 0.21, "learning_rate": 2.9667670416426132e-05, "loss": 0.804, "step": 4550 }, { "epoch": 0.21, "learning_rate": 2.9667202346590112e-05, "loss": 0.3194, "step": 4555 }, { "epoch": 0.21, "learning_rate": 2.9666734276754095e-05, "loss": 0.1735, "step": 4560 }, { "epoch": 0.21, "learning_rate": 2.9666266206918075e-05, "loss": 0.1628, "step": 4565 }, { "epoch": 0.21, "learning_rate": 2.966579813708205e-05, "loss": 0.2651, "step": 4570 }, { "epoch": 0.21, "learning_rate": 2.966533006724603e-05, "loss": 0.2258, "step": 4575 }, { "epoch": 0.21, "learning_rate": 2.9664861997410014e-05, "loss": 0.2476, "step": 4580 }, { "epoch": 0.21, "learning_rate": 2.9664393927573994e-05, "loss": 0.3759, "step": 4585 }, { "epoch": 0.21, "learning_rate": 2.9663925857737974e-05, "loss": 0.4086, "step": 4590 }, { "epoch": 0.21, "learning_rate": 2.9663457787901954e-05, "loss": 0.4154, "step": 4595 }, { "epoch": 0.21, "learning_rate": 2.9662989718065937e-05, "loss": 0.5595, "step": 4600 }, { "epoch": 0.21, "learning_rate": 2.9662521648229917e-05, "loss": 0.3738, "step": 4605 }, { "epoch": 0.22, "learning_rate": 2.9662053578393897e-05, "loss": 0.1435, "step": 4610 }, { "epoch": 0.22, "learning_rate": 2.966158550855788e-05, "loss": 0.2051, "step": 4615 }, { "epoch": 0.22, "learning_rate": 2.966111743872186e-05, "loss": 0.1334, "step": 4620 }, { "epoch": 0.22, "learning_rate": 2.966064936888584e-05, "loss": 0.4194, "step": 4625 }, { "epoch": 0.22, "learning_rate": 2.966018129904982e-05, "loss": 0.2923, "step": 4630 }, { "epoch": 0.22, "learning_rate": 2.96597132292138e-05, "loss": 0.3075, "step": 4635 }, { "epoch": 0.22, "learning_rate": 2.965924515937778e-05, "loss": 0.4637, "step": 4640 }, { "epoch": 0.22, "learning_rate": 2.965877708954176e-05, "loss": 0.5122, "step": 4645 }, { "epoch": 0.22, "learning_rate": 2.965830901970574e-05, "loss": 0.6531, "step": 4650 }, { "epoch": 0.22, "learning_rate": 2.9657840949869722e-05, "loss": 0.3787, "step": 4655 }, { "epoch": 0.22, "learning_rate": 2.96573728800337e-05, "loss": 0.1451, "step": 4660 }, { "epoch": 0.22, "learning_rate": 2.965690481019768e-05, "loss": 0.244, "step": 4665 }, { "epoch": 0.22, "learning_rate": 2.9656436740361665e-05, "loss": 0.2495, "step": 4670 }, { "epoch": 0.22, "learning_rate": 2.9655968670525644e-05, "loss": 0.3045, "step": 4675 }, { "epoch": 0.22, "learning_rate": 2.9655500600689624e-05, "loss": 0.189, "step": 4680 }, { "epoch": 0.22, "learning_rate": 2.9655032530853604e-05, "loss": 0.3852, "step": 4685 }, { "epoch": 0.22, "learning_rate": 2.9654564461017587e-05, "loss": 0.4357, "step": 4690 }, { "epoch": 0.22, "learning_rate": 2.9654096391181567e-05, "loss": 0.3422, "step": 4695 }, { "epoch": 0.22, "learning_rate": 2.9653628321345544e-05, "loss": 0.6239, "step": 4700 }, { "epoch": 0.22, "learning_rate": 2.9653160251509523e-05, "loss": 0.4151, "step": 4705 }, { "epoch": 0.22, "learning_rate": 2.9652692181673507e-05, "loss": 0.1538, "step": 4710 }, { "epoch": 0.22, "learning_rate": 2.9652224111837486e-05, "loss": 0.2, "step": 4715 }, { "epoch": 0.22, "learning_rate": 2.9651756042001466e-05, "loss": 0.2349, "step": 4720 }, { "epoch": 0.22, "learning_rate": 2.9651287972165446e-05, "loss": 0.2839, "step": 4725 }, { "epoch": 0.22, "learning_rate": 2.965081990232943e-05, "loss": 0.2757, "step": 4730 }, { "epoch": 0.22, "learning_rate": 2.965035183249341e-05, "loss": 0.7267, "step": 4735 }, { "epoch": 0.22, "learning_rate": 2.964988376265739e-05, "loss": 0.3978, "step": 4740 }, { "epoch": 0.22, "learning_rate": 2.9649415692821372e-05, "loss": 0.3665, "step": 4745 }, { "epoch": 0.22, "learning_rate": 2.9648947622985352e-05, "loss": 0.7552, "step": 4750 }, { "epoch": 0.22, "learning_rate": 2.9648479553149332e-05, "loss": 0.352, "step": 4755 }, { "epoch": 0.22, "learning_rate": 2.9648011483313308e-05, "loss": 0.1796, "step": 4760 }, { "epoch": 0.22, "learning_rate": 2.964754341347729e-05, "loss": 0.2277, "step": 4765 }, { "epoch": 0.22, "learning_rate": 2.964707534364127e-05, "loss": 0.2142, "step": 4770 }, { "epoch": 0.22, "learning_rate": 2.964660727380525e-05, "loss": 0.2721, "step": 4775 }, { "epoch": 0.22, "learning_rate": 2.964613920396923e-05, "loss": 0.2925, "step": 4780 }, { "epoch": 0.22, "learning_rate": 2.9645671134133214e-05, "loss": 0.1989, "step": 4785 }, { "epoch": 0.22, "learning_rate": 2.9645203064297194e-05, "loss": 0.3872, "step": 4790 }, { "epoch": 0.22, "learning_rate": 2.9644734994461174e-05, "loss": 0.4552, "step": 4795 }, { "epoch": 0.22, "learning_rate": 2.9644266924625157e-05, "loss": 0.6281, "step": 4800 }, { "epoch": 0.22, "learning_rate": 2.9643798854789137e-05, "loss": 0.3205, "step": 4805 }, { "epoch": 0.22, "learning_rate": 2.9643330784953116e-05, "loss": 0.1648, "step": 4810 }, { "epoch": 0.22, "learning_rate": 2.9642862715117096e-05, "loss": 0.1366, "step": 4815 }, { "epoch": 0.22, "learning_rate": 2.964239464528108e-05, "loss": 0.2225, "step": 4820 }, { "epoch": 0.23, "learning_rate": 2.9641926575445056e-05, "loss": 0.2482, "step": 4825 }, { "epoch": 0.23, "learning_rate": 2.9641458505609036e-05, "loss": 0.2454, "step": 4830 }, { "epoch": 0.23, "learning_rate": 2.9640990435773016e-05, "loss": 0.4022, "step": 4835 }, { "epoch": 0.23, "learning_rate": 2.9640522365937e-05, "loss": 0.3712, "step": 4840 }, { "epoch": 0.23, "learning_rate": 2.964005429610098e-05, "loss": 0.4711, "step": 4845 }, { "epoch": 0.23, "learning_rate": 2.963958622626496e-05, "loss": 0.704, "step": 4850 }, { "epoch": 0.23, "learning_rate": 2.963911815642894e-05, "loss": 0.3407, "step": 4855 }, { "epoch": 0.23, "learning_rate": 2.963865008659292e-05, "loss": 0.1218, "step": 4860 }, { "epoch": 0.23, "learning_rate": 2.96381820167569e-05, "loss": 0.1539, "step": 4865 }, { "epoch": 0.23, "learning_rate": 2.963771394692088e-05, "loss": 0.2058, "step": 4870 }, { "epoch": 0.23, "learning_rate": 2.9637245877084864e-05, "loss": 0.3064, "step": 4875 }, { "epoch": 0.23, "learning_rate": 2.9636777807248844e-05, "loss": 0.298, "step": 4880 }, { "epoch": 0.23, "learning_rate": 2.9636309737412824e-05, "loss": 0.2942, "step": 4885 }, { "epoch": 0.23, "learning_rate": 2.96358416675768e-05, "loss": 0.376, "step": 4890 }, { "epoch": 0.23, "learning_rate": 2.9635373597740784e-05, "loss": 0.4057, "step": 4895 }, { "epoch": 0.23, "learning_rate": 2.9634905527904763e-05, "loss": 0.8207, "step": 4900 }, { "epoch": 0.23, "learning_rate": 2.9634437458068743e-05, "loss": 0.332, "step": 4905 }, { "epoch": 0.23, "learning_rate": 2.9633969388232726e-05, "loss": 0.1467, "step": 4910 }, { "epoch": 0.23, "learning_rate": 2.9633501318396706e-05, "loss": 0.1282, "step": 4915 }, { "epoch": 0.23, "learning_rate": 2.9633033248560686e-05, "loss": 0.2048, "step": 4920 }, { "epoch": 0.23, "learning_rate": 2.9632565178724666e-05, "loss": 0.26, "step": 4925 }, { "epoch": 0.23, "learning_rate": 2.963209710888865e-05, "loss": 0.3335, "step": 4930 }, { "epoch": 0.23, "learning_rate": 2.963162903905263e-05, "loss": 0.2966, "step": 4935 }, { "epoch": 0.23, "learning_rate": 2.963116096921661e-05, "loss": 0.3337, "step": 4940 }, { "epoch": 0.23, "learning_rate": 2.963069289938059e-05, "loss": 0.4369, "step": 4945 }, { "epoch": 0.23, "learning_rate": 2.963022482954457e-05, "loss": 0.6286, "step": 4950 }, { "epoch": 0.23, "learning_rate": 2.9629756759708548e-05, "loss": 0.3736, "step": 4955 }, { "epoch": 0.23, "learning_rate": 2.9629288689872528e-05, "loss": 0.0903, "step": 4960 }, { "epoch": 0.23, "learning_rate": 2.9628820620036508e-05, "loss": 0.2495, "step": 4965 }, { "epoch": 0.23, "learning_rate": 2.962835255020049e-05, "loss": 0.3487, "step": 4970 }, { "epoch": 0.23, "learning_rate": 2.962788448036447e-05, "loss": 0.2783, "step": 4975 }, { "epoch": 0.23, "learning_rate": 2.962741641052845e-05, "loss": 0.3847, "step": 4980 }, { "epoch": 0.23, "learning_rate": 2.9626948340692434e-05, "loss": 0.3038, "step": 4985 }, { "epoch": 0.23, "learning_rate": 2.9626480270856414e-05, "loss": 0.5606, "step": 4990 }, { "epoch": 0.23, "learning_rate": 2.9626012201020393e-05, "loss": 0.4234, "step": 4995 }, { "epoch": 0.23, "learning_rate": 2.9625544131184373e-05, "loss": 0.8685, "step": 5000 }, { "epoch": 0.23, "learning_rate": 2.9625076061348356e-05, "loss": 0.3391, "step": 5005 }, { "epoch": 0.23, "learning_rate": 2.9624607991512336e-05, "loss": 0.2322, "step": 5010 }, { "epoch": 0.23, "learning_rate": 2.9624139921676313e-05, "loss": 0.2068, "step": 5015 }, { "epoch": 0.23, "learning_rate": 2.9623671851840293e-05, "loss": 0.257, "step": 5020 }, { "epoch": 0.23, "learning_rate": 2.9623203782004276e-05, "loss": 0.3483, "step": 5025 }, { "epoch": 0.23, "learning_rate": 2.9622735712168256e-05, "loss": 0.1754, "step": 5030 }, { "epoch": 0.23, "learning_rate": 2.9622267642332235e-05, "loss": 0.4093, "step": 5035 }, { "epoch": 0.24, "learning_rate": 2.962179957249622e-05, "loss": 0.3109, "step": 5040 }, { "epoch": 0.24, "learning_rate": 2.96213315026602e-05, "loss": 0.4415, "step": 5045 }, { "epoch": 0.24, "learning_rate": 2.9620863432824178e-05, "loss": 0.8288, "step": 5050 }, { "epoch": 0.24, "learning_rate": 2.9620395362988158e-05, "loss": 0.3313, "step": 5055 }, { "epoch": 0.24, "learning_rate": 2.961992729315214e-05, "loss": 0.1508, "step": 5060 }, { "epoch": 0.24, "learning_rate": 2.961945922331612e-05, "loss": 0.1846, "step": 5065 }, { "epoch": 0.24, "learning_rate": 2.96189911534801e-05, "loss": 0.1819, "step": 5070 }, { "epoch": 0.24, "learning_rate": 2.9618523083644077e-05, "loss": 0.255, "step": 5075 }, { "epoch": 0.24, "learning_rate": 2.961805501380806e-05, "loss": 0.2447, "step": 5080 }, { "epoch": 0.24, "learning_rate": 2.961758694397204e-05, "loss": 0.3081, "step": 5085 }, { "epoch": 0.24, "learning_rate": 2.961711887413602e-05, "loss": 0.2883, "step": 5090 }, { "epoch": 0.24, "learning_rate": 2.9616650804300003e-05, "loss": 0.5553, "step": 5095 }, { "epoch": 0.24, "learning_rate": 2.9616182734463983e-05, "loss": 0.5633, "step": 5100 }, { "epoch": 0.24, "learning_rate": 2.9615714664627963e-05, "loss": 0.3469, "step": 5105 }, { "epoch": 0.24, "learning_rate": 2.9615246594791943e-05, "loss": 0.0976, "step": 5110 }, { "epoch": 0.24, "learning_rate": 2.9614778524955926e-05, "loss": 0.1514, "step": 5115 }, { "epoch": 0.24, "learning_rate": 2.9614310455119906e-05, "loss": 0.2376, "step": 5120 }, { "epoch": 0.24, "learning_rate": 2.9613842385283886e-05, "loss": 0.2272, "step": 5125 }, { "epoch": 0.24, "learning_rate": 2.9613374315447865e-05, "loss": 0.2956, "step": 5130 }, { "epoch": 0.24, "learning_rate": 2.961290624561185e-05, "loss": 0.505, "step": 5135 }, { "epoch": 0.24, "learning_rate": 2.9612438175775825e-05, "loss": 0.279, "step": 5140 }, { "epoch": 0.24, "learning_rate": 2.9611970105939805e-05, "loss": 0.5588, "step": 5145 }, { "epoch": 0.24, "learning_rate": 2.9611502036103785e-05, "loss": 0.7466, "step": 5150 }, { "epoch": 0.24, "learning_rate": 2.9611033966267768e-05, "loss": 0.3743, "step": 5155 }, { "epoch": 0.24, "learning_rate": 2.9610565896431748e-05, "loss": 0.1645, "step": 5160 }, { "epoch": 0.24, "learning_rate": 2.9610097826595728e-05, "loss": 0.145, "step": 5165 }, { "epoch": 0.24, "learning_rate": 2.960962975675971e-05, "loss": 0.1843, "step": 5170 }, { "epoch": 0.24, "learning_rate": 2.960916168692369e-05, "loss": 0.225, "step": 5175 }, { "epoch": 0.24, "learning_rate": 2.960869361708767e-05, "loss": 0.2798, "step": 5180 }, { "epoch": 0.24, "learning_rate": 2.960822554725165e-05, "loss": 0.3423, "step": 5185 }, { "epoch": 0.24, "learning_rate": 2.9607757477415633e-05, "loss": 0.3232, "step": 5190 }, { "epoch": 0.24, "learning_rate": 2.9607289407579613e-05, "loss": 0.3699, "step": 5195 }, { "epoch": 0.24, "learning_rate": 2.9606821337743593e-05, "loss": 0.7137, "step": 5200 }, { "epoch": 0.24, "learning_rate": 2.960635326790757e-05, "loss": 0.3644, "step": 5205 }, { "epoch": 0.24, "learning_rate": 2.9605885198071553e-05, "loss": 0.1612, "step": 5210 }, { "epoch": 0.24, "learning_rate": 2.9605417128235533e-05, "loss": 0.2446, "step": 5215 }, { "epoch": 0.24, "learning_rate": 2.9604949058399512e-05, "loss": 0.219, "step": 5220 }, { "epoch": 0.24, "learning_rate": 2.9604480988563496e-05, "loss": 0.2171, "step": 5225 }, { "epoch": 0.24, "learning_rate": 2.9604012918727475e-05, "loss": 0.2166, "step": 5230 }, { "epoch": 0.24, "learning_rate": 2.9603544848891455e-05, "loss": 0.3706, "step": 5235 }, { "epoch": 0.24, "learning_rate": 2.9603076779055435e-05, "loss": 0.3719, "step": 5240 }, { "epoch": 0.24, "learning_rate": 2.9602608709219418e-05, "loss": 0.6137, "step": 5245 }, { "epoch": 0.24, "learning_rate": 2.9602140639383398e-05, "loss": 0.6093, "step": 5250 }, { "epoch": 0.25, "learning_rate": 2.9601672569547378e-05, "loss": 0.3343, "step": 5255 }, { "epoch": 0.25, "learning_rate": 2.9601204499711358e-05, "loss": 0.1317, "step": 5260 }, { "epoch": 0.25, "learning_rate": 2.9600736429875337e-05, "loss": 0.1908, "step": 5265 }, { "epoch": 0.25, "learning_rate": 2.9600268360039317e-05, "loss": 0.291, "step": 5270 }, { "epoch": 0.25, "learning_rate": 2.9599800290203297e-05, "loss": 0.2974, "step": 5275 }, { "epoch": 0.25, "learning_rate": 2.959933222036728e-05, "loss": 0.2858, "step": 5280 }, { "epoch": 0.25, "learning_rate": 2.959886415053126e-05, "loss": 0.3804, "step": 5285 }, { "epoch": 0.25, "learning_rate": 2.959839608069524e-05, "loss": 0.3363, "step": 5290 }, { "epoch": 0.25, "learning_rate": 2.959792801085922e-05, "loss": 0.4221, "step": 5295 }, { "epoch": 0.25, "learning_rate": 2.9597459941023203e-05, "loss": 0.584, "step": 5300 }, { "epoch": 0.25, "learning_rate": 2.9596991871187183e-05, "loss": 0.2908, "step": 5305 }, { "epoch": 0.25, "learning_rate": 2.9596523801351163e-05, "loss": 0.1704, "step": 5310 }, { "epoch": 0.25, "learning_rate": 2.9596055731515142e-05, "loss": 0.1326, "step": 5315 }, { "epoch": 0.25, "learning_rate": 2.9595587661679126e-05, "loss": 0.2063, "step": 5320 }, { "epoch": 0.25, "learning_rate": 2.9595119591843105e-05, "loss": 0.2612, "step": 5325 }, { "epoch": 0.25, "learning_rate": 2.9594651522007082e-05, "loss": 0.2536, "step": 5330 }, { "epoch": 0.25, "learning_rate": 2.9594183452171062e-05, "loss": 0.3799, "step": 5335 }, { "epoch": 0.25, "learning_rate": 2.9593715382335045e-05, "loss": 0.3562, "step": 5340 }, { "epoch": 0.25, "learning_rate": 2.9593247312499025e-05, "loss": 0.4044, "step": 5345 }, { "epoch": 0.25, "learning_rate": 2.9592779242663005e-05, "loss": 0.8405, "step": 5350 }, { "epoch": 0.25, "learning_rate": 2.9592311172826988e-05, "loss": 0.258, "step": 5355 }, { "epoch": 0.25, "learning_rate": 2.9591843102990968e-05, "loss": 0.134, "step": 5360 }, { "epoch": 0.25, "learning_rate": 2.9591375033154947e-05, "loss": 0.1787, "step": 5365 }, { "epoch": 0.25, "learning_rate": 2.9590906963318927e-05, "loss": 0.227, "step": 5370 }, { "epoch": 0.25, "learning_rate": 2.959043889348291e-05, "loss": 0.1671, "step": 5375 }, { "epoch": 0.25, "learning_rate": 2.958997082364689e-05, "loss": 0.1537, "step": 5380 }, { "epoch": 0.25, "learning_rate": 2.958950275381087e-05, "loss": 0.3416, "step": 5385 }, { "epoch": 0.25, "learning_rate": 2.958903468397485e-05, "loss": 0.4624, "step": 5390 }, { "epoch": 0.25, "learning_rate": 2.958856661413883e-05, "loss": 0.4494, "step": 5395 }, { "epoch": 0.25, "learning_rate": 2.958809854430281e-05, "loss": 0.4924, "step": 5400 }, { "epoch": 0.25, "learning_rate": 2.958763047446679e-05, "loss": 0.3339, "step": 5405 }, { "epoch": 0.25, "learning_rate": 2.9587162404630773e-05, "loss": 0.1303, "step": 5410 }, { "epoch": 0.25, "learning_rate": 2.9586694334794752e-05, "loss": 0.1634, "step": 5415 }, { "epoch": 0.25, "learning_rate": 2.9586226264958732e-05, "loss": 0.1576, "step": 5420 }, { "epoch": 0.25, "learning_rate": 2.9585758195122712e-05, "loss": 0.1873, "step": 5425 }, { "epoch": 0.25, "learning_rate": 2.9585290125286695e-05, "loss": 0.2938, "step": 5430 }, { "epoch": 0.25, "learning_rate": 2.9584822055450675e-05, "loss": 0.4097, "step": 5435 }, { "epoch": 0.25, "learning_rate": 2.9584353985614655e-05, "loss": 0.2995, "step": 5440 }, { "epoch": 0.25, "learning_rate": 2.9583885915778635e-05, "loss": 0.6294, "step": 5445 }, { "epoch": 0.25, "learning_rate": 2.9583417845942618e-05, "loss": 0.7594, "step": 5450 }, { "epoch": 0.25, "learning_rate": 2.9582949776106594e-05, "loss": 0.3177, "step": 5455 }, { "epoch": 0.25, "learning_rate": 2.9582481706270574e-05, "loss": 0.1492, "step": 5460 }, { "epoch": 0.26, "learning_rate": 2.9582013636434557e-05, "loss": 0.1839, "step": 5465 }, { "epoch": 0.26, "learning_rate": 2.9581545566598537e-05, "loss": 0.183, "step": 5470 }, { "epoch": 0.26, "learning_rate": 2.9581077496762517e-05, "loss": 0.251, "step": 5475 }, { "epoch": 0.26, "learning_rate": 2.9580609426926497e-05, "loss": 0.3105, "step": 5480 }, { "epoch": 0.26, "learning_rate": 2.958014135709048e-05, "loss": 0.3272, "step": 5485 }, { "epoch": 0.26, "learning_rate": 2.957967328725446e-05, "loss": 0.3467, "step": 5490 }, { "epoch": 0.26, "learning_rate": 2.957920521741844e-05, "loss": 0.4998, "step": 5495 }, { "epoch": 0.26, "learning_rate": 2.957873714758242e-05, "loss": 0.5874, "step": 5500 }, { "epoch": 0.26, "learning_rate": 2.9578269077746403e-05, "loss": 0.4178, "step": 5505 }, { "epoch": 0.26, "learning_rate": 2.9577801007910382e-05, "loss": 0.0913, "step": 5510 }, { "epoch": 0.26, "learning_rate": 2.9577332938074362e-05, "loss": 0.1883, "step": 5515 }, { "epoch": 0.26, "learning_rate": 2.9576864868238342e-05, "loss": 0.2468, "step": 5520 }, { "epoch": 0.26, "learning_rate": 2.9576396798402322e-05, "loss": 0.1948, "step": 5525 }, { "epoch": 0.26, "learning_rate": 2.9575928728566302e-05, "loss": 0.203, "step": 5530 }, { "epoch": 0.26, "learning_rate": 2.957546065873028e-05, "loss": 0.2808, "step": 5535 }, { "epoch": 0.26, "learning_rate": 2.9574992588894265e-05, "loss": 0.3499, "step": 5540 }, { "epoch": 0.26, "learning_rate": 2.9574524519058245e-05, "loss": 0.4629, "step": 5545 }, { "epoch": 0.26, "learning_rate": 2.9574056449222224e-05, "loss": 0.6269, "step": 5550 }, { "epoch": 0.26, "learning_rate": 2.9573588379386204e-05, "loss": 0.3445, "step": 5555 }, { "epoch": 0.26, "learning_rate": 2.9573120309550187e-05, "loss": 0.1888, "step": 5560 }, { "epoch": 0.26, "learning_rate": 2.9572652239714167e-05, "loss": 0.2382, "step": 5565 }, { "epoch": 0.26, "learning_rate": 2.9572184169878147e-05, "loss": 0.2049, "step": 5570 }, { "epoch": 0.26, "learning_rate": 2.9571716100042127e-05, "loss": 0.2395, "step": 5575 }, { "epoch": 0.26, "learning_rate": 2.9571248030206107e-05, "loss": 0.2236, "step": 5580 }, { "epoch": 0.26, "learning_rate": 2.9570779960370086e-05, "loss": 0.2339, "step": 5585 }, { "epoch": 0.26, "learning_rate": 2.9570311890534066e-05, "loss": 0.2487, "step": 5590 }, { "epoch": 0.26, "learning_rate": 2.956984382069805e-05, "loss": 0.3833, "step": 5595 }, { "epoch": 0.26, "learning_rate": 2.956937575086203e-05, "loss": 0.5892, "step": 5600 }, { "epoch": 0.26, "learning_rate": 2.956890768102601e-05, "loss": 0.3309, "step": 5605 }, { "epoch": 0.26, "learning_rate": 2.956843961118999e-05, "loss": 0.1288, "step": 5610 }, { "epoch": 0.26, "learning_rate": 2.9567971541353972e-05, "loss": 0.1725, "step": 5615 }, { "epoch": 0.26, "learning_rate": 2.9567503471517952e-05, "loss": 0.2006, "step": 5620 }, { "epoch": 0.26, "learning_rate": 2.9567035401681932e-05, "loss": 0.2777, "step": 5625 }, { "epoch": 0.26, "learning_rate": 2.956656733184591e-05, "loss": 0.2572, "step": 5630 }, { "epoch": 0.26, "learning_rate": 2.9566099262009895e-05, "loss": 0.3117, "step": 5635 }, { "epoch": 0.26, "learning_rate": 2.9565631192173875e-05, "loss": 0.2662, "step": 5640 }, { "epoch": 0.26, "learning_rate": 2.956516312233785e-05, "loss": 0.4618, "step": 5645 }, { "epoch": 0.26, "learning_rate": 2.9564695052501834e-05, "loss": 0.638, "step": 5650 }, { "epoch": 0.26, "learning_rate": 2.9564226982665814e-05, "loss": 0.3631, "step": 5655 }, { "epoch": 0.26, "learning_rate": 2.9563758912829794e-05, "loss": 0.1522, "step": 5660 }, { "epoch": 0.26, "learning_rate": 2.9563290842993774e-05, "loss": 0.2193, "step": 5665 }, { "epoch": 0.26, "learning_rate": 2.9562822773157757e-05, "loss": 0.2372, "step": 5670 }, { "epoch": 0.26, "learning_rate": 2.9562354703321737e-05, "loss": 0.2513, "step": 5675 }, { "epoch": 0.27, "learning_rate": 2.9561886633485717e-05, "loss": 0.1802, "step": 5680 }, { "epoch": 0.27, "learning_rate": 2.9561418563649696e-05, "loss": 0.2209, "step": 5685 }, { "epoch": 0.27, "learning_rate": 2.956095049381368e-05, "loss": 0.3681, "step": 5690 }, { "epoch": 0.27, "learning_rate": 2.956048242397766e-05, "loss": 0.4923, "step": 5695 }, { "epoch": 0.27, "learning_rate": 2.956001435414164e-05, "loss": 0.5072, "step": 5700 }, { "epoch": 0.27, "learning_rate": 2.9559546284305622e-05, "loss": 0.3223, "step": 5705 }, { "epoch": 0.27, "learning_rate": 2.95590782144696e-05, "loss": 0.1211, "step": 5710 }, { "epoch": 0.27, "learning_rate": 2.955861014463358e-05, "loss": 0.2507, "step": 5715 }, { "epoch": 0.27, "learning_rate": 2.955814207479756e-05, "loss": 0.1866, "step": 5720 }, { "epoch": 0.27, "learning_rate": 2.9557674004961542e-05, "loss": 0.2197, "step": 5725 }, { "epoch": 0.27, "learning_rate": 2.955720593512552e-05, "loss": 0.2306, "step": 5730 }, { "epoch": 0.27, "learning_rate": 2.95567378652895e-05, "loss": 0.2568, "step": 5735 }, { "epoch": 0.27, "learning_rate": 2.955626979545348e-05, "loss": 0.2844, "step": 5740 }, { "epoch": 0.27, "learning_rate": 2.9555801725617464e-05, "loss": 0.346, "step": 5745 }, { "epoch": 0.27, "learning_rate": 2.9555333655781444e-05, "loss": 0.545, "step": 5750 }, { "epoch": 0.27, "learning_rate": 2.9554865585945424e-05, "loss": 0.3282, "step": 5755 }, { "epoch": 0.27, "learning_rate": 2.9554397516109404e-05, "loss": 0.1594, "step": 5760 }, { "epoch": 0.27, "learning_rate": 2.9553929446273387e-05, "loss": 0.1354, "step": 5765 }, { "epoch": 0.27, "learning_rate": 2.9553461376437363e-05, "loss": 0.2147, "step": 5770 }, { "epoch": 0.27, "learning_rate": 2.9552993306601343e-05, "loss": 0.2238, "step": 5775 }, { "epoch": 0.27, "learning_rate": 2.9552525236765326e-05, "loss": 0.2884, "step": 5780 }, { "epoch": 0.27, "learning_rate": 2.9552057166929306e-05, "loss": 0.1929, "step": 5785 }, { "epoch": 0.27, "learning_rate": 2.9551589097093286e-05, "loss": 0.3401, "step": 5790 }, { "epoch": 0.27, "learning_rate": 2.9551121027257266e-05, "loss": 0.5252, "step": 5795 }, { "epoch": 0.27, "learning_rate": 2.955065295742125e-05, "loss": 0.7707, "step": 5800 }, { "epoch": 0.27, "learning_rate": 2.955018488758523e-05, "loss": 0.3297, "step": 5805 }, { "epoch": 0.27, "learning_rate": 2.954971681774921e-05, "loss": 0.1562, "step": 5810 }, { "epoch": 0.27, "learning_rate": 2.954924874791319e-05, "loss": 0.1504, "step": 5815 }, { "epoch": 0.27, "learning_rate": 2.9548780678077172e-05, "loss": 0.1697, "step": 5820 }, { "epoch": 0.27, "learning_rate": 2.954831260824115e-05, "loss": 0.3208, "step": 5825 }, { "epoch": 0.27, "learning_rate": 2.954784453840513e-05, "loss": 0.2643, "step": 5830 }, { "epoch": 0.27, "learning_rate": 2.954737646856911e-05, "loss": 0.3444, "step": 5835 }, { "epoch": 0.27, "learning_rate": 2.954690839873309e-05, "loss": 0.3433, "step": 5840 }, { "epoch": 0.27, "learning_rate": 2.954644032889707e-05, "loss": 0.42, "step": 5845 }, { "epoch": 0.27, "learning_rate": 2.954597225906105e-05, "loss": 0.5641, "step": 5850 }, { "epoch": 0.27, "learning_rate": 2.9545504189225034e-05, "loss": 0.2726, "step": 5855 }, { "epoch": 0.27, "learning_rate": 2.9545036119389014e-05, "loss": 0.1108, "step": 5860 }, { "epoch": 0.27, "learning_rate": 2.9544568049552994e-05, "loss": 0.2075, "step": 5865 }, { "epoch": 0.27, "learning_rate": 2.9544099979716973e-05, "loss": 0.1593, "step": 5870 }, { "epoch": 0.27, "learning_rate": 2.9543631909880957e-05, "loss": 0.2366, "step": 5875 }, { "epoch": 0.27, "learning_rate": 2.9543163840044936e-05, "loss": 0.3362, "step": 5880 }, { "epoch": 0.27, "learning_rate": 2.9542695770208916e-05, "loss": 0.3316, "step": 5885 }, { "epoch": 0.27, "learning_rate": 2.95422277003729e-05, "loss": 0.3774, "step": 5890 }, { "epoch": 0.28, "learning_rate": 2.954175963053688e-05, "loss": 0.4141, "step": 5895 }, { "epoch": 0.28, "learning_rate": 2.9541291560700856e-05, "loss": 0.7402, "step": 5900 }, { "epoch": 0.28, "learning_rate": 2.9540823490864835e-05, "loss": 0.3221, "step": 5905 }, { "epoch": 0.28, "learning_rate": 2.954035542102882e-05, "loss": 0.1264, "step": 5910 }, { "epoch": 0.28, "learning_rate": 2.95398873511928e-05, "loss": 0.1677, "step": 5915 }, { "epoch": 0.28, "learning_rate": 2.953941928135678e-05, "loss": 0.2109, "step": 5920 }, { "epoch": 0.28, "learning_rate": 2.9538951211520758e-05, "loss": 0.2547, "step": 5925 }, { "epoch": 0.28, "learning_rate": 2.953848314168474e-05, "loss": 0.2995, "step": 5930 }, { "epoch": 0.28, "learning_rate": 2.953801507184872e-05, "loss": 0.3493, "step": 5935 }, { "epoch": 0.28, "learning_rate": 2.95375470020127e-05, "loss": 0.367, "step": 5940 }, { "epoch": 0.28, "learning_rate": 2.953707893217668e-05, "loss": 0.5445, "step": 5945 }, { "epoch": 0.28, "learning_rate": 2.9536610862340664e-05, "loss": 0.6944, "step": 5950 }, { "epoch": 0.28, "learning_rate": 2.9536142792504644e-05, "loss": 0.3739, "step": 5955 }, { "epoch": 0.28, "learning_rate": 2.953567472266862e-05, "loss": 0.1107, "step": 5960 }, { "epoch": 0.28, "learning_rate": 2.9535206652832603e-05, "loss": 0.2579, "step": 5965 }, { "epoch": 0.28, "learning_rate": 2.9534738582996583e-05, "loss": 0.1703, "step": 5970 }, { "epoch": 0.28, "learning_rate": 2.9534270513160563e-05, "loss": 0.2861, "step": 5975 }, { "epoch": 0.28, "learning_rate": 2.9533802443324543e-05, "loss": 0.3869, "step": 5980 }, { "epoch": 0.28, "learning_rate": 2.9533334373488526e-05, "loss": 0.2519, "step": 5985 }, { "epoch": 0.28, "learning_rate": 2.9532866303652506e-05, "loss": 0.3323, "step": 5990 }, { "epoch": 0.28, "learning_rate": 2.9532398233816486e-05, "loss": 0.4471, "step": 5995 }, { "epoch": 0.28, "learning_rate": 2.9531930163980466e-05, "loss": 0.6038, "step": 6000 }, { "epoch": 0.28, "learning_rate": 2.953146209414445e-05, "loss": 0.3086, "step": 6005 }, { "epoch": 0.28, "learning_rate": 2.953099402430843e-05, "loss": 0.214, "step": 6010 }, { "epoch": 0.28, "learning_rate": 2.953052595447241e-05, "loss": 0.198, "step": 6015 }, { "epoch": 0.28, "learning_rate": 2.953005788463639e-05, "loss": 0.1601, "step": 6020 }, { "epoch": 0.28, "learning_rate": 2.9529589814800368e-05, "loss": 0.1962, "step": 6025 }, { "epoch": 0.28, "learning_rate": 2.9529121744964348e-05, "loss": 0.3013, "step": 6030 }, { "epoch": 0.28, "learning_rate": 2.9528653675128328e-05, "loss": 0.4289, "step": 6035 }, { "epoch": 0.28, "learning_rate": 2.952818560529231e-05, "loss": 0.2674, "step": 6040 }, { "epoch": 0.28, "learning_rate": 2.952771753545629e-05, "loss": 0.4578, "step": 6045 }, { "epoch": 0.28, "learning_rate": 2.952724946562027e-05, "loss": 0.7972, "step": 6050 }, { "epoch": 0.28, "learning_rate": 2.952678139578425e-05, "loss": 0.3369, "step": 6055 }, { "epoch": 0.28, "learning_rate": 2.9526313325948234e-05, "loss": 0.1701, "step": 6060 }, { "epoch": 0.28, "learning_rate": 2.9525845256112213e-05, "loss": 0.228, "step": 6065 }, { "epoch": 0.28, "learning_rate": 2.9525377186276193e-05, "loss": 0.2577, "step": 6070 }, { "epoch": 0.28, "learning_rate": 2.9524909116440176e-05, "loss": 0.2919, "step": 6075 }, { "epoch": 0.28, "learning_rate": 2.9524441046604156e-05, "loss": 0.1989, "step": 6080 }, { "epoch": 0.28, "learning_rate": 2.9523972976768136e-05, "loss": 0.3102, "step": 6085 }, { "epoch": 0.28, "learning_rate": 2.9523504906932112e-05, "loss": 0.413, "step": 6090 }, { "epoch": 0.28, "learning_rate": 2.9523036837096096e-05, "loss": 0.3839, "step": 6095 }, { "epoch": 0.28, "learning_rate": 2.9522568767260075e-05, "loss": 0.751, "step": 6100 }, { "epoch": 0.28, "learning_rate": 2.9522100697424055e-05, "loss": 0.2672, "step": 6105 }, { "epoch": 0.29, "learning_rate": 2.9521632627588035e-05, "loss": 0.1398, "step": 6110 }, { "epoch": 0.29, "learning_rate": 2.952116455775202e-05, "loss": 0.2228, "step": 6115 }, { "epoch": 0.29, "learning_rate": 2.9520696487915998e-05, "loss": 0.1747, "step": 6120 }, { "epoch": 0.29, "learning_rate": 2.9520228418079978e-05, "loss": 0.2612, "step": 6125 }, { "epoch": 0.29, "learning_rate": 2.9519760348243958e-05, "loss": 0.261, "step": 6130 }, { "epoch": 0.29, "learning_rate": 2.951929227840794e-05, "loss": 0.3475, "step": 6135 }, { "epoch": 0.29, "learning_rate": 2.951882420857192e-05, "loss": 0.3063, "step": 6140 }, { "epoch": 0.29, "learning_rate": 2.95183561387359e-05, "loss": 0.524, "step": 6145 }, { "epoch": 0.29, "learning_rate": 2.951788806889988e-05, "loss": 0.7232, "step": 6150 }, { "epoch": 0.29, "learning_rate": 2.951741999906386e-05, "loss": 0.2241, "step": 6155 }, { "epoch": 0.29, "learning_rate": 2.951695192922784e-05, "loss": 0.1908, "step": 6160 }, { "epoch": 0.29, "learning_rate": 2.951648385939182e-05, "loss": 0.1571, "step": 6165 }, { "epoch": 0.29, "learning_rate": 2.9516015789555803e-05, "loss": 0.2126, "step": 6170 }, { "epoch": 0.29, "learning_rate": 2.9515547719719783e-05, "loss": 0.2266, "step": 6175 }, { "epoch": 0.29, "learning_rate": 2.9515079649883763e-05, "loss": 0.3131, "step": 6180 }, { "epoch": 0.29, "learning_rate": 2.9514611580047743e-05, "loss": 0.2176, "step": 6185 }, { "epoch": 0.29, "learning_rate": 2.9514143510211726e-05, "loss": 0.307, "step": 6190 }, { "epoch": 0.29, "learning_rate": 2.9513675440375706e-05, "loss": 0.4321, "step": 6195 }, { "epoch": 0.29, "learning_rate": 2.9513207370539685e-05, "loss": 0.6908, "step": 6200 }, { "epoch": 0.29, "learning_rate": 2.951273930070367e-05, "loss": 0.3091, "step": 6205 }, { "epoch": 0.29, "learning_rate": 2.951227123086765e-05, "loss": 0.1268, "step": 6210 }, { "epoch": 0.29, "learning_rate": 2.9511803161031625e-05, "loss": 0.1892, "step": 6215 }, { "epoch": 0.29, "learning_rate": 2.9511335091195605e-05, "loss": 0.2528, "step": 6220 }, { "epoch": 0.29, "learning_rate": 2.9510867021359588e-05, "loss": 0.2706, "step": 6225 }, { "epoch": 0.29, "learning_rate": 2.9510398951523568e-05, "loss": 0.2767, "step": 6230 }, { "epoch": 0.29, "learning_rate": 2.9509930881687547e-05, "loss": 0.3146, "step": 6235 }, { "epoch": 0.29, "learning_rate": 2.9509462811851527e-05, "loss": 0.3389, "step": 6240 }, { "epoch": 0.29, "learning_rate": 2.950899474201551e-05, "loss": 0.4984, "step": 6245 }, { "epoch": 0.29, "learning_rate": 2.950852667217949e-05, "loss": 0.5109, "step": 6250 }, { "epoch": 0.29, "learning_rate": 2.950805860234347e-05, "loss": 0.2741, "step": 6255 }, { "epoch": 0.29, "learning_rate": 2.9507590532507453e-05, "loss": 0.1101, "step": 6260 }, { "epoch": 0.29, "learning_rate": 2.9507122462671433e-05, "loss": 0.1227, "step": 6265 }, { "epoch": 0.29, "learning_rate": 2.9506654392835413e-05, "loss": 0.2644, "step": 6270 }, { "epoch": 0.29, "learning_rate": 2.9506186322999393e-05, "loss": 0.1347, "step": 6275 }, { "epoch": 0.29, "learning_rate": 2.9505718253163373e-05, "loss": 0.2553, "step": 6280 }, { "epoch": 0.29, "learning_rate": 2.9505250183327352e-05, "loss": 0.2612, "step": 6285 }, { "epoch": 0.29, "learning_rate": 2.9504782113491332e-05, "loss": 0.4636, "step": 6290 }, { "epoch": 0.29, "learning_rate": 2.9504314043655312e-05, "loss": 0.3807, "step": 6295 }, { "epoch": 0.29, "learning_rate": 2.9503845973819295e-05, "loss": 0.8262, "step": 6300 }, { "epoch": 0.29, "learning_rate": 2.9503377903983275e-05, "loss": 0.2826, "step": 6305 }, { "epoch": 0.29, "learning_rate": 2.9502909834147255e-05, "loss": 0.1274, "step": 6310 }, { "epoch": 0.29, "learning_rate": 2.9502441764311238e-05, "loss": 0.177, "step": 6315 }, { "epoch": 0.29, "learning_rate": 2.9501973694475218e-05, "loss": 0.2008, "step": 6320 }, { "epoch": 0.3, "learning_rate": 2.9501505624639198e-05, "loss": 0.2339, "step": 6325 }, { "epoch": 0.3, "learning_rate": 2.9501037554803178e-05, "loss": 0.2365, "step": 6330 }, { "epoch": 0.3, "learning_rate": 2.950056948496716e-05, "loss": 0.2761, "step": 6335 }, { "epoch": 0.3, "learning_rate": 2.9500101415131137e-05, "loss": 0.3615, "step": 6340 }, { "epoch": 0.3, "learning_rate": 2.9499633345295117e-05, "loss": 0.4044, "step": 6345 }, { "epoch": 0.3, "learning_rate": 2.9499165275459097e-05, "loss": 0.6302, "step": 6350 }, { "epoch": 0.3, "learning_rate": 2.949869720562308e-05, "loss": 0.3552, "step": 6355 }, { "epoch": 0.3, "learning_rate": 2.949822913578706e-05, "loss": 0.1164, "step": 6360 }, { "epoch": 0.3, "learning_rate": 2.949776106595104e-05, "loss": 0.1787, "step": 6365 }, { "epoch": 0.3, "learning_rate": 2.949729299611502e-05, "loss": 0.1635, "step": 6370 }, { "epoch": 0.3, "learning_rate": 2.9496824926279003e-05, "loss": 0.2198, "step": 6375 }, { "epoch": 0.3, "learning_rate": 2.9496356856442983e-05, "loss": 0.1736, "step": 6380 }, { "epoch": 0.3, "learning_rate": 2.9495888786606962e-05, "loss": 0.1955, "step": 6385 }, { "epoch": 0.3, "learning_rate": 2.9495420716770946e-05, "loss": 0.4086, "step": 6390 }, { "epoch": 0.3, "learning_rate": 2.9494952646934925e-05, "loss": 0.3732, "step": 6395 }, { "epoch": 0.3, "learning_rate": 2.9494484577098905e-05, "loss": 0.581, "step": 6400 }, { "epoch": 0.3, "learning_rate": 2.949401650726288e-05, "loss": 0.2786, "step": 6405 }, { "epoch": 0.3, "learning_rate": 2.9493548437426865e-05, "loss": 0.1187, "step": 6410 }, { "epoch": 0.3, "learning_rate": 2.9493080367590845e-05, "loss": 0.1062, "step": 6415 }, { "epoch": 0.3, "learning_rate": 2.9492612297754824e-05, "loss": 0.1636, "step": 6420 }, { "epoch": 0.3, "learning_rate": 2.9492144227918804e-05, "loss": 0.2069, "step": 6425 }, { "epoch": 0.3, "learning_rate": 2.9491676158082787e-05, "loss": 0.2534, "step": 6430 }, { "epoch": 0.3, "learning_rate": 2.9491208088246767e-05, "loss": 0.2974, "step": 6435 }, { "epoch": 0.3, "learning_rate": 2.9490740018410747e-05, "loss": 0.3632, "step": 6440 }, { "epoch": 0.3, "learning_rate": 2.949027194857473e-05, "loss": 0.4625, "step": 6445 }, { "epoch": 0.3, "learning_rate": 2.948980387873871e-05, "loss": 0.4981, "step": 6450 }, { "epoch": 0.3, "learning_rate": 2.948933580890269e-05, "loss": 0.307, "step": 6455 }, { "epoch": 0.3, "learning_rate": 2.948886773906667e-05, "loss": 0.1518, "step": 6460 }, { "epoch": 0.3, "learning_rate": 2.948839966923065e-05, "loss": 0.1984, "step": 6465 }, { "epoch": 0.3, "learning_rate": 2.948793159939463e-05, "loss": 0.2891, "step": 6470 }, { "epoch": 0.3, "learning_rate": 2.948746352955861e-05, "loss": 0.2248, "step": 6475 }, { "epoch": 0.3, "learning_rate": 2.948699545972259e-05, "loss": 0.2901, "step": 6480 }, { "epoch": 0.3, "learning_rate": 2.9486527389886572e-05, "loss": 0.33, "step": 6485 }, { "epoch": 0.3, "learning_rate": 2.9486059320050552e-05, "loss": 0.2767, "step": 6490 }, { "epoch": 0.3, "learning_rate": 2.9485591250214532e-05, "loss": 0.3612, "step": 6495 }, { "epoch": 0.3, "learning_rate": 2.9485123180378515e-05, "loss": 0.644, "step": 6500 }, { "epoch": 0.3, "learning_rate": 2.9484655110542495e-05, "loss": 0.3238, "step": 6505 }, { "epoch": 0.3, "learning_rate": 2.9484187040706475e-05, "loss": 0.1715, "step": 6510 }, { "epoch": 0.3, "learning_rate": 2.9483718970870455e-05, "loss": 0.1669, "step": 6515 }, { "epoch": 0.3, "learning_rate": 2.9483250901034438e-05, "loss": 0.2612, "step": 6520 }, { "epoch": 0.3, "learning_rate": 2.9482782831198418e-05, "loss": 0.2686, "step": 6525 }, { "epoch": 0.3, "learning_rate": 2.9482314761362394e-05, "loss": 0.2823, "step": 6530 }, { "epoch": 0.3, "learning_rate": 2.9481846691526374e-05, "loss": 0.3895, "step": 6535 }, { "epoch": 0.31, "learning_rate": 2.9481378621690357e-05, "loss": 0.3165, "step": 6540 }, { "epoch": 0.31, "learning_rate": 2.9480910551854337e-05, "loss": 0.439, "step": 6545 }, { "epoch": 0.31, "learning_rate": 2.9480442482018317e-05, "loss": 0.6556, "step": 6550 }, { "epoch": 0.31, "learning_rate": 2.9479974412182296e-05, "loss": 0.3243, "step": 6555 }, { "epoch": 0.31, "learning_rate": 2.947950634234628e-05, "loss": 0.1045, "step": 6560 }, { "epoch": 0.31, "learning_rate": 2.947903827251026e-05, "loss": 0.1191, "step": 6565 }, { "epoch": 0.31, "learning_rate": 2.947857020267424e-05, "loss": 0.1821, "step": 6570 }, { "epoch": 0.31, "learning_rate": 2.9478102132838223e-05, "loss": 0.2167, "step": 6575 }, { "epoch": 0.31, "learning_rate": 2.9477634063002202e-05, "loss": 0.2829, "step": 6580 }, { "epoch": 0.31, "learning_rate": 2.9477165993166182e-05, "loss": 0.3564, "step": 6585 }, { "epoch": 0.31, "learning_rate": 2.9476697923330162e-05, "loss": 0.5041, "step": 6590 }, { "epoch": 0.31, "learning_rate": 2.9476229853494142e-05, "loss": 0.3571, "step": 6595 }, { "epoch": 0.31, "learning_rate": 2.947576178365812e-05, "loss": 0.6696, "step": 6600 }, { "epoch": 0.31, "learning_rate": 2.94752937138221e-05, "loss": 0.2935, "step": 6605 }, { "epoch": 0.31, "learning_rate": 2.947482564398608e-05, "loss": 0.1073, "step": 6610 }, { "epoch": 0.31, "learning_rate": 2.9474357574150064e-05, "loss": 0.1227, "step": 6615 }, { "epoch": 0.31, "learning_rate": 2.9473889504314044e-05, "loss": 0.1901, "step": 6620 }, { "epoch": 0.31, "learning_rate": 2.9473421434478024e-05, "loss": 0.1586, "step": 6625 }, { "epoch": 0.31, "learning_rate": 2.9472953364642007e-05, "loss": 0.2283, "step": 6630 }, { "epoch": 0.31, "learning_rate": 2.9472485294805987e-05, "loss": 0.2978, "step": 6635 }, { "epoch": 0.31, "learning_rate": 2.9472017224969967e-05, "loss": 0.309, "step": 6640 }, { "epoch": 0.31, "learning_rate": 2.9471549155133947e-05, "loss": 0.4962, "step": 6645 }, { "epoch": 0.31, "learning_rate": 2.947108108529793e-05, "loss": 0.639, "step": 6650 }, { "epoch": 0.31, "learning_rate": 2.9470613015461906e-05, "loss": 0.2725, "step": 6655 }, { "epoch": 0.31, "learning_rate": 2.9470144945625886e-05, "loss": 0.1545, "step": 6660 }, { "epoch": 0.31, "learning_rate": 2.9469676875789866e-05, "loss": 0.1472, "step": 6665 }, { "epoch": 0.31, "learning_rate": 2.946920880595385e-05, "loss": 0.1253, "step": 6670 }, { "epoch": 0.31, "learning_rate": 2.946874073611783e-05, "loss": 0.2014, "step": 6675 }, { "epoch": 0.31, "learning_rate": 2.946827266628181e-05, "loss": 0.2666, "step": 6680 }, { "epoch": 0.31, "learning_rate": 2.9467804596445792e-05, "loss": 0.2151, "step": 6685 }, { "epoch": 0.31, "learning_rate": 2.9467336526609772e-05, "loss": 0.3208, "step": 6690 }, { "epoch": 0.31, "learning_rate": 2.9466868456773752e-05, "loss": 0.3027, "step": 6695 }, { "epoch": 0.31, "learning_rate": 2.946640038693773e-05, "loss": 0.6314, "step": 6700 }, { "epoch": 0.31, "learning_rate": 2.9465932317101715e-05, "loss": 0.3624, "step": 6705 }, { "epoch": 0.31, "learning_rate": 2.9465464247265695e-05, "loss": 0.1625, "step": 6710 }, { "epoch": 0.31, "learning_rate": 2.9464996177429674e-05, "loss": 0.1313, "step": 6715 }, { "epoch": 0.31, "learning_rate": 2.946452810759365e-05, "loss": 0.2024, "step": 6720 }, { "epoch": 0.31, "learning_rate": 2.9464060037757634e-05, "loss": 0.1811, "step": 6725 }, { "epoch": 0.31, "learning_rate": 2.9463591967921614e-05, "loss": 0.3074, "step": 6730 }, { "epoch": 0.31, "learning_rate": 2.9463123898085594e-05, "loss": 0.3057, "step": 6735 }, { "epoch": 0.31, "learning_rate": 2.9462655828249573e-05, "loss": 0.4035, "step": 6740 }, { "epoch": 0.31, "learning_rate": 2.9462187758413557e-05, "loss": 0.4331, "step": 6745 }, { "epoch": 0.31, "learning_rate": 2.9461719688577536e-05, "loss": 0.5612, "step": 6750 }, { "epoch": 0.32, "learning_rate": 2.9461251618741516e-05, "loss": 0.2139, "step": 6755 }, { "epoch": 0.32, "learning_rate": 2.94607835489055e-05, "loss": 0.1331, "step": 6760 }, { "epoch": 0.32, "learning_rate": 2.946031547906948e-05, "loss": 0.1882, "step": 6765 }, { "epoch": 0.32, "learning_rate": 2.945984740923346e-05, "loss": 0.2629, "step": 6770 }, { "epoch": 0.32, "learning_rate": 2.945937933939744e-05, "loss": 0.2592, "step": 6775 }, { "epoch": 0.32, "learning_rate": 2.945891126956142e-05, "loss": 0.3825, "step": 6780 }, { "epoch": 0.32, "learning_rate": 2.94584431997254e-05, "loss": 0.2151, "step": 6785 }, { "epoch": 0.32, "learning_rate": 2.945797512988938e-05, "loss": 0.2689, "step": 6790 }, { "epoch": 0.32, "learning_rate": 2.9457507060053358e-05, "loss": 0.5211, "step": 6795 }, { "epoch": 0.32, "learning_rate": 2.945703899021734e-05, "loss": 0.4612, "step": 6800 }, { "epoch": 0.32, "learning_rate": 2.945657092038132e-05, "loss": 0.3253, "step": 6805 }, { "epoch": 0.32, "learning_rate": 2.94561028505453e-05, "loss": 0.08, "step": 6810 }, { "epoch": 0.32, "learning_rate": 2.9455634780709284e-05, "loss": 0.138, "step": 6815 }, { "epoch": 0.32, "learning_rate": 2.9455166710873264e-05, "loss": 0.2012, "step": 6820 }, { "epoch": 0.32, "learning_rate": 2.9454698641037244e-05, "loss": 0.2973, "step": 6825 }, { "epoch": 0.32, "learning_rate": 2.9454230571201224e-05, "loss": 0.3456, "step": 6830 }, { "epoch": 0.32, "learning_rate": 2.9453762501365207e-05, "loss": 0.2108, "step": 6835 }, { "epoch": 0.32, "learning_rate": 2.9453294431529187e-05, "loss": 0.2841, "step": 6840 }, { "epoch": 0.32, "learning_rate": 2.9452826361693163e-05, "loss": 0.4899, "step": 6845 }, { "epoch": 0.32, "learning_rate": 2.9452358291857143e-05, "loss": 0.6438, "step": 6850 }, { "epoch": 0.32, "learning_rate": 2.9451890222021126e-05, "loss": 0.3009, "step": 6855 }, { "epoch": 0.32, "learning_rate": 2.9451422152185106e-05, "loss": 0.1273, "step": 6860 }, { "epoch": 0.32, "learning_rate": 2.9450954082349086e-05, "loss": 0.168, "step": 6865 }, { "epoch": 0.32, "learning_rate": 2.945048601251307e-05, "loss": 0.1879, "step": 6870 }, { "epoch": 0.32, "learning_rate": 2.945001794267705e-05, "loss": 0.2848, "step": 6875 }, { "epoch": 0.32, "learning_rate": 2.944954987284103e-05, "loss": 0.2592, "step": 6880 }, { "epoch": 0.32, "learning_rate": 2.944908180300501e-05, "loss": 0.3403, "step": 6885 }, { "epoch": 0.32, "learning_rate": 2.944861373316899e-05, "loss": 0.4345, "step": 6890 }, { "epoch": 0.32, "learning_rate": 2.944814566333297e-05, "loss": 0.3656, "step": 6895 }, { "epoch": 0.32, "learning_rate": 2.944767759349695e-05, "loss": 0.4829, "step": 6900 }, { "epoch": 0.32, "learning_rate": 2.944720952366093e-05, "loss": 0.4176, "step": 6905 }, { "epoch": 0.32, "learning_rate": 2.944674145382491e-05, "loss": 0.1538, "step": 6910 }, { "epoch": 0.32, "learning_rate": 2.944627338398889e-05, "loss": 0.0787, "step": 6915 }, { "epoch": 0.32, "learning_rate": 2.944580531415287e-05, "loss": 0.159, "step": 6920 }, { "epoch": 0.32, "learning_rate": 2.9445337244316854e-05, "loss": 0.1508, "step": 6925 }, { "epoch": 0.32, "learning_rate": 2.9444869174480834e-05, "loss": 0.2665, "step": 6930 }, { "epoch": 0.32, "learning_rate": 2.9444401104644813e-05, "loss": 0.3905, "step": 6935 }, { "epoch": 0.32, "learning_rate": 2.9443933034808793e-05, "loss": 0.3491, "step": 6940 }, { "epoch": 0.32, "learning_rate": 2.9443464964972776e-05, "loss": 0.4199, "step": 6945 }, { "epoch": 0.32, "learning_rate": 2.9442996895136756e-05, "loss": 0.6787, "step": 6950 }, { "epoch": 0.32, "learning_rate": 2.9442528825300736e-05, "loss": 0.3151, "step": 6955 }, { "epoch": 0.32, "learning_rate": 2.9442060755464716e-05, "loss": 0.1328, "step": 6960 }, { "epoch": 0.32, "learning_rate": 2.94415926856287e-05, "loss": 0.1392, "step": 6965 }, { "epoch": 0.33, "learning_rate": 2.9441124615792676e-05, "loss": 0.1369, "step": 6970 }, { "epoch": 0.33, "learning_rate": 2.9440656545956655e-05, "loss": 0.2611, "step": 6975 }, { "epoch": 0.33, "learning_rate": 2.9440188476120635e-05, "loss": 0.2943, "step": 6980 }, { "epoch": 0.33, "learning_rate": 2.943972040628462e-05, "loss": 0.2458, "step": 6985 }, { "epoch": 0.33, "learning_rate": 2.9439252336448598e-05, "loss": 0.3252, "step": 6990 }, { "epoch": 0.33, "learning_rate": 2.9438784266612578e-05, "loss": 0.4069, "step": 6995 }, { "epoch": 0.33, "learning_rate": 2.943831619677656e-05, "loss": 0.6211, "step": 7000 }, { "epoch": 0.33, "learning_rate": 2.943784812694054e-05, "loss": 0.3158, "step": 7005 }, { "epoch": 0.33, "learning_rate": 2.943738005710452e-05, "loss": 0.1283, "step": 7010 }, { "epoch": 0.33, "learning_rate": 2.94369119872685e-05, "loss": 0.1746, "step": 7015 }, { "epoch": 0.33, "learning_rate": 2.9436443917432484e-05, "loss": 0.2119, "step": 7020 }, { "epoch": 0.33, "learning_rate": 2.9435975847596464e-05, "loss": 0.182, "step": 7025 }, { "epoch": 0.33, "learning_rate": 2.9435507777760444e-05, "loss": 0.2626, "step": 7030 }, { "epoch": 0.33, "learning_rate": 2.943503970792442e-05, "loss": 0.2755, "step": 7035 }, { "epoch": 0.33, "learning_rate": 2.9434571638088403e-05, "loss": 0.2745, "step": 7040 }, { "epoch": 0.33, "learning_rate": 2.9434103568252383e-05, "loss": 0.4722, "step": 7045 }, { "epoch": 0.33, "learning_rate": 2.9433635498416363e-05, "loss": 0.7391, "step": 7050 }, { "epoch": 0.33, "learning_rate": 2.9433167428580346e-05, "loss": 0.3168, "step": 7055 }, { "epoch": 0.33, "learning_rate": 2.9432699358744326e-05, "loss": 0.1312, "step": 7060 }, { "epoch": 0.33, "learning_rate": 2.9432231288908306e-05, "loss": 0.156, "step": 7065 }, { "epoch": 0.33, "learning_rate": 2.9431763219072285e-05, "loss": 0.185, "step": 7070 }, { "epoch": 0.33, "learning_rate": 2.943129514923627e-05, "loss": 0.2077, "step": 7075 }, { "epoch": 0.33, "learning_rate": 2.943082707940025e-05, "loss": 0.22, "step": 7080 }, { "epoch": 0.33, "learning_rate": 2.9430359009564228e-05, "loss": 0.357, "step": 7085 }, { "epoch": 0.33, "learning_rate": 2.9429890939728208e-05, "loss": 0.3937, "step": 7090 }, { "epoch": 0.33, "learning_rate": 2.942942286989219e-05, "loss": 0.3502, "step": 7095 }, { "epoch": 0.33, "learning_rate": 2.9428954800056168e-05, "loss": 0.6648, "step": 7100 }, { "epoch": 0.33, "learning_rate": 2.9428486730220148e-05, "loss": 0.3005, "step": 7105 }, { "epoch": 0.33, "learning_rate": 2.942801866038413e-05, "loss": 0.1217, "step": 7110 }, { "epoch": 0.33, "learning_rate": 2.942755059054811e-05, "loss": 0.1151, "step": 7115 }, { "epoch": 0.33, "learning_rate": 2.942708252071209e-05, "loss": 0.0922, "step": 7120 }, { "epoch": 0.33, "learning_rate": 2.942661445087607e-05, "loss": 0.1953, "step": 7125 }, { "epoch": 0.33, "learning_rate": 2.9426146381040053e-05, "loss": 0.2047, "step": 7130 }, { "epoch": 0.33, "learning_rate": 2.9425678311204033e-05, "loss": 0.2647, "step": 7135 }, { "epoch": 0.33, "learning_rate": 2.9425210241368013e-05, "loss": 0.314, "step": 7140 }, { "epoch": 0.33, "learning_rate": 2.9424742171531993e-05, "loss": 0.4359, "step": 7145 }, { "epoch": 0.33, "learning_rate": 2.9424274101695976e-05, "loss": 0.5848, "step": 7150 }, { "epoch": 0.33, "learning_rate": 2.9423806031859956e-05, "loss": 0.2917, "step": 7155 }, { "epoch": 0.33, "learning_rate": 2.9423337962023932e-05, "loss": 0.1425, "step": 7160 }, { "epoch": 0.33, "learning_rate": 2.9422869892187912e-05, "loss": 0.0683, "step": 7165 }, { "epoch": 0.33, "learning_rate": 2.9422401822351895e-05, "loss": 0.1217, "step": 7170 }, { "epoch": 0.33, "learning_rate": 2.9421933752515875e-05, "loss": 0.1867, "step": 7175 }, { "epoch": 0.34, "learning_rate": 2.9421465682679855e-05, "loss": 0.2212, "step": 7180 }, { "epoch": 0.34, "learning_rate": 2.9420997612843838e-05, "loss": 0.2069, "step": 7185 }, { "epoch": 0.34, "learning_rate": 2.9420529543007818e-05, "loss": 0.3088, "step": 7190 }, { "epoch": 0.34, "learning_rate": 2.9420061473171798e-05, "loss": 0.396, "step": 7195 }, { "epoch": 0.34, "learning_rate": 2.9419593403335778e-05, "loss": 0.648, "step": 7200 }, { "epoch": 0.34, "learning_rate": 2.941912533349976e-05, "loss": 0.2814, "step": 7205 }, { "epoch": 0.34, "learning_rate": 2.941865726366374e-05, "loss": 0.1857, "step": 7210 }, { "epoch": 0.34, "learning_rate": 2.941818919382772e-05, "loss": 0.2256, "step": 7215 }, { "epoch": 0.34, "learning_rate": 2.94177211239917e-05, "loss": 0.1774, "step": 7220 }, { "epoch": 0.34, "learning_rate": 2.941725305415568e-05, "loss": 0.2119, "step": 7225 }, { "epoch": 0.34, "learning_rate": 2.941678498431966e-05, "loss": 0.2456, "step": 7230 }, { "epoch": 0.34, "learning_rate": 2.941631691448364e-05, "loss": 0.3269, "step": 7235 }, { "epoch": 0.34, "learning_rate": 2.9415848844647623e-05, "loss": 0.312, "step": 7240 }, { "epoch": 0.34, "learning_rate": 2.9415380774811603e-05, "loss": 0.4873, "step": 7245 }, { "epoch": 0.34, "learning_rate": 2.9414912704975583e-05, "loss": 0.5289, "step": 7250 }, { "epoch": 0.34, "learning_rate": 2.9414444635139562e-05, "loss": 0.2919, "step": 7255 }, { "epoch": 0.34, "learning_rate": 2.9413976565303546e-05, "loss": 0.1547, "step": 7260 }, { "epoch": 0.34, "learning_rate": 2.9413508495467525e-05, "loss": 0.1351, "step": 7265 }, { "epoch": 0.34, "learning_rate": 2.9413040425631505e-05, "loss": 0.1993, "step": 7270 }, { "epoch": 0.34, "learning_rate": 2.9412572355795485e-05, "loss": 0.2175, "step": 7275 }, { "epoch": 0.34, "learning_rate": 2.9412104285959468e-05, "loss": 0.2232, "step": 7280 }, { "epoch": 0.34, "learning_rate": 2.9411636216123448e-05, "loss": 0.236, "step": 7285 }, { "epoch": 0.34, "learning_rate": 2.9411168146287425e-05, "loss": 0.315, "step": 7290 }, { "epoch": 0.34, "learning_rate": 2.9410700076451408e-05, "loss": 0.4131, "step": 7295 }, { "epoch": 0.34, "learning_rate": 2.9410232006615388e-05, "loss": 0.6558, "step": 7300 }, { "epoch": 0.34, "learning_rate": 2.9409763936779367e-05, "loss": 0.2861, "step": 7305 }, { "epoch": 0.34, "learning_rate": 2.9409295866943347e-05, "loss": 0.0903, "step": 7310 }, { "epoch": 0.34, "learning_rate": 2.940882779710733e-05, "loss": 0.1395, "step": 7315 }, { "epoch": 0.34, "learning_rate": 2.940835972727131e-05, "loss": 0.1971, "step": 7320 }, { "epoch": 0.34, "learning_rate": 2.940789165743529e-05, "loss": 0.2326, "step": 7325 }, { "epoch": 0.34, "learning_rate": 2.940742358759927e-05, "loss": 0.2499, "step": 7330 }, { "epoch": 0.34, "learning_rate": 2.9406955517763253e-05, "loss": 0.2246, "step": 7335 }, { "epoch": 0.34, "learning_rate": 2.9406487447927233e-05, "loss": 0.2401, "step": 7340 }, { "epoch": 0.34, "learning_rate": 2.9406019378091213e-05, "loss": 0.3713, "step": 7345 }, { "epoch": 0.34, "learning_rate": 2.940555130825519e-05, "loss": 0.4214, "step": 7350 }, { "epoch": 0.34, "learning_rate": 2.9405083238419172e-05, "loss": 0.232, "step": 7355 }, { "epoch": 0.34, "learning_rate": 2.9404615168583152e-05, "loss": 0.1076, "step": 7360 }, { "epoch": 0.34, "learning_rate": 2.9404147098747132e-05, "loss": 0.2181, "step": 7365 }, { "epoch": 0.34, "learning_rate": 2.9403679028911115e-05, "loss": 0.2323, "step": 7370 }, { "epoch": 0.34, "learning_rate": 2.9403210959075095e-05, "loss": 0.2025, "step": 7375 }, { "epoch": 0.34, "learning_rate": 2.9402742889239075e-05, "loss": 0.3211, "step": 7380 }, { "epoch": 0.34, "learning_rate": 2.9402274819403055e-05, "loss": 0.2685, "step": 7385 }, { "epoch": 0.34, "learning_rate": 2.9401806749567038e-05, "loss": 0.3723, "step": 7390 }, { "epoch": 0.35, "learning_rate": 2.9401338679731018e-05, "loss": 0.5123, "step": 7395 }, { "epoch": 0.35, "learning_rate": 2.9400870609894997e-05, "loss": 0.6605, "step": 7400 }, { "epoch": 0.35, "learning_rate": 2.9400402540058977e-05, "loss": 0.3077, "step": 7405 }, { "epoch": 0.35, "learning_rate": 2.939993447022296e-05, "loss": 0.1301, "step": 7410 }, { "epoch": 0.35, "learning_rate": 2.9399466400386937e-05, "loss": 0.1506, "step": 7415 }, { "epoch": 0.35, "learning_rate": 2.9398998330550917e-05, "loss": 0.1235, "step": 7420 }, { "epoch": 0.35, "learning_rate": 2.93985302607149e-05, "loss": 0.2549, "step": 7425 }, { "epoch": 0.35, "learning_rate": 2.939806219087888e-05, "loss": 0.1672, "step": 7430 }, { "epoch": 0.35, "learning_rate": 2.939759412104286e-05, "loss": 0.174, "step": 7435 }, { "epoch": 0.35, "learning_rate": 2.939712605120684e-05, "loss": 0.1782, "step": 7440 }, { "epoch": 0.35, "learning_rate": 2.9396657981370823e-05, "loss": 0.4178, "step": 7445 }, { "epoch": 0.35, "learning_rate": 2.9396189911534802e-05, "loss": 0.6901, "step": 7450 }, { "epoch": 0.35, "learning_rate": 2.9395721841698782e-05, "loss": 0.2847, "step": 7455 }, { "epoch": 0.35, "learning_rate": 2.9395253771862762e-05, "loss": 0.0785, "step": 7460 }, { "epoch": 0.35, "learning_rate": 2.9394785702026745e-05, "loss": 0.2106, "step": 7465 }, { "epoch": 0.35, "learning_rate": 2.9394317632190725e-05, "loss": 0.1722, "step": 7470 }, { "epoch": 0.35, "learning_rate": 2.9393849562354705e-05, "loss": 0.1965, "step": 7475 }, { "epoch": 0.35, "learning_rate": 2.9393381492518685e-05, "loss": 0.2786, "step": 7480 }, { "epoch": 0.35, "learning_rate": 2.9392913422682665e-05, "loss": 0.3831, "step": 7485 }, { "epoch": 0.35, "learning_rate": 2.9392445352846644e-05, "loss": 0.3972, "step": 7490 }, { "epoch": 0.35, "learning_rate": 2.9391977283010624e-05, "loss": 0.3598, "step": 7495 }, { "epoch": 0.35, "learning_rate": 2.9391509213174607e-05, "loss": 0.7829, "step": 7500 }, { "epoch": 0.35, "learning_rate": 2.9391041143338587e-05, "loss": 0.3237, "step": 7505 }, { "epoch": 0.35, "learning_rate": 2.9390573073502567e-05, "loss": 0.1234, "step": 7510 }, { "epoch": 0.35, "learning_rate": 2.9390105003666547e-05, "loss": 0.0924, "step": 7515 }, { "epoch": 0.35, "learning_rate": 2.938963693383053e-05, "loss": 0.1817, "step": 7520 }, { "epoch": 0.35, "learning_rate": 2.938916886399451e-05, "loss": 0.1511, "step": 7525 }, { "epoch": 0.35, "learning_rate": 2.938870079415849e-05, "loss": 0.1858, "step": 7530 }, { "epoch": 0.35, "learning_rate": 2.938823272432247e-05, "loss": 0.319, "step": 7535 }, { "epoch": 0.35, "learning_rate": 2.938776465448645e-05, "loss": 0.1937, "step": 7540 }, { "epoch": 0.35, "learning_rate": 2.938729658465043e-05, "loss": 0.3744, "step": 7545 }, { "epoch": 0.35, "learning_rate": 2.938682851481441e-05, "loss": 0.6982, "step": 7550 }, { "epoch": 0.35, "learning_rate": 2.9386360444978392e-05, "loss": 0.2617, "step": 7555 }, { "epoch": 0.35, "learning_rate": 2.9385892375142372e-05, "loss": 0.1418, "step": 7560 }, { "epoch": 0.35, "learning_rate": 2.9385424305306352e-05, "loss": 0.2089, "step": 7565 }, { "epoch": 0.35, "learning_rate": 2.938495623547033e-05, "loss": 0.1712, "step": 7570 }, { "epoch": 0.35, "learning_rate": 2.9384488165634315e-05, "loss": 0.2747, "step": 7575 }, { "epoch": 0.35, "learning_rate": 2.9384020095798295e-05, "loss": 0.3313, "step": 7580 }, { "epoch": 0.35, "learning_rate": 2.9383552025962274e-05, "loss": 0.2143, "step": 7585 }, { "epoch": 0.35, "learning_rate": 2.9383083956126254e-05, "loss": 0.2866, "step": 7590 }, { "epoch": 0.35, "learning_rate": 2.9382615886290237e-05, "loss": 0.4264, "step": 7595 }, { "epoch": 0.35, "learning_rate": 2.9382147816454217e-05, "loss": 0.7221, "step": 7600 }, { "epoch": 0.35, "learning_rate": 2.9381679746618194e-05, "loss": 0.334, "step": 7605 }, { "epoch": 0.36, "learning_rate": 2.9381211676782177e-05, "loss": 0.1238, "step": 7610 }, { "epoch": 0.36, "learning_rate": 2.9380743606946157e-05, "loss": 0.1912, "step": 7615 }, { "epoch": 0.36, "learning_rate": 2.9380275537110137e-05, "loss": 0.1523, "step": 7620 }, { "epoch": 0.36, "learning_rate": 2.9379807467274116e-05, "loss": 0.2195, "step": 7625 }, { "epoch": 0.36, "learning_rate": 2.93793393974381e-05, "loss": 0.2804, "step": 7630 }, { "epoch": 0.36, "learning_rate": 2.937887132760208e-05, "loss": 0.2835, "step": 7635 }, { "epoch": 0.36, "learning_rate": 2.937840325776606e-05, "loss": 0.3109, "step": 7640 }, { "epoch": 0.36, "learning_rate": 2.937793518793004e-05, "loss": 0.459, "step": 7645 }, { "epoch": 0.36, "learning_rate": 2.9377467118094022e-05, "loss": 0.7011, "step": 7650 }, { "epoch": 0.36, "learning_rate": 2.9376999048258002e-05, "loss": 0.2537, "step": 7655 }, { "epoch": 0.36, "learning_rate": 2.9376530978421982e-05, "loss": 0.1177, "step": 7660 }, { "epoch": 0.36, "learning_rate": 2.937606290858596e-05, "loss": 0.1364, "step": 7665 }, { "epoch": 0.36, "learning_rate": 2.937559483874994e-05, "loss": 0.1555, "step": 7670 }, { "epoch": 0.36, "learning_rate": 2.937512676891392e-05, "loss": 0.1555, "step": 7675 }, { "epoch": 0.36, "learning_rate": 2.93746586990779e-05, "loss": 0.2303, "step": 7680 }, { "epoch": 0.36, "learning_rate": 2.9374190629241884e-05, "loss": 0.267, "step": 7685 }, { "epoch": 0.36, "learning_rate": 2.9373722559405864e-05, "loss": 0.389, "step": 7690 }, { "epoch": 0.36, "learning_rate": 2.9373254489569844e-05, "loss": 0.3087, "step": 7695 }, { "epoch": 0.36, "learning_rate": 2.9372786419733824e-05, "loss": 0.5691, "step": 7700 }, { "epoch": 0.36, "learning_rate": 2.9372318349897807e-05, "loss": 0.2561, "step": 7705 }, { "epoch": 0.36, "learning_rate": 2.9371850280061787e-05, "loss": 0.1445, "step": 7710 }, { "epoch": 0.36, "learning_rate": 2.9371382210225767e-05, "loss": 0.1372, "step": 7715 }, { "epoch": 0.36, "learning_rate": 2.937091414038975e-05, "loss": 0.2023, "step": 7720 }, { "epoch": 0.36, "learning_rate": 2.937044607055373e-05, "loss": 0.2601, "step": 7725 }, { "epoch": 0.36, "learning_rate": 2.9369978000717706e-05, "loss": 0.3441, "step": 7730 }, { "epoch": 0.36, "learning_rate": 2.9369509930881686e-05, "loss": 0.2752, "step": 7735 }, { "epoch": 0.36, "learning_rate": 2.936904186104567e-05, "loss": 0.5722, "step": 7740 }, { "epoch": 0.36, "learning_rate": 2.936857379120965e-05, "loss": 0.4032, "step": 7745 }, { "epoch": 0.36, "learning_rate": 2.936810572137363e-05, "loss": 0.69, "step": 7750 }, { "epoch": 0.36, "learning_rate": 2.936763765153761e-05, "loss": 0.2623, "step": 7755 }, { "epoch": 0.36, "learning_rate": 2.9367169581701592e-05, "loss": 0.1181, "step": 7760 }, { "epoch": 0.36, "learning_rate": 2.936670151186557e-05, "loss": 0.1879, "step": 7765 }, { "epoch": 0.36, "learning_rate": 2.936623344202955e-05, "loss": 0.1422, "step": 7770 }, { "epoch": 0.36, "learning_rate": 2.936576537219353e-05, "loss": 0.2175, "step": 7775 }, { "epoch": 0.36, "learning_rate": 2.9365297302357514e-05, "loss": 0.2454, "step": 7780 }, { "epoch": 0.36, "learning_rate": 2.9364829232521494e-05, "loss": 0.231, "step": 7785 }, { "epoch": 0.36, "learning_rate": 2.9364361162685474e-05, "loss": 0.3173, "step": 7790 }, { "epoch": 0.36, "learning_rate": 2.9363893092849454e-05, "loss": 0.3939, "step": 7795 }, { "epoch": 0.36, "learning_rate": 2.9363425023013434e-05, "loss": 0.5128, "step": 7800 }, { "epoch": 0.36, "learning_rate": 2.9362956953177414e-05, "loss": 0.3327, "step": 7805 }, { "epoch": 0.36, "learning_rate": 2.9362488883341393e-05, "loss": 0.1581, "step": 7810 }, { "epoch": 0.36, "learning_rate": 2.9362020813505377e-05, "loss": 0.1292, "step": 7815 }, { "epoch": 0.36, "learning_rate": 2.9361552743669356e-05, "loss": 0.2084, "step": 7820 }, { "epoch": 0.37, "learning_rate": 2.9361084673833336e-05, "loss": 0.2552, "step": 7825 }, { "epoch": 0.37, "learning_rate": 2.9360616603997316e-05, "loss": 0.2811, "step": 7830 }, { "epoch": 0.37, "learning_rate": 2.93601485341613e-05, "loss": 0.2133, "step": 7835 }, { "epoch": 0.37, "learning_rate": 2.935968046432528e-05, "loss": 0.3318, "step": 7840 }, { "epoch": 0.37, "learning_rate": 2.935921239448926e-05, "loss": 0.4631, "step": 7845 }, { "epoch": 0.37, "learning_rate": 2.9358744324653242e-05, "loss": 0.5992, "step": 7850 }, { "epoch": 0.37, "learning_rate": 2.935827625481722e-05, "loss": 0.3598, "step": 7855 }, { "epoch": 0.37, "learning_rate": 2.9357808184981198e-05, "loss": 0.1455, "step": 7860 }, { "epoch": 0.37, "learning_rate": 2.9357340115145178e-05, "loss": 0.1451, "step": 7865 }, { "epoch": 0.37, "learning_rate": 2.935687204530916e-05, "loss": 0.1478, "step": 7870 }, { "epoch": 0.37, "learning_rate": 2.935640397547314e-05, "loss": 0.1385, "step": 7875 }, { "epoch": 0.37, "learning_rate": 2.935593590563712e-05, "loss": 0.2764, "step": 7880 }, { "epoch": 0.37, "learning_rate": 2.93554678358011e-05, "loss": 0.2472, "step": 7885 }, { "epoch": 0.37, "learning_rate": 2.9354999765965084e-05, "loss": 0.2704, "step": 7890 }, { "epoch": 0.37, "learning_rate": 2.9354531696129064e-05, "loss": 0.2523, "step": 7895 }, { "epoch": 0.37, "learning_rate": 2.9354063626293044e-05, "loss": 0.8368, "step": 7900 }, { "epoch": 0.37, "learning_rate": 2.9353595556457027e-05, "loss": 0.2962, "step": 7905 }, { "epoch": 0.37, "learning_rate": 2.9353127486621007e-05, "loss": 0.1155, "step": 7910 }, { "epoch": 0.37, "learning_rate": 2.9352659416784986e-05, "loss": 0.0999, "step": 7915 }, { "epoch": 0.37, "learning_rate": 2.9352191346948963e-05, "loss": 0.1697, "step": 7920 }, { "epoch": 0.37, "learning_rate": 2.9351723277112946e-05, "loss": 0.2106, "step": 7925 }, { "epoch": 0.37, "learning_rate": 2.9351255207276926e-05, "loss": 0.227, "step": 7930 }, { "epoch": 0.37, "learning_rate": 2.9350787137440906e-05, "loss": 0.2881, "step": 7935 }, { "epoch": 0.37, "learning_rate": 2.9350319067604886e-05, "loss": 0.4414, "step": 7940 }, { "epoch": 0.37, "learning_rate": 2.934985099776887e-05, "loss": 0.3858, "step": 7945 }, { "epoch": 0.37, "learning_rate": 2.934938292793285e-05, "loss": 0.7588, "step": 7950 }, { "epoch": 0.37, "learning_rate": 2.934891485809683e-05, "loss": 0.245, "step": 7955 }, { "epoch": 0.37, "learning_rate": 2.9348446788260808e-05, "loss": 0.1279, "step": 7960 }, { "epoch": 0.37, "learning_rate": 2.934797871842479e-05, "loss": 0.1447, "step": 7965 }, { "epoch": 0.37, "learning_rate": 2.934751064858877e-05, "loss": 0.2412, "step": 7970 }, { "epoch": 0.37, "learning_rate": 2.934704257875275e-05, "loss": 0.2052, "step": 7975 }, { "epoch": 0.37, "learning_rate": 2.934657450891673e-05, "loss": 0.2586, "step": 7980 }, { "epoch": 0.37, "learning_rate": 2.934610643908071e-05, "loss": 0.2186, "step": 7985 }, { "epoch": 0.37, "learning_rate": 2.934563836924469e-05, "loss": 0.3543, "step": 7990 }, { "epoch": 0.37, "learning_rate": 2.934517029940867e-05, "loss": 0.4752, "step": 7995 }, { "epoch": 0.37, "learning_rate": 2.9344702229572654e-05, "loss": 0.6421, "step": 8000 }, { "epoch": 0.37, "learning_rate": 2.9344234159736633e-05, "loss": 0.1856, "step": 8005 }, { "epoch": 0.37, "learning_rate": 2.9343766089900613e-05, "loss": 0.0823, "step": 8010 }, { "epoch": 0.37, "learning_rate": 2.9343298020064593e-05, "loss": 0.1514, "step": 8015 }, { "epoch": 0.37, "learning_rate": 2.9342829950228576e-05, "loss": 0.2636, "step": 8020 }, { "epoch": 0.37, "learning_rate": 2.9342361880392556e-05, "loss": 0.2093, "step": 8025 }, { "epoch": 0.37, "learning_rate": 2.9341893810556536e-05, "loss": 0.1602, "step": 8030 }, { "epoch": 0.37, "learning_rate": 2.934142574072052e-05, "loss": 0.4032, "step": 8035 }, { "epoch": 0.38, "learning_rate": 2.93409576708845e-05, "loss": 0.3007, "step": 8040 }, { "epoch": 0.38, "learning_rate": 2.9340489601048475e-05, "loss": 0.4056, "step": 8045 }, { "epoch": 0.38, "learning_rate": 2.9340021531212455e-05, "loss": 0.7518, "step": 8050 }, { "epoch": 0.38, "learning_rate": 2.9339553461376438e-05, "loss": 0.2775, "step": 8055 }, { "epoch": 0.38, "learning_rate": 2.9339085391540418e-05, "loss": 0.1476, "step": 8060 }, { "epoch": 0.38, "learning_rate": 2.9338617321704398e-05, "loss": 0.1085, "step": 8065 }, { "epoch": 0.38, "learning_rate": 2.9338149251868378e-05, "loss": 0.2244, "step": 8070 }, { "epoch": 0.38, "learning_rate": 2.933768118203236e-05, "loss": 0.1919, "step": 8075 }, { "epoch": 0.38, "learning_rate": 2.933721311219634e-05, "loss": 0.3301, "step": 8080 }, { "epoch": 0.38, "learning_rate": 2.933674504236032e-05, "loss": 0.2832, "step": 8085 }, { "epoch": 0.38, "learning_rate": 2.9336276972524304e-05, "loss": 0.3272, "step": 8090 }, { "epoch": 0.38, "learning_rate": 2.9335808902688284e-05, "loss": 0.4913, "step": 8095 }, { "epoch": 0.38, "learning_rate": 2.9335340832852263e-05, "loss": 0.6859, "step": 8100 }, { "epoch": 0.38, "learning_rate": 2.9334872763016243e-05, "loss": 0.2242, "step": 8105 }, { "epoch": 0.38, "learning_rate": 2.9334404693180223e-05, "loss": 0.1239, "step": 8110 }, { "epoch": 0.38, "learning_rate": 2.9333936623344203e-05, "loss": 0.1162, "step": 8115 }, { "epoch": 0.38, "learning_rate": 2.9333468553508183e-05, "loss": 0.1759, "step": 8120 }, { "epoch": 0.38, "learning_rate": 2.9333000483672163e-05, "loss": 0.2751, "step": 8125 }, { "epoch": 0.38, "learning_rate": 2.9332532413836146e-05, "loss": 0.3035, "step": 8130 }, { "epoch": 0.38, "learning_rate": 2.9332064344000126e-05, "loss": 0.2707, "step": 8135 }, { "epoch": 0.38, "learning_rate": 2.9331596274164105e-05, "loss": 0.3818, "step": 8140 }, { "epoch": 0.38, "learning_rate": 2.9331128204328085e-05, "loss": 0.5258, "step": 8145 }, { "epoch": 0.38, "learning_rate": 2.933066013449207e-05, "loss": 0.4677, "step": 8150 }, { "epoch": 0.38, "learning_rate": 2.9330192064656048e-05, "loss": 0.3592, "step": 8155 }, { "epoch": 0.38, "learning_rate": 2.9329723994820028e-05, "loss": 0.13, "step": 8160 }, { "epoch": 0.38, "learning_rate": 2.932925592498401e-05, "loss": 0.1378, "step": 8165 }, { "epoch": 0.38, "learning_rate": 2.9328787855147988e-05, "loss": 0.2196, "step": 8170 }, { "epoch": 0.38, "learning_rate": 2.9328319785311967e-05, "loss": 0.2147, "step": 8175 }, { "epoch": 0.38, "learning_rate": 2.9327851715475947e-05, "loss": 0.1999, "step": 8180 }, { "epoch": 0.38, "learning_rate": 2.932738364563993e-05, "loss": 0.3967, "step": 8185 }, { "epoch": 0.38, "learning_rate": 2.932691557580391e-05, "loss": 0.4154, "step": 8190 }, { "epoch": 0.38, "learning_rate": 2.932644750596789e-05, "loss": 0.3932, "step": 8195 }, { "epoch": 0.38, "learning_rate": 2.932597943613187e-05, "loss": 0.5204, "step": 8200 }, { "epoch": 0.38, "learning_rate": 2.9325511366295853e-05, "loss": 0.2393, "step": 8205 }, { "epoch": 0.38, "learning_rate": 2.9325043296459833e-05, "loss": 0.0981, "step": 8210 }, { "epoch": 0.38, "learning_rate": 2.9324575226623813e-05, "loss": 0.0835, "step": 8215 }, { "epoch": 0.38, "learning_rate": 2.9324107156787796e-05, "loss": 0.1849, "step": 8220 }, { "epoch": 0.38, "learning_rate": 2.9323639086951776e-05, "loss": 0.2397, "step": 8225 }, { "epoch": 0.38, "learning_rate": 2.9323171017115756e-05, "loss": 0.1983, "step": 8230 }, { "epoch": 0.38, "learning_rate": 2.9322702947279732e-05, "loss": 0.2821, "step": 8235 }, { "epoch": 0.38, "learning_rate": 2.9322234877443715e-05, "loss": 0.2998, "step": 8240 }, { "epoch": 0.38, "learning_rate": 2.9321766807607695e-05, "loss": 0.3615, "step": 8245 }, { "epoch": 0.38, "learning_rate": 2.9321298737771675e-05, "loss": 0.442, "step": 8250 }, { "epoch": 0.39, "learning_rate": 2.9320830667935655e-05, "loss": 0.3459, "step": 8255 }, { "epoch": 0.39, "learning_rate": 2.9320362598099638e-05, "loss": 0.0708, "step": 8260 }, { "epoch": 0.39, "learning_rate": 2.9319894528263618e-05, "loss": 0.128, "step": 8265 }, { "epoch": 0.39, "learning_rate": 2.9319426458427598e-05, "loss": 0.1367, "step": 8270 }, { "epoch": 0.39, "learning_rate": 2.931895838859158e-05, "loss": 0.1983, "step": 8275 }, { "epoch": 0.39, "learning_rate": 2.931849031875556e-05, "loss": 0.262, "step": 8280 }, { "epoch": 0.39, "learning_rate": 2.931802224891954e-05, "loss": 0.1638, "step": 8285 }, { "epoch": 0.39, "learning_rate": 2.931755417908352e-05, "loss": 0.2613, "step": 8290 }, { "epoch": 0.39, "learning_rate": 2.9317086109247503e-05, "loss": 0.3276, "step": 8295 }, { "epoch": 0.39, "learning_rate": 2.931661803941148e-05, "loss": 0.5973, "step": 8300 }, { "epoch": 0.39, "learning_rate": 2.931614996957546e-05, "loss": 0.3419, "step": 8305 }, { "epoch": 0.39, "learning_rate": 2.931568189973944e-05, "loss": 0.0987, "step": 8310 }, { "epoch": 0.39, "learning_rate": 2.9315213829903423e-05, "loss": 0.1584, "step": 8315 }, { "epoch": 0.39, "learning_rate": 2.9314745760067403e-05, "loss": 0.1905, "step": 8320 }, { "epoch": 0.39, "learning_rate": 2.9314277690231382e-05, "loss": 0.1759, "step": 8325 }, { "epoch": 0.39, "learning_rate": 2.9313809620395366e-05, "loss": 0.2104, "step": 8330 }, { "epoch": 0.39, "learning_rate": 2.9313341550559345e-05, "loss": 0.2476, "step": 8335 }, { "epoch": 0.39, "learning_rate": 2.9312873480723325e-05, "loss": 0.2675, "step": 8340 }, { "epoch": 0.39, "learning_rate": 2.9312405410887305e-05, "loss": 0.2632, "step": 8345 }, { "epoch": 0.39, "learning_rate": 2.9311937341051288e-05, "loss": 0.5837, "step": 8350 }, { "epoch": 0.39, "learning_rate": 2.9311469271215268e-05, "loss": 0.2755, "step": 8355 }, { "epoch": 0.39, "learning_rate": 2.9311001201379244e-05, "loss": 0.0873, "step": 8360 }, { "epoch": 0.39, "learning_rate": 2.9310533131543224e-05, "loss": 0.1184, "step": 8365 }, { "epoch": 0.39, "learning_rate": 2.9310065061707207e-05, "loss": 0.161, "step": 8370 }, { "epoch": 0.39, "learning_rate": 2.9309596991871187e-05, "loss": 0.1826, "step": 8375 }, { "epoch": 0.39, "learning_rate": 2.9309128922035167e-05, "loss": 0.2666, "step": 8380 }, { "epoch": 0.39, "learning_rate": 2.9308660852199147e-05, "loss": 0.2212, "step": 8385 }, { "epoch": 0.39, "learning_rate": 2.930819278236313e-05, "loss": 0.3762, "step": 8390 }, { "epoch": 0.39, "learning_rate": 2.930772471252711e-05, "loss": 0.4746, "step": 8395 }, { "epoch": 0.39, "learning_rate": 2.930725664269109e-05, "loss": 0.8668, "step": 8400 }, { "epoch": 0.39, "learning_rate": 2.9306788572855073e-05, "loss": 0.3019, "step": 8405 }, { "epoch": 0.39, "learning_rate": 2.9306320503019053e-05, "loss": 0.096, "step": 8410 }, { "epoch": 0.39, "learning_rate": 2.9305852433183033e-05, "loss": 0.1645, "step": 8415 }, { "epoch": 0.39, "learning_rate": 2.9305384363347012e-05, "loss": 0.181, "step": 8420 }, { "epoch": 0.39, "learning_rate": 2.9304916293510992e-05, "loss": 0.1734, "step": 8425 }, { "epoch": 0.39, "learning_rate": 2.9304448223674972e-05, "loss": 0.2393, "step": 8430 }, { "epoch": 0.39, "learning_rate": 2.9303980153838952e-05, "loss": 0.292, "step": 8435 }, { "epoch": 0.39, "learning_rate": 2.930351208400293e-05, "loss": 0.2921, "step": 8440 }, { "epoch": 0.39, "learning_rate": 2.9303044014166915e-05, "loss": 0.5353, "step": 8445 }, { "epoch": 0.39, "learning_rate": 2.9302575944330895e-05, "loss": 0.707, "step": 8450 }, { "epoch": 0.39, "learning_rate": 2.9302107874494875e-05, "loss": 0.2303, "step": 8455 }, { "epoch": 0.39, "learning_rate": 2.9301639804658858e-05, "loss": 0.1254, "step": 8460 }, { "epoch": 0.39, "learning_rate": 2.9301171734822838e-05, "loss": 0.1365, "step": 8465 }, { "epoch": 0.4, "learning_rate": 2.9300703664986817e-05, "loss": 0.0969, "step": 8470 }, { "epoch": 0.4, "learning_rate": 2.9300235595150797e-05, "loss": 0.1487, "step": 8475 }, { "epoch": 0.4, "learning_rate": 2.929976752531478e-05, "loss": 0.2461, "step": 8480 }, { "epoch": 0.4, "learning_rate": 2.929929945547876e-05, "loss": 0.3552, "step": 8485 }, { "epoch": 0.4, "learning_rate": 2.9298831385642737e-05, "loss": 0.3688, "step": 8490 }, { "epoch": 0.4, "learning_rate": 2.9298363315806716e-05, "loss": 0.3737, "step": 8495 }, { "epoch": 0.4, "learning_rate": 2.92978952459707e-05, "loss": 0.4919, "step": 8500 }, { "epoch": 0.4, "learning_rate": 2.929742717613468e-05, "loss": 0.24, "step": 8505 }, { "epoch": 0.4, "learning_rate": 2.929695910629866e-05, "loss": 0.1171, "step": 8510 }, { "epoch": 0.4, "learning_rate": 2.9296491036462643e-05, "loss": 0.0792, "step": 8515 }, { "epoch": 0.4, "learning_rate": 2.9296022966626622e-05, "loss": 0.1682, "step": 8520 }, { "epoch": 0.4, "learning_rate": 2.9295554896790602e-05, "loss": 0.1375, "step": 8525 }, { "epoch": 0.4, "learning_rate": 2.9295086826954582e-05, "loss": 0.2226, "step": 8530 }, { "epoch": 0.4, "learning_rate": 2.9294618757118565e-05, "loss": 0.2332, "step": 8535 }, { "epoch": 0.4, "learning_rate": 2.9294150687282545e-05, "loss": 0.2816, "step": 8540 }, { "epoch": 0.4, "learning_rate": 2.9293682617446525e-05, "loss": 0.4733, "step": 8545 }, { "epoch": 0.4, "learning_rate": 2.92932145476105e-05, "loss": 0.6968, "step": 8550 }, { "epoch": 0.4, "learning_rate": 2.9292746477774484e-05, "loss": 0.3567, "step": 8555 }, { "epoch": 0.4, "learning_rate": 2.9292278407938464e-05, "loss": 0.1067, "step": 8560 }, { "epoch": 0.4, "learning_rate": 2.9291810338102444e-05, "loss": 0.0944, "step": 8565 }, { "epoch": 0.4, "learning_rate": 2.9291342268266424e-05, "loss": 0.1015, "step": 8570 }, { "epoch": 0.4, "learning_rate": 2.9290874198430407e-05, "loss": 0.172, "step": 8575 }, { "epoch": 0.4, "learning_rate": 2.9290406128594387e-05, "loss": 0.2308, "step": 8580 }, { "epoch": 0.4, "learning_rate": 2.9289938058758367e-05, "loss": 0.2653, "step": 8585 }, { "epoch": 0.4, "learning_rate": 2.928946998892235e-05, "loss": 0.2944, "step": 8590 }, { "epoch": 0.4, "learning_rate": 2.928900191908633e-05, "loss": 0.5232, "step": 8595 }, { "epoch": 0.4, "learning_rate": 2.928853384925031e-05, "loss": 0.8223, "step": 8600 }, { "epoch": 0.4, "learning_rate": 2.928806577941429e-05, "loss": 0.3007, "step": 8605 }, { "epoch": 0.4, "learning_rate": 2.9287597709578273e-05, "loss": 0.1118, "step": 8610 }, { "epoch": 0.4, "learning_rate": 2.928712963974225e-05, "loss": 0.2052, "step": 8615 }, { "epoch": 0.4, "learning_rate": 2.928666156990623e-05, "loss": 0.2462, "step": 8620 }, { "epoch": 0.4, "learning_rate": 2.928619350007021e-05, "loss": 0.254, "step": 8625 }, { "epoch": 0.4, "learning_rate": 2.9285725430234192e-05, "loss": 0.1789, "step": 8630 }, { "epoch": 0.4, "learning_rate": 2.928525736039817e-05, "loss": 0.2189, "step": 8635 }, { "epoch": 0.4, "learning_rate": 2.928478929056215e-05, "loss": 0.3401, "step": 8640 }, { "epoch": 0.4, "learning_rate": 2.9284321220726135e-05, "loss": 0.2369, "step": 8645 }, { "epoch": 0.4, "learning_rate": 2.9283853150890115e-05, "loss": 0.6526, "step": 8650 }, { "epoch": 0.4, "learning_rate": 2.9283385081054094e-05, "loss": 0.2264, "step": 8655 }, { "epoch": 0.4, "learning_rate": 2.9282917011218074e-05, "loss": 0.098, "step": 8660 }, { "epoch": 0.4, "learning_rate": 2.9282448941382057e-05, "loss": 0.1507, "step": 8665 }, { "epoch": 0.4, "learning_rate": 2.9281980871546037e-05, "loss": 0.1024, "step": 8670 }, { "epoch": 0.4, "learning_rate": 2.9281512801710017e-05, "loss": 0.1547, "step": 8675 }, { "epoch": 0.41, "learning_rate": 2.9281044731873993e-05, "loss": 0.2488, "step": 8680 }, { "epoch": 0.41, "learning_rate": 2.9280576662037977e-05, "loss": 0.3293, "step": 8685 }, { "epoch": 0.41, "learning_rate": 2.9280108592201956e-05, "loss": 0.2928, "step": 8690 }, { "epoch": 0.41, "learning_rate": 2.9279640522365936e-05, "loss": 0.4896, "step": 8695 }, { "epoch": 0.41, "learning_rate": 2.927917245252992e-05, "loss": 0.451, "step": 8700 }, { "epoch": 0.41, "learning_rate": 2.92787043826939e-05, "loss": 0.2863, "step": 8705 }, { "epoch": 0.41, "learning_rate": 2.927823631285788e-05, "loss": 0.1178, "step": 8710 }, { "epoch": 0.41, "learning_rate": 2.927776824302186e-05, "loss": 0.1274, "step": 8715 }, { "epoch": 0.41, "learning_rate": 2.9277300173185842e-05, "loss": 0.1832, "step": 8720 }, { "epoch": 0.41, "learning_rate": 2.9276832103349822e-05, "loss": 0.2102, "step": 8725 }, { "epoch": 0.41, "learning_rate": 2.9276364033513802e-05, "loss": 0.2072, "step": 8730 }, { "epoch": 0.41, "learning_rate": 2.927589596367778e-05, "loss": 0.4095, "step": 8735 }, { "epoch": 0.41, "learning_rate": 2.927542789384176e-05, "loss": 0.2432, "step": 8740 }, { "epoch": 0.41, "learning_rate": 2.927495982400574e-05, "loss": 0.3224, "step": 8745 }, { "epoch": 0.41, "learning_rate": 2.927449175416972e-05, "loss": 0.5451, "step": 8750 }, { "epoch": 0.41, "learning_rate": 2.92740236843337e-05, "loss": 0.175, "step": 8755 }, { "epoch": 0.41, "learning_rate": 2.9273555614497684e-05, "loss": 0.1485, "step": 8760 }, { "epoch": 0.41, "learning_rate": 2.9273087544661664e-05, "loss": 0.157, "step": 8765 }, { "epoch": 0.41, "learning_rate": 2.9272619474825644e-05, "loss": 0.174, "step": 8770 }, { "epoch": 0.41, "learning_rate": 2.9272151404989627e-05, "loss": 0.1748, "step": 8775 }, { "epoch": 0.41, "learning_rate": 2.9271683335153607e-05, "loss": 0.2344, "step": 8780 }, { "epoch": 0.41, "learning_rate": 2.9271215265317587e-05, "loss": 0.2776, "step": 8785 }, { "epoch": 0.41, "learning_rate": 2.9270747195481566e-05, "loss": 0.3259, "step": 8790 }, { "epoch": 0.41, "learning_rate": 2.927027912564555e-05, "loss": 0.3422, "step": 8795 }, { "epoch": 0.41, "learning_rate": 2.926981105580953e-05, "loss": 0.5633, "step": 8800 }, { "epoch": 0.41, "learning_rate": 2.9269342985973506e-05, "loss": 0.2588, "step": 8805 }, { "epoch": 0.41, "learning_rate": 2.9268874916137486e-05, "loss": 0.172, "step": 8810 }, { "epoch": 0.41, "learning_rate": 2.926840684630147e-05, "loss": 0.1366, "step": 8815 }, { "epoch": 0.41, "learning_rate": 2.926793877646545e-05, "loss": 0.2007, "step": 8820 }, { "epoch": 0.41, "learning_rate": 2.926747070662943e-05, "loss": 0.2308, "step": 8825 }, { "epoch": 0.41, "learning_rate": 2.926700263679341e-05, "loss": 0.2118, "step": 8830 }, { "epoch": 0.41, "learning_rate": 2.926653456695739e-05, "loss": 0.3322, "step": 8835 }, { "epoch": 0.41, "learning_rate": 2.926606649712137e-05, "loss": 0.3329, "step": 8840 }, { "epoch": 0.41, "learning_rate": 2.926559842728535e-05, "loss": 0.346, "step": 8845 }, { "epoch": 0.41, "learning_rate": 2.9265130357449334e-05, "loss": 0.7412, "step": 8850 }, { "epoch": 0.41, "learning_rate": 2.9264662287613314e-05, "loss": 0.2103, "step": 8855 }, { "epoch": 0.41, "learning_rate": 2.9264194217777294e-05, "loss": 0.1274, "step": 8860 }, { "epoch": 0.41, "learning_rate": 2.9263726147941274e-05, "loss": 0.1788, "step": 8865 }, { "epoch": 0.41, "learning_rate": 2.9263258078105254e-05, "loss": 0.1556, "step": 8870 }, { "epoch": 0.41, "learning_rate": 2.9262790008269233e-05, "loss": 0.18, "step": 8875 }, { "epoch": 0.41, "learning_rate": 2.9262321938433213e-05, "loss": 0.1629, "step": 8880 }, { "epoch": 0.41, "learning_rate": 2.9261853868597196e-05, "loss": 0.2814, "step": 8885 }, { "epoch": 0.41, "learning_rate": 2.9261385798761176e-05, "loss": 0.4083, "step": 8890 }, { "epoch": 0.42, "learning_rate": 2.9260917728925156e-05, "loss": 0.2613, "step": 8895 }, { "epoch": 0.42, "learning_rate": 2.9260449659089136e-05, "loss": 0.4846, "step": 8900 }, { "epoch": 0.42, "learning_rate": 2.925998158925312e-05, "loss": 0.2465, "step": 8905 }, { "epoch": 0.42, "learning_rate": 2.92595135194171e-05, "loss": 0.1036, "step": 8910 }, { "epoch": 0.42, "learning_rate": 2.925904544958108e-05, "loss": 0.1551, "step": 8915 }, { "epoch": 0.42, "learning_rate": 2.925857737974506e-05, "loss": 0.1265, "step": 8920 }, { "epoch": 0.42, "learning_rate": 2.9258109309909042e-05, "loss": 0.2116, "step": 8925 }, { "epoch": 0.42, "learning_rate": 2.9257641240073018e-05, "loss": 0.3118, "step": 8930 }, { "epoch": 0.42, "learning_rate": 2.9257173170236998e-05, "loss": 0.4265, "step": 8935 }, { "epoch": 0.42, "learning_rate": 2.925670510040098e-05, "loss": 0.3197, "step": 8940 }, { "epoch": 0.42, "learning_rate": 2.925623703056496e-05, "loss": 0.2795, "step": 8945 }, { "epoch": 0.42, "learning_rate": 2.925576896072894e-05, "loss": 0.5113, "step": 8950 }, { "epoch": 0.42, "learning_rate": 2.925530089089292e-05, "loss": 0.3283, "step": 8955 }, { "epoch": 0.42, "learning_rate": 2.9254832821056904e-05, "loss": 0.0701, "step": 8960 }, { "epoch": 0.42, "learning_rate": 2.9254364751220884e-05, "loss": 0.1698, "step": 8965 }, { "epoch": 0.42, "learning_rate": 2.9253896681384864e-05, "loss": 0.1525, "step": 8970 }, { "epoch": 0.42, "learning_rate": 2.9253428611548843e-05, "loss": 0.1762, "step": 8975 }, { "epoch": 0.42, "learning_rate": 2.9252960541712827e-05, "loss": 0.2388, "step": 8980 }, { "epoch": 0.42, "learning_rate": 2.9252492471876806e-05, "loss": 0.3278, "step": 8985 }, { "epoch": 0.42, "learning_rate": 2.9252024402040786e-05, "loss": 0.3177, "step": 8990 }, { "epoch": 0.42, "learning_rate": 2.9251556332204763e-05, "loss": 0.4042, "step": 8995 }, { "epoch": 0.42, "learning_rate": 2.9251088262368746e-05, "loss": 0.6997, "step": 9000 }, { "epoch": 0.42, "learning_rate": 2.9250620192532726e-05, "loss": 0.2166, "step": 9005 }, { "epoch": 0.42, "learning_rate": 2.9250152122696705e-05, "loss": 0.087, "step": 9010 }, { "epoch": 0.42, "learning_rate": 2.924968405286069e-05, "loss": 0.1007, "step": 9015 }, { "epoch": 0.42, "learning_rate": 2.924921598302467e-05, "loss": 0.1181, "step": 9020 }, { "epoch": 0.42, "learning_rate": 2.9248747913188648e-05, "loss": 0.2034, "step": 9025 }, { "epoch": 0.42, "learning_rate": 2.9248279843352628e-05, "loss": 0.1965, "step": 9030 }, { "epoch": 0.42, "learning_rate": 2.924781177351661e-05, "loss": 0.2305, "step": 9035 }, { "epoch": 0.42, "learning_rate": 2.924734370368059e-05, "loss": 0.3461, "step": 9040 }, { "epoch": 0.42, "learning_rate": 2.924687563384457e-05, "loss": 0.4492, "step": 9045 }, { "epoch": 0.42, "learning_rate": 2.924640756400855e-05, "loss": 0.5317, "step": 9050 }, { "epoch": 0.42, "learning_rate": 2.924593949417253e-05, "loss": 0.2327, "step": 9055 }, { "epoch": 0.42, "learning_rate": 2.924547142433651e-05, "loss": 0.1141, "step": 9060 }, { "epoch": 0.42, "learning_rate": 2.924500335450049e-05, "loss": 0.2636, "step": 9065 }, { "epoch": 0.42, "learning_rate": 2.9244535284664473e-05, "loss": 0.1519, "step": 9070 }, { "epoch": 0.42, "learning_rate": 2.9244067214828453e-05, "loss": 0.2387, "step": 9075 }, { "epoch": 0.42, "learning_rate": 2.9243599144992433e-05, "loss": 0.3108, "step": 9080 }, { "epoch": 0.42, "learning_rate": 2.9243131075156413e-05, "loss": 0.341, "step": 9085 }, { "epoch": 0.42, "learning_rate": 2.9242663005320396e-05, "loss": 0.3498, "step": 9090 }, { "epoch": 0.42, "learning_rate": 2.9242194935484376e-05, "loss": 0.3984, "step": 9095 }, { "epoch": 0.42, "learning_rate": 2.9241726865648356e-05, "loss": 0.4773, "step": 9100 }, { "epoch": 0.42, "learning_rate": 2.9241258795812336e-05, "loss": 0.2201, "step": 9105 }, { "epoch": 0.43, "learning_rate": 2.924079072597632e-05, "loss": 0.1281, "step": 9110 }, { "epoch": 0.43, "learning_rate": 2.92403226561403e-05, "loss": 0.1381, "step": 9115 }, { "epoch": 0.43, "learning_rate": 2.9239854586304275e-05, "loss": 0.113, "step": 9120 }, { "epoch": 0.43, "learning_rate": 2.9239386516468258e-05, "loss": 0.1503, "step": 9125 }, { "epoch": 0.43, "learning_rate": 2.9238918446632238e-05, "loss": 0.2912, "step": 9130 }, { "epoch": 0.43, "learning_rate": 2.9238450376796218e-05, "loss": 0.2596, "step": 9135 }, { "epoch": 0.43, "learning_rate": 2.9237982306960198e-05, "loss": 0.3625, "step": 9140 }, { "epoch": 0.43, "learning_rate": 2.923751423712418e-05, "loss": 0.2824, "step": 9145 }, { "epoch": 0.43, "learning_rate": 2.923704616728816e-05, "loss": 0.8002, "step": 9150 }, { "epoch": 0.43, "learning_rate": 2.923657809745214e-05, "loss": 0.2374, "step": 9155 }, { "epoch": 0.43, "learning_rate": 2.923611002761612e-05, "loss": 0.1549, "step": 9160 }, { "epoch": 0.43, "learning_rate": 2.9235641957780104e-05, "loss": 0.0864, "step": 9165 }, { "epoch": 0.43, "learning_rate": 2.9235173887944083e-05, "loss": 0.1367, "step": 9170 }, { "epoch": 0.43, "learning_rate": 2.9234705818108063e-05, "loss": 0.1108, "step": 9175 }, { "epoch": 0.43, "learning_rate": 2.9234237748272043e-05, "loss": 0.1935, "step": 9180 }, { "epoch": 0.43, "learning_rate": 2.9233769678436023e-05, "loss": 0.1717, "step": 9185 }, { "epoch": 0.43, "learning_rate": 2.9233301608600003e-05, "loss": 0.2344, "step": 9190 }, { "epoch": 0.43, "learning_rate": 2.9232833538763982e-05, "loss": 0.2817, "step": 9195 }, { "epoch": 0.43, "learning_rate": 2.9232365468927966e-05, "loss": 0.6679, "step": 9200 }, { "epoch": 0.43, "learning_rate": 2.9231897399091945e-05, "loss": 0.2419, "step": 9205 }, { "epoch": 0.43, "learning_rate": 2.9231429329255925e-05, "loss": 0.1268, "step": 9210 }, { "epoch": 0.43, "learning_rate": 2.9230961259419905e-05, "loss": 0.1094, "step": 9215 }, { "epoch": 0.43, "learning_rate": 2.9230493189583888e-05, "loss": 0.1455, "step": 9220 }, { "epoch": 0.43, "learning_rate": 2.9230025119747868e-05, "loss": 0.2385, "step": 9225 }, { "epoch": 0.43, "learning_rate": 2.9229557049911848e-05, "loss": 0.1942, "step": 9230 }, { "epoch": 0.43, "learning_rate": 2.9229088980075828e-05, "loss": 0.2294, "step": 9235 }, { "epoch": 0.43, "learning_rate": 2.922862091023981e-05, "loss": 0.2644, "step": 9240 }, { "epoch": 0.43, "learning_rate": 2.9228152840403787e-05, "loss": 0.2645, "step": 9245 }, { "epoch": 0.43, "learning_rate": 2.9227684770567767e-05, "loss": 0.503, "step": 9250 }, { "epoch": 0.43, "learning_rate": 2.922721670073175e-05, "loss": 0.3093, "step": 9255 }, { "epoch": 0.43, "learning_rate": 2.922674863089573e-05, "loss": 0.0982, "step": 9260 }, { "epoch": 0.43, "learning_rate": 2.922628056105971e-05, "loss": 0.0931, "step": 9265 }, { "epoch": 0.43, "learning_rate": 2.922581249122369e-05, "loss": 0.2207, "step": 9270 }, { "epoch": 0.43, "learning_rate": 2.9225344421387673e-05, "loss": 0.2538, "step": 9275 }, { "epoch": 0.43, "learning_rate": 2.9224876351551653e-05, "loss": 0.2333, "step": 9280 }, { "epoch": 0.43, "learning_rate": 2.9224408281715633e-05, "loss": 0.1845, "step": 9285 }, { "epoch": 0.43, "learning_rate": 2.9223940211879612e-05, "loss": 0.2484, "step": 9290 }, { "epoch": 0.43, "learning_rate": 2.9223472142043596e-05, "loss": 0.3841, "step": 9295 }, { "epoch": 0.43, "learning_rate": 2.9223004072207576e-05, "loss": 0.5719, "step": 9300 }, { "epoch": 0.43, "learning_rate": 2.9222536002371555e-05, "loss": 0.2552, "step": 9305 }, { "epoch": 0.43, "learning_rate": 2.9222067932535535e-05, "loss": 0.0922, "step": 9310 }, { "epoch": 0.43, "learning_rate": 2.9221599862699515e-05, "loss": 0.1027, "step": 9315 }, { "epoch": 0.43, "learning_rate": 2.9221131792863495e-05, "loss": 0.2423, "step": 9320 }, { "epoch": 0.44, "learning_rate": 2.9220663723027475e-05, "loss": 0.2209, "step": 9325 }, { "epoch": 0.44, "learning_rate": 2.9220195653191458e-05, "loss": 0.1603, "step": 9330 }, { "epoch": 0.44, "learning_rate": 2.9219727583355438e-05, "loss": 0.2639, "step": 9335 }, { "epoch": 0.44, "learning_rate": 2.9219259513519417e-05, "loss": 0.3166, "step": 9340 }, { "epoch": 0.44, "learning_rate": 2.9218791443683397e-05, "loss": 0.3233, "step": 9345 }, { "epoch": 0.44, "learning_rate": 2.921832337384738e-05, "loss": 0.6509, "step": 9350 }, { "epoch": 0.44, "learning_rate": 2.921785530401136e-05, "loss": 0.2402, "step": 9355 }, { "epoch": 0.44, "learning_rate": 2.921738723417534e-05, "loss": 0.0998, "step": 9360 }, { "epoch": 0.44, "learning_rate": 2.921691916433932e-05, "loss": 0.121, "step": 9365 }, { "epoch": 0.44, "learning_rate": 2.92164510945033e-05, "loss": 0.1643, "step": 9370 }, { "epoch": 0.44, "learning_rate": 2.921598302466728e-05, "loss": 0.1153, "step": 9375 }, { "epoch": 0.44, "learning_rate": 2.921551495483126e-05, "loss": 0.2076, "step": 9380 }, { "epoch": 0.44, "learning_rate": 2.9215046884995243e-05, "loss": 0.2147, "step": 9385 }, { "epoch": 0.44, "learning_rate": 2.9214578815159222e-05, "loss": 0.3021, "step": 9390 }, { "epoch": 0.44, "learning_rate": 2.9214110745323202e-05, "loss": 0.4945, "step": 9395 }, { "epoch": 0.44, "learning_rate": 2.9213642675487182e-05, "loss": 0.7701, "step": 9400 }, { "epoch": 0.44, "learning_rate": 2.9213174605651165e-05, "loss": 0.2791, "step": 9405 }, { "epoch": 0.44, "learning_rate": 2.9212706535815145e-05, "loss": 0.1013, "step": 9410 }, { "epoch": 0.44, "learning_rate": 2.9212238465979125e-05, "loss": 0.1797, "step": 9415 }, { "epoch": 0.44, "learning_rate": 2.9211770396143105e-05, "loss": 0.196, "step": 9420 }, { "epoch": 0.44, "learning_rate": 2.9211302326307088e-05, "loss": 0.1942, "step": 9425 }, { "epoch": 0.44, "learning_rate": 2.9210834256471068e-05, "loss": 0.2213, "step": 9430 }, { "epoch": 0.44, "learning_rate": 2.9210366186635044e-05, "loss": 0.2232, "step": 9435 }, { "epoch": 0.44, "learning_rate": 2.9209898116799027e-05, "loss": 0.2296, "step": 9440 }, { "epoch": 0.44, "learning_rate": 2.9209430046963007e-05, "loss": 0.3679, "step": 9445 }, { "epoch": 0.44, "learning_rate": 2.9208961977126987e-05, "loss": 0.5784, "step": 9450 }, { "epoch": 0.44, "learning_rate": 2.9208493907290967e-05, "loss": 0.3411, "step": 9455 }, { "epoch": 0.44, "learning_rate": 2.920802583745495e-05, "loss": 0.0989, "step": 9460 }, { "epoch": 0.44, "learning_rate": 2.920755776761893e-05, "loss": 0.0827, "step": 9465 }, { "epoch": 0.44, "learning_rate": 2.920708969778291e-05, "loss": 0.1644, "step": 9470 }, { "epoch": 0.44, "learning_rate": 2.920662162794689e-05, "loss": 0.1767, "step": 9475 }, { "epoch": 0.44, "learning_rate": 2.9206153558110873e-05, "loss": 0.2319, "step": 9480 }, { "epoch": 0.44, "learning_rate": 2.9205685488274852e-05, "loss": 0.2015, "step": 9485 }, { "epoch": 0.44, "learning_rate": 2.9205217418438832e-05, "loss": 0.3777, "step": 9490 }, { "epoch": 0.44, "learning_rate": 2.9204749348602816e-05, "loss": 0.3779, "step": 9495 }, { "epoch": 0.44, "learning_rate": 2.9204281278766792e-05, "loss": 0.5985, "step": 9500 }, { "epoch": 0.44, "learning_rate": 2.9203813208930772e-05, "loss": 0.2136, "step": 9505 }, { "epoch": 0.44, "learning_rate": 2.920334513909475e-05, "loss": 0.1769, "step": 9510 }, { "epoch": 0.44, "learning_rate": 2.9202877069258735e-05, "loss": 0.2131, "step": 9515 }, { "epoch": 0.44, "learning_rate": 2.9202408999422715e-05, "loss": 0.2437, "step": 9520 }, { "epoch": 0.44, "learning_rate": 2.9201940929586694e-05, "loss": 0.1986, "step": 9525 }, { "epoch": 0.44, "learning_rate": 2.9201472859750674e-05, "loss": 0.1832, "step": 9530 }, { "epoch": 0.44, "learning_rate": 2.9201004789914657e-05, "loss": 0.3835, "step": 9535 }, { "epoch": 0.45, "learning_rate": 2.9200536720078637e-05, "loss": 0.3604, "step": 9540 }, { "epoch": 0.45, "learning_rate": 2.9200068650242617e-05, "loss": 0.4509, "step": 9545 }, { "epoch": 0.45, "learning_rate": 2.9199600580406597e-05, "loss": 0.4497, "step": 9550 }, { "epoch": 0.45, "learning_rate": 2.919913251057058e-05, "loss": 0.265, "step": 9555 }, { "epoch": 0.45, "learning_rate": 2.9198664440734557e-05, "loss": 0.1075, "step": 9560 }, { "epoch": 0.45, "learning_rate": 2.9198196370898536e-05, "loss": 0.246, "step": 9565 }, { "epoch": 0.45, "learning_rate": 2.919772830106252e-05, "loss": 0.1737, "step": 9570 }, { "epoch": 0.45, "learning_rate": 2.91972602312265e-05, "loss": 0.1635, "step": 9575 }, { "epoch": 0.45, "learning_rate": 2.919679216139048e-05, "loss": 0.1547, "step": 9580 }, { "epoch": 0.45, "learning_rate": 2.919632409155446e-05, "loss": 0.2552, "step": 9585 }, { "epoch": 0.45, "learning_rate": 2.9195856021718442e-05, "loss": 0.363, "step": 9590 }, { "epoch": 0.45, "learning_rate": 2.9195387951882422e-05, "loss": 0.2709, "step": 9595 }, { "epoch": 0.45, "learning_rate": 2.9194919882046402e-05, "loss": 0.5096, "step": 9600 }, { "epoch": 0.45, "learning_rate": 2.919445181221038e-05, "loss": 0.2245, "step": 9605 }, { "epoch": 0.45, "learning_rate": 2.9193983742374365e-05, "loss": 0.0959, "step": 9610 }, { "epoch": 0.45, "learning_rate": 2.9193515672538345e-05, "loss": 0.1363, "step": 9615 }, { "epoch": 0.45, "learning_rate": 2.9193047602702325e-05, "loss": 0.1798, "step": 9620 }, { "epoch": 0.45, "learning_rate": 2.9192579532866304e-05, "loss": 0.1677, "step": 9625 }, { "epoch": 0.45, "learning_rate": 2.9192111463030284e-05, "loss": 0.1836, "step": 9630 }, { "epoch": 0.45, "learning_rate": 2.9191643393194264e-05, "loss": 0.2168, "step": 9635 }, { "epoch": 0.45, "learning_rate": 2.9191175323358244e-05, "loss": 0.3292, "step": 9640 }, { "epoch": 0.45, "learning_rate": 2.9190707253522227e-05, "loss": 0.3587, "step": 9645 }, { "epoch": 0.45, "learning_rate": 2.9190239183686207e-05, "loss": 0.3841, "step": 9650 }, { "epoch": 0.45, "learning_rate": 2.9189771113850187e-05, "loss": 0.2487, "step": 9655 }, { "epoch": 0.45, "learning_rate": 2.9189303044014166e-05, "loss": 0.0936, "step": 9660 }, { "epoch": 0.45, "learning_rate": 2.918883497417815e-05, "loss": 0.137, "step": 9665 }, { "epoch": 0.45, "learning_rate": 2.918836690434213e-05, "loss": 0.1499, "step": 9670 }, { "epoch": 0.45, "learning_rate": 2.918789883450611e-05, "loss": 0.1184, "step": 9675 }, { "epoch": 0.45, "learning_rate": 2.9187430764670092e-05, "loss": 0.2235, "step": 9680 }, { "epoch": 0.45, "learning_rate": 2.9186962694834072e-05, "loss": 0.2345, "step": 9685 }, { "epoch": 0.45, "learning_rate": 2.918649462499805e-05, "loss": 0.2049, "step": 9690 }, { "epoch": 0.45, "learning_rate": 2.918602655516203e-05, "loss": 0.314, "step": 9695 }, { "epoch": 0.45, "learning_rate": 2.9185558485326012e-05, "loss": 0.6398, "step": 9700 }, { "epoch": 0.45, "learning_rate": 2.918509041548999e-05, "loss": 0.2686, "step": 9705 }, { "epoch": 0.45, "learning_rate": 2.918462234565397e-05, "loss": 0.0707, "step": 9710 }, { "epoch": 0.45, "learning_rate": 2.918415427581795e-05, "loss": 0.1081, "step": 9715 }, { "epoch": 0.45, "learning_rate": 2.9183686205981934e-05, "loss": 0.1142, "step": 9720 }, { "epoch": 0.45, "learning_rate": 2.9183218136145914e-05, "loss": 0.2019, "step": 9725 }, { "epoch": 0.45, "learning_rate": 2.9182750066309894e-05, "loss": 0.1826, "step": 9730 }, { "epoch": 0.45, "learning_rate": 2.9182281996473877e-05, "loss": 0.2352, "step": 9735 }, { "epoch": 0.45, "learning_rate": 2.9181813926637857e-05, "loss": 0.3715, "step": 9740 }, { "epoch": 0.45, "learning_rate": 2.9181345856801837e-05, "loss": 0.4708, "step": 9745 }, { "epoch": 0.45, "learning_rate": 2.9180877786965813e-05, "loss": 0.5184, "step": 9750 }, { "epoch": 0.46, "learning_rate": 2.9180409717129797e-05, "loss": 0.2204, "step": 9755 }, { "epoch": 0.46, "learning_rate": 2.9179941647293776e-05, "loss": 0.1351, "step": 9760 }, { "epoch": 0.46, "learning_rate": 2.9179473577457756e-05, "loss": 0.1712, "step": 9765 }, { "epoch": 0.46, "learning_rate": 2.9179005507621736e-05, "loss": 0.1685, "step": 9770 }, { "epoch": 0.46, "learning_rate": 2.917853743778572e-05, "loss": 0.1494, "step": 9775 }, { "epoch": 0.46, "learning_rate": 2.91780693679497e-05, "loss": 0.2308, "step": 9780 }, { "epoch": 0.46, "learning_rate": 2.917760129811368e-05, "loss": 0.3646, "step": 9785 }, { "epoch": 0.46, "learning_rate": 2.917713322827766e-05, "loss": 0.2776, "step": 9790 }, { "epoch": 0.46, "learning_rate": 2.9176665158441642e-05, "loss": 0.317, "step": 9795 }, { "epoch": 0.46, "learning_rate": 2.917619708860562e-05, "loss": 0.4983, "step": 9800 }, { "epoch": 0.46, "learning_rate": 2.91757290187696e-05, "loss": 0.2496, "step": 9805 }, { "epoch": 0.46, "learning_rate": 2.9175260948933585e-05, "loss": 0.0691, "step": 9810 }, { "epoch": 0.46, "learning_rate": 2.917479287909756e-05, "loss": 0.157, "step": 9815 }, { "epoch": 0.46, "learning_rate": 2.917432480926154e-05, "loss": 0.1621, "step": 9820 }, { "epoch": 0.46, "learning_rate": 2.917385673942552e-05, "loss": 0.1488, "step": 9825 }, { "epoch": 0.46, "learning_rate": 2.9173388669589504e-05, "loss": 0.2045, "step": 9830 }, { "epoch": 0.46, "learning_rate": 2.9172920599753484e-05, "loss": 0.2649, "step": 9835 }, { "epoch": 0.46, "learning_rate": 2.9172452529917464e-05, "loss": 0.2739, "step": 9840 }, { "epoch": 0.46, "learning_rate": 2.9171984460081443e-05, "loss": 0.4539, "step": 9845 }, { "epoch": 0.46, "learning_rate": 2.9171516390245427e-05, "loss": 0.458, "step": 9850 }, { "epoch": 0.46, "learning_rate": 2.9171048320409406e-05, "loss": 0.2403, "step": 9855 }, { "epoch": 0.46, "learning_rate": 2.9170580250573386e-05, "loss": 0.0825, "step": 9860 }, { "epoch": 0.46, "learning_rate": 2.917011218073737e-05, "loss": 0.1144, "step": 9865 }, { "epoch": 0.46, "learning_rate": 2.916964411090135e-05, "loss": 0.1379, "step": 9870 }, { "epoch": 0.46, "learning_rate": 2.916917604106533e-05, "loss": 0.1643, "step": 9875 }, { "epoch": 0.46, "learning_rate": 2.9168707971229306e-05, "loss": 0.1821, "step": 9880 }, { "epoch": 0.46, "learning_rate": 2.916823990139329e-05, "loss": 0.2061, "step": 9885 }, { "epoch": 0.46, "learning_rate": 2.916777183155727e-05, "loss": 0.2136, "step": 9890 }, { "epoch": 0.46, "learning_rate": 2.916730376172125e-05, "loss": 0.2821, "step": 9895 }, { "epoch": 0.46, "learning_rate": 2.9166835691885228e-05, "loss": 0.6509, "step": 9900 }, { "epoch": 0.46, "learning_rate": 2.916636762204921e-05, "loss": 0.241, "step": 9905 }, { "epoch": 0.46, "learning_rate": 2.916589955221319e-05, "loss": 0.1254, "step": 9910 }, { "epoch": 0.46, "learning_rate": 2.916543148237717e-05, "loss": 0.1624, "step": 9915 }, { "epoch": 0.46, "learning_rate": 2.9164963412541154e-05, "loss": 0.1871, "step": 9920 }, { "epoch": 0.46, "learning_rate": 2.9164495342705134e-05, "loss": 0.1965, "step": 9925 }, { "epoch": 0.46, "learning_rate": 2.9164027272869114e-05, "loss": 0.2924, "step": 9930 }, { "epoch": 0.46, "learning_rate": 2.9163559203033094e-05, "loss": 0.2327, "step": 9935 }, { "epoch": 0.46, "learning_rate": 2.9163091133197073e-05, "loss": 0.3731, "step": 9940 }, { "epoch": 0.46, "learning_rate": 2.9162623063361053e-05, "loss": 0.411, "step": 9945 }, { "epoch": 0.46, "learning_rate": 2.9162154993525033e-05, "loss": 0.4525, "step": 9950 }, { "epoch": 0.46, "learning_rate": 2.9161686923689013e-05, "loss": 0.1787, "step": 9955 }, { "epoch": 0.46, "learning_rate": 2.9161218853852996e-05, "loss": 0.1399, "step": 9960 }, { "epoch": 0.46, "learning_rate": 2.9160750784016976e-05, "loss": 0.1591, "step": 9965 }, { "epoch": 0.47, "learning_rate": 2.9160282714180956e-05, "loss": 0.2047, "step": 9970 }, { "epoch": 0.47, "learning_rate": 2.9159814644344936e-05, "loss": 0.1774, "step": 9975 }, { "epoch": 0.47, "learning_rate": 2.915934657450892e-05, "loss": 0.2215, "step": 9980 }, { "epoch": 0.47, "learning_rate": 2.91588785046729e-05, "loss": 0.2193, "step": 9985 }, { "epoch": 0.47, "learning_rate": 2.915841043483688e-05, "loss": 0.2754, "step": 9990 }, { "epoch": 0.47, "learning_rate": 2.915794236500086e-05, "loss": 0.3376, "step": 9995 }, { "epoch": 0.47, "learning_rate": 2.915747429516484e-05, "loss": 0.5115, "step": 10000 }, { "epoch": 0.47, "learning_rate": 2.9157006225328818e-05, "loss": 0.2713, "step": 10005 }, { "epoch": 0.47, "learning_rate": 2.9156538155492798e-05, "loss": 0.1161, "step": 10010 }, { "epoch": 0.47, "learning_rate": 2.915607008565678e-05, "loss": 0.1312, "step": 10015 }, { "epoch": 0.47, "learning_rate": 2.915560201582076e-05, "loss": 0.1382, "step": 10020 }, { "epoch": 0.47, "learning_rate": 2.915513394598474e-05, "loss": 0.2032, "step": 10025 }, { "epoch": 0.47, "learning_rate": 2.915466587614872e-05, "loss": 0.2508, "step": 10030 }, { "epoch": 0.47, "learning_rate": 2.9154197806312704e-05, "loss": 0.3753, "step": 10035 }, { "epoch": 0.47, "learning_rate": 2.9153729736476683e-05, "loss": 0.3759, "step": 10040 }, { "epoch": 0.47, "learning_rate": 2.9153261666640663e-05, "loss": 0.3609, "step": 10045 }, { "epoch": 0.47, "learning_rate": 2.9152793596804646e-05, "loss": 0.5963, "step": 10050 }, { "epoch": 0.47, "learning_rate": 2.9152325526968626e-05, "loss": 0.2251, "step": 10055 }, { "epoch": 0.47, "learning_rate": 2.9151857457132606e-05, "loss": 0.119, "step": 10060 }, { "epoch": 0.47, "learning_rate": 2.9151389387296586e-05, "loss": 0.0838, "step": 10065 }, { "epoch": 0.47, "learning_rate": 2.9150921317460566e-05, "loss": 0.0799, "step": 10070 }, { "epoch": 0.47, "learning_rate": 2.9150453247624546e-05, "loss": 0.2083, "step": 10075 }, { "epoch": 0.47, "learning_rate": 2.9149985177788525e-05, "loss": 0.258, "step": 10080 }, { "epoch": 0.47, "learning_rate": 2.9149517107952505e-05, "loss": 0.2772, "step": 10085 }, { "epoch": 0.47, "learning_rate": 2.914904903811649e-05, "loss": 0.2897, "step": 10090 }, { "epoch": 0.47, "learning_rate": 2.9148580968280468e-05, "loss": 0.4434, "step": 10095 }, { "epoch": 0.47, "learning_rate": 2.9148112898444448e-05, "loss": 0.5291, "step": 10100 }, { "epoch": 0.47, "learning_rate": 2.914764482860843e-05, "loss": 0.295, "step": 10105 }, { "epoch": 0.47, "learning_rate": 2.914717675877241e-05, "loss": 0.1425, "step": 10110 }, { "epoch": 0.47, "learning_rate": 2.914670868893639e-05, "loss": 0.1711, "step": 10115 }, { "epoch": 0.47, "learning_rate": 2.914624061910037e-05, "loss": 0.136, "step": 10120 }, { "epoch": 0.47, "learning_rate": 2.9145772549264354e-05, "loss": 0.1828, "step": 10125 }, { "epoch": 0.47, "learning_rate": 2.914530447942833e-05, "loss": 0.2204, "step": 10130 }, { "epoch": 0.47, "learning_rate": 2.914483640959231e-05, "loss": 0.1504, "step": 10135 }, { "epoch": 0.47, "learning_rate": 2.914436833975629e-05, "loss": 0.1957, "step": 10140 }, { "epoch": 0.47, "learning_rate": 2.9143900269920273e-05, "loss": 0.3153, "step": 10145 }, { "epoch": 0.47, "learning_rate": 2.9143432200084253e-05, "loss": 0.4669, "step": 10150 }, { "epoch": 0.47, "learning_rate": 2.9142964130248233e-05, "loss": 0.2938, "step": 10155 }, { "epoch": 0.47, "learning_rate": 2.9142496060412213e-05, "loss": 0.0995, "step": 10160 }, { "epoch": 0.47, "learning_rate": 2.9142027990576196e-05, "loss": 0.1582, "step": 10165 }, { "epoch": 0.47, "learning_rate": 2.9141559920740176e-05, "loss": 0.1777, "step": 10170 }, { "epoch": 0.47, "learning_rate": 2.9141091850904155e-05, "loss": 0.0941, "step": 10175 }, { "epoch": 0.48, "learning_rate": 2.914062378106814e-05, "loss": 0.2079, "step": 10180 }, { "epoch": 0.48, "learning_rate": 2.914015571123212e-05, "loss": 0.3759, "step": 10185 }, { "epoch": 0.48, "learning_rate": 2.9139687641396098e-05, "loss": 0.2368, "step": 10190 }, { "epoch": 0.48, "learning_rate": 2.9139219571560075e-05, "loss": 0.3776, "step": 10195 }, { "epoch": 0.48, "learning_rate": 2.9138751501724058e-05, "loss": 0.6228, "step": 10200 }, { "epoch": 0.48, "learning_rate": 2.9138283431888038e-05, "loss": 0.1931, "step": 10205 }, { "epoch": 0.48, "learning_rate": 2.9137815362052018e-05, "loss": 0.0513, "step": 10210 }, { "epoch": 0.48, "learning_rate": 2.9137347292215997e-05, "loss": 0.1184, "step": 10215 }, { "epoch": 0.48, "learning_rate": 2.913687922237998e-05, "loss": 0.1752, "step": 10220 }, { "epoch": 0.48, "learning_rate": 2.913641115254396e-05, "loss": 0.1899, "step": 10225 }, { "epoch": 0.48, "learning_rate": 2.913594308270794e-05, "loss": 0.2375, "step": 10230 }, { "epoch": 0.48, "learning_rate": 2.9135475012871923e-05, "loss": 0.2201, "step": 10235 }, { "epoch": 0.48, "learning_rate": 2.9135006943035903e-05, "loss": 0.3693, "step": 10240 }, { "epoch": 0.48, "learning_rate": 2.9134538873199883e-05, "loss": 0.3074, "step": 10245 }, { "epoch": 0.48, "learning_rate": 2.9134070803363863e-05, "loss": 0.5954, "step": 10250 }, { "epoch": 0.48, "learning_rate": 2.9133602733527843e-05, "loss": 0.2433, "step": 10255 }, { "epoch": 0.48, "learning_rate": 2.9133134663691822e-05, "loss": 0.0825, "step": 10260 }, { "epoch": 0.48, "learning_rate": 2.9132666593855802e-05, "loss": 0.1623, "step": 10265 }, { "epoch": 0.48, "learning_rate": 2.9132198524019782e-05, "loss": 0.1625, "step": 10270 }, { "epoch": 0.48, "learning_rate": 2.9131730454183765e-05, "loss": 0.2033, "step": 10275 }, { "epoch": 0.48, "learning_rate": 2.9131262384347745e-05, "loss": 0.303, "step": 10280 }, { "epoch": 0.48, "learning_rate": 2.9130794314511725e-05, "loss": 0.4226, "step": 10285 }, { "epoch": 0.48, "learning_rate": 2.9130326244675708e-05, "loss": 0.2968, "step": 10290 }, { "epoch": 0.48, "learning_rate": 2.9129858174839688e-05, "loss": 0.4019, "step": 10295 }, { "epoch": 0.48, "learning_rate": 2.9129390105003668e-05, "loss": 0.5179, "step": 10300 }, { "epoch": 0.48, "learning_rate": 2.9128922035167648e-05, "loss": 0.3199, "step": 10305 }, { "epoch": 0.48, "learning_rate": 2.912845396533163e-05, "loss": 0.1018, "step": 10310 }, { "epoch": 0.48, "learning_rate": 2.912798589549561e-05, "loss": 0.0952, "step": 10315 }, { "epoch": 0.48, "learning_rate": 2.9127517825659587e-05, "loss": 0.0911, "step": 10320 }, { "epoch": 0.48, "learning_rate": 2.9127049755823567e-05, "loss": 0.1483, "step": 10325 }, { "epoch": 0.48, "learning_rate": 2.912658168598755e-05, "loss": 0.1974, "step": 10330 }, { "epoch": 0.48, "learning_rate": 2.912611361615153e-05, "loss": 0.2161, "step": 10335 }, { "epoch": 0.48, "learning_rate": 2.912564554631551e-05, "loss": 0.3492, "step": 10340 }, { "epoch": 0.48, "learning_rate": 2.9125177476479493e-05, "loss": 0.4528, "step": 10345 }, { "epoch": 0.48, "learning_rate": 2.9124709406643473e-05, "loss": 0.6735, "step": 10350 }, { "epoch": 0.48, "learning_rate": 2.9124241336807453e-05, "loss": 0.2464, "step": 10355 }, { "epoch": 0.48, "learning_rate": 2.9123773266971432e-05, "loss": 0.1387, "step": 10360 }, { "epoch": 0.48, "learning_rate": 2.9123305197135416e-05, "loss": 0.1129, "step": 10365 }, { "epoch": 0.48, "learning_rate": 2.9122837127299395e-05, "loss": 0.149, "step": 10370 }, { "epoch": 0.48, "learning_rate": 2.9122369057463375e-05, "loss": 0.1488, "step": 10375 }, { "epoch": 0.48, "learning_rate": 2.9121900987627355e-05, "loss": 0.2723, "step": 10380 }, { "epoch": 0.48, "learning_rate": 2.9121432917791335e-05, "loss": 0.2941, "step": 10385 }, { "epoch": 0.48, "learning_rate": 2.9120964847955315e-05, "loss": 0.289, "step": 10390 }, { "epoch": 0.49, "learning_rate": 2.9120496778119295e-05, "loss": 0.2986, "step": 10395 }, { "epoch": 0.49, "learning_rate": 2.9120028708283274e-05, "loss": 0.4768, "step": 10400 }, { "epoch": 0.49, "learning_rate": 2.9119560638447258e-05, "loss": 0.2318, "step": 10405 }, { "epoch": 0.49, "learning_rate": 2.9119092568611237e-05, "loss": 0.087, "step": 10410 }, { "epoch": 0.49, "learning_rate": 2.9118624498775217e-05, "loss": 0.1399, "step": 10415 }, { "epoch": 0.49, "learning_rate": 2.91181564289392e-05, "loss": 0.1293, "step": 10420 }, { "epoch": 0.49, "learning_rate": 2.911768835910318e-05, "loss": 0.1353, "step": 10425 }, { "epoch": 0.49, "learning_rate": 2.911722028926716e-05, "loss": 0.2022, "step": 10430 }, { "epoch": 0.49, "learning_rate": 2.911675221943114e-05, "loss": 0.289, "step": 10435 }, { "epoch": 0.49, "learning_rate": 2.9116284149595123e-05, "loss": 0.5087, "step": 10440 }, { "epoch": 0.49, "learning_rate": 2.91158160797591e-05, "loss": 0.3206, "step": 10445 }, { "epoch": 0.49, "learning_rate": 2.911534800992308e-05, "loss": 0.597, "step": 10450 }, { "epoch": 0.49, "learning_rate": 2.911487994008706e-05, "loss": 0.3211, "step": 10455 }, { "epoch": 0.49, "learning_rate": 2.9114411870251042e-05, "loss": 0.1773, "step": 10460 }, { "epoch": 0.49, "learning_rate": 2.9113943800415022e-05, "loss": 0.1689, "step": 10465 }, { "epoch": 0.49, "learning_rate": 2.9113475730579002e-05, "loss": 0.1645, "step": 10470 }, { "epoch": 0.49, "learning_rate": 2.9113007660742985e-05, "loss": 0.2574, "step": 10475 }, { "epoch": 0.49, "learning_rate": 2.9112539590906965e-05, "loss": 0.2319, "step": 10480 }, { "epoch": 0.49, "learning_rate": 2.9112071521070945e-05, "loss": 0.2182, "step": 10485 }, { "epoch": 0.49, "learning_rate": 2.9111603451234925e-05, "loss": 0.2716, "step": 10490 }, { "epoch": 0.49, "learning_rate": 2.9111135381398908e-05, "loss": 0.2271, "step": 10495 }, { "epoch": 0.49, "learning_rate": 2.9110667311562888e-05, "loss": 0.6028, "step": 10500 }, { "epoch": 0.49, "learning_rate": 2.9110199241726867e-05, "loss": 0.2078, "step": 10505 }, { "epoch": 0.49, "learning_rate": 2.9109731171890844e-05, "loss": 0.1231, "step": 10510 }, { "epoch": 0.49, "learning_rate": 2.9109263102054827e-05, "loss": 0.1728, "step": 10515 }, { "epoch": 0.49, "learning_rate": 2.9108795032218807e-05, "loss": 0.1528, "step": 10520 }, { "epoch": 0.49, "learning_rate": 2.9108326962382787e-05, "loss": 0.2504, "step": 10525 }, { "epoch": 0.49, "learning_rate": 2.910785889254677e-05, "loss": 0.2285, "step": 10530 }, { "epoch": 0.49, "learning_rate": 2.910739082271075e-05, "loss": 0.1583, "step": 10535 }, { "epoch": 0.49, "learning_rate": 2.910692275287473e-05, "loss": 0.1986, "step": 10540 }, { "epoch": 0.49, "learning_rate": 2.910645468303871e-05, "loss": 0.3468, "step": 10545 }, { "epoch": 0.49, "learning_rate": 2.9105986613202693e-05, "loss": 0.3821, "step": 10550 }, { "epoch": 0.49, "learning_rate": 2.9105518543366672e-05, "loss": 0.2888, "step": 10555 }, { "epoch": 0.49, "learning_rate": 2.9105050473530652e-05, "loss": 0.1296, "step": 10560 }, { "epoch": 0.49, "learning_rate": 2.9104582403694632e-05, "loss": 0.149, "step": 10565 }, { "epoch": 0.49, "learning_rate": 2.9104114333858615e-05, "loss": 0.1573, "step": 10570 }, { "epoch": 0.49, "learning_rate": 2.910364626402259e-05, "loss": 0.1666, "step": 10575 }, { "epoch": 0.49, "learning_rate": 2.910317819418657e-05, "loss": 0.2131, "step": 10580 }, { "epoch": 0.49, "learning_rate": 2.910271012435055e-05, "loss": 0.3106, "step": 10585 }, { "epoch": 0.49, "learning_rate": 2.9102242054514534e-05, "loss": 0.2393, "step": 10590 }, { "epoch": 0.49, "learning_rate": 2.9101773984678514e-05, "loss": 0.3747, "step": 10595 }, { "epoch": 0.49, "learning_rate": 2.9101305914842494e-05, "loss": 0.5276, "step": 10600 }, { "epoch": 0.49, "learning_rate": 2.9100837845006477e-05, "loss": 0.2109, "step": 10605 }, { "epoch": 0.5, "learning_rate": 2.9100369775170457e-05, "loss": 0.1697, "step": 10610 }, { "epoch": 0.5, "learning_rate": 2.9099901705334437e-05, "loss": 0.1781, "step": 10615 }, { "epoch": 0.5, "learning_rate": 2.9099433635498417e-05, "loss": 0.1481, "step": 10620 }, { "epoch": 0.5, "learning_rate": 2.90989655656624e-05, "loss": 0.1875, "step": 10625 }, { "epoch": 0.5, "learning_rate": 2.909849749582638e-05, "loss": 0.23, "step": 10630 }, { "epoch": 0.5, "learning_rate": 2.9098029425990356e-05, "loss": 0.2837, "step": 10635 }, { "epoch": 0.5, "learning_rate": 2.9097561356154336e-05, "loss": 0.3611, "step": 10640 }, { "epoch": 0.5, "learning_rate": 2.909709328631832e-05, "loss": 0.3145, "step": 10645 }, { "epoch": 0.5, "learning_rate": 2.90966252164823e-05, "loss": 0.7858, "step": 10650 }, { "epoch": 0.5, "learning_rate": 2.909615714664628e-05, "loss": 0.2585, "step": 10655 }, { "epoch": 0.5, "learning_rate": 2.9095689076810262e-05, "loss": 0.1072, "step": 10660 }, { "epoch": 0.5, "learning_rate": 2.9095221006974242e-05, "loss": 0.0863, "step": 10665 }, { "epoch": 0.5, "learning_rate": 2.9094752937138222e-05, "loss": 0.2089, "step": 10670 }, { "epoch": 0.5, "learning_rate": 2.90942848673022e-05, "loss": 0.1946, "step": 10675 }, { "epoch": 0.5, "learning_rate": 2.9093816797466185e-05, "loss": 0.1804, "step": 10680 }, { "epoch": 0.5, "learning_rate": 2.9093348727630165e-05, "loss": 0.2488, "step": 10685 }, { "epoch": 0.5, "learning_rate": 2.9092880657794144e-05, "loss": 0.3446, "step": 10690 }, { "epoch": 0.5, "learning_rate": 2.9092412587958124e-05, "loss": 0.4518, "step": 10695 }, { "epoch": 0.5, "learning_rate": 2.9091944518122104e-05, "loss": 0.5246, "step": 10700 }, { "epoch": 0.5, "learning_rate": 2.9091476448286084e-05, "loss": 0.2917, "step": 10705 }, { "epoch": 0.5, "learning_rate": 2.9091008378450064e-05, "loss": 0.1728, "step": 10710 }, { "epoch": 0.5, "learning_rate": 2.9090540308614047e-05, "loss": 0.1577, "step": 10715 }, { "epoch": 0.5, "learning_rate": 2.9090072238778027e-05, "loss": 0.0824, "step": 10720 }, { "epoch": 0.5, "learning_rate": 2.9089604168942007e-05, "loss": 0.1226, "step": 10725 }, { "epoch": 0.5, "learning_rate": 2.9089136099105986e-05, "loss": 0.2118, "step": 10730 }, { "epoch": 0.5, "learning_rate": 2.908866802926997e-05, "loss": 0.162, "step": 10735 }, { "epoch": 0.5, "learning_rate": 2.908819995943395e-05, "loss": 0.3048, "step": 10740 }, { "epoch": 0.5, "learning_rate": 2.908773188959793e-05, "loss": 0.2798, "step": 10745 }, { "epoch": 0.5, "learning_rate": 2.908726381976191e-05, "loss": 0.4306, "step": 10750 }, { "epoch": 0.5, "learning_rate": 2.9086795749925892e-05, "loss": 0.2259, "step": 10755 }, { "epoch": 0.5, "learning_rate": 2.908632768008987e-05, "loss": 0.116, "step": 10760 }, { "epoch": 0.5, "learning_rate": 2.908585961025385e-05, "loss": 0.108, "step": 10765 }, { "epoch": 0.5, "learning_rate": 2.9085391540417828e-05, "loss": 0.1246, "step": 10770 }, { "epoch": 0.5, "learning_rate": 2.908492347058181e-05, "loss": 0.1518, "step": 10775 }, { "epoch": 0.5, "learning_rate": 2.908445540074579e-05, "loss": 0.1694, "step": 10780 }, { "epoch": 0.5, "learning_rate": 2.908398733090977e-05, "loss": 0.3221, "step": 10785 }, { "epoch": 0.5, "learning_rate": 2.9083519261073754e-05, "loss": 0.2946, "step": 10790 }, { "epoch": 0.5, "learning_rate": 2.9083051191237734e-05, "loss": 0.4544, "step": 10795 }, { "epoch": 0.5, "learning_rate": 2.9082583121401714e-05, "loss": 0.6625, "step": 10800 }, { "epoch": 0.5, "learning_rate": 2.9082115051565694e-05, "loss": 0.3402, "step": 10805 }, { "epoch": 0.5, "learning_rate": 2.9081646981729677e-05, "loss": 0.1606, "step": 10810 }, { "epoch": 0.5, "learning_rate": 2.9081178911893657e-05, "loss": 0.1431, "step": 10815 }, { "epoch": 0.5, "learning_rate": 2.9080710842057637e-05, "loss": 0.1159, "step": 10820 }, { "epoch": 0.51, "learning_rate": 2.9080242772221613e-05, "loss": 0.171, "step": 10825 }, { "epoch": 0.51, "learning_rate": 2.9079774702385596e-05, "loss": 0.143, "step": 10830 }, { "epoch": 0.51, "learning_rate": 2.9079306632549576e-05, "loss": 0.3442, "step": 10835 }, { "epoch": 0.51, "learning_rate": 2.9078838562713556e-05, "loss": 0.3745, "step": 10840 }, { "epoch": 0.51, "learning_rate": 2.907837049287754e-05, "loss": 0.2238, "step": 10845 }, { "epoch": 0.51, "learning_rate": 2.907790242304152e-05, "loss": 0.558, "step": 10850 }, { "epoch": 0.51, "learning_rate": 2.90774343532055e-05, "loss": 0.1763, "step": 10855 }, { "epoch": 0.51, "learning_rate": 2.907696628336948e-05, "loss": 0.0905, "step": 10860 }, { "epoch": 0.51, "learning_rate": 2.9076498213533462e-05, "loss": 0.1248, "step": 10865 }, { "epoch": 0.51, "learning_rate": 2.907603014369744e-05, "loss": 0.0964, "step": 10870 }, { "epoch": 0.51, "learning_rate": 2.907556207386142e-05, "loss": 0.1588, "step": 10875 }, { "epoch": 0.51, "learning_rate": 2.90750940040254e-05, "loss": 0.2716, "step": 10880 }, { "epoch": 0.51, "learning_rate": 2.9074625934189384e-05, "loss": 0.1772, "step": 10885 }, { "epoch": 0.51, "learning_rate": 2.907415786435336e-05, "loss": 0.3076, "step": 10890 }, { "epoch": 0.51, "learning_rate": 2.907368979451734e-05, "loss": 0.3516, "step": 10895 }, { "epoch": 0.51, "learning_rate": 2.9073221724681324e-05, "loss": 0.651, "step": 10900 }, { "epoch": 0.51, "learning_rate": 2.9072753654845304e-05, "loss": 0.377, "step": 10905 }, { "epoch": 0.51, "learning_rate": 2.9072285585009283e-05, "loss": 0.1039, "step": 10910 }, { "epoch": 0.51, "learning_rate": 2.9071817515173263e-05, "loss": 0.158, "step": 10915 }, { "epoch": 0.51, "learning_rate": 2.9071349445337247e-05, "loss": 0.1531, "step": 10920 }, { "epoch": 0.51, "learning_rate": 2.9070881375501226e-05, "loss": 0.1671, "step": 10925 }, { "epoch": 0.51, "learning_rate": 2.9070413305665206e-05, "loss": 0.164, "step": 10930 }, { "epoch": 0.51, "learning_rate": 2.9069945235829186e-05, "loss": 0.2341, "step": 10935 }, { "epoch": 0.51, "learning_rate": 2.906947716599317e-05, "loss": 0.3567, "step": 10940 }, { "epoch": 0.51, "learning_rate": 2.906900909615715e-05, "loss": 0.4844, "step": 10945 }, { "epoch": 0.51, "learning_rate": 2.9068541026321125e-05, "loss": 0.5868, "step": 10950 }, { "epoch": 0.51, "learning_rate": 2.9068072956485105e-05, "loss": 0.2721, "step": 10955 }, { "epoch": 0.51, "learning_rate": 2.906760488664909e-05, "loss": 0.0986, "step": 10960 }, { "epoch": 0.51, "learning_rate": 2.9067136816813068e-05, "loss": 0.128, "step": 10965 }, { "epoch": 0.51, "learning_rate": 2.9066668746977048e-05, "loss": 0.2164, "step": 10970 }, { "epoch": 0.51, "learning_rate": 2.906620067714103e-05, "loss": 0.1356, "step": 10975 }, { "epoch": 0.51, "learning_rate": 2.906573260730501e-05, "loss": 0.1868, "step": 10980 }, { "epoch": 0.51, "learning_rate": 2.906526453746899e-05, "loss": 0.1627, "step": 10985 }, { "epoch": 0.51, "learning_rate": 2.906479646763297e-05, "loss": 0.2136, "step": 10990 }, { "epoch": 0.51, "learning_rate": 2.9064328397796954e-05, "loss": 0.3246, "step": 10995 }, { "epoch": 0.51, "learning_rate": 2.9063860327960934e-05, "loss": 0.6251, "step": 11000 }, { "epoch": 0.51, "learning_rate": 2.9063392258124914e-05, "loss": 0.2857, "step": 11005 }, { "epoch": 0.51, "learning_rate": 2.9062924188288893e-05, "loss": 0.0919, "step": 11010 }, { "epoch": 0.51, "learning_rate": 2.9062456118452873e-05, "loss": 0.1208, "step": 11015 }, { "epoch": 0.51, "learning_rate": 2.9061988048616853e-05, "loss": 0.2048, "step": 11020 }, { "epoch": 0.51, "learning_rate": 2.9061519978780833e-05, "loss": 0.1597, "step": 11025 }, { "epoch": 0.51, "learning_rate": 2.9061051908944816e-05, "loss": 0.1806, "step": 11030 }, { "epoch": 0.51, "learning_rate": 2.9060583839108796e-05, "loss": 0.2313, "step": 11035 }, { "epoch": 0.52, "learning_rate": 2.9060115769272776e-05, "loss": 0.3312, "step": 11040 }, { "epoch": 0.52, "learning_rate": 2.9059647699436756e-05, "loss": 0.3499, "step": 11045 }, { "epoch": 0.52, "learning_rate": 2.905917962960074e-05, "loss": 0.502, "step": 11050 }, { "epoch": 0.52, "learning_rate": 2.905871155976472e-05, "loss": 0.2127, "step": 11055 }, { "epoch": 0.52, "learning_rate": 2.90582434899287e-05, "loss": 0.0773, "step": 11060 }, { "epoch": 0.52, "learning_rate": 2.9057775420092678e-05, "loss": 0.2171, "step": 11065 }, { "epoch": 0.52, "learning_rate": 2.905730735025666e-05, "loss": 0.1632, "step": 11070 }, { "epoch": 0.52, "learning_rate": 2.905683928042064e-05, "loss": 0.1804, "step": 11075 }, { "epoch": 0.52, "learning_rate": 2.9056371210584618e-05, "loss": 0.1888, "step": 11080 }, { "epoch": 0.52, "learning_rate": 2.90559031407486e-05, "loss": 0.1782, "step": 11085 }, { "epoch": 0.52, "learning_rate": 2.905543507091258e-05, "loss": 0.2457, "step": 11090 }, { "epoch": 0.52, "learning_rate": 2.905496700107656e-05, "loss": 0.3533, "step": 11095 }, { "epoch": 0.52, "learning_rate": 2.905449893124054e-05, "loss": 0.662, "step": 11100 }, { "epoch": 0.52, "learning_rate": 2.9054030861404523e-05, "loss": 0.2686, "step": 11105 }, { "epoch": 0.52, "learning_rate": 2.9053562791568503e-05, "loss": 0.0685, "step": 11110 }, { "epoch": 0.52, "learning_rate": 2.9053094721732483e-05, "loss": 0.1569, "step": 11115 }, { "epoch": 0.52, "learning_rate": 2.9052626651896463e-05, "loss": 0.1152, "step": 11120 }, { "epoch": 0.52, "learning_rate": 2.9052158582060446e-05, "loss": 0.1515, "step": 11125 }, { "epoch": 0.52, "learning_rate": 2.9051690512224426e-05, "loss": 0.242, "step": 11130 }, { "epoch": 0.52, "learning_rate": 2.9051222442388406e-05, "loss": 0.2395, "step": 11135 }, { "epoch": 0.52, "learning_rate": 2.9050754372552386e-05, "loss": 0.2851, "step": 11140 }, { "epoch": 0.52, "learning_rate": 2.9050286302716365e-05, "loss": 0.347, "step": 11145 }, { "epoch": 0.52, "learning_rate": 2.9049818232880345e-05, "loss": 0.4843, "step": 11150 }, { "epoch": 0.52, "learning_rate": 2.9049350163044325e-05, "loss": 0.225, "step": 11155 }, { "epoch": 0.52, "learning_rate": 2.9048882093208308e-05, "loss": 0.1013, "step": 11160 }, { "epoch": 0.52, "learning_rate": 2.9048414023372288e-05, "loss": 0.1425, "step": 11165 }, { "epoch": 0.52, "learning_rate": 2.9047945953536268e-05, "loss": 0.2072, "step": 11170 }, { "epoch": 0.52, "learning_rate": 2.9047477883700248e-05, "loss": 0.1754, "step": 11175 }, { "epoch": 0.52, "learning_rate": 2.904700981386423e-05, "loss": 0.1523, "step": 11180 }, { "epoch": 0.52, "learning_rate": 2.904654174402821e-05, "loss": 0.2163, "step": 11185 }, { "epoch": 0.52, "learning_rate": 2.904607367419219e-05, "loss": 0.3456, "step": 11190 }, { "epoch": 0.52, "learning_rate": 2.904560560435617e-05, "loss": 0.3785, "step": 11195 }, { "epoch": 0.52, "learning_rate": 2.9045137534520154e-05, "loss": 0.571, "step": 11200 }, { "epoch": 0.52, "learning_rate": 2.904466946468413e-05, "loss": 0.2442, "step": 11205 }, { "epoch": 0.52, "learning_rate": 2.904420139484811e-05, "loss": 0.1094, "step": 11210 }, { "epoch": 0.52, "learning_rate": 2.9043733325012093e-05, "loss": 0.1536, "step": 11215 }, { "epoch": 0.52, "learning_rate": 2.9043265255176073e-05, "loss": 0.1979, "step": 11220 }, { "epoch": 0.52, "learning_rate": 2.9042797185340053e-05, "loss": 0.1617, "step": 11225 }, { "epoch": 0.52, "learning_rate": 2.9042329115504032e-05, "loss": 0.2681, "step": 11230 }, { "epoch": 0.52, "learning_rate": 2.9041861045668016e-05, "loss": 0.2328, "step": 11235 }, { "epoch": 0.52, "learning_rate": 2.9041392975831995e-05, "loss": 0.2601, "step": 11240 }, { "epoch": 0.52, "learning_rate": 2.9040924905995975e-05, "loss": 0.2401, "step": 11245 }, { "epoch": 0.52, "learning_rate": 2.9040456836159955e-05, "loss": 0.5136, "step": 11250 }, { "epoch": 0.53, "learning_rate": 2.903998876632394e-05, "loss": 0.2174, "step": 11255 }, { "epoch": 0.53, "learning_rate": 2.9039520696487918e-05, "loss": 0.0978, "step": 11260 }, { "epoch": 0.53, "learning_rate": 2.9039052626651898e-05, "loss": 0.1548, "step": 11265 }, { "epoch": 0.53, "learning_rate": 2.9038584556815878e-05, "loss": 0.1512, "step": 11270 }, { "epoch": 0.53, "learning_rate": 2.9038116486979858e-05, "loss": 0.0844, "step": 11275 }, { "epoch": 0.53, "learning_rate": 2.9037648417143837e-05, "loss": 0.2364, "step": 11280 }, { "epoch": 0.53, "learning_rate": 2.9037180347307817e-05, "loss": 0.2016, "step": 11285 }, { "epoch": 0.53, "learning_rate": 2.90367122774718e-05, "loss": 0.3631, "step": 11290 }, { "epoch": 0.53, "learning_rate": 2.903624420763578e-05, "loss": 0.3152, "step": 11295 }, { "epoch": 0.53, "learning_rate": 2.903577613779976e-05, "loss": 0.6023, "step": 11300 }, { "epoch": 0.53, "learning_rate": 2.903530806796374e-05, "loss": 0.2934, "step": 11305 }, { "epoch": 0.53, "learning_rate": 2.9034839998127723e-05, "loss": 0.0976, "step": 11310 }, { "epoch": 0.53, "learning_rate": 2.9034371928291703e-05, "loss": 0.124, "step": 11315 }, { "epoch": 0.53, "learning_rate": 2.9033903858455683e-05, "loss": 0.2261, "step": 11320 }, { "epoch": 0.53, "learning_rate": 2.9033435788619666e-05, "loss": 0.1269, "step": 11325 }, { "epoch": 0.53, "learning_rate": 2.9032967718783642e-05, "loss": 0.2572, "step": 11330 }, { "epoch": 0.53, "learning_rate": 2.9032499648947622e-05, "loss": 0.233, "step": 11335 }, { "epoch": 0.53, "learning_rate": 2.9032031579111602e-05, "loss": 0.3301, "step": 11340 }, { "epoch": 0.53, "learning_rate": 2.9031563509275585e-05, "loss": 0.339, "step": 11345 }, { "epoch": 0.53, "learning_rate": 2.9031095439439565e-05, "loss": 0.5173, "step": 11350 }, { "epoch": 0.53, "learning_rate": 2.9030627369603545e-05, "loss": 0.2303, "step": 11355 }, { "epoch": 0.53, "learning_rate": 2.9030159299767525e-05, "loss": 0.2254, "step": 11360 }, { "epoch": 0.53, "learning_rate": 2.9029691229931508e-05, "loss": 0.1155, "step": 11365 }, { "epoch": 0.53, "learning_rate": 2.9029223160095488e-05, "loss": 0.2024, "step": 11370 }, { "epoch": 0.53, "learning_rate": 2.9028755090259468e-05, "loss": 0.1095, "step": 11375 }, { "epoch": 0.53, "learning_rate": 2.9028287020423447e-05, "loss": 0.2487, "step": 11380 }, { "epoch": 0.53, "learning_rate": 2.902781895058743e-05, "loss": 0.2335, "step": 11385 }, { "epoch": 0.53, "learning_rate": 2.902735088075141e-05, "loss": 0.2552, "step": 11390 }, { "epoch": 0.53, "learning_rate": 2.9026882810915387e-05, "loss": 0.4556, "step": 11395 }, { "epoch": 0.53, "learning_rate": 2.902641474107937e-05, "loss": 0.4673, "step": 11400 }, { "epoch": 0.53, "learning_rate": 2.902594667124335e-05, "loss": 0.2411, "step": 11405 }, { "epoch": 0.53, "learning_rate": 2.902547860140733e-05, "loss": 0.1411, "step": 11410 }, { "epoch": 0.53, "learning_rate": 2.902501053157131e-05, "loss": 0.1266, "step": 11415 }, { "epoch": 0.53, "learning_rate": 2.9024542461735293e-05, "loss": 0.1989, "step": 11420 }, { "epoch": 0.53, "learning_rate": 2.9024074391899272e-05, "loss": 0.1909, "step": 11425 }, { "epoch": 0.53, "learning_rate": 2.9023606322063252e-05, "loss": 0.2077, "step": 11430 }, { "epoch": 0.53, "learning_rate": 2.9023138252227232e-05, "loss": 0.2693, "step": 11435 }, { "epoch": 0.53, "learning_rate": 2.9022670182391215e-05, "loss": 0.2329, "step": 11440 }, { "epoch": 0.53, "learning_rate": 2.9022202112555195e-05, "loss": 0.3578, "step": 11445 }, { "epoch": 0.53, "learning_rate": 2.9021734042719175e-05, "loss": 0.5947, "step": 11450 }, { "epoch": 0.53, "learning_rate": 2.9021265972883155e-05, "loss": 0.2367, "step": 11455 }, { "epoch": 0.53, "learning_rate": 2.9020797903047135e-05, "loss": 0.0884, "step": 11460 }, { "epoch": 0.53, "learning_rate": 2.9020329833211114e-05, "loss": 0.1093, "step": 11465 }, { "epoch": 0.54, "learning_rate": 2.9019861763375094e-05, "loss": 0.1203, "step": 11470 }, { "epoch": 0.54, "learning_rate": 2.9019393693539077e-05, "loss": 0.1763, "step": 11475 }, { "epoch": 0.54, "learning_rate": 2.9018925623703057e-05, "loss": 0.2753, "step": 11480 }, { "epoch": 0.54, "learning_rate": 2.9018457553867037e-05, "loss": 0.2258, "step": 11485 }, { "epoch": 0.54, "learning_rate": 2.9017989484031017e-05, "loss": 0.2626, "step": 11490 }, { "epoch": 0.54, "learning_rate": 2.9017521414195e-05, "loss": 0.3519, "step": 11495 }, { "epoch": 0.54, "learning_rate": 2.901705334435898e-05, "loss": 0.508, "step": 11500 }, { "epoch": 0.54, "learning_rate": 2.901658527452296e-05, "loss": 0.1452, "step": 11505 }, { "epoch": 0.54, "learning_rate": 2.9016117204686943e-05, "loss": 0.0792, "step": 11510 }, { "epoch": 0.54, "learning_rate": 2.9015649134850923e-05, "loss": 0.1266, "step": 11515 }, { "epoch": 0.54, "learning_rate": 2.90151810650149e-05, "loss": 0.1531, "step": 11520 }, { "epoch": 0.54, "learning_rate": 2.901471299517888e-05, "loss": 0.1746, "step": 11525 }, { "epoch": 0.54, "learning_rate": 2.9014244925342862e-05, "loss": 0.2665, "step": 11530 }, { "epoch": 0.54, "learning_rate": 2.9013776855506842e-05, "loss": 0.3047, "step": 11535 }, { "epoch": 0.54, "learning_rate": 2.9013308785670822e-05, "loss": 0.304, "step": 11540 }, { "epoch": 0.54, "learning_rate": 2.90128407158348e-05, "loss": 0.3622, "step": 11545 }, { "epoch": 0.54, "learning_rate": 2.9012372645998785e-05, "loss": 0.56, "step": 11550 }, { "epoch": 0.54, "learning_rate": 2.9011904576162765e-05, "loss": 0.2347, "step": 11555 }, { "epoch": 0.54, "learning_rate": 2.9011436506326744e-05, "loss": 0.0708, "step": 11560 }, { "epoch": 0.54, "learning_rate": 2.9010968436490724e-05, "loss": 0.0918, "step": 11565 }, { "epoch": 0.54, "learning_rate": 2.9010500366654708e-05, "loss": 0.148, "step": 11570 }, { "epoch": 0.54, "learning_rate": 2.9010032296818687e-05, "loss": 0.2018, "step": 11575 }, { "epoch": 0.54, "learning_rate": 2.9009564226982667e-05, "loss": 0.1662, "step": 11580 }, { "epoch": 0.54, "learning_rate": 2.9009096157146647e-05, "loss": 0.2973, "step": 11585 }, { "epoch": 0.54, "learning_rate": 2.9008628087310627e-05, "loss": 0.2854, "step": 11590 }, { "epoch": 0.54, "learning_rate": 2.9008160017474607e-05, "loss": 0.3845, "step": 11595 }, { "epoch": 0.54, "learning_rate": 2.9007691947638586e-05, "loss": 0.5031, "step": 11600 }, { "epoch": 0.54, "learning_rate": 2.900722387780257e-05, "loss": 0.2475, "step": 11605 }, { "epoch": 0.54, "learning_rate": 2.900675580796655e-05, "loss": 0.1119, "step": 11610 }, { "epoch": 0.54, "learning_rate": 2.900628773813053e-05, "loss": 0.1501, "step": 11615 }, { "epoch": 0.54, "learning_rate": 2.900581966829451e-05, "loss": 0.102, "step": 11620 }, { "epoch": 0.54, "learning_rate": 2.9005351598458492e-05, "loss": 0.1555, "step": 11625 }, { "epoch": 0.54, "learning_rate": 2.9004883528622472e-05, "loss": 0.2079, "step": 11630 }, { "epoch": 0.54, "learning_rate": 2.9004415458786452e-05, "loss": 0.2517, "step": 11635 }, { "epoch": 0.54, "learning_rate": 2.9003947388950435e-05, "loss": 0.3671, "step": 11640 }, { "epoch": 0.54, "learning_rate": 2.900347931911441e-05, "loss": 0.3443, "step": 11645 }, { "epoch": 0.54, "learning_rate": 2.900301124927839e-05, "loss": 0.4903, "step": 11650 }, { "epoch": 0.54, "learning_rate": 2.900254317944237e-05, "loss": 0.2725, "step": 11655 }, { "epoch": 0.54, "learning_rate": 2.9002075109606354e-05, "loss": 0.0895, "step": 11660 }, { "epoch": 0.54, "learning_rate": 2.9001607039770334e-05, "loss": 0.1666, "step": 11665 }, { "epoch": 0.54, "learning_rate": 2.9001138969934314e-05, "loss": 0.0981, "step": 11670 }, { "epoch": 0.54, "learning_rate": 2.9000670900098294e-05, "loss": 0.1788, "step": 11675 }, { "epoch": 0.55, "learning_rate": 2.9000202830262277e-05, "loss": 0.1948, "step": 11680 }, { "epoch": 0.55, "learning_rate": 2.8999734760426257e-05, "loss": 0.1784, "step": 11685 }, { "epoch": 0.55, "learning_rate": 2.8999266690590237e-05, "loss": 0.2079, "step": 11690 }, { "epoch": 0.55, "learning_rate": 2.899879862075422e-05, "loss": 0.2909, "step": 11695 }, { "epoch": 0.55, "learning_rate": 2.89983305509182e-05, "loss": 0.5101, "step": 11700 }, { "epoch": 0.55, "learning_rate": 2.899786248108218e-05, "loss": 0.2351, "step": 11705 }, { "epoch": 0.55, "learning_rate": 2.8997394411246156e-05, "loss": 0.1133, "step": 11710 }, { "epoch": 0.55, "learning_rate": 2.899692634141014e-05, "loss": 0.1062, "step": 11715 }, { "epoch": 0.55, "learning_rate": 2.899645827157412e-05, "loss": 0.1585, "step": 11720 }, { "epoch": 0.55, "learning_rate": 2.89959902017381e-05, "loss": 0.2545, "step": 11725 }, { "epoch": 0.55, "learning_rate": 2.899552213190208e-05, "loss": 0.2166, "step": 11730 }, { "epoch": 0.55, "learning_rate": 2.8995054062066062e-05, "loss": 0.1675, "step": 11735 }, { "epoch": 0.55, "learning_rate": 2.899458599223004e-05, "loss": 0.3697, "step": 11740 }, { "epoch": 0.55, "learning_rate": 2.899411792239402e-05, "loss": 0.4677, "step": 11745 }, { "epoch": 0.55, "learning_rate": 2.8993649852558005e-05, "loss": 0.5517, "step": 11750 }, { "epoch": 0.55, "learning_rate": 2.8993181782721984e-05, "loss": 0.2434, "step": 11755 }, { "epoch": 0.55, "learning_rate": 2.8992713712885964e-05, "loss": 0.0643, "step": 11760 }, { "epoch": 0.55, "learning_rate": 2.8992245643049944e-05, "loss": 0.1173, "step": 11765 }, { "epoch": 0.55, "learning_rate": 2.8991777573213927e-05, "loss": 0.1006, "step": 11770 }, { "epoch": 0.55, "learning_rate": 2.8991309503377904e-05, "loss": 0.1823, "step": 11775 }, { "epoch": 0.55, "learning_rate": 2.8990841433541884e-05, "loss": 0.1258, "step": 11780 }, { "epoch": 0.55, "learning_rate": 2.8990373363705863e-05, "loss": 0.2484, "step": 11785 }, { "epoch": 0.55, "learning_rate": 2.8989905293869847e-05, "loss": 0.2762, "step": 11790 }, { "epoch": 0.55, "learning_rate": 2.8989437224033826e-05, "loss": 0.3407, "step": 11795 }, { "epoch": 0.55, "learning_rate": 2.8988969154197806e-05, "loss": 0.679, "step": 11800 }, { "epoch": 0.55, "learning_rate": 2.8988501084361786e-05, "loss": 0.1905, "step": 11805 }, { "epoch": 0.55, "learning_rate": 2.898803301452577e-05, "loss": 0.0999, "step": 11810 }, { "epoch": 0.55, "learning_rate": 2.898756494468975e-05, "loss": 0.1598, "step": 11815 }, { "epoch": 0.55, "learning_rate": 2.898709687485373e-05, "loss": 0.0974, "step": 11820 }, { "epoch": 0.55, "learning_rate": 2.8986628805017712e-05, "loss": 0.0905, "step": 11825 }, { "epoch": 0.55, "learning_rate": 2.8986160735181692e-05, "loss": 0.1875, "step": 11830 }, { "epoch": 0.55, "learning_rate": 2.898569266534567e-05, "loss": 0.1852, "step": 11835 }, { "epoch": 0.55, "learning_rate": 2.8985224595509648e-05, "loss": 0.2225, "step": 11840 }, { "epoch": 0.55, "learning_rate": 2.898475652567363e-05, "loss": 0.2743, "step": 11845 }, { "epoch": 0.55, "learning_rate": 2.898428845583761e-05, "loss": 0.5935, "step": 11850 }, { "epoch": 0.55, "learning_rate": 2.898382038600159e-05, "loss": 0.2495, "step": 11855 }, { "epoch": 0.55, "learning_rate": 2.898335231616557e-05, "loss": 0.0954, "step": 11860 }, { "epoch": 0.55, "learning_rate": 2.8982884246329554e-05, "loss": 0.1585, "step": 11865 }, { "epoch": 0.55, "learning_rate": 2.8982416176493534e-05, "loss": 0.1473, "step": 11870 }, { "epoch": 0.55, "learning_rate": 2.8981948106657514e-05, "loss": 0.1356, "step": 11875 }, { "epoch": 0.55, "learning_rate": 2.8981480036821497e-05, "loss": 0.1425, "step": 11880 }, { "epoch": 0.55, "learning_rate": 2.8981011966985477e-05, "loss": 0.2479, "step": 11885 }, { "epoch": 0.55, "learning_rate": 2.8980543897149456e-05, "loss": 0.2751, "step": 11890 }, { "epoch": 0.56, "learning_rate": 2.8980075827313436e-05, "loss": 0.343, "step": 11895 }, { "epoch": 0.56, "learning_rate": 2.8979607757477416e-05, "loss": 0.7812, "step": 11900 }, { "epoch": 0.56, "learning_rate": 2.8979139687641396e-05, "loss": 0.219, "step": 11905 }, { "epoch": 0.56, "learning_rate": 2.8978671617805376e-05, "loss": 0.051, "step": 11910 }, { "epoch": 0.56, "learning_rate": 2.8978203547969356e-05, "loss": 0.0918, "step": 11915 }, { "epoch": 0.56, "learning_rate": 2.897773547813334e-05, "loss": 0.1408, "step": 11920 }, { "epoch": 0.56, "learning_rate": 2.897726740829732e-05, "loss": 0.1209, "step": 11925 }, { "epoch": 0.56, "learning_rate": 2.89767993384613e-05, "loss": 0.1562, "step": 11930 }, { "epoch": 0.56, "learning_rate": 2.897633126862528e-05, "loss": 0.1601, "step": 11935 }, { "epoch": 0.56, "learning_rate": 2.897586319878926e-05, "loss": 0.3487, "step": 11940 }, { "epoch": 0.56, "learning_rate": 2.897539512895324e-05, "loss": 0.3111, "step": 11945 }, { "epoch": 0.56, "learning_rate": 2.897492705911722e-05, "loss": 0.4768, "step": 11950 }, { "epoch": 0.56, "learning_rate": 2.8974458989281204e-05, "loss": 0.2605, "step": 11955 }, { "epoch": 0.56, "learning_rate": 2.897399091944518e-05, "loss": 0.1403, "step": 11960 }, { "epoch": 0.56, "learning_rate": 2.897352284960916e-05, "loss": 0.0971, "step": 11965 }, { "epoch": 0.56, "learning_rate": 2.897305477977314e-05, "loss": 0.2108, "step": 11970 }, { "epoch": 0.56, "learning_rate": 2.8972586709937124e-05, "loss": 0.1406, "step": 11975 }, { "epoch": 0.56, "learning_rate": 2.8972118640101103e-05, "loss": 0.2211, "step": 11980 }, { "epoch": 0.56, "learning_rate": 2.8971650570265083e-05, "loss": 0.2503, "step": 11985 }, { "epoch": 0.56, "learning_rate": 2.8971182500429063e-05, "loss": 0.2778, "step": 11990 }, { "epoch": 0.56, "learning_rate": 2.8970714430593046e-05, "loss": 0.379, "step": 11995 }, { "epoch": 0.56, "learning_rate": 2.8970246360757026e-05, "loss": 0.5211, "step": 12000 }, { "epoch": 0.56, "learning_rate": 2.8969778290921006e-05, "loss": 0.1406, "step": 12005 }, { "epoch": 0.56, "learning_rate": 2.896931022108499e-05, "loss": 0.0509, "step": 12010 }, { "epoch": 0.56, "learning_rate": 2.896884215124897e-05, "loss": 0.0842, "step": 12015 }, { "epoch": 0.56, "learning_rate": 2.896837408141295e-05, "loss": 0.2049, "step": 12020 }, { "epoch": 0.56, "learning_rate": 2.8967906011576925e-05, "loss": 0.1409, "step": 12025 }, { "epoch": 0.56, "learning_rate": 2.896743794174091e-05, "loss": 0.1995, "step": 12030 }, { "epoch": 0.56, "learning_rate": 2.8966969871904888e-05, "loss": 0.1705, "step": 12035 }, { "epoch": 0.56, "learning_rate": 2.8966501802068868e-05, "loss": 0.3089, "step": 12040 }, { "epoch": 0.56, "learning_rate": 2.8966033732232848e-05, "loss": 0.2439, "step": 12045 }, { "epoch": 0.56, "learning_rate": 2.896556566239683e-05, "loss": 0.3632, "step": 12050 }, { "epoch": 0.56, "learning_rate": 2.896509759256081e-05, "loss": 0.2364, "step": 12055 }, { "epoch": 0.56, "learning_rate": 2.896462952272479e-05, "loss": 0.1063, "step": 12060 }, { "epoch": 0.56, "learning_rate": 2.8964161452888774e-05, "loss": 0.1204, "step": 12065 }, { "epoch": 0.56, "learning_rate": 2.8963693383052754e-05, "loss": 0.105, "step": 12070 }, { "epoch": 0.56, "learning_rate": 2.8963225313216733e-05, "loss": 0.1662, "step": 12075 }, { "epoch": 0.56, "learning_rate": 2.8962757243380713e-05, "loss": 0.238, "step": 12080 }, { "epoch": 0.56, "learning_rate": 2.8962289173544696e-05, "loss": 0.2373, "step": 12085 }, { "epoch": 0.56, "learning_rate": 2.8961821103708673e-05, "loss": 0.3565, "step": 12090 }, { "epoch": 0.56, "learning_rate": 2.8961353033872653e-05, "loss": 0.2394, "step": 12095 }, { "epoch": 0.56, "learning_rate": 2.8960884964036633e-05, "loss": 0.5335, "step": 12100 }, { "epoch": 0.56, "learning_rate": 2.8960416894200616e-05, "loss": 0.2562, "step": 12105 }, { "epoch": 0.57, "learning_rate": 2.8959948824364596e-05, "loss": 0.111, "step": 12110 }, { "epoch": 0.57, "learning_rate": 2.8959480754528575e-05, "loss": 0.1321, "step": 12115 }, { "epoch": 0.57, "learning_rate": 2.895901268469256e-05, "loss": 0.1541, "step": 12120 }, { "epoch": 0.57, "learning_rate": 2.895854461485654e-05, "loss": 0.1097, "step": 12125 }, { "epoch": 0.57, "learning_rate": 2.8958076545020518e-05, "loss": 0.3175, "step": 12130 }, { "epoch": 0.57, "learning_rate": 2.8957608475184498e-05, "loss": 0.2379, "step": 12135 }, { "epoch": 0.57, "learning_rate": 2.895714040534848e-05, "loss": 0.317, "step": 12140 }, { "epoch": 0.57, "learning_rate": 2.895667233551246e-05, "loss": 0.2476, "step": 12145 }, { "epoch": 0.57, "learning_rate": 2.8956204265676438e-05, "loss": 0.6227, "step": 12150 }, { "epoch": 0.57, "learning_rate": 2.8955736195840417e-05, "loss": 0.2231, "step": 12155 }, { "epoch": 0.57, "learning_rate": 2.89552681260044e-05, "loss": 0.0913, "step": 12160 }, { "epoch": 0.57, "learning_rate": 2.895480005616838e-05, "loss": 0.1101, "step": 12165 }, { "epoch": 0.57, "learning_rate": 2.895433198633236e-05, "loss": 0.1475, "step": 12170 }, { "epoch": 0.57, "learning_rate": 2.895386391649634e-05, "loss": 0.1494, "step": 12175 }, { "epoch": 0.57, "learning_rate": 2.8953395846660323e-05, "loss": 0.1823, "step": 12180 }, { "epoch": 0.57, "learning_rate": 2.8952927776824303e-05, "loss": 0.2334, "step": 12185 }, { "epoch": 0.57, "learning_rate": 2.8952459706988283e-05, "loss": 0.2612, "step": 12190 }, { "epoch": 0.57, "learning_rate": 2.8951991637152266e-05, "loss": 0.3328, "step": 12195 }, { "epoch": 0.57, "learning_rate": 2.8951523567316246e-05, "loss": 0.5815, "step": 12200 }, { "epoch": 0.57, "learning_rate": 2.8951055497480226e-05, "loss": 0.2459, "step": 12205 }, { "epoch": 0.57, "learning_rate": 2.8950587427644205e-05, "loss": 0.0765, "step": 12210 }, { "epoch": 0.57, "learning_rate": 2.8950119357808185e-05, "loss": 0.0804, "step": 12215 }, { "epoch": 0.57, "learning_rate": 2.8949651287972165e-05, "loss": 0.1282, "step": 12220 }, { "epoch": 0.57, "learning_rate": 2.8949183218136145e-05, "loss": 0.1609, "step": 12225 }, { "epoch": 0.57, "learning_rate": 2.8948715148300125e-05, "loss": 0.11, "step": 12230 }, { "epoch": 0.57, "learning_rate": 2.8948247078464108e-05, "loss": 0.2662, "step": 12235 }, { "epoch": 0.57, "learning_rate": 2.8947779008628088e-05, "loss": 0.3361, "step": 12240 }, { "epoch": 0.57, "learning_rate": 2.8947310938792068e-05, "loss": 0.3896, "step": 12245 }, { "epoch": 0.57, "learning_rate": 2.894684286895605e-05, "loss": 0.5672, "step": 12250 }, { "epoch": 0.57, "learning_rate": 2.894637479912003e-05, "loss": 0.2442, "step": 12255 }, { "epoch": 0.57, "learning_rate": 2.894590672928401e-05, "loss": 0.0852, "step": 12260 }, { "epoch": 0.57, "learning_rate": 2.894543865944799e-05, "loss": 0.086, "step": 12265 }, { "epoch": 0.57, "learning_rate": 2.8944970589611973e-05, "loss": 0.1488, "step": 12270 }, { "epoch": 0.57, "learning_rate": 2.8944502519775953e-05, "loss": 0.1247, "step": 12275 }, { "epoch": 0.57, "learning_rate": 2.894403444993993e-05, "loss": 0.1959, "step": 12280 }, { "epoch": 0.57, "learning_rate": 2.894356638010391e-05, "loss": 0.1784, "step": 12285 }, { "epoch": 0.57, "learning_rate": 2.8943098310267893e-05, "loss": 0.2463, "step": 12290 }, { "epoch": 0.57, "learning_rate": 2.8942630240431873e-05, "loss": 0.451, "step": 12295 }, { "epoch": 0.57, "learning_rate": 2.8942162170595852e-05, "loss": 0.4983, "step": 12300 }, { "epoch": 0.57, "learning_rate": 2.8941694100759836e-05, "loss": 0.2711, "step": 12305 }, { "epoch": 0.57, "learning_rate": 2.8941226030923815e-05, "loss": 0.1113, "step": 12310 }, { "epoch": 0.57, "learning_rate": 2.8940757961087795e-05, "loss": 0.1236, "step": 12315 }, { "epoch": 0.57, "learning_rate": 2.8940289891251775e-05, "loss": 0.0896, "step": 12320 }, { "epoch": 0.58, "learning_rate": 2.8939821821415758e-05, "loss": 0.1486, "step": 12325 }, { "epoch": 0.58, "learning_rate": 2.8939353751579738e-05, "loss": 0.173, "step": 12330 }, { "epoch": 0.58, "learning_rate": 2.8938885681743718e-05, "loss": 0.1792, "step": 12335 }, { "epoch": 0.58, "learning_rate": 2.8938417611907694e-05, "loss": 0.2277, "step": 12340 }, { "epoch": 0.58, "learning_rate": 2.8937949542071678e-05, "loss": 0.335, "step": 12345 }, { "epoch": 0.58, "learning_rate": 2.8937481472235657e-05, "loss": 0.4859, "step": 12350 }, { "epoch": 0.58, "learning_rate": 2.8937013402399637e-05, "loss": 0.2462, "step": 12355 }, { "epoch": 0.58, "learning_rate": 2.8936545332563617e-05, "loss": 0.1036, "step": 12360 }, { "epoch": 0.58, "learning_rate": 2.89360772627276e-05, "loss": 0.0784, "step": 12365 }, { "epoch": 0.58, "learning_rate": 2.893560919289158e-05, "loss": 0.2005, "step": 12370 }, { "epoch": 0.58, "learning_rate": 2.893514112305556e-05, "loss": 0.1206, "step": 12375 }, { "epoch": 0.58, "learning_rate": 2.8934673053219543e-05, "loss": 0.1557, "step": 12380 }, { "epoch": 0.58, "learning_rate": 2.8934204983383523e-05, "loss": 0.2565, "step": 12385 }, { "epoch": 0.58, "learning_rate": 2.8933736913547503e-05, "loss": 0.2539, "step": 12390 }, { "epoch": 0.58, "learning_rate": 2.8933268843711482e-05, "loss": 0.4199, "step": 12395 }, { "epoch": 0.58, "learning_rate": 2.8932800773875466e-05, "loss": 0.5691, "step": 12400 }, { "epoch": 0.58, "learning_rate": 2.8932332704039442e-05, "loss": 0.1846, "step": 12405 }, { "epoch": 0.58, "learning_rate": 2.8931864634203422e-05, "loss": 0.0847, "step": 12410 }, { "epoch": 0.58, "learning_rate": 2.8931396564367402e-05, "loss": 0.0395, "step": 12415 }, { "epoch": 0.58, "learning_rate": 2.8930928494531385e-05, "loss": 0.1356, "step": 12420 }, { "epoch": 0.58, "learning_rate": 2.8930460424695365e-05, "loss": 0.1799, "step": 12425 }, { "epoch": 0.58, "learning_rate": 2.8929992354859345e-05, "loss": 0.2015, "step": 12430 }, { "epoch": 0.58, "learning_rate": 2.8929524285023328e-05, "loss": 0.1996, "step": 12435 }, { "epoch": 0.58, "learning_rate": 2.8929056215187308e-05, "loss": 0.2419, "step": 12440 }, { "epoch": 0.58, "learning_rate": 2.8928588145351287e-05, "loss": 0.3582, "step": 12445 }, { "epoch": 0.58, "learning_rate": 2.8928120075515267e-05, "loss": 0.4335, "step": 12450 }, { "epoch": 0.58, "learning_rate": 2.892765200567925e-05, "loss": 0.2001, "step": 12455 }, { "epoch": 0.58, "learning_rate": 2.892718393584323e-05, "loss": 0.0701, "step": 12460 }, { "epoch": 0.58, "learning_rate": 2.892671586600721e-05, "loss": 0.0676, "step": 12465 }, { "epoch": 0.58, "learning_rate": 2.8926247796171186e-05, "loss": 0.2049, "step": 12470 }, { "epoch": 0.58, "learning_rate": 2.892577972633517e-05, "loss": 0.1933, "step": 12475 }, { "epoch": 0.58, "learning_rate": 2.892531165649915e-05, "loss": 0.2288, "step": 12480 }, { "epoch": 0.58, "learning_rate": 2.892484358666313e-05, "loss": 0.2598, "step": 12485 }, { "epoch": 0.58, "learning_rate": 2.8924375516827113e-05, "loss": 0.2414, "step": 12490 }, { "epoch": 0.58, "learning_rate": 2.8923907446991092e-05, "loss": 0.2961, "step": 12495 }, { "epoch": 0.58, "learning_rate": 2.8923439377155072e-05, "loss": 0.4972, "step": 12500 }, { "epoch": 0.58, "learning_rate": 2.8922971307319052e-05, "loss": 0.2474, "step": 12505 }, { "epoch": 0.58, "learning_rate": 2.8922503237483035e-05, "loss": 0.1236, "step": 12510 }, { "epoch": 0.58, "learning_rate": 2.8922035167647015e-05, "loss": 0.1226, "step": 12515 }, { "epoch": 0.58, "learning_rate": 2.8921567097810995e-05, "loss": 0.1345, "step": 12520 }, { "epoch": 0.58, "learning_rate": 2.8921099027974975e-05, "loss": 0.1681, "step": 12525 }, { "epoch": 0.58, "learning_rate": 2.8920630958138954e-05, "loss": 0.2256, "step": 12530 }, { "epoch": 0.58, "learning_rate": 2.8920162888302934e-05, "loss": 0.2, "step": 12535 }, { "epoch": 0.59, "learning_rate": 2.8919694818466914e-05, "loss": 0.2551, "step": 12540 }, { "epoch": 0.59, "learning_rate": 2.8919226748630897e-05, "loss": 0.2243, "step": 12545 }, { "epoch": 0.59, "learning_rate": 2.8918758678794877e-05, "loss": 0.4802, "step": 12550 }, { "epoch": 0.59, "learning_rate": 2.8918290608958857e-05, "loss": 0.1943, "step": 12555 }, { "epoch": 0.59, "learning_rate": 2.8917822539122837e-05, "loss": 0.0714, "step": 12560 }, { "epoch": 0.59, "learning_rate": 2.891735446928682e-05, "loss": 0.2009, "step": 12565 }, { "epoch": 0.59, "learning_rate": 2.89168863994508e-05, "loss": 0.1655, "step": 12570 }, { "epoch": 0.59, "learning_rate": 2.891641832961478e-05, "loss": 0.1689, "step": 12575 }, { "epoch": 0.59, "learning_rate": 2.891595025977876e-05, "loss": 0.1921, "step": 12580 }, { "epoch": 0.59, "learning_rate": 2.8915482189942743e-05, "loss": 0.2718, "step": 12585 }, { "epoch": 0.59, "learning_rate": 2.8915014120106722e-05, "loss": 0.2729, "step": 12590 }, { "epoch": 0.59, "learning_rate": 2.89145460502707e-05, "loss": 0.3708, "step": 12595 }, { "epoch": 0.59, "learning_rate": 2.891407798043468e-05, "loss": 0.5397, "step": 12600 }, { "epoch": 0.59, "learning_rate": 2.8913609910598662e-05, "loss": 0.219, "step": 12605 }, { "epoch": 0.59, "learning_rate": 2.8913141840762642e-05, "loss": 0.0587, "step": 12610 }, { "epoch": 0.59, "learning_rate": 2.891267377092662e-05, "loss": 0.0868, "step": 12615 }, { "epoch": 0.59, "learning_rate": 2.8912205701090605e-05, "loss": 0.1228, "step": 12620 }, { "epoch": 0.59, "learning_rate": 2.8911737631254585e-05, "loss": 0.2103, "step": 12625 }, { "epoch": 0.59, "learning_rate": 2.8911269561418564e-05, "loss": 0.1224, "step": 12630 }, { "epoch": 0.59, "learning_rate": 2.8910801491582544e-05, "loss": 0.1422, "step": 12635 }, { "epoch": 0.59, "learning_rate": 2.8910333421746527e-05, "loss": 0.1755, "step": 12640 }, { "epoch": 0.59, "learning_rate": 2.8909865351910507e-05, "loss": 0.3288, "step": 12645 }, { "epoch": 0.59, "learning_rate": 2.8909397282074487e-05, "loss": 0.4658, "step": 12650 }, { "epoch": 0.59, "learning_rate": 2.8908929212238467e-05, "loss": 0.2296, "step": 12655 }, { "epoch": 0.59, "learning_rate": 2.8908461142402447e-05, "loss": 0.1478, "step": 12660 }, { "epoch": 0.59, "learning_rate": 2.8907993072566426e-05, "loss": 0.122, "step": 12665 }, { "epoch": 0.59, "learning_rate": 2.8907525002730406e-05, "loss": 0.1071, "step": 12670 }, { "epoch": 0.59, "learning_rate": 2.890705693289439e-05, "loss": 0.1918, "step": 12675 }, { "epoch": 0.59, "learning_rate": 2.890658886305837e-05, "loss": 0.1592, "step": 12680 }, { "epoch": 0.59, "learning_rate": 2.890612079322235e-05, "loss": 0.2755, "step": 12685 }, { "epoch": 0.59, "learning_rate": 2.890565272338633e-05, "loss": 0.3408, "step": 12690 }, { "epoch": 0.59, "learning_rate": 2.8905184653550312e-05, "loss": 0.2939, "step": 12695 }, { "epoch": 0.59, "learning_rate": 2.8904716583714292e-05, "loss": 0.52, "step": 12700 }, { "epoch": 0.59, "learning_rate": 2.8904248513878272e-05, "loss": 0.2381, "step": 12705 }, { "epoch": 0.59, "learning_rate": 2.890378044404225e-05, "loss": 0.1097, "step": 12710 }, { "epoch": 0.59, "learning_rate": 2.8903312374206235e-05, "loss": 0.1028, "step": 12715 }, { "epoch": 0.59, "learning_rate": 2.890284430437021e-05, "loss": 0.0656, "step": 12720 }, { "epoch": 0.59, "learning_rate": 2.890237623453419e-05, "loss": 0.1366, "step": 12725 }, { "epoch": 0.59, "learning_rate": 2.8901908164698174e-05, "loss": 0.1737, "step": 12730 }, { "epoch": 0.59, "learning_rate": 2.8901440094862154e-05, "loss": 0.2923, "step": 12735 }, { "epoch": 0.59, "learning_rate": 2.8900972025026134e-05, "loss": 0.2551, "step": 12740 }, { "epoch": 0.59, "learning_rate": 2.8900503955190114e-05, "loss": 0.2689, "step": 12745 }, { "epoch": 0.59, "learning_rate": 2.8900035885354097e-05, "loss": 0.4539, "step": 12750 }, { "epoch": 0.6, "learning_rate": 2.8899567815518077e-05, "loss": 0.2615, "step": 12755 }, { "epoch": 0.6, "learning_rate": 2.8899099745682057e-05, "loss": 0.0602, "step": 12760 }, { "epoch": 0.6, "learning_rate": 2.8898631675846036e-05, "loss": 0.0877, "step": 12765 }, { "epoch": 0.6, "learning_rate": 2.889816360601002e-05, "loss": 0.075, "step": 12770 }, { "epoch": 0.6, "learning_rate": 2.8897695536174e-05, "loss": 0.118, "step": 12775 }, { "epoch": 0.6, "learning_rate": 2.889722746633798e-05, "loss": 0.1976, "step": 12780 }, { "epoch": 0.6, "learning_rate": 2.8896759396501956e-05, "loss": 0.2382, "step": 12785 }, { "epoch": 0.6, "learning_rate": 2.889629132666594e-05, "loss": 0.2397, "step": 12790 }, { "epoch": 0.6, "learning_rate": 2.889582325682992e-05, "loss": 0.2587, "step": 12795 }, { "epoch": 0.6, "learning_rate": 2.88953551869939e-05, "loss": 0.4395, "step": 12800 }, { "epoch": 0.6, "learning_rate": 2.8894887117157882e-05, "loss": 0.2838, "step": 12805 }, { "epoch": 0.6, "learning_rate": 2.889441904732186e-05, "loss": 0.1163, "step": 12810 }, { "epoch": 0.6, "learning_rate": 2.889395097748584e-05, "loss": 0.089, "step": 12815 }, { "epoch": 0.6, "learning_rate": 2.889348290764982e-05, "loss": 0.1262, "step": 12820 }, { "epoch": 0.6, "learning_rate": 2.8893014837813804e-05, "loss": 0.2655, "step": 12825 }, { "epoch": 0.6, "learning_rate": 2.8892546767977784e-05, "loss": 0.1867, "step": 12830 }, { "epoch": 0.6, "learning_rate": 2.8892078698141764e-05, "loss": 0.2756, "step": 12835 }, { "epoch": 0.6, "learning_rate": 2.8891610628305744e-05, "loss": 0.2579, "step": 12840 }, { "epoch": 0.6, "learning_rate": 2.8891142558469724e-05, "loss": 0.4167, "step": 12845 }, { "epoch": 0.6, "learning_rate": 2.8890674488633703e-05, "loss": 0.6041, "step": 12850 }, { "epoch": 0.6, "learning_rate": 2.8890206418797683e-05, "loss": 0.2392, "step": 12855 }, { "epoch": 0.6, "learning_rate": 2.8889738348961666e-05, "loss": 0.1081, "step": 12860 }, { "epoch": 0.6, "learning_rate": 2.8889270279125646e-05, "loss": 0.0972, "step": 12865 }, { "epoch": 0.6, "learning_rate": 2.8888802209289626e-05, "loss": 0.1922, "step": 12870 }, { "epoch": 0.6, "learning_rate": 2.8888334139453606e-05, "loss": 0.1852, "step": 12875 }, { "epoch": 0.6, "learning_rate": 2.888786606961759e-05, "loss": 0.2465, "step": 12880 }, { "epoch": 0.6, "learning_rate": 2.888739799978157e-05, "loss": 0.2227, "step": 12885 }, { "epoch": 0.6, "learning_rate": 2.888692992994555e-05, "loss": 0.3551, "step": 12890 }, { "epoch": 0.6, "learning_rate": 2.888646186010953e-05, "loss": 0.3309, "step": 12895 }, { "epoch": 0.6, "learning_rate": 2.8885993790273512e-05, "loss": 0.645, "step": 12900 }, { "epoch": 0.6, "learning_rate": 2.888552572043749e-05, "loss": 0.2844, "step": 12905 }, { "epoch": 0.6, "learning_rate": 2.8885057650601468e-05, "loss": 0.1013, "step": 12910 }, { "epoch": 0.6, "learning_rate": 2.888458958076545e-05, "loss": 0.1039, "step": 12915 }, { "epoch": 0.6, "learning_rate": 2.888412151092943e-05, "loss": 0.1857, "step": 12920 }, { "epoch": 0.6, "learning_rate": 2.888365344109341e-05, "loss": 0.189, "step": 12925 }, { "epoch": 0.6, "learning_rate": 2.888318537125739e-05, "loss": 0.2779, "step": 12930 }, { "epoch": 0.6, "learning_rate": 2.8882717301421374e-05, "loss": 0.2452, "step": 12935 }, { "epoch": 0.6, "learning_rate": 2.8882249231585354e-05, "loss": 0.2025, "step": 12940 }, { "epoch": 0.6, "learning_rate": 2.8881781161749334e-05, "loss": 0.3729, "step": 12945 }, { "epoch": 0.6, "learning_rate": 2.8881313091913313e-05, "loss": 0.5273, "step": 12950 }, { "epoch": 0.6, "learning_rate": 2.8880845022077297e-05, "loss": 0.2253, "step": 12955 }, { "epoch": 0.6, "learning_rate": 2.8880376952241276e-05, "loss": 0.0965, "step": 12960 }, { "epoch": 0.6, "learning_rate": 2.8879908882405256e-05, "loss": 0.0847, "step": 12965 }, { "epoch": 0.61, "learning_rate": 2.8879440812569236e-05, "loss": 0.1271, "step": 12970 }, { "epoch": 0.61, "learning_rate": 2.8878972742733216e-05, "loss": 0.1779, "step": 12975 }, { "epoch": 0.61, "learning_rate": 2.8878504672897196e-05, "loss": 0.1899, "step": 12980 }, { "epoch": 0.61, "learning_rate": 2.8878036603061175e-05, "loss": 0.1681, "step": 12985 }, { "epoch": 0.61, "learning_rate": 2.887756853322516e-05, "loss": 0.2269, "step": 12990 }, { "epoch": 0.61, "learning_rate": 2.887710046338914e-05, "loss": 0.2698, "step": 12995 }, { "epoch": 0.61, "learning_rate": 2.887663239355312e-05, "loss": 0.5955, "step": 13000 }, { "epoch": 0.61, "learning_rate": 2.8876164323717098e-05, "loss": 0.2408, "step": 13005 }, { "epoch": 0.61, "learning_rate": 2.887569625388108e-05, "loss": 0.0575, "step": 13010 }, { "epoch": 0.61, "learning_rate": 2.887522818404506e-05, "loss": 0.0674, "step": 13015 }, { "epoch": 0.61, "learning_rate": 2.887476011420904e-05, "loss": 0.2003, "step": 13020 }, { "epoch": 0.61, "learning_rate": 2.887429204437302e-05, "loss": 0.2053, "step": 13025 }, { "epoch": 0.61, "learning_rate": 2.8873823974537004e-05, "loss": 0.1924, "step": 13030 }, { "epoch": 0.61, "learning_rate": 2.887335590470098e-05, "loss": 0.2305, "step": 13035 }, { "epoch": 0.61, "learning_rate": 2.887288783486496e-05, "loss": 0.2027, "step": 13040 }, { "epoch": 0.61, "learning_rate": 2.8872419765028943e-05, "loss": 0.3431, "step": 13045 }, { "epoch": 0.61, "learning_rate": 2.8871951695192923e-05, "loss": 0.486, "step": 13050 }, { "epoch": 0.61, "learning_rate": 2.8871483625356903e-05, "loss": 0.1978, "step": 13055 }, { "epoch": 0.61, "learning_rate": 2.8871015555520883e-05, "loss": 0.1411, "step": 13060 }, { "epoch": 0.61, "learning_rate": 2.8870547485684866e-05, "loss": 0.1373, "step": 13065 }, { "epoch": 0.61, "learning_rate": 2.8870079415848846e-05, "loss": 0.141, "step": 13070 }, { "epoch": 0.61, "learning_rate": 2.8869611346012826e-05, "loss": 0.1704, "step": 13075 }, { "epoch": 0.61, "learning_rate": 2.8869143276176806e-05, "loss": 0.2197, "step": 13080 }, { "epoch": 0.61, "learning_rate": 2.886867520634079e-05, "loss": 0.2301, "step": 13085 }, { "epoch": 0.61, "learning_rate": 2.886820713650477e-05, "loss": 0.2399, "step": 13090 }, { "epoch": 0.61, "learning_rate": 2.886773906666875e-05, "loss": 0.3613, "step": 13095 }, { "epoch": 0.61, "learning_rate": 2.8867270996832728e-05, "loss": 0.5423, "step": 13100 }, { "epoch": 0.61, "learning_rate": 2.8866802926996708e-05, "loss": 0.2183, "step": 13105 }, { "epoch": 0.61, "learning_rate": 2.8866334857160688e-05, "loss": 0.0634, "step": 13110 }, { "epoch": 0.61, "learning_rate": 2.8865866787324668e-05, "loss": 0.0862, "step": 13115 }, { "epoch": 0.61, "learning_rate": 2.886539871748865e-05, "loss": 0.1013, "step": 13120 }, { "epoch": 0.61, "learning_rate": 2.886493064765263e-05, "loss": 0.1501, "step": 13125 }, { "epoch": 0.61, "learning_rate": 2.886446257781661e-05, "loss": 0.2243, "step": 13130 }, { "epoch": 0.61, "learning_rate": 2.886399450798059e-05, "loss": 0.2639, "step": 13135 }, { "epoch": 0.61, "learning_rate": 2.8863526438144574e-05, "loss": 0.2846, "step": 13140 }, { "epoch": 0.61, "learning_rate": 2.8863058368308553e-05, "loss": 0.4036, "step": 13145 }, { "epoch": 0.61, "learning_rate": 2.8862590298472533e-05, "loss": 0.5067, "step": 13150 }, { "epoch": 0.61, "learning_rate": 2.8862122228636516e-05, "loss": 0.3654, "step": 13155 }, { "epoch": 0.61, "learning_rate": 2.8861654158800496e-05, "loss": 0.0709, "step": 13160 }, { "epoch": 0.61, "learning_rate": 2.8861186088964473e-05, "loss": 0.1663, "step": 13165 }, { "epoch": 0.61, "learning_rate": 2.8860718019128452e-05, "loss": 0.1669, "step": 13170 }, { "epoch": 0.61, "learning_rate": 2.8860249949292436e-05, "loss": 0.1563, "step": 13175 }, { "epoch": 0.61, "learning_rate": 2.8859781879456415e-05, "loss": 0.1988, "step": 13180 }, { "epoch": 0.62, "learning_rate": 2.8859313809620395e-05, "loss": 0.2369, "step": 13185 }, { "epoch": 0.62, "learning_rate": 2.8858845739784375e-05, "loss": 0.2254, "step": 13190 }, { "epoch": 0.62, "learning_rate": 2.885837766994836e-05, "loss": 0.288, "step": 13195 }, { "epoch": 0.62, "learning_rate": 2.8857909600112338e-05, "loss": 0.5139, "step": 13200 }, { "epoch": 0.62, "learning_rate": 2.8857441530276318e-05, "loss": 0.2087, "step": 13205 }, { "epoch": 0.62, "learning_rate": 2.8856973460440298e-05, "loss": 0.0958, "step": 13210 }, { "epoch": 0.62, "learning_rate": 2.885650539060428e-05, "loss": 0.1304, "step": 13215 }, { "epoch": 0.62, "learning_rate": 2.885603732076826e-05, "loss": 0.1667, "step": 13220 }, { "epoch": 0.62, "learning_rate": 2.8855569250932237e-05, "loss": 0.2053, "step": 13225 }, { "epoch": 0.62, "learning_rate": 2.885510118109622e-05, "loss": 0.1388, "step": 13230 }, { "epoch": 0.62, "learning_rate": 2.88546331112602e-05, "loss": 0.1431, "step": 13235 }, { "epoch": 0.62, "learning_rate": 2.885416504142418e-05, "loss": 0.2002, "step": 13240 }, { "epoch": 0.62, "learning_rate": 2.885369697158816e-05, "loss": 0.3338, "step": 13245 }, { "epoch": 0.62, "learning_rate": 2.8853228901752143e-05, "loss": 0.7644, "step": 13250 }, { "epoch": 0.62, "learning_rate": 2.8852760831916123e-05, "loss": 0.1865, "step": 13255 }, { "epoch": 0.62, "learning_rate": 2.8852292762080103e-05, "loss": 0.0992, "step": 13260 }, { "epoch": 0.62, "learning_rate": 2.8851824692244083e-05, "loss": 0.1559, "step": 13265 }, { "epoch": 0.62, "learning_rate": 2.8851356622408066e-05, "loss": 0.1212, "step": 13270 }, { "epoch": 0.62, "learning_rate": 2.8850888552572046e-05, "loss": 0.2575, "step": 13275 }, { "epoch": 0.62, "learning_rate": 2.8850420482736025e-05, "loss": 0.1288, "step": 13280 }, { "epoch": 0.62, "learning_rate": 2.884995241290001e-05, "loss": 0.2245, "step": 13285 }, { "epoch": 0.62, "learning_rate": 2.8849484343063985e-05, "loss": 0.2046, "step": 13290 }, { "epoch": 0.62, "learning_rate": 2.8849016273227965e-05, "loss": 0.2378, "step": 13295 }, { "epoch": 0.62, "learning_rate": 2.8848548203391945e-05, "loss": 0.4623, "step": 13300 }, { "epoch": 0.62, "learning_rate": 2.8848080133555928e-05, "loss": 0.2265, "step": 13305 }, { "epoch": 0.62, "learning_rate": 2.8847612063719908e-05, "loss": 0.0554, "step": 13310 }, { "epoch": 0.62, "learning_rate": 2.8847143993883887e-05, "loss": 0.125, "step": 13315 }, { "epoch": 0.62, "learning_rate": 2.8846675924047867e-05, "loss": 0.1131, "step": 13320 }, { "epoch": 0.62, "learning_rate": 2.884620785421185e-05, "loss": 0.1376, "step": 13325 }, { "epoch": 0.62, "learning_rate": 2.884573978437583e-05, "loss": 0.1802, "step": 13330 }, { "epoch": 0.62, "learning_rate": 2.884527171453981e-05, "loss": 0.1885, "step": 13335 }, { "epoch": 0.62, "learning_rate": 2.8844803644703793e-05, "loss": 0.2701, "step": 13340 }, { "epoch": 0.62, "learning_rate": 2.8844335574867773e-05, "loss": 0.2813, "step": 13345 }, { "epoch": 0.62, "learning_rate": 2.884386750503175e-05, "loss": 0.5205, "step": 13350 }, { "epoch": 0.62, "learning_rate": 2.884339943519573e-05, "loss": 0.1893, "step": 13355 }, { "epoch": 0.62, "learning_rate": 2.8842931365359713e-05, "loss": 0.0986, "step": 13360 }, { "epoch": 0.62, "learning_rate": 2.8842463295523692e-05, "loss": 0.1085, "step": 13365 }, { "epoch": 0.62, "learning_rate": 2.8841995225687672e-05, "loss": 0.1641, "step": 13370 }, { "epoch": 0.62, "learning_rate": 2.8841527155851652e-05, "loss": 0.1011, "step": 13375 }, { "epoch": 0.62, "learning_rate": 2.8841059086015635e-05, "loss": 0.2169, "step": 13380 }, { "epoch": 0.62, "learning_rate": 2.8840591016179615e-05, "loss": 0.1129, "step": 13385 }, { "epoch": 0.62, "learning_rate": 2.8840122946343595e-05, "loss": 0.2633, "step": 13390 }, { "epoch": 0.63, "learning_rate": 2.8839654876507575e-05, "loss": 0.3939, "step": 13395 }, { "epoch": 0.63, "learning_rate": 2.8839186806671558e-05, "loss": 0.4045, "step": 13400 }, { "epoch": 0.63, "learning_rate": 2.8838718736835538e-05, "loss": 0.2207, "step": 13405 }, { "epoch": 0.63, "learning_rate": 2.8838250666999518e-05, "loss": 0.0585, "step": 13410 }, { "epoch": 0.63, "learning_rate": 2.8837782597163497e-05, "loss": 0.1279, "step": 13415 }, { "epoch": 0.63, "learning_rate": 2.8837314527327477e-05, "loss": 0.1611, "step": 13420 }, { "epoch": 0.63, "learning_rate": 2.8836846457491457e-05, "loss": 0.1943, "step": 13425 }, { "epoch": 0.63, "learning_rate": 2.8836378387655437e-05, "loss": 0.1545, "step": 13430 }, { "epoch": 0.63, "learning_rate": 2.883591031781942e-05, "loss": 0.1712, "step": 13435 }, { "epoch": 0.63, "learning_rate": 2.88354422479834e-05, "loss": 0.1363, "step": 13440 }, { "epoch": 0.63, "learning_rate": 2.883497417814738e-05, "loss": 0.3044, "step": 13445 }, { "epoch": 0.63, "learning_rate": 2.883450610831136e-05, "loss": 0.4945, "step": 13450 }, { "epoch": 0.63, "learning_rate": 2.8834038038475343e-05, "loss": 0.1929, "step": 13455 }, { "epoch": 0.63, "learning_rate": 2.8833569968639323e-05, "loss": 0.1058, "step": 13460 }, { "epoch": 0.63, "learning_rate": 2.8833101898803302e-05, "loss": 0.0493, "step": 13465 }, { "epoch": 0.63, "learning_rate": 2.8832633828967286e-05, "loss": 0.1075, "step": 13470 }, { "epoch": 0.63, "learning_rate": 2.8832165759131265e-05, "loss": 0.1647, "step": 13475 }, { "epoch": 0.63, "learning_rate": 2.8831697689295242e-05, "loss": 0.2417, "step": 13480 }, { "epoch": 0.63, "learning_rate": 2.883122961945922e-05, "loss": 0.2051, "step": 13485 }, { "epoch": 0.63, "learning_rate": 2.8830761549623205e-05, "loss": 0.2409, "step": 13490 }, { "epoch": 0.63, "learning_rate": 2.8830293479787185e-05, "loss": 0.2244, "step": 13495 }, { "epoch": 0.63, "learning_rate": 2.8829825409951164e-05, "loss": 0.4758, "step": 13500 }, { "epoch": 0.63, "learning_rate": 2.8829357340115144e-05, "loss": 0.3203, "step": 13505 }, { "epoch": 0.63, "learning_rate": 2.8828889270279127e-05, "loss": 0.0932, "step": 13510 }, { "epoch": 0.63, "learning_rate": 2.8828421200443107e-05, "loss": 0.148, "step": 13515 }, { "epoch": 0.63, "learning_rate": 2.8827953130607087e-05, "loss": 0.108, "step": 13520 }, { "epoch": 0.63, "learning_rate": 2.882748506077107e-05, "loss": 0.1227, "step": 13525 }, { "epoch": 0.63, "learning_rate": 2.882701699093505e-05, "loss": 0.2322, "step": 13530 }, { "epoch": 0.63, "learning_rate": 2.882654892109903e-05, "loss": 0.1955, "step": 13535 }, { "epoch": 0.63, "learning_rate": 2.8826080851263006e-05, "loss": 0.2372, "step": 13540 }, { "epoch": 0.63, "learning_rate": 2.882561278142699e-05, "loss": 0.2577, "step": 13545 }, { "epoch": 0.63, "learning_rate": 2.882514471159097e-05, "loss": 0.6495, "step": 13550 }, { "epoch": 0.63, "learning_rate": 2.882467664175495e-05, "loss": 0.3233, "step": 13555 }, { "epoch": 0.63, "learning_rate": 2.882420857191893e-05, "loss": 0.0876, "step": 13560 }, { "epoch": 0.63, "learning_rate": 2.8823740502082912e-05, "loss": 0.0832, "step": 13565 }, { "epoch": 0.63, "learning_rate": 2.8823272432246892e-05, "loss": 0.1516, "step": 13570 }, { "epoch": 0.63, "learning_rate": 2.8822804362410872e-05, "loss": 0.1306, "step": 13575 }, { "epoch": 0.63, "learning_rate": 2.8822336292574852e-05, "loss": 0.189, "step": 13580 }, { "epoch": 0.63, "learning_rate": 2.8821868222738835e-05, "loss": 0.2558, "step": 13585 }, { "epoch": 0.63, "learning_rate": 2.8821400152902815e-05, "loss": 0.2561, "step": 13590 }, { "epoch": 0.63, "learning_rate": 2.8820932083066795e-05, "loss": 0.2606, "step": 13595 }, { "epoch": 0.63, "learning_rate": 2.8820464013230778e-05, "loss": 0.4794, "step": 13600 }, { "epoch": 0.63, "learning_rate": 2.8819995943394754e-05, "loss": 0.1903, "step": 13605 }, { "epoch": 0.64, "learning_rate": 2.8819527873558734e-05, "loss": 0.1164, "step": 13610 }, { "epoch": 0.64, "learning_rate": 2.8819059803722714e-05, "loss": 0.115, "step": 13615 }, { "epoch": 0.64, "learning_rate": 2.8818591733886697e-05, "loss": 0.0967, "step": 13620 }, { "epoch": 0.64, "learning_rate": 2.8818123664050677e-05, "loss": 0.1154, "step": 13625 }, { "epoch": 0.64, "learning_rate": 2.8817655594214657e-05, "loss": 0.247, "step": 13630 }, { "epoch": 0.64, "learning_rate": 2.8817187524378636e-05, "loss": 0.1664, "step": 13635 }, { "epoch": 0.64, "learning_rate": 2.881671945454262e-05, "loss": 0.2397, "step": 13640 }, { "epoch": 0.64, "learning_rate": 2.88162513847066e-05, "loss": 0.1994, "step": 13645 }, { "epoch": 0.64, "learning_rate": 2.881578331487058e-05, "loss": 0.4377, "step": 13650 }, { "epoch": 0.64, "learning_rate": 2.8815315245034563e-05, "loss": 0.2532, "step": 13655 }, { "epoch": 0.64, "learning_rate": 2.8814847175198542e-05, "loss": 0.1245, "step": 13660 }, { "epoch": 0.64, "learning_rate": 2.8814379105362522e-05, "loss": 0.0604, "step": 13665 }, { "epoch": 0.64, "learning_rate": 2.88139110355265e-05, "loss": 0.0789, "step": 13670 }, { "epoch": 0.64, "learning_rate": 2.8813442965690482e-05, "loss": 0.1269, "step": 13675 }, { "epoch": 0.64, "learning_rate": 2.881297489585446e-05, "loss": 0.1837, "step": 13680 }, { "epoch": 0.64, "learning_rate": 2.881250682601844e-05, "loss": 0.1308, "step": 13685 }, { "epoch": 0.64, "learning_rate": 2.881203875618242e-05, "loss": 0.2554, "step": 13690 }, { "epoch": 0.64, "learning_rate": 2.8811570686346404e-05, "loss": 0.2574, "step": 13695 }, { "epoch": 0.64, "learning_rate": 2.8811102616510384e-05, "loss": 0.4211, "step": 13700 }, { "epoch": 0.64, "learning_rate": 2.8810634546674364e-05, "loss": 0.2948, "step": 13705 }, { "epoch": 0.64, "learning_rate": 2.8810166476838347e-05, "loss": 0.0887, "step": 13710 }, { "epoch": 0.64, "learning_rate": 2.8809698407002327e-05, "loss": 0.0908, "step": 13715 }, { "epoch": 0.64, "learning_rate": 2.8809230337166307e-05, "loss": 0.1765, "step": 13720 }, { "epoch": 0.64, "learning_rate": 2.8808762267330287e-05, "loss": 0.2056, "step": 13725 }, { "epoch": 0.64, "learning_rate": 2.8808294197494267e-05, "loss": 0.2093, "step": 13730 }, { "epoch": 0.64, "learning_rate": 2.8807826127658246e-05, "loss": 0.1898, "step": 13735 }, { "epoch": 0.64, "learning_rate": 2.8807358057822226e-05, "loss": 0.3098, "step": 13740 }, { "epoch": 0.64, "learning_rate": 2.8806889987986206e-05, "loss": 0.433, "step": 13745 }, { "epoch": 0.64, "learning_rate": 2.880642191815019e-05, "loss": 0.3335, "step": 13750 }, { "epoch": 0.64, "learning_rate": 2.880595384831417e-05, "loss": 0.2012, "step": 13755 }, { "epoch": 0.64, "learning_rate": 2.880548577847815e-05, "loss": 0.0612, "step": 13760 }, { "epoch": 0.64, "learning_rate": 2.8805017708642132e-05, "loss": 0.0931, "step": 13765 }, { "epoch": 0.64, "learning_rate": 2.8804549638806112e-05, "loss": 0.122, "step": 13770 }, { "epoch": 0.64, "learning_rate": 2.8804081568970092e-05, "loss": 0.1118, "step": 13775 }, { "epoch": 0.64, "learning_rate": 2.880361349913407e-05, "loss": 0.2083, "step": 13780 }, { "epoch": 0.64, "learning_rate": 2.8803145429298055e-05, "loss": 0.1495, "step": 13785 }, { "epoch": 0.64, "learning_rate": 2.8802677359462035e-05, "loss": 0.1985, "step": 13790 }, { "epoch": 0.64, "learning_rate": 2.880220928962601e-05, "loss": 0.2896, "step": 13795 }, { "epoch": 0.64, "learning_rate": 2.880174121978999e-05, "loss": 0.6363, "step": 13800 }, { "epoch": 0.64, "learning_rate": 2.8801273149953974e-05, "loss": 0.1096, "step": 13805 }, { "epoch": 0.64, "learning_rate": 2.8800805080117954e-05, "loss": 0.0928, "step": 13810 }, { "epoch": 0.64, "learning_rate": 2.8800337010281934e-05, "loss": 0.1624, "step": 13815 }, { "epoch": 0.64, "learning_rate": 2.8799868940445913e-05, "loss": 0.1412, "step": 13820 }, { "epoch": 0.65, "learning_rate": 2.8799400870609897e-05, "loss": 0.1277, "step": 13825 }, { "epoch": 0.65, "learning_rate": 2.8798932800773876e-05, "loss": 0.2675, "step": 13830 }, { "epoch": 0.65, "learning_rate": 2.8798464730937856e-05, "loss": 0.3349, "step": 13835 }, { "epoch": 0.65, "learning_rate": 2.879799666110184e-05, "loss": 0.2923, "step": 13840 }, { "epoch": 0.65, "learning_rate": 2.879752859126582e-05, "loss": 0.3276, "step": 13845 }, { "epoch": 0.65, "learning_rate": 2.87970605214298e-05, "loss": 0.606, "step": 13850 }, { "epoch": 0.65, "learning_rate": 2.879659245159378e-05, "loss": 0.2451, "step": 13855 }, { "epoch": 0.65, "learning_rate": 2.879612438175776e-05, "loss": 0.1395, "step": 13860 }, { "epoch": 0.65, "learning_rate": 2.879565631192174e-05, "loss": 0.0936, "step": 13865 }, { "epoch": 0.65, "learning_rate": 2.879518824208572e-05, "loss": 0.164, "step": 13870 }, { "epoch": 0.65, "learning_rate": 2.8794720172249698e-05, "loss": 0.1364, "step": 13875 }, { "epoch": 0.65, "learning_rate": 2.879425210241368e-05, "loss": 0.132, "step": 13880 }, { "epoch": 0.65, "learning_rate": 2.879378403257766e-05, "loss": 0.1665, "step": 13885 }, { "epoch": 0.65, "learning_rate": 2.879331596274164e-05, "loss": 0.285, "step": 13890 }, { "epoch": 0.65, "learning_rate": 2.8792847892905624e-05, "loss": 0.4156, "step": 13895 }, { "epoch": 0.65, "learning_rate": 2.8792379823069604e-05, "loss": 0.6868, "step": 13900 }, { "epoch": 0.65, "learning_rate": 2.8791911753233584e-05, "loss": 0.1785, "step": 13905 }, { "epoch": 0.65, "learning_rate": 2.8791443683397564e-05, "loss": 0.0816, "step": 13910 }, { "epoch": 0.65, "learning_rate": 2.8790975613561547e-05, "loss": 0.0749, "step": 13915 }, { "epoch": 0.65, "learning_rate": 2.8790507543725523e-05, "loss": 0.1498, "step": 13920 }, { "epoch": 0.65, "learning_rate": 2.8790039473889503e-05, "loss": 0.1049, "step": 13925 }, { "epoch": 0.65, "learning_rate": 2.8789571404053483e-05, "loss": 0.1121, "step": 13930 }, { "epoch": 0.65, "learning_rate": 2.8789103334217466e-05, "loss": 0.1455, "step": 13935 }, { "epoch": 0.65, "learning_rate": 2.8788635264381446e-05, "loss": 0.2959, "step": 13940 }, { "epoch": 0.65, "learning_rate": 2.8788167194545426e-05, "loss": 0.2852, "step": 13945 }, { "epoch": 0.65, "learning_rate": 2.878769912470941e-05, "loss": 0.6112, "step": 13950 }, { "epoch": 0.65, "learning_rate": 2.878723105487339e-05, "loss": 0.324, "step": 13955 }, { "epoch": 0.65, "learning_rate": 2.878676298503737e-05, "loss": 0.1254, "step": 13960 }, { "epoch": 0.65, "learning_rate": 2.878629491520135e-05, "loss": 0.0945, "step": 13965 }, { "epoch": 0.65, "learning_rate": 2.8785826845365332e-05, "loss": 0.1694, "step": 13970 }, { "epoch": 0.65, "learning_rate": 2.878535877552931e-05, "loss": 0.1078, "step": 13975 }, { "epoch": 0.65, "learning_rate": 2.878489070569329e-05, "loss": 0.2042, "step": 13980 }, { "epoch": 0.65, "learning_rate": 2.8784422635857268e-05, "loss": 0.2052, "step": 13985 }, { "epoch": 0.65, "learning_rate": 2.878395456602125e-05, "loss": 0.2638, "step": 13990 }, { "epoch": 0.65, "learning_rate": 2.878348649618523e-05, "loss": 0.3099, "step": 13995 }, { "epoch": 0.65, "learning_rate": 2.878301842634921e-05, "loss": 0.3693, "step": 14000 }, { "epoch": 0.65, "learning_rate": 2.878255035651319e-05, "loss": 0.1832, "step": 14005 }, { "epoch": 0.65, "learning_rate": 2.8782082286677174e-05, "loss": 0.0897, "step": 14010 }, { "epoch": 0.65, "learning_rate": 2.8781614216841153e-05, "loss": 0.1162, "step": 14015 }, { "epoch": 0.65, "learning_rate": 2.8781146147005133e-05, "loss": 0.0888, "step": 14020 }, { "epoch": 0.65, "learning_rate": 2.8780678077169116e-05, "loss": 0.0888, "step": 14025 }, { "epoch": 0.65, "learning_rate": 2.8780210007333096e-05, "loss": 0.1766, "step": 14030 }, { "epoch": 0.65, "learning_rate": 2.8779741937497076e-05, "loss": 0.1924, "step": 14035 }, { "epoch": 0.66, "learning_rate": 2.8779273867661056e-05, "loss": 0.2636, "step": 14040 }, { "epoch": 0.66, "learning_rate": 2.8778805797825036e-05, "loss": 0.2952, "step": 14045 }, { "epoch": 0.66, "learning_rate": 2.8778337727989016e-05, "loss": 0.5158, "step": 14050 }, { "epoch": 0.66, "learning_rate": 2.8777869658152995e-05, "loss": 0.223, "step": 14055 }, { "epoch": 0.66, "learning_rate": 2.8777401588316975e-05, "loss": 0.0951, "step": 14060 }, { "epoch": 0.66, "learning_rate": 2.877693351848096e-05, "loss": 0.0751, "step": 14065 }, { "epoch": 0.66, "learning_rate": 2.8776465448644938e-05, "loss": 0.09, "step": 14070 }, { "epoch": 0.66, "learning_rate": 2.8775997378808918e-05, "loss": 0.1597, "step": 14075 }, { "epoch": 0.66, "learning_rate": 2.87755293089729e-05, "loss": 0.0965, "step": 14080 }, { "epoch": 0.66, "learning_rate": 2.877506123913688e-05, "loss": 0.2149, "step": 14085 }, { "epoch": 0.66, "learning_rate": 2.877459316930086e-05, "loss": 0.3045, "step": 14090 }, { "epoch": 0.66, "learning_rate": 2.877412509946484e-05, "loss": 0.3601, "step": 14095 }, { "epoch": 0.66, "learning_rate": 2.8773657029628824e-05, "loss": 0.8148, "step": 14100 }, { "epoch": 0.66, "learning_rate": 2.8773188959792804e-05, "loss": 0.2421, "step": 14105 }, { "epoch": 0.66, "learning_rate": 2.877272088995678e-05, "loss": 0.0657, "step": 14110 }, { "epoch": 0.66, "learning_rate": 2.877225282012076e-05, "loss": 0.097, "step": 14115 }, { "epoch": 0.66, "learning_rate": 2.8771784750284743e-05, "loss": 0.1607, "step": 14120 }, { "epoch": 0.66, "learning_rate": 2.8771316680448723e-05, "loss": 0.1174, "step": 14125 }, { "epoch": 0.66, "learning_rate": 2.8770848610612703e-05, "loss": 0.1596, "step": 14130 }, { "epoch": 0.66, "learning_rate": 2.8770380540776686e-05, "loss": 0.2451, "step": 14135 }, { "epoch": 0.66, "learning_rate": 2.8769912470940666e-05, "loss": 0.2717, "step": 14140 }, { "epoch": 0.66, "learning_rate": 2.8769444401104646e-05, "loss": 0.4318, "step": 14145 }, { "epoch": 0.66, "learning_rate": 2.8768976331268625e-05, "loss": 0.591, "step": 14150 }, { "epoch": 0.66, "learning_rate": 2.876850826143261e-05, "loss": 0.2317, "step": 14155 }, { "epoch": 0.66, "learning_rate": 2.876804019159659e-05, "loss": 0.0934, "step": 14160 }, { "epoch": 0.66, "learning_rate": 2.8767572121760568e-05, "loss": 0.0979, "step": 14165 }, { "epoch": 0.66, "learning_rate": 2.8767104051924548e-05, "loss": 0.0935, "step": 14170 }, { "epoch": 0.66, "learning_rate": 2.8766635982088528e-05, "loss": 0.1455, "step": 14175 }, { "epoch": 0.66, "learning_rate": 2.8766167912252508e-05, "loss": 0.207, "step": 14180 }, { "epoch": 0.66, "learning_rate": 2.8765699842416488e-05, "loss": 0.1849, "step": 14185 }, { "epoch": 0.66, "learning_rate": 2.8765231772580467e-05, "loss": 0.2923, "step": 14190 }, { "epoch": 0.66, "learning_rate": 2.876476370274445e-05, "loss": 0.2744, "step": 14195 }, { "epoch": 0.66, "learning_rate": 2.876429563290843e-05, "loss": 0.4343, "step": 14200 }, { "epoch": 0.66, "learning_rate": 2.876382756307241e-05, "loss": 0.184, "step": 14205 }, { "epoch": 0.66, "learning_rate": 2.8763359493236393e-05, "loss": 0.0398, "step": 14210 }, { "epoch": 0.66, "learning_rate": 2.8762891423400373e-05, "loss": 0.1524, "step": 14215 }, { "epoch": 0.66, "learning_rate": 2.8762423353564353e-05, "loss": 0.1415, "step": 14220 }, { "epoch": 0.66, "learning_rate": 2.8761955283728333e-05, "loss": 0.1488, "step": 14225 }, { "epoch": 0.66, "learning_rate": 2.8761487213892316e-05, "loss": 0.1139, "step": 14230 }, { "epoch": 0.66, "learning_rate": 2.8761019144056293e-05, "loss": 0.2217, "step": 14235 }, { "epoch": 0.66, "learning_rate": 2.8760551074220272e-05, "loss": 0.27, "step": 14240 }, { "epoch": 0.66, "learning_rate": 2.8760083004384252e-05, "loss": 0.3859, "step": 14245 }, { "epoch": 0.66, "learning_rate": 2.8759614934548235e-05, "loss": 0.4615, "step": 14250 }, { "epoch": 0.67, "learning_rate": 2.8759146864712215e-05, "loss": 0.2342, "step": 14255 }, { "epoch": 0.67, "learning_rate": 2.8758678794876195e-05, "loss": 0.0786, "step": 14260 }, { "epoch": 0.67, "learning_rate": 2.8758210725040178e-05, "loss": 0.1323, "step": 14265 }, { "epoch": 0.67, "learning_rate": 2.8757742655204158e-05, "loss": 0.1346, "step": 14270 }, { "epoch": 0.67, "learning_rate": 2.8757274585368138e-05, "loss": 0.115, "step": 14275 }, { "epoch": 0.67, "learning_rate": 2.8756806515532118e-05, "loss": 0.1903, "step": 14280 }, { "epoch": 0.67, "learning_rate": 2.87563384456961e-05, "loss": 0.2855, "step": 14285 }, { "epoch": 0.67, "learning_rate": 2.875587037586008e-05, "loss": 0.2659, "step": 14290 }, { "epoch": 0.67, "learning_rate": 2.875540230602406e-05, "loss": 0.5062, "step": 14295 }, { "epoch": 0.67, "learning_rate": 2.8754934236188037e-05, "loss": 0.5901, "step": 14300 }, { "epoch": 0.67, "learning_rate": 2.875446616635202e-05, "loss": 0.1554, "step": 14305 }, { "epoch": 0.67, "learning_rate": 2.8753998096516e-05, "loss": 0.1165, "step": 14310 }, { "epoch": 0.67, "learning_rate": 2.875353002667998e-05, "loss": 0.0867, "step": 14315 }, { "epoch": 0.67, "learning_rate": 2.8753061956843963e-05, "loss": 0.1441, "step": 14320 }, { "epoch": 0.67, "learning_rate": 2.8752593887007943e-05, "loss": 0.1574, "step": 14325 }, { "epoch": 0.67, "learning_rate": 2.8752125817171923e-05, "loss": 0.1679, "step": 14330 }, { "epoch": 0.67, "learning_rate": 2.8751657747335902e-05, "loss": 0.1935, "step": 14335 }, { "epoch": 0.67, "learning_rate": 2.8751189677499886e-05, "loss": 0.1993, "step": 14340 }, { "epoch": 0.67, "learning_rate": 2.8750721607663865e-05, "loss": 0.295, "step": 14345 }, { "epoch": 0.67, "learning_rate": 2.8750253537827845e-05, "loss": 0.6586, "step": 14350 }, { "epoch": 0.67, "learning_rate": 2.8749785467991825e-05, "loss": 0.1668, "step": 14355 }, { "epoch": 0.67, "learning_rate": 2.8749317398155808e-05, "loss": 0.0989, "step": 14360 }, { "epoch": 0.67, "learning_rate": 2.8748849328319785e-05, "loss": 0.0756, "step": 14365 }, { "epoch": 0.67, "learning_rate": 2.8748381258483765e-05, "loss": 0.0983, "step": 14370 }, { "epoch": 0.67, "learning_rate": 2.8747913188647744e-05, "loss": 0.1155, "step": 14375 }, { "epoch": 0.67, "learning_rate": 2.8747445118811728e-05, "loss": 0.1271, "step": 14380 }, { "epoch": 0.67, "learning_rate": 2.8746977048975707e-05, "loss": 0.2332, "step": 14385 }, { "epoch": 0.67, "learning_rate": 2.8746508979139687e-05, "loss": 0.1807, "step": 14390 }, { "epoch": 0.67, "learning_rate": 2.874604090930367e-05, "loss": 0.3818, "step": 14395 }, { "epoch": 0.67, "learning_rate": 2.874557283946765e-05, "loss": 0.5533, "step": 14400 }, { "epoch": 0.67, "learning_rate": 2.874510476963163e-05, "loss": 0.2287, "step": 14405 }, { "epoch": 0.67, "learning_rate": 2.874463669979561e-05, "loss": 0.0701, "step": 14410 }, { "epoch": 0.67, "learning_rate": 2.8744168629959593e-05, "loss": 0.0954, "step": 14415 }, { "epoch": 0.67, "learning_rate": 2.8743700560123573e-05, "loss": 0.1512, "step": 14420 }, { "epoch": 0.67, "learning_rate": 2.874323249028755e-05, "loss": 0.156, "step": 14425 }, { "epoch": 0.67, "learning_rate": 2.874276442045153e-05, "loss": 0.1778, "step": 14430 }, { "epoch": 0.67, "learning_rate": 2.8742296350615512e-05, "loss": 0.1784, "step": 14435 }, { "epoch": 0.67, "learning_rate": 2.8741828280779492e-05, "loss": 0.1927, "step": 14440 }, { "epoch": 0.67, "learning_rate": 2.8741360210943472e-05, "loss": 0.2273, "step": 14445 }, { "epoch": 0.67, "learning_rate": 2.8740892141107455e-05, "loss": 0.4766, "step": 14450 }, { "epoch": 0.67, "learning_rate": 2.8740424071271435e-05, "loss": 0.2449, "step": 14455 }, { "epoch": 0.67, "learning_rate": 2.8739956001435415e-05, "loss": 0.0797, "step": 14460 }, { "epoch": 0.67, "learning_rate": 2.8739487931599395e-05, "loss": 0.1437, "step": 14465 }, { "epoch": 0.68, "learning_rate": 2.8739019861763378e-05, "loss": 0.1084, "step": 14470 }, { "epoch": 0.68, "learning_rate": 2.8738551791927358e-05, "loss": 0.1042, "step": 14475 }, { "epoch": 0.68, "learning_rate": 2.8738083722091337e-05, "loss": 0.2117, "step": 14480 }, { "epoch": 0.68, "learning_rate": 2.8737615652255317e-05, "loss": 0.2622, "step": 14485 }, { "epoch": 0.68, "learning_rate": 2.8737147582419297e-05, "loss": 0.2691, "step": 14490 }, { "epoch": 0.68, "learning_rate": 2.8736679512583277e-05, "loss": 0.3481, "step": 14495 }, { "epoch": 0.68, "learning_rate": 2.8736211442747257e-05, "loss": 0.3504, "step": 14500 }, { "epoch": 0.68, "learning_rate": 2.873574337291124e-05, "loss": 0.264, "step": 14505 }, { "epoch": 0.68, "learning_rate": 2.873527530307522e-05, "loss": 0.0905, "step": 14510 }, { "epoch": 0.68, "learning_rate": 2.87348072332392e-05, "loss": 0.07, "step": 14515 }, { "epoch": 0.68, "learning_rate": 2.873433916340318e-05, "loss": 0.1167, "step": 14520 }, { "epoch": 0.68, "learning_rate": 2.8733871093567163e-05, "loss": 0.1427, "step": 14525 }, { "epoch": 0.68, "learning_rate": 2.8733403023731142e-05, "loss": 0.1946, "step": 14530 }, { "epoch": 0.68, "learning_rate": 2.8732934953895122e-05, "loss": 0.2287, "step": 14535 }, { "epoch": 0.68, "learning_rate": 2.8732466884059102e-05, "loss": 0.2621, "step": 14540 }, { "epoch": 0.68, "learning_rate": 2.8731998814223085e-05, "loss": 0.3953, "step": 14545 }, { "epoch": 0.68, "learning_rate": 2.8731530744387065e-05, "loss": 0.5614, "step": 14550 }, { "epoch": 0.68, "learning_rate": 2.873106267455104e-05, "loss": 0.1823, "step": 14555 }, { "epoch": 0.68, "learning_rate": 2.8730594604715025e-05, "loss": 0.0456, "step": 14560 }, { "epoch": 0.68, "learning_rate": 2.8730126534879005e-05, "loss": 0.0692, "step": 14565 }, { "epoch": 0.68, "learning_rate": 2.8729658465042984e-05, "loss": 0.1398, "step": 14570 }, { "epoch": 0.68, "learning_rate": 2.8729190395206964e-05, "loss": 0.1686, "step": 14575 }, { "epoch": 0.68, "learning_rate": 2.8728722325370947e-05, "loss": 0.0955, "step": 14580 }, { "epoch": 0.68, "learning_rate": 2.8728254255534927e-05, "loss": 0.1732, "step": 14585 }, { "epoch": 0.68, "learning_rate": 2.8727786185698907e-05, "loss": 0.2574, "step": 14590 }, { "epoch": 0.68, "learning_rate": 2.8727318115862887e-05, "loss": 0.3676, "step": 14595 }, { "epoch": 0.68, "learning_rate": 2.872685004602687e-05, "loss": 0.29, "step": 14600 }, { "epoch": 0.68, "learning_rate": 2.872638197619085e-05, "loss": 0.199, "step": 14605 }, { "epoch": 0.68, "learning_rate": 2.872591390635483e-05, "loss": 0.073, "step": 14610 }, { "epoch": 0.68, "learning_rate": 2.8725445836518806e-05, "loss": 0.0835, "step": 14615 }, { "epoch": 0.68, "learning_rate": 2.872497776668279e-05, "loss": 0.1706, "step": 14620 }, { "epoch": 0.68, "learning_rate": 2.872450969684677e-05, "loss": 0.1481, "step": 14625 }, { "epoch": 0.68, "learning_rate": 2.872404162701075e-05, "loss": 0.1746, "step": 14630 }, { "epoch": 0.68, "learning_rate": 2.8723573557174732e-05, "loss": 0.1933, "step": 14635 }, { "epoch": 0.68, "learning_rate": 2.8723105487338712e-05, "loss": 0.2433, "step": 14640 }, { "epoch": 0.68, "learning_rate": 2.8722637417502692e-05, "loss": 0.3401, "step": 14645 }, { "epoch": 0.68, "learning_rate": 2.872216934766667e-05, "loss": 0.3509, "step": 14650 }, { "epoch": 0.68, "learning_rate": 2.8721701277830655e-05, "loss": 0.2608, "step": 14655 }, { "epoch": 0.68, "learning_rate": 2.8721233207994635e-05, "loss": 0.0762, "step": 14660 }, { "epoch": 0.68, "learning_rate": 2.8720765138158614e-05, "loss": 0.1259, "step": 14665 }, { "epoch": 0.68, "learning_rate": 2.8720297068322594e-05, "loss": 0.1109, "step": 14670 }, { "epoch": 0.68, "learning_rate": 2.8719828998486577e-05, "loss": 0.0788, "step": 14675 }, { "epoch": 0.68, "learning_rate": 2.8719360928650554e-05, "loss": 0.2035, "step": 14680 }, { "epoch": 0.69, "learning_rate": 2.8718892858814534e-05, "loss": 0.2886, "step": 14685 }, { "epoch": 0.69, "learning_rate": 2.8718424788978517e-05, "loss": 0.1857, "step": 14690 }, { "epoch": 0.69, "learning_rate": 2.8717956719142497e-05, "loss": 0.3754, "step": 14695 }, { "epoch": 0.69, "learning_rate": 2.8717488649306477e-05, "loss": 0.6902, "step": 14700 }, { "epoch": 0.69, "learning_rate": 2.8717020579470456e-05, "loss": 0.2214, "step": 14705 }, { "epoch": 0.69, "learning_rate": 2.871655250963444e-05, "loss": 0.0739, "step": 14710 }, { "epoch": 0.69, "learning_rate": 2.871608443979842e-05, "loss": 0.0544, "step": 14715 }, { "epoch": 0.69, "learning_rate": 2.87156163699624e-05, "loss": 0.0543, "step": 14720 }, { "epoch": 0.69, "learning_rate": 2.871514830012638e-05, "loss": 0.1311, "step": 14725 }, { "epoch": 0.69, "learning_rate": 2.8714680230290362e-05, "loss": 0.1319, "step": 14730 }, { "epoch": 0.69, "learning_rate": 2.8714212160454342e-05, "loss": 0.1843, "step": 14735 }, { "epoch": 0.69, "learning_rate": 2.871374409061832e-05, "loss": 0.3472, "step": 14740 }, { "epoch": 0.69, "learning_rate": 2.8713276020782302e-05, "loss": 0.2921, "step": 14745 }, { "epoch": 0.69, "learning_rate": 2.871280795094628e-05, "loss": 0.5515, "step": 14750 }, { "epoch": 0.69, "learning_rate": 2.871233988111026e-05, "loss": 0.2132, "step": 14755 }, { "epoch": 0.69, "learning_rate": 2.871187181127424e-05, "loss": 0.1453, "step": 14760 }, { "epoch": 0.69, "learning_rate": 2.8711403741438224e-05, "loss": 0.1433, "step": 14765 }, { "epoch": 0.69, "learning_rate": 2.8710935671602204e-05, "loss": 0.1084, "step": 14770 }, { "epoch": 0.69, "learning_rate": 2.8710467601766184e-05, "loss": 0.1699, "step": 14775 }, { "epoch": 0.69, "learning_rate": 2.8709999531930164e-05, "loss": 0.1344, "step": 14780 }, { "epoch": 0.69, "learning_rate": 2.8709531462094147e-05, "loss": 0.1827, "step": 14785 }, { "epoch": 0.69, "learning_rate": 2.8709063392258127e-05, "loss": 0.2047, "step": 14790 }, { "epoch": 0.69, "learning_rate": 2.8708595322422107e-05, "loss": 0.4195, "step": 14795 }, { "epoch": 0.69, "learning_rate": 2.8708127252586086e-05, "loss": 0.6423, "step": 14800 }, { "epoch": 0.69, "learning_rate": 2.8707659182750066e-05, "loss": 0.2532, "step": 14805 }, { "epoch": 0.69, "learning_rate": 2.8707191112914046e-05, "loss": 0.0983, "step": 14810 }, { "epoch": 0.69, "learning_rate": 2.8706723043078026e-05, "loss": 0.1019, "step": 14815 }, { "epoch": 0.69, "learning_rate": 2.870625497324201e-05, "loss": 0.0828, "step": 14820 }, { "epoch": 0.69, "learning_rate": 2.870578690340599e-05, "loss": 0.1249, "step": 14825 }, { "epoch": 0.69, "learning_rate": 2.870531883356997e-05, "loss": 0.1412, "step": 14830 }, { "epoch": 0.69, "learning_rate": 2.870485076373395e-05, "loss": 0.3927, "step": 14835 }, { "epoch": 0.69, "learning_rate": 2.8704382693897932e-05, "loss": 0.2275, "step": 14840 }, { "epoch": 0.69, "learning_rate": 2.870391462406191e-05, "loss": 0.2967, "step": 14845 }, { "epoch": 0.69, "learning_rate": 2.870344655422589e-05, "loss": 0.5893, "step": 14850 }, { "epoch": 0.69, "learning_rate": 2.870297848438987e-05, "loss": 0.3111, "step": 14855 }, { "epoch": 0.69, "learning_rate": 2.8702510414553854e-05, "loss": 0.0802, "step": 14860 }, { "epoch": 0.69, "learning_rate": 2.8702042344717834e-05, "loss": 0.0921, "step": 14865 }, { "epoch": 0.69, "learning_rate": 2.870157427488181e-05, "loss": 0.0865, "step": 14870 }, { "epoch": 0.69, "learning_rate": 2.8701106205045794e-05, "loss": 0.2425, "step": 14875 }, { "epoch": 0.69, "learning_rate": 2.8700638135209774e-05, "loss": 0.1841, "step": 14880 }, { "epoch": 0.69, "learning_rate": 2.8700170065373754e-05, "loss": 0.1941, "step": 14885 }, { "epoch": 0.69, "learning_rate": 2.8699701995537733e-05, "loss": 0.2457, "step": 14890 }, { "epoch": 0.7, "learning_rate": 2.8699233925701717e-05, "loss": 0.278, "step": 14895 }, { "epoch": 0.7, "learning_rate": 2.8698765855865696e-05, "loss": 0.4692, "step": 14900 }, { "epoch": 0.7, "learning_rate": 2.8698297786029676e-05, "loss": 0.2133, "step": 14905 }, { "epoch": 0.7, "learning_rate": 2.8697829716193656e-05, "loss": 0.065, "step": 14910 }, { "epoch": 0.7, "learning_rate": 2.869736164635764e-05, "loss": 0.105, "step": 14915 }, { "epoch": 0.7, "learning_rate": 2.869689357652162e-05, "loss": 0.0833, "step": 14920 }, { "epoch": 0.7, "learning_rate": 2.86964255066856e-05, "loss": 0.1978, "step": 14925 }, { "epoch": 0.7, "learning_rate": 2.869595743684958e-05, "loss": 0.1104, "step": 14930 }, { "epoch": 0.7, "learning_rate": 2.869548936701356e-05, "loss": 0.1582, "step": 14935 }, { "epoch": 0.7, "learning_rate": 2.8695021297177538e-05, "loss": 0.2012, "step": 14940 }, { "epoch": 0.7, "learning_rate": 2.8694553227341518e-05, "loss": 0.3984, "step": 14945 }, { "epoch": 0.7, "learning_rate": 2.86940851575055e-05, "loss": 0.7813, "step": 14950 }, { "epoch": 0.7, "learning_rate": 2.869361708766948e-05, "loss": 0.1963, "step": 14955 }, { "epoch": 0.7, "learning_rate": 2.869314901783346e-05, "loss": 0.0924, "step": 14960 }, { "epoch": 0.7, "learning_rate": 2.869268094799744e-05, "loss": 0.1491, "step": 14965 }, { "epoch": 0.7, "learning_rate": 2.8692212878161424e-05, "loss": 0.0754, "step": 14970 }, { "epoch": 0.7, "learning_rate": 2.8691744808325404e-05, "loss": 0.1767, "step": 14975 }, { "epoch": 0.7, "learning_rate": 2.8691276738489384e-05, "loss": 0.2417, "step": 14980 }, { "epoch": 0.7, "learning_rate": 2.8690808668653363e-05, "loss": 0.1983, "step": 14985 }, { "epoch": 0.7, "learning_rate": 2.8690340598817347e-05, "loss": 0.2171, "step": 14990 }, { "epoch": 0.7, "learning_rate": 2.8689872528981323e-05, "loss": 0.3203, "step": 14995 }, { "epoch": 0.7, "learning_rate": 2.8689404459145303e-05, "loss": 0.6601, "step": 15000 }, { "epoch": 0.7, "learning_rate": 2.8688936389309286e-05, "loss": 0.3103, "step": 15005 }, { "epoch": 0.7, "learning_rate": 2.8688468319473266e-05, "loss": 0.0684, "step": 15010 }, { "epoch": 0.7, "learning_rate": 2.8688000249637246e-05, "loss": 0.157, "step": 15015 }, { "epoch": 0.7, "learning_rate": 2.8687532179801226e-05, "loss": 0.1252, "step": 15020 }, { "epoch": 0.7, "learning_rate": 2.868706410996521e-05, "loss": 0.1809, "step": 15025 }, { "epoch": 0.7, "learning_rate": 2.868659604012919e-05, "loss": 0.1075, "step": 15030 }, { "epoch": 0.7, "learning_rate": 2.868612797029317e-05, "loss": 0.1389, "step": 15035 }, { "epoch": 0.7, "learning_rate": 2.8685659900457148e-05, "loss": 0.1951, "step": 15040 }, { "epoch": 0.7, "learning_rate": 2.868519183062113e-05, "loss": 0.2842, "step": 15045 }, { "epoch": 0.7, "learning_rate": 2.868472376078511e-05, "loss": 0.51, "step": 15050 }, { "epoch": 0.7, "learning_rate": 2.868425569094909e-05, "loss": 0.2096, "step": 15055 }, { "epoch": 0.7, "learning_rate": 2.868378762111307e-05, "loss": 0.1046, "step": 15060 }, { "epoch": 0.7, "learning_rate": 2.868331955127705e-05, "loss": 0.0639, "step": 15065 }, { "epoch": 0.7, "learning_rate": 2.868285148144103e-05, "loss": 0.1285, "step": 15070 }, { "epoch": 0.7, "learning_rate": 2.868238341160501e-05, "loss": 0.1548, "step": 15075 }, { "epoch": 0.7, "learning_rate": 2.8681915341768994e-05, "loss": 0.2482, "step": 15080 }, { "epoch": 0.7, "learning_rate": 2.8681447271932973e-05, "loss": 0.1444, "step": 15085 }, { "epoch": 0.7, "learning_rate": 2.8680979202096953e-05, "loss": 0.2597, "step": 15090 }, { "epoch": 0.7, "learning_rate": 2.8680511132260933e-05, "loss": 0.2534, "step": 15095 }, { "epoch": 0.7, "learning_rate": 2.8680043062424916e-05, "loss": 0.4651, "step": 15100 }, { "epoch": 0.7, "learning_rate": 2.8679574992588896e-05, "loss": 0.2288, "step": 15105 }, { "epoch": 0.71, "learning_rate": 2.8679106922752876e-05, "loss": 0.0982, "step": 15110 }, { "epoch": 0.71, "learning_rate": 2.867863885291686e-05, "loss": 0.0942, "step": 15115 }, { "epoch": 0.71, "learning_rate": 2.8678170783080835e-05, "loss": 0.1433, "step": 15120 }, { "epoch": 0.71, "learning_rate": 2.8677702713244815e-05, "loss": 0.1679, "step": 15125 }, { "epoch": 0.71, "learning_rate": 2.8677234643408795e-05, "loss": 0.1222, "step": 15130 }, { "epoch": 0.71, "learning_rate": 2.8676766573572778e-05, "loss": 0.2887, "step": 15135 }, { "epoch": 0.71, "learning_rate": 2.8676298503736758e-05, "loss": 0.2913, "step": 15140 }, { "epoch": 0.71, "learning_rate": 2.8675830433900738e-05, "loss": 0.4461, "step": 15145 }, { "epoch": 0.71, "learning_rate": 2.8675362364064718e-05, "loss": 0.5593, "step": 15150 }, { "epoch": 0.71, "learning_rate": 2.86748942942287e-05, "loss": 0.2748, "step": 15155 }, { "epoch": 0.71, "learning_rate": 2.867442622439268e-05, "loss": 0.0439, "step": 15160 }, { "epoch": 0.71, "learning_rate": 2.867395815455666e-05, "loss": 0.08, "step": 15165 }, { "epoch": 0.71, "learning_rate": 2.8673490084720644e-05, "loss": 0.1478, "step": 15170 }, { "epoch": 0.71, "learning_rate": 2.8673022014884624e-05, "loss": 0.1772, "step": 15175 }, { "epoch": 0.71, "learning_rate": 2.8672553945048603e-05, "loss": 0.2313, "step": 15180 }, { "epoch": 0.71, "learning_rate": 2.867208587521258e-05, "loss": 0.1585, "step": 15185 }, { "epoch": 0.71, "learning_rate": 2.8671617805376563e-05, "loss": 0.2704, "step": 15190 }, { "epoch": 0.71, "learning_rate": 2.8671149735540543e-05, "loss": 0.3115, "step": 15195 }, { "epoch": 0.71, "learning_rate": 2.8670681665704523e-05, "loss": 0.5487, "step": 15200 }, { "epoch": 0.71, "learning_rate": 2.8670213595868503e-05, "loss": 0.2093, "step": 15205 }, { "epoch": 0.71, "learning_rate": 2.8669745526032486e-05, "loss": 0.0895, "step": 15210 }, { "epoch": 0.71, "learning_rate": 2.8669277456196466e-05, "loss": 0.0847, "step": 15215 }, { "epoch": 0.71, "learning_rate": 2.8668809386360445e-05, "loss": 0.0671, "step": 15220 }, { "epoch": 0.71, "learning_rate": 2.8668341316524425e-05, "loss": 0.1134, "step": 15225 }, { "epoch": 0.71, "learning_rate": 2.866787324668841e-05, "loss": 0.191, "step": 15230 }, { "epoch": 0.71, "learning_rate": 2.8667405176852388e-05, "loss": 0.2068, "step": 15235 }, { "epoch": 0.71, "learning_rate": 2.8666937107016368e-05, "loss": 0.2486, "step": 15240 }, { "epoch": 0.71, "learning_rate": 2.8666469037180348e-05, "loss": 0.3443, "step": 15245 }, { "epoch": 0.71, "learning_rate": 2.8666000967344328e-05, "loss": 0.4892, "step": 15250 }, { "epoch": 0.71, "learning_rate": 2.8665532897508307e-05, "loss": 0.2362, "step": 15255 }, { "epoch": 0.71, "learning_rate": 2.8665064827672287e-05, "loss": 0.1597, "step": 15260 }, { "epoch": 0.71, "learning_rate": 2.866459675783627e-05, "loss": 0.061, "step": 15265 }, { "epoch": 0.71, "learning_rate": 2.866412868800025e-05, "loss": 0.0928, "step": 15270 }, { "epoch": 0.71, "learning_rate": 2.866366061816423e-05, "loss": 0.1497, "step": 15275 }, { "epoch": 0.71, "learning_rate": 2.866319254832821e-05, "loss": 0.3184, "step": 15280 }, { "epoch": 0.71, "learning_rate": 2.8662724478492193e-05, "loss": 0.2595, "step": 15285 }, { "epoch": 0.71, "learning_rate": 2.8662256408656173e-05, "loss": 0.3638, "step": 15290 }, { "epoch": 0.71, "learning_rate": 2.8661788338820153e-05, "loss": 0.2282, "step": 15295 }, { "epoch": 0.71, "learning_rate": 2.8661320268984136e-05, "loss": 0.3601, "step": 15300 }, { "epoch": 0.71, "learning_rate": 2.8660852199148116e-05, "loss": 0.256, "step": 15305 }, { "epoch": 0.71, "learning_rate": 2.8660384129312092e-05, "loss": 0.0989, "step": 15310 }, { "epoch": 0.71, "learning_rate": 2.8659916059476072e-05, "loss": 0.0888, "step": 15315 }, { "epoch": 0.71, "learning_rate": 2.8659447989640055e-05, "loss": 0.093, "step": 15320 }, { "epoch": 0.72, "learning_rate": 2.8658979919804035e-05, "loss": 0.1509, "step": 15325 }, { "epoch": 0.72, "learning_rate": 2.8658511849968015e-05, "loss": 0.192, "step": 15330 }, { "epoch": 0.72, "learning_rate": 2.8658043780131995e-05, "loss": 0.1538, "step": 15335 }, { "epoch": 0.72, "learning_rate": 2.8657575710295978e-05, "loss": 0.28, "step": 15340 }, { "epoch": 0.72, "learning_rate": 2.8657107640459958e-05, "loss": 0.3267, "step": 15345 }, { "epoch": 0.72, "learning_rate": 2.8656639570623938e-05, "loss": 0.3868, "step": 15350 }, { "epoch": 0.72, "learning_rate": 2.865617150078792e-05, "loss": 0.1887, "step": 15355 }, { "epoch": 0.72, "learning_rate": 2.86557034309519e-05, "loss": 0.0915, "step": 15360 }, { "epoch": 0.72, "learning_rate": 2.865523536111588e-05, "loss": 0.1361, "step": 15365 }, { "epoch": 0.72, "learning_rate": 2.865476729127986e-05, "loss": 0.1772, "step": 15370 }, { "epoch": 0.72, "learning_rate": 2.865429922144384e-05, "loss": 0.0874, "step": 15375 }, { "epoch": 0.72, "learning_rate": 2.865383115160782e-05, "loss": 0.1635, "step": 15380 }, { "epoch": 0.72, "learning_rate": 2.86533630817718e-05, "loss": 0.1755, "step": 15385 }, { "epoch": 0.72, "learning_rate": 2.865289501193578e-05, "loss": 0.3083, "step": 15390 }, { "epoch": 0.72, "learning_rate": 2.8652426942099763e-05, "loss": 0.3658, "step": 15395 }, { "epoch": 0.72, "learning_rate": 2.8651958872263743e-05, "loss": 0.4214, "step": 15400 }, { "epoch": 0.72, "learning_rate": 2.8651490802427722e-05, "loss": 0.1958, "step": 15405 }, { "epoch": 0.72, "learning_rate": 2.8651022732591702e-05, "loss": 0.0651, "step": 15410 }, { "epoch": 0.72, "learning_rate": 2.8650554662755685e-05, "loss": 0.0996, "step": 15415 }, { "epoch": 0.72, "learning_rate": 2.8650086592919665e-05, "loss": 0.1533, "step": 15420 }, { "epoch": 0.72, "learning_rate": 2.8649618523083645e-05, "loss": 0.151, "step": 15425 }, { "epoch": 0.72, "learning_rate": 2.8649150453247628e-05, "loss": 0.2203, "step": 15430 }, { "epoch": 0.72, "learning_rate": 2.8648682383411605e-05, "loss": 0.242, "step": 15435 }, { "epoch": 0.72, "learning_rate": 2.8648214313575584e-05, "loss": 0.1803, "step": 15440 }, { "epoch": 0.72, "learning_rate": 2.8647746243739564e-05, "loss": 0.3412, "step": 15445 }, { "epoch": 0.72, "learning_rate": 2.8647278173903547e-05, "loss": 0.4736, "step": 15450 }, { "epoch": 0.72, "learning_rate": 2.8646810104067527e-05, "loss": 0.2068, "step": 15455 }, { "epoch": 0.72, "learning_rate": 2.8646342034231507e-05, "loss": 0.1948, "step": 15460 }, { "epoch": 0.72, "learning_rate": 2.8645873964395487e-05, "loss": 0.0944, "step": 15465 }, { "epoch": 0.72, "learning_rate": 2.864540589455947e-05, "loss": 0.1068, "step": 15470 }, { "epoch": 0.72, "learning_rate": 2.864493782472345e-05, "loss": 0.2815, "step": 15475 }, { "epoch": 0.72, "learning_rate": 2.864446975488743e-05, "loss": 0.1155, "step": 15480 }, { "epoch": 0.72, "learning_rate": 2.8644001685051413e-05, "loss": 0.1846, "step": 15485 }, { "epoch": 0.72, "learning_rate": 2.8643533615215393e-05, "loss": 0.2807, "step": 15490 }, { "epoch": 0.72, "learning_rate": 2.8643065545379373e-05, "loss": 0.3434, "step": 15495 }, { "epoch": 0.72, "learning_rate": 2.864259747554335e-05, "loss": 0.6789, "step": 15500 }, { "epoch": 0.72, "learning_rate": 2.8642129405707332e-05, "loss": 0.2136, "step": 15505 }, { "epoch": 0.72, "learning_rate": 2.8641661335871312e-05, "loss": 0.0901, "step": 15510 }, { "epoch": 0.72, "learning_rate": 2.8641193266035292e-05, "loss": 0.1024, "step": 15515 }, { "epoch": 0.72, "learning_rate": 2.864072519619927e-05, "loss": 0.1646, "step": 15520 }, { "epoch": 0.72, "learning_rate": 2.8640257126363255e-05, "loss": 0.223, "step": 15525 }, { "epoch": 0.72, "learning_rate": 2.8639789056527235e-05, "loss": 0.1807, "step": 15530 }, { "epoch": 0.72, "learning_rate": 2.8639320986691215e-05, "loss": 0.1904, "step": 15535 }, { "epoch": 0.73, "learning_rate": 2.8638852916855198e-05, "loss": 0.236, "step": 15540 }, { "epoch": 0.73, "learning_rate": 2.8638384847019178e-05, "loss": 0.2811, "step": 15545 }, { "epoch": 0.73, "learning_rate": 2.8637916777183157e-05, "loss": 0.6167, "step": 15550 }, { "epoch": 0.73, "learning_rate": 2.8637448707347137e-05, "loss": 0.1889, "step": 15555 }, { "epoch": 0.73, "learning_rate": 2.863698063751112e-05, "loss": 0.1408, "step": 15560 }, { "epoch": 0.73, "learning_rate": 2.8636512567675097e-05, "loss": 0.0766, "step": 15565 }, { "epoch": 0.73, "learning_rate": 2.8636044497839077e-05, "loss": 0.1584, "step": 15570 }, { "epoch": 0.73, "learning_rate": 2.8635576428003056e-05, "loss": 0.1382, "step": 15575 }, { "epoch": 0.73, "learning_rate": 2.863510835816704e-05, "loss": 0.1553, "step": 15580 }, { "epoch": 0.73, "learning_rate": 2.863464028833102e-05, "loss": 0.1733, "step": 15585 }, { "epoch": 0.73, "learning_rate": 2.8634172218495e-05, "loss": 0.1479, "step": 15590 }, { "epoch": 0.73, "learning_rate": 2.863370414865898e-05, "loss": 0.4117, "step": 15595 }, { "epoch": 0.73, "learning_rate": 2.8633236078822962e-05, "loss": 0.3843, "step": 15600 }, { "epoch": 0.73, "learning_rate": 2.8632768008986942e-05, "loss": 0.2137, "step": 15605 }, { "epoch": 0.73, "learning_rate": 2.8632299939150922e-05, "loss": 0.0529, "step": 15610 }, { "epoch": 0.73, "learning_rate": 2.8631831869314905e-05, "loss": 0.1212, "step": 15615 }, { "epoch": 0.73, "learning_rate": 2.8631363799478885e-05, "loss": 0.1558, "step": 15620 }, { "epoch": 0.73, "learning_rate": 2.863089572964286e-05, "loss": 0.1929, "step": 15625 }, { "epoch": 0.73, "learning_rate": 2.863042765980684e-05, "loss": 0.1506, "step": 15630 }, { "epoch": 0.73, "learning_rate": 2.8629959589970824e-05, "loss": 0.2172, "step": 15635 }, { "epoch": 0.73, "learning_rate": 2.8629491520134804e-05, "loss": 0.2056, "step": 15640 }, { "epoch": 0.73, "learning_rate": 2.8629023450298784e-05, "loss": 0.3093, "step": 15645 }, { "epoch": 0.73, "learning_rate": 2.8628555380462764e-05, "loss": 0.4196, "step": 15650 }, { "epoch": 0.73, "learning_rate": 2.8628087310626747e-05, "loss": 0.2582, "step": 15655 }, { "epoch": 0.73, "learning_rate": 2.8627619240790727e-05, "loss": 0.0894, "step": 15660 }, { "epoch": 0.73, "learning_rate": 2.8627151170954707e-05, "loss": 0.1122, "step": 15665 }, { "epoch": 0.73, "learning_rate": 2.862668310111869e-05, "loss": 0.1188, "step": 15670 }, { "epoch": 0.73, "learning_rate": 2.862621503128267e-05, "loss": 0.1679, "step": 15675 }, { "epoch": 0.73, "learning_rate": 2.862574696144665e-05, "loss": 0.1441, "step": 15680 }, { "epoch": 0.73, "learning_rate": 2.862527889161063e-05, "loss": 0.1572, "step": 15685 }, { "epoch": 0.73, "learning_rate": 2.862481082177461e-05, "loss": 0.3004, "step": 15690 }, { "epoch": 0.73, "learning_rate": 2.862434275193859e-05, "loss": 0.1968, "step": 15695 }, { "epoch": 0.73, "learning_rate": 2.862387468210257e-05, "loss": 0.3735, "step": 15700 }, { "epoch": 0.73, "learning_rate": 2.862340661226655e-05, "loss": 0.1962, "step": 15705 }, { "epoch": 0.73, "learning_rate": 2.8622938542430532e-05, "loss": 0.0614, "step": 15710 }, { "epoch": 0.73, "learning_rate": 2.862247047259451e-05, "loss": 0.2023, "step": 15715 }, { "epoch": 0.73, "learning_rate": 2.862200240275849e-05, "loss": 0.1172, "step": 15720 }, { "epoch": 0.73, "learning_rate": 2.8621534332922475e-05, "loss": 0.1431, "step": 15725 }, { "epoch": 0.73, "learning_rate": 2.8621066263086455e-05, "loss": 0.1741, "step": 15730 }, { "epoch": 0.73, "learning_rate": 2.8620598193250434e-05, "loss": 0.2025, "step": 15735 }, { "epoch": 0.73, "learning_rate": 2.8620130123414414e-05, "loss": 0.3104, "step": 15740 }, { "epoch": 0.73, "learning_rate": 2.8619662053578397e-05, "loss": 0.2903, "step": 15745 }, { "epoch": 0.73, "learning_rate": 2.8619193983742377e-05, "loss": 0.5286, "step": 15750 }, { "epoch": 0.74, "learning_rate": 2.8618725913906354e-05, "loss": 0.2759, "step": 15755 }, { "epoch": 0.74, "learning_rate": 2.8618257844070333e-05, "loss": 0.0952, "step": 15760 }, { "epoch": 0.74, "learning_rate": 2.8617789774234317e-05, "loss": 0.1035, "step": 15765 }, { "epoch": 0.74, "learning_rate": 2.8617321704398296e-05, "loss": 0.0883, "step": 15770 }, { "epoch": 0.74, "learning_rate": 2.8616853634562276e-05, "loss": 0.1798, "step": 15775 }, { "epoch": 0.74, "learning_rate": 2.8616385564726256e-05, "loss": 0.2773, "step": 15780 }, { "epoch": 0.74, "learning_rate": 2.861591749489024e-05, "loss": 0.1251, "step": 15785 }, { "epoch": 0.74, "learning_rate": 2.861544942505422e-05, "loss": 0.1566, "step": 15790 }, { "epoch": 0.74, "learning_rate": 2.86149813552182e-05, "loss": 0.3412, "step": 15795 }, { "epoch": 0.74, "learning_rate": 2.8614513285382182e-05, "loss": 0.5419, "step": 15800 }, { "epoch": 0.74, "learning_rate": 2.8614045215546162e-05, "loss": 0.1462, "step": 15805 }, { "epoch": 0.74, "learning_rate": 2.8613577145710142e-05, "loss": 0.0858, "step": 15810 }, { "epoch": 0.74, "learning_rate": 2.8613109075874118e-05, "loss": 0.0801, "step": 15815 }, { "epoch": 0.74, "learning_rate": 2.86126410060381e-05, "loss": 0.1397, "step": 15820 }, { "epoch": 0.74, "learning_rate": 2.861217293620208e-05, "loss": 0.0733, "step": 15825 }, { "epoch": 0.74, "learning_rate": 2.861170486636606e-05, "loss": 0.1209, "step": 15830 }, { "epoch": 0.74, "learning_rate": 2.861123679653004e-05, "loss": 0.3169, "step": 15835 }, { "epoch": 0.74, "learning_rate": 2.8610768726694024e-05, "loss": 0.1338, "step": 15840 }, { "epoch": 0.74, "learning_rate": 2.8610300656858004e-05, "loss": 0.4793, "step": 15845 }, { "epoch": 0.74, "learning_rate": 2.8609832587021984e-05, "loss": 0.4244, "step": 15850 }, { "epoch": 0.74, "learning_rate": 2.8609364517185967e-05, "loss": 0.1958, "step": 15855 }, { "epoch": 0.74, "learning_rate": 2.8608896447349947e-05, "loss": 0.1052, "step": 15860 }, { "epoch": 0.74, "learning_rate": 2.8608428377513927e-05, "loss": 0.1194, "step": 15865 }, { "epoch": 0.74, "learning_rate": 2.8607960307677906e-05, "loss": 0.1594, "step": 15870 }, { "epoch": 0.74, "learning_rate": 2.860749223784189e-05, "loss": 0.1061, "step": 15875 }, { "epoch": 0.74, "learning_rate": 2.8607024168005866e-05, "loss": 0.1318, "step": 15880 }, { "epoch": 0.74, "learning_rate": 2.8606556098169846e-05, "loss": 0.2331, "step": 15885 }, { "epoch": 0.74, "learning_rate": 2.8606088028333826e-05, "loss": 0.304, "step": 15890 }, { "epoch": 0.74, "learning_rate": 2.860561995849781e-05, "loss": 0.318, "step": 15895 }, { "epoch": 0.74, "learning_rate": 2.860515188866179e-05, "loss": 0.4271, "step": 15900 }, { "epoch": 0.74, "learning_rate": 2.860468381882577e-05, "loss": 0.2185, "step": 15905 }, { "epoch": 0.74, "learning_rate": 2.860421574898975e-05, "loss": 0.1382, "step": 15910 }, { "epoch": 0.74, "learning_rate": 2.860374767915373e-05, "loss": 0.1087, "step": 15915 }, { "epoch": 0.74, "learning_rate": 2.860327960931771e-05, "loss": 0.2021, "step": 15920 }, { "epoch": 0.74, "learning_rate": 2.860281153948169e-05, "loss": 0.1516, "step": 15925 }, { "epoch": 0.74, "learning_rate": 2.8602343469645674e-05, "loss": 0.1961, "step": 15930 }, { "epoch": 0.74, "learning_rate": 2.8601875399809654e-05, "loss": 0.2074, "step": 15935 }, { "epoch": 0.74, "learning_rate": 2.8601407329973634e-05, "loss": 0.2017, "step": 15940 }, { "epoch": 0.74, "learning_rate": 2.860093926013761e-05, "loss": 0.1713, "step": 15945 }, { "epoch": 0.74, "learning_rate": 2.8600471190301594e-05, "loss": 0.4773, "step": 15950 }, { "epoch": 0.74, "learning_rate": 2.8600003120465573e-05, "loss": 0.229, "step": 15955 }, { "epoch": 0.74, "learning_rate": 2.8599535050629553e-05, "loss": 0.0675, "step": 15960 }, { "epoch": 0.74, "learning_rate": 2.8599066980793536e-05, "loss": 0.1082, "step": 15965 }, { "epoch": 0.75, "learning_rate": 2.8598598910957516e-05, "loss": 0.1686, "step": 15970 }, { "epoch": 0.75, "learning_rate": 2.8598130841121496e-05, "loss": 0.1475, "step": 15975 }, { "epoch": 0.75, "learning_rate": 2.8597662771285476e-05, "loss": 0.1345, "step": 15980 }, { "epoch": 0.75, "learning_rate": 2.859719470144946e-05, "loss": 0.1773, "step": 15985 }, { "epoch": 0.75, "learning_rate": 2.859672663161344e-05, "loss": 0.1989, "step": 15990 }, { "epoch": 0.75, "learning_rate": 2.859625856177742e-05, "loss": 0.2659, "step": 15995 }, { "epoch": 0.75, "learning_rate": 2.85957904919414e-05, "loss": 0.5006, "step": 16000 }, { "epoch": 0.75, "learning_rate": 2.859532242210538e-05, "loss": 0.2812, "step": 16005 }, { "epoch": 0.75, "learning_rate": 2.8594854352269358e-05, "loss": 0.0904, "step": 16010 }, { "epoch": 0.75, "learning_rate": 2.8594386282433338e-05, "loss": 0.1094, "step": 16015 }, { "epoch": 0.75, "learning_rate": 2.8593918212597318e-05, "loss": 0.1808, "step": 16020 }, { "epoch": 0.75, "learning_rate": 2.85934501427613e-05, "loss": 0.1391, "step": 16025 }, { "epoch": 0.75, "learning_rate": 2.859298207292528e-05, "loss": 0.1167, "step": 16030 }, { "epoch": 0.75, "learning_rate": 2.859251400308926e-05, "loss": 0.1936, "step": 16035 }, { "epoch": 0.75, "learning_rate": 2.8592045933253244e-05, "loss": 0.2442, "step": 16040 }, { "epoch": 0.75, "learning_rate": 2.8591577863417224e-05, "loss": 0.2471, "step": 16045 }, { "epoch": 0.75, "learning_rate": 2.8591109793581204e-05, "loss": 0.514, "step": 16050 }, { "epoch": 0.75, "learning_rate": 2.8590641723745183e-05, "loss": 0.2005, "step": 16055 }, { "epoch": 0.75, "learning_rate": 2.8590173653909167e-05, "loss": 0.0913, "step": 16060 }, { "epoch": 0.75, "learning_rate": 2.8589705584073146e-05, "loss": 0.0698, "step": 16065 }, { "epoch": 0.75, "learning_rate": 2.8589237514237123e-05, "loss": 0.1241, "step": 16070 }, { "epoch": 0.75, "learning_rate": 2.8588769444401103e-05, "loss": 0.1931, "step": 16075 }, { "epoch": 0.75, "learning_rate": 2.8588301374565086e-05, "loss": 0.102, "step": 16080 }, { "epoch": 0.75, "learning_rate": 2.8587833304729066e-05, "loss": 0.1887, "step": 16085 }, { "epoch": 0.75, "learning_rate": 2.8587365234893045e-05, "loss": 0.3126, "step": 16090 }, { "epoch": 0.75, "learning_rate": 2.858689716505703e-05, "loss": 0.295, "step": 16095 }, { "epoch": 0.75, "learning_rate": 2.858642909522101e-05, "loss": 0.5638, "step": 16100 }, { "epoch": 0.75, "learning_rate": 2.8585961025384988e-05, "loss": 0.2382, "step": 16105 }, { "epoch": 0.75, "learning_rate": 2.8585492955548968e-05, "loss": 0.0758, "step": 16110 }, { "epoch": 0.75, "learning_rate": 2.858502488571295e-05, "loss": 0.1077, "step": 16115 }, { "epoch": 0.75, "learning_rate": 2.858455681587693e-05, "loss": 0.1388, "step": 16120 }, { "epoch": 0.75, "learning_rate": 2.858408874604091e-05, "loss": 0.1855, "step": 16125 }, { "epoch": 0.75, "learning_rate": 2.8583620676204887e-05, "loss": 0.1516, "step": 16130 }, { "epoch": 0.75, "learning_rate": 2.858315260636887e-05, "loss": 0.1732, "step": 16135 }, { "epoch": 0.75, "learning_rate": 2.858268453653285e-05, "loss": 0.2748, "step": 16140 }, { "epoch": 0.75, "learning_rate": 2.858221646669683e-05, "loss": 0.4037, "step": 16145 }, { "epoch": 0.75, "learning_rate": 2.8581748396860813e-05, "loss": 0.5582, "step": 16150 }, { "epoch": 0.75, "learning_rate": 2.8581280327024793e-05, "loss": 0.1921, "step": 16155 }, { "epoch": 0.75, "learning_rate": 2.8580812257188773e-05, "loss": 0.1022, "step": 16160 }, { "epoch": 0.75, "learning_rate": 2.8580344187352753e-05, "loss": 0.0983, "step": 16165 }, { "epoch": 0.75, "learning_rate": 2.8579876117516736e-05, "loss": 0.1711, "step": 16170 }, { "epoch": 0.75, "learning_rate": 2.8579408047680716e-05, "loss": 0.1264, "step": 16175 }, { "epoch": 0.75, "learning_rate": 2.8578939977844696e-05, "loss": 0.129, "step": 16180 }, { "epoch": 0.76, "learning_rate": 2.8578471908008676e-05, "loss": 0.227, "step": 16185 }, { "epoch": 0.76, "learning_rate": 2.857800383817266e-05, "loss": 0.2669, "step": 16190 }, { "epoch": 0.76, "learning_rate": 2.8577535768336635e-05, "loss": 0.3455, "step": 16195 }, { "epoch": 0.76, "learning_rate": 2.8577067698500615e-05, "loss": 0.4634, "step": 16200 }, { "epoch": 0.76, "learning_rate": 2.8576599628664595e-05, "loss": 0.2328, "step": 16205 }, { "epoch": 0.76, "learning_rate": 2.8576131558828578e-05, "loss": 0.0769, "step": 16210 }, { "epoch": 0.76, "learning_rate": 2.8575663488992558e-05, "loss": 0.074, "step": 16215 }, { "epoch": 0.76, "learning_rate": 2.8575195419156538e-05, "loss": 0.0922, "step": 16220 }, { "epoch": 0.76, "learning_rate": 2.857472734932052e-05, "loss": 0.1451, "step": 16225 }, { "epoch": 0.76, "learning_rate": 2.85742592794845e-05, "loss": 0.2113, "step": 16230 }, { "epoch": 0.76, "learning_rate": 2.857379120964848e-05, "loss": 0.2035, "step": 16235 }, { "epoch": 0.76, "learning_rate": 2.857332313981246e-05, "loss": 0.2921, "step": 16240 }, { "epoch": 0.76, "learning_rate": 2.8572855069976444e-05, "loss": 0.313, "step": 16245 }, { "epoch": 0.76, "learning_rate": 2.8572387000140423e-05, "loss": 0.4806, "step": 16250 }, { "epoch": 0.76, "learning_rate": 2.8571918930304403e-05, "loss": 0.3072, "step": 16255 }, { "epoch": 0.76, "learning_rate": 2.857145086046838e-05, "loss": 0.1016, "step": 16260 }, { "epoch": 0.76, "learning_rate": 2.8570982790632363e-05, "loss": 0.0676, "step": 16265 }, { "epoch": 0.76, "learning_rate": 2.8570514720796343e-05, "loss": 0.1592, "step": 16270 }, { "epoch": 0.76, "learning_rate": 2.8570046650960322e-05, "loss": 0.1448, "step": 16275 }, { "epoch": 0.76, "learning_rate": 2.8569578581124306e-05, "loss": 0.1645, "step": 16280 }, { "epoch": 0.76, "learning_rate": 2.8569110511288285e-05, "loss": 0.1477, "step": 16285 }, { "epoch": 0.76, "learning_rate": 2.8568642441452265e-05, "loss": 0.2008, "step": 16290 }, { "epoch": 0.76, "learning_rate": 2.8568174371616245e-05, "loss": 0.3682, "step": 16295 }, { "epoch": 0.76, "learning_rate": 2.8567706301780228e-05, "loss": 0.5006, "step": 16300 }, { "epoch": 0.76, "learning_rate": 2.8567238231944208e-05, "loss": 0.182, "step": 16305 }, { "epoch": 0.76, "learning_rate": 2.8566770162108188e-05, "loss": 0.0546, "step": 16310 }, { "epoch": 0.76, "learning_rate": 2.8566302092272168e-05, "loss": 0.1098, "step": 16315 }, { "epoch": 0.76, "learning_rate": 2.8565834022436148e-05, "loss": 0.109, "step": 16320 }, { "epoch": 0.76, "learning_rate": 2.8565365952600127e-05, "loss": 0.1537, "step": 16325 }, { "epoch": 0.76, "learning_rate": 2.8564897882764107e-05, "loss": 0.1976, "step": 16330 }, { "epoch": 0.76, "learning_rate": 2.856442981292809e-05, "loss": 0.2869, "step": 16335 }, { "epoch": 0.76, "learning_rate": 2.856396174309207e-05, "loss": 0.2232, "step": 16340 }, { "epoch": 0.76, "learning_rate": 2.856349367325605e-05, "loss": 0.2059, "step": 16345 }, { "epoch": 0.76, "learning_rate": 2.856302560342003e-05, "loss": 0.5648, "step": 16350 }, { "epoch": 0.76, "learning_rate": 2.8562557533584013e-05, "loss": 0.1993, "step": 16355 }, { "epoch": 0.76, "learning_rate": 2.8562089463747993e-05, "loss": 0.0903, "step": 16360 }, { "epoch": 0.76, "learning_rate": 2.8561621393911973e-05, "loss": 0.087, "step": 16365 }, { "epoch": 0.76, "learning_rate": 2.8561153324075953e-05, "loss": 0.1336, "step": 16370 }, { "epoch": 0.76, "learning_rate": 2.8560685254239936e-05, "loss": 0.1008, "step": 16375 }, { "epoch": 0.76, "learning_rate": 2.8560217184403916e-05, "loss": 0.15, "step": 16380 }, { "epoch": 0.76, "learning_rate": 2.8559749114567892e-05, "loss": 0.142, "step": 16385 }, { "epoch": 0.76, "learning_rate": 2.8559281044731872e-05, "loss": 0.2956, "step": 16390 }, { "epoch": 0.77, "learning_rate": 2.8558812974895855e-05, "loss": 0.3568, "step": 16395 }, { "epoch": 0.77, "learning_rate": 2.8558344905059835e-05, "loss": 0.3688, "step": 16400 }, { "epoch": 0.77, "learning_rate": 2.8557876835223815e-05, "loss": 0.1329, "step": 16405 }, { "epoch": 0.77, "learning_rate": 2.8557408765387798e-05, "loss": 0.0728, "step": 16410 }, { "epoch": 0.77, "learning_rate": 2.8556940695551778e-05, "loss": 0.0862, "step": 16415 }, { "epoch": 0.77, "learning_rate": 2.8556472625715757e-05, "loss": 0.1549, "step": 16420 }, { "epoch": 0.77, "learning_rate": 2.8556004555879737e-05, "loss": 0.1165, "step": 16425 }, { "epoch": 0.77, "learning_rate": 2.855553648604372e-05, "loss": 0.1744, "step": 16430 }, { "epoch": 0.77, "learning_rate": 2.85550684162077e-05, "loss": 0.2404, "step": 16435 }, { "epoch": 0.77, "learning_rate": 2.855460034637168e-05, "loss": 0.2093, "step": 16440 }, { "epoch": 0.77, "learning_rate": 2.855413227653566e-05, "loss": 0.2035, "step": 16445 }, { "epoch": 0.77, "learning_rate": 2.855366420669964e-05, "loss": 0.509, "step": 16450 }, { "epoch": 0.77, "learning_rate": 2.855319613686362e-05, "loss": 0.2632, "step": 16455 }, { "epoch": 0.77, "learning_rate": 2.85527280670276e-05, "loss": 0.1014, "step": 16460 }, { "epoch": 0.77, "learning_rate": 2.8552259997191583e-05, "loss": 0.1104, "step": 16465 }, { "epoch": 0.77, "learning_rate": 2.8551791927355562e-05, "loss": 0.1301, "step": 16470 }, { "epoch": 0.77, "learning_rate": 2.8551323857519542e-05, "loss": 0.1182, "step": 16475 }, { "epoch": 0.77, "learning_rate": 2.8550855787683522e-05, "loss": 0.1053, "step": 16480 }, { "epoch": 0.77, "learning_rate": 2.8550387717847505e-05, "loss": 0.169, "step": 16485 }, { "epoch": 0.77, "learning_rate": 2.8549919648011485e-05, "loss": 0.1742, "step": 16490 }, { "epoch": 0.77, "learning_rate": 2.8549451578175465e-05, "loss": 0.2934, "step": 16495 }, { "epoch": 0.77, "learning_rate": 2.8548983508339445e-05, "loss": 0.4648, "step": 16500 }, { "epoch": 0.77, "learning_rate": 2.8548515438503428e-05, "loss": 0.2497, "step": 16505 }, { "epoch": 0.77, "learning_rate": 2.8548047368667404e-05, "loss": 0.0621, "step": 16510 }, { "epoch": 0.77, "learning_rate": 2.8547579298831384e-05, "loss": 0.0941, "step": 16515 }, { "epoch": 0.77, "learning_rate": 2.8547111228995367e-05, "loss": 0.1163, "step": 16520 }, { "epoch": 0.77, "learning_rate": 2.8546643159159347e-05, "loss": 0.1363, "step": 16525 }, { "epoch": 0.77, "learning_rate": 2.8546175089323327e-05, "loss": 0.1571, "step": 16530 }, { "epoch": 0.77, "learning_rate": 2.8545707019487307e-05, "loss": 0.2624, "step": 16535 }, { "epoch": 0.77, "learning_rate": 2.854523894965129e-05, "loss": 0.178, "step": 16540 }, { "epoch": 0.77, "learning_rate": 2.854477087981527e-05, "loss": 0.3534, "step": 16545 }, { "epoch": 0.77, "learning_rate": 2.854430280997925e-05, "loss": 0.5826, "step": 16550 }, { "epoch": 0.77, "learning_rate": 2.854383474014323e-05, "loss": 0.1824, "step": 16555 }, { "epoch": 0.77, "learning_rate": 2.8543366670307213e-05, "loss": 0.0982, "step": 16560 }, { "epoch": 0.77, "learning_rate": 2.8542898600471192e-05, "loss": 0.0746, "step": 16565 }, { "epoch": 0.77, "learning_rate": 2.8542430530635172e-05, "loss": 0.1469, "step": 16570 }, { "epoch": 0.77, "learning_rate": 2.8541962460799152e-05, "loss": 0.1967, "step": 16575 }, { "epoch": 0.77, "learning_rate": 2.8541494390963132e-05, "loss": 0.1434, "step": 16580 }, { "epoch": 0.77, "learning_rate": 2.8541026321127112e-05, "loss": 0.296, "step": 16585 }, { "epoch": 0.77, "learning_rate": 2.854055825129109e-05, "loss": 0.2408, "step": 16590 }, { "epoch": 0.77, "learning_rate": 2.8540090181455075e-05, "loss": 0.1945, "step": 16595 }, { "epoch": 0.77, "learning_rate": 2.8539622111619055e-05, "loss": 0.4325, "step": 16600 }, { "epoch": 0.77, "learning_rate": 2.8539154041783034e-05, "loss": 0.2406, "step": 16605 }, { "epoch": 0.78, "learning_rate": 2.8538685971947014e-05, "loss": 0.0676, "step": 16610 }, { "epoch": 0.78, "learning_rate": 2.8538217902110997e-05, "loss": 0.0867, "step": 16615 }, { "epoch": 0.78, "learning_rate": 2.8537749832274977e-05, "loss": 0.1002, "step": 16620 }, { "epoch": 0.78, "learning_rate": 2.8537281762438957e-05, "loss": 0.1743, "step": 16625 }, { "epoch": 0.78, "learning_rate": 2.8536813692602937e-05, "loss": 0.2309, "step": 16630 }, { "epoch": 0.78, "learning_rate": 2.8536345622766917e-05, "loss": 0.2388, "step": 16635 }, { "epoch": 0.78, "learning_rate": 2.8535877552930897e-05, "loss": 0.1545, "step": 16640 }, { "epoch": 0.78, "learning_rate": 2.8535409483094876e-05, "loss": 0.287, "step": 16645 }, { "epoch": 0.78, "learning_rate": 2.853494141325886e-05, "loss": 0.5251, "step": 16650 }, { "epoch": 0.78, "learning_rate": 2.853447334342284e-05, "loss": 0.2706, "step": 16655 }, { "epoch": 0.78, "learning_rate": 2.853400527358682e-05, "loss": 0.0825, "step": 16660 }, { "epoch": 0.78, "learning_rate": 2.85335372037508e-05, "loss": 0.0998, "step": 16665 }, { "epoch": 0.78, "learning_rate": 2.8533069133914782e-05, "loss": 0.1043, "step": 16670 }, { "epoch": 0.78, "learning_rate": 2.8532601064078762e-05, "loss": 0.1212, "step": 16675 }, { "epoch": 0.78, "learning_rate": 2.8532132994242742e-05, "loss": 0.1515, "step": 16680 }, { "epoch": 0.78, "learning_rate": 2.853166492440672e-05, "loss": 0.1651, "step": 16685 }, { "epoch": 0.78, "learning_rate": 2.8531196854570705e-05, "loss": 0.1988, "step": 16690 }, { "epoch": 0.78, "learning_rate": 2.8530728784734685e-05, "loss": 0.2775, "step": 16695 }, { "epoch": 0.78, "learning_rate": 2.853026071489866e-05, "loss": 0.3739, "step": 16700 }, { "epoch": 0.78, "learning_rate": 2.8529792645062644e-05, "loss": 0.2159, "step": 16705 }, { "epoch": 0.78, "learning_rate": 2.8529324575226624e-05, "loss": 0.0993, "step": 16710 }, { "epoch": 0.78, "learning_rate": 2.8528856505390604e-05, "loss": 0.0733, "step": 16715 }, { "epoch": 0.78, "learning_rate": 2.8528388435554584e-05, "loss": 0.1262, "step": 16720 }, { "epoch": 0.78, "learning_rate": 2.8527920365718567e-05, "loss": 0.1416, "step": 16725 }, { "epoch": 0.78, "learning_rate": 2.8527452295882547e-05, "loss": 0.1235, "step": 16730 }, { "epoch": 0.78, "learning_rate": 2.8526984226046527e-05, "loss": 0.178, "step": 16735 }, { "epoch": 0.78, "learning_rate": 2.8526516156210506e-05, "loss": 0.2262, "step": 16740 }, { "epoch": 0.78, "learning_rate": 2.852604808637449e-05, "loss": 0.389, "step": 16745 }, { "epoch": 0.78, "learning_rate": 2.852558001653847e-05, "loss": 0.4825, "step": 16750 }, { "epoch": 0.78, "learning_rate": 2.852511194670245e-05, "loss": 0.1922, "step": 16755 }, { "epoch": 0.78, "learning_rate": 2.8524643876866432e-05, "loss": 0.0468, "step": 16760 }, { "epoch": 0.78, "learning_rate": 2.852417580703041e-05, "loss": 0.1222, "step": 16765 }, { "epoch": 0.78, "learning_rate": 2.852370773719439e-05, "loss": 0.1168, "step": 16770 }, { "epoch": 0.78, "learning_rate": 2.852323966735837e-05, "loss": 0.1204, "step": 16775 }, { "epoch": 0.78, "learning_rate": 2.8522771597522352e-05, "loss": 0.2345, "step": 16780 }, { "epoch": 0.78, "learning_rate": 2.852230352768633e-05, "loss": 0.1697, "step": 16785 }, { "epoch": 0.78, "learning_rate": 2.852183545785031e-05, "loss": 0.2511, "step": 16790 }, { "epoch": 0.78, "learning_rate": 2.852136738801429e-05, "loss": 0.3124, "step": 16795 }, { "epoch": 0.78, "learning_rate": 2.8520899318178274e-05, "loss": 0.4245, "step": 16800 }, { "epoch": 0.78, "learning_rate": 2.8520431248342254e-05, "loss": 0.172, "step": 16805 }, { "epoch": 0.78, "learning_rate": 2.8519963178506234e-05, "loss": 0.0788, "step": 16810 }, { "epoch": 0.78, "learning_rate": 2.8519495108670214e-05, "loss": 0.083, "step": 16815 }, { "epoch": 0.78, "learning_rate": 2.8519027038834197e-05, "loss": 0.1405, "step": 16820 }, { "epoch": 0.79, "learning_rate": 2.8518558968998174e-05, "loss": 0.1144, "step": 16825 }, { "epoch": 0.79, "learning_rate": 2.8518090899162153e-05, "loss": 0.1346, "step": 16830 }, { "epoch": 0.79, "learning_rate": 2.8517622829326137e-05, "loss": 0.1765, "step": 16835 }, { "epoch": 0.79, "learning_rate": 2.8517154759490116e-05, "loss": 0.3672, "step": 16840 }, { "epoch": 0.79, "learning_rate": 2.8516686689654096e-05, "loss": 0.2694, "step": 16845 }, { "epoch": 0.79, "learning_rate": 2.8516218619818076e-05, "loss": 0.6112, "step": 16850 }, { "epoch": 0.79, "learning_rate": 2.851575054998206e-05, "loss": 0.2137, "step": 16855 }, { "epoch": 0.79, "learning_rate": 2.851528248014604e-05, "loss": 0.0569, "step": 16860 }, { "epoch": 0.79, "learning_rate": 2.851481441031002e-05, "loss": 0.1371, "step": 16865 }, { "epoch": 0.79, "learning_rate": 2.8514346340474e-05, "loss": 0.0837, "step": 16870 }, { "epoch": 0.79, "learning_rate": 2.8513878270637982e-05, "loss": 0.1527, "step": 16875 }, { "epoch": 0.79, "learning_rate": 2.851341020080196e-05, "loss": 0.1685, "step": 16880 }, { "epoch": 0.79, "learning_rate": 2.851294213096594e-05, "loss": 0.2899, "step": 16885 }, { "epoch": 0.79, "learning_rate": 2.851247406112992e-05, "loss": 0.2131, "step": 16890 }, { "epoch": 0.79, "learning_rate": 2.85120059912939e-05, "loss": 0.2218, "step": 16895 }, { "epoch": 0.79, "learning_rate": 2.851153792145788e-05, "loss": 0.6848, "step": 16900 }, { "epoch": 0.79, "learning_rate": 2.851106985162186e-05, "loss": 0.111, "step": 16905 }, { "epoch": 0.79, "learning_rate": 2.8510601781785844e-05, "loss": 0.11, "step": 16910 }, { "epoch": 0.79, "learning_rate": 2.8510133711949824e-05, "loss": 0.0791, "step": 16915 }, { "epoch": 0.79, "learning_rate": 2.8509665642113804e-05, "loss": 0.0977, "step": 16920 }, { "epoch": 0.79, "learning_rate": 2.8509197572277783e-05, "loss": 0.1874, "step": 16925 }, { "epoch": 0.79, "learning_rate": 2.8508729502441767e-05, "loss": 0.1482, "step": 16930 }, { "epoch": 0.79, "learning_rate": 2.8508261432605746e-05, "loss": 0.2308, "step": 16935 }, { "epoch": 0.79, "learning_rate": 2.8507793362769726e-05, "loss": 0.2865, "step": 16940 }, { "epoch": 0.79, "learning_rate": 2.850732529293371e-05, "loss": 0.3884, "step": 16945 }, { "epoch": 0.79, "learning_rate": 2.850685722309769e-05, "loss": 0.5079, "step": 16950 }, { "epoch": 0.79, "learning_rate": 2.8506389153261666e-05, "loss": 0.2105, "step": 16955 }, { "epoch": 0.79, "learning_rate": 2.8505921083425646e-05, "loss": 0.0665, "step": 16960 }, { "epoch": 0.79, "learning_rate": 2.850545301358963e-05, "loss": 0.0669, "step": 16965 }, { "epoch": 0.79, "learning_rate": 2.850498494375361e-05, "loss": 0.0971, "step": 16970 }, { "epoch": 0.79, "learning_rate": 2.850451687391759e-05, "loss": 0.0863, "step": 16975 }, { "epoch": 0.79, "learning_rate": 2.8504048804081568e-05, "loss": 0.2074, "step": 16980 }, { "epoch": 0.79, "learning_rate": 2.850358073424555e-05, "loss": 0.2256, "step": 16985 }, { "epoch": 0.79, "learning_rate": 2.850311266440953e-05, "loss": 0.1474, "step": 16990 }, { "epoch": 0.79, "learning_rate": 2.850264459457351e-05, "loss": 0.3712, "step": 16995 }, { "epoch": 0.79, "learning_rate": 2.850217652473749e-05, "loss": 0.4077, "step": 17000 }, { "epoch": 0.79, "learning_rate": 2.8501708454901474e-05, "loss": 0.2453, "step": 17005 }, { "epoch": 0.79, "learning_rate": 2.8501240385065454e-05, "loss": 0.0503, "step": 17010 }, { "epoch": 0.79, "learning_rate": 2.850077231522943e-05, "loss": 0.0805, "step": 17015 }, { "epoch": 0.79, "learning_rate": 2.8500304245393414e-05, "loss": 0.1867, "step": 17020 }, { "epoch": 0.79, "learning_rate": 2.8499836175557393e-05, "loss": 0.0975, "step": 17025 }, { "epoch": 0.79, "learning_rate": 2.8499368105721373e-05, "loss": 0.1772, "step": 17030 }, { "epoch": 0.79, "learning_rate": 2.8498900035885353e-05, "loss": 0.1767, "step": 17035 }, { "epoch": 0.8, "learning_rate": 2.8498431966049336e-05, "loss": 0.2192, "step": 17040 }, { "epoch": 0.8, "learning_rate": 2.8497963896213316e-05, "loss": 0.3021, "step": 17045 }, { "epoch": 0.8, "learning_rate": 2.8497495826377296e-05, "loss": 0.6028, "step": 17050 }, { "epoch": 0.8, "learning_rate": 2.8497027756541276e-05, "loss": 0.2188, "step": 17055 }, { "epoch": 0.8, "learning_rate": 2.849655968670526e-05, "loss": 0.1277, "step": 17060 }, { "epoch": 0.8, "learning_rate": 2.849609161686924e-05, "loss": 0.0907, "step": 17065 }, { "epoch": 0.8, "learning_rate": 2.849562354703322e-05, "loss": 0.1664, "step": 17070 }, { "epoch": 0.8, "learning_rate": 2.84951554771972e-05, "loss": 0.1506, "step": 17075 }, { "epoch": 0.8, "learning_rate": 2.8494687407361178e-05, "loss": 0.2175, "step": 17080 }, { "epoch": 0.8, "learning_rate": 2.8494219337525158e-05, "loss": 0.1628, "step": 17085 }, { "epoch": 0.8, "learning_rate": 2.8493751267689138e-05, "loss": 0.2535, "step": 17090 }, { "epoch": 0.8, "learning_rate": 2.849328319785312e-05, "loss": 0.3497, "step": 17095 }, { "epoch": 0.8, "learning_rate": 2.84928151280171e-05, "loss": 0.5314, "step": 17100 }, { "epoch": 0.8, "learning_rate": 2.849234705818108e-05, "loss": 0.2614, "step": 17105 }, { "epoch": 0.8, "learning_rate": 2.849187898834506e-05, "loss": 0.0965, "step": 17110 }, { "epoch": 0.8, "learning_rate": 2.8491410918509044e-05, "loss": 0.1758, "step": 17115 }, { "epoch": 0.8, "learning_rate": 2.8490942848673023e-05, "loss": 0.1704, "step": 17120 }, { "epoch": 0.8, "learning_rate": 2.8490474778837003e-05, "loss": 0.1588, "step": 17125 }, { "epoch": 0.8, "learning_rate": 2.8490006709000986e-05, "loss": 0.1558, "step": 17130 }, { "epoch": 0.8, "learning_rate": 2.8489538639164966e-05, "loss": 0.1528, "step": 17135 }, { "epoch": 0.8, "learning_rate": 2.8489070569328946e-05, "loss": 0.2789, "step": 17140 }, { "epoch": 0.8, "learning_rate": 2.8488602499492922e-05, "loss": 0.2574, "step": 17145 }, { "epoch": 0.8, "learning_rate": 2.8488134429656906e-05, "loss": 0.4975, "step": 17150 }, { "epoch": 0.8, "learning_rate": 2.8487666359820886e-05, "loss": 0.1706, "step": 17155 }, { "epoch": 0.8, "learning_rate": 2.8487198289984865e-05, "loss": 0.0641, "step": 17160 }, { "epoch": 0.8, "learning_rate": 2.8486730220148845e-05, "loss": 0.0686, "step": 17165 }, { "epoch": 0.8, "learning_rate": 2.848626215031283e-05, "loss": 0.1254, "step": 17170 }, { "epoch": 0.8, "learning_rate": 2.8485794080476808e-05, "loss": 0.1117, "step": 17175 }, { "epoch": 0.8, "learning_rate": 2.8485326010640788e-05, "loss": 0.1488, "step": 17180 }, { "epoch": 0.8, "learning_rate": 2.8484857940804768e-05, "loss": 0.2334, "step": 17185 }, { "epoch": 0.8, "learning_rate": 2.848438987096875e-05, "loss": 0.2169, "step": 17190 }, { "epoch": 0.8, "learning_rate": 2.848392180113273e-05, "loss": 0.2871, "step": 17195 }, { "epoch": 0.8, "learning_rate": 2.848345373129671e-05, "loss": 0.557, "step": 17200 }, { "epoch": 0.8, "learning_rate": 2.848298566146069e-05, "loss": 0.2358, "step": 17205 }, { "epoch": 0.8, "learning_rate": 2.848251759162467e-05, "loss": 0.0385, "step": 17210 }, { "epoch": 0.8, "learning_rate": 2.848204952178865e-05, "loss": 0.1046, "step": 17215 }, { "epoch": 0.8, "learning_rate": 2.848158145195263e-05, "loss": 0.1113, "step": 17220 }, { "epoch": 0.8, "learning_rate": 2.8481113382116613e-05, "loss": 0.1022, "step": 17225 }, { "epoch": 0.8, "learning_rate": 2.8480645312280593e-05, "loss": 0.1602, "step": 17230 }, { "epoch": 0.8, "learning_rate": 2.8480177242444573e-05, "loss": 0.2675, "step": 17235 }, { "epoch": 0.8, "learning_rate": 2.8479709172608553e-05, "loss": 0.2947, "step": 17240 }, { "epoch": 0.8, "learning_rate": 2.8479241102772536e-05, "loss": 0.3125, "step": 17245 }, { "epoch": 0.8, "learning_rate": 2.8478773032936516e-05, "loss": 0.6016, "step": 17250 }, { "epoch": 0.81, "learning_rate": 2.8478304963100495e-05, "loss": 0.2517, "step": 17255 }, { "epoch": 0.81, "learning_rate": 2.847783689326448e-05, "loss": 0.1022, "step": 17260 }, { "epoch": 0.81, "learning_rate": 2.847736882342846e-05, "loss": 0.074, "step": 17265 }, { "epoch": 0.81, "learning_rate": 2.8476900753592435e-05, "loss": 0.0836, "step": 17270 }, { "epoch": 0.81, "learning_rate": 2.8476432683756415e-05, "loss": 0.2013, "step": 17275 }, { "epoch": 0.81, "learning_rate": 2.8475964613920398e-05, "loss": 0.1588, "step": 17280 }, { "epoch": 0.81, "learning_rate": 2.8475496544084378e-05, "loss": 0.3324, "step": 17285 }, { "epoch": 0.81, "learning_rate": 2.8475028474248358e-05, "loss": 0.2227, "step": 17290 }, { "epoch": 0.81, "learning_rate": 2.8474560404412337e-05, "loss": 0.3555, "step": 17295 }, { "epoch": 0.81, "learning_rate": 2.847409233457632e-05, "loss": 0.4057, "step": 17300 }, { "epoch": 0.81, "learning_rate": 2.84736242647403e-05, "loss": 0.2916, "step": 17305 }, { "epoch": 0.81, "learning_rate": 2.847315619490428e-05, "loss": 0.1287, "step": 17310 }, { "epoch": 0.81, "learning_rate": 2.8472688125068263e-05, "loss": 0.1233, "step": 17315 }, { "epoch": 0.81, "learning_rate": 2.8472220055232243e-05, "loss": 0.0903, "step": 17320 }, { "epoch": 0.81, "learning_rate": 2.8471751985396223e-05, "loss": 0.1557, "step": 17325 }, { "epoch": 0.81, "learning_rate": 2.84712839155602e-05, "loss": 0.1055, "step": 17330 }, { "epoch": 0.81, "learning_rate": 2.8470815845724183e-05, "loss": 0.2877, "step": 17335 }, { "epoch": 0.81, "learning_rate": 2.8470347775888162e-05, "loss": 0.1888, "step": 17340 }, { "epoch": 0.81, "learning_rate": 2.8469879706052142e-05, "loss": 0.446, "step": 17345 }, { "epoch": 0.81, "learning_rate": 2.8469411636216122e-05, "loss": 0.5535, "step": 17350 }, { "epoch": 0.81, "learning_rate": 2.8468943566380105e-05, "loss": 0.1888, "step": 17355 }, { "epoch": 0.81, "learning_rate": 2.8468475496544085e-05, "loss": 0.0642, "step": 17360 }, { "epoch": 0.81, "learning_rate": 2.8468007426708065e-05, "loss": 0.1229, "step": 17365 }, { "epoch": 0.81, "learning_rate": 2.8467539356872048e-05, "loss": 0.0824, "step": 17370 }, { "epoch": 0.81, "learning_rate": 2.8467071287036028e-05, "loss": 0.1548, "step": 17375 }, { "epoch": 0.81, "learning_rate": 2.8466603217200008e-05, "loss": 0.1229, "step": 17380 }, { "epoch": 0.81, "learning_rate": 2.8466135147363988e-05, "loss": 0.2513, "step": 17385 }, { "epoch": 0.81, "learning_rate": 2.846566707752797e-05, "loss": 0.3038, "step": 17390 }, { "epoch": 0.81, "learning_rate": 2.8465199007691947e-05, "loss": 0.289, "step": 17395 }, { "epoch": 0.81, "learning_rate": 2.8464730937855927e-05, "loss": 0.5119, "step": 17400 }, { "epoch": 0.81, "learning_rate": 2.8464262868019907e-05, "loss": 0.1253, "step": 17405 }, { "epoch": 0.81, "learning_rate": 2.846379479818389e-05, "loss": 0.0963, "step": 17410 }, { "epoch": 0.81, "learning_rate": 2.846332672834787e-05, "loss": 0.0999, "step": 17415 }, { "epoch": 0.81, "learning_rate": 2.846285865851185e-05, "loss": 0.099, "step": 17420 }, { "epoch": 0.81, "learning_rate": 2.846239058867583e-05, "loss": 0.158, "step": 17425 }, { "epoch": 0.81, "learning_rate": 2.8461922518839813e-05, "loss": 0.189, "step": 17430 }, { "epoch": 0.81, "learning_rate": 2.8461454449003793e-05, "loss": 0.1467, "step": 17435 }, { "epoch": 0.81, "learning_rate": 2.8460986379167772e-05, "loss": 0.2386, "step": 17440 }, { "epoch": 0.81, "learning_rate": 2.8460518309331756e-05, "loss": 0.2599, "step": 17445 }, { "epoch": 0.81, "learning_rate": 2.8460050239495735e-05, "loss": 0.5057, "step": 17450 }, { "epoch": 0.81, "learning_rate": 2.8459582169659715e-05, "loss": 0.2334, "step": 17455 }, { "epoch": 0.81, "learning_rate": 2.845911409982369e-05, "loss": 0.0933, "step": 17460 }, { "epoch": 0.81, "learning_rate": 2.8458646029987675e-05, "loss": 0.1363, "step": 17465 }, { "epoch": 0.82, "learning_rate": 2.8458177960151655e-05, "loss": 0.0681, "step": 17470 }, { "epoch": 0.82, "learning_rate": 2.8457709890315635e-05, "loss": 0.1143, "step": 17475 }, { "epoch": 0.82, "learning_rate": 2.8457241820479614e-05, "loss": 0.1788, "step": 17480 }, { "epoch": 0.82, "learning_rate": 2.8456773750643598e-05, "loss": 0.0861, "step": 17485 }, { "epoch": 0.82, "learning_rate": 2.8456305680807577e-05, "loss": 0.2347, "step": 17490 }, { "epoch": 0.82, "learning_rate": 2.8455837610971557e-05, "loss": 0.4229, "step": 17495 }, { "epoch": 0.82, "learning_rate": 2.845536954113554e-05, "loss": 0.4567, "step": 17500 }, { "epoch": 0.82, "learning_rate": 2.845490147129952e-05, "loss": 0.2445, "step": 17505 }, { "epoch": 0.82, "learning_rate": 2.84544334014635e-05, "loss": 0.0806, "step": 17510 }, { "epoch": 0.82, "learning_rate": 2.845396533162748e-05, "loss": 0.0552, "step": 17515 }, { "epoch": 0.82, "learning_rate": 2.845349726179146e-05, "loss": 0.1101, "step": 17520 }, { "epoch": 0.82, "learning_rate": 2.845302919195544e-05, "loss": 0.1413, "step": 17525 }, { "epoch": 0.82, "learning_rate": 2.845256112211942e-05, "loss": 0.1538, "step": 17530 }, { "epoch": 0.82, "learning_rate": 2.84520930522834e-05, "loss": 0.2205, "step": 17535 }, { "epoch": 0.82, "learning_rate": 2.8451624982447382e-05, "loss": 0.2771, "step": 17540 }, { "epoch": 0.82, "learning_rate": 2.8451156912611362e-05, "loss": 0.2482, "step": 17545 }, { "epoch": 0.82, "learning_rate": 2.8450688842775342e-05, "loss": 0.5521, "step": 17550 }, { "epoch": 0.82, "learning_rate": 2.8450220772939325e-05, "loss": 0.2345, "step": 17555 }, { "epoch": 0.82, "learning_rate": 2.8449752703103305e-05, "loss": 0.0546, "step": 17560 }, { "epoch": 0.82, "learning_rate": 2.8449284633267285e-05, "loss": 0.115, "step": 17565 }, { "epoch": 0.82, "learning_rate": 2.8448816563431265e-05, "loss": 0.1554, "step": 17570 }, { "epoch": 0.82, "learning_rate": 2.8448348493595248e-05, "loss": 0.1924, "step": 17575 }, { "epoch": 0.82, "learning_rate": 2.8447880423759228e-05, "loss": 0.1633, "step": 17580 }, { "epoch": 0.82, "learning_rate": 2.8447412353923204e-05, "loss": 0.2276, "step": 17585 }, { "epoch": 0.82, "learning_rate": 2.8446944284087184e-05, "loss": 0.2455, "step": 17590 }, { "epoch": 0.82, "learning_rate": 2.8446476214251167e-05, "loss": 0.3305, "step": 17595 }, { "epoch": 0.82, "learning_rate": 2.8446008144415147e-05, "loss": 0.4453, "step": 17600 }, { "epoch": 0.82, "learning_rate": 2.8445540074579127e-05, "loss": 0.1797, "step": 17605 }, { "epoch": 0.82, "learning_rate": 2.8445072004743107e-05, "loss": 0.0646, "step": 17610 }, { "epoch": 0.82, "learning_rate": 2.844460393490709e-05, "loss": 0.0597, "step": 17615 }, { "epoch": 0.82, "learning_rate": 2.844413586507107e-05, "loss": 0.1145, "step": 17620 }, { "epoch": 0.82, "learning_rate": 2.844366779523505e-05, "loss": 0.1437, "step": 17625 }, { "epoch": 0.82, "learning_rate": 2.8443199725399033e-05, "loss": 0.1537, "step": 17630 }, { "epoch": 0.82, "learning_rate": 2.8442731655563012e-05, "loss": 0.1481, "step": 17635 }, { "epoch": 0.82, "learning_rate": 2.8442263585726992e-05, "loss": 0.3613, "step": 17640 }, { "epoch": 0.82, "learning_rate": 2.8441795515890972e-05, "loss": 0.2935, "step": 17645 }, { "epoch": 0.82, "learning_rate": 2.8441327446054952e-05, "loss": 0.4195, "step": 17650 }, { "epoch": 0.82, "learning_rate": 2.844085937621893e-05, "loss": 0.2121, "step": 17655 }, { "epoch": 0.82, "learning_rate": 2.844039130638291e-05, "loss": 0.1253, "step": 17660 }, { "epoch": 0.82, "learning_rate": 2.843992323654689e-05, "loss": 0.0747, "step": 17665 }, { "epoch": 0.82, "learning_rate": 2.8439455166710875e-05, "loss": 0.095, "step": 17670 }, { "epoch": 0.82, "learning_rate": 2.8438987096874854e-05, "loss": 0.193, "step": 17675 }, { "epoch": 0.82, "learning_rate": 2.8438519027038834e-05, "loss": 0.2413, "step": 17680 }, { "epoch": 0.83, "learning_rate": 2.8438050957202817e-05, "loss": 0.2313, "step": 17685 }, { "epoch": 0.83, "learning_rate": 2.8437582887366797e-05, "loss": 0.2927, "step": 17690 }, { "epoch": 0.83, "learning_rate": 2.8437114817530777e-05, "loss": 0.2861, "step": 17695 }, { "epoch": 0.83, "learning_rate": 2.8436646747694757e-05, "loss": 0.3195, "step": 17700 }, { "epoch": 0.83, "learning_rate": 2.843617867785874e-05, "loss": 0.1636, "step": 17705 }, { "epoch": 0.83, "learning_rate": 2.8435710608022716e-05, "loss": 0.0918, "step": 17710 }, { "epoch": 0.83, "learning_rate": 2.8435242538186696e-05, "loss": 0.0742, "step": 17715 }, { "epoch": 0.83, "learning_rate": 2.8434774468350676e-05, "loss": 0.1572, "step": 17720 }, { "epoch": 0.83, "learning_rate": 2.843430639851466e-05, "loss": 0.1237, "step": 17725 }, { "epoch": 0.83, "learning_rate": 2.843383832867864e-05, "loss": 0.1644, "step": 17730 }, { "epoch": 0.83, "learning_rate": 2.843337025884262e-05, "loss": 0.1167, "step": 17735 }, { "epoch": 0.83, "learning_rate": 2.8432902189006602e-05, "loss": 0.157, "step": 17740 }, { "epoch": 0.83, "learning_rate": 2.8432434119170582e-05, "loss": 0.3453, "step": 17745 }, { "epoch": 0.83, "learning_rate": 2.8431966049334562e-05, "loss": 0.46, "step": 17750 }, { "epoch": 0.83, "learning_rate": 2.843149797949854e-05, "loss": 0.2216, "step": 17755 }, { "epoch": 0.83, "learning_rate": 2.8431029909662525e-05, "loss": 0.0829, "step": 17760 }, { "epoch": 0.83, "learning_rate": 2.8430561839826505e-05, "loss": 0.0597, "step": 17765 }, { "epoch": 0.83, "learning_rate": 2.8430093769990484e-05, "loss": 0.1267, "step": 17770 }, { "epoch": 0.83, "learning_rate": 2.842962570015446e-05, "loss": 0.2343, "step": 17775 }, { "epoch": 0.83, "learning_rate": 2.8429157630318444e-05, "loss": 0.1144, "step": 17780 }, { "epoch": 0.83, "learning_rate": 2.8428689560482424e-05, "loss": 0.1421, "step": 17785 }, { "epoch": 0.83, "learning_rate": 2.8428221490646404e-05, "loss": 0.1425, "step": 17790 }, { "epoch": 0.83, "learning_rate": 2.8427753420810383e-05, "loss": 0.2553, "step": 17795 }, { "epoch": 0.83, "learning_rate": 2.8427285350974367e-05, "loss": 0.4929, "step": 17800 }, { "epoch": 0.83, "learning_rate": 2.8426817281138347e-05, "loss": 0.1698, "step": 17805 }, { "epoch": 0.83, "learning_rate": 2.8426349211302326e-05, "loss": 0.0677, "step": 17810 }, { "epoch": 0.83, "learning_rate": 2.842588114146631e-05, "loss": 0.0714, "step": 17815 }, { "epoch": 0.83, "learning_rate": 2.842541307163029e-05, "loss": 0.1241, "step": 17820 }, { "epoch": 0.83, "learning_rate": 2.842494500179427e-05, "loss": 0.0877, "step": 17825 }, { "epoch": 0.83, "learning_rate": 2.842447693195825e-05, "loss": 0.2265, "step": 17830 }, { "epoch": 0.83, "learning_rate": 2.842400886212223e-05, "loss": 0.1535, "step": 17835 }, { "epoch": 0.83, "learning_rate": 2.842354079228621e-05, "loss": 0.3066, "step": 17840 }, { "epoch": 0.83, "learning_rate": 2.842307272245019e-05, "loss": 0.247, "step": 17845 }, { "epoch": 0.83, "learning_rate": 2.8422604652614168e-05, "loss": 0.7187, "step": 17850 }, { "epoch": 0.83, "learning_rate": 2.842213658277815e-05, "loss": 0.1971, "step": 17855 }, { "epoch": 0.83, "learning_rate": 2.842166851294213e-05, "loss": 0.1076, "step": 17860 }, { "epoch": 0.83, "learning_rate": 2.842120044310611e-05, "loss": 0.1249, "step": 17865 }, { "epoch": 0.83, "learning_rate": 2.8420732373270094e-05, "loss": 0.1041, "step": 17870 }, { "epoch": 0.83, "learning_rate": 2.8420264303434074e-05, "loss": 0.1278, "step": 17875 }, { "epoch": 0.83, "learning_rate": 2.8419796233598054e-05, "loss": 0.1483, "step": 17880 }, { "epoch": 0.83, "learning_rate": 2.8419328163762034e-05, "loss": 0.2494, "step": 17885 }, { "epoch": 0.83, "learning_rate": 2.8418860093926017e-05, "loss": 0.1923, "step": 17890 }, { "epoch": 0.84, "learning_rate": 2.8418392024089997e-05, "loss": 0.3253, "step": 17895 }, { "epoch": 0.84, "learning_rate": 2.8417923954253973e-05, "loss": 0.4543, "step": 17900 }, { "epoch": 0.84, "learning_rate": 2.8417455884417953e-05, "loss": 0.158, "step": 17905 }, { "epoch": 0.84, "learning_rate": 2.8416987814581936e-05, "loss": 0.1329, "step": 17910 }, { "epoch": 0.84, "learning_rate": 2.8416519744745916e-05, "loss": 0.1538, "step": 17915 }, { "epoch": 0.84, "learning_rate": 2.8416051674909896e-05, "loss": 0.1384, "step": 17920 }, { "epoch": 0.84, "learning_rate": 2.841558360507388e-05, "loss": 0.0706, "step": 17925 }, { "epoch": 0.84, "learning_rate": 2.841511553523786e-05, "loss": 0.1279, "step": 17930 }, { "epoch": 0.84, "learning_rate": 2.841464746540184e-05, "loss": 0.2132, "step": 17935 }, { "epoch": 0.84, "learning_rate": 2.841417939556582e-05, "loss": 0.2513, "step": 17940 }, { "epoch": 0.84, "learning_rate": 2.8413711325729802e-05, "loss": 0.2697, "step": 17945 }, { "epoch": 0.84, "learning_rate": 2.841324325589378e-05, "loss": 0.3846, "step": 17950 }, { "epoch": 0.84, "learning_rate": 2.841277518605776e-05, "loss": 0.204, "step": 17955 }, { "epoch": 0.84, "learning_rate": 2.841230711622174e-05, "loss": 0.0675, "step": 17960 }, { "epoch": 0.84, "learning_rate": 2.841183904638572e-05, "loss": 0.1023, "step": 17965 }, { "epoch": 0.84, "learning_rate": 2.84113709765497e-05, "loss": 0.1322, "step": 17970 }, { "epoch": 0.84, "learning_rate": 2.841090290671368e-05, "loss": 0.0832, "step": 17975 }, { "epoch": 0.84, "learning_rate": 2.8410434836877664e-05, "loss": 0.2682, "step": 17980 }, { "epoch": 0.84, "learning_rate": 2.8409966767041644e-05, "loss": 0.252, "step": 17985 }, { "epoch": 0.84, "learning_rate": 2.8409498697205623e-05, "loss": 0.3592, "step": 17990 }, { "epoch": 0.84, "learning_rate": 2.8409030627369603e-05, "loss": 0.3841, "step": 17995 }, { "epoch": 0.84, "learning_rate": 2.8408562557533587e-05, "loss": 0.4768, "step": 18000 }, { "epoch": 0.84, "learning_rate": 2.8408094487697566e-05, "loss": 0.1898, "step": 18005 }, { "epoch": 0.84, "learning_rate": 2.8407626417861546e-05, "loss": 0.0701, "step": 18010 }, { "epoch": 0.84, "learning_rate": 2.8407158348025526e-05, "loss": 0.0812, "step": 18015 }, { "epoch": 0.84, "learning_rate": 2.840669027818951e-05, "loss": 0.0757, "step": 18020 }, { "epoch": 0.84, "learning_rate": 2.8406222208353486e-05, "loss": 0.125, "step": 18025 }, { "epoch": 0.84, "learning_rate": 2.8405754138517465e-05, "loss": 0.1982, "step": 18030 }, { "epoch": 0.84, "learning_rate": 2.8405286068681445e-05, "loss": 0.1307, "step": 18035 }, { "epoch": 0.84, "learning_rate": 2.840481799884543e-05, "loss": 0.3637, "step": 18040 }, { "epoch": 0.84, "learning_rate": 2.8404349929009408e-05, "loss": 0.2655, "step": 18045 }, { "epoch": 0.84, "learning_rate": 2.8403881859173388e-05, "loss": 0.3529, "step": 18050 }, { "epoch": 0.84, "learning_rate": 2.840341378933737e-05, "loss": 0.2616, "step": 18055 }, { "epoch": 0.84, "learning_rate": 2.840294571950135e-05, "loss": 0.0918, "step": 18060 }, { "epoch": 0.84, "learning_rate": 2.840247764966533e-05, "loss": 0.1122, "step": 18065 }, { "epoch": 0.84, "learning_rate": 2.840200957982931e-05, "loss": 0.1235, "step": 18070 }, { "epoch": 0.84, "learning_rate": 2.8401541509993294e-05, "loss": 0.1412, "step": 18075 }, { "epoch": 0.84, "learning_rate": 2.8401073440157274e-05, "loss": 0.1196, "step": 18080 }, { "epoch": 0.84, "learning_rate": 2.8400605370321254e-05, "loss": 0.231, "step": 18085 }, { "epoch": 0.84, "learning_rate": 2.840013730048523e-05, "loss": 0.216, "step": 18090 }, { "epoch": 0.84, "learning_rate": 2.8399669230649213e-05, "loss": 0.3107, "step": 18095 }, { "epoch": 0.84, "learning_rate": 2.8399201160813193e-05, "loss": 0.3715, "step": 18100 }, { "epoch": 0.84, "learning_rate": 2.8398733090977173e-05, "loss": 0.1537, "step": 18105 }, { "epoch": 0.85, "learning_rate": 2.8398265021141156e-05, "loss": 0.09, "step": 18110 }, { "epoch": 0.85, "learning_rate": 2.8397796951305136e-05, "loss": 0.0968, "step": 18115 }, { "epoch": 0.85, "learning_rate": 2.8397328881469116e-05, "loss": 0.1648, "step": 18120 }, { "epoch": 0.85, "learning_rate": 2.8396860811633096e-05, "loss": 0.1818, "step": 18125 }, { "epoch": 0.85, "learning_rate": 2.839639274179708e-05, "loss": 0.2729, "step": 18130 }, { "epoch": 0.85, "learning_rate": 2.839592467196106e-05, "loss": 0.2332, "step": 18135 }, { "epoch": 0.85, "learning_rate": 2.839545660212504e-05, "loss": 0.2418, "step": 18140 }, { "epoch": 0.85, "learning_rate": 2.8394988532289018e-05, "loss": 0.2807, "step": 18145 }, { "epoch": 0.85, "learning_rate": 2.8394520462453e-05, "loss": 0.3192, "step": 18150 }, { "epoch": 0.85, "learning_rate": 2.8394052392616978e-05, "loss": 0.1612, "step": 18155 }, { "epoch": 0.85, "learning_rate": 2.8393584322780958e-05, "loss": 0.0834, "step": 18160 }, { "epoch": 0.85, "learning_rate": 2.839311625294494e-05, "loss": 0.1163, "step": 18165 }, { "epoch": 0.85, "learning_rate": 2.839264818310892e-05, "loss": 0.1613, "step": 18170 }, { "epoch": 0.85, "learning_rate": 2.83921801132729e-05, "loss": 0.0896, "step": 18175 }, { "epoch": 0.85, "learning_rate": 2.839171204343688e-05, "loss": 0.136, "step": 18180 }, { "epoch": 0.85, "learning_rate": 2.8391243973600863e-05, "loss": 0.2184, "step": 18185 }, { "epoch": 0.85, "learning_rate": 2.8390775903764843e-05, "loss": 0.1982, "step": 18190 }, { "epoch": 0.85, "learning_rate": 2.8390307833928823e-05, "loss": 0.3135, "step": 18195 }, { "epoch": 0.85, "learning_rate": 2.8389839764092803e-05, "loss": 0.4263, "step": 18200 }, { "epoch": 0.85, "learning_rate": 2.8389371694256786e-05, "loss": 0.192, "step": 18205 }, { "epoch": 0.85, "learning_rate": 2.8388903624420766e-05, "loss": 0.0525, "step": 18210 }, { "epoch": 0.85, "learning_rate": 2.8388435554584742e-05, "loss": 0.076, "step": 18215 }, { "epoch": 0.85, "learning_rate": 2.8387967484748722e-05, "loss": 0.1511, "step": 18220 }, { "epoch": 0.85, "learning_rate": 2.8387499414912705e-05, "loss": 0.1484, "step": 18225 }, { "epoch": 0.85, "learning_rate": 2.8387031345076685e-05, "loss": 0.2002, "step": 18230 }, { "epoch": 0.85, "learning_rate": 2.8386563275240665e-05, "loss": 0.1409, "step": 18235 }, { "epoch": 0.85, "learning_rate": 2.8386095205404648e-05, "loss": 0.1672, "step": 18240 }, { "epoch": 0.85, "learning_rate": 2.8385627135568628e-05, "loss": 0.3112, "step": 18245 }, { "epoch": 0.85, "learning_rate": 2.8385159065732608e-05, "loss": 0.7144, "step": 18250 }, { "epoch": 0.85, "learning_rate": 2.8384690995896588e-05, "loss": 0.2201, "step": 18255 }, { "epoch": 0.85, "learning_rate": 2.838422292606057e-05, "loss": 0.1169, "step": 18260 }, { "epoch": 0.85, "learning_rate": 2.838375485622455e-05, "loss": 0.0843, "step": 18265 }, { "epoch": 0.85, "learning_rate": 2.838328678638853e-05, "loss": 0.1896, "step": 18270 }, { "epoch": 0.85, "learning_rate": 2.838281871655251e-05, "loss": 0.2242, "step": 18275 }, { "epoch": 0.85, "learning_rate": 2.838235064671649e-05, "loss": 0.0801, "step": 18280 }, { "epoch": 0.85, "learning_rate": 2.838188257688047e-05, "loss": 0.0778, "step": 18285 }, { "epoch": 0.85, "learning_rate": 2.838141450704445e-05, "loss": 0.1371, "step": 18290 }, { "epoch": 0.85, "learning_rate": 2.8380946437208433e-05, "loss": 0.2982, "step": 18295 }, { "epoch": 0.85, "learning_rate": 2.8380478367372413e-05, "loss": 0.549, "step": 18300 }, { "epoch": 0.85, "learning_rate": 2.8380010297536393e-05, "loss": 0.1805, "step": 18305 }, { "epoch": 0.85, "learning_rate": 2.8379542227700372e-05, "loss": 0.0731, "step": 18310 }, { "epoch": 0.85, "learning_rate": 2.8379074157864356e-05, "loss": 0.0572, "step": 18315 }, { "epoch": 0.85, "learning_rate": 2.8378606088028336e-05, "loss": 0.2219, "step": 18320 }, { "epoch": 0.86, "learning_rate": 2.8378138018192315e-05, "loss": 0.1155, "step": 18325 }, { "epoch": 0.86, "learning_rate": 2.8377669948356295e-05, "loss": 0.2212, "step": 18330 }, { "epoch": 0.86, "learning_rate": 2.837720187852028e-05, "loss": 0.1358, "step": 18335 }, { "epoch": 0.86, "learning_rate": 2.8376733808684258e-05, "loss": 0.3169, "step": 18340 }, { "epoch": 0.86, "learning_rate": 2.8376265738848235e-05, "loss": 0.2658, "step": 18345 }, { "epoch": 0.86, "learning_rate": 2.8375797669012218e-05, "loss": 0.6011, "step": 18350 }, { "epoch": 0.86, "learning_rate": 2.8375329599176198e-05, "loss": 0.2064, "step": 18355 }, { "epoch": 0.86, "learning_rate": 2.8374861529340177e-05, "loss": 0.0559, "step": 18360 }, { "epoch": 0.86, "learning_rate": 2.8374393459504157e-05, "loss": 0.1031, "step": 18365 }, { "epoch": 0.86, "learning_rate": 2.837392538966814e-05, "loss": 0.1769, "step": 18370 }, { "epoch": 0.86, "learning_rate": 2.837345731983212e-05, "loss": 0.0809, "step": 18375 }, { "epoch": 0.86, "learning_rate": 2.83729892499961e-05, "loss": 0.2196, "step": 18380 }, { "epoch": 0.86, "learning_rate": 2.837252118016008e-05, "loss": 0.1496, "step": 18385 }, { "epoch": 0.86, "learning_rate": 2.8372053110324063e-05, "loss": 0.3154, "step": 18390 }, { "epoch": 0.86, "learning_rate": 2.8371585040488043e-05, "loss": 0.2883, "step": 18395 }, { "epoch": 0.86, "learning_rate": 2.8371116970652023e-05, "loss": 0.6507, "step": 18400 }, { "epoch": 0.86, "learning_rate": 2.8370648900816e-05, "loss": 0.2271, "step": 18405 }, { "epoch": 0.86, "learning_rate": 2.8370180830979982e-05, "loss": 0.0465, "step": 18410 }, { "epoch": 0.86, "learning_rate": 2.8369712761143962e-05, "loss": 0.09, "step": 18415 }, { "epoch": 0.86, "learning_rate": 2.8369244691307942e-05, "loss": 0.1439, "step": 18420 }, { "epoch": 0.86, "learning_rate": 2.8368776621471925e-05, "loss": 0.1548, "step": 18425 }, { "epoch": 0.86, "learning_rate": 2.8368308551635905e-05, "loss": 0.1617, "step": 18430 }, { "epoch": 0.86, "learning_rate": 2.8367840481799885e-05, "loss": 0.1673, "step": 18435 }, { "epoch": 0.86, "learning_rate": 2.8367372411963865e-05, "loss": 0.1674, "step": 18440 }, { "epoch": 0.86, "learning_rate": 2.8366904342127848e-05, "loss": 0.2876, "step": 18445 }, { "epoch": 0.86, "learning_rate": 2.8366436272291828e-05, "loss": 0.3987, "step": 18450 }, { "epoch": 0.86, "learning_rate": 2.8365968202455808e-05, "loss": 0.2224, "step": 18455 }, { "epoch": 0.86, "learning_rate": 2.8365500132619787e-05, "loss": 0.0738, "step": 18460 }, { "epoch": 0.86, "learning_rate": 2.836503206278377e-05, "loss": 0.1386, "step": 18465 }, { "epoch": 0.86, "learning_rate": 2.8364563992947747e-05, "loss": 0.0821, "step": 18470 }, { "epoch": 0.86, "learning_rate": 2.8364095923111727e-05, "loss": 0.2388, "step": 18475 }, { "epoch": 0.86, "learning_rate": 2.836362785327571e-05, "loss": 0.1524, "step": 18480 }, { "epoch": 0.86, "learning_rate": 2.836315978343969e-05, "loss": 0.1129, "step": 18485 }, { "epoch": 0.86, "learning_rate": 2.836269171360367e-05, "loss": 0.2458, "step": 18490 }, { "epoch": 0.86, "learning_rate": 2.836222364376765e-05, "loss": 0.2152, "step": 18495 }, { "epoch": 0.86, "learning_rate": 2.8361755573931633e-05, "loss": 0.3406, "step": 18500 }, { "epoch": 0.86, "learning_rate": 2.8361287504095612e-05, "loss": 0.2004, "step": 18505 }, { "epoch": 0.86, "learning_rate": 2.8360819434259592e-05, "loss": 0.0556, "step": 18510 }, { "epoch": 0.86, "learning_rate": 2.8360351364423572e-05, "loss": 0.0763, "step": 18515 }, { "epoch": 0.86, "learning_rate": 2.8359883294587555e-05, "loss": 0.0865, "step": 18520 }, { "epoch": 0.86, "learning_rate": 2.8359415224751535e-05, "loss": 0.1009, "step": 18525 }, { "epoch": 0.86, "learning_rate": 2.8358947154915515e-05, "loss": 0.1363, "step": 18530 }, { "epoch": 0.86, "learning_rate": 2.8358479085079495e-05, "loss": 0.1338, "step": 18535 }, { "epoch": 0.87, "learning_rate": 2.8358011015243475e-05, "loss": 0.2195, "step": 18540 }, { "epoch": 0.87, "learning_rate": 2.8357542945407454e-05, "loss": 0.3171, "step": 18545 }, { "epoch": 0.87, "learning_rate": 2.8357074875571434e-05, "loss": 0.5755, "step": 18550 }, { "epoch": 0.87, "learning_rate": 2.8356606805735417e-05, "loss": 0.2008, "step": 18555 }, { "epoch": 0.87, "learning_rate": 2.8356138735899397e-05, "loss": 0.0802, "step": 18560 }, { "epoch": 0.87, "learning_rate": 2.8355670666063377e-05, "loss": 0.0548, "step": 18565 }, { "epoch": 0.87, "learning_rate": 2.8355202596227357e-05, "loss": 0.1332, "step": 18570 }, { "epoch": 0.87, "learning_rate": 2.835473452639134e-05, "loss": 0.083, "step": 18575 }, { "epoch": 0.87, "learning_rate": 2.835426645655532e-05, "loss": 0.1904, "step": 18580 }, { "epoch": 0.87, "learning_rate": 2.83537983867193e-05, "loss": 0.194, "step": 18585 }, { "epoch": 0.87, "learning_rate": 2.8353330316883283e-05, "loss": 0.1367, "step": 18590 }, { "epoch": 0.87, "learning_rate": 2.835286224704726e-05, "loss": 0.28, "step": 18595 }, { "epoch": 0.87, "learning_rate": 2.835239417721124e-05, "loss": 0.451, "step": 18600 }, { "epoch": 0.87, "learning_rate": 2.835192610737522e-05, "loss": 0.1892, "step": 18605 }, { "epoch": 0.87, "learning_rate": 2.8351458037539202e-05, "loss": 0.0842, "step": 18610 }, { "epoch": 0.87, "learning_rate": 2.8350989967703182e-05, "loss": 0.1046, "step": 18615 }, { "epoch": 0.87, "learning_rate": 2.8350521897867162e-05, "loss": 0.1042, "step": 18620 }, { "epoch": 0.87, "learning_rate": 2.835005382803114e-05, "loss": 0.107, "step": 18625 }, { "epoch": 0.87, "learning_rate": 2.8349585758195125e-05, "loss": 0.1507, "step": 18630 }, { "epoch": 0.87, "learning_rate": 2.8349117688359105e-05, "loss": 0.209, "step": 18635 }, { "epoch": 0.87, "learning_rate": 2.8348649618523084e-05, "loss": 0.2198, "step": 18640 }, { "epoch": 0.87, "learning_rate": 2.8348181548687064e-05, "loss": 0.1924, "step": 18645 }, { "epoch": 0.87, "learning_rate": 2.8347713478851048e-05, "loss": 0.4928, "step": 18650 }, { "epoch": 0.87, "learning_rate": 2.8347245409015027e-05, "loss": 0.1856, "step": 18655 }, { "epoch": 0.87, "learning_rate": 2.8346777339179004e-05, "loss": 0.0468, "step": 18660 }, { "epoch": 0.87, "learning_rate": 2.8346309269342987e-05, "loss": 0.0885, "step": 18665 }, { "epoch": 0.87, "learning_rate": 2.8345841199506967e-05, "loss": 0.1245, "step": 18670 }, { "epoch": 0.87, "learning_rate": 2.8345373129670947e-05, "loss": 0.1097, "step": 18675 }, { "epoch": 0.87, "learning_rate": 2.8344905059834926e-05, "loss": 0.1906, "step": 18680 }, { "epoch": 0.87, "learning_rate": 2.834443698999891e-05, "loss": 0.2069, "step": 18685 }, { "epoch": 0.87, "learning_rate": 2.834396892016289e-05, "loss": 0.2415, "step": 18690 }, { "epoch": 0.87, "learning_rate": 2.834350085032687e-05, "loss": 0.5044, "step": 18695 }, { "epoch": 0.87, "learning_rate": 2.834303278049085e-05, "loss": 0.3405, "step": 18700 }, { "epoch": 0.87, "learning_rate": 2.8342564710654832e-05, "loss": 0.2299, "step": 18705 }, { "epoch": 0.87, "learning_rate": 2.8342096640818812e-05, "loss": 0.0958, "step": 18710 }, { "epoch": 0.87, "learning_rate": 2.8341628570982792e-05, "loss": 0.0939, "step": 18715 }, { "epoch": 0.87, "learning_rate": 2.8341160501146772e-05, "loss": 0.1363, "step": 18720 }, { "epoch": 0.87, "learning_rate": 2.834069243131075e-05, "loss": 0.1064, "step": 18725 }, { "epoch": 0.87, "learning_rate": 2.834022436147473e-05, "loss": 0.2211, "step": 18730 }, { "epoch": 0.87, "learning_rate": 2.833975629163871e-05, "loss": 0.2835, "step": 18735 }, { "epoch": 0.87, "learning_rate": 2.8339288221802694e-05, "loss": 0.1807, "step": 18740 }, { "epoch": 0.87, "learning_rate": 2.8338820151966674e-05, "loss": 0.4162, "step": 18745 }, { "epoch": 0.87, "learning_rate": 2.8338352082130654e-05, "loss": 0.5333, "step": 18750 }, { "epoch": 0.88, "learning_rate": 2.8337884012294634e-05, "loss": 0.2408, "step": 18755 }, { "epoch": 0.88, "learning_rate": 2.8337415942458617e-05, "loss": 0.0875, "step": 18760 }, { "epoch": 0.88, "learning_rate": 2.8336947872622597e-05, "loss": 0.0799, "step": 18765 }, { "epoch": 0.88, "learning_rate": 2.8336479802786577e-05, "loss": 0.1036, "step": 18770 }, { "epoch": 0.88, "learning_rate": 2.833601173295056e-05, "loss": 0.1331, "step": 18775 }, { "epoch": 0.88, "learning_rate": 2.833554366311454e-05, "loss": 0.1381, "step": 18780 }, { "epoch": 0.88, "learning_rate": 2.8335075593278516e-05, "loss": 0.1336, "step": 18785 }, { "epoch": 0.88, "learning_rate": 2.8334607523442496e-05, "loss": 0.2163, "step": 18790 }, { "epoch": 0.88, "learning_rate": 2.833413945360648e-05, "loss": 0.2485, "step": 18795 }, { "epoch": 0.88, "learning_rate": 2.833367138377046e-05, "loss": 0.4374, "step": 18800 }, { "epoch": 0.88, "learning_rate": 2.833320331393444e-05, "loss": 0.2362, "step": 18805 }, { "epoch": 0.88, "learning_rate": 2.833273524409842e-05, "loss": 0.0922, "step": 18810 }, { "epoch": 0.88, "learning_rate": 2.8332267174262402e-05, "loss": 0.0773, "step": 18815 }, { "epoch": 0.88, "learning_rate": 2.833179910442638e-05, "loss": 0.1277, "step": 18820 }, { "epoch": 0.88, "learning_rate": 2.833133103459036e-05, "loss": 0.1718, "step": 18825 }, { "epoch": 0.88, "learning_rate": 2.833086296475434e-05, "loss": 0.2076, "step": 18830 }, { "epoch": 0.88, "learning_rate": 2.8330394894918324e-05, "loss": 0.2233, "step": 18835 }, { "epoch": 0.88, "learning_rate": 2.8329926825082304e-05, "loss": 0.2237, "step": 18840 }, { "epoch": 0.88, "learning_rate": 2.8329458755246284e-05, "loss": 0.2385, "step": 18845 }, { "epoch": 0.88, "learning_rate": 2.8328990685410264e-05, "loss": 0.4141, "step": 18850 }, { "epoch": 0.88, "learning_rate": 2.8328522615574244e-05, "loss": 0.1521, "step": 18855 }, { "epoch": 0.88, "learning_rate": 2.8328054545738224e-05, "loss": 0.0794, "step": 18860 }, { "epoch": 0.88, "learning_rate": 2.8327586475902203e-05, "loss": 0.2141, "step": 18865 }, { "epoch": 0.88, "learning_rate": 2.8327118406066187e-05, "loss": 0.168, "step": 18870 }, { "epoch": 0.88, "learning_rate": 2.8326650336230166e-05, "loss": 0.1151, "step": 18875 }, { "epoch": 0.88, "learning_rate": 2.8326182266394146e-05, "loss": 0.3286, "step": 18880 }, { "epoch": 0.88, "learning_rate": 2.8325714196558126e-05, "loss": 0.173, "step": 18885 }, { "epoch": 0.88, "learning_rate": 2.832524612672211e-05, "loss": 0.1535, "step": 18890 }, { "epoch": 0.88, "learning_rate": 2.832477805688609e-05, "loss": 0.2278, "step": 18895 }, { "epoch": 0.88, "learning_rate": 2.832430998705007e-05, "loss": 0.3436, "step": 18900 }, { "epoch": 0.88, "learning_rate": 2.8323841917214052e-05, "loss": 0.187, "step": 18905 }, { "epoch": 0.88, "learning_rate": 2.832337384737803e-05, "loss": 0.0451, "step": 18910 }, { "epoch": 0.88, "learning_rate": 2.832290577754201e-05, "loss": 0.1106, "step": 18915 }, { "epoch": 0.88, "learning_rate": 2.8322437707705988e-05, "loss": 0.1392, "step": 18920 }, { "epoch": 0.88, "learning_rate": 2.832196963786997e-05, "loss": 0.1006, "step": 18925 }, { "epoch": 0.88, "learning_rate": 2.832150156803395e-05, "loss": 0.1918, "step": 18930 }, { "epoch": 0.88, "learning_rate": 2.832103349819793e-05, "loss": 0.1577, "step": 18935 }, { "epoch": 0.88, "learning_rate": 2.832056542836191e-05, "loss": 0.2796, "step": 18940 }, { "epoch": 0.88, "learning_rate": 2.8320097358525894e-05, "loss": 0.282, "step": 18945 }, { "epoch": 0.88, "learning_rate": 2.8319629288689874e-05, "loss": 0.5019, "step": 18950 }, { "epoch": 0.88, "learning_rate": 2.8319161218853854e-05, "loss": 0.2091, "step": 18955 }, { "epoch": 0.88, "learning_rate": 2.8318693149017837e-05, "loss": 0.0502, "step": 18960 }, { "epoch": 0.88, "learning_rate": 2.8318225079181817e-05, "loss": 0.0924, "step": 18965 }, { "epoch": 0.89, "learning_rate": 2.8317757009345797e-05, "loss": 0.1117, "step": 18970 }, { "epoch": 0.89, "learning_rate": 2.8317288939509773e-05, "loss": 0.1084, "step": 18975 }, { "epoch": 0.89, "learning_rate": 2.8316820869673756e-05, "loss": 0.138, "step": 18980 }, { "epoch": 0.89, "learning_rate": 2.8316352799837736e-05, "loss": 0.1402, "step": 18985 }, { "epoch": 0.89, "learning_rate": 2.8315884730001716e-05, "loss": 0.1729, "step": 18990 }, { "epoch": 0.89, "learning_rate": 2.8315416660165696e-05, "loss": 0.2369, "step": 18995 }, { "epoch": 0.89, "learning_rate": 2.831494859032968e-05, "loss": 0.414, "step": 19000 }, { "epoch": 0.89, "learning_rate": 2.831448052049366e-05, "loss": 0.223, "step": 19005 }, { "epoch": 0.89, "learning_rate": 2.831401245065764e-05, "loss": 0.1283, "step": 19010 }, { "epoch": 0.89, "learning_rate": 2.8313544380821618e-05, "loss": 0.1046, "step": 19015 }, { "epoch": 0.89, "learning_rate": 2.83130763109856e-05, "loss": 0.1075, "step": 19020 }, { "epoch": 0.89, "learning_rate": 2.831260824114958e-05, "loss": 0.1584, "step": 19025 }, { "epoch": 0.89, "learning_rate": 2.831214017131356e-05, "loss": 0.1721, "step": 19030 }, { "epoch": 0.89, "learning_rate": 2.831167210147754e-05, "loss": 0.1674, "step": 19035 }, { "epoch": 0.89, "learning_rate": 2.831120403164152e-05, "loss": 0.21, "step": 19040 }, { "epoch": 0.89, "learning_rate": 2.83107359618055e-05, "loss": 0.2495, "step": 19045 }, { "epoch": 0.89, "learning_rate": 2.831026789196948e-05, "loss": 0.6011, "step": 19050 }, { "epoch": 0.89, "learning_rate": 2.8309799822133464e-05, "loss": 0.1744, "step": 19055 }, { "epoch": 0.89, "learning_rate": 2.8309331752297443e-05, "loss": 0.0949, "step": 19060 }, { "epoch": 0.89, "learning_rate": 2.8308863682461423e-05, "loss": 0.0569, "step": 19065 }, { "epoch": 0.89, "learning_rate": 2.8308395612625403e-05, "loss": 0.109, "step": 19070 }, { "epoch": 0.89, "learning_rate": 2.8307927542789386e-05, "loss": 0.1346, "step": 19075 }, { "epoch": 0.89, "learning_rate": 2.8307459472953366e-05, "loss": 0.1163, "step": 19080 }, { "epoch": 0.89, "learning_rate": 2.8306991403117346e-05, "loss": 0.1234, "step": 19085 }, { "epoch": 0.89, "learning_rate": 2.830652333328133e-05, "loss": 0.2103, "step": 19090 }, { "epoch": 0.89, "learning_rate": 2.830605526344531e-05, "loss": 0.3757, "step": 19095 }, { "epoch": 0.89, "learning_rate": 2.8305587193609285e-05, "loss": 0.5541, "step": 19100 }, { "epoch": 0.89, "learning_rate": 2.8305119123773265e-05, "loss": 0.1623, "step": 19105 }, { "epoch": 0.89, "learning_rate": 2.830465105393725e-05, "loss": 0.0796, "step": 19110 }, { "epoch": 0.89, "learning_rate": 2.8304182984101228e-05, "loss": 0.1489, "step": 19115 }, { "epoch": 0.89, "learning_rate": 2.8303714914265208e-05, "loss": 0.0964, "step": 19120 }, { "epoch": 0.89, "learning_rate": 2.8303246844429188e-05, "loss": 0.1179, "step": 19125 }, { "epoch": 0.89, "learning_rate": 2.830277877459317e-05, "loss": 0.2085, "step": 19130 }, { "epoch": 0.89, "learning_rate": 2.830231070475715e-05, "loss": 0.168, "step": 19135 }, { "epoch": 0.89, "learning_rate": 2.830184263492113e-05, "loss": 0.2215, "step": 19140 }, { "epoch": 0.89, "learning_rate": 2.8301374565085114e-05, "loss": 0.2907, "step": 19145 }, { "epoch": 0.89, "learning_rate": 2.8300906495249094e-05, "loss": 0.6909, "step": 19150 }, { "epoch": 0.89, "learning_rate": 2.8300438425413073e-05, "loss": 0.1187, "step": 19155 }, { "epoch": 0.89, "learning_rate": 2.8299970355577053e-05, "loss": 0.0872, "step": 19160 }, { "epoch": 0.89, "learning_rate": 2.8299502285741033e-05, "loss": 0.1098, "step": 19165 }, { "epoch": 0.89, "learning_rate": 2.8299034215905013e-05, "loss": 0.1312, "step": 19170 }, { "epoch": 0.89, "learning_rate": 2.8298566146068993e-05, "loss": 0.0822, "step": 19175 }, { "epoch": 0.89, "learning_rate": 2.8298098076232973e-05, "loss": 0.1598, "step": 19180 }, { "epoch": 0.9, "learning_rate": 2.8297630006396956e-05, "loss": 0.152, "step": 19185 }, { "epoch": 0.9, "learning_rate": 2.8297161936560936e-05, "loss": 0.2295, "step": 19190 }, { "epoch": 0.9, "learning_rate": 2.8296693866724915e-05, "loss": 0.3206, "step": 19195 }, { "epoch": 0.9, "learning_rate": 2.8296225796888895e-05, "loss": 0.4797, "step": 19200 }, { "epoch": 0.9, "learning_rate": 2.829575772705288e-05, "loss": 0.1981, "step": 19205 }, { "epoch": 0.9, "learning_rate": 2.8295289657216858e-05, "loss": 0.0623, "step": 19210 }, { "epoch": 0.9, "learning_rate": 2.8294821587380838e-05, "loss": 0.1144, "step": 19215 }, { "epoch": 0.9, "learning_rate": 2.829435351754482e-05, "loss": 0.1894, "step": 19220 }, { "epoch": 0.9, "learning_rate": 2.8293885447708798e-05, "loss": 0.1042, "step": 19225 }, { "epoch": 0.9, "learning_rate": 2.8293417377872778e-05, "loss": 0.1333, "step": 19230 }, { "epoch": 0.9, "learning_rate": 2.8292949308036757e-05, "loss": 0.1473, "step": 19235 }, { "epoch": 0.9, "learning_rate": 2.829248123820074e-05, "loss": 0.1897, "step": 19240 }, { "epoch": 0.9, "learning_rate": 2.829201316836472e-05, "loss": 0.4412, "step": 19245 }, { "epoch": 0.9, "learning_rate": 2.82915450985287e-05, "loss": 0.5611, "step": 19250 }, { "epoch": 0.9, "learning_rate": 2.829107702869268e-05, "loss": 0.2042, "step": 19255 }, { "epoch": 0.9, "learning_rate": 2.8290608958856663e-05, "loss": 0.0395, "step": 19260 }, { "epoch": 0.9, "learning_rate": 2.8290140889020643e-05, "loss": 0.1283, "step": 19265 }, { "epoch": 0.9, "learning_rate": 2.8289672819184623e-05, "loss": 0.0516, "step": 19270 }, { "epoch": 0.9, "learning_rate": 2.8289204749348606e-05, "loss": 0.0904, "step": 19275 }, { "epoch": 0.9, "learning_rate": 2.8288736679512586e-05, "loss": 0.1615, "step": 19280 }, { "epoch": 0.9, "learning_rate": 2.8288268609676566e-05, "loss": 0.2146, "step": 19285 }, { "epoch": 0.9, "learning_rate": 2.8287800539840542e-05, "loss": 0.2287, "step": 19290 }, { "epoch": 0.9, "learning_rate": 2.8287332470004525e-05, "loss": 0.2794, "step": 19295 }, { "epoch": 0.9, "learning_rate": 2.8286864400168505e-05, "loss": 0.3963, "step": 19300 }, { "epoch": 0.9, "learning_rate": 2.8286396330332485e-05, "loss": 0.2743, "step": 19305 }, { "epoch": 0.9, "learning_rate": 2.8285928260496465e-05, "loss": 0.0492, "step": 19310 }, { "epoch": 0.9, "learning_rate": 2.8285460190660448e-05, "loss": 0.1209, "step": 19315 }, { "epoch": 0.9, "learning_rate": 2.8284992120824428e-05, "loss": 0.091, "step": 19320 }, { "epoch": 0.9, "learning_rate": 2.8284524050988408e-05, "loss": 0.1206, "step": 19325 }, { "epoch": 0.9, "learning_rate": 2.828405598115239e-05, "loss": 0.1549, "step": 19330 }, { "epoch": 0.9, "learning_rate": 2.828358791131637e-05, "loss": 0.2142, "step": 19335 }, { "epoch": 0.9, "learning_rate": 2.828311984148035e-05, "loss": 0.1949, "step": 19340 }, { "epoch": 0.9, "learning_rate": 2.828265177164433e-05, "loss": 0.2274, "step": 19345 }, { "epoch": 0.9, "learning_rate": 2.8282183701808313e-05, "loss": 0.3556, "step": 19350 }, { "epoch": 0.9, "learning_rate": 2.828171563197229e-05, "loss": 0.1303, "step": 19355 }, { "epoch": 0.9, "learning_rate": 2.828124756213627e-05, "loss": 0.1132, "step": 19360 }, { "epoch": 0.9, "learning_rate": 2.828077949230025e-05, "loss": 0.1327, "step": 19365 }, { "epoch": 0.9, "learning_rate": 2.8280311422464233e-05, "loss": 0.0916, "step": 19370 }, { "epoch": 0.9, "learning_rate": 2.8279843352628213e-05, "loss": 0.1513, "step": 19375 }, { "epoch": 0.9, "learning_rate": 2.8279375282792192e-05, "loss": 0.124, "step": 19380 }, { "epoch": 0.9, "learning_rate": 2.8278907212956176e-05, "loss": 0.1891, "step": 19385 }, { "epoch": 0.9, "learning_rate": 2.8278439143120155e-05, "loss": 0.1877, "step": 19390 }, { "epoch": 0.9, "learning_rate": 2.8277971073284135e-05, "loss": 0.2175, "step": 19395 }, { "epoch": 0.91, "learning_rate": 2.8277503003448115e-05, "loss": 0.4662, "step": 19400 }, { "epoch": 0.91, "learning_rate": 2.8277034933612098e-05, "loss": 0.1913, "step": 19405 }, { "epoch": 0.91, "learning_rate": 2.8276566863776078e-05, "loss": 0.0924, "step": 19410 }, { "epoch": 0.91, "learning_rate": 2.8276098793940054e-05, "loss": 0.1196, "step": 19415 }, { "epoch": 0.91, "learning_rate": 2.8275630724104034e-05, "loss": 0.151, "step": 19420 }, { "epoch": 0.91, "learning_rate": 2.8275162654268018e-05, "loss": 0.0961, "step": 19425 }, { "epoch": 0.91, "learning_rate": 2.8274694584431997e-05, "loss": 0.1317, "step": 19430 }, { "epoch": 0.91, "learning_rate": 2.8274226514595977e-05, "loss": 0.1752, "step": 19435 }, { "epoch": 0.91, "learning_rate": 2.8273758444759957e-05, "loss": 0.2372, "step": 19440 }, { "epoch": 0.91, "learning_rate": 2.827329037492394e-05, "loss": 0.2799, "step": 19445 }, { "epoch": 0.91, "learning_rate": 2.827282230508792e-05, "loss": 0.5546, "step": 19450 }, { "epoch": 0.91, "learning_rate": 2.82723542352519e-05, "loss": 0.1655, "step": 19455 }, { "epoch": 0.91, "learning_rate": 2.8271886165415883e-05, "loss": 0.0895, "step": 19460 }, { "epoch": 0.91, "learning_rate": 2.8271418095579863e-05, "loss": 0.1524, "step": 19465 }, { "epoch": 0.91, "learning_rate": 2.8270950025743843e-05, "loss": 0.1442, "step": 19470 }, { "epoch": 0.91, "learning_rate": 2.8270481955907822e-05, "loss": 0.223, "step": 19475 }, { "epoch": 0.91, "learning_rate": 2.8270013886071802e-05, "loss": 0.175, "step": 19480 }, { "epoch": 0.91, "learning_rate": 2.8269545816235782e-05, "loss": 0.2154, "step": 19485 }, { "epoch": 0.91, "learning_rate": 2.8269077746399762e-05, "loss": 0.3557, "step": 19490 }, { "epoch": 0.91, "learning_rate": 2.8268609676563742e-05, "loss": 0.2562, "step": 19495 }, { "epoch": 0.91, "learning_rate": 2.8268141606727725e-05, "loss": 0.4082, "step": 19500 }, { "epoch": 0.91, "learning_rate": 2.8267673536891705e-05, "loss": 0.2267, "step": 19505 }, { "epoch": 0.91, "learning_rate": 2.8267205467055685e-05, "loss": 0.0698, "step": 19510 }, { "epoch": 0.91, "learning_rate": 2.8266737397219668e-05, "loss": 0.1245, "step": 19515 }, { "epoch": 0.91, "learning_rate": 2.8266269327383648e-05, "loss": 0.1111, "step": 19520 }, { "epoch": 0.91, "learning_rate": 2.8265801257547627e-05, "loss": 0.1291, "step": 19525 }, { "epoch": 0.91, "learning_rate": 2.8265333187711607e-05, "loss": 0.1354, "step": 19530 }, { "epoch": 0.91, "learning_rate": 2.826486511787559e-05, "loss": 0.1767, "step": 19535 }, { "epoch": 0.91, "learning_rate": 2.826439704803957e-05, "loss": 0.1922, "step": 19540 }, { "epoch": 0.91, "learning_rate": 2.8263928978203547e-05, "loss": 0.2851, "step": 19545 }, { "epoch": 0.91, "learning_rate": 2.8263460908367527e-05, "loss": 0.4096, "step": 19550 }, { "epoch": 0.91, "learning_rate": 2.826299283853151e-05, "loss": 0.1669, "step": 19555 }, { "epoch": 0.91, "learning_rate": 2.826252476869549e-05, "loss": 0.0812, "step": 19560 }, { "epoch": 0.91, "learning_rate": 2.826205669885947e-05, "loss": 0.1695, "step": 19565 }, { "epoch": 0.91, "learning_rate": 2.8261588629023453e-05, "loss": 0.1234, "step": 19570 }, { "epoch": 0.91, "learning_rate": 2.8261120559187432e-05, "loss": 0.0983, "step": 19575 }, { "epoch": 0.91, "learning_rate": 2.8260652489351412e-05, "loss": 0.1572, "step": 19580 }, { "epoch": 0.91, "learning_rate": 2.8260184419515392e-05, "loss": 0.1783, "step": 19585 }, { "epoch": 0.91, "learning_rate": 2.8259716349679375e-05, "loss": 0.2718, "step": 19590 }, { "epoch": 0.91, "learning_rate": 2.8259248279843355e-05, "loss": 0.2048, "step": 19595 }, { "epoch": 0.91, "learning_rate": 2.8258780210007335e-05, "loss": 0.6031, "step": 19600 }, { "epoch": 0.91, "learning_rate": 2.825831214017131e-05, "loss": 0.1302, "step": 19605 }, { "epoch": 0.92, "learning_rate": 2.8257844070335294e-05, "loss": 0.0921, "step": 19610 }, { "epoch": 0.92, "learning_rate": 2.8257376000499274e-05, "loss": 0.1714, "step": 19615 }, { "epoch": 0.92, "learning_rate": 2.8256907930663254e-05, "loss": 0.088, "step": 19620 }, { "epoch": 0.92, "learning_rate": 2.8256439860827234e-05, "loss": 0.1346, "step": 19625 }, { "epoch": 0.92, "learning_rate": 2.8255971790991217e-05, "loss": 0.121, "step": 19630 }, { "epoch": 0.92, "learning_rate": 2.8255503721155197e-05, "loss": 0.2561, "step": 19635 }, { "epoch": 0.92, "learning_rate": 2.8255035651319177e-05, "loss": 0.1892, "step": 19640 }, { "epoch": 0.92, "learning_rate": 2.825456758148316e-05, "loss": 0.309, "step": 19645 }, { "epoch": 0.92, "learning_rate": 2.825409951164714e-05, "loss": 0.5683, "step": 19650 }, { "epoch": 0.92, "learning_rate": 2.825363144181112e-05, "loss": 0.2245, "step": 19655 }, { "epoch": 0.92, "learning_rate": 2.82531633719751e-05, "loss": 0.0395, "step": 19660 }, { "epoch": 0.92, "learning_rate": 2.8252695302139083e-05, "loss": 0.0715, "step": 19665 }, { "epoch": 0.92, "learning_rate": 2.825222723230306e-05, "loss": 0.064, "step": 19670 }, { "epoch": 0.92, "learning_rate": 2.825175916246704e-05, "loss": 0.1127, "step": 19675 }, { "epoch": 0.92, "learning_rate": 2.825129109263102e-05, "loss": 0.2057, "step": 19680 }, { "epoch": 0.92, "learning_rate": 2.8250823022795002e-05, "loss": 0.2353, "step": 19685 }, { "epoch": 0.92, "learning_rate": 2.8250354952958982e-05, "loss": 0.1715, "step": 19690 }, { "epoch": 0.92, "learning_rate": 2.824988688312296e-05, "loss": 0.2828, "step": 19695 }, { "epoch": 0.92, "learning_rate": 2.8249418813286945e-05, "loss": 0.4615, "step": 19700 }, { "epoch": 0.92, "learning_rate": 2.8248950743450925e-05, "loss": 0.2102, "step": 19705 }, { "epoch": 0.92, "learning_rate": 2.8248482673614904e-05, "loss": 0.1068, "step": 19710 }, { "epoch": 0.92, "learning_rate": 2.8248014603778884e-05, "loss": 0.0895, "step": 19715 }, { "epoch": 0.92, "learning_rate": 2.8247546533942867e-05, "loss": 0.1568, "step": 19720 }, { "epoch": 0.92, "learning_rate": 2.8247078464106847e-05, "loss": 0.1136, "step": 19725 }, { "epoch": 0.92, "learning_rate": 2.8246610394270827e-05, "loss": 0.1937, "step": 19730 }, { "epoch": 0.92, "learning_rate": 2.8246142324434803e-05, "loss": 0.195, "step": 19735 }, { "epoch": 0.92, "learning_rate": 2.8245674254598787e-05, "loss": 0.1874, "step": 19740 }, { "epoch": 0.92, "learning_rate": 2.8245206184762766e-05, "loss": 0.151, "step": 19745 }, { "epoch": 0.92, "learning_rate": 2.8244738114926746e-05, "loss": 0.4252, "step": 19750 }, { "epoch": 0.92, "learning_rate": 2.824427004509073e-05, "loss": 0.1778, "step": 19755 }, { "epoch": 0.92, "learning_rate": 2.824380197525471e-05, "loss": 0.0692, "step": 19760 }, { "epoch": 0.92, "learning_rate": 2.824333390541869e-05, "loss": 0.0619, "step": 19765 }, { "epoch": 0.92, "learning_rate": 2.824286583558267e-05, "loss": 0.0554, "step": 19770 }, { "epoch": 0.92, "learning_rate": 2.8242397765746652e-05, "loss": 0.1056, "step": 19775 }, { "epoch": 0.92, "learning_rate": 2.8241929695910632e-05, "loss": 0.1334, "step": 19780 }, { "epoch": 0.92, "learning_rate": 2.8241461626074612e-05, "loss": 0.2277, "step": 19785 }, { "epoch": 0.92, "learning_rate": 2.824099355623859e-05, "loss": 0.2113, "step": 19790 }, { "epoch": 0.92, "learning_rate": 2.824052548640257e-05, "loss": 0.3137, "step": 19795 }, { "epoch": 0.92, "learning_rate": 2.824005741656655e-05, "loss": 0.6677, "step": 19800 }, { "epoch": 0.92, "learning_rate": 2.823958934673053e-05, "loss": 0.1618, "step": 19805 }, { "epoch": 0.92, "learning_rate": 2.823912127689451e-05, "loss": 0.0787, "step": 19810 }, { "epoch": 0.92, "learning_rate": 2.8238653207058494e-05, "loss": 0.0507, "step": 19815 }, { "epoch": 0.92, "learning_rate": 2.8238185137222474e-05, "loss": 0.0941, "step": 19820 }, { "epoch": 0.93, "learning_rate": 2.8237717067386454e-05, "loss": 0.1134, "step": 19825 }, { "epoch": 0.93, "learning_rate": 2.8237248997550437e-05, "loss": 0.1516, "step": 19830 }, { "epoch": 0.93, "learning_rate": 2.8236780927714417e-05, "loss": 0.159, "step": 19835 }, { "epoch": 0.93, "learning_rate": 2.8236312857878397e-05, "loss": 0.221, "step": 19840 }, { "epoch": 0.93, "learning_rate": 2.8235844788042376e-05, "loss": 0.2901, "step": 19845 }, { "epoch": 0.93, "learning_rate": 2.823537671820636e-05, "loss": 0.5143, "step": 19850 }, { "epoch": 0.93, "learning_rate": 2.823490864837034e-05, "loss": 0.1777, "step": 19855 }, { "epoch": 0.93, "learning_rate": 2.8234440578534316e-05, "loss": 0.118, "step": 19860 }, { "epoch": 0.93, "learning_rate": 2.8233972508698296e-05, "loss": 0.087, "step": 19865 }, { "epoch": 0.93, "learning_rate": 2.823350443886228e-05, "loss": 0.0556, "step": 19870 }, { "epoch": 0.93, "learning_rate": 2.823303636902626e-05, "loss": 0.0683, "step": 19875 }, { "epoch": 0.93, "learning_rate": 2.823256829919024e-05, "loss": 0.1517, "step": 19880 }, { "epoch": 0.93, "learning_rate": 2.8232100229354222e-05, "loss": 0.1644, "step": 19885 }, { "epoch": 0.93, "learning_rate": 2.82316321595182e-05, "loss": 0.2846, "step": 19890 }, { "epoch": 0.93, "learning_rate": 2.823116408968218e-05, "loss": 0.1794, "step": 19895 }, { "epoch": 0.93, "learning_rate": 2.823069601984616e-05, "loss": 0.5903, "step": 19900 }, { "epoch": 0.93, "learning_rate": 2.8230227950010144e-05, "loss": 0.2026, "step": 19905 }, { "epoch": 0.93, "learning_rate": 2.8229759880174124e-05, "loss": 0.0524, "step": 19910 }, { "epoch": 0.93, "learning_rate": 2.8229291810338104e-05, "loss": 0.0721, "step": 19915 }, { "epoch": 0.93, "learning_rate": 2.8228823740502084e-05, "loss": 0.1443, "step": 19920 }, { "epoch": 0.93, "learning_rate": 2.8228355670666064e-05, "loss": 0.1695, "step": 19925 }, { "epoch": 0.93, "learning_rate": 2.8227887600830043e-05, "loss": 0.1654, "step": 19930 }, { "epoch": 0.93, "learning_rate": 2.8227419530994023e-05, "loss": 0.1993, "step": 19935 }, { "epoch": 0.93, "learning_rate": 2.8226951461158006e-05, "loss": 0.1515, "step": 19940 }, { "epoch": 0.93, "learning_rate": 2.8226483391321986e-05, "loss": 0.2342, "step": 19945 }, { "epoch": 0.93, "learning_rate": 2.8226015321485966e-05, "loss": 0.4621, "step": 19950 }, { "epoch": 0.93, "learning_rate": 2.8225547251649946e-05, "loss": 0.1994, "step": 19955 }, { "epoch": 0.93, "learning_rate": 2.822507918181393e-05, "loss": 0.0835, "step": 19960 }, { "epoch": 0.93, "learning_rate": 2.822461111197791e-05, "loss": 0.0973, "step": 19965 }, { "epoch": 0.93, "learning_rate": 2.822414304214189e-05, "loss": 0.0795, "step": 19970 }, { "epoch": 0.93, "learning_rate": 2.822367497230587e-05, "loss": 0.1491, "step": 19975 }, { "epoch": 0.93, "learning_rate": 2.8223206902469852e-05, "loss": 0.1642, "step": 19980 }, { "epoch": 0.93, "learning_rate": 2.8222738832633828e-05, "loss": 0.135, "step": 19985 }, { "epoch": 0.93, "learning_rate": 2.8222270762797808e-05, "loss": 0.2358, "step": 19990 }, { "epoch": 0.93, "learning_rate": 2.822180269296179e-05, "loss": 0.1661, "step": 19995 }, { "epoch": 0.93, "learning_rate": 2.822133462312577e-05, "loss": 0.4391, "step": 20000 }, { "epoch": 0.93, "learning_rate": 2.822086655328975e-05, "loss": 0.1482, "step": 20005 }, { "epoch": 0.93, "learning_rate": 2.822039848345373e-05, "loss": 0.0925, "step": 20010 }, { "epoch": 0.93, "learning_rate": 2.8219930413617714e-05, "loss": 0.0681, "step": 20015 }, { "epoch": 0.93, "learning_rate": 2.8219462343781694e-05, "loss": 0.082, "step": 20020 }, { "epoch": 0.93, "learning_rate": 2.8218994273945674e-05, "loss": 0.11, "step": 20025 }, { "epoch": 0.93, "learning_rate": 2.8218526204109653e-05, "loss": 0.1559, "step": 20030 }, { "epoch": 0.93, "learning_rate": 2.8218058134273637e-05, "loss": 0.1582, "step": 20035 }, { "epoch": 0.94, "learning_rate": 2.8217590064437616e-05, "loss": 0.1825, "step": 20040 }, { "epoch": 0.94, "learning_rate": 2.8217121994601596e-05, "loss": 0.347, "step": 20045 }, { "epoch": 0.94, "learning_rate": 2.8216653924765573e-05, "loss": 0.5309, "step": 20050 }, { "epoch": 0.94, "learning_rate": 2.8216185854929556e-05, "loss": 0.1357, "step": 20055 }, { "epoch": 0.94, "learning_rate": 2.8215717785093536e-05, "loss": 0.0813, "step": 20060 }, { "epoch": 0.94, "learning_rate": 2.8215249715257515e-05, "loss": 0.0771, "step": 20065 }, { "epoch": 0.94, "learning_rate": 2.82147816454215e-05, "loss": 0.1573, "step": 20070 }, { "epoch": 0.94, "learning_rate": 2.821431357558548e-05, "loss": 0.1563, "step": 20075 }, { "epoch": 0.94, "learning_rate": 2.821384550574946e-05, "loss": 0.1515, "step": 20080 }, { "epoch": 0.94, "learning_rate": 2.8213377435913438e-05, "loss": 0.1877, "step": 20085 }, { "epoch": 0.94, "learning_rate": 2.821290936607742e-05, "loss": 0.2433, "step": 20090 }, { "epoch": 0.94, "learning_rate": 2.82124412962414e-05, "loss": 0.2405, "step": 20095 }, { "epoch": 0.94, "learning_rate": 2.821197322640538e-05, "loss": 0.4775, "step": 20100 }, { "epoch": 0.94, "learning_rate": 2.821150515656936e-05, "loss": 0.209, "step": 20105 }, { "epoch": 0.94, "learning_rate": 2.821103708673334e-05, "loss": 0.0433, "step": 20110 }, { "epoch": 0.94, "learning_rate": 2.821056901689732e-05, "loss": 0.1083, "step": 20115 }, { "epoch": 0.94, "learning_rate": 2.82101009470613e-05, "loss": 0.139, "step": 20120 }, { "epoch": 0.94, "learning_rate": 2.8209632877225283e-05, "loss": 0.1155, "step": 20125 }, { "epoch": 0.94, "learning_rate": 2.8209164807389263e-05, "loss": 0.1707, "step": 20130 }, { "epoch": 0.94, "learning_rate": 2.8208696737553243e-05, "loss": 0.183, "step": 20135 }, { "epoch": 0.94, "learning_rate": 2.8208228667717223e-05, "loss": 0.1512, "step": 20140 }, { "epoch": 0.94, "learning_rate": 2.8207760597881206e-05, "loss": 0.2248, "step": 20145 }, { "epoch": 0.94, "learning_rate": 2.8207292528045186e-05, "loss": 0.5306, "step": 20150 }, { "epoch": 0.94, "learning_rate": 2.8206824458209166e-05, "loss": 0.2311, "step": 20155 }, { "epoch": 0.94, "learning_rate": 2.8206356388373146e-05, "loss": 0.0609, "step": 20160 }, { "epoch": 0.94, "learning_rate": 2.820588831853713e-05, "loss": 0.1172, "step": 20165 }, { "epoch": 0.94, "learning_rate": 2.820542024870111e-05, "loss": 0.0703, "step": 20170 }, { "epoch": 0.94, "learning_rate": 2.8204952178865085e-05, "loss": 0.1642, "step": 20175 }, { "epoch": 0.94, "learning_rate": 2.8204484109029068e-05, "loss": 0.1785, "step": 20180 }, { "epoch": 0.94, "learning_rate": 2.8204016039193048e-05, "loss": 0.26, "step": 20185 }, { "epoch": 0.94, "learning_rate": 2.8203547969357028e-05, "loss": 0.132, "step": 20190 }, { "epoch": 0.94, "learning_rate": 2.8203079899521008e-05, "loss": 0.2015, "step": 20195 }, { "epoch": 0.94, "learning_rate": 2.820261182968499e-05, "loss": 0.4543, "step": 20200 }, { "epoch": 0.94, "learning_rate": 2.820214375984897e-05, "loss": 0.1331, "step": 20205 }, { "epoch": 0.94, "learning_rate": 2.820167569001295e-05, "loss": 0.0367, "step": 20210 }, { "epoch": 0.94, "learning_rate": 2.820120762017693e-05, "loss": 0.1203, "step": 20215 }, { "epoch": 0.94, "learning_rate": 2.8200739550340914e-05, "loss": 0.1229, "step": 20220 }, { "epoch": 0.94, "learning_rate": 2.8200271480504893e-05, "loss": 0.1072, "step": 20225 }, { "epoch": 0.94, "learning_rate": 2.8199803410668873e-05, "loss": 0.2379, "step": 20230 }, { "epoch": 0.94, "learning_rate": 2.8199335340832853e-05, "loss": 0.1161, "step": 20235 }, { "epoch": 0.94, "learning_rate": 2.8198867270996833e-05, "loss": 0.2194, "step": 20240 }, { "epoch": 0.94, "learning_rate": 2.8198399201160813e-05, "loss": 0.2563, "step": 20245 }, { "epoch": 0.94, "learning_rate": 2.8197931131324792e-05, "loss": 0.6662, "step": 20250 }, { "epoch": 0.95, "learning_rate": 2.8197463061488776e-05, "loss": 0.1148, "step": 20255 }, { "epoch": 0.95, "learning_rate": 2.8196994991652755e-05, "loss": 0.0444, "step": 20260 }, { "epoch": 0.95, "learning_rate": 2.8196526921816735e-05, "loss": 0.0626, "step": 20265 }, { "epoch": 0.95, "learning_rate": 2.8196058851980715e-05, "loss": 0.1184, "step": 20270 }, { "epoch": 0.95, "learning_rate": 2.81955907821447e-05, "loss": 0.0785, "step": 20275 }, { "epoch": 0.95, "learning_rate": 2.8195122712308678e-05, "loss": 0.098, "step": 20280 }, { "epoch": 0.95, "learning_rate": 2.8194654642472658e-05, "loss": 0.2003, "step": 20285 }, { "epoch": 0.95, "learning_rate": 2.8194186572636638e-05, "loss": 0.1674, "step": 20290 }, { "epoch": 0.95, "learning_rate": 2.819371850280062e-05, "loss": 0.3642, "step": 20295 }, { "epoch": 0.95, "learning_rate": 2.8193250432964597e-05, "loss": 0.4801, "step": 20300 }, { "epoch": 0.95, "learning_rate": 2.8192782363128577e-05, "loss": 0.2435, "step": 20305 }, { "epoch": 0.95, "learning_rate": 2.819231429329256e-05, "loss": 0.0955, "step": 20310 }, { "epoch": 0.95, "learning_rate": 2.819184622345654e-05, "loss": 0.0778, "step": 20315 }, { "epoch": 0.95, "learning_rate": 2.819137815362052e-05, "loss": 0.1128, "step": 20320 }, { "epoch": 0.95, "learning_rate": 2.81909100837845e-05, "loss": 0.1563, "step": 20325 }, { "epoch": 0.95, "learning_rate": 2.8190442013948483e-05, "loss": 0.0633, "step": 20330 }, { "epoch": 0.95, "learning_rate": 2.8189973944112463e-05, "loss": 0.1201, "step": 20335 }, { "epoch": 0.95, "learning_rate": 2.8189505874276443e-05, "loss": 0.1788, "step": 20340 }, { "epoch": 0.95, "learning_rate": 2.8189037804440423e-05, "loss": 0.2385, "step": 20345 }, { "epoch": 0.95, "learning_rate": 2.8188569734604406e-05, "loss": 0.662, "step": 20350 }, { "epoch": 0.95, "learning_rate": 2.8188101664768386e-05, "loss": 0.2491, "step": 20355 }, { "epoch": 0.95, "learning_rate": 2.8187633594932365e-05, "loss": 0.1046, "step": 20360 }, { "epoch": 0.95, "learning_rate": 2.8187165525096345e-05, "loss": 0.0764, "step": 20365 }, { "epoch": 0.95, "learning_rate": 2.8186697455260325e-05, "loss": 0.1108, "step": 20370 }, { "epoch": 0.95, "learning_rate": 2.8186229385424305e-05, "loss": 0.1127, "step": 20375 }, { "epoch": 0.95, "learning_rate": 2.8185761315588285e-05, "loss": 0.1035, "step": 20380 }, { "epoch": 0.95, "learning_rate": 2.8185293245752268e-05, "loss": 0.1857, "step": 20385 }, { "epoch": 0.95, "learning_rate": 2.8184825175916248e-05, "loss": 0.2943, "step": 20390 }, { "epoch": 0.95, "learning_rate": 2.8184357106080227e-05, "loss": 0.3479, "step": 20395 }, { "epoch": 0.95, "learning_rate": 2.8183889036244207e-05, "loss": 0.5486, "step": 20400 }, { "epoch": 0.95, "learning_rate": 2.818342096640819e-05, "loss": 0.1338, "step": 20405 }, { "epoch": 0.95, "learning_rate": 2.818295289657217e-05, "loss": 0.0795, "step": 20410 }, { "epoch": 0.95, "learning_rate": 2.818248482673615e-05, "loss": 0.065, "step": 20415 }, { "epoch": 0.95, "learning_rate": 2.818201675690013e-05, "loss": 0.1033, "step": 20420 }, { "epoch": 0.95, "learning_rate": 2.818154868706411e-05, "loss": 0.1537, "step": 20425 }, { "epoch": 0.95, "learning_rate": 2.818108061722809e-05, "loss": 0.1581, "step": 20430 }, { "epoch": 0.95, "learning_rate": 2.818061254739207e-05, "loss": 0.1775, "step": 20435 }, { "epoch": 0.95, "learning_rate": 2.8180144477556053e-05, "loss": 0.1888, "step": 20440 }, { "epoch": 0.95, "learning_rate": 2.8179676407720032e-05, "loss": 0.2106, "step": 20445 }, { "epoch": 0.95, "learning_rate": 2.8179208337884012e-05, "loss": 0.3352, "step": 20450 }, { "epoch": 0.95, "learning_rate": 2.8178740268047992e-05, "loss": 0.1744, "step": 20455 }, { "epoch": 0.95, "learning_rate": 2.8178272198211975e-05, "loss": 0.0895, "step": 20460 }, { "epoch": 0.95, "learning_rate": 2.8177804128375955e-05, "loss": 0.0839, "step": 20465 }, { "epoch": 0.96, "learning_rate": 2.8177336058539935e-05, "loss": 0.2121, "step": 20470 }, { "epoch": 0.96, "learning_rate": 2.8176867988703915e-05, "loss": 0.1547, "step": 20475 }, { "epoch": 0.96, "learning_rate": 2.8176399918867898e-05, "loss": 0.1767, "step": 20480 }, { "epoch": 0.96, "learning_rate": 2.8175931849031878e-05, "loss": 0.1461, "step": 20485 }, { "epoch": 0.96, "learning_rate": 2.8175463779195854e-05, "loss": 0.1666, "step": 20490 }, { "epoch": 0.96, "learning_rate": 2.8174995709359837e-05, "loss": 0.2874, "step": 20495 }, { "epoch": 0.96, "learning_rate": 2.8174527639523817e-05, "loss": 0.3162, "step": 20500 }, { "epoch": 0.96, "learning_rate": 2.8174059569687797e-05, "loss": 0.226, "step": 20505 }, { "epoch": 0.96, "learning_rate": 2.8173591499851777e-05, "loss": 0.0568, "step": 20510 }, { "epoch": 0.96, "learning_rate": 2.817312343001576e-05, "loss": 0.0747, "step": 20515 }, { "epoch": 0.96, "learning_rate": 2.817265536017974e-05, "loss": 0.0523, "step": 20520 }, { "epoch": 0.96, "learning_rate": 2.817218729034372e-05, "loss": 0.1438, "step": 20525 }, { "epoch": 0.96, "learning_rate": 2.81717192205077e-05, "loss": 0.109, "step": 20530 }, { "epoch": 0.96, "learning_rate": 2.8171251150671683e-05, "loss": 0.1728, "step": 20535 }, { "epoch": 0.96, "learning_rate": 2.8170783080835663e-05, "loss": 0.2249, "step": 20540 }, { "epoch": 0.96, "learning_rate": 2.8170315010999642e-05, "loss": 0.2156, "step": 20545 }, { "epoch": 0.96, "learning_rate": 2.8169846941163626e-05, "loss": 0.6314, "step": 20550 }, { "epoch": 0.96, "learning_rate": 2.8169378871327602e-05, "loss": 0.1749, "step": 20555 }, { "epoch": 0.96, "learning_rate": 2.8168910801491582e-05, "loss": 0.07, "step": 20560 }, { "epoch": 0.96, "learning_rate": 2.816844273165556e-05, "loss": 0.1051, "step": 20565 }, { "epoch": 0.96, "learning_rate": 2.8167974661819545e-05, "loss": 0.0834, "step": 20570 }, { "epoch": 0.96, "learning_rate": 2.8167506591983525e-05, "loss": 0.104, "step": 20575 }, { "epoch": 0.96, "learning_rate": 2.8167038522147504e-05, "loss": 0.1638, "step": 20580 }, { "epoch": 0.96, "learning_rate": 2.8166570452311484e-05, "loss": 0.1461, "step": 20585 }, { "epoch": 0.96, "learning_rate": 2.8166102382475467e-05, "loss": 0.2532, "step": 20590 }, { "epoch": 0.96, "learning_rate": 2.8165634312639447e-05, "loss": 0.2593, "step": 20595 }, { "epoch": 0.96, "learning_rate": 2.8165166242803427e-05, "loss": 0.6777, "step": 20600 }, { "epoch": 0.96, "learning_rate": 2.8164698172967407e-05, "loss": 0.1545, "step": 20605 }, { "epoch": 0.96, "learning_rate": 2.816423010313139e-05, "loss": 0.0451, "step": 20610 }, { "epoch": 0.96, "learning_rate": 2.8163762033295367e-05, "loss": 0.0954, "step": 20615 }, { "epoch": 0.96, "learning_rate": 2.8163293963459346e-05, "loss": 0.1675, "step": 20620 }, { "epoch": 0.96, "learning_rate": 2.816282589362333e-05, "loss": 0.1969, "step": 20625 }, { "epoch": 0.96, "learning_rate": 2.816235782378731e-05, "loss": 0.1834, "step": 20630 }, { "epoch": 0.96, "learning_rate": 2.816188975395129e-05, "loss": 0.2388, "step": 20635 }, { "epoch": 0.96, "learning_rate": 2.816142168411527e-05, "loss": 0.2778, "step": 20640 }, { "epoch": 0.96, "learning_rate": 2.8160953614279252e-05, "loss": 0.1899, "step": 20645 }, { "epoch": 0.96, "learning_rate": 2.8160485544443232e-05, "loss": 0.556, "step": 20650 }, { "epoch": 0.96, "learning_rate": 2.8160017474607212e-05, "loss": 0.1431, "step": 20655 }, { "epoch": 0.96, "learning_rate": 2.8159549404771192e-05, "loss": 0.0746, "step": 20660 }, { "epoch": 0.96, "learning_rate": 2.8159081334935175e-05, "loss": 0.0774, "step": 20665 }, { "epoch": 0.96, "learning_rate": 2.8158613265099155e-05, "loss": 0.143, "step": 20670 }, { "epoch": 0.96, "learning_rate": 2.8158145195263135e-05, "loss": 0.1597, "step": 20675 }, { "epoch": 0.96, "learning_rate": 2.8157677125427114e-05, "loss": 0.168, "step": 20680 }, { "epoch": 0.97, "learning_rate": 2.8157209055591094e-05, "loss": 0.3065, "step": 20685 }, { "epoch": 0.97, "learning_rate": 2.8156740985755074e-05, "loss": 0.1927, "step": 20690 }, { "epoch": 0.97, "learning_rate": 2.8156272915919054e-05, "loss": 0.2088, "step": 20695 }, { "epoch": 0.97, "learning_rate": 2.8155804846083037e-05, "loss": 0.574, "step": 20700 }, { "epoch": 0.97, "learning_rate": 2.8155336776247017e-05, "loss": 0.1484, "step": 20705 }, { "epoch": 0.97, "learning_rate": 2.8154868706410997e-05, "loss": 0.078, "step": 20710 }, { "epoch": 0.97, "learning_rate": 2.8154400636574976e-05, "loss": 0.0621, "step": 20715 }, { "epoch": 0.97, "learning_rate": 2.815393256673896e-05, "loss": 0.0976, "step": 20720 }, { "epoch": 0.97, "learning_rate": 2.815346449690294e-05, "loss": 0.1303, "step": 20725 }, { "epoch": 0.97, "learning_rate": 2.815299642706692e-05, "loss": 0.1321, "step": 20730 }, { "epoch": 0.97, "learning_rate": 2.8152528357230903e-05, "loss": 0.1589, "step": 20735 }, { "epoch": 0.97, "learning_rate": 2.8152060287394882e-05, "loss": 0.2044, "step": 20740 }, { "epoch": 0.97, "learning_rate": 2.815159221755886e-05, "loss": 0.3798, "step": 20745 }, { "epoch": 0.97, "learning_rate": 2.815112414772284e-05, "loss": 0.4709, "step": 20750 }, { "epoch": 0.97, "learning_rate": 2.8150656077886822e-05, "loss": 0.196, "step": 20755 }, { "epoch": 0.97, "learning_rate": 2.81501880080508e-05, "loss": 0.0599, "step": 20760 }, { "epoch": 0.97, "learning_rate": 2.814971993821478e-05, "loss": 0.0847, "step": 20765 }, { "epoch": 0.97, "learning_rate": 2.814925186837876e-05, "loss": 0.088, "step": 20770 }, { "epoch": 0.97, "learning_rate": 2.8148783798542744e-05, "loss": 0.0987, "step": 20775 }, { "epoch": 0.97, "learning_rate": 2.8148315728706724e-05, "loss": 0.2123, "step": 20780 }, { "epoch": 0.97, "learning_rate": 2.8147847658870704e-05, "loss": 0.2164, "step": 20785 }, { "epoch": 0.97, "learning_rate": 2.8147379589034687e-05, "loss": 0.1814, "step": 20790 }, { "epoch": 0.97, "learning_rate": 2.8146911519198667e-05, "loss": 0.3309, "step": 20795 }, { "epoch": 0.97, "learning_rate": 2.8146443449362647e-05, "loss": 0.4629, "step": 20800 }, { "epoch": 0.97, "learning_rate": 2.8145975379526623e-05, "loss": 0.1562, "step": 20805 }, { "epoch": 0.97, "learning_rate": 2.8145507309690607e-05, "loss": 0.027, "step": 20810 }, { "epoch": 0.97, "learning_rate": 2.8145039239854586e-05, "loss": 0.1902, "step": 20815 }, { "epoch": 0.97, "learning_rate": 2.8144571170018566e-05, "loss": 0.1324, "step": 20820 }, { "epoch": 0.97, "learning_rate": 2.8144103100182546e-05, "loss": 0.0776, "step": 20825 }, { "epoch": 0.97, "learning_rate": 2.814363503034653e-05, "loss": 0.1613, "step": 20830 }, { "epoch": 0.97, "learning_rate": 2.814316696051051e-05, "loss": 0.1971, "step": 20835 }, { "epoch": 0.97, "learning_rate": 2.814269889067449e-05, "loss": 0.2176, "step": 20840 }, { "epoch": 0.97, "learning_rate": 2.814223082083847e-05, "loss": 0.3287, "step": 20845 }, { "epoch": 0.97, "learning_rate": 2.8141762751002452e-05, "loss": 0.4665, "step": 20850 }, { "epoch": 0.97, "learning_rate": 2.8141294681166432e-05, "loss": 0.2047, "step": 20855 }, { "epoch": 0.97, "learning_rate": 2.814082661133041e-05, "loss": 0.0807, "step": 20860 }, { "epoch": 0.97, "learning_rate": 2.8140358541494395e-05, "loss": 0.1344, "step": 20865 }, { "epoch": 0.97, "learning_rate": 2.813989047165837e-05, "loss": 0.1654, "step": 20870 }, { "epoch": 0.97, "learning_rate": 2.813942240182235e-05, "loss": 0.0912, "step": 20875 }, { "epoch": 0.97, "learning_rate": 2.813895433198633e-05, "loss": 0.1348, "step": 20880 }, { "epoch": 0.97, "learning_rate": 2.8138486262150314e-05, "loss": 0.1929, "step": 20885 }, { "epoch": 0.97, "learning_rate": 2.8138018192314294e-05, "loss": 0.1943, "step": 20890 }, { "epoch": 0.97, "learning_rate": 2.8137550122478274e-05, "loss": 0.1947, "step": 20895 }, { "epoch": 0.98, "learning_rate": 2.8137082052642253e-05, "loss": 0.4322, "step": 20900 }, { "epoch": 0.98, "learning_rate": 2.8136613982806237e-05, "loss": 0.1952, "step": 20905 }, { "epoch": 0.98, "learning_rate": 2.8136145912970216e-05, "loss": 0.0827, "step": 20910 }, { "epoch": 0.98, "learning_rate": 2.8135677843134196e-05, "loss": 0.1029, "step": 20915 }, { "epoch": 0.98, "learning_rate": 2.813520977329818e-05, "loss": 0.0621, "step": 20920 }, { "epoch": 0.98, "learning_rate": 2.813474170346216e-05, "loss": 0.1272, "step": 20925 }, { "epoch": 0.98, "learning_rate": 2.813427363362614e-05, "loss": 0.1242, "step": 20930 }, { "epoch": 0.98, "learning_rate": 2.8133805563790116e-05, "loss": 0.2298, "step": 20935 }, { "epoch": 0.98, "learning_rate": 2.81333374939541e-05, "loss": 0.1909, "step": 20940 }, { "epoch": 0.98, "learning_rate": 2.813286942411808e-05, "loss": 0.2004, "step": 20945 }, { "epoch": 0.98, "learning_rate": 2.813240135428206e-05, "loss": 0.397, "step": 20950 }, { "epoch": 0.98, "learning_rate": 2.8131933284446038e-05, "loss": 0.1798, "step": 20955 }, { "epoch": 0.98, "learning_rate": 2.813146521461002e-05, "loss": 0.0624, "step": 20960 }, { "epoch": 0.98, "learning_rate": 2.8130997144774e-05, "loss": 0.0655, "step": 20965 }, { "epoch": 0.98, "learning_rate": 2.813052907493798e-05, "loss": 0.1675, "step": 20970 }, { "epoch": 0.98, "learning_rate": 2.8130061005101964e-05, "loss": 0.1494, "step": 20975 }, { "epoch": 0.98, "learning_rate": 2.8129592935265944e-05, "loss": 0.24, "step": 20980 }, { "epoch": 0.98, "learning_rate": 2.8129124865429924e-05, "loss": 0.3086, "step": 20985 }, { "epoch": 0.98, "learning_rate": 2.8128656795593904e-05, "loss": 0.1123, "step": 20990 }, { "epoch": 0.98, "learning_rate": 2.8128188725757884e-05, "loss": 0.2502, "step": 20995 }, { "epoch": 0.98, "learning_rate": 2.8127720655921863e-05, "loss": 0.4047, "step": 21000 }, { "epoch": 0.98, "learning_rate": 2.8127252586085843e-05, "loss": 0.2248, "step": 21005 }, { "epoch": 0.98, "learning_rate": 2.8126784516249823e-05, "loss": 0.043, "step": 21010 }, { "epoch": 0.98, "learning_rate": 2.8126316446413806e-05, "loss": 0.0671, "step": 21015 }, { "epoch": 0.98, "learning_rate": 2.8125848376577786e-05, "loss": 0.0699, "step": 21020 }, { "epoch": 0.98, "learning_rate": 2.8125380306741766e-05, "loss": 0.1004, "step": 21025 }, { "epoch": 0.98, "learning_rate": 2.8124912236905746e-05, "loss": 0.0738, "step": 21030 }, { "epoch": 0.98, "learning_rate": 2.812444416706973e-05, "loss": 0.1827, "step": 21035 }, { "epoch": 0.98, "learning_rate": 2.812397609723371e-05, "loss": 0.2244, "step": 21040 }, { "epoch": 0.98, "learning_rate": 2.812350802739769e-05, "loss": 0.3415, "step": 21045 }, { "epoch": 0.98, "learning_rate": 2.8123039957561672e-05, "loss": 0.4996, "step": 21050 }, { "epoch": 0.98, "learning_rate": 2.812257188772565e-05, "loss": 0.2611, "step": 21055 }, { "epoch": 0.98, "learning_rate": 2.8122103817889628e-05, "loss": 0.0686, "step": 21060 }, { "epoch": 0.98, "learning_rate": 2.8121635748053608e-05, "loss": 0.0538, "step": 21065 }, { "epoch": 0.98, "learning_rate": 2.812116767821759e-05, "loss": 0.0752, "step": 21070 }, { "epoch": 0.98, "learning_rate": 2.812069960838157e-05, "loss": 0.1292, "step": 21075 }, { "epoch": 0.98, "learning_rate": 2.812023153854555e-05, "loss": 0.1622, "step": 21080 }, { "epoch": 0.98, "learning_rate": 2.811976346870953e-05, "loss": 0.1767, "step": 21085 }, { "epoch": 0.98, "learning_rate": 2.8119295398873514e-05, "loss": 0.1789, "step": 21090 }, { "epoch": 0.98, "learning_rate": 2.8118827329037493e-05, "loss": 0.3069, "step": 21095 }, { "epoch": 0.98, "learning_rate": 2.8118359259201473e-05, "loss": 0.3624, "step": 21100 }, { "epoch": 0.98, "learning_rate": 2.8117891189365456e-05, "loss": 0.168, "step": 21105 }, { "epoch": 0.99, "learning_rate": 2.8117423119529436e-05, "loss": 0.148, "step": 21110 }, { "epoch": 0.99, "learning_rate": 2.8116955049693416e-05, "loss": 0.0757, "step": 21115 }, { "epoch": 0.99, "learning_rate": 2.8116486979857396e-05, "loss": 0.0868, "step": 21120 }, { "epoch": 0.99, "learning_rate": 2.8116018910021376e-05, "loss": 0.1677, "step": 21125 }, { "epoch": 0.99, "learning_rate": 2.8115550840185356e-05, "loss": 0.0994, "step": 21130 }, { "epoch": 0.99, "learning_rate": 2.8115082770349335e-05, "loss": 0.1551, "step": 21135 }, { "epoch": 0.99, "learning_rate": 2.8114614700513315e-05, "loss": 0.2855, "step": 21140 }, { "epoch": 0.99, "learning_rate": 2.81141466306773e-05, "loss": 0.2286, "step": 21145 }, { "epoch": 0.99, "learning_rate": 2.8113678560841278e-05, "loss": 0.413, "step": 21150 }, { "epoch": 0.99, "learning_rate": 2.8113210491005258e-05, "loss": 0.1841, "step": 21155 }, { "epoch": 0.99, "learning_rate": 2.811274242116924e-05, "loss": 0.0742, "step": 21160 }, { "epoch": 0.99, "learning_rate": 2.811227435133322e-05, "loss": 0.0712, "step": 21165 }, { "epoch": 0.99, "learning_rate": 2.81118062814972e-05, "loss": 0.0768, "step": 21170 }, { "epoch": 0.99, "learning_rate": 2.811133821166118e-05, "loss": 0.0907, "step": 21175 }, { "epoch": 0.99, "learning_rate": 2.8110870141825164e-05, "loss": 0.1989, "step": 21180 }, { "epoch": 0.99, "learning_rate": 2.811040207198914e-05, "loss": 0.2145, "step": 21185 }, { "epoch": 0.99, "learning_rate": 2.810993400215312e-05, "loss": 0.2252, "step": 21190 }, { "epoch": 0.99, "learning_rate": 2.81094659323171e-05, "loss": 0.238, "step": 21195 }, { "epoch": 0.99, "learning_rate": 2.8108997862481083e-05, "loss": 0.4606, "step": 21200 }, { "epoch": 0.99, "learning_rate": 2.8108529792645063e-05, "loss": 0.2023, "step": 21205 }, { "epoch": 0.99, "learning_rate": 2.8108061722809043e-05, "loss": 0.089, "step": 21210 }, { "epoch": 0.99, "learning_rate": 2.8107593652973023e-05, "loss": 0.1199, "step": 21215 }, { "epoch": 0.99, "learning_rate": 2.8107125583137006e-05, "loss": 0.152, "step": 21220 }, { "epoch": 0.99, "learning_rate": 2.8106657513300986e-05, "loss": 0.1005, "step": 21225 }, { "epoch": 0.99, "learning_rate": 2.8106189443464965e-05, "loss": 0.1959, "step": 21230 }, { "epoch": 0.99, "learning_rate": 2.810572137362895e-05, "loss": 0.1537, "step": 21235 }, { "epoch": 0.99, "learning_rate": 2.810525330379293e-05, "loss": 0.1304, "step": 21240 }, { "epoch": 0.99, "learning_rate": 2.810478523395691e-05, "loss": 0.4672, "step": 21245 }, { "epoch": 0.99, "learning_rate": 2.8104317164120885e-05, "loss": 0.4464, "step": 21250 }, { "epoch": 0.99, "learning_rate": 2.8103849094284868e-05, "loss": 0.1754, "step": 21255 }, { "epoch": 0.99, "learning_rate": 2.8103381024448848e-05, "loss": 0.0353, "step": 21260 }, { "epoch": 0.99, "learning_rate": 2.8102912954612828e-05, "loss": 0.1236, "step": 21265 }, { "epoch": 0.99, "learning_rate": 2.8102444884776807e-05, "loss": 0.0951, "step": 21270 }, { "epoch": 0.99, "learning_rate": 2.810197681494079e-05, "loss": 0.1122, "step": 21275 }, { "epoch": 0.99, "learning_rate": 2.810150874510477e-05, "loss": 0.1565, "step": 21280 }, { "epoch": 0.99, "learning_rate": 2.810104067526875e-05, "loss": 0.1991, "step": 21285 }, { "epoch": 0.99, "learning_rate": 2.8100572605432733e-05, "loss": 0.298, "step": 21290 }, { "epoch": 0.99, "learning_rate": 2.8100104535596713e-05, "loss": 0.3901, "step": 21295 }, { "epoch": 0.99, "learning_rate": 2.8099636465760693e-05, "loss": 0.5585, "step": 21300 }, { "epoch": 0.99, "learning_rate": 2.8099168395924673e-05, "loss": 0.2222, "step": 21305 }, { "epoch": 0.99, "learning_rate": 2.8098700326088653e-05, "loss": 0.0505, "step": 21310 }, { "epoch": 0.99, "learning_rate": 2.8098232256252633e-05, "loss": 0.0697, "step": 21315 }, { "epoch": 0.99, "learning_rate": 2.8097764186416612e-05, "loss": 0.0983, "step": 21320 }, { "epoch": 1.0, "learning_rate": 2.8097296116580592e-05, "loss": 0.1085, "step": 21325 }, { "epoch": 1.0, "learning_rate": 2.8096828046744575e-05, "loss": 0.0795, "step": 21330 }, { "epoch": 1.0, "learning_rate": 2.8096359976908555e-05, "loss": 0.1129, "step": 21335 }, { "epoch": 1.0, "learning_rate": 2.8095891907072535e-05, "loss": 0.186, "step": 21340 }, { "epoch": 1.0, "learning_rate": 2.8095423837236518e-05, "loss": 0.3137, "step": 21345 }, { "epoch": 1.0, "learning_rate": 2.8094955767400498e-05, "loss": 0.4092, "step": 21350 }, { "epoch": 1.0, "learning_rate": 2.8094487697564478e-05, "loss": 0.2431, "step": 21355 }, { "epoch": 1.0, "learning_rate": 2.8094019627728458e-05, "loss": 0.0776, "step": 21360 }, { "epoch": 1.0, "learning_rate": 2.809355155789244e-05, "loss": 0.0519, "step": 21365 }, { "epoch": 1.0, "learning_rate": 2.809308348805642e-05, "loss": 0.1305, "step": 21370 }, { "epoch": 1.0, "learning_rate": 2.8092615418220397e-05, "loss": 0.1244, "step": 21375 }, { "epoch": 1.0, "learning_rate": 2.8092147348384377e-05, "loss": 0.1205, "step": 21380 }, { "epoch": 1.0, "learning_rate": 2.809167927854836e-05, "loss": 0.2034, "step": 21385 }, { "epoch": 1.0, "learning_rate": 2.809121120871234e-05, "loss": 0.1759, "step": 21390 }, { "epoch": 1.0, "learning_rate": 2.809074313887632e-05, "loss": 0.2639, "step": 21395 }, { "epoch": 1.0, "learning_rate": 2.8090275069040303e-05, "loss": 0.3853, "step": 21400 }, { "epoch": 1.0, "learning_rate": 2.8089806999204283e-05, "loss": 0.1772, "step": 21405 }, { "epoch": 1.0, "learning_rate": 2.8089338929368263e-05, "loss": 0.0547, "step": 21410 }, { "epoch": 1.0, "learning_rate": 2.8088870859532242e-05, "loss": 0.1344, "step": 21415 }, { "epoch": 1.0, "learning_rate": 2.8088402789696226e-05, "loss": 0.1098, "step": 21420 }, { "epoch": 1.0, "learning_rate": 2.8087934719860205e-05, "loss": 0.2273, "step": 21425 }, { "epoch": 1.0, "learning_rate": 2.8087466650024185e-05, "loss": 0.2281, "step": 21430 }, { "epoch": 1.0, "eval_cer": 0.017324029687760277, "eval_loss": 0.044368449598550797, "eval_runtime": 417.3503, "eval_samples_per_second": 45.645, "eval_steps_per_second": 11.412, "eval_wer": 0.1453773151691413, "step": 21431 }, { "epoch": 1.0, "learning_rate": 2.8086998580188165e-05, "loss": 0.3569, "step": 21435 }, { "epoch": 1.0, "learning_rate": 2.8086530510352145e-05, "loss": 0.0547, "step": 21440 }, { "epoch": 1.0, "learning_rate": 2.8086062440516125e-05, "loss": 0.1129, "step": 21445 }, { "epoch": 1.0, "learning_rate": 2.8085594370680105e-05, "loss": 0.1289, "step": 21450 }, { "epoch": 1.0, "learning_rate": 2.8085126300844084e-05, "loss": 0.136, "step": 21455 }, { "epoch": 1.0, "learning_rate": 2.8084658231008068e-05, "loss": 0.1492, "step": 21460 }, { "epoch": 1.0, "learning_rate": 2.8084190161172047e-05, "loss": 0.1638, "step": 21465 }, { "epoch": 1.0, "learning_rate": 2.8083722091336027e-05, "loss": 0.2593, "step": 21470 }, { "epoch": 1.0, "learning_rate": 2.808325402150001e-05, "loss": 0.2652, "step": 21475 }, { "epoch": 1.0, "learning_rate": 2.808278595166399e-05, "loss": 0.349, "step": 21480 }, { "epoch": 1.0, "learning_rate": 2.808231788182797e-05, "loss": 0.2494, "step": 21485 }, { "epoch": 1.0, "learning_rate": 2.808184981199195e-05, "loss": 0.0287, "step": 21490 }, { "epoch": 1.0, "learning_rate": 2.8081381742155933e-05, "loss": 0.1174, "step": 21495 }, { "epoch": 1.0, "learning_rate": 2.808091367231991e-05, "loss": 0.084, "step": 21500 }, { "epoch": 1.0, "learning_rate": 2.808044560248389e-05, "loss": 0.1471, "step": 21505 }, { "epoch": 1.0, "learning_rate": 2.807997753264787e-05, "loss": 0.0932, "step": 21510 }, { "epoch": 1.0, "learning_rate": 2.8079509462811852e-05, "loss": 0.1504, "step": 21515 }, { "epoch": 1.0, "learning_rate": 2.8079041392975832e-05, "loss": 0.1676, "step": 21520 }, { "epoch": 1.0, "learning_rate": 2.8078573323139812e-05, "loss": 0.2544, "step": 21525 }, { "epoch": 1.0, "learning_rate": 2.8078105253303795e-05, "loss": 0.5146, "step": 21530 }, { "epoch": 1.0, "learning_rate": 2.8077637183467775e-05, "loss": 0.2322, "step": 21535 }, { "epoch": 1.01, "learning_rate": 2.8077169113631755e-05, "loss": 0.0566, "step": 21540 }, { "epoch": 1.01, "learning_rate": 2.8076701043795735e-05, "loss": 0.0943, "step": 21545 }, { "epoch": 1.01, "learning_rate": 2.8076232973959718e-05, "loss": 0.1057, "step": 21550 }, { "epoch": 1.01, "learning_rate": 2.8075764904123698e-05, "loss": 0.118, "step": 21555 }, { "epoch": 1.01, "learning_rate": 2.8075296834287677e-05, "loss": 0.1078, "step": 21560 }, { "epoch": 1.01, "learning_rate": 2.8074828764451654e-05, "loss": 0.1053, "step": 21565 }, { "epoch": 1.01, "learning_rate": 2.8074360694615637e-05, "loss": 0.189, "step": 21570 }, { "epoch": 1.01, "learning_rate": 2.8073892624779617e-05, "loss": 0.1968, "step": 21575 }, { "epoch": 1.01, "learning_rate": 2.8073424554943597e-05, "loss": 0.3459, "step": 21580 }, { "epoch": 1.01, "learning_rate": 2.807295648510758e-05, "loss": 0.3776, "step": 21585 }, { "epoch": 1.01, "learning_rate": 2.807248841527156e-05, "loss": 0.0604, "step": 21590 }, { "epoch": 1.01, "learning_rate": 2.807202034543554e-05, "loss": 0.0643, "step": 21595 }, { "epoch": 1.01, "learning_rate": 2.807155227559952e-05, "loss": 0.1255, "step": 21600 }, { "epoch": 1.01, "learning_rate": 2.8071084205763503e-05, "loss": 0.1175, "step": 21605 }, { "epoch": 1.01, "learning_rate": 2.8070616135927482e-05, "loss": 0.1656, "step": 21610 }, { "epoch": 1.01, "learning_rate": 2.8070148066091462e-05, "loss": 0.2271, "step": 21615 }, { "epoch": 1.01, "learning_rate": 2.8069679996255442e-05, "loss": 0.2579, "step": 21620 }, { "epoch": 1.01, "learning_rate": 2.8069211926419422e-05, "loss": 0.1913, "step": 21625 }, { "epoch": 1.01, "learning_rate": 2.8068743856583402e-05, "loss": 0.4454, "step": 21630 }, { "epoch": 1.01, "learning_rate": 2.806827578674738e-05, "loss": 0.3754, "step": 21635 }, { "epoch": 1.01, "learning_rate": 2.806780771691136e-05, "loss": 0.0788, "step": 21640 }, { "epoch": 1.01, "learning_rate": 2.8067339647075345e-05, "loss": 0.0852, "step": 21645 }, { "epoch": 1.01, "learning_rate": 2.8066871577239324e-05, "loss": 0.1291, "step": 21650 }, { "epoch": 1.01, "learning_rate": 2.8066403507403304e-05, "loss": 0.12, "step": 21655 }, { "epoch": 1.01, "learning_rate": 2.8065935437567287e-05, "loss": 0.169, "step": 21660 }, { "epoch": 1.01, "learning_rate": 2.8065467367731267e-05, "loss": 0.1362, "step": 21665 }, { "epoch": 1.01, "learning_rate": 2.8064999297895247e-05, "loss": 0.2041, "step": 21670 }, { "epoch": 1.01, "learning_rate": 2.8064531228059227e-05, "loss": 0.254, "step": 21675 }, { "epoch": 1.01, "learning_rate": 2.806406315822321e-05, "loss": 0.2922, "step": 21680 }, { "epoch": 1.01, "learning_rate": 2.806359508838719e-05, "loss": 0.3923, "step": 21685 }, { "epoch": 1.01, "learning_rate": 2.8063127018551166e-05, "loss": 0.0997, "step": 21690 }, { "epoch": 1.01, "learning_rate": 2.8062658948715146e-05, "loss": 0.076, "step": 21695 }, { "epoch": 1.01, "learning_rate": 2.806219087887913e-05, "loss": 0.1171, "step": 21700 }, { "epoch": 1.01, "learning_rate": 2.806172280904311e-05, "loss": 0.0969, "step": 21705 }, { "epoch": 1.01, "learning_rate": 2.806125473920709e-05, "loss": 0.089, "step": 21710 }, { "epoch": 1.01, "learning_rate": 2.8060786669371072e-05, "loss": 0.1608, "step": 21715 }, { "epoch": 1.01, "learning_rate": 2.8060318599535052e-05, "loss": 0.1848, "step": 21720 }, { "epoch": 1.01, "learning_rate": 2.8059850529699032e-05, "loss": 0.2496, "step": 21725 }, { "epoch": 1.01, "learning_rate": 2.805938245986301e-05, "loss": 0.3286, "step": 21730 }, { "epoch": 1.01, "learning_rate": 2.8058914390026995e-05, "loss": 0.3203, "step": 21735 }, { "epoch": 1.01, "learning_rate": 2.8058446320190975e-05, "loss": 0.0839, "step": 21740 }, { "epoch": 1.01, "learning_rate": 2.8057978250354954e-05, "loss": 0.0886, "step": 21745 }, { "epoch": 1.01, "learning_rate": 2.8057510180518934e-05, "loss": 0.1171, "step": 21750 }, { "epoch": 1.02, "learning_rate": 2.8057042110682914e-05, "loss": 0.0864, "step": 21755 }, { "epoch": 1.02, "learning_rate": 2.8056574040846894e-05, "loss": 0.1462, "step": 21760 }, { "epoch": 1.02, "learning_rate": 2.8056105971010874e-05, "loss": 0.2781, "step": 21765 }, { "epoch": 1.02, "learning_rate": 2.8055637901174857e-05, "loss": 0.155, "step": 21770 }, { "epoch": 1.02, "learning_rate": 2.8055169831338837e-05, "loss": 0.2439, "step": 21775 }, { "epoch": 1.02, "learning_rate": 2.8054701761502817e-05, "loss": 0.4164, "step": 21780 }, { "epoch": 1.02, "learning_rate": 2.8054233691666796e-05, "loss": 0.2868, "step": 21785 }, { "epoch": 1.02, "learning_rate": 2.805376562183078e-05, "loss": 0.0581, "step": 21790 }, { "epoch": 1.02, "learning_rate": 2.805329755199476e-05, "loss": 0.0677, "step": 21795 }, { "epoch": 1.02, "learning_rate": 2.805282948215874e-05, "loss": 0.0646, "step": 21800 }, { "epoch": 1.02, "learning_rate": 2.805236141232272e-05, "loss": 0.0902, "step": 21805 }, { "epoch": 1.02, "learning_rate": 2.8051893342486702e-05, "loss": 0.166, "step": 21810 }, { "epoch": 1.02, "learning_rate": 2.805142527265068e-05, "loss": 0.1674, "step": 21815 }, { "epoch": 1.02, "learning_rate": 2.805095720281466e-05, "loss": 0.183, "step": 21820 }, { "epoch": 1.02, "learning_rate": 2.805048913297864e-05, "loss": 0.2956, "step": 21825 }, { "epoch": 1.02, "learning_rate": 2.805002106314262e-05, "loss": 0.341, "step": 21830 }, { "epoch": 1.02, "learning_rate": 2.80495529933066e-05, "loss": 0.2537, "step": 21835 }, { "epoch": 1.02, "learning_rate": 2.804908492347058e-05, "loss": 0.0434, "step": 21840 }, { "epoch": 1.02, "learning_rate": 2.8048616853634564e-05, "loss": 0.0873, "step": 21845 }, { "epoch": 1.02, "learning_rate": 2.8048148783798544e-05, "loss": 0.135, "step": 21850 }, { "epoch": 1.02, "learning_rate": 2.8047680713962524e-05, "loss": 0.1317, "step": 21855 }, { "epoch": 1.02, "learning_rate": 2.8047212644126504e-05, "loss": 0.1786, "step": 21860 }, { "epoch": 1.02, "learning_rate": 2.8046744574290487e-05, "loss": 0.1707, "step": 21865 }, { "epoch": 1.02, "learning_rate": 2.8046276504454467e-05, "loss": 0.2086, "step": 21870 }, { "epoch": 1.02, "learning_rate": 2.8045808434618447e-05, "loss": 0.3222, "step": 21875 }, { "epoch": 1.02, "learning_rate": 2.8045340364782423e-05, "loss": 0.3771, "step": 21880 }, { "epoch": 1.02, "learning_rate": 2.8044872294946406e-05, "loss": 0.3114, "step": 21885 }, { "epoch": 1.02, "learning_rate": 2.8044404225110386e-05, "loss": 0.1046, "step": 21890 }, { "epoch": 1.02, "learning_rate": 2.8043936155274366e-05, "loss": 0.1055, "step": 21895 }, { "epoch": 1.02, "learning_rate": 2.804346808543835e-05, "loss": 0.1283, "step": 21900 }, { "epoch": 1.02, "learning_rate": 2.804300001560233e-05, "loss": 0.147, "step": 21905 }, { "epoch": 1.02, "learning_rate": 2.804253194576631e-05, "loss": 0.1537, "step": 21910 }, { "epoch": 1.02, "learning_rate": 2.804206387593029e-05, "loss": 0.1984, "step": 21915 }, { "epoch": 1.02, "learning_rate": 2.8041595806094272e-05, "loss": 0.2843, "step": 21920 }, { "epoch": 1.02, "learning_rate": 2.804112773625825e-05, "loss": 0.2247, "step": 21925 }, { "epoch": 1.02, "learning_rate": 2.804065966642223e-05, "loss": 0.3886, "step": 21930 }, { "epoch": 1.02, "learning_rate": 2.804019159658621e-05, "loss": 0.2921, "step": 21935 }, { "epoch": 1.02, "learning_rate": 2.8039723526750194e-05, "loss": 0.1001, "step": 21940 }, { "epoch": 1.02, "learning_rate": 2.803925545691417e-05, "loss": 0.0521, "step": 21945 }, { "epoch": 1.02, "learning_rate": 2.803878738707815e-05, "loss": 0.137, "step": 21950 }, { "epoch": 1.02, "learning_rate": 2.8038319317242134e-05, "loss": 0.1007, "step": 21955 }, { "epoch": 1.02, "learning_rate": 2.8037851247406114e-05, "loss": 0.2019, "step": 21960 }, { "epoch": 1.02, "learning_rate": 2.8037383177570094e-05, "loss": 0.1404, "step": 21965 }, { "epoch": 1.03, "learning_rate": 2.8036915107734073e-05, "loss": 0.2021, "step": 21970 }, { "epoch": 1.03, "learning_rate": 2.8036447037898057e-05, "loss": 0.1831, "step": 21975 }, { "epoch": 1.03, "learning_rate": 2.8035978968062036e-05, "loss": 0.3806, "step": 21980 }, { "epoch": 1.03, "learning_rate": 2.8035510898226016e-05, "loss": 0.2251, "step": 21985 }, { "epoch": 1.03, "learning_rate": 2.8035042828389996e-05, "loss": 0.0822, "step": 21990 }, { "epoch": 1.03, "learning_rate": 2.803457475855398e-05, "loss": 0.1029, "step": 21995 }, { "epoch": 1.03, "learning_rate": 2.803410668871796e-05, "loss": 0.1215, "step": 22000 }, { "epoch": 1.03, "learning_rate": 2.8033638618881935e-05, "loss": 0.1059, "step": 22005 }, { "epoch": 1.03, "learning_rate": 2.8033170549045915e-05, "loss": 0.1448, "step": 22010 }, { "epoch": 1.03, "learning_rate": 2.80327024792099e-05, "loss": 0.1495, "step": 22015 }, { "epoch": 1.03, "learning_rate": 2.8032234409373878e-05, "loss": 0.3297, "step": 22020 }, { "epoch": 1.03, "learning_rate": 2.8031766339537858e-05, "loss": 0.314, "step": 22025 }, { "epoch": 1.03, "learning_rate": 2.803129826970184e-05, "loss": 0.345, "step": 22030 }, { "epoch": 1.03, "learning_rate": 2.803083019986582e-05, "loss": 0.3261, "step": 22035 }, { "epoch": 1.03, "learning_rate": 2.80303621300298e-05, "loss": 0.0667, "step": 22040 }, { "epoch": 1.03, "learning_rate": 2.802989406019378e-05, "loss": 0.0971, "step": 22045 }, { "epoch": 1.03, "learning_rate": 2.8029425990357764e-05, "loss": 0.0779, "step": 22050 }, { "epoch": 1.03, "learning_rate": 2.8028957920521744e-05, "loss": 0.1409, "step": 22055 }, { "epoch": 1.03, "learning_rate": 2.8028489850685724e-05, "loss": 0.1598, "step": 22060 }, { "epoch": 1.03, "learning_rate": 2.8028021780849703e-05, "loss": 0.1167, "step": 22065 }, { "epoch": 1.03, "learning_rate": 2.8027553711013683e-05, "loss": 0.156, "step": 22070 }, { "epoch": 1.03, "learning_rate": 2.8027085641177663e-05, "loss": 0.2611, "step": 22075 }, { "epoch": 1.03, "learning_rate": 2.8026617571341643e-05, "loss": 0.4057, "step": 22080 }, { "epoch": 1.03, "learning_rate": 2.8026149501505626e-05, "loss": 0.2867, "step": 22085 }, { "epoch": 1.03, "learning_rate": 2.8025681431669606e-05, "loss": 0.0698, "step": 22090 }, { "epoch": 1.03, "learning_rate": 2.8025213361833586e-05, "loss": 0.0762, "step": 22095 }, { "epoch": 1.03, "learning_rate": 2.8024745291997566e-05, "loss": 0.1022, "step": 22100 }, { "epoch": 1.03, "learning_rate": 2.802427722216155e-05, "loss": 0.0939, "step": 22105 }, { "epoch": 1.03, "learning_rate": 2.802380915232553e-05, "loss": 0.1342, "step": 22110 }, { "epoch": 1.03, "learning_rate": 2.802334108248951e-05, "loss": 0.1678, "step": 22115 }, { "epoch": 1.03, "learning_rate": 2.8022873012653488e-05, "loss": 0.223, "step": 22120 }, { "epoch": 1.03, "learning_rate": 2.802240494281747e-05, "loss": 0.1388, "step": 22125 }, { "epoch": 1.03, "learning_rate": 2.802193687298145e-05, "loss": 0.4349, "step": 22130 }, { "epoch": 1.03, "learning_rate": 2.8021468803145428e-05, "loss": 0.3072, "step": 22135 }, { "epoch": 1.03, "learning_rate": 2.802100073330941e-05, "loss": 0.0539, "step": 22140 }, { "epoch": 1.03, "learning_rate": 2.802053266347339e-05, "loss": 0.0851, "step": 22145 }, { "epoch": 1.03, "learning_rate": 2.802006459363737e-05, "loss": 0.1224, "step": 22150 }, { "epoch": 1.03, "learning_rate": 2.801959652380135e-05, "loss": 0.0958, "step": 22155 }, { "epoch": 1.03, "learning_rate": 2.8019128453965334e-05, "loss": 0.1109, "step": 22160 }, { "epoch": 1.03, "learning_rate": 2.8018660384129313e-05, "loss": 0.1401, "step": 22165 }, { "epoch": 1.03, "learning_rate": 2.8018192314293293e-05, "loss": 0.161, "step": 22170 }, { "epoch": 1.03, "learning_rate": 2.8017724244457273e-05, "loss": 0.295, "step": 22175 }, { "epoch": 1.03, "learning_rate": 2.8017256174621256e-05, "loss": 0.3682, "step": 22180 }, { "epoch": 1.04, "learning_rate": 2.8016788104785236e-05, "loss": 0.3074, "step": 22185 }, { "epoch": 1.04, "learning_rate": 2.8016320034949216e-05, "loss": 0.1186, "step": 22190 }, { "epoch": 1.04, "learning_rate": 2.8015851965113196e-05, "loss": 0.0638, "step": 22195 }, { "epoch": 1.04, "learning_rate": 2.8015383895277175e-05, "loss": 0.1235, "step": 22200 }, { "epoch": 1.04, "learning_rate": 2.8014915825441155e-05, "loss": 0.0682, "step": 22205 }, { "epoch": 1.04, "learning_rate": 2.8014447755605135e-05, "loss": 0.086, "step": 22210 }, { "epoch": 1.04, "learning_rate": 2.8013979685769118e-05, "loss": 0.1951, "step": 22215 }, { "epoch": 1.04, "learning_rate": 2.8013511615933098e-05, "loss": 0.2108, "step": 22220 }, { "epoch": 1.04, "learning_rate": 2.8013043546097078e-05, "loss": 0.2215, "step": 22225 }, { "epoch": 1.04, "learning_rate": 2.8012575476261058e-05, "loss": 0.3945, "step": 22230 }, { "epoch": 1.04, "learning_rate": 2.801210740642504e-05, "loss": 0.3312, "step": 22235 }, { "epoch": 1.04, "learning_rate": 2.801163933658902e-05, "loss": 0.0904, "step": 22240 }, { "epoch": 1.04, "learning_rate": 2.8011171266753e-05, "loss": 0.0737, "step": 22245 }, { "epoch": 1.04, "learning_rate": 2.801070319691698e-05, "loss": 0.0878, "step": 22250 }, { "epoch": 1.04, "learning_rate": 2.8010235127080964e-05, "loss": 0.0806, "step": 22255 }, { "epoch": 1.04, "learning_rate": 2.800976705724494e-05, "loss": 0.1336, "step": 22260 }, { "epoch": 1.04, "learning_rate": 2.800929898740892e-05, "loss": 0.216, "step": 22265 }, { "epoch": 1.04, "learning_rate": 2.8008830917572903e-05, "loss": 0.2947, "step": 22270 }, { "epoch": 1.04, "learning_rate": 2.8008362847736883e-05, "loss": 0.2117, "step": 22275 }, { "epoch": 1.04, "learning_rate": 2.8007894777900863e-05, "loss": 0.3475, "step": 22280 }, { "epoch": 1.04, "learning_rate": 2.8007426708064843e-05, "loss": 0.2453, "step": 22285 }, { "epoch": 1.04, "learning_rate": 2.8006958638228826e-05, "loss": 0.1209, "step": 22290 }, { "epoch": 1.04, "learning_rate": 2.8006490568392806e-05, "loss": 0.1034, "step": 22295 }, { "epoch": 1.04, "learning_rate": 2.8006022498556785e-05, "loss": 0.0383, "step": 22300 }, { "epoch": 1.04, "learning_rate": 2.8005554428720765e-05, "loss": 0.1362, "step": 22305 }, { "epoch": 1.04, "learning_rate": 2.800508635888475e-05, "loss": 0.1696, "step": 22310 }, { "epoch": 1.04, "learning_rate": 2.8004618289048728e-05, "loss": 0.1563, "step": 22315 }, { "epoch": 1.04, "learning_rate": 2.8004150219212708e-05, "loss": 0.265, "step": 22320 }, { "epoch": 1.04, "learning_rate": 2.8003682149376688e-05, "loss": 0.2571, "step": 22325 }, { "epoch": 1.04, "learning_rate": 2.8003214079540668e-05, "loss": 0.3525, "step": 22330 }, { "epoch": 1.04, "learning_rate": 2.8002746009704647e-05, "loss": 0.2973, "step": 22335 }, { "epoch": 1.04, "learning_rate": 2.8002277939868627e-05, "loss": 0.0616, "step": 22340 }, { "epoch": 1.04, "learning_rate": 2.800180987003261e-05, "loss": 0.0927, "step": 22345 }, { "epoch": 1.04, "learning_rate": 2.800134180019659e-05, "loss": 0.0686, "step": 22350 }, { "epoch": 1.04, "learning_rate": 2.800087373036057e-05, "loss": 0.154, "step": 22355 }, { "epoch": 1.04, "learning_rate": 2.800040566052455e-05, "loss": 0.1362, "step": 22360 }, { "epoch": 1.04, "learning_rate": 2.7999937590688533e-05, "loss": 0.1925, "step": 22365 }, { "epoch": 1.04, "learning_rate": 2.7999469520852513e-05, "loss": 0.262, "step": 22370 }, { "epoch": 1.04, "learning_rate": 2.7999001451016493e-05, "loss": 0.2398, "step": 22375 }, { "epoch": 1.04, "learning_rate": 2.7998533381180476e-05, "loss": 0.414, "step": 22380 }, { "epoch": 1.04, "learning_rate": 2.7998065311344452e-05, "loss": 0.3237, "step": 22385 }, { "epoch": 1.04, "learning_rate": 2.7997597241508432e-05, "loss": 0.0841, "step": 22390 }, { "epoch": 1.04, "learning_rate": 2.7997129171672412e-05, "loss": 0.0513, "step": 22395 }, { "epoch": 1.05, "learning_rate": 2.7996661101836395e-05, "loss": 0.0424, "step": 22400 }, { "epoch": 1.05, "learning_rate": 2.7996193032000375e-05, "loss": 0.1236, "step": 22405 }, { "epoch": 1.05, "learning_rate": 2.7995724962164355e-05, "loss": 0.1358, "step": 22410 }, { "epoch": 1.05, "learning_rate": 2.7995256892328335e-05, "loss": 0.19, "step": 22415 }, { "epoch": 1.05, "learning_rate": 2.7994788822492318e-05, "loss": 0.1605, "step": 22420 }, { "epoch": 1.05, "learning_rate": 2.7994320752656298e-05, "loss": 0.1292, "step": 22425 }, { "epoch": 1.05, "learning_rate": 2.7993852682820278e-05, "loss": 0.3035, "step": 22430 }, { "epoch": 1.05, "learning_rate": 2.7993384612984257e-05, "loss": 0.3058, "step": 22435 }, { "epoch": 1.05, "learning_rate": 2.799291654314824e-05, "loss": 0.0367, "step": 22440 }, { "epoch": 1.05, "learning_rate": 2.799244847331222e-05, "loss": 0.0674, "step": 22445 }, { "epoch": 1.05, "learning_rate": 2.7991980403476197e-05, "loss": 0.1236, "step": 22450 }, { "epoch": 1.05, "learning_rate": 2.799151233364018e-05, "loss": 0.1277, "step": 22455 }, { "epoch": 1.05, "learning_rate": 2.799104426380416e-05, "loss": 0.1916, "step": 22460 }, { "epoch": 1.05, "learning_rate": 2.799057619396814e-05, "loss": 0.2015, "step": 22465 }, { "epoch": 1.05, "learning_rate": 2.799010812413212e-05, "loss": 0.3296, "step": 22470 }, { "epoch": 1.05, "learning_rate": 2.7989640054296103e-05, "loss": 0.2275, "step": 22475 }, { "epoch": 1.05, "learning_rate": 2.7989171984460083e-05, "loss": 0.3404, "step": 22480 }, { "epoch": 1.05, "learning_rate": 2.7988703914624062e-05, "loss": 0.3814, "step": 22485 }, { "epoch": 1.05, "learning_rate": 2.7988235844788042e-05, "loss": 0.065, "step": 22490 }, { "epoch": 1.05, "learning_rate": 2.7987767774952025e-05, "loss": 0.0983, "step": 22495 }, { "epoch": 1.05, "learning_rate": 2.7987299705116005e-05, "loss": 0.0914, "step": 22500 }, { "epoch": 1.05, "learning_rate": 2.7986831635279985e-05, "loss": 0.0641, "step": 22505 }, { "epoch": 1.05, "learning_rate": 2.7986363565443965e-05, "loss": 0.1114, "step": 22510 }, { "epoch": 1.05, "learning_rate": 2.7985895495607945e-05, "loss": 0.1769, "step": 22515 }, { "epoch": 1.05, "learning_rate": 2.7985427425771924e-05, "loss": 0.2426, "step": 22520 }, { "epoch": 1.05, "learning_rate": 2.7984959355935904e-05, "loss": 0.3089, "step": 22525 }, { "epoch": 1.05, "learning_rate": 2.7984491286099887e-05, "loss": 0.428, "step": 22530 }, { "epoch": 1.05, "learning_rate": 2.7984023216263867e-05, "loss": 0.3321, "step": 22535 }, { "epoch": 1.05, "learning_rate": 2.7983555146427847e-05, "loss": 0.0933, "step": 22540 }, { "epoch": 1.05, "learning_rate": 2.7983087076591827e-05, "loss": 0.0801, "step": 22545 }, { "epoch": 1.05, "learning_rate": 2.798261900675581e-05, "loss": 0.0415, "step": 22550 }, { "epoch": 1.05, "learning_rate": 2.798215093691979e-05, "loss": 0.0557, "step": 22555 }, { "epoch": 1.05, "learning_rate": 2.798168286708377e-05, "loss": 0.0842, "step": 22560 }, { "epoch": 1.05, "learning_rate": 2.7981214797247753e-05, "loss": 0.1134, "step": 22565 }, { "epoch": 1.05, "learning_rate": 2.7980746727411733e-05, "loss": 0.1539, "step": 22570 }, { "epoch": 1.05, "learning_rate": 2.798027865757571e-05, "loss": 0.2321, "step": 22575 }, { "epoch": 1.05, "learning_rate": 2.797981058773969e-05, "loss": 0.352, "step": 22580 }, { "epoch": 1.05, "learning_rate": 2.7979342517903672e-05, "loss": 0.2247, "step": 22585 }, { "epoch": 1.05, "learning_rate": 2.7978874448067652e-05, "loss": 0.0818, "step": 22590 }, { "epoch": 1.05, "learning_rate": 2.7978406378231632e-05, "loss": 0.1145, "step": 22595 }, { "epoch": 1.05, "learning_rate": 2.7977938308395612e-05, "loss": 0.12, "step": 22600 }, { "epoch": 1.05, "learning_rate": 2.7977470238559595e-05, "loss": 0.0897, "step": 22605 }, { "epoch": 1.06, "learning_rate": 2.7977002168723575e-05, "loss": 0.1287, "step": 22610 }, { "epoch": 1.06, "learning_rate": 2.7976534098887555e-05, "loss": 0.2098, "step": 22615 }, { "epoch": 1.06, "learning_rate": 2.7976066029051534e-05, "loss": 0.2627, "step": 22620 }, { "epoch": 1.06, "learning_rate": 2.7975597959215518e-05, "loss": 0.3518, "step": 22625 }, { "epoch": 1.06, "learning_rate": 2.7975129889379497e-05, "loss": 0.4095, "step": 22630 }, { "epoch": 1.06, "learning_rate": 2.7974661819543477e-05, "loss": 0.4317, "step": 22635 }, { "epoch": 1.06, "learning_rate": 2.7974193749707457e-05, "loss": 0.0738, "step": 22640 }, { "epoch": 1.06, "learning_rate": 2.7973725679871437e-05, "loss": 0.1177, "step": 22645 }, { "epoch": 1.06, "learning_rate": 2.7973257610035417e-05, "loss": 0.1492, "step": 22650 }, { "epoch": 1.06, "learning_rate": 2.7972789540199396e-05, "loss": 0.1138, "step": 22655 }, { "epoch": 1.06, "learning_rate": 2.797232147036338e-05, "loss": 0.0826, "step": 22660 }, { "epoch": 1.06, "learning_rate": 2.797185340052736e-05, "loss": 0.2128, "step": 22665 }, { "epoch": 1.06, "learning_rate": 2.797138533069134e-05, "loss": 0.1339, "step": 22670 }, { "epoch": 1.06, "learning_rate": 2.797091726085532e-05, "loss": 0.181, "step": 22675 }, { "epoch": 1.06, "learning_rate": 2.7970449191019302e-05, "loss": 0.4068, "step": 22680 }, { "epoch": 1.06, "learning_rate": 2.7969981121183282e-05, "loss": 0.2568, "step": 22685 }, { "epoch": 1.06, "learning_rate": 2.7969513051347262e-05, "loss": 0.0896, "step": 22690 }, { "epoch": 1.06, "learning_rate": 2.7969044981511245e-05, "loss": 0.0615, "step": 22695 }, { "epoch": 1.06, "learning_rate": 2.796857691167522e-05, "loss": 0.0534, "step": 22700 }, { "epoch": 1.06, "learning_rate": 2.79681088418392e-05, "loss": 0.1399, "step": 22705 }, { "epoch": 1.06, "learning_rate": 2.796764077200318e-05, "loss": 0.1286, "step": 22710 }, { "epoch": 1.06, "learning_rate": 2.7967172702167164e-05, "loss": 0.1565, "step": 22715 }, { "epoch": 1.06, "learning_rate": 2.7966704632331144e-05, "loss": 0.2351, "step": 22720 }, { "epoch": 1.06, "learning_rate": 2.7966236562495124e-05, "loss": 0.2478, "step": 22725 }, { "epoch": 1.06, "learning_rate": 2.7965768492659104e-05, "loss": 0.346, "step": 22730 }, { "epoch": 1.06, "learning_rate": 2.7965300422823087e-05, "loss": 0.308, "step": 22735 }, { "epoch": 1.06, "learning_rate": 2.7964832352987067e-05, "loss": 0.0745, "step": 22740 }, { "epoch": 1.06, "learning_rate": 2.7964364283151047e-05, "loss": 0.0747, "step": 22745 }, { "epoch": 1.06, "learning_rate": 2.796389621331503e-05, "loss": 0.094, "step": 22750 }, { "epoch": 1.06, "learning_rate": 2.796342814347901e-05, "loss": 0.1546, "step": 22755 }, { "epoch": 1.06, "learning_rate": 2.796296007364299e-05, "loss": 0.1048, "step": 22760 }, { "epoch": 1.06, "learning_rate": 2.7962492003806966e-05, "loss": 0.0816, "step": 22765 }, { "epoch": 1.06, "learning_rate": 2.796202393397095e-05, "loss": 0.1823, "step": 22770 }, { "epoch": 1.06, "learning_rate": 2.796155586413493e-05, "loss": 0.1807, "step": 22775 }, { "epoch": 1.06, "learning_rate": 2.796108779429891e-05, "loss": 0.2225, "step": 22780 }, { "epoch": 1.06, "learning_rate": 2.796061972446289e-05, "loss": 0.451, "step": 22785 }, { "epoch": 1.06, "learning_rate": 2.7960151654626872e-05, "loss": 0.0427, "step": 22790 }, { "epoch": 1.06, "learning_rate": 2.795968358479085e-05, "loss": 0.1386, "step": 22795 }, { "epoch": 1.06, "learning_rate": 2.795921551495483e-05, "loss": 0.1519, "step": 22800 }, { "epoch": 1.06, "learning_rate": 2.7958747445118815e-05, "loss": 0.1064, "step": 22805 }, { "epoch": 1.06, "learning_rate": 2.7958279375282795e-05, "loss": 0.1201, "step": 22810 }, { "epoch": 1.06, "learning_rate": 2.7957811305446774e-05, "loss": 0.158, "step": 22815 }, { "epoch": 1.06, "learning_rate": 2.7957343235610754e-05, "loss": 0.2632, "step": 22820 }, { "epoch": 1.07, "learning_rate": 2.7956875165774737e-05, "loss": 0.2432, "step": 22825 }, { "epoch": 1.07, "learning_rate": 2.7956407095938714e-05, "loss": 0.4555, "step": 22830 }, { "epoch": 1.07, "learning_rate": 2.7955939026102694e-05, "loss": 0.3108, "step": 22835 }, { "epoch": 1.07, "learning_rate": 2.7955470956266673e-05, "loss": 0.0548, "step": 22840 }, { "epoch": 1.07, "learning_rate": 2.7955002886430657e-05, "loss": 0.1005, "step": 22845 }, { "epoch": 1.07, "learning_rate": 2.7954534816594636e-05, "loss": 0.063, "step": 22850 }, { "epoch": 1.07, "learning_rate": 2.7954066746758616e-05, "loss": 0.1485, "step": 22855 }, { "epoch": 1.07, "learning_rate": 2.7953598676922596e-05, "loss": 0.1224, "step": 22860 }, { "epoch": 1.07, "learning_rate": 2.795313060708658e-05, "loss": 0.1332, "step": 22865 }, { "epoch": 1.07, "learning_rate": 2.795266253725056e-05, "loss": 0.1694, "step": 22870 }, { "epoch": 1.07, "learning_rate": 2.795219446741454e-05, "loss": 0.2077, "step": 22875 }, { "epoch": 1.07, "learning_rate": 2.7951726397578522e-05, "loss": 0.4646, "step": 22880 }, { "epoch": 1.07, "learning_rate": 2.7951258327742502e-05, "loss": 0.2959, "step": 22885 }, { "epoch": 1.07, "learning_rate": 2.795079025790648e-05, "loss": 0.1028, "step": 22890 }, { "epoch": 1.07, "learning_rate": 2.7950322188070458e-05, "loss": 0.1081, "step": 22895 }, { "epoch": 1.07, "learning_rate": 2.794985411823444e-05, "loss": 0.079, "step": 22900 }, { "epoch": 1.07, "learning_rate": 2.794938604839842e-05, "loss": 0.0909, "step": 22905 }, { "epoch": 1.07, "learning_rate": 2.79489179785624e-05, "loss": 0.1335, "step": 22910 }, { "epoch": 1.07, "learning_rate": 2.794844990872638e-05, "loss": 0.1752, "step": 22915 }, { "epoch": 1.07, "learning_rate": 2.7947981838890364e-05, "loss": 0.1535, "step": 22920 }, { "epoch": 1.07, "learning_rate": 2.7947513769054344e-05, "loss": 0.2721, "step": 22925 }, { "epoch": 1.07, "learning_rate": 2.7947045699218324e-05, "loss": 0.3652, "step": 22930 }, { "epoch": 1.07, "learning_rate": 2.7946577629382307e-05, "loss": 0.3648, "step": 22935 }, { "epoch": 1.07, "learning_rate": 2.7946109559546287e-05, "loss": 0.0835, "step": 22940 }, { "epoch": 1.07, "learning_rate": 2.7945641489710267e-05, "loss": 0.0997, "step": 22945 }, { "epoch": 1.07, "learning_rate": 2.7945173419874246e-05, "loss": 0.0862, "step": 22950 }, { "epoch": 1.07, "learning_rate": 2.7944705350038226e-05, "loss": 0.1361, "step": 22955 }, { "epoch": 1.07, "learning_rate": 2.7944237280202206e-05, "loss": 0.1504, "step": 22960 }, { "epoch": 1.07, "learning_rate": 2.7943769210366186e-05, "loss": 0.1236, "step": 22965 }, { "epoch": 1.07, "learning_rate": 2.7943301140530166e-05, "loss": 0.1653, "step": 22970 }, { "epoch": 1.07, "learning_rate": 2.794283307069415e-05, "loss": 0.2567, "step": 22975 }, { "epoch": 1.07, "learning_rate": 2.794236500085813e-05, "loss": 0.4976, "step": 22980 }, { "epoch": 1.07, "learning_rate": 2.794189693102211e-05, "loss": 0.3687, "step": 22985 }, { "epoch": 1.07, "learning_rate": 2.794142886118609e-05, "loss": 0.0721, "step": 22990 }, { "epoch": 1.07, "learning_rate": 2.794096079135007e-05, "loss": 0.0776, "step": 22995 }, { "epoch": 1.07, "learning_rate": 2.794049272151405e-05, "loss": 0.0999, "step": 23000 }, { "epoch": 1.07, "learning_rate": 2.794002465167803e-05, "loss": 0.0748, "step": 23005 }, { "epoch": 1.07, "learning_rate": 2.7939556581842014e-05, "loss": 0.1476, "step": 23010 }, { "epoch": 1.07, "learning_rate": 2.793908851200599e-05, "loss": 0.1064, "step": 23015 }, { "epoch": 1.07, "learning_rate": 2.793862044216997e-05, "loss": 0.2505, "step": 23020 }, { "epoch": 1.07, "learning_rate": 2.793815237233395e-05, "loss": 0.2573, "step": 23025 }, { "epoch": 1.07, "learning_rate": 2.7937684302497934e-05, "loss": 0.1895, "step": 23030 }, { "epoch": 1.07, "learning_rate": 2.7937216232661913e-05, "loss": 0.204, "step": 23035 }, { "epoch": 1.08, "learning_rate": 2.7936748162825893e-05, "loss": 0.1203, "step": 23040 }, { "epoch": 1.08, "learning_rate": 2.7936280092989873e-05, "loss": 0.0748, "step": 23045 }, { "epoch": 1.08, "learning_rate": 2.7935812023153856e-05, "loss": 0.0602, "step": 23050 }, { "epoch": 1.08, "learning_rate": 2.7935343953317836e-05, "loss": 0.1381, "step": 23055 }, { "epoch": 1.08, "learning_rate": 2.7934875883481816e-05, "loss": 0.133, "step": 23060 }, { "epoch": 1.08, "learning_rate": 2.79344078136458e-05, "loss": 0.1485, "step": 23065 }, { "epoch": 1.08, "learning_rate": 2.793393974380978e-05, "loss": 0.167, "step": 23070 }, { "epoch": 1.08, "learning_rate": 2.793347167397376e-05, "loss": 0.3623, "step": 23075 }, { "epoch": 1.08, "learning_rate": 2.7933003604137735e-05, "loss": 0.5458, "step": 23080 }, { "epoch": 1.08, "learning_rate": 2.793253553430172e-05, "loss": 0.3545, "step": 23085 }, { "epoch": 1.08, "learning_rate": 2.7932067464465698e-05, "loss": 0.0485, "step": 23090 }, { "epoch": 1.08, "learning_rate": 2.7931599394629678e-05, "loss": 0.0781, "step": 23095 }, { "epoch": 1.08, "learning_rate": 2.7931131324793658e-05, "loss": 0.1369, "step": 23100 }, { "epoch": 1.08, "learning_rate": 2.793066325495764e-05, "loss": 0.0773, "step": 23105 }, { "epoch": 1.08, "learning_rate": 2.793019518512162e-05, "loss": 0.1182, "step": 23110 }, { "epoch": 1.08, "learning_rate": 2.79297271152856e-05, "loss": 0.1371, "step": 23115 }, { "epoch": 1.08, "learning_rate": 2.7929259045449584e-05, "loss": 0.1681, "step": 23120 }, { "epoch": 1.08, "learning_rate": 2.7928790975613564e-05, "loss": 0.2092, "step": 23125 }, { "epoch": 1.08, "learning_rate": 2.7928322905777544e-05, "loss": 0.3097, "step": 23130 }, { "epoch": 1.08, "learning_rate": 2.7927854835941523e-05, "loss": 0.2789, "step": 23135 }, { "epoch": 1.08, "learning_rate": 2.7927386766105507e-05, "loss": 0.0853, "step": 23140 }, { "epoch": 1.08, "learning_rate": 2.7926918696269483e-05, "loss": 0.1227, "step": 23145 }, { "epoch": 1.08, "learning_rate": 2.7926450626433463e-05, "loss": 0.1103, "step": 23150 }, { "epoch": 1.08, "learning_rate": 2.7925982556597443e-05, "loss": 0.1223, "step": 23155 }, { "epoch": 1.08, "learning_rate": 2.7925514486761426e-05, "loss": 0.1283, "step": 23160 }, { "epoch": 1.08, "learning_rate": 2.7925046416925406e-05, "loss": 0.1876, "step": 23165 }, { "epoch": 1.08, "learning_rate": 2.7924578347089385e-05, "loss": 0.2919, "step": 23170 }, { "epoch": 1.08, "learning_rate": 2.792411027725337e-05, "loss": 0.2611, "step": 23175 }, { "epoch": 1.08, "learning_rate": 2.792364220741735e-05, "loss": 0.3863, "step": 23180 }, { "epoch": 1.08, "learning_rate": 2.7923174137581328e-05, "loss": 0.2966, "step": 23185 }, { "epoch": 1.08, "learning_rate": 2.7922706067745308e-05, "loss": 0.0797, "step": 23190 }, { "epoch": 1.08, "learning_rate": 2.792223799790929e-05, "loss": 0.0962, "step": 23195 }, { "epoch": 1.08, "learning_rate": 2.792176992807327e-05, "loss": 0.0764, "step": 23200 }, { "epoch": 1.08, "learning_rate": 2.7921301858237248e-05, "loss": 0.1265, "step": 23205 }, { "epoch": 1.08, "learning_rate": 2.7920833788401227e-05, "loss": 0.1923, "step": 23210 }, { "epoch": 1.08, "learning_rate": 2.792036571856521e-05, "loss": 0.1019, "step": 23215 }, { "epoch": 1.08, "learning_rate": 2.791989764872919e-05, "loss": 0.1793, "step": 23220 }, { "epoch": 1.08, "learning_rate": 2.791942957889317e-05, "loss": 0.2424, "step": 23225 }, { "epoch": 1.08, "learning_rate": 2.791896150905715e-05, "loss": 0.3228, "step": 23230 }, { "epoch": 1.08, "learning_rate": 2.7918493439221133e-05, "loss": 0.3051, "step": 23235 }, { "epoch": 1.08, "learning_rate": 2.7918025369385113e-05, "loss": 0.0913, "step": 23240 }, { "epoch": 1.08, "learning_rate": 2.7917557299549093e-05, "loss": 0.1234, "step": 23245 }, { "epoch": 1.08, "learning_rate": 2.7917089229713076e-05, "loss": 0.0642, "step": 23250 }, { "epoch": 1.09, "learning_rate": 2.7916621159877056e-05, "loss": 0.177, "step": 23255 }, { "epoch": 1.09, "learning_rate": 2.7916153090041036e-05, "loss": 0.1408, "step": 23260 }, { "epoch": 1.09, "learning_rate": 2.7915685020205016e-05, "loss": 0.1713, "step": 23265 }, { "epoch": 1.09, "learning_rate": 2.7915216950368995e-05, "loss": 0.1815, "step": 23270 }, { "epoch": 1.09, "learning_rate": 2.7914748880532975e-05, "loss": 0.2341, "step": 23275 }, { "epoch": 1.09, "learning_rate": 2.7914280810696955e-05, "loss": 0.2827, "step": 23280 }, { "epoch": 1.09, "learning_rate": 2.7913812740860935e-05, "loss": 0.3039, "step": 23285 }, { "epoch": 1.09, "learning_rate": 2.7913344671024918e-05, "loss": 0.1035, "step": 23290 }, { "epoch": 1.09, "learning_rate": 2.7912876601188898e-05, "loss": 0.0578, "step": 23295 }, { "epoch": 1.09, "learning_rate": 2.7912408531352878e-05, "loss": 0.0686, "step": 23300 }, { "epoch": 1.09, "learning_rate": 2.791194046151686e-05, "loss": 0.1239, "step": 23305 }, { "epoch": 1.09, "learning_rate": 2.791147239168084e-05, "loss": 0.1145, "step": 23310 }, { "epoch": 1.09, "learning_rate": 2.791100432184482e-05, "loss": 0.149, "step": 23315 }, { "epoch": 1.09, "learning_rate": 2.79105362520088e-05, "loss": 0.2288, "step": 23320 }, { "epoch": 1.09, "learning_rate": 2.7910068182172784e-05, "loss": 0.2054, "step": 23325 }, { "epoch": 1.09, "learning_rate": 2.7909600112336763e-05, "loss": 0.4812, "step": 23330 }, { "epoch": 1.09, "learning_rate": 2.790913204250074e-05, "loss": 0.3454, "step": 23335 }, { "epoch": 1.09, "learning_rate": 2.790866397266472e-05, "loss": 0.0272, "step": 23340 }, { "epoch": 1.09, "learning_rate": 2.7908195902828703e-05, "loss": 0.12, "step": 23345 }, { "epoch": 1.09, "learning_rate": 2.7907727832992683e-05, "loss": 0.0455, "step": 23350 }, { "epoch": 1.09, "learning_rate": 2.7907259763156662e-05, "loss": 0.0957, "step": 23355 }, { "epoch": 1.09, "learning_rate": 2.7906791693320646e-05, "loss": 0.1513, "step": 23360 }, { "epoch": 1.09, "learning_rate": 2.7906323623484625e-05, "loss": 0.2252, "step": 23365 }, { "epoch": 1.09, "learning_rate": 2.7905855553648605e-05, "loss": 0.2206, "step": 23370 }, { "epoch": 1.09, "learning_rate": 2.7905387483812585e-05, "loss": 0.2257, "step": 23375 }, { "epoch": 1.09, "learning_rate": 2.7904919413976568e-05, "loss": 0.3453, "step": 23380 }, { "epoch": 1.09, "learning_rate": 2.7904451344140548e-05, "loss": 0.2708, "step": 23385 }, { "epoch": 1.09, "learning_rate": 2.7903983274304528e-05, "loss": 0.1018, "step": 23390 }, { "epoch": 1.09, "learning_rate": 2.7903515204468504e-05, "loss": 0.0812, "step": 23395 }, { "epoch": 1.09, "learning_rate": 2.7903047134632488e-05, "loss": 0.1129, "step": 23400 }, { "epoch": 1.09, "learning_rate": 2.7902579064796467e-05, "loss": 0.0758, "step": 23405 }, { "epoch": 1.09, "learning_rate": 2.7902110994960447e-05, "loss": 0.1734, "step": 23410 }, { "epoch": 1.09, "learning_rate": 2.790164292512443e-05, "loss": 0.1572, "step": 23415 }, { "epoch": 1.09, "learning_rate": 2.790117485528841e-05, "loss": 0.2247, "step": 23420 }, { "epoch": 1.09, "learning_rate": 2.790070678545239e-05, "loss": 0.3071, "step": 23425 }, { "epoch": 1.09, "learning_rate": 2.790023871561637e-05, "loss": 0.292, "step": 23430 }, { "epoch": 1.09, "learning_rate": 2.7899770645780353e-05, "loss": 0.2463, "step": 23435 }, { "epoch": 1.09, "learning_rate": 2.7899302575944333e-05, "loss": 0.0202, "step": 23440 }, { "epoch": 1.09, "learning_rate": 2.7898834506108313e-05, "loss": 0.1184, "step": 23445 }, { "epoch": 1.09, "learning_rate": 2.7898366436272293e-05, "loss": 0.0944, "step": 23450 }, { "epoch": 1.09, "learning_rate": 2.7897898366436276e-05, "loss": 0.176, "step": 23455 }, { "epoch": 1.09, "learning_rate": 2.7897430296600252e-05, "loss": 0.2143, "step": 23460 }, { "epoch": 1.09, "learning_rate": 2.7896962226764232e-05, "loss": 0.1107, "step": 23465 }, { "epoch": 1.1, "learning_rate": 2.7896494156928212e-05, "loss": 0.1465, "step": 23470 }, { "epoch": 1.1, "learning_rate": 2.7896026087092195e-05, "loss": 0.332, "step": 23475 }, { "epoch": 1.1, "learning_rate": 2.7895558017256175e-05, "loss": 0.455, "step": 23480 }, { "epoch": 1.1, "learning_rate": 2.7895089947420155e-05, "loss": 0.3153, "step": 23485 }, { "epoch": 1.1, "learning_rate": 2.7894621877584138e-05, "loss": 0.0421, "step": 23490 }, { "epoch": 1.1, "learning_rate": 2.7894153807748118e-05, "loss": 0.1043, "step": 23495 }, { "epoch": 1.1, "learning_rate": 2.7893685737912097e-05, "loss": 0.1151, "step": 23500 }, { "epoch": 1.1, "learning_rate": 2.7893217668076077e-05, "loss": 0.1384, "step": 23505 }, { "epoch": 1.1, "learning_rate": 2.789274959824006e-05, "loss": 0.1662, "step": 23510 }, { "epoch": 1.1, "learning_rate": 2.789228152840404e-05, "loss": 0.1708, "step": 23515 }, { "epoch": 1.1, "learning_rate": 2.789181345856802e-05, "loss": 0.1553, "step": 23520 }, { "epoch": 1.1, "learning_rate": 2.7891345388731997e-05, "loss": 0.1907, "step": 23525 }, { "epoch": 1.1, "learning_rate": 2.789087731889598e-05, "loss": 0.3254, "step": 23530 }, { "epoch": 1.1, "learning_rate": 2.789040924905996e-05, "loss": 0.3018, "step": 23535 }, { "epoch": 1.1, "learning_rate": 2.788994117922394e-05, "loss": 0.0582, "step": 23540 }, { "epoch": 1.1, "learning_rate": 2.7889473109387923e-05, "loss": 0.0943, "step": 23545 }, { "epoch": 1.1, "learning_rate": 2.7889005039551902e-05, "loss": 0.0854, "step": 23550 }, { "epoch": 1.1, "learning_rate": 2.7888536969715882e-05, "loss": 0.086, "step": 23555 }, { "epoch": 1.1, "learning_rate": 2.7888068899879862e-05, "loss": 0.0775, "step": 23560 }, { "epoch": 1.1, "learning_rate": 2.7887600830043845e-05, "loss": 0.1291, "step": 23565 }, { "epoch": 1.1, "learning_rate": 2.7887132760207825e-05, "loss": 0.2555, "step": 23570 }, { "epoch": 1.1, "learning_rate": 2.7886664690371805e-05, "loss": 0.2474, "step": 23575 }, { "epoch": 1.1, "learning_rate": 2.7886196620535785e-05, "loss": 0.2907, "step": 23580 }, { "epoch": 1.1, "learning_rate": 2.7885728550699765e-05, "loss": 0.219, "step": 23585 }, { "epoch": 1.1, "learning_rate": 2.7885260480863744e-05, "loss": 0.026, "step": 23590 }, { "epoch": 1.1, "learning_rate": 2.7884792411027724e-05, "loss": 0.0669, "step": 23595 }, { "epoch": 1.1, "learning_rate": 2.7884324341191707e-05, "loss": 0.0977, "step": 23600 }, { "epoch": 1.1, "learning_rate": 2.7883856271355687e-05, "loss": 0.0636, "step": 23605 }, { "epoch": 1.1, "learning_rate": 2.7883388201519667e-05, "loss": 0.1187, "step": 23610 }, { "epoch": 1.1, "learning_rate": 2.7882920131683647e-05, "loss": 0.1748, "step": 23615 }, { "epoch": 1.1, "learning_rate": 2.788245206184763e-05, "loss": 0.1696, "step": 23620 }, { "epoch": 1.1, "learning_rate": 2.788198399201161e-05, "loss": 0.1567, "step": 23625 }, { "epoch": 1.1, "learning_rate": 2.788151592217559e-05, "loss": 0.5083, "step": 23630 }, { "epoch": 1.1, "learning_rate": 2.788104785233957e-05, "loss": 0.2237, "step": 23635 }, { "epoch": 1.1, "learning_rate": 2.7880579782503553e-05, "loss": 0.0791, "step": 23640 }, { "epoch": 1.1, "learning_rate": 2.7880111712667533e-05, "loss": 0.0885, "step": 23645 }, { "epoch": 1.1, "learning_rate": 2.787964364283151e-05, "loss": 0.066, "step": 23650 }, { "epoch": 1.1, "learning_rate": 2.787917557299549e-05, "loss": 0.1573, "step": 23655 }, { "epoch": 1.1, "learning_rate": 2.7878707503159472e-05, "loss": 0.1412, "step": 23660 }, { "epoch": 1.1, "learning_rate": 2.7878239433323452e-05, "loss": 0.2119, "step": 23665 }, { "epoch": 1.1, "learning_rate": 2.787777136348743e-05, "loss": 0.1254, "step": 23670 }, { "epoch": 1.1, "learning_rate": 2.7877303293651415e-05, "loss": 0.2836, "step": 23675 }, { "epoch": 1.1, "learning_rate": 2.7876835223815395e-05, "loss": 0.3298, "step": 23680 }, { "epoch": 1.11, "learning_rate": 2.7876367153979374e-05, "loss": 0.284, "step": 23685 }, { "epoch": 1.11, "learning_rate": 2.7875899084143354e-05, "loss": 0.0646, "step": 23690 }, { "epoch": 1.11, "learning_rate": 2.7875431014307337e-05, "loss": 0.087, "step": 23695 }, { "epoch": 1.11, "learning_rate": 2.7874962944471317e-05, "loss": 0.0796, "step": 23700 }, { "epoch": 1.11, "learning_rate": 2.7874494874635297e-05, "loss": 0.0773, "step": 23705 }, { "epoch": 1.11, "learning_rate": 2.7874026804799277e-05, "loss": 0.1057, "step": 23710 }, { "epoch": 1.11, "learning_rate": 2.7873558734963257e-05, "loss": 0.1404, "step": 23715 }, { "epoch": 1.11, "learning_rate": 2.7873090665127237e-05, "loss": 0.1837, "step": 23720 }, { "epoch": 1.11, "learning_rate": 2.7872622595291216e-05, "loss": 0.1783, "step": 23725 }, { "epoch": 1.11, "learning_rate": 2.78721545254552e-05, "loss": 0.4584, "step": 23730 }, { "epoch": 1.11, "learning_rate": 2.787168645561918e-05, "loss": 0.367, "step": 23735 }, { "epoch": 1.11, "learning_rate": 2.787121838578316e-05, "loss": 0.0392, "step": 23740 }, { "epoch": 1.11, "learning_rate": 2.787075031594714e-05, "loss": 0.0638, "step": 23745 }, { "epoch": 1.11, "learning_rate": 2.7870282246111122e-05, "loss": 0.1059, "step": 23750 }, { "epoch": 1.11, "learning_rate": 2.7869814176275102e-05, "loss": 0.1247, "step": 23755 }, { "epoch": 1.11, "learning_rate": 2.7869346106439082e-05, "loss": 0.0934, "step": 23760 }, { "epoch": 1.11, "learning_rate": 2.786887803660306e-05, "loss": 0.1016, "step": 23765 }, { "epoch": 1.11, "learning_rate": 2.7868409966767045e-05, "loss": 0.3392, "step": 23770 }, { "epoch": 1.11, "learning_rate": 2.786794189693102e-05, "loss": 0.205, "step": 23775 }, { "epoch": 1.11, "learning_rate": 2.7867473827095e-05, "loss": 0.4122, "step": 23780 }, { "epoch": 1.11, "learning_rate": 2.7867005757258984e-05, "loss": 0.2699, "step": 23785 }, { "epoch": 1.11, "learning_rate": 2.7866537687422964e-05, "loss": 0.0749, "step": 23790 }, { "epoch": 1.11, "learning_rate": 2.7866069617586944e-05, "loss": 0.0608, "step": 23795 }, { "epoch": 1.11, "learning_rate": 2.7865601547750924e-05, "loss": 0.1115, "step": 23800 }, { "epoch": 1.11, "learning_rate": 2.7865133477914907e-05, "loss": 0.0917, "step": 23805 }, { "epoch": 1.11, "learning_rate": 2.7864665408078887e-05, "loss": 0.1316, "step": 23810 }, { "epoch": 1.11, "learning_rate": 2.7864197338242867e-05, "loss": 0.0842, "step": 23815 }, { "epoch": 1.11, "learning_rate": 2.7863729268406846e-05, "loss": 0.1673, "step": 23820 }, { "epoch": 1.11, "learning_rate": 2.786326119857083e-05, "loss": 0.2385, "step": 23825 }, { "epoch": 1.11, "learning_rate": 2.786279312873481e-05, "loss": 0.2514, "step": 23830 }, { "epoch": 1.11, "learning_rate": 2.786232505889879e-05, "loss": 0.3098, "step": 23835 }, { "epoch": 1.11, "learning_rate": 2.7861856989062766e-05, "loss": 0.0623, "step": 23840 }, { "epoch": 1.11, "learning_rate": 2.786138891922675e-05, "loss": 0.0985, "step": 23845 }, { "epoch": 1.11, "learning_rate": 2.786092084939073e-05, "loss": 0.0456, "step": 23850 }, { "epoch": 1.11, "learning_rate": 2.786045277955471e-05, "loss": 0.1148, "step": 23855 }, { "epoch": 1.11, "learning_rate": 2.7859984709718692e-05, "loss": 0.1061, "step": 23860 }, { "epoch": 1.11, "learning_rate": 2.785951663988267e-05, "loss": 0.1392, "step": 23865 }, { "epoch": 1.11, "learning_rate": 2.785904857004665e-05, "loss": 0.1639, "step": 23870 }, { "epoch": 1.11, "learning_rate": 2.785858050021063e-05, "loss": 0.2892, "step": 23875 }, { "epoch": 1.11, "learning_rate": 2.7858112430374614e-05, "loss": 0.3988, "step": 23880 }, { "epoch": 1.11, "learning_rate": 2.7857644360538594e-05, "loss": 0.2502, "step": 23885 }, { "epoch": 1.11, "learning_rate": 2.7857176290702574e-05, "loss": 0.0541, "step": 23890 }, { "epoch": 1.11, "learning_rate": 2.7856708220866554e-05, "loss": 0.0181, "step": 23895 }, { "epoch": 1.12, "learning_rate": 2.7856240151030534e-05, "loss": 0.1351, "step": 23900 }, { "epoch": 1.12, "learning_rate": 2.7855772081194514e-05, "loss": 0.1144, "step": 23905 }, { "epoch": 1.12, "learning_rate": 2.7855304011358493e-05, "loss": 0.0976, "step": 23910 }, { "epoch": 1.12, "learning_rate": 2.7854835941522477e-05, "loss": 0.162, "step": 23915 }, { "epoch": 1.12, "learning_rate": 2.7854367871686456e-05, "loss": 0.1764, "step": 23920 }, { "epoch": 1.12, "learning_rate": 2.7853899801850436e-05, "loss": 0.2787, "step": 23925 }, { "epoch": 1.12, "learning_rate": 2.7853431732014416e-05, "loss": 0.255, "step": 23930 }, { "epoch": 1.12, "learning_rate": 2.78529636621784e-05, "loss": 0.2853, "step": 23935 }, { "epoch": 1.12, "learning_rate": 2.785249559234238e-05, "loss": 0.0753, "step": 23940 }, { "epoch": 1.12, "learning_rate": 2.785202752250636e-05, "loss": 0.0491, "step": 23945 }, { "epoch": 1.12, "learning_rate": 2.785155945267034e-05, "loss": 0.0862, "step": 23950 }, { "epoch": 1.12, "learning_rate": 2.7851091382834322e-05, "loss": 0.141, "step": 23955 }, { "epoch": 1.12, "learning_rate": 2.78506233129983e-05, "loss": 0.1506, "step": 23960 }, { "epoch": 1.12, "learning_rate": 2.7850155243162278e-05, "loss": 0.1839, "step": 23965 }, { "epoch": 1.12, "learning_rate": 2.784968717332626e-05, "loss": 0.1658, "step": 23970 }, { "epoch": 1.12, "learning_rate": 2.784921910349024e-05, "loss": 0.2148, "step": 23975 }, { "epoch": 1.12, "learning_rate": 2.784875103365422e-05, "loss": 0.3744, "step": 23980 }, { "epoch": 1.12, "learning_rate": 2.78482829638182e-05, "loss": 0.4416, "step": 23985 }, { "epoch": 1.12, "learning_rate": 2.7847814893982184e-05, "loss": 0.0708, "step": 23990 }, { "epoch": 1.12, "learning_rate": 2.7847346824146164e-05, "loss": 0.0429, "step": 23995 }, { "epoch": 1.12, "learning_rate": 2.7846878754310144e-05, "loss": 0.0519, "step": 24000 }, { "epoch": 1.12, "learning_rate": 2.7846410684474123e-05, "loss": 0.076, "step": 24005 }, { "epoch": 1.12, "learning_rate": 2.7845942614638107e-05, "loss": 0.0836, "step": 24010 }, { "epoch": 1.12, "learning_rate": 2.7845474544802086e-05, "loss": 0.2031, "step": 24015 }, { "epoch": 1.12, "learning_rate": 2.7845006474966066e-05, "loss": 0.1877, "step": 24020 }, { "epoch": 1.12, "learning_rate": 2.7844538405130046e-05, "loss": 0.1343, "step": 24025 }, { "epoch": 1.12, "learning_rate": 2.7844070335294026e-05, "loss": 0.4115, "step": 24030 }, { "epoch": 1.12, "learning_rate": 2.7843602265458006e-05, "loss": 0.1969, "step": 24035 }, { "epoch": 1.12, "learning_rate": 2.7843134195621986e-05, "loss": 0.0571, "step": 24040 }, { "epoch": 1.12, "learning_rate": 2.784266612578597e-05, "loss": 0.0702, "step": 24045 }, { "epoch": 1.12, "learning_rate": 2.784219805594995e-05, "loss": 0.0763, "step": 24050 }, { "epoch": 1.12, "learning_rate": 2.784172998611393e-05, "loss": 0.0857, "step": 24055 }, { "epoch": 1.12, "learning_rate": 2.7841261916277908e-05, "loss": 0.1207, "step": 24060 }, { "epoch": 1.12, "learning_rate": 2.784079384644189e-05, "loss": 0.2118, "step": 24065 }, { "epoch": 1.12, "learning_rate": 2.784032577660587e-05, "loss": 0.1536, "step": 24070 }, { "epoch": 1.12, "learning_rate": 2.783985770676985e-05, "loss": 0.3481, "step": 24075 }, { "epoch": 1.12, "learning_rate": 2.783938963693383e-05, "loss": 0.2797, "step": 24080 }, { "epoch": 1.12, "learning_rate": 2.7838921567097814e-05, "loss": 0.3007, "step": 24085 }, { "epoch": 1.12, "learning_rate": 2.783845349726179e-05, "loss": 0.0347, "step": 24090 }, { "epoch": 1.12, "learning_rate": 2.783798542742577e-05, "loss": 0.0338, "step": 24095 }, { "epoch": 1.12, "learning_rate": 2.7837517357589754e-05, "loss": 0.1197, "step": 24100 }, { "epoch": 1.12, "learning_rate": 2.7837049287753733e-05, "loss": 0.105, "step": 24105 }, { "epoch": 1.13, "learning_rate": 2.7836581217917713e-05, "loss": 0.0927, "step": 24110 }, { "epoch": 1.13, "learning_rate": 2.7836113148081693e-05, "loss": 0.1928, "step": 24115 }, { "epoch": 1.13, "learning_rate": 2.7835645078245676e-05, "loss": 0.1256, "step": 24120 }, { "epoch": 1.13, "learning_rate": 2.7835177008409656e-05, "loss": 0.2703, "step": 24125 }, { "epoch": 1.13, "learning_rate": 2.7834708938573636e-05, "loss": 0.5215, "step": 24130 }, { "epoch": 1.13, "learning_rate": 2.7834240868737616e-05, "loss": 0.4436, "step": 24135 }, { "epoch": 1.13, "learning_rate": 2.78337727989016e-05, "loss": 0.0458, "step": 24140 }, { "epoch": 1.13, "learning_rate": 2.783330472906558e-05, "loss": 0.059, "step": 24145 }, { "epoch": 1.13, "learning_rate": 2.783283665922956e-05, "loss": 0.0762, "step": 24150 }, { "epoch": 1.13, "learning_rate": 2.7832368589393538e-05, "loss": 0.0905, "step": 24155 }, { "epoch": 1.13, "learning_rate": 2.7831900519557518e-05, "loss": 0.0656, "step": 24160 }, { "epoch": 1.13, "learning_rate": 2.7831432449721498e-05, "loss": 0.1279, "step": 24165 }, { "epoch": 1.13, "learning_rate": 2.7830964379885478e-05, "loss": 0.1636, "step": 24170 }, { "epoch": 1.13, "learning_rate": 2.783049631004946e-05, "loss": 0.2044, "step": 24175 }, { "epoch": 1.13, "learning_rate": 2.783002824021344e-05, "loss": 0.3608, "step": 24180 }, { "epoch": 1.13, "learning_rate": 2.782956017037742e-05, "loss": 0.3159, "step": 24185 }, { "epoch": 1.13, "learning_rate": 2.78290921005414e-05, "loss": 0.0909, "step": 24190 }, { "epoch": 1.13, "learning_rate": 2.7828624030705384e-05, "loss": 0.0536, "step": 24195 }, { "epoch": 1.13, "learning_rate": 2.7828155960869363e-05, "loss": 0.0767, "step": 24200 }, { "epoch": 1.13, "learning_rate": 2.7827687891033343e-05, "loss": 0.0605, "step": 24205 }, { "epoch": 1.13, "learning_rate": 2.7827219821197326e-05, "loss": 0.1344, "step": 24210 }, { "epoch": 1.13, "learning_rate": 2.7826751751361303e-05, "loss": 0.1458, "step": 24215 }, { "epoch": 1.13, "learning_rate": 2.7826283681525283e-05, "loss": 0.2126, "step": 24220 }, { "epoch": 1.13, "learning_rate": 2.7825815611689263e-05, "loss": 0.213, "step": 24225 }, { "epoch": 1.13, "learning_rate": 2.7825347541853246e-05, "loss": 0.4755, "step": 24230 }, { "epoch": 1.13, "learning_rate": 2.7824879472017226e-05, "loss": 0.3213, "step": 24235 }, { "epoch": 1.13, "learning_rate": 2.7824411402181205e-05, "loss": 0.0693, "step": 24240 }, { "epoch": 1.13, "learning_rate": 2.7823943332345185e-05, "loss": 0.0591, "step": 24245 }, { "epoch": 1.13, "learning_rate": 2.782347526250917e-05, "loss": 0.0504, "step": 24250 }, { "epoch": 1.13, "learning_rate": 2.7823007192673148e-05, "loss": 0.0842, "step": 24255 }, { "epoch": 1.13, "learning_rate": 2.7822539122837128e-05, "loss": 0.1454, "step": 24260 }, { "epoch": 1.13, "learning_rate": 2.7822071053001108e-05, "loss": 0.1549, "step": 24265 }, { "epoch": 1.13, "learning_rate": 2.782160298316509e-05, "loss": 0.1323, "step": 24270 }, { "epoch": 1.13, "learning_rate": 2.782113491332907e-05, "loss": 0.265, "step": 24275 }, { "epoch": 1.13, "learning_rate": 2.7820666843493047e-05, "loss": 0.4209, "step": 24280 }, { "epoch": 1.13, "learning_rate": 2.782019877365703e-05, "loss": 0.2853, "step": 24285 }, { "epoch": 1.13, "learning_rate": 2.781973070382101e-05, "loss": 0.101, "step": 24290 }, { "epoch": 1.13, "learning_rate": 2.781926263398499e-05, "loss": 0.022, "step": 24295 }, { "epoch": 1.13, "learning_rate": 2.781879456414897e-05, "loss": 0.0391, "step": 24300 }, { "epoch": 1.13, "learning_rate": 2.7818326494312953e-05, "loss": 0.0911, "step": 24305 }, { "epoch": 1.13, "learning_rate": 2.7817858424476933e-05, "loss": 0.1841, "step": 24310 }, { "epoch": 1.13, "learning_rate": 2.7817390354640913e-05, "loss": 0.144, "step": 24315 }, { "epoch": 1.13, "learning_rate": 2.7816922284804893e-05, "loss": 0.1671, "step": 24320 }, { "epoch": 1.14, "learning_rate": 2.7816454214968876e-05, "loss": 0.3196, "step": 24325 }, { "epoch": 1.14, "learning_rate": 2.7815986145132856e-05, "loss": 0.4916, "step": 24330 }, { "epoch": 1.14, "learning_rate": 2.7815518075296835e-05, "loss": 0.4262, "step": 24335 }, { "epoch": 1.14, "learning_rate": 2.781505000546082e-05, "loss": 0.1215, "step": 24340 }, { "epoch": 1.14, "learning_rate": 2.7814581935624795e-05, "loss": 0.0956, "step": 24345 }, { "epoch": 1.14, "learning_rate": 2.7814113865788775e-05, "loss": 0.1143, "step": 24350 }, { "epoch": 1.14, "learning_rate": 2.7813645795952755e-05, "loss": 0.0758, "step": 24355 }, { "epoch": 1.14, "learning_rate": 2.7813177726116738e-05, "loss": 0.1008, "step": 24360 }, { "epoch": 1.14, "learning_rate": 2.7812709656280718e-05, "loss": 0.1616, "step": 24365 }, { "epoch": 1.14, "learning_rate": 2.7812241586444698e-05, "loss": 0.1988, "step": 24370 }, { "epoch": 1.14, "learning_rate": 2.7811773516608677e-05, "loss": 0.1654, "step": 24375 }, { "epoch": 1.14, "learning_rate": 2.781130544677266e-05, "loss": 0.3107, "step": 24380 }, { "epoch": 1.14, "learning_rate": 2.781083737693664e-05, "loss": 0.2834, "step": 24385 }, { "epoch": 1.14, "learning_rate": 2.781036930710062e-05, "loss": 0.1004, "step": 24390 }, { "epoch": 1.14, "learning_rate": 2.7809901237264603e-05, "loss": 0.0424, "step": 24395 }, { "epoch": 1.14, "learning_rate": 2.7809433167428583e-05, "loss": 0.1437, "step": 24400 }, { "epoch": 1.14, "learning_rate": 2.780896509759256e-05, "loss": 0.1121, "step": 24405 }, { "epoch": 1.14, "learning_rate": 2.780849702775654e-05, "loss": 0.1602, "step": 24410 }, { "epoch": 1.14, "learning_rate": 2.7808028957920523e-05, "loss": 0.2569, "step": 24415 }, { "epoch": 1.14, "learning_rate": 2.7807560888084502e-05, "loss": 0.1752, "step": 24420 }, { "epoch": 1.14, "learning_rate": 2.7807092818248482e-05, "loss": 0.2632, "step": 24425 }, { "epoch": 1.14, "learning_rate": 2.7806624748412462e-05, "loss": 0.3705, "step": 24430 }, { "epoch": 1.14, "learning_rate": 2.7806156678576445e-05, "loss": 0.3963, "step": 24435 }, { "epoch": 1.14, "learning_rate": 2.7805688608740425e-05, "loss": 0.0922, "step": 24440 }, { "epoch": 1.14, "learning_rate": 2.7805220538904405e-05, "loss": 0.0646, "step": 24445 }, { "epoch": 1.14, "learning_rate": 2.7804752469068385e-05, "loss": 0.036, "step": 24450 }, { "epoch": 1.14, "learning_rate": 2.7804284399232368e-05, "loss": 0.1185, "step": 24455 }, { "epoch": 1.14, "learning_rate": 2.7803816329396348e-05, "loss": 0.1575, "step": 24460 }, { "epoch": 1.14, "learning_rate": 2.7803348259560328e-05, "loss": 0.0933, "step": 24465 }, { "epoch": 1.14, "learning_rate": 2.7802880189724307e-05, "loss": 0.1831, "step": 24470 }, { "epoch": 1.14, "learning_rate": 2.7802412119888287e-05, "loss": 0.2376, "step": 24475 }, { "epoch": 1.14, "learning_rate": 2.7801944050052267e-05, "loss": 0.3066, "step": 24480 }, { "epoch": 1.14, "learning_rate": 2.7801475980216247e-05, "loss": 0.3635, "step": 24485 }, { "epoch": 1.14, "learning_rate": 2.780100791038023e-05, "loss": 0.0318, "step": 24490 }, { "epoch": 1.14, "learning_rate": 2.780053984054421e-05, "loss": 0.0678, "step": 24495 }, { "epoch": 1.14, "learning_rate": 2.780007177070819e-05, "loss": 0.1153, "step": 24500 }, { "epoch": 1.14, "learning_rate": 2.779960370087217e-05, "loss": 0.1895, "step": 24505 }, { "epoch": 1.14, "learning_rate": 2.7799135631036153e-05, "loss": 0.1513, "step": 24510 }, { "epoch": 1.14, "learning_rate": 2.7798667561200133e-05, "loss": 0.0862, "step": 24515 }, { "epoch": 1.14, "learning_rate": 2.7798199491364112e-05, "loss": 0.2052, "step": 24520 }, { "epoch": 1.14, "learning_rate": 2.7797731421528096e-05, "loss": 0.2323, "step": 24525 }, { "epoch": 1.14, "learning_rate": 2.7797263351692075e-05, "loss": 0.4008, "step": 24530 }, { "epoch": 1.14, "learning_rate": 2.7796795281856052e-05, "loss": 0.3173, "step": 24535 }, { "epoch": 1.15, "learning_rate": 2.779632721202003e-05, "loss": 0.0315, "step": 24540 }, { "epoch": 1.15, "learning_rate": 2.7795859142184015e-05, "loss": 0.1739, "step": 24545 }, { "epoch": 1.15, "learning_rate": 2.7795391072347995e-05, "loss": 0.0783, "step": 24550 }, { "epoch": 1.15, "learning_rate": 2.7794923002511975e-05, "loss": 0.1348, "step": 24555 }, { "epoch": 1.15, "learning_rate": 2.7794454932675954e-05, "loss": 0.1921, "step": 24560 }, { "epoch": 1.15, "learning_rate": 2.7793986862839938e-05, "loss": 0.1798, "step": 24565 }, { "epoch": 1.15, "learning_rate": 2.7793518793003917e-05, "loss": 0.1721, "step": 24570 }, { "epoch": 1.15, "learning_rate": 2.7793050723167897e-05, "loss": 0.2046, "step": 24575 }, { "epoch": 1.15, "learning_rate": 2.779258265333188e-05, "loss": 0.3995, "step": 24580 }, { "epoch": 1.15, "learning_rate": 2.779211458349586e-05, "loss": 0.3411, "step": 24585 }, { "epoch": 1.15, "learning_rate": 2.779164651365984e-05, "loss": 0.0392, "step": 24590 }, { "epoch": 1.15, "learning_rate": 2.7791178443823816e-05, "loss": 0.0342, "step": 24595 }, { "epoch": 1.15, "learning_rate": 2.77907103739878e-05, "loss": 0.03, "step": 24600 }, { "epoch": 1.15, "learning_rate": 2.779024230415178e-05, "loss": 0.0792, "step": 24605 }, { "epoch": 1.15, "learning_rate": 2.778977423431576e-05, "loss": 0.1117, "step": 24610 }, { "epoch": 1.15, "learning_rate": 2.778930616447974e-05, "loss": 0.186, "step": 24615 }, { "epoch": 1.15, "learning_rate": 2.7788838094643722e-05, "loss": 0.2702, "step": 24620 }, { "epoch": 1.15, "learning_rate": 2.7788370024807702e-05, "loss": 0.2745, "step": 24625 }, { "epoch": 1.15, "learning_rate": 2.7787901954971682e-05, "loss": 0.3906, "step": 24630 }, { "epoch": 1.15, "learning_rate": 2.7787433885135662e-05, "loss": 0.2601, "step": 24635 }, { "epoch": 1.15, "learning_rate": 2.7786965815299645e-05, "loss": 0.0918, "step": 24640 }, { "epoch": 1.15, "learning_rate": 2.7786497745463625e-05, "loss": 0.0976, "step": 24645 }, { "epoch": 1.15, "learning_rate": 2.7786029675627605e-05, "loss": 0.0672, "step": 24650 }, { "epoch": 1.15, "learning_rate": 2.7785561605791588e-05, "loss": 0.078, "step": 24655 }, { "epoch": 1.15, "learning_rate": 2.7785093535955564e-05, "loss": 0.0773, "step": 24660 }, { "epoch": 1.15, "learning_rate": 2.7784625466119544e-05, "loss": 0.127, "step": 24665 }, { "epoch": 1.15, "learning_rate": 2.7784157396283524e-05, "loss": 0.1797, "step": 24670 }, { "epoch": 1.15, "learning_rate": 2.7783689326447507e-05, "loss": 0.2975, "step": 24675 }, { "epoch": 1.15, "learning_rate": 2.7783221256611487e-05, "loss": 0.461, "step": 24680 }, { "epoch": 1.15, "learning_rate": 2.7782753186775467e-05, "loss": 0.2972, "step": 24685 }, { "epoch": 1.15, "learning_rate": 2.7782285116939447e-05, "loss": 0.1419, "step": 24690 }, { "epoch": 1.15, "learning_rate": 2.778181704710343e-05, "loss": 0.1212, "step": 24695 }, { "epoch": 1.15, "learning_rate": 2.778134897726741e-05, "loss": 0.1443, "step": 24700 }, { "epoch": 1.15, "learning_rate": 2.778088090743139e-05, "loss": 0.1086, "step": 24705 }, { "epoch": 1.15, "learning_rate": 2.7780412837595373e-05, "loss": 0.1421, "step": 24710 }, { "epoch": 1.15, "learning_rate": 2.7779944767759352e-05, "loss": 0.1977, "step": 24715 }, { "epoch": 1.15, "learning_rate": 2.7779476697923332e-05, "loss": 0.2013, "step": 24720 }, { "epoch": 1.15, "learning_rate": 2.777900862808731e-05, "loss": 0.2399, "step": 24725 }, { "epoch": 1.15, "learning_rate": 2.7778540558251292e-05, "loss": 0.3368, "step": 24730 }, { "epoch": 1.15, "learning_rate": 2.777807248841527e-05, "loss": 0.3361, "step": 24735 }, { "epoch": 1.15, "learning_rate": 2.777760441857925e-05, "loss": 0.0392, "step": 24740 }, { "epoch": 1.15, "learning_rate": 2.777713634874323e-05, "loss": 0.116, "step": 24745 }, { "epoch": 1.15, "learning_rate": 2.7776668278907215e-05, "loss": 0.1077, "step": 24750 }, { "epoch": 1.16, "learning_rate": 2.7776200209071194e-05, "loss": 0.1151, "step": 24755 }, { "epoch": 1.16, "learning_rate": 2.7775732139235174e-05, "loss": 0.1359, "step": 24760 }, { "epoch": 1.16, "learning_rate": 2.7775264069399157e-05, "loss": 0.1645, "step": 24765 }, { "epoch": 1.16, "learning_rate": 2.7774795999563137e-05, "loss": 0.2649, "step": 24770 }, { "epoch": 1.16, "learning_rate": 2.7774327929727117e-05, "loss": 0.2646, "step": 24775 }, { "epoch": 1.16, "learning_rate": 2.7773859859891097e-05, "loss": 0.4178, "step": 24780 }, { "epoch": 1.16, "learning_rate": 2.7773391790055077e-05, "loss": 0.2718, "step": 24785 }, { "epoch": 1.16, "learning_rate": 2.7772923720219056e-05, "loss": 0.0642, "step": 24790 }, { "epoch": 1.16, "learning_rate": 2.7772455650383036e-05, "loss": 0.1047, "step": 24795 }, { "epoch": 1.16, "learning_rate": 2.7771987580547016e-05, "loss": 0.1553, "step": 24800 }, { "epoch": 1.16, "learning_rate": 2.7771519510711e-05, "loss": 0.147, "step": 24805 }, { "epoch": 1.16, "learning_rate": 2.777105144087498e-05, "loss": 0.1093, "step": 24810 }, { "epoch": 1.16, "learning_rate": 2.777058337103896e-05, "loss": 0.157, "step": 24815 }, { "epoch": 1.16, "learning_rate": 2.7770115301202942e-05, "loss": 0.2539, "step": 24820 }, { "epoch": 1.16, "learning_rate": 2.7769647231366922e-05, "loss": 0.1965, "step": 24825 }, { "epoch": 1.16, "learning_rate": 2.7769179161530902e-05, "loss": 0.2952, "step": 24830 }, { "epoch": 1.16, "learning_rate": 2.776871109169488e-05, "loss": 0.2187, "step": 24835 }, { "epoch": 1.16, "learning_rate": 2.7768243021858865e-05, "loss": 0.0681, "step": 24840 }, { "epoch": 1.16, "learning_rate": 2.7767774952022845e-05, "loss": 0.0781, "step": 24845 }, { "epoch": 1.16, "learning_rate": 2.776730688218682e-05, "loss": 0.1127, "step": 24850 }, { "epoch": 1.16, "learning_rate": 2.77668388123508e-05, "loss": 0.1195, "step": 24855 }, { "epoch": 1.16, "learning_rate": 2.7766370742514784e-05, "loss": 0.1037, "step": 24860 }, { "epoch": 1.16, "learning_rate": 2.7765902672678764e-05, "loss": 0.16, "step": 24865 }, { "epoch": 1.16, "learning_rate": 2.7765434602842744e-05, "loss": 0.249, "step": 24870 }, { "epoch": 1.16, "learning_rate": 2.7764966533006724e-05, "loss": 0.2931, "step": 24875 }, { "epoch": 1.16, "learning_rate": 2.7764498463170707e-05, "loss": 0.4896, "step": 24880 }, { "epoch": 1.16, "learning_rate": 2.7764030393334687e-05, "loss": 0.1716, "step": 24885 }, { "epoch": 1.16, "learning_rate": 2.7763562323498666e-05, "loss": 0.0742, "step": 24890 }, { "epoch": 1.16, "learning_rate": 2.776309425366265e-05, "loss": 0.1061, "step": 24895 }, { "epoch": 1.16, "learning_rate": 2.776262618382663e-05, "loss": 0.1011, "step": 24900 }, { "epoch": 1.16, "learning_rate": 2.776215811399061e-05, "loss": 0.1582, "step": 24905 }, { "epoch": 1.16, "learning_rate": 2.776169004415459e-05, "loss": 0.1836, "step": 24910 }, { "epoch": 1.16, "learning_rate": 2.776122197431857e-05, "loss": 0.1387, "step": 24915 }, { "epoch": 1.16, "learning_rate": 2.776075390448255e-05, "loss": 0.2108, "step": 24920 }, { "epoch": 1.16, "learning_rate": 2.776028583464653e-05, "loss": 0.2764, "step": 24925 }, { "epoch": 1.16, "learning_rate": 2.7759817764810508e-05, "loss": 0.3323, "step": 24930 }, { "epoch": 1.16, "learning_rate": 2.775934969497449e-05, "loss": 0.2756, "step": 24935 }, { "epoch": 1.16, "learning_rate": 2.775888162513847e-05, "loss": 0.0872, "step": 24940 }, { "epoch": 1.16, "learning_rate": 2.775841355530245e-05, "loss": 0.1389, "step": 24945 }, { "epoch": 1.16, "learning_rate": 2.7757945485466434e-05, "loss": 0.0912, "step": 24950 }, { "epoch": 1.16, "learning_rate": 2.7757477415630414e-05, "loss": 0.0729, "step": 24955 }, { "epoch": 1.16, "learning_rate": 2.7757009345794394e-05, "loss": 0.0942, "step": 24960 }, { "epoch": 1.16, "learning_rate": 2.7756541275958374e-05, "loss": 0.1227, "step": 24965 }, { "epoch": 1.17, "learning_rate": 2.7756073206122357e-05, "loss": 0.1822, "step": 24970 }, { "epoch": 1.17, "learning_rate": 2.7755605136286333e-05, "loss": 0.2144, "step": 24975 }, { "epoch": 1.17, "learning_rate": 2.7755137066450313e-05, "loss": 0.5167, "step": 24980 }, { "epoch": 1.17, "learning_rate": 2.7754668996614293e-05, "loss": 0.2825, "step": 24985 }, { "epoch": 1.17, "learning_rate": 2.7754200926778276e-05, "loss": 0.045, "step": 24990 }, { "epoch": 1.17, "learning_rate": 2.7753732856942256e-05, "loss": 0.0452, "step": 24995 }, { "epoch": 1.17, "learning_rate": 2.7753264787106236e-05, "loss": 0.071, "step": 25000 }, { "epoch": 1.17, "learning_rate": 2.775279671727022e-05, "loss": 0.1491, "step": 25005 }, { "epoch": 1.17, "learning_rate": 2.77523286474342e-05, "loss": 0.1434, "step": 25010 }, { "epoch": 1.17, "learning_rate": 2.775186057759818e-05, "loss": 0.151, "step": 25015 }, { "epoch": 1.17, "learning_rate": 2.775139250776216e-05, "loss": 0.1819, "step": 25020 }, { "epoch": 1.17, "learning_rate": 2.7750924437926142e-05, "loss": 0.2723, "step": 25025 }, { "epoch": 1.17, "learning_rate": 2.775045636809012e-05, "loss": 0.2587, "step": 25030 }, { "epoch": 1.17, "learning_rate": 2.77499882982541e-05, "loss": 0.2854, "step": 25035 }, { "epoch": 1.17, "learning_rate": 2.7749520228418078e-05, "loss": 0.0922, "step": 25040 }, { "epoch": 1.17, "learning_rate": 2.774905215858206e-05, "loss": 0.0191, "step": 25045 }, { "epoch": 1.17, "learning_rate": 2.774858408874604e-05, "loss": 0.0941, "step": 25050 }, { "epoch": 1.17, "learning_rate": 2.774811601891002e-05, "loss": 0.1208, "step": 25055 }, { "epoch": 1.17, "learning_rate": 2.7747647949074e-05, "loss": 0.1192, "step": 25060 }, { "epoch": 1.17, "learning_rate": 2.7747179879237984e-05, "loss": 0.1651, "step": 25065 }, { "epoch": 1.17, "learning_rate": 2.7746711809401963e-05, "loss": 0.2296, "step": 25070 }, { "epoch": 1.17, "learning_rate": 2.7746243739565943e-05, "loss": 0.21, "step": 25075 }, { "epoch": 1.17, "learning_rate": 2.7745775669729927e-05, "loss": 0.3513, "step": 25080 }, { "epoch": 1.17, "learning_rate": 2.7745307599893906e-05, "loss": 0.3436, "step": 25085 }, { "epoch": 1.17, "learning_rate": 2.7744839530057886e-05, "loss": 0.0526, "step": 25090 }, { "epoch": 1.17, "learning_rate": 2.7744371460221866e-05, "loss": 0.0514, "step": 25095 }, { "epoch": 1.17, "learning_rate": 2.7743903390385846e-05, "loss": 0.1136, "step": 25100 }, { "epoch": 1.17, "learning_rate": 2.7743435320549826e-05, "loss": 0.0964, "step": 25105 }, { "epoch": 1.17, "learning_rate": 2.7742967250713805e-05, "loss": 0.0994, "step": 25110 }, { "epoch": 1.17, "learning_rate": 2.7742499180877785e-05, "loss": 0.1668, "step": 25115 }, { "epoch": 1.17, "learning_rate": 2.774203111104177e-05, "loss": 0.172, "step": 25120 }, { "epoch": 1.17, "learning_rate": 2.7741563041205748e-05, "loss": 0.2633, "step": 25125 }, { "epoch": 1.17, "learning_rate": 2.7741094971369728e-05, "loss": 0.4466, "step": 25130 }, { "epoch": 1.17, "learning_rate": 2.774062690153371e-05, "loss": 0.3911, "step": 25135 }, { "epoch": 1.17, "learning_rate": 2.774015883169769e-05, "loss": 0.0396, "step": 25140 }, { "epoch": 1.17, "learning_rate": 2.773969076186167e-05, "loss": 0.0798, "step": 25145 }, { "epoch": 1.17, "learning_rate": 2.773922269202565e-05, "loss": 0.0737, "step": 25150 }, { "epoch": 1.17, "learning_rate": 2.7738754622189634e-05, "loss": 0.0927, "step": 25155 }, { "epoch": 1.17, "learning_rate": 2.7738286552353614e-05, "loss": 0.1029, "step": 25160 }, { "epoch": 1.17, "learning_rate": 2.773781848251759e-05, "loss": 0.1644, "step": 25165 }, { "epoch": 1.17, "learning_rate": 2.773735041268157e-05, "loss": 0.2245, "step": 25170 }, { "epoch": 1.17, "learning_rate": 2.7736882342845553e-05, "loss": 0.21, "step": 25175 }, { "epoch": 1.17, "learning_rate": 2.7736414273009533e-05, "loss": 0.2846, "step": 25180 }, { "epoch": 1.18, "learning_rate": 2.7735946203173513e-05, "loss": 0.3117, "step": 25185 }, { "epoch": 1.18, "learning_rate": 2.7735478133337496e-05, "loss": 0.0897, "step": 25190 }, { "epoch": 1.18, "learning_rate": 2.7735010063501476e-05, "loss": 0.0676, "step": 25195 }, { "epoch": 1.18, "learning_rate": 2.7734541993665456e-05, "loss": 0.0712, "step": 25200 }, { "epoch": 1.18, "learning_rate": 2.7734073923829436e-05, "loss": 0.0796, "step": 25205 }, { "epoch": 1.18, "learning_rate": 2.773360585399342e-05, "loss": 0.089, "step": 25210 }, { "epoch": 1.18, "learning_rate": 2.77331377841574e-05, "loss": 0.1672, "step": 25215 }, { "epoch": 1.18, "learning_rate": 2.773266971432138e-05, "loss": 0.2702, "step": 25220 }, { "epoch": 1.18, "learning_rate": 2.7732201644485358e-05, "loss": 0.2624, "step": 25225 }, { "epoch": 1.18, "learning_rate": 2.7731733574649338e-05, "loss": 0.3468, "step": 25230 }, { "epoch": 1.18, "learning_rate": 2.7731265504813318e-05, "loss": 0.3082, "step": 25235 }, { "epoch": 1.18, "learning_rate": 2.7730797434977298e-05, "loss": 0.0881, "step": 25240 }, { "epoch": 1.18, "learning_rate": 2.7730329365141277e-05, "loss": 0.0347, "step": 25245 }, { "epoch": 1.18, "learning_rate": 2.772986129530526e-05, "loss": 0.105, "step": 25250 }, { "epoch": 1.18, "learning_rate": 2.772939322546924e-05, "loss": 0.0889, "step": 25255 }, { "epoch": 1.18, "learning_rate": 2.772892515563322e-05, "loss": 0.0521, "step": 25260 }, { "epoch": 1.18, "learning_rate": 2.7728457085797203e-05, "loss": 0.1286, "step": 25265 }, { "epoch": 1.18, "learning_rate": 2.7727989015961183e-05, "loss": 0.1604, "step": 25270 }, { "epoch": 1.18, "learning_rate": 2.7727520946125163e-05, "loss": 0.2217, "step": 25275 }, { "epoch": 1.18, "learning_rate": 2.7727052876289143e-05, "loss": 0.4269, "step": 25280 }, { "epoch": 1.18, "learning_rate": 2.7726584806453126e-05, "loss": 0.3102, "step": 25285 }, { "epoch": 1.18, "learning_rate": 2.7726116736617103e-05, "loss": 0.0935, "step": 25290 }, { "epoch": 1.18, "learning_rate": 2.7725648666781082e-05, "loss": 0.0656, "step": 25295 }, { "epoch": 1.18, "learning_rate": 2.7725180596945062e-05, "loss": 0.0585, "step": 25300 }, { "epoch": 1.18, "learning_rate": 2.7724712527109045e-05, "loss": 0.0882, "step": 25305 }, { "epoch": 1.18, "learning_rate": 2.7724244457273025e-05, "loss": 0.155, "step": 25310 }, { "epoch": 1.18, "learning_rate": 2.7723776387437005e-05, "loss": 0.1676, "step": 25315 }, { "epoch": 1.18, "learning_rate": 2.7723308317600988e-05, "loss": 0.1588, "step": 25320 }, { "epoch": 1.18, "learning_rate": 2.7722840247764968e-05, "loss": 0.2572, "step": 25325 }, { "epoch": 1.18, "learning_rate": 2.7722372177928948e-05, "loss": 0.375, "step": 25330 }, { "epoch": 1.18, "learning_rate": 2.7721904108092928e-05, "loss": 0.2919, "step": 25335 }, { "epoch": 1.18, "learning_rate": 2.772143603825691e-05, "loss": 0.0853, "step": 25340 }, { "epoch": 1.18, "learning_rate": 2.772096796842089e-05, "loss": 0.1015, "step": 25345 }, { "epoch": 1.18, "learning_rate": 2.772049989858487e-05, "loss": 0.0854, "step": 25350 }, { "epoch": 1.18, "learning_rate": 2.7720031828748847e-05, "loss": 0.1539, "step": 25355 }, { "epoch": 1.18, "learning_rate": 2.771956375891283e-05, "loss": 0.102, "step": 25360 }, { "epoch": 1.18, "learning_rate": 2.771909568907681e-05, "loss": 0.1338, "step": 25365 }, { "epoch": 1.18, "learning_rate": 2.771862761924079e-05, "loss": 0.1894, "step": 25370 }, { "epoch": 1.18, "learning_rate": 2.7718159549404773e-05, "loss": 0.1858, "step": 25375 }, { "epoch": 1.18, "learning_rate": 2.7717691479568753e-05, "loss": 0.5424, "step": 25380 }, { "epoch": 1.18, "learning_rate": 2.7717223409732733e-05, "loss": 0.3205, "step": 25385 }, { "epoch": 1.18, "learning_rate": 2.7716755339896712e-05, "loss": 0.061, "step": 25390 }, { "epoch": 1.18, "learning_rate": 2.7716287270060696e-05, "loss": 0.0828, "step": 25395 }, { "epoch": 1.19, "learning_rate": 2.7715819200224676e-05, "loss": 0.1036, "step": 25400 }, { "epoch": 1.19, "learning_rate": 2.7715351130388655e-05, "loss": 0.1288, "step": 25405 }, { "epoch": 1.19, "learning_rate": 2.7714883060552635e-05, "loss": 0.134, "step": 25410 }, { "epoch": 1.19, "learning_rate": 2.771441499071662e-05, "loss": 0.2133, "step": 25415 }, { "epoch": 1.19, "learning_rate": 2.7713946920880595e-05, "loss": 0.2697, "step": 25420 }, { "epoch": 1.19, "learning_rate": 2.7713478851044575e-05, "loss": 0.2125, "step": 25425 }, { "epoch": 1.19, "learning_rate": 2.7713010781208554e-05, "loss": 0.4802, "step": 25430 }, { "epoch": 1.19, "learning_rate": 2.7712542711372538e-05, "loss": 0.2832, "step": 25435 }, { "epoch": 1.19, "learning_rate": 2.7712074641536517e-05, "loss": 0.0427, "step": 25440 }, { "epoch": 1.19, "learning_rate": 2.7711606571700497e-05, "loss": 0.0404, "step": 25445 }, { "epoch": 1.19, "learning_rate": 2.771113850186448e-05, "loss": 0.1058, "step": 25450 }, { "epoch": 1.19, "learning_rate": 2.771067043202846e-05, "loss": 0.1017, "step": 25455 }, { "epoch": 1.19, "learning_rate": 2.771020236219244e-05, "loss": 0.1037, "step": 25460 }, { "epoch": 1.19, "learning_rate": 2.770973429235642e-05, "loss": 0.2486, "step": 25465 }, { "epoch": 1.19, "learning_rate": 2.7709266222520403e-05, "loss": 0.179, "step": 25470 }, { "epoch": 1.19, "learning_rate": 2.7708798152684383e-05, "loss": 0.265, "step": 25475 }, { "epoch": 1.19, "learning_rate": 2.770833008284836e-05, "loss": 0.4503, "step": 25480 }, { "epoch": 1.19, "learning_rate": 2.770786201301234e-05, "loss": 0.4062, "step": 25485 }, { "epoch": 1.19, "learning_rate": 2.7707393943176322e-05, "loss": 0.101, "step": 25490 }, { "epoch": 1.19, "learning_rate": 2.7706925873340302e-05, "loss": 0.0385, "step": 25495 }, { "epoch": 1.19, "learning_rate": 2.7706457803504282e-05, "loss": 0.0892, "step": 25500 }, { "epoch": 1.19, "learning_rate": 2.7705989733668265e-05, "loss": 0.1343, "step": 25505 }, { "epoch": 1.19, "learning_rate": 2.7705521663832245e-05, "loss": 0.0809, "step": 25510 }, { "epoch": 1.19, "learning_rate": 2.7705053593996225e-05, "loss": 0.1952, "step": 25515 }, { "epoch": 1.19, "learning_rate": 2.7704585524160205e-05, "loss": 0.2179, "step": 25520 }, { "epoch": 1.19, "learning_rate": 2.7704117454324188e-05, "loss": 0.2581, "step": 25525 }, { "epoch": 1.19, "learning_rate": 2.7703649384488168e-05, "loss": 0.3992, "step": 25530 }, { "epoch": 1.19, "learning_rate": 2.7703181314652148e-05, "loss": 0.3823, "step": 25535 }, { "epoch": 1.19, "learning_rate": 2.7702713244816127e-05, "loss": 0.0789, "step": 25540 }, { "epoch": 1.19, "learning_rate": 2.7702245174980107e-05, "loss": 0.0499, "step": 25545 }, { "epoch": 1.19, "learning_rate": 2.7701777105144087e-05, "loss": 0.073, "step": 25550 }, { "epoch": 1.19, "learning_rate": 2.7701309035308067e-05, "loss": 0.1008, "step": 25555 }, { "epoch": 1.19, "learning_rate": 2.770084096547205e-05, "loss": 0.1142, "step": 25560 }, { "epoch": 1.19, "learning_rate": 2.770037289563603e-05, "loss": 0.1204, "step": 25565 }, { "epoch": 1.19, "learning_rate": 2.769990482580001e-05, "loss": 0.1725, "step": 25570 }, { "epoch": 1.19, "learning_rate": 2.769943675596399e-05, "loss": 0.1671, "step": 25575 }, { "epoch": 1.19, "learning_rate": 2.7698968686127973e-05, "loss": 0.4282, "step": 25580 }, { "epoch": 1.19, "learning_rate": 2.7698500616291952e-05, "loss": 0.2291, "step": 25585 }, { "epoch": 1.19, "learning_rate": 2.7698032546455932e-05, "loss": 0.0799, "step": 25590 }, { "epoch": 1.19, "learning_rate": 2.7697564476619912e-05, "loss": 0.1047, "step": 25595 }, { "epoch": 1.19, "learning_rate": 2.7697096406783895e-05, "loss": 0.1213, "step": 25600 }, { "epoch": 1.19, "learning_rate": 2.7696628336947872e-05, "loss": 0.1082, "step": 25605 }, { "epoch": 1.19, "learning_rate": 2.769616026711185e-05, "loss": 0.1425, "step": 25610 }, { "epoch": 1.2, "learning_rate": 2.7695692197275835e-05, "loss": 0.1121, "step": 25615 }, { "epoch": 1.2, "learning_rate": 2.7695224127439815e-05, "loss": 0.225, "step": 25620 }, { "epoch": 1.2, "learning_rate": 2.7694756057603794e-05, "loss": 0.2783, "step": 25625 }, { "epoch": 1.2, "learning_rate": 2.7694287987767774e-05, "loss": 0.3588, "step": 25630 }, { "epoch": 1.2, "learning_rate": 2.7693819917931757e-05, "loss": 0.3846, "step": 25635 }, { "epoch": 1.2, "learning_rate": 2.7693351848095737e-05, "loss": 0.0537, "step": 25640 }, { "epoch": 1.2, "learning_rate": 2.7692883778259717e-05, "loss": 0.0729, "step": 25645 }, { "epoch": 1.2, "learning_rate": 2.7692415708423697e-05, "loss": 0.1688, "step": 25650 }, { "epoch": 1.2, "learning_rate": 2.769194763858768e-05, "loss": 0.1576, "step": 25655 }, { "epoch": 1.2, "learning_rate": 2.769147956875166e-05, "loss": 0.1189, "step": 25660 }, { "epoch": 1.2, "learning_rate": 2.769101149891564e-05, "loss": 0.1946, "step": 25665 }, { "epoch": 1.2, "learning_rate": 2.7690543429079616e-05, "loss": 0.2424, "step": 25670 }, { "epoch": 1.2, "learning_rate": 2.76900753592436e-05, "loss": 0.266, "step": 25675 }, { "epoch": 1.2, "learning_rate": 2.768960728940758e-05, "loss": 0.4794, "step": 25680 }, { "epoch": 1.2, "learning_rate": 2.768913921957156e-05, "loss": 0.2383, "step": 25685 }, { "epoch": 1.2, "learning_rate": 2.7688671149735542e-05, "loss": 0.0965, "step": 25690 }, { "epoch": 1.2, "learning_rate": 2.7688203079899522e-05, "loss": 0.0779, "step": 25695 }, { "epoch": 1.2, "learning_rate": 2.7687735010063502e-05, "loss": 0.1129, "step": 25700 }, { "epoch": 1.2, "learning_rate": 2.768726694022748e-05, "loss": 0.1068, "step": 25705 }, { "epoch": 1.2, "learning_rate": 2.7686798870391465e-05, "loss": 0.0992, "step": 25710 }, { "epoch": 1.2, "learning_rate": 2.7686330800555445e-05, "loss": 0.1579, "step": 25715 }, { "epoch": 1.2, "learning_rate": 2.7685862730719424e-05, "loss": 0.1385, "step": 25720 }, { "epoch": 1.2, "learning_rate": 2.7685394660883404e-05, "loss": 0.2129, "step": 25725 }, { "epoch": 1.2, "learning_rate": 2.7684926591047388e-05, "loss": 0.4603, "step": 25730 }, { "epoch": 1.2, "learning_rate": 2.7684458521211364e-05, "loss": 0.2509, "step": 25735 }, { "epoch": 1.2, "learning_rate": 2.7683990451375344e-05, "loss": 0.0231, "step": 25740 }, { "epoch": 1.2, "learning_rate": 2.7683522381539327e-05, "loss": 0.1146, "step": 25745 }, { "epoch": 1.2, "learning_rate": 2.7683054311703307e-05, "loss": 0.0884, "step": 25750 }, { "epoch": 1.2, "learning_rate": 2.7682586241867287e-05, "loss": 0.1142, "step": 25755 }, { "epoch": 1.2, "learning_rate": 2.7682118172031266e-05, "loss": 0.0975, "step": 25760 }, { "epoch": 1.2, "learning_rate": 2.768165010219525e-05, "loss": 0.1035, "step": 25765 }, { "epoch": 1.2, "learning_rate": 2.768118203235923e-05, "loss": 0.2098, "step": 25770 }, { "epoch": 1.2, "learning_rate": 2.768071396252321e-05, "loss": 0.2113, "step": 25775 }, { "epoch": 1.2, "learning_rate": 2.768024589268719e-05, "loss": 0.2928, "step": 25780 }, { "epoch": 1.2, "learning_rate": 2.7679777822851172e-05, "loss": 0.2965, "step": 25785 }, { "epoch": 1.2, "learning_rate": 2.7679309753015152e-05, "loss": 0.0334, "step": 25790 }, { "epoch": 1.2, "learning_rate": 2.767884168317913e-05, "loss": 0.0585, "step": 25795 }, { "epoch": 1.2, "learning_rate": 2.7678373613343112e-05, "loss": 0.0588, "step": 25800 }, { "epoch": 1.2, "learning_rate": 2.767790554350709e-05, "loss": 0.1113, "step": 25805 }, { "epoch": 1.2, "learning_rate": 2.767743747367107e-05, "loss": 0.1197, "step": 25810 }, { "epoch": 1.2, "learning_rate": 2.767696940383505e-05, "loss": 0.1716, "step": 25815 }, { "epoch": 1.2, "learning_rate": 2.7676501333999034e-05, "loss": 0.2334, "step": 25820 }, { "epoch": 1.21, "learning_rate": 2.7676033264163014e-05, "loss": 0.2301, "step": 25825 }, { "epoch": 1.21, "learning_rate": 2.7675565194326994e-05, "loss": 0.2914, "step": 25830 }, { "epoch": 1.21, "learning_rate": 2.7675097124490974e-05, "loss": 0.2337, "step": 25835 }, { "epoch": 1.21, "learning_rate": 2.7674629054654957e-05, "loss": 0.0694, "step": 25840 }, { "epoch": 1.21, "learning_rate": 2.7674160984818937e-05, "loss": 0.075, "step": 25845 }, { "epoch": 1.21, "learning_rate": 2.7673692914982917e-05, "loss": 0.1596, "step": 25850 }, { "epoch": 1.21, "learning_rate": 2.7673224845146897e-05, "loss": 0.0813, "step": 25855 }, { "epoch": 1.21, "learning_rate": 2.7672756775310876e-05, "loss": 0.1492, "step": 25860 }, { "epoch": 1.21, "learning_rate": 2.7672288705474856e-05, "loss": 0.2187, "step": 25865 }, { "epoch": 1.21, "learning_rate": 2.7671820635638836e-05, "loss": 0.2334, "step": 25870 }, { "epoch": 1.21, "learning_rate": 2.767135256580282e-05, "loss": 0.2039, "step": 25875 }, { "epoch": 1.21, "learning_rate": 2.76708844959668e-05, "loss": 0.459, "step": 25880 }, { "epoch": 1.21, "learning_rate": 2.767041642613078e-05, "loss": 0.4232, "step": 25885 }, { "epoch": 1.21, "learning_rate": 2.766994835629476e-05, "loss": 0.0428, "step": 25890 }, { "epoch": 1.21, "learning_rate": 2.7669480286458742e-05, "loss": 0.093, "step": 25895 }, { "epoch": 1.21, "learning_rate": 2.766901221662272e-05, "loss": 0.0938, "step": 25900 }, { "epoch": 1.21, "learning_rate": 2.76685441467867e-05, "loss": 0.1127, "step": 25905 }, { "epoch": 1.21, "learning_rate": 2.766807607695068e-05, "loss": 0.1284, "step": 25910 }, { "epoch": 1.21, "learning_rate": 2.7667608007114664e-05, "loss": 0.1795, "step": 25915 }, { "epoch": 1.21, "learning_rate": 2.7667139937278644e-05, "loss": 0.2153, "step": 25920 }, { "epoch": 1.21, "learning_rate": 2.766667186744262e-05, "loss": 0.1978, "step": 25925 }, { "epoch": 1.21, "learning_rate": 2.7666203797606604e-05, "loss": 0.4028, "step": 25930 }, { "epoch": 1.21, "learning_rate": 2.7665735727770584e-05, "loss": 0.3551, "step": 25935 }, { "epoch": 1.21, "learning_rate": 2.7665267657934564e-05, "loss": 0.0499, "step": 25940 }, { "epoch": 1.21, "learning_rate": 2.7664799588098543e-05, "loss": 0.0746, "step": 25945 }, { "epoch": 1.21, "learning_rate": 2.7664331518262527e-05, "loss": 0.0827, "step": 25950 }, { "epoch": 1.21, "learning_rate": 2.7663863448426506e-05, "loss": 0.0656, "step": 25955 }, { "epoch": 1.21, "learning_rate": 2.7663395378590486e-05, "loss": 0.0846, "step": 25960 }, { "epoch": 1.21, "learning_rate": 2.7662927308754466e-05, "loss": 0.2031, "step": 25965 }, { "epoch": 1.21, "learning_rate": 2.766245923891845e-05, "loss": 0.162, "step": 25970 }, { "epoch": 1.21, "learning_rate": 2.766199116908243e-05, "loss": 0.2951, "step": 25975 }, { "epoch": 1.21, "learning_rate": 2.766152309924641e-05, "loss": 0.3014, "step": 25980 }, { "epoch": 1.21, "learning_rate": 2.766105502941039e-05, "loss": 0.2334, "step": 25985 }, { "epoch": 1.21, "learning_rate": 2.766058695957437e-05, "loss": 0.0926, "step": 25990 }, { "epoch": 1.21, "learning_rate": 2.766011888973835e-05, "loss": 0.0547, "step": 25995 }, { "epoch": 1.21, "learning_rate": 2.7659650819902328e-05, "loss": 0.0435, "step": 26000 }, { "epoch": 1.21, "learning_rate": 2.765918275006631e-05, "loss": 0.0942, "step": 26005 }, { "epoch": 1.21, "learning_rate": 2.765871468023029e-05, "loss": 0.1461, "step": 26010 }, { "epoch": 1.21, "learning_rate": 2.765824661039427e-05, "loss": 0.0797, "step": 26015 }, { "epoch": 1.21, "learning_rate": 2.765777854055825e-05, "loss": 0.2177, "step": 26020 }, { "epoch": 1.21, "learning_rate": 2.7657310470722234e-05, "loss": 0.1941, "step": 26025 }, { "epoch": 1.21, "learning_rate": 2.7656842400886214e-05, "loss": 0.3445, "step": 26030 }, { "epoch": 1.21, "learning_rate": 2.7656374331050194e-05, "loss": 0.2635, "step": 26035 }, { "epoch": 1.22, "learning_rate": 2.7655906261214173e-05, "loss": 0.1444, "step": 26040 }, { "epoch": 1.22, "learning_rate": 2.7655438191378157e-05, "loss": 0.0868, "step": 26045 }, { "epoch": 1.22, "learning_rate": 2.7654970121542133e-05, "loss": 0.1137, "step": 26050 }, { "epoch": 1.22, "learning_rate": 2.7654502051706113e-05, "loss": 0.0746, "step": 26055 }, { "epoch": 1.22, "learning_rate": 2.7654033981870096e-05, "loss": 0.0848, "step": 26060 }, { "epoch": 1.22, "learning_rate": 2.7653565912034076e-05, "loss": 0.131, "step": 26065 }, { "epoch": 1.22, "learning_rate": 2.7653097842198056e-05, "loss": 0.2067, "step": 26070 }, { "epoch": 1.22, "learning_rate": 2.7652629772362036e-05, "loss": 0.2435, "step": 26075 }, { "epoch": 1.22, "learning_rate": 2.765216170252602e-05, "loss": 0.2184, "step": 26080 }, { "epoch": 1.22, "learning_rate": 2.765169363269e-05, "loss": 0.3624, "step": 26085 }, { "epoch": 1.22, "learning_rate": 2.765122556285398e-05, "loss": 0.097, "step": 26090 }, { "epoch": 1.22, "learning_rate": 2.7650757493017958e-05, "loss": 0.1063, "step": 26095 }, { "epoch": 1.22, "learning_rate": 2.765028942318194e-05, "loss": 0.1098, "step": 26100 }, { "epoch": 1.22, "learning_rate": 2.764982135334592e-05, "loss": 0.1686, "step": 26105 }, { "epoch": 1.22, "learning_rate": 2.76493532835099e-05, "loss": 0.1633, "step": 26110 }, { "epoch": 1.22, "learning_rate": 2.764888521367388e-05, "loss": 0.1439, "step": 26115 }, { "epoch": 1.22, "learning_rate": 2.764841714383786e-05, "loss": 0.1674, "step": 26120 }, { "epoch": 1.22, "learning_rate": 2.764794907400184e-05, "loss": 0.1911, "step": 26125 }, { "epoch": 1.22, "learning_rate": 2.764748100416582e-05, "loss": 0.2752, "step": 26130 }, { "epoch": 1.22, "learning_rate": 2.7647012934329804e-05, "loss": 0.4117, "step": 26135 }, { "epoch": 1.22, "learning_rate": 2.7646544864493783e-05, "loss": 0.0602, "step": 26140 }, { "epoch": 1.22, "learning_rate": 2.7646076794657763e-05, "loss": 0.0713, "step": 26145 }, { "epoch": 1.22, "learning_rate": 2.7645608724821743e-05, "loss": 0.0761, "step": 26150 }, { "epoch": 1.22, "learning_rate": 2.7645140654985726e-05, "loss": 0.0877, "step": 26155 }, { "epoch": 1.22, "learning_rate": 2.7644672585149706e-05, "loss": 0.1915, "step": 26160 }, { "epoch": 1.22, "learning_rate": 2.7644204515313686e-05, "loss": 0.1301, "step": 26165 }, { "epoch": 1.22, "learning_rate": 2.764373644547767e-05, "loss": 0.1413, "step": 26170 }, { "epoch": 1.22, "learning_rate": 2.7643268375641646e-05, "loss": 0.2631, "step": 26175 }, { "epoch": 1.22, "learning_rate": 2.7642800305805625e-05, "loss": 0.3781, "step": 26180 }, { "epoch": 1.22, "learning_rate": 2.7642332235969605e-05, "loss": 0.2615, "step": 26185 }, { "epoch": 1.22, "learning_rate": 2.764186416613359e-05, "loss": 0.1036, "step": 26190 }, { "epoch": 1.22, "learning_rate": 2.7641396096297568e-05, "loss": 0.1248, "step": 26195 }, { "epoch": 1.22, "learning_rate": 2.7640928026461548e-05, "loss": 0.1448, "step": 26200 }, { "epoch": 1.22, "learning_rate": 2.7640459956625528e-05, "loss": 0.1155, "step": 26205 }, { "epoch": 1.22, "learning_rate": 2.763999188678951e-05, "loss": 0.1462, "step": 26210 }, { "epoch": 1.22, "learning_rate": 2.763952381695349e-05, "loss": 0.0997, "step": 26215 }, { "epoch": 1.22, "learning_rate": 2.763905574711747e-05, "loss": 0.238, "step": 26220 }, { "epoch": 1.22, "learning_rate": 2.7638587677281454e-05, "loss": 0.214, "step": 26225 }, { "epoch": 1.22, "learning_rate": 2.7638119607445434e-05, "loss": 0.2839, "step": 26230 }, { "epoch": 1.22, "learning_rate": 2.7637651537609413e-05, "loss": 0.2874, "step": 26235 }, { "epoch": 1.22, "learning_rate": 2.763718346777339e-05, "loss": 0.0405, "step": 26240 }, { "epoch": 1.22, "learning_rate": 2.7636715397937373e-05, "loss": 0.0466, "step": 26245 }, { "epoch": 1.22, "learning_rate": 2.7636247328101353e-05, "loss": 0.0732, "step": 26250 }, { "epoch": 1.23, "learning_rate": 2.7635779258265333e-05, "loss": 0.1521, "step": 26255 }, { "epoch": 1.23, "learning_rate": 2.7635311188429313e-05, "loss": 0.1306, "step": 26260 }, { "epoch": 1.23, "learning_rate": 2.7634843118593296e-05, "loss": 0.1572, "step": 26265 }, { "epoch": 1.23, "learning_rate": 2.7634375048757276e-05, "loss": 0.1458, "step": 26270 }, { "epoch": 1.23, "learning_rate": 2.7633906978921255e-05, "loss": 0.1574, "step": 26275 }, { "epoch": 1.23, "learning_rate": 2.7633438909085235e-05, "loss": 0.2982, "step": 26280 }, { "epoch": 1.23, "learning_rate": 2.763297083924922e-05, "loss": 0.2551, "step": 26285 }, { "epoch": 1.23, "learning_rate": 2.7632502769413198e-05, "loss": 0.0704, "step": 26290 }, { "epoch": 1.23, "learning_rate": 2.7632034699577178e-05, "loss": 0.0801, "step": 26295 }, { "epoch": 1.23, "learning_rate": 2.7631566629741158e-05, "loss": 0.099, "step": 26300 }, { "epoch": 1.23, "learning_rate": 2.7631098559905138e-05, "loss": 0.1362, "step": 26305 }, { "epoch": 1.23, "learning_rate": 2.7630630490069118e-05, "loss": 0.1717, "step": 26310 }, { "epoch": 1.23, "learning_rate": 2.7630162420233097e-05, "loss": 0.2557, "step": 26315 }, { "epoch": 1.23, "learning_rate": 2.762969435039708e-05, "loss": 0.158, "step": 26320 }, { "epoch": 1.23, "learning_rate": 2.762922628056106e-05, "loss": 0.2529, "step": 26325 }, { "epoch": 1.23, "learning_rate": 2.762875821072504e-05, "loss": 0.3616, "step": 26330 }, { "epoch": 1.23, "learning_rate": 2.762829014088902e-05, "loss": 0.2514, "step": 26335 }, { "epoch": 1.23, "learning_rate": 2.7627822071053003e-05, "loss": 0.0665, "step": 26340 }, { "epoch": 1.23, "learning_rate": 2.7627354001216983e-05, "loss": 0.0737, "step": 26345 }, { "epoch": 1.23, "learning_rate": 2.7626885931380963e-05, "loss": 0.1017, "step": 26350 }, { "epoch": 1.23, "learning_rate": 2.7626417861544946e-05, "loss": 0.0793, "step": 26355 }, { "epoch": 1.23, "learning_rate": 2.7625949791708926e-05, "loss": 0.1185, "step": 26360 }, { "epoch": 1.23, "learning_rate": 2.7625481721872902e-05, "loss": 0.1498, "step": 26365 }, { "epoch": 1.23, "learning_rate": 2.7625013652036882e-05, "loss": 0.0816, "step": 26370 }, { "epoch": 1.23, "learning_rate": 2.7624545582200865e-05, "loss": 0.2107, "step": 26375 }, { "epoch": 1.23, "learning_rate": 2.7624077512364845e-05, "loss": 0.361, "step": 26380 }, { "epoch": 1.23, "learning_rate": 2.7623609442528825e-05, "loss": 0.2103, "step": 26385 }, { "epoch": 1.23, "learning_rate": 2.7623141372692805e-05, "loss": 0.0649, "step": 26390 }, { "epoch": 1.23, "learning_rate": 2.7622673302856788e-05, "loss": 0.0719, "step": 26395 }, { "epoch": 1.23, "learning_rate": 2.7622205233020768e-05, "loss": 0.095, "step": 26400 }, { "epoch": 1.23, "learning_rate": 2.7621737163184748e-05, "loss": 0.1392, "step": 26405 }, { "epoch": 1.23, "learning_rate": 2.762126909334873e-05, "loss": 0.1261, "step": 26410 }, { "epoch": 1.23, "learning_rate": 2.762080102351271e-05, "loss": 0.2235, "step": 26415 }, { "epoch": 1.23, "learning_rate": 2.762033295367669e-05, "loss": 0.1772, "step": 26420 }, { "epoch": 1.23, "learning_rate": 2.761986488384067e-05, "loss": 0.2098, "step": 26425 }, { "epoch": 1.23, "learning_rate": 2.761939681400465e-05, "loss": 0.5498, "step": 26430 }, { "epoch": 1.23, "learning_rate": 2.761892874416863e-05, "loss": 0.283, "step": 26435 }, { "epoch": 1.23, "learning_rate": 2.761846067433261e-05, "loss": 0.0199, "step": 26440 }, { "epoch": 1.23, "learning_rate": 2.761799260449659e-05, "loss": 0.0774, "step": 26445 }, { "epoch": 1.23, "learning_rate": 2.7617524534660573e-05, "loss": 0.0804, "step": 26450 }, { "epoch": 1.23, "learning_rate": 2.7617056464824553e-05, "loss": 0.054, "step": 26455 }, { "epoch": 1.23, "learning_rate": 2.7616588394988532e-05, "loss": 0.1484, "step": 26460 }, { "epoch": 1.23, "learning_rate": 2.7616120325152512e-05, "loss": 0.1441, "step": 26465 }, { "epoch": 1.24, "learning_rate": 2.7615652255316495e-05, "loss": 0.18, "step": 26470 }, { "epoch": 1.24, "learning_rate": 2.7615184185480475e-05, "loss": 0.1979, "step": 26475 }, { "epoch": 1.24, "learning_rate": 2.7614716115644455e-05, "loss": 0.2328, "step": 26480 }, { "epoch": 1.24, "learning_rate": 2.7614248045808438e-05, "loss": 0.3263, "step": 26485 }, { "epoch": 1.24, "learning_rate": 2.7613779975972415e-05, "loss": 0.0549, "step": 26490 }, { "epoch": 1.24, "learning_rate": 2.7613311906136394e-05, "loss": 0.0571, "step": 26495 }, { "epoch": 1.24, "learning_rate": 2.7612843836300374e-05, "loss": 0.0743, "step": 26500 }, { "epoch": 1.24, "learning_rate": 2.7612375766464358e-05, "loss": 0.0909, "step": 26505 }, { "epoch": 1.24, "learning_rate": 2.7611907696628337e-05, "loss": 0.1282, "step": 26510 }, { "epoch": 1.24, "learning_rate": 2.7611439626792317e-05, "loss": 0.1184, "step": 26515 }, { "epoch": 1.24, "learning_rate": 2.7610971556956297e-05, "loss": 0.1973, "step": 26520 }, { "epoch": 1.24, "learning_rate": 2.761050348712028e-05, "loss": 0.1799, "step": 26525 }, { "epoch": 1.24, "learning_rate": 2.761003541728426e-05, "loss": 0.3093, "step": 26530 }, { "epoch": 1.24, "learning_rate": 2.760956734744824e-05, "loss": 0.3023, "step": 26535 }, { "epoch": 1.24, "learning_rate": 2.7609099277612223e-05, "loss": 0.054, "step": 26540 }, { "epoch": 1.24, "learning_rate": 2.7608631207776203e-05, "loss": 0.0806, "step": 26545 }, { "epoch": 1.24, "learning_rate": 2.7608163137940183e-05, "loss": 0.0825, "step": 26550 }, { "epoch": 1.24, "learning_rate": 2.760769506810416e-05, "loss": 0.0995, "step": 26555 }, { "epoch": 1.24, "learning_rate": 2.7607226998268142e-05, "loss": 0.0734, "step": 26560 }, { "epoch": 1.24, "learning_rate": 2.7606758928432122e-05, "loss": 0.1258, "step": 26565 }, { "epoch": 1.24, "learning_rate": 2.7606290858596102e-05, "loss": 0.2489, "step": 26570 }, { "epoch": 1.24, "learning_rate": 2.7605822788760082e-05, "loss": 0.2083, "step": 26575 }, { "epoch": 1.24, "learning_rate": 2.7605354718924065e-05, "loss": 0.3093, "step": 26580 }, { "epoch": 1.24, "learning_rate": 2.7604886649088045e-05, "loss": 0.3207, "step": 26585 }, { "epoch": 1.24, "learning_rate": 2.7604418579252025e-05, "loss": 0.0797, "step": 26590 }, { "epoch": 1.24, "learning_rate": 2.7603950509416008e-05, "loss": 0.1105, "step": 26595 }, { "epoch": 1.24, "learning_rate": 2.7603482439579988e-05, "loss": 0.0829, "step": 26600 }, { "epoch": 1.24, "learning_rate": 2.7603014369743967e-05, "loss": 0.097, "step": 26605 }, { "epoch": 1.24, "learning_rate": 2.7602546299907947e-05, "loss": 0.0403, "step": 26610 }, { "epoch": 1.24, "learning_rate": 2.760207823007193e-05, "loss": 0.1353, "step": 26615 }, { "epoch": 1.24, "learning_rate": 2.7601610160235907e-05, "loss": 0.2431, "step": 26620 }, { "epoch": 1.24, "learning_rate": 2.7601142090399887e-05, "loss": 0.2593, "step": 26625 }, { "epoch": 1.24, "learning_rate": 2.7600674020563867e-05, "loss": 0.4024, "step": 26630 }, { "epoch": 1.24, "learning_rate": 2.760020595072785e-05, "loss": 0.2005, "step": 26635 }, { "epoch": 1.24, "learning_rate": 2.759973788089183e-05, "loss": 0.0299, "step": 26640 }, { "epoch": 1.24, "learning_rate": 2.759926981105581e-05, "loss": 0.0994, "step": 26645 }, { "epoch": 1.24, "learning_rate": 2.759880174121979e-05, "loss": 0.0605, "step": 26650 }, { "epoch": 1.24, "learning_rate": 2.7598333671383772e-05, "loss": 0.1023, "step": 26655 }, { "epoch": 1.24, "learning_rate": 2.7597865601547752e-05, "loss": 0.0889, "step": 26660 }, { "epoch": 1.24, "learning_rate": 2.7597397531711732e-05, "loss": 0.1332, "step": 26665 }, { "epoch": 1.24, "learning_rate": 2.7596929461875715e-05, "loss": 0.2111, "step": 26670 }, { "epoch": 1.24, "learning_rate": 2.7596461392039695e-05, "loss": 0.2561, "step": 26675 }, { "epoch": 1.24, "learning_rate": 2.759599332220367e-05, "loss": 0.319, "step": 26680 }, { "epoch": 1.25, "learning_rate": 2.759552525236765e-05, "loss": 0.2339, "step": 26685 }, { "epoch": 1.25, "learning_rate": 2.7595057182531634e-05, "loss": 0.0774, "step": 26690 }, { "epoch": 1.25, "learning_rate": 2.7594589112695614e-05, "loss": 0.044, "step": 26695 }, { "epoch": 1.25, "learning_rate": 2.7594121042859594e-05, "loss": 0.0623, "step": 26700 }, { "epoch": 1.25, "learning_rate": 2.7593652973023574e-05, "loss": 0.132, "step": 26705 }, { "epoch": 1.25, "learning_rate": 2.7593184903187557e-05, "loss": 0.1076, "step": 26710 }, { "epoch": 1.25, "learning_rate": 2.7592716833351537e-05, "loss": 0.0853, "step": 26715 }, { "epoch": 1.25, "learning_rate": 2.7592248763515517e-05, "loss": 0.2748, "step": 26720 }, { "epoch": 1.25, "learning_rate": 2.75917806936795e-05, "loss": 0.2715, "step": 26725 }, { "epoch": 1.25, "learning_rate": 2.759131262384348e-05, "loss": 0.1765, "step": 26730 }, { "epoch": 1.25, "learning_rate": 2.759084455400746e-05, "loss": 0.3085, "step": 26735 }, { "epoch": 1.25, "learning_rate": 2.759037648417144e-05, "loss": 0.0501, "step": 26740 }, { "epoch": 1.25, "learning_rate": 2.758990841433542e-05, "loss": 0.0654, "step": 26745 }, { "epoch": 1.25, "learning_rate": 2.75894403444994e-05, "loss": 0.0644, "step": 26750 }, { "epoch": 1.25, "learning_rate": 2.758897227466338e-05, "loss": 0.0433, "step": 26755 }, { "epoch": 1.25, "learning_rate": 2.758850420482736e-05, "loss": 0.1507, "step": 26760 }, { "epoch": 1.25, "learning_rate": 2.7588036134991342e-05, "loss": 0.1453, "step": 26765 }, { "epoch": 1.25, "learning_rate": 2.7587568065155322e-05, "loss": 0.1428, "step": 26770 }, { "epoch": 1.25, "learning_rate": 2.75870999953193e-05, "loss": 0.1616, "step": 26775 }, { "epoch": 1.25, "learning_rate": 2.7586631925483285e-05, "loss": 0.2276, "step": 26780 }, { "epoch": 1.25, "learning_rate": 2.7586163855647265e-05, "loss": 0.2257, "step": 26785 }, { "epoch": 1.25, "learning_rate": 2.7585695785811244e-05, "loss": 0.0895, "step": 26790 }, { "epoch": 1.25, "learning_rate": 2.7585227715975224e-05, "loss": 0.1216, "step": 26795 }, { "epoch": 1.25, "learning_rate": 2.7584759646139207e-05, "loss": 0.0717, "step": 26800 }, { "epoch": 1.25, "learning_rate": 2.7584291576303187e-05, "loss": 0.0901, "step": 26805 }, { "epoch": 1.25, "learning_rate": 2.7583823506467164e-05, "loss": 0.1055, "step": 26810 }, { "epoch": 1.25, "learning_rate": 2.7583355436631143e-05, "loss": 0.1052, "step": 26815 }, { "epoch": 1.25, "learning_rate": 2.7582887366795127e-05, "loss": 0.1673, "step": 26820 }, { "epoch": 1.25, "learning_rate": 2.7582419296959107e-05, "loss": 0.2426, "step": 26825 }, { "epoch": 1.25, "learning_rate": 2.7581951227123086e-05, "loss": 0.3206, "step": 26830 }, { "epoch": 1.25, "learning_rate": 2.7581483157287066e-05, "loss": 0.3846, "step": 26835 }, { "epoch": 1.25, "learning_rate": 2.758101508745105e-05, "loss": 0.0558, "step": 26840 }, { "epoch": 1.25, "learning_rate": 2.758054701761503e-05, "loss": 0.133, "step": 26845 }, { "epoch": 1.25, "learning_rate": 2.758007894777901e-05, "loss": 0.0974, "step": 26850 }, { "epoch": 1.25, "learning_rate": 2.7579610877942992e-05, "loss": 0.1085, "step": 26855 }, { "epoch": 1.25, "learning_rate": 2.7579142808106972e-05, "loss": 0.1434, "step": 26860 }, { "epoch": 1.25, "learning_rate": 2.7578674738270952e-05, "loss": 0.159, "step": 26865 }, { "epoch": 1.25, "learning_rate": 2.7578206668434928e-05, "loss": 0.2478, "step": 26870 }, { "epoch": 1.25, "learning_rate": 2.757773859859891e-05, "loss": 0.2168, "step": 26875 }, { "epoch": 1.25, "learning_rate": 2.757727052876289e-05, "loss": 0.5468, "step": 26880 }, { "epoch": 1.25, "learning_rate": 2.757680245892687e-05, "loss": 0.2508, "step": 26885 }, { "epoch": 1.25, "learning_rate": 2.757633438909085e-05, "loss": 0.0252, "step": 26890 }, { "epoch": 1.25, "learning_rate": 2.7575866319254834e-05, "loss": 0.0561, "step": 26895 }, { "epoch": 1.26, "learning_rate": 2.7575398249418814e-05, "loss": 0.096, "step": 26900 }, { "epoch": 1.26, "learning_rate": 2.7574930179582794e-05, "loss": 0.1182, "step": 26905 }, { "epoch": 1.26, "learning_rate": 2.7574462109746777e-05, "loss": 0.227, "step": 26910 }, { "epoch": 1.26, "learning_rate": 2.7573994039910757e-05, "loss": 0.2021, "step": 26915 }, { "epoch": 1.26, "learning_rate": 2.7573525970074737e-05, "loss": 0.2733, "step": 26920 }, { "epoch": 1.26, "learning_rate": 2.7573057900238716e-05, "loss": 0.284, "step": 26925 }, { "epoch": 1.26, "learning_rate": 2.75725898304027e-05, "loss": 0.4521, "step": 26930 }, { "epoch": 1.26, "learning_rate": 2.7572121760566676e-05, "loss": 0.2281, "step": 26935 }, { "epoch": 1.26, "learning_rate": 2.7571653690730656e-05, "loss": 0.1074, "step": 26940 }, { "epoch": 1.26, "learning_rate": 2.7571185620894636e-05, "loss": 0.1161, "step": 26945 }, { "epoch": 1.26, "learning_rate": 2.757071755105862e-05, "loss": 0.0888, "step": 26950 }, { "epoch": 1.26, "learning_rate": 2.75702494812226e-05, "loss": 0.1003, "step": 26955 }, { "epoch": 1.26, "learning_rate": 2.756978141138658e-05, "loss": 0.1212, "step": 26960 }, { "epoch": 1.26, "learning_rate": 2.7569313341550562e-05, "loss": 0.1769, "step": 26965 }, { "epoch": 1.26, "learning_rate": 2.756884527171454e-05, "loss": 0.2336, "step": 26970 }, { "epoch": 1.26, "learning_rate": 2.756837720187852e-05, "loss": 0.2811, "step": 26975 }, { "epoch": 1.26, "learning_rate": 2.75679091320425e-05, "loss": 0.3899, "step": 26980 }, { "epoch": 1.26, "learning_rate": 2.7567441062206484e-05, "loss": 0.2493, "step": 26985 }, { "epoch": 1.26, "learning_rate": 2.7566972992370464e-05, "loss": 0.0727, "step": 26990 }, { "epoch": 1.26, "learning_rate": 2.756650492253444e-05, "loss": 0.0782, "step": 26995 }, { "epoch": 1.26, "learning_rate": 2.756603685269842e-05, "loss": 0.0617, "step": 27000 }, { "epoch": 1.26, "learning_rate": 2.7565568782862404e-05, "loss": 0.1509, "step": 27005 }, { "epoch": 1.26, "learning_rate": 2.7565100713026383e-05, "loss": 0.1272, "step": 27010 }, { "epoch": 1.26, "learning_rate": 2.7564632643190363e-05, "loss": 0.1887, "step": 27015 }, { "epoch": 1.26, "learning_rate": 2.7564164573354346e-05, "loss": 0.1928, "step": 27020 }, { "epoch": 1.26, "learning_rate": 2.7563696503518326e-05, "loss": 0.206, "step": 27025 }, { "epoch": 1.26, "learning_rate": 2.7563228433682306e-05, "loss": 0.3126, "step": 27030 }, { "epoch": 1.26, "learning_rate": 2.7562760363846286e-05, "loss": 0.2151, "step": 27035 }, { "epoch": 1.26, "learning_rate": 2.756229229401027e-05, "loss": 0.0827, "step": 27040 }, { "epoch": 1.26, "learning_rate": 2.756182422417425e-05, "loss": 0.0308, "step": 27045 }, { "epoch": 1.26, "learning_rate": 2.756135615433823e-05, "loss": 0.0792, "step": 27050 }, { "epoch": 1.26, "learning_rate": 2.756088808450221e-05, "loss": 0.1393, "step": 27055 }, { "epoch": 1.26, "learning_rate": 2.756042001466619e-05, "loss": 0.096, "step": 27060 }, { "epoch": 1.26, "learning_rate": 2.7559951944830168e-05, "loss": 0.1089, "step": 27065 }, { "epoch": 1.26, "learning_rate": 2.7559483874994148e-05, "loss": 0.1666, "step": 27070 }, { "epoch": 1.26, "learning_rate": 2.7559015805158128e-05, "loss": 0.2164, "step": 27075 }, { "epoch": 1.26, "learning_rate": 2.755854773532211e-05, "loss": 0.3704, "step": 27080 }, { "epoch": 1.26, "learning_rate": 2.755807966548609e-05, "loss": 0.218, "step": 27085 }, { "epoch": 1.26, "learning_rate": 2.755761159565007e-05, "loss": 0.0739, "step": 27090 }, { "epoch": 1.26, "learning_rate": 2.7557143525814054e-05, "loss": 0.1344, "step": 27095 }, { "epoch": 1.26, "learning_rate": 2.7556675455978034e-05, "loss": 0.1578, "step": 27100 }, { "epoch": 1.26, "learning_rate": 2.7556207386142014e-05, "loss": 0.2067, "step": 27105 }, { "epoch": 1.26, "learning_rate": 2.7555739316305993e-05, "loss": 0.1024, "step": 27110 }, { "epoch": 1.27, "learning_rate": 2.7555271246469977e-05, "loss": 0.1475, "step": 27115 }, { "epoch": 1.27, "learning_rate": 2.7554803176633956e-05, "loss": 0.1925, "step": 27120 }, { "epoch": 1.27, "learning_rate": 2.7554335106797933e-05, "loss": 0.3717, "step": 27125 }, { "epoch": 1.27, "learning_rate": 2.7553867036961913e-05, "loss": 0.3782, "step": 27130 }, { "epoch": 1.27, "learning_rate": 2.7553398967125896e-05, "loss": 0.2524, "step": 27135 }, { "epoch": 1.27, "learning_rate": 2.7552930897289876e-05, "loss": 0.0635, "step": 27140 }, { "epoch": 1.27, "learning_rate": 2.7552462827453855e-05, "loss": 0.0427, "step": 27145 }, { "epoch": 1.27, "learning_rate": 2.755199475761784e-05, "loss": 0.0806, "step": 27150 }, { "epoch": 1.27, "learning_rate": 2.755152668778182e-05, "loss": 0.1157, "step": 27155 }, { "epoch": 1.27, "learning_rate": 2.75510586179458e-05, "loss": 0.1437, "step": 27160 }, { "epoch": 1.27, "learning_rate": 2.7550590548109778e-05, "loss": 0.1757, "step": 27165 }, { "epoch": 1.27, "learning_rate": 2.755012247827376e-05, "loss": 0.1635, "step": 27170 }, { "epoch": 1.27, "learning_rate": 2.754965440843774e-05, "loss": 0.2316, "step": 27175 }, { "epoch": 1.27, "learning_rate": 2.754918633860172e-05, "loss": 0.3009, "step": 27180 }, { "epoch": 1.27, "learning_rate": 2.7548718268765697e-05, "loss": 0.2999, "step": 27185 }, { "epoch": 1.27, "learning_rate": 2.754825019892968e-05, "loss": 0.1279, "step": 27190 }, { "epoch": 1.27, "learning_rate": 2.754778212909366e-05, "loss": 0.0824, "step": 27195 }, { "epoch": 1.27, "learning_rate": 2.754731405925764e-05, "loss": 0.0984, "step": 27200 }, { "epoch": 1.27, "learning_rate": 2.7546845989421623e-05, "loss": 0.0874, "step": 27205 }, { "epoch": 1.27, "learning_rate": 2.7546377919585603e-05, "loss": 0.0781, "step": 27210 }, { "epoch": 1.27, "learning_rate": 2.7545909849749583e-05, "loss": 0.0944, "step": 27215 }, { "epoch": 1.27, "learning_rate": 2.7545441779913563e-05, "loss": 0.1486, "step": 27220 }, { "epoch": 1.27, "learning_rate": 2.7544973710077546e-05, "loss": 0.2477, "step": 27225 }, { "epoch": 1.27, "learning_rate": 2.7544505640241526e-05, "loss": 0.3283, "step": 27230 }, { "epoch": 1.27, "learning_rate": 2.7544037570405506e-05, "loss": 0.3542, "step": 27235 }, { "epoch": 1.27, "learning_rate": 2.7543569500569486e-05, "loss": 0.0884, "step": 27240 }, { "epoch": 1.27, "learning_rate": 2.754310143073347e-05, "loss": 0.0939, "step": 27245 }, { "epoch": 1.27, "learning_rate": 2.7542633360897445e-05, "loss": 0.1591, "step": 27250 }, { "epoch": 1.27, "learning_rate": 2.7542165291061425e-05, "loss": 0.0898, "step": 27255 }, { "epoch": 1.27, "learning_rate": 2.7541697221225405e-05, "loss": 0.1615, "step": 27260 }, { "epoch": 1.27, "learning_rate": 2.7541229151389388e-05, "loss": 0.1586, "step": 27265 }, { "epoch": 1.27, "learning_rate": 2.7540761081553368e-05, "loss": 0.1517, "step": 27270 }, { "epoch": 1.27, "learning_rate": 2.7540293011717348e-05, "loss": 0.1381, "step": 27275 }, { "epoch": 1.27, "learning_rate": 2.753982494188133e-05, "loss": 0.3747, "step": 27280 }, { "epoch": 1.27, "learning_rate": 2.753935687204531e-05, "loss": 0.3374, "step": 27285 }, { "epoch": 1.27, "learning_rate": 2.753888880220929e-05, "loss": 0.0837, "step": 27290 }, { "epoch": 1.27, "learning_rate": 2.753842073237327e-05, "loss": 0.097, "step": 27295 }, { "epoch": 1.27, "learning_rate": 2.7537952662537254e-05, "loss": 0.118, "step": 27300 }, { "epoch": 1.27, "learning_rate": 2.7537484592701233e-05, "loss": 0.0809, "step": 27305 }, { "epoch": 1.27, "learning_rate": 2.7537016522865213e-05, "loss": 0.1227, "step": 27310 }, { "epoch": 1.27, "learning_rate": 2.753654845302919e-05, "loss": 0.2199, "step": 27315 }, { "epoch": 1.27, "learning_rate": 2.7536080383193173e-05, "loss": 0.2727, "step": 27320 }, { "epoch": 1.28, "learning_rate": 2.7535612313357153e-05, "loss": 0.1952, "step": 27325 }, { "epoch": 1.28, "learning_rate": 2.7535144243521132e-05, "loss": 0.3614, "step": 27330 }, { "epoch": 1.28, "learning_rate": 2.7534676173685116e-05, "loss": 0.2547, "step": 27335 }, { "epoch": 1.28, "learning_rate": 2.7534208103849095e-05, "loss": 0.0119, "step": 27340 }, { "epoch": 1.28, "learning_rate": 2.7533740034013075e-05, "loss": 0.0574, "step": 27345 }, { "epoch": 1.28, "learning_rate": 2.7533271964177055e-05, "loss": 0.0443, "step": 27350 }, { "epoch": 1.28, "learning_rate": 2.753280389434104e-05, "loss": 0.0841, "step": 27355 }, { "epoch": 1.28, "learning_rate": 2.7532335824505018e-05, "loss": 0.0898, "step": 27360 }, { "epoch": 1.28, "learning_rate": 2.7531867754668998e-05, "loss": 0.1118, "step": 27365 }, { "epoch": 1.28, "learning_rate": 2.7531399684832978e-05, "loss": 0.1412, "step": 27370 }, { "epoch": 1.28, "learning_rate": 2.7530931614996958e-05, "loss": 0.2171, "step": 27375 }, { "epoch": 1.28, "learning_rate": 2.7530463545160937e-05, "loss": 0.365, "step": 27380 }, { "epoch": 1.28, "learning_rate": 2.7529995475324917e-05, "loss": 0.208, "step": 27385 }, { "epoch": 1.28, "learning_rate": 2.75295274054889e-05, "loss": 0.0513, "step": 27390 }, { "epoch": 1.28, "learning_rate": 2.752905933565288e-05, "loss": 0.1209, "step": 27395 }, { "epoch": 1.28, "learning_rate": 2.752859126581686e-05, "loss": 0.0542, "step": 27400 }, { "epoch": 1.28, "learning_rate": 2.752812319598084e-05, "loss": 0.1033, "step": 27405 }, { "epoch": 1.28, "learning_rate": 2.7527655126144823e-05, "loss": 0.1326, "step": 27410 }, { "epoch": 1.28, "learning_rate": 2.7527187056308803e-05, "loss": 0.1027, "step": 27415 }, { "epoch": 1.28, "learning_rate": 2.7526718986472783e-05, "loss": 0.1504, "step": 27420 }, { "epoch": 1.28, "learning_rate": 2.7526250916636763e-05, "loss": 0.1953, "step": 27425 }, { "epoch": 1.28, "learning_rate": 2.7525782846800746e-05, "loss": 0.3413, "step": 27430 }, { "epoch": 1.28, "learning_rate": 2.7525314776964726e-05, "loss": 0.3712, "step": 27435 }, { "epoch": 1.28, "learning_rate": 2.7524846707128702e-05, "loss": 0.0262, "step": 27440 }, { "epoch": 1.28, "learning_rate": 2.7524378637292682e-05, "loss": 0.0384, "step": 27445 }, { "epoch": 1.28, "learning_rate": 2.7523910567456665e-05, "loss": 0.0576, "step": 27450 }, { "epoch": 1.28, "learning_rate": 2.7523442497620645e-05, "loss": 0.0831, "step": 27455 }, { "epoch": 1.28, "learning_rate": 2.7522974427784625e-05, "loss": 0.1302, "step": 27460 }, { "epoch": 1.28, "learning_rate": 2.7522506357948608e-05, "loss": 0.1653, "step": 27465 }, { "epoch": 1.28, "learning_rate": 2.7522038288112588e-05, "loss": 0.1907, "step": 27470 }, { "epoch": 1.28, "learning_rate": 2.7521570218276568e-05, "loss": 0.2658, "step": 27475 }, { "epoch": 1.28, "learning_rate": 2.7521102148440547e-05, "loss": 0.3106, "step": 27480 }, { "epoch": 1.28, "learning_rate": 2.752063407860453e-05, "loss": 0.331, "step": 27485 }, { "epoch": 1.28, "learning_rate": 2.752016600876851e-05, "loss": 0.0404, "step": 27490 }, { "epoch": 1.28, "learning_rate": 2.751969793893249e-05, "loss": 0.0581, "step": 27495 }, { "epoch": 1.28, "learning_rate": 2.751922986909647e-05, "loss": 0.1581, "step": 27500 }, { "epoch": 1.28, "learning_rate": 2.751876179926045e-05, "loss": 0.1493, "step": 27505 }, { "epoch": 1.28, "learning_rate": 2.751829372942443e-05, "loss": 0.102, "step": 27510 }, { "epoch": 1.28, "learning_rate": 2.751782565958841e-05, "loss": 0.1639, "step": 27515 }, { "epoch": 1.28, "learning_rate": 2.7517357589752393e-05, "loss": 0.0845, "step": 27520 }, { "epoch": 1.28, "learning_rate": 2.7516889519916372e-05, "loss": 0.2207, "step": 27525 }, { "epoch": 1.28, "learning_rate": 2.7516421450080352e-05, "loss": 0.3716, "step": 27530 }, { "epoch": 1.28, "learning_rate": 2.7515953380244332e-05, "loss": 0.2371, "step": 27535 }, { "epoch": 1.29, "learning_rate": 2.7515485310408315e-05, "loss": 0.0889, "step": 27540 }, { "epoch": 1.29, "learning_rate": 2.7515017240572295e-05, "loss": 0.0583, "step": 27545 }, { "epoch": 1.29, "learning_rate": 2.7514549170736275e-05, "loss": 0.1231, "step": 27550 }, { "epoch": 1.29, "learning_rate": 2.7514081100900255e-05, "loss": 0.1132, "step": 27555 }, { "epoch": 1.29, "learning_rate": 2.7513613031064238e-05, "loss": 0.155, "step": 27560 }, { "epoch": 1.29, "learning_rate": 2.7513144961228214e-05, "loss": 0.1438, "step": 27565 }, { "epoch": 1.29, "learning_rate": 2.7512676891392194e-05, "loss": 0.2467, "step": 27570 }, { "epoch": 1.29, "learning_rate": 2.7512208821556177e-05, "loss": 0.3728, "step": 27575 }, { "epoch": 1.29, "learning_rate": 2.7511740751720157e-05, "loss": 0.386, "step": 27580 }, { "epoch": 1.29, "learning_rate": 2.7511272681884137e-05, "loss": 0.2282, "step": 27585 }, { "epoch": 1.29, "learning_rate": 2.7510804612048117e-05, "loss": 0.0428, "step": 27590 }, { "epoch": 1.29, "learning_rate": 2.75103365422121e-05, "loss": 0.0724, "step": 27595 }, { "epoch": 1.29, "learning_rate": 2.750986847237608e-05, "loss": 0.0553, "step": 27600 }, { "epoch": 1.29, "learning_rate": 2.750940040254006e-05, "loss": 0.1176, "step": 27605 }, { "epoch": 1.29, "learning_rate": 2.750893233270404e-05, "loss": 0.1243, "step": 27610 }, { "epoch": 1.29, "learning_rate": 2.7508464262868023e-05, "loss": 0.1583, "step": 27615 }, { "epoch": 1.29, "learning_rate": 2.7507996193032003e-05, "loss": 0.1511, "step": 27620 }, { "epoch": 1.29, "learning_rate": 2.7507528123195982e-05, "loss": 0.211, "step": 27625 }, { "epoch": 1.29, "learning_rate": 2.7507060053359962e-05, "loss": 0.3121, "step": 27630 }, { "epoch": 1.29, "learning_rate": 2.7506591983523942e-05, "loss": 0.2638, "step": 27635 }, { "epoch": 1.29, "learning_rate": 2.7506123913687922e-05, "loss": 0.0356, "step": 27640 }, { "epoch": 1.29, "learning_rate": 2.75056558438519e-05, "loss": 0.0703, "step": 27645 }, { "epoch": 1.29, "learning_rate": 2.7505187774015885e-05, "loss": 0.085, "step": 27650 }, { "epoch": 1.29, "learning_rate": 2.7504719704179865e-05, "loss": 0.0835, "step": 27655 }, { "epoch": 1.29, "learning_rate": 2.7504251634343844e-05, "loss": 0.0965, "step": 27660 }, { "epoch": 1.29, "learning_rate": 2.7503783564507824e-05, "loss": 0.1107, "step": 27665 }, { "epoch": 1.29, "learning_rate": 2.7503315494671807e-05, "loss": 0.1891, "step": 27670 }, { "epoch": 1.29, "learning_rate": 2.7502847424835787e-05, "loss": 0.3491, "step": 27675 }, { "epoch": 1.29, "learning_rate": 2.7502379354999767e-05, "loss": 0.1841, "step": 27680 }, { "epoch": 1.29, "learning_rate": 2.7501911285163747e-05, "loss": 0.2994, "step": 27685 }, { "epoch": 1.29, "learning_rate": 2.7501443215327727e-05, "loss": 0.0297, "step": 27690 }, { "epoch": 1.29, "learning_rate": 2.7500975145491707e-05, "loss": 0.1179, "step": 27695 }, { "epoch": 1.29, "learning_rate": 2.7500507075655686e-05, "loss": 0.0643, "step": 27700 }, { "epoch": 1.29, "learning_rate": 2.750003900581967e-05, "loss": 0.073, "step": 27705 }, { "epoch": 1.29, "learning_rate": 2.749957093598365e-05, "loss": 0.0744, "step": 27710 }, { "epoch": 1.29, "learning_rate": 2.749910286614763e-05, "loss": 0.1319, "step": 27715 }, { "epoch": 1.29, "learning_rate": 2.749863479631161e-05, "loss": 0.3513, "step": 27720 }, { "epoch": 1.29, "learning_rate": 2.7498166726475592e-05, "loss": 0.1999, "step": 27725 }, { "epoch": 1.29, "learning_rate": 2.7497698656639572e-05, "loss": 0.5361, "step": 27730 }, { "epoch": 1.29, "learning_rate": 2.7497230586803552e-05, "loss": 0.2125, "step": 27735 }, { "epoch": 1.29, "learning_rate": 2.7496762516967532e-05, "loss": 0.0869, "step": 27740 }, { "epoch": 1.29, "learning_rate": 2.7496294447131515e-05, "loss": 0.0586, "step": 27745 }, { "epoch": 1.29, "learning_rate": 2.7495826377295495e-05, "loss": 0.0497, "step": 27750 }, { "epoch": 1.3, "learning_rate": 2.749535830745947e-05, "loss": 0.112, "step": 27755 }, { "epoch": 1.3, "learning_rate": 2.7494890237623454e-05, "loss": 0.135, "step": 27760 }, { "epoch": 1.3, "learning_rate": 2.7494422167787434e-05, "loss": 0.0718, "step": 27765 }, { "epoch": 1.3, "learning_rate": 2.7493954097951414e-05, "loss": 0.1493, "step": 27770 }, { "epoch": 1.3, "learning_rate": 2.7493486028115394e-05, "loss": 0.1885, "step": 27775 }, { "epoch": 1.3, "learning_rate": 2.7493017958279377e-05, "loss": 0.3673, "step": 27780 }, { "epoch": 1.3, "learning_rate": 2.7492549888443357e-05, "loss": 0.3758, "step": 27785 }, { "epoch": 1.3, "learning_rate": 2.7492081818607337e-05, "loss": 0.0475, "step": 27790 }, { "epoch": 1.3, "learning_rate": 2.7491613748771316e-05, "loss": 0.0395, "step": 27795 }, { "epoch": 1.3, "learning_rate": 2.74911456789353e-05, "loss": 0.137, "step": 27800 }, { "epoch": 1.3, "learning_rate": 2.749067760909928e-05, "loss": 0.1077, "step": 27805 }, { "epoch": 1.3, "learning_rate": 2.749020953926326e-05, "loss": 0.1306, "step": 27810 }, { "epoch": 1.3, "learning_rate": 2.7489741469427243e-05, "loss": 0.0834, "step": 27815 }, { "epoch": 1.3, "learning_rate": 2.748927339959122e-05, "loss": 0.1719, "step": 27820 }, { "epoch": 1.3, "learning_rate": 2.74888053297552e-05, "loss": 0.2288, "step": 27825 }, { "epoch": 1.3, "learning_rate": 2.748833725991918e-05, "loss": 0.2699, "step": 27830 }, { "epoch": 1.3, "learning_rate": 2.7487869190083162e-05, "loss": 0.3177, "step": 27835 }, { "epoch": 1.3, "learning_rate": 2.748740112024714e-05, "loss": 0.0738, "step": 27840 }, { "epoch": 1.3, "learning_rate": 2.748693305041112e-05, "loss": 0.0389, "step": 27845 }, { "epoch": 1.3, "learning_rate": 2.74864649805751e-05, "loss": 0.0625, "step": 27850 }, { "epoch": 1.3, "learning_rate": 2.7485996910739084e-05, "loss": 0.1434, "step": 27855 }, { "epoch": 1.3, "learning_rate": 2.7485528840903064e-05, "loss": 0.1452, "step": 27860 }, { "epoch": 1.3, "learning_rate": 2.7485060771067044e-05, "loss": 0.1811, "step": 27865 }, { "epoch": 1.3, "learning_rate": 2.7484592701231024e-05, "loss": 0.2638, "step": 27870 }, { "epoch": 1.3, "learning_rate": 2.7484124631395007e-05, "loss": 0.34, "step": 27875 }, { "epoch": 1.3, "learning_rate": 2.7483656561558984e-05, "loss": 0.328, "step": 27880 }, { "epoch": 1.3, "learning_rate": 2.7483188491722963e-05, "loss": 0.2302, "step": 27885 }, { "epoch": 1.3, "learning_rate": 2.7482720421886947e-05, "loss": 0.0415, "step": 27890 }, { "epoch": 1.3, "learning_rate": 2.7482252352050926e-05, "loss": 0.0742, "step": 27895 }, { "epoch": 1.3, "learning_rate": 2.7481784282214906e-05, "loss": 0.1024, "step": 27900 }, { "epoch": 1.3, "learning_rate": 2.7481316212378886e-05, "loss": 0.1901, "step": 27905 }, { "epoch": 1.3, "learning_rate": 2.748084814254287e-05, "loss": 0.1882, "step": 27910 }, { "epoch": 1.3, "learning_rate": 2.748038007270685e-05, "loss": 0.1124, "step": 27915 }, { "epoch": 1.3, "learning_rate": 2.747991200287083e-05, "loss": 0.2339, "step": 27920 }, { "epoch": 1.3, "learning_rate": 2.747944393303481e-05, "loss": 0.2302, "step": 27925 }, { "epoch": 1.3, "learning_rate": 2.7478975863198792e-05, "loss": 0.2955, "step": 27930 }, { "epoch": 1.3, "learning_rate": 2.7478507793362772e-05, "loss": 0.2843, "step": 27935 }, { "epoch": 1.3, "learning_rate": 2.747803972352675e-05, "loss": 0.0291, "step": 27940 }, { "epoch": 1.3, "learning_rate": 2.747757165369073e-05, "loss": 0.0432, "step": 27945 }, { "epoch": 1.3, "learning_rate": 2.747710358385471e-05, "loss": 0.066, "step": 27950 }, { "epoch": 1.3, "learning_rate": 2.747663551401869e-05, "loss": 0.1124, "step": 27955 }, { "epoch": 1.3, "learning_rate": 2.747616744418267e-05, "loss": 0.0902, "step": 27960 }, { "epoch": 1.3, "learning_rate": 2.7475699374346654e-05, "loss": 0.0392, "step": 27965 }, { "epoch": 1.31, "learning_rate": 2.7475231304510634e-05, "loss": 0.1585, "step": 27970 }, { "epoch": 1.31, "learning_rate": 2.7474763234674614e-05, "loss": 0.1677, "step": 27975 }, { "epoch": 1.31, "learning_rate": 2.7474295164838593e-05, "loss": 0.2895, "step": 27980 }, { "epoch": 1.31, "learning_rate": 2.7473827095002577e-05, "loss": 0.2902, "step": 27985 }, { "epoch": 1.31, "learning_rate": 2.7473359025166556e-05, "loss": 0.1316, "step": 27990 }, { "epoch": 1.31, "learning_rate": 2.7472890955330536e-05, "loss": 0.0687, "step": 27995 }, { "epoch": 1.31, "learning_rate": 2.747242288549452e-05, "loss": 0.1078, "step": 28000 }, { "epoch": 1.31, "learning_rate": 2.74719548156585e-05, "loss": 0.1177, "step": 28005 }, { "epoch": 1.31, "learning_rate": 2.7471486745822476e-05, "loss": 0.1082, "step": 28010 }, { "epoch": 1.31, "learning_rate": 2.7471018675986456e-05, "loss": 0.089, "step": 28015 }, { "epoch": 1.31, "learning_rate": 2.747055060615044e-05, "loss": 0.249, "step": 28020 }, { "epoch": 1.31, "learning_rate": 2.747008253631442e-05, "loss": 0.1919, "step": 28025 }, { "epoch": 1.31, "learning_rate": 2.74696144664784e-05, "loss": 0.2358, "step": 28030 }, { "epoch": 1.31, "learning_rate": 2.7469146396642378e-05, "loss": 0.3767, "step": 28035 }, { "epoch": 1.31, "learning_rate": 2.746867832680636e-05, "loss": 0.099, "step": 28040 }, { "epoch": 1.31, "learning_rate": 2.746821025697034e-05, "loss": 0.0876, "step": 28045 }, { "epoch": 1.31, "learning_rate": 2.746774218713432e-05, "loss": 0.1126, "step": 28050 }, { "epoch": 1.31, "learning_rate": 2.74672741172983e-05, "loss": 0.1157, "step": 28055 }, { "epoch": 1.31, "learning_rate": 2.7466806047462284e-05, "loss": 0.084, "step": 28060 }, { "epoch": 1.31, "learning_rate": 2.7466337977626264e-05, "loss": 0.1668, "step": 28065 }, { "epoch": 1.31, "learning_rate": 2.746586990779024e-05, "loss": 0.1968, "step": 28070 }, { "epoch": 1.31, "learning_rate": 2.7465401837954224e-05, "loss": 0.2437, "step": 28075 }, { "epoch": 1.31, "learning_rate": 2.7464933768118203e-05, "loss": 0.3091, "step": 28080 }, { "epoch": 1.31, "learning_rate": 2.7464465698282183e-05, "loss": 0.2536, "step": 28085 }, { "epoch": 1.31, "learning_rate": 2.7463997628446163e-05, "loss": 0.0452, "step": 28090 }, { "epoch": 1.31, "learning_rate": 2.7463529558610146e-05, "loss": 0.0926, "step": 28095 }, { "epoch": 1.31, "learning_rate": 2.7463061488774126e-05, "loss": 0.0483, "step": 28100 }, { "epoch": 1.31, "learning_rate": 2.7462593418938106e-05, "loss": 0.0627, "step": 28105 }, { "epoch": 1.31, "learning_rate": 2.7462125349102086e-05, "loss": 0.1334, "step": 28110 }, { "epoch": 1.31, "learning_rate": 2.746165727926607e-05, "loss": 0.1587, "step": 28115 }, { "epoch": 1.31, "learning_rate": 2.746118920943005e-05, "loss": 0.1338, "step": 28120 }, { "epoch": 1.31, "learning_rate": 2.746072113959403e-05, "loss": 0.2125, "step": 28125 }, { "epoch": 1.31, "learning_rate": 2.7460253069758012e-05, "loss": 0.2883, "step": 28130 }, { "epoch": 1.31, "learning_rate": 2.7459784999921988e-05, "loss": 0.2835, "step": 28135 }, { "epoch": 1.31, "learning_rate": 2.7459316930085968e-05, "loss": 0.0499, "step": 28140 }, { "epoch": 1.31, "learning_rate": 2.7458848860249948e-05, "loss": 0.0252, "step": 28145 }, { "epoch": 1.31, "learning_rate": 2.745838079041393e-05, "loss": 0.1248, "step": 28150 }, { "epoch": 1.31, "learning_rate": 2.745791272057791e-05, "loss": 0.1053, "step": 28155 }, { "epoch": 1.31, "learning_rate": 2.745744465074189e-05, "loss": 0.1143, "step": 28160 }, { "epoch": 1.31, "learning_rate": 2.745697658090587e-05, "loss": 0.2063, "step": 28165 }, { "epoch": 1.31, "learning_rate": 2.7456508511069854e-05, "loss": 0.2292, "step": 28170 }, { "epoch": 1.31, "learning_rate": 2.7456040441233833e-05, "loss": 0.2822, "step": 28175 }, { "epoch": 1.31, "learning_rate": 2.7455572371397813e-05, "loss": 0.5754, "step": 28180 }, { "epoch": 1.32, "learning_rate": 2.7455104301561796e-05, "loss": 0.2625, "step": 28185 }, { "epoch": 1.32, "learning_rate": 2.7454636231725776e-05, "loss": 0.0868, "step": 28190 }, { "epoch": 1.32, "learning_rate": 2.7454168161889756e-05, "loss": 0.0331, "step": 28195 }, { "epoch": 1.32, "learning_rate": 2.7453700092053733e-05, "loss": 0.097, "step": 28200 }, { "epoch": 1.32, "learning_rate": 2.7453232022217716e-05, "loss": 0.0772, "step": 28205 }, { "epoch": 1.32, "learning_rate": 2.7452763952381696e-05, "loss": 0.1231, "step": 28210 }, { "epoch": 1.32, "learning_rate": 2.7452295882545675e-05, "loss": 0.0905, "step": 28215 }, { "epoch": 1.32, "learning_rate": 2.7451827812709655e-05, "loss": 0.2014, "step": 28220 }, { "epoch": 1.32, "learning_rate": 2.745135974287364e-05, "loss": 0.2087, "step": 28225 }, { "epoch": 1.32, "learning_rate": 2.7450891673037618e-05, "loss": 0.3961, "step": 28230 }, { "epoch": 1.32, "learning_rate": 2.7450423603201598e-05, "loss": 0.2582, "step": 28235 }, { "epoch": 1.32, "learning_rate": 2.744995553336558e-05, "loss": 0.0715, "step": 28240 }, { "epoch": 1.32, "learning_rate": 2.744948746352956e-05, "loss": 0.083, "step": 28245 }, { "epoch": 1.32, "learning_rate": 2.744901939369354e-05, "loss": 0.1078, "step": 28250 }, { "epoch": 1.32, "learning_rate": 2.744855132385752e-05, "loss": 0.0949, "step": 28255 }, { "epoch": 1.32, "learning_rate": 2.74480832540215e-05, "loss": 0.1244, "step": 28260 }, { "epoch": 1.32, "learning_rate": 2.744761518418548e-05, "loss": 0.188, "step": 28265 }, { "epoch": 1.32, "learning_rate": 2.744714711434946e-05, "loss": 0.2029, "step": 28270 }, { "epoch": 1.32, "learning_rate": 2.744667904451344e-05, "loss": 0.2647, "step": 28275 }, { "epoch": 1.32, "learning_rate": 2.7446210974677423e-05, "loss": 0.3943, "step": 28280 }, { "epoch": 1.32, "learning_rate": 2.7445742904841403e-05, "loss": 0.4138, "step": 28285 }, { "epoch": 1.32, "learning_rate": 2.7445274835005383e-05, "loss": 0.0202, "step": 28290 }, { "epoch": 1.32, "learning_rate": 2.7444806765169363e-05, "loss": 0.0484, "step": 28295 }, { "epoch": 1.32, "learning_rate": 2.7444338695333346e-05, "loss": 0.0667, "step": 28300 }, { "epoch": 1.32, "learning_rate": 2.7443870625497326e-05, "loss": 0.0925, "step": 28305 }, { "epoch": 1.32, "learning_rate": 2.7443402555661305e-05, "loss": 0.181, "step": 28310 }, { "epoch": 1.32, "learning_rate": 2.744293448582529e-05, "loss": 0.1262, "step": 28315 }, { "epoch": 1.32, "learning_rate": 2.744246641598927e-05, "loss": 0.2171, "step": 28320 }, { "epoch": 1.32, "learning_rate": 2.7441998346153245e-05, "loss": 0.1984, "step": 28325 }, { "epoch": 1.32, "learning_rate": 2.7441530276317225e-05, "loss": 0.3275, "step": 28330 }, { "epoch": 1.32, "learning_rate": 2.7441062206481208e-05, "loss": 0.2852, "step": 28335 }, { "epoch": 1.32, "learning_rate": 2.7440594136645188e-05, "loss": 0.062, "step": 28340 }, { "epoch": 1.32, "learning_rate": 2.7440126066809168e-05, "loss": 0.0677, "step": 28345 }, { "epoch": 1.32, "learning_rate": 2.7439657996973147e-05, "loss": 0.1333, "step": 28350 }, { "epoch": 1.32, "learning_rate": 2.743918992713713e-05, "loss": 0.128, "step": 28355 }, { "epoch": 1.32, "learning_rate": 2.743872185730111e-05, "loss": 0.1954, "step": 28360 }, { "epoch": 1.32, "learning_rate": 2.743825378746509e-05, "loss": 0.1572, "step": 28365 }, { "epoch": 1.32, "learning_rate": 2.7437785717629073e-05, "loss": 0.134, "step": 28370 }, { "epoch": 1.32, "learning_rate": 2.7437317647793053e-05, "loss": 0.2102, "step": 28375 }, { "epoch": 1.32, "learning_rate": 2.7436849577957033e-05, "loss": 0.3999, "step": 28380 }, { "epoch": 1.32, "learning_rate": 2.743638150812101e-05, "loss": 0.1921, "step": 28385 }, { "epoch": 1.32, "learning_rate": 2.7435913438284993e-05, "loss": 0.0599, "step": 28390 }, { "epoch": 1.32, "learning_rate": 2.7435445368448973e-05, "loss": 0.0365, "step": 28395 }, { "epoch": 1.33, "learning_rate": 2.7434977298612952e-05, "loss": 0.1112, "step": 28400 }, { "epoch": 1.33, "learning_rate": 2.7434509228776932e-05, "loss": 0.0971, "step": 28405 }, { "epoch": 1.33, "learning_rate": 2.7434041158940915e-05, "loss": 0.0944, "step": 28410 }, { "epoch": 1.33, "learning_rate": 2.7433573089104895e-05, "loss": 0.1911, "step": 28415 }, { "epoch": 1.33, "learning_rate": 2.7433105019268875e-05, "loss": 0.0776, "step": 28420 }, { "epoch": 1.33, "learning_rate": 2.7432636949432858e-05, "loss": 0.1998, "step": 28425 }, { "epoch": 1.33, "learning_rate": 2.7432168879596838e-05, "loss": 0.3238, "step": 28430 }, { "epoch": 1.33, "learning_rate": 2.7431700809760818e-05, "loss": 0.1711, "step": 28435 }, { "epoch": 1.33, "learning_rate": 2.7431232739924798e-05, "loss": 0.0681, "step": 28440 }, { "epoch": 1.33, "learning_rate": 2.743076467008878e-05, "loss": 0.0965, "step": 28445 }, { "epoch": 1.33, "learning_rate": 2.7430296600252757e-05, "loss": 0.0806, "step": 28450 }, { "epoch": 1.33, "learning_rate": 2.7429828530416737e-05, "loss": 0.0958, "step": 28455 }, { "epoch": 1.33, "learning_rate": 2.7429360460580717e-05, "loss": 0.2168, "step": 28460 }, { "epoch": 1.33, "learning_rate": 2.74288923907447e-05, "loss": 0.1312, "step": 28465 }, { "epoch": 1.33, "learning_rate": 2.742842432090868e-05, "loss": 0.2123, "step": 28470 }, { "epoch": 1.33, "learning_rate": 2.742795625107266e-05, "loss": 0.2767, "step": 28475 }, { "epoch": 1.33, "learning_rate": 2.742748818123664e-05, "loss": 0.3819, "step": 28480 }, { "epoch": 1.33, "learning_rate": 2.7427020111400623e-05, "loss": 0.2376, "step": 28485 }, { "epoch": 1.33, "learning_rate": 2.7426552041564603e-05, "loss": 0.0465, "step": 28490 }, { "epoch": 1.33, "learning_rate": 2.7426083971728582e-05, "loss": 0.0668, "step": 28495 }, { "epoch": 1.33, "learning_rate": 2.7425615901892566e-05, "loss": 0.1181, "step": 28500 }, { "epoch": 1.33, "learning_rate": 2.7425147832056545e-05, "loss": 0.0504, "step": 28505 }, { "epoch": 1.33, "learning_rate": 2.7424679762220525e-05, "loss": 0.1084, "step": 28510 }, { "epoch": 1.33, "learning_rate": 2.7424211692384502e-05, "loss": 0.1506, "step": 28515 }, { "epoch": 1.33, "learning_rate": 2.7423743622548485e-05, "loss": 0.1575, "step": 28520 }, { "epoch": 1.33, "learning_rate": 2.7423275552712465e-05, "loss": 0.2008, "step": 28525 }, { "epoch": 1.33, "learning_rate": 2.7422807482876445e-05, "loss": 0.3201, "step": 28530 }, { "epoch": 1.33, "learning_rate": 2.7422339413040424e-05, "loss": 0.2395, "step": 28535 }, { "epoch": 1.33, "learning_rate": 2.7421871343204408e-05, "loss": 0.1258, "step": 28540 }, { "epoch": 1.33, "learning_rate": 2.7421403273368387e-05, "loss": 0.0798, "step": 28545 }, { "epoch": 1.33, "learning_rate": 2.7420935203532367e-05, "loss": 0.0968, "step": 28550 }, { "epoch": 1.33, "learning_rate": 2.742046713369635e-05, "loss": 0.0886, "step": 28555 }, { "epoch": 1.33, "learning_rate": 2.741999906386033e-05, "loss": 0.2289, "step": 28560 }, { "epoch": 1.33, "learning_rate": 2.741953099402431e-05, "loss": 0.2077, "step": 28565 }, { "epoch": 1.33, "learning_rate": 2.741906292418829e-05, "loss": 0.163, "step": 28570 }, { "epoch": 1.33, "learning_rate": 2.741859485435227e-05, "loss": 0.217, "step": 28575 }, { "epoch": 1.33, "learning_rate": 2.741812678451625e-05, "loss": 0.3247, "step": 28580 }, { "epoch": 1.33, "learning_rate": 2.741765871468023e-05, "loss": 0.3186, "step": 28585 }, { "epoch": 1.33, "learning_rate": 2.741719064484421e-05, "loss": 0.0384, "step": 28590 }, { "epoch": 1.33, "learning_rate": 2.7416722575008192e-05, "loss": 0.0504, "step": 28595 }, { "epoch": 1.33, "learning_rate": 2.7416254505172172e-05, "loss": 0.0692, "step": 28600 }, { "epoch": 1.33, "learning_rate": 2.7415786435336152e-05, "loss": 0.1006, "step": 28605 }, { "epoch": 1.33, "learning_rate": 2.7415318365500135e-05, "loss": 0.1761, "step": 28610 }, { "epoch": 1.34, "learning_rate": 2.7414850295664115e-05, "loss": 0.1248, "step": 28615 }, { "epoch": 1.34, "learning_rate": 2.7414382225828095e-05, "loss": 0.19, "step": 28620 }, { "epoch": 1.34, "learning_rate": 2.7413914155992075e-05, "loss": 0.1107, "step": 28625 }, { "epoch": 1.34, "learning_rate": 2.7413446086156058e-05, "loss": 0.293, "step": 28630 }, { "epoch": 1.34, "learning_rate": 2.7412978016320038e-05, "loss": 0.2144, "step": 28635 }, { "epoch": 1.34, "learning_rate": 2.7412509946484014e-05, "loss": 0.0996, "step": 28640 }, { "epoch": 1.34, "learning_rate": 2.7412041876647994e-05, "loss": 0.0689, "step": 28645 }, { "epoch": 1.34, "learning_rate": 2.7411573806811977e-05, "loss": 0.0643, "step": 28650 }, { "epoch": 1.34, "learning_rate": 2.7411105736975957e-05, "loss": 0.0736, "step": 28655 }, { "epoch": 1.34, "learning_rate": 2.7410637667139937e-05, "loss": 0.1032, "step": 28660 }, { "epoch": 1.34, "learning_rate": 2.7410169597303917e-05, "loss": 0.1954, "step": 28665 }, { "epoch": 1.34, "learning_rate": 2.74097015274679e-05, "loss": 0.1054, "step": 28670 }, { "epoch": 1.34, "learning_rate": 2.740923345763188e-05, "loss": 0.3436, "step": 28675 }, { "epoch": 1.34, "learning_rate": 2.740876538779586e-05, "loss": 0.2841, "step": 28680 }, { "epoch": 1.34, "learning_rate": 2.7408297317959843e-05, "loss": 0.2925, "step": 28685 }, { "epoch": 1.34, "learning_rate": 2.7407829248123822e-05, "loss": 0.0757, "step": 28690 }, { "epoch": 1.34, "learning_rate": 2.7407361178287802e-05, "loss": 0.0521, "step": 28695 }, { "epoch": 1.34, "learning_rate": 2.7406893108451782e-05, "loss": 0.0722, "step": 28700 }, { "epoch": 1.34, "learning_rate": 2.7406425038615762e-05, "loss": 0.1159, "step": 28705 }, { "epoch": 1.34, "learning_rate": 2.7405956968779742e-05, "loss": 0.1192, "step": 28710 }, { "epoch": 1.34, "learning_rate": 2.740548889894372e-05, "loss": 0.1786, "step": 28715 }, { "epoch": 1.34, "learning_rate": 2.74050208291077e-05, "loss": 0.2406, "step": 28720 }, { "epoch": 1.34, "learning_rate": 2.7404552759271685e-05, "loss": 0.202, "step": 28725 }, { "epoch": 1.34, "learning_rate": 2.7404084689435664e-05, "loss": 0.2896, "step": 28730 }, { "epoch": 1.34, "learning_rate": 2.7403616619599644e-05, "loss": 0.2428, "step": 28735 }, { "epoch": 1.34, "learning_rate": 2.7403148549763627e-05, "loss": 0.0399, "step": 28740 }, { "epoch": 1.34, "learning_rate": 2.7402680479927607e-05, "loss": 0.0714, "step": 28745 }, { "epoch": 1.34, "learning_rate": 2.7402212410091587e-05, "loss": 0.047, "step": 28750 }, { "epoch": 1.34, "learning_rate": 2.7401744340255567e-05, "loss": 0.0778, "step": 28755 }, { "epoch": 1.34, "learning_rate": 2.740127627041955e-05, "loss": 0.0955, "step": 28760 }, { "epoch": 1.34, "learning_rate": 2.7400808200583526e-05, "loss": 0.099, "step": 28765 }, { "epoch": 1.34, "learning_rate": 2.7400340130747506e-05, "loss": 0.1675, "step": 28770 }, { "epoch": 1.34, "learning_rate": 2.7399872060911486e-05, "loss": 0.1985, "step": 28775 }, { "epoch": 1.34, "learning_rate": 2.739940399107547e-05, "loss": 0.4462, "step": 28780 }, { "epoch": 1.34, "learning_rate": 2.739893592123945e-05, "loss": 0.2544, "step": 28785 }, { "epoch": 1.34, "learning_rate": 2.739846785140343e-05, "loss": 0.043, "step": 28790 }, { "epoch": 1.34, "learning_rate": 2.7397999781567412e-05, "loss": 0.0404, "step": 28795 }, { "epoch": 1.34, "learning_rate": 2.7397531711731392e-05, "loss": 0.073, "step": 28800 }, { "epoch": 1.34, "learning_rate": 2.7397063641895372e-05, "loss": 0.0875, "step": 28805 }, { "epoch": 1.34, "learning_rate": 2.739659557205935e-05, "loss": 0.0974, "step": 28810 }, { "epoch": 1.34, "learning_rate": 2.7396127502223335e-05, "loss": 0.2016, "step": 28815 }, { "epoch": 1.34, "learning_rate": 2.7395659432387315e-05, "loss": 0.2177, "step": 28820 }, { "epoch": 1.35, "learning_rate": 2.7395191362551294e-05, "loss": 0.2592, "step": 28825 }, { "epoch": 1.35, "learning_rate": 2.739472329271527e-05, "loss": 0.4144, "step": 28830 }, { "epoch": 1.35, "learning_rate": 2.7394255222879254e-05, "loss": 0.2577, "step": 28835 }, { "epoch": 1.35, "learning_rate": 2.7393787153043234e-05, "loss": 0.1039, "step": 28840 }, { "epoch": 1.35, "learning_rate": 2.7393319083207214e-05, "loss": 0.0974, "step": 28845 }, { "epoch": 1.35, "learning_rate": 2.7392851013371194e-05, "loss": 0.075, "step": 28850 }, { "epoch": 1.35, "learning_rate": 2.7392382943535177e-05, "loss": 0.1481, "step": 28855 }, { "epoch": 1.35, "learning_rate": 2.7391914873699157e-05, "loss": 0.1497, "step": 28860 }, { "epoch": 1.35, "learning_rate": 2.7391446803863136e-05, "loss": 0.114, "step": 28865 }, { "epoch": 1.35, "learning_rate": 2.739097873402712e-05, "loss": 0.1186, "step": 28870 }, { "epoch": 1.35, "learning_rate": 2.73905106641911e-05, "loss": 0.218, "step": 28875 }, { "epoch": 1.35, "learning_rate": 2.739004259435508e-05, "loss": 0.4271, "step": 28880 }, { "epoch": 1.35, "learning_rate": 2.738957452451906e-05, "loss": 0.3903, "step": 28885 }, { "epoch": 1.35, "learning_rate": 2.738910645468304e-05, "loss": 0.0907, "step": 28890 }, { "epoch": 1.35, "learning_rate": 2.738863838484702e-05, "loss": 0.0537, "step": 28895 }, { "epoch": 1.35, "learning_rate": 2.7388170315011e-05, "loss": 0.0873, "step": 28900 }, { "epoch": 1.35, "learning_rate": 2.738770224517498e-05, "loss": 0.0824, "step": 28905 }, { "epoch": 1.35, "learning_rate": 2.738723417533896e-05, "loss": 0.2234, "step": 28910 }, { "epoch": 1.35, "learning_rate": 2.738676610550294e-05, "loss": 0.1144, "step": 28915 }, { "epoch": 1.35, "learning_rate": 2.738629803566692e-05, "loss": 0.1043, "step": 28920 }, { "epoch": 1.35, "learning_rate": 2.7385829965830904e-05, "loss": 0.3015, "step": 28925 }, { "epoch": 1.35, "learning_rate": 2.7385361895994884e-05, "loss": 0.414, "step": 28930 }, { "epoch": 1.35, "learning_rate": 2.7384893826158864e-05, "loss": 0.2836, "step": 28935 }, { "epoch": 1.35, "learning_rate": 2.7384425756322844e-05, "loss": 0.0647, "step": 28940 }, { "epoch": 1.35, "learning_rate": 2.7383957686486827e-05, "loss": 0.1362, "step": 28945 }, { "epoch": 1.35, "learning_rate": 2.7383489616650807e-05, "loss": 0.0785, "step": 28950 }, { "epoch": 1.35, "learning_rate": 2.7383021546814783e-05, "loss": 0.1368, "step": 28955 }, { "epoch": 1.35, "learning_rate": 2.7382553476978763e-05, "loss": 0.0939, "step": 28960 }, { "epoch": 1.35, "learning_rate": 2.7382085407142746e-05, "loss": 0.1351, "step": 28965 }, { "epoch": 1.35, "learning_rate": 2.7381617337306726e-05, "loss": 0.1164, "step": 28970 }, { "epoch": 1.35, "learning_rate": 2.7381149267470706e-05, "loss": 0.2935, "step": 28975 }, { "epoch": 1.35, "learning_rate": 2.738068119763469e-05, "loss": 0.3958, "step": 28980 }, { "epoch": 1.35, "learning_rate": 2.738021312779867e-05, "loss": 0.2515, "step": 28985 }, { "epoch": 1.35, "learning_rate": 2.737974505796265e-05, "loss": 0.05, "step": 28990 }, { "epoch": 1.35, "learning_rate": 2.737927698812663e-05, "loss": 0.0642, "step": 28995 }, { "epoch": 1.35, "learning_rate": 2.7378808918290612e-05, "loss": 0.1265, "step": 29000 }, { "epoch": 1.35, "learning_rate": 2.737834084845459e-05, "loss": 0.2096, "step": 29005 }, { "epoch": 1.35, "learning_rate": 2.737787277861857e-05, "loss": 0.0948, "step": 29010 }, { "epoch": 1.35, "learning_rate": 2.737740470878255e-05, "loss": 0.229, "step": 29015 }, { "epoch": 1.35, "learning_rate": 2.737693663894653e-05, "loss": 0.1108, "step": 29020 }, { "epoch": 1.35, "learning_rate": 2.737646856911051e-05, "loss": 0.1859, "step": 29025 }, { "epoch": 1.35, "learning_rate": 2.737600049927449e-05, "loss": 0.2927, "step": 29030 }, { "epoch": 1.35, "learning_rate": 2.7375532429438474e-05, "loss": 0.3511, "step": 29035 }, { "epoch": 1.36, "learning_rate": 2.7375064359602454e-05, "loss": 0.0414, "step": 29040 }, { "epoch": 1.36, "learning_rate": 2.7374596289766434e-05, "loss": 0.0578, "step": 29045 }, { "epoch": 1.36, "learning_rate": 2.7374128219930413e-05, "loss": 0.0689, "step": 29050 }, { "epoch": 1.36, "learning_rate": 2.7373660150094397e-05, "loss": 0.1956, "step": 29055 }, { "epoch": 1.36, "learning_rate": 2.7373192080258376e-05, "loss": 0.1061, "step": 29060 }, { "epoch": 1.36, "learning_rate": 2.7372724010422356e-05, "loss": 0.0428, "step": 29065 }, { "epoch": 1.36, "learning_rate": 2.7372255940586336e-05, "loss": 0.2066, "step": 29070 }, { "epoch": 1.36, "learning_rate": 2.737178787075032e-05, "loss": 0.1464, "step": 29075 }, { "epoch": 1.36, "learning_rate": 2.7371319800914296e-05, "loss": 0.2196, "step": 29080 }, { "epoch": 1.36, "learning_rate": 2.7370851731078275e-05, "loss": 0.3011, "step": 29085 }, { "epoch": 1.36, "learning_rate": 2.7370383661242255e-05, "loss": 0.0939, "step": 29090 }, { "epoch": 1.36, "learning_rate": 2.736991559140624e-05, "loss": 0.0834, "step": 29095 }, { "epoch": 1.36, "learning_rate": 2.736944752157022e-05, "loss": 0.0629, "step": 29100 }, { "epoch": 1.36, "learning_rate": 2.7368979451734198e-05, "loss": 0.0954, "step": 29105 }, { "epoch": 1.36, "learning_rate": 2.736851138189818e-05, "loss": 0.1095, "step": 29110 }, { "epoch": 1.36, "learning_rate": 2.736804331206216e-05, "loss": 0.1453, "step": 29115 }, { "epoch": 1.36, "learning_rate": 2.736757524222614e-05, "loss": 0.2119, "step": 29120 }, { "epoch": 1.36, "learning_rate": 2.736710717239012e-05, "loss": 0.278, "step": 29125 }, { "epoch": 1.36, "learning_rate": 2.7366639102554104e-05, "loss": 0.2615, "step": 29130 }, { "epoch": 1.36, "learning_rate": 2.7366171032718084e-05, "loss": 0.3442, "step": 29135 }, { "epoch": 1.36, "learning_rate": 2.7365702962882064e-05, "loss": 0.0458, "step": 29140 }, { "epoch": 1.36, "learning_rate": 2.736523489304604e-05, "loss": 0.0862, "step": 29145 }, { "epoch": 1.36, "learning_rate": 2.7364766823210023e-05, "loss": 0.0631, "step": 29150 }, { "epoch": 1.36, "learning_rate": 2.7364298753374003e-05, "loss": 0.033, "step": 29155 }, { "epoch": 1.36, "learning_rate": 2.7363830683537983e-05, "loss": 0.0917, "step": 29160 }, { "epoch": 1.36, "learning_rate": 2.7363362613701966e-05, "loss": 0.1551, "step": 29165 }, { "epoch": 1.36, "learning_rate": 2.7362894543865946e-05, "loss": 0.1307, "step": 29170 }, { "epoch": 1.36, "learning_rate": 2.7362426474029926e-05, "loss": 0.2624, "step": 29175 }, { "epoch": 1.36, "learning_rate": 2.7361958404193906e-05, "loss": 0.3407, "step": 29180 }, { "epoch": 1.36, "learning_rate": 2.736149033435789e-05, "loss": 0.2735, "step": 29185 }, { "epoch": 1.36, "learning_rate": 2.736102226452187e-05, "loss": 0.0644, "step": 29190 }, { "epoch": 1.36, "learning_rate": 2.736055419468585e-05, "loss": 0.0732, "step": 29195 }, { "epoch": 1.36, "learning_rate": 2.7360086124849828e-05, "loss": 0.0568, "step": 29200 }, { "epoch": 1.36, "learning_rate": 2.735961805501381e-05, "loss": 0.0936, "step": 29205 }, { "epoch": 1.36, "learning_rate": 2.7359149985177788e-05, "loss": 0.0682, "step": 29210 }, { "epoch": 1.36, "learning_rate": 2.7358681915341768e-05, "loss": 0.2085, "step": 29215 }, { "epoch": 1.36, "learning_rate": 2.735821384550575e-05, "loss": 0.1262, "step": 29220 }, { "epoch": 1.36, "learning_rate": 2.735774577566973e-05, "loss": 0.1878, "step": 29225 }, { "epoch": 1.36, "learning_rate": 2.735727770583371e-05, "loss": 0.3107, "step": 29230 }, { "epoch": 1.36, "learning_rate": 2.735680963599769e-05, "loss": 0.3274, "step": 29235 }, { "epoch": 1.36, "learning_rate": 2.7356341566161674e-05, "loss": 0.0723, "step": 29240 }, { "epoch": 1.36, "learning_rate": 2.7355873496325653e-05, "loss": 0.0393, "step": 29245 }, { "epoch": 1.36, "learning_rate": 2.7355405426489633e-05, "loss": 0.0706, "step": 29250 }, { "epoch": 1.37, "learning_rate": 2.7354937356653613e-05, "loss": 0.1295, "step": 29255 }, { "epoch": 1.37, "learning_rate": 2.7354469286817596e-05, "loss": 0.1177, "step": 29260 }, { "epoch": 1.37, "learning_rate": 2.7354001216981576e-05, "loss": 0.1223, "step": 29265 }, { "epoch": 1.37, "learning_rate": 2.7353533147145552e-05, "loss": 0.2594, "step": 29270 }, { "epoch": 1.37, "learning_rate": 2.7353065077309532e-05, "loss": 0.1623, "step": 29275 }, { "epoch": 1.37, "learning_rate": 2.7352597007473515e-05, "loss": 0.3219, "step": 29280 }, { "epoch": 1.37, "learning_rate": 2.7352128937637495e-05, "loss": 0.2786, "step": 29285 }, { "epoch": 1.37, "learning_rate": 2.7351660867801475e-05, "loss": 0.0947, "step": 29290 }, { "epoch": 1.37, "learning_rate": 2.7351192797965458e-05, "loss": 0.05, "step": 29295 }, { "epoch": 1.37, "learning_rate": 2.7350724728129438e-05, "loss": 0.0553, "step": 29300 }, { "epoch": 1.37, "learning_rate": 2.7350256658293418e-05, "loss": 0.1859, "step": 29305 }, { "epoch": 1.37, "learning_rate": 2.7349788588457398e-05, "loss": 0.1293, "step": 29310 }, { "epoch": 1.37, "learning_rate": 2.734932051862138e-05, "loss": 0.1753, "step": 29315 }, { "epoch": 1.37, "learning_rate": 2.734885244878536e-05, "loss": 0.2847, "step": 29320 }, { "epoch": 1.37, "learning_rate": 2.734838437894934e-05, "loss": 0.3092, "step": 29325 }, { "epoch": 1.37, "learning_rate": 2.734791630911332e-05, "loss": 0.2419, "step": 29330 }, { "epoch": 1.37, "learning_rate": 2.73474482392773e-05, "loss": 0.3635, "step": 29335 }, { "epoch": 1.37, "learning_rate": 2.734698016944128e-05, "loss": 0.0397, "step": 29340 }, { "epoch": 1.37, "learning_rate": 2.734651209960526e-05, "loss": 0.0611, "step": 29345 }, { "epoch": 1.37, "learning_rate": 2.7346044029769243e-05, "loss": 0.074, "step": 29350 }, { "epoch": 1.37, "learning_rate": 2.7345575959933223e-05, "loss": 0.0998, "step": 29355 }, { "epoch": 1.37, "learning_rate": 2.7345107890097203e-05, "loss": 0.1043, "step": 29360 }, { "epoch": 1.37, "learning_rate": 2.7344639820261183e-05, "loss": 0.1502, "step": 29365 }, { "epoch": 1.37, "learning_rate": 2.7344171750425166e-05, "loss": 0.1774, "step": 29370 }, { "epoch": 1.37, "learning_rate": 2.7343703680589146e-05, "loss": 0.2189, "step": 29375 }, { "epoch": 1.37, "learning_rate": 2.7343235610753125e-05, "loss": 0.3716, "step": 29380 }, { "epoch": 1.37, "learning_rate": 2.7342767540917105e-05, "loss": 0.2224, "step": 29385 }, { "epoch": 1.37, "learning_rate": 2.734229947108109e-05, "loss": 0.0402, "step": 29390 }, { "epoch": 1.37, "learning_rate": 2.7341831401245068e-05, "loss": 0.127, "step": 29395 }, { "epoch": 1.37, "learning_rate": 2.7341363331409045e-05, "loss": 0.1143, "step": 29400 }, { "epoch": 1.37, "learning_rate": 2.7340895261573028e-05, "loss": 0.1461, "step": 29405 }, { "epoch": 1.37, "learning_rate": 2.7340427191737008e-05, "loss": 0.1332, "step": 29410 }, { "epoch": 1.37, "learning_rate": 2.7339959121900987e-05, "loss": 0.1332, "step": 29415 }, { "epoch": 1.37, "learning_rate": 2.7339491052064967e-05, "loss": 0.0964, "step": 29420 }, { "epoch": 1.37, "learning_rate": 2.733902298222895e-05, "loss": 0.2352, "step": 29425 }, { "epoch": 1.37, "learning_rate": 2.733855491239293e-05, "loss": 0.3506, "step": 29430 }, { "epoch": 1.37, "learning_rate": 2.733808684255691e-05, "loss": 0.2323, "step": 29435 }, { "epoch": 1.37, "learning_rate": 2.733761877272089e-05, "loss": 0.0665, "step": 29440 }, { "epoch": 1.37, "learning_rate": 2.7337150702884873e-05, "loss": 0.0533, "step": 29445 }, { "epoch": 1.37, "learning_rate": 2.7336682633048853e-05, "loss": 0.1063, "step": 29450 }, { "epoch": 1.37, "learning_rate": 2.7336214563212833e-05, "loss": 0.1773, "step": 29455 }, { "epoch": 1.37, "learning_rate": 2.733574649337681e-05, "loss": 0.0982, "step": 29460 }, { "epoch": 1.37, "learning_rate": 2.7335278423540792e-05, "loss": 0.112, "step": 29465 }, { "epoch": 1.38, "learning_rate": 2.7334810353704772e-05, "loss": 0.2218, "step": 29470 }, { "epoch": 1.38, "learning_rate": 2.7334342283868752e-05, "loss": 0.1661, "step": 29475 }, { "epoch": 1.38, "learning_rate": 2.7333874214032735e-05, "loss": 0.2736, "step": 29480 }, { "epoch": 1.38, "learning_rate": 2.7333406144196715e-05, "loss": 0.2688, "step": 29485 }, { "epoch": 1.38, "learning_rate": 2.7332938074360695e-05, "loss": 0.0811, "step": 29490 }, { "epoch": 1.38, "learning_rate": 2.7332470004524675e-05, "loss": 0.0763, "step": 29495 }, { "epoch": 1.38, "learning_rate": 2.7332001934688658e-05, "loss": 0.0626, "step": 29500 }, { "epoch": 1.38, "learning_rate": 2.7331533864852638e-05, "loss": 0.1125, "step": 29505 }, { "epoch": 1.38, "learning_rate": 2.7331065795016618e-05, "loss": 0.1497, "step": 29510 }, { "epoch": 1.38, "learning_rate": 2.7330597725180597e-05, "loss": 0.1313, "step": 29515 }, { "epoch": 1.38, "learning_rate": 2.733012965534458e-05, "loss": 0.1472, "step": 29520 }, { "epoch": 1.38, "learning_rate": 2.7329661585508557e-05, "loss": 0.2164, "step": 29525 }, { "epoch": 1.38, "learning_rate": 2.7329193515672537e-05, "loss": 0.3979, "step": 29530 }, { "epoch": 1.38, "learning_rate": 2.732872544583652e-05, "loss": 0.3194, "step": 29535 }, { "epoch": 1.38, "learning_rate": 2.73282573760005e-05, "loss": 0.0541, "step": 29540 }, { "epoch": 1.38, "learning_rate": 2.732778930616448e-05, "loss": 0.0625, "step": 29545 }, { "epoch": 1.38, "learning_rate": 2.732732123632846e-05, "loss": 0.0817, "step": 29550 }, { "epoch": 1.38, "learning_rate": 2.7326853166492443e-05, "loss": 0.0807, "step": 29555 }, { "epoch": 1.38, "learning_rate": 2.7326385096656423e-05, "loss": 0.1262, "step": 29560 }, { "epoch": 1.38, "learning_rate": 2.7325917026820402e-05, "loss": 0.234, "step": 29565 }, { "epoch": 1.38, "learning_rate": 2.7325448956984382e-05, "loss": 0.1336, "step": 29570 }, { "epoch": 1.38, "learning_rate": 2.7324980887148365e-05, "loss": 0.2085, "step": 29575 }, { "epoch": 1.38, "learning_rate": 2.7324512817312345e-05, "loss": 0.4152, "step": 29580 }, { "epoch": 1.38, "learning_rate": 2.7324044747476325e-05, "loss": 0.2356, "step": 29585 }, { "epoch": 1.38, "learning_rate": 2.7323576677640305e-05, "loss": 0.0428, "step": 29590 }, { "epoch": 1.38, "learning_rate": 2.7323108607804285e-05, "loss": 0.0449, "step": 29595 }, { "epoch": 1.38, "learning_rate": 2.7322640537968264e-05, "loss": 0.067, "step": 29600 }, { "epoch": 1.38, "learning_rate": 2.7322172468132244e-05, "loss": 0.0813, "step": 29605 }, { "epoch": 1.38, "learning_rate": 2.7321704398296227e-05, "loss": 0.1189, "step": 29610 }, { "epoch": 1.38, "learning_rate": 2.7321236328460207e-05, "loss": 0.1537, "step": 29615 }, { "epoch": 1.38, "learning_rate": 2.7320768258624187e-05, "loss": 0.1681, "step": 29620 }, { "epoch": 1.38, "learning_rate": 2.7320300188788167e-05, "loss": 0.1898, "step": 29625 }, { "epoch": 1.38, "learning_rate": 2.731983211895215e-05, "loss": 0.3343, "step": 29630 }, { "epoch": 1.38, "learning_rate": 2.731936404911613e-05, "loss": 0.2255, "step": 29635 }, { "epoch": 1.38, "learning_rate": 2.731889597928011e-05, "loss": 0.0407, "step": 29640 }, { "epoch": 1.38, "learning_rate": 2.7318427909444093e-05, "loss": 0.0766, "step": 29645 }, { "epoch": 1.38, "learning_rate": 2.731795983960807e-05, "loss": 0.089, "step": 29650 }, { "epoch": 1.38, "learning_rate": 2.731749176977205e-05, "loss": 0.0971, "step": 29655 }, { "epoch": 1.38, "learning_rate": 2.731702369993603e-05, "loss": 0.0827, "step": 29660 }, { "epoch": 1.38, "learning_rate": 2.7316555630100012e-05, "loss": 0.1651, "step": 29665 }, { "epoch": 1.38, "learning_rate": 2.7316087560263992e-05, "loss": 0.2248, "step": 29670 }, { "epoch": 1.38, "learning_rate": 2.7315619490427972e-05, "loss": 0.1914, "step": 29675 }, { "epoch": 1.38, "learning_rate": 2.7315151420591952e-05, "loss": 0.3703, "step": 29680 }, { "epoch": 1.39, "learning_rate": 2.7314683350755935e-05, "loss": 0.2352, "step": 29685 }, { "epoch": 1.39, "learning_rate": 2.7314215280919915e-05, "loss": 0.0698, "step": 29690 }, { "epoch": 1.39, "learning_rate": 2.7313747211083895e-05, "loss": 0.0649, "step": 29695 }, { "epoch": 1.39, "learning_rate": 2.7313279141247874e-05, "loss": 0.1088, "step": 29700 }, { "epoch": 1.39, "learning_rate": 2.7312811071411858e-05, "loss": 0.117, "step": 29705 }, { "epoch": 1.39, "learning_rate": 2.7312343001575837e-05, "loss": 0.0865, "step": 29710 }, { "epoch": 1.39, "learning_rate": 2.7311874931739814e-05, "loss": 0.19, "step": 29715 }, { "epoch": 1.39, "learning_rate": 2.7311406861903797e-05, "loss": 0.1368, "step": 29720 }, { "epoch": 1.39, "learning_rate": 2.7310938792067777e-05, "loss": 0.2456, "step": 29725 }, { "epoch": 1.39, "learning_rate": 2.7310470722231757e-05, "loss": 0.2816, "step": 29730 }, { "epoch": 1.39, "learning_rate": 2.7310002652395736e-05, "loss": 0.2871, "step": 29735 }, { "epoch": 1.39, "learning_rate": 2.730953458255972e-05, "loss": 0.0209, "step": 29740 }, { "epoch": 1.39, "learning_rate": 2.73090665127237e-05, "loss": 0.0584, "step": 29745 }, { "epoch": 1.39, "learning_rate": 2.730859844288768e-05, "loss": 0.0264, "step": 29750 }, { "epoch": 1.39, "learning_rate": 2.730813037305166e-05, "loss": 0.0675, "step": 29755 }, { "epoch": 1.39, "learning_rate": 2.7307662303215642e-05, "loss": 0.1525, "step": 29760 }, { "epoch": 1.39, "learning_rate": 2.7307194233379622e-05, "loss": 0.1282, "step": 29765 }, { "epoch": 1.39, "learning_rate": 2.7306726163543602e-05, "loss": 0.1607, "step": 29770 }, { "epoch": 1.39, "learning_rate": 2.7306258093707582e-05, "loss": 0.1726, "step": 29775 }, { "epoch": 1.39, "learning_rate": 2.730579002387156e-05, "loss": 0.3153, "step": 29780 }, { "epoch": 1.39, "learning_rate": 2.730532195403554e-05, "loss": 0.3799, "step": 29785 }, { "epoch": 1.39, "learning_rate": 2.730485388419952e-05, "loss": 0.0519, "step": 29790 }, { "epoch": 1.39, "learning_rate": 2.7304385814363504e-05, "loss": 0.0807, "step": 29795 }, { "epoch": 1.39, "learning_rate": 2.7303917744527484e-05, "loss": 0.0703, "step": 29800 }, { "epoch": 1.39, "learning_rate": 2.7303449674691464e-05, "loss": 0.1323, "step": 29805 }, { "epoch": 1.39, "learning_rate": 2.7302981604855444e-05, "loss": 0.1327, "step": 29810 }, { "epoch": 1.39, "learning_rate": 2.7302513535019427e-05, "loss": 0.1007, "step": 29815 }, { "epoch": 1.39, "learning_rate": 2.7302045465183407e-05, "loss": 0.1576, "step": 29820 }, { "epoch": 1.39, "learning_rate": 2.7301577395347387e-05, "loss": 0.303, "step": 29825 }, { "epoch": 1.39, "learning_rate": 2.730110932551137e-05, "loss": 0.4647, "step": 29830 }, { "epoch": 1.39, "learning_rate": 2.730064125567535e-05, "loss": 0.2118, "step": 29835 }, { "epoch": 1.39, "learning_rate": 2.7300173185839326e-05, "loss": 0.0858, "step": 29840 }, { "epoch": 1.39, "learning_rate": 2.7299705116003306e-05, "loss": 0.1213, "step": 29845 }, { "epoch": 1.39, "learning_rate": 2.729923704616729e-05, "loss": 0.067, "step": 29850 }, { "epoch": 1.39, "learning_rate": 2.729876897633127e-05, "loss": 0.1014, "step": 29855 }, { "epoch": 1.39, "learning_rate": 2.729830090649525e-05, "loss": 0.0869, "step": 29860 }, { "epoch": 1.39, "learning_rate": 2.729783283665923e-05, "loss": 0.0445, "step": 29865 }, { "epoch": 1.39, "learning_rate": 2.7297364766823212e-05, "loss": 0.1358, "step": 29870 }, { "epoch": 1.39, "learning_rate": 2.729689669698719e-05, "loss": 0.1666, "step": 29875 }, { "epoch": 1.39, "learning_rate": 2.729642862715117e-05, "loss": 0.2982, "step": 29880 }, { "epoch": 1.39, "learning_rate": 2.729596055731515e-05, "loss": 0.2161, "step": 29885 }, { "epoch": 1.39, "learning_rate": 2.7295492487479135e-05, "loss": 0.0664, "step": 29890 }, { "epoch": 1.39, "learning_rate": 2.7295024417643114e-05, "loss": 0.0384, "step": 29895 }, { "epoch": 1.4, "learning_rate": 2.7294556347807094e-05, "loss": 0.0762, "step": 29900 }, { "epoch": 1.4, "learning_rate": 2.7294088277971074e-05, "loss": 0.0866, "step": 29905 }, { "epoch": 1.4, "learning_rate": 2.7293620208135054e-05, "loss": 0.103, "step": 29910 }, { "epoch": 1.4, "learning_rate": 2.7293152138299034e-05, "loss": 0.1467, "step": 29915 }, { "epoch": 1.4, "learning_rate": 2.7292684068463013e-05, "loss": 0.142, "step": 29920 }, { "epoch": 1.4, "learning_rate": 2.7292215998626997e-05, "loss": 0.1852, "step": 29925 }, { "epoch": 1.4, "learning_rate": 2.7291747928790976e-05, "loss": 0.3194, "step": 29930 }, { "epoch": 1.4, "learning_rate": 2.7291279858954956e-05, "loss": 0.2362, "step": 29935 }, { "epoch": 1.4, "learning_rate": 2.7290811789118936e-05, "loss": 0.0778, "step": 29940 }, { "epoch": 1.4, "learning_rate": 2.729034371928292e-05, "loss": 0.0882, "step": 29945 }, { "epoch": 1.4, "learning_rate": 2.72898756494469e-05, "loss": 0.046, "step": 29950 }, { "epoch": 1.4, "learning_rate": 2.728940757961088e-05, "loss": 0.1595, "step": 29955 }, { "epoch": 1.4, "learning_rate": 2.7288939509774862e-05, "loss": 0.1162, "step": 29960 }, { "epoch": 1.4, "learning_rate": 2.728847143993884e-05, "loss": 0.1399, "step": 29965 }, { "epoch": 1.4, "learning_rate": 2.728800337010282e-05, "loss": 0.1863, "step": 29970 }, { "epoch": 1.4, "learning_rate": 2.7287535300266798e-05, "loss": 0.1735, "step": 29975 }, { "epoch": 1.4, "learning_rate": 2.728706723043078e-05, "loss": 0.4845, "step": 29980 }, { "epoch": 1.4, "learning_rate": 2.728659916059476e-05, "loss": 0.3567, "step": 29985 }, { "epoch": 1.4, "learning_rate": 2.728613109075874e-05, "loss": 0.054, "step": 29990 }, { "epoch": 1.4, "learning_rate": 2.728566302092272e-05, "loss": 0.0953, "step": 29995 }, { "epoch": 1.4, "learning_rate": 2.7285194951086704e-05, "loss": 0.111, "step": 30000 }, { "epoch": 1.4, "learning_rate": 2.7284726881250684e-05, "loss": 0.0884, "step": 30005 }, { "epoch": 1.4, "learning_rate": 2.7284258811414664e-05, "loss": 0.1144, "step": 30010 }, { "epoch": 1.4, "learning_rate": 2.7283790741578647e-05, "loss": 0.1874, "step": 30015 }, { "epoch": 1.4, "learning_rate": 2.7283322671742627e-05, "loss": 0.181, "step": 30020 }, { "epoch": 1.4, "learning_rate": 2.7282854601906607e-05, "loss": 0.2215, "step": 30025 }, { "epoch": 1.4, "learning_rate": 2.7282386532070583e-05, "loss": 0.23, "step": 30030 }, { "epoch": 1.4, "learning_rate": 2.7281918462234566e-05, "loss": 0.2539, "step": 30035 }, { "epoch": 1.4, "learning_rate": 2.7281450392398546e-05, "loss": 0.0296, "step": 30040 }, { "epoch": 1.4, "learning_rate": 2.7280982322562526e-05, "loss": 0.1145, "step": 30045 }, { "epoch": 1.4, "learning_rate": 2.7280514252726506e-05, "loss": 0.052, "step": 30050 }, { "epoch": 1.4, "learning_rate": 2.728004618289049e-05, "loss": 0.0763, "step": 30055 }, { "epoch": 1.4, "learning_rate": 2.727957811305447e-05, "loss": 0.093, "step": 30060 }, { "epoch": 1.4, "learning_rate": 2.727911004321845e-05, "loss": 0.132, "step": 30065 }, { "epoch": 1.4, "learning_rate": 2.7278641973382428e-05, "loss": 0.1771, "step": 30070 }, { "epoch": 1.4, "learning_rate": 2.727817390354641e-05, "loss": 0.2489, "step": 30075 }, { "epoch": 1.4, "learning_rate": 2.727770583371039e-05, "loss": 0.4328, "step": 30080 }, { "epoch": 1.4, "learning_rate": 2.727723776387437e-05, "loss": 0.278, "step": 30085 }, { "epoch": 1.4, "learning_rate": 2.727676969403835e-05, "loss": 0.1109, "step": 30090 }, { "epoch": 1.4, "learning_rate": 2.727630162420233e-05, "loss": 0.1077, "step": 30095 }, { "epoch": 1.4, "learning_rate": 2.727583355436631e-05, "loss": 0.0888, "step": 30100 }, { "epoch": 1.4, "learning_rate": 2.727536548453029e-05, "loss": 0.0535, "step": 30105 }, { "epoch": 1.4, "learning_rate": 2.7274897414694274e-05, "loss": 0.1247, "step": 30110 }, { "epoch": 1.41, "learning_rate": 2.7274429344858253e-05, "loss": 0.2781, "step": 30115 }, { "epoch": 1.41, "learning_rate": 2.7273961275022233e-05, "loss": 0.1777, "step": 30120 }, { "epoch": 1.41, "learning_rate": 2.7273493205186213e-05, "loss": 0.1703, "step": 30125 }, { "epoch": 1.41, "learning_rate": 2.7273025135350196e-05, "loss": 0.2506, "step": 30130 }, { "epoch": 1.41, "learning_rate": 2.7272557065514176e-05, "loss": 0.2557, "step": 30135 }, { "epoch": 1.41, "learning_rate": 2.7272088995678156e-05, "loss": 0.057, "step": 30140 }, { "epoch": 1.41, "learning_rate": 2.727162092584214e-05, "loss": 0.0928, "step": 30145 }, { "epoch": 1.41, "learning_rate": 2.727115285600612e-05, "loss": 0.1044, "step": 30150 }, { "epoch": 1.41, "learning_rate": 2.7270684786170095e-05, "loss": 0.0798, "step": 30155 }, { "epoch": 1.41, "learning_rate": 2.7270216716334075e-05, "loss": 0.0734, "step": 30160 }, { "epoch": 1.41, "learning_rate": 2.726974864649806e-05, "loss": 0.1477, "step": 30165 }, { "epoch": 1.41, "learning_rate": 2.7269280576662038e-05, "loss": 0.11, "step": 30170 }, { "epoch": 1.41, "learning_rate": 2.7268812506826018e-05, "loss": 0.3176, "step": 30175 }, { "epoch": 1.41, "learning_rate": 2.7268344436989998e-05, "loss": 0.4192, "step": 30180 }, { "epoch": 1.41, "learning_rate": 2.726787636715398e-05, "loss": 0.3313, "step": 30185 }, { "epoch": 1.41, "learning_rate": 2.726740829731796e-05, "loss": 0.0438, "step": 30190 }, { "epoch": 1.41, "learning_rate": 2.726694022748194e-05, "loss": 0.0737, "step": 30195 }, { "epoch": 1.41, "learning_rate": 2.7266472157645924e-05, "loss": 0.084, "step": 30200 }, { "epoch": 1.41, "learning_rate": 2.7266004087809904e-05, "loss": 0.0911, "step": 30205 }, { "epoch": 1.41, "learning_rate": 2.7265536017973884e-05, "loss": 0.0846, "step": 30210 }, { "epoch": 1.41, "learning_rate": 2.7265067948137863e-05, "loss": 0.237, "step": 30215 }, { "epoch": 1.41, "learning_rate": 2.7264599878301843e-05, "loss": 0.1959, "step": 30220 }, { "epoch": 1.41, "learning_rate": 2.7264131808465823e-05, "loss": 0.2003, "step": 30225 }, { "epoch": 1.41, "learning_rate": 2.7263663738629803e-05, "loss": 0.3959, "step": 30230 }, { "epoch": 1.41, "learning_rate": 2.7263195668793783e-05, "loss": 0.2974, "step": 30235 }, { "epoch": 1.41, "learning_rate": 2.7262727598957766e-05, "loss": 0.0595, "step": 30240 }, { "epoch": 1.41, "learning_rate": 2.7262259529121746e-05, "loss": 0.0366, "step": 30245 }, { "epoch": 1.41, "learning_rate": 2.7261791459285725e-05, "loss": 0.0932, "step": 30250 }, { "epoch": 1.41, "learning_rate": 2.7261323389449705e-05, "loss": 0.1064, "step": 30255 }, { "epoch": 1.41, "learning_rate": 2.726085531961369e-05, "loss": 0.0823, "step": 30260 }, { "epoch": 1.41, "learning_rate": 2.7260387249777668e-05, "loss": 0.169, "step": 30265 }, { "epoch": 1.41, "learning_rate": 2.7259919179941648e-05, "loss": 0.11, "step": 30270 }, { "epoch": 1.41, "learning_rate": 2.725945111010563e-05, "loss": 0.287, "step": 30275 }, { "epoch": 1.41, "learning_rate": 2.7258983040269608e-05, "loss": 0.4121, "step": 30280 }, { "epoch": 1.41, "learning_rate": 2.7258514970433588e-05, "loss": 0.296, "step": 30285 }, { "epoch": 1.41, "learning_rate": 2.7258046900597567e-05, "loss": 0.1055, "step": 30290 }, { "epoch": 1.41, "learning_rate": 2.725757883076155e-05, "loss": 0.085, "step": 30295 }, { "epoch": 1.41, "learning_rate": 2.725711076092553e-05, "loss": 0.1009, "step": 30300 }, { "epoch": 1.41, "learning_rate": 2.725664269108951e-05, "loss": 0.1298, "step": 30305 }, { "epoch": 1.41, "learning_rate": 2.725617462125349e-05, "loss": 0.1291, "step": 30310 }, { "epoch": 1.41, "learning_rate": 2.7255706551417473e-05, "loss": 0.0958, "step": 30315 }, { "epoch": 1.41, "learning_rate": 2.7255238481581453e-05, "loss": 0.2243, "step": 30320 }, { "epoch": 1.42, "learning_rate": 2.7254770411745433e-05, "loss": 0.2349, "step": 30325 }, { "epoch": 1.42, "learning_rate": 2.7254302341909416e-05, "loss": 0.3058, "step": 30330 }, { "epoch": 1.42, "learning_rate": 2.7253834272073396e-05, "loss": 0.2994, "step": 30335 }, { "epoch": 1.42, "learning_rate": 2.7253366202237376e-05, "loss": 0.0751, "step": 30340 }, { "epoch": 1.42, "learning_rate": 2.7252898132401352e-05, "loss": 0.0397, "step": 30345 }, { "epoch": 1.42, "learning_rate": 2.7252430062565335e-05, "loss": 0.0997, "step": 30350 }, { "epoch": 1.42, "learning_rate": 2.7251961992729315e-05, "loss": 0.125, "step": 30355 }, { "epoch": 1.42, "learning_rate": 2.7251493922893295e-05, "loss": 0.148, "step": 30360 }, { "epoch": 1.42, "learning_rate": 2.7251025853057275e-05, "loss": 0.1239, "step": 30365 }, { "epoch": 1.42, "learning_rate": 2.7250557783221258e-05, "loss": 0.1708, "step": 30370 }, { "epoch": 1.42, "learning_rate": 2.7250089713385238e-05, "loss": 0.2306, "step": 30375 }, { "epoch": 1.42, "learning_rate": 2.7249621643549218e-05, "loss": 0.3325, "step": 30380 }, { "epoch": 1.42, "learning_rate": 2.72491535737132e-05, "loss": 0.2819, "step": 30385 }, { "epoch": 1.42, "learning_rate": 2.724868550387718e-05, "loss": 0.0532, "step": 30390 }, { "epoch": 1.42, "learning_rate": 2.724821743404116e-05, "loss": 0.0423, "step": 30395 }, { "epoch": 1.42, "learning_rate": 2.724774936420514e-05, "loss": 0.1486, "step": 30400 }, { "epoch": 1.42, "learning_rate": 2.7247281294369124e-05, "loss": 0.1546, "step": 30405 }, { "epoch": 1.42, "learning_rate": 2.72468132245331e-05, "loss": 0.067, "step": 30410 }, { "epoch": 1.42, "learning_rate": 2.724634515469708e-05, "loss": 0.1597, "step": 30415 }, { "epoch": 1.42, "learning_rate": 2.724587708486106e-05, "loss": 0.2715, "step": 30420 }, { "epoch": 1.42, "learning_rate": 2.7245409015025043e-05, "loss": 0.2488, "step": 30425 }, { "epoch": 1.42, "learning_rate": 2.7244940945189023e-05, "loss": 0.4374, "step": 30430 }, { "epoch": 1.42, "learning_rate": 2.7244472875353002e-05, "loss": 0.2353, "step": 30435 }, { "epoch": 1.42, "learning_rate": 2.7244004805516986e-05, "loss": 0.1057, "step": 30440 }, { "epoch": 1.42, "learning_rate": 2.7243536735680965e-05, "loss": 0.0851, "step": 30445 }, { "epoch": 1.42, "learning_rate": 2.7243068665844945e-05, "loss": 0.0867, "step": 30450 }, { "epoch": 1.42, "learning_rate": 2.7242600596008925e-05, "loss": 0.1265, "step": 30455 }, { "epoch": 1.42, "learning_rate": 2.7242132526172908e-05, "loss": 0.0953, "step": 30460 }, { "epoch": 1.42, "learning_rate": 2.7241664456336888e-05, "loss": 0.0935, "step": 30465 }, { "epoch": 1.42, "learning_rate": 2.7241196386500865e-05, "loss": 0.1259, "step": 30470 }, { "epoch": 1.42, "learning_rate": 2.7240728316664844e-05, "loss": 0.1751, "step": 30475 }, { "epoch": 1.42, "learning_rate": 2.7240260246828828e-05, "loss": 0.2389, "step": 30480 }, { "epoch": 1.42, "learning_rate": 2.7239792176992807e-05, "loss": 0.307, "step": 30485 }, { "epoch": 1.42, "learning_rate": 2.7239324107156787e-05, "loss": 0.0281, "step": 30490 }, { "epoch": 1.42, "learning_rate": 2.7238856037320767e-05, "loss": 0.0688, "step": 30495 }, { "epoch": 1.42, "learning_rate": 2.723838796748475e-05, "loss": 0.115, "step": 30500 }, { "epoch": 1.42, "learning_rate": 2.723791989764873e-05, "loss": 0.0956, "step": 30505 }, { "epoch": 1.42, "learning_rate": 2.723745182781271e-05, "loss": 0.1199, "step": 30510 }, { "epoch": 1.42, "learning_rate": 2.7236983757976693e-05, "loss": 0.1294, "step": 30515 }, { "epoch": 1.42, "learning_rate": 2.7236515688140673e-05, "loss": 0.2323, "step": 30520 }, { "epoch": 1.42, "learning_rate": 2.7236047618304653e-05, "loss": 0.1501, "step": 30525 }, { "epoch": 1.42, "learning_rate": 2.7235579548468633e-05, "loss": 0.2528, "step": 30530 }, { "epoch": 1.42, "learning_rate": 2.7235111478632612e-05, "loss": 0.2352, "step": 30535 }, { "epoch": 1.43, "learning_rate": 2.7234643408796592e-05, "loss": 0.0412, "step": 30540 }, { "epoch": 1.43, "learning_rate": 2.7234175338960572e-05, "loss": 0.0744, "step": 30545 }, { "epoch": 1.43, "learning_rate": 2.7233707269124552e-05, "loss": 0.0957, "step": 30550 }, { "epoch": 1.43, "learning_rate": 2.7233239199288535e-05, "loss": 0.1033, "step": 30555 }, { "epoch": 1.43, "learning_rate": 2.7232771129452515e-05, "loss": 0.1381, "step": 30560 }, { "epoch": 1.43, "learning_rate": 2.7232303059616495e-05, "loss": 0.1634, "step": 30565 }, { "epoch": 1.43, "learning_rate": 2.7231834989780478e-05, "loss": 0.2235, "step": 30570 }, { "epoch": 1.43, "learning_rate": 2.7231366919944458e-05, "loss": 0.244, "step": 30575 }, { "epoch": 1.43, "learning_rate": 2.7230898850108437e-05, "loss": 0.5258, "step": 30580 }, { "epoch": 1.43, "learning_rate": 2.7230430780272417e-05, "loss": 0.3767, "step": 30585 }, { "epoch": 1.43, "learning_rate": 2.72299627104364e-05, "loss": 0.0356, "step": 30590 }, { "epoch": 1.43, "learning_rate": 2.722949464060038e-05, "loss": 0.0709, "step": 30595 }, { "epoch": 1.43, "learning_rate": 2.7229026570764357e-05, "loss": 0.0632, "step": 30600 }, { "epoch": 1.43, "learning_rate": 2.7228558500928337e-05, "loss": 0.1599, "step": 30605 }, { "epoch": 1.43, "learning_rate": 2.722809043109232e-05, "loss": 0.1475, "step": 30610 }, { "epoch": 1.43, "learning_rate": 2.72276223612563e-05, "loss": 0.1541, "step": 30615 }, { "epoch": 1.43, "learning_rate": 2.722715429142028e-05, "loss": 0.0917, "step": 30620 }, { "epoch": 1.43, "learning_rate": 2.7226686221584263e-05, "loss": 0.2968, "step": 30625 }, { "epoch": 1.43, "learning_rate": 2.7226218151748242e-05, "loss": 0.3078, "step": 30630 }, { "epoch": 1.43, "learning_rate": 2.7225750081912222e-05, "loss": 0.1851, "step": 30635 }, { "epoch": 1.43, "learning_rate": 2.7225282012076202e-05, "loss": 0.0619, "step": 30640 }, { "epoch": 1.43, "learning_rate": 2.7224813942240185e-05, "loss": 0.1107, "step": 30645 }, { "epoch": 1.43, "learning_rate": 2.7224345872404165e-05, "loss": 0.0636, "step": 30650 }, { "epoch": 1.43, "learning_rate": 2.7223877802568145e-05, "loss": 0.1222, "step": 30655 }, { "epoch": 1.43, "learning_rate": 2.722340973273212e-05, "loss": 0.072, "step": 30660 }, { "epoch": 1.43, "learning_rate": 2.7222941662896105e-05, "loss": 0.2491, "step": 30665 }, { "epoch": 1.43, "learning_rate": 2.7222473593060084e-05, "loss": 0.1339, "step": 30670 }, { "epoch": 1.43, "learning_rate": 2.7222005523224064e-05, "loss": 0.1526, "step": 30675 }, { "epoch": 1.43, "learning_rate": 2.7221537453388044e-05, "loss": 0.4271, "step": 30680 }, { "epoch": 1.43, "learning_rate": 2.7221069383552027e-05, "loss": 0.1767, "step": 30685 }, { "epoch": 1.43, "learning_rate": 2.7220601313716007e-05, "loss": 0.0591, "step": 30690 }, { "epoch": 1.43, "learning_rate": 2.7220133243879987e-05, "loss": 0.0743, "step": 30695 }, { "epoch": 1.43, "learning_rate": 2.721966517404397e-05, "loss": 0.1219, "step": 30700 }, { "epoch": 1.43, "learning_rate": 2.721919710420795e-05, "loss": 0.1162, "step": 30705 }, { "epoch": 1.43, "learning_rate": 2.721872903437193e-05, "loss": 0.0511, "step": 30710 }, { "epoch": 1.43, "learning_rate": 2.721826096453591e-05, "loss": 0.2201, "step": 30715 }, { "epoch": 1.43, "learning_rate": 2.7217792894699893e-05, "loss": 0.1748, "step": 30720 }, { "epoch": 1.43, "learning_rate": 2.721732482486387e-05, "loss": 0.2107, "step": 30725 }, { "epoch": 1.43, "learning_rate": 2.721685675502785e-05, "loss": 0.1747, "step": 30730 }, { "epoch": 1.43, "learning_rate": 2.721638868519183e-05, "loss": 0.2163, "step": 30735 }, { "epoch": 1.43, "learning_rate": 2.7215920615355812e-05, "loss": 0.0622, "step": 30740 }, { "epoch": 1.43, "learning_rate": 2.7215452545519792e-05, "loss": 0.0323, "step": 30745 }, { "epoch": 1.43, "learning_rate": 2.721498447568377e-05, "loss": 0.0959, "step": 30750 }, { "epoch": 1.44, "learning_rate": 2.7214516405847755e-05, "loss": 0.1716, "step": 30755 }, { "epoch": 1.44, "learning_rate": 2.7214048336011735e-05, "loss": 0.1084, "step": 30760 }, { "epoch": 1.44, "learning_rate": 2.7213580266175714e-05, "loss": 0.2736, "step": 30765 }, { "epoch": 1.44, "learning_rate": 2.7213112196339694e-05, "loss": 0.1093, "step": 30770 }, { "epoch": 1.44, "learning_rate": 2.7212644126503677e-05, "loss": 0.1853, "step": 30775 }, { "epoch": 1.44, "learning_rate": 2.7212176056667657e-05, "loss": 0.3013, "step": 30780 }, { "epoch": 1.44, "learning_rate": 2.7211707986831637e-05, "loss": 0.2163, "step": 30785 }, { "epoch": 1.44, "learning_rate": 2.7211239916995614e-05, "loss": 0.0327, "step": 30790 }, { "epoch": 1.44, "learning_rate": 2.7210771847159597e-05, "loss": 0.079, "step": 30795 }, { "epoch": 1.44, "learning_rate": 2.7210303777323577e-05, "loss": 0.1031, "step": 30800 }, { "epoch": 1.44, "learning_rate": 2.7209835707487556e-05, "loss": 0.0917, "step": 30805 }, { "epoch": 1.44, "learning_rate": 2.720936763765154e-05, "loss": 0.1494, "step": 30810 }, { "epoch": 1.44, "learning_rate": 2.720889956781552e-05, "loss": 0.1752, "step": 30815 }, { "epoch": 1.44, "learning_rate": 2.72084314979795e-05, "loss": 0.2704, "step": 30820 }, { "epoch": 1.44, "learning_rate": 2.720796342814348e-05, "loss": 0.2556, "step": 30825 }, { "epoch": 1.44, "learning_rate": 2.7207495358307462e-05, "loss": 0.3796, "step": 30830 }, { "epoch": 1.44, "learning_rate": 2.7207027288471442e-05, "loss": 0.2472, "step": 30835 }, { "epoch": 1.44, "learning_rate": 2.7206559218635422e-05, "loss": 0.0347, "step": 30840 }, { "epoch": 1.44, "learning_rate": 2.72060911487994e-05, "loss": 0.0353, "step": 30845 }, { "epoch": 1.44, "learning_rate": 2.720562307896338e-05, "loss": 0.1022, "step": 30850 }, { "epoch": 1.44, "learning_rate": 2.720515500912736e-05, "loss": 0.1397, "step": 30855 }, { "epoch": 1.44, "learning_rate": 2.720468693929134e-05, "loss": 0.2229, "step": 30860 }, { "epoch": 1.44, "learning_rate": 2.720421886945532e-05, "loss": 0.1854, "step": 30865 }, { "epoch": 1.44, "learning_rate": 2.7203750799619304e-05, "loss": 0.2439, "step": 30870 }, { "epoch": 1.44, "learning_rate": 2.7203282729783284e-05, "loss": 0.173, "step": 30875 }, { "epoch": 1.44, "learning_rate": 2.7202814659947264e-05, "loss": 0.3904, "step": 30880 }, { "epoch": 1.44, "learning_rate": 2.7202346590111247e-05, "loss": 0.2966, "step": 30885 }, { "epoch": 1.44, "learning_rate": 2.7201878520275227e-05, "loss": 0.0616, "step": 30890 }, { "epoch": 1.44, "learning_rate": 2.7201410450439207e-05, "loss": 0.0956, "step": 30895 }, { "epoch": 1.44, "learning_rate": 2.7200942380603186e-05, "loss": 0.0862, "step": 30900 }, { "epoch": 1.44, "learning_rate": 2.720047431076717e-05, "loss": 0.0854, "step": 30905 }, { "epoch": 1.44, "learning_rate": 2.720000624093115e-05, "loss": 0.125, "step": 30910 }, { "epoch": 1.44, "learning_rate": 2.7199538171095126e-05, "loss": 0.148, "step": 30915 }, { "epoch": 1.44, "learning_rate": 2.7199070101259106e-05, "loss": 0.1018, "step": 30920 }, { "epoch": 1.44, "learning_rate": 2.719860203142309e-05, "loss": 0.1813, "step": 30925 }, { "epoch": 1.44, "learning_rate": 2.719813396158707e-05, "loss": 0.3915, "step": 30930 }, { "epoch": 1.44, "learning_rate": 2.719766589175105e-05, "loss": 0.3807, "step": 30935 }, { "epoch": 1.44, "learning_rate": 2.7197197821915032e-05, "loss": 0.0512, "step": 30940 }, { "epoch": 1.44, "learning_rate": 2.719672975207901e-05, "loss": 0.0802, "step": 30945 }, { "epoch": 1.44, "learning_rate": 2.719626168224299e-05, "loss": 0.0591, "step": 30950 }, { "epoch": 1.44, "learning_rate": 2.719579361240697e-05, "loss": 0.1057, "step": 30955 }, { "epoch": 1.44, "learning_rate": 2.7195325542570954e-05, "loss": 0.0571, "step": 30960 }, { "epoch": 1.44, "learning_rate": 2.7194857472734934e-05, "loss": 0.1816, "step": 30965 }, { "epoch": 1.45, "learning_rate": 2.7194389402898914e-05, "loss": 0.1923, "step": 30970 }, { "epoch": 1.45, "learning_rate": 2.719392133306289e-05, "loss": 0.2293, "step": 30975 }, { "epoch": 1.45, "learning_rate": 2.7193453263226874e-05, "loss": 0.2999, "step": 30980 }, { "epoch": 1.45, "learning_rate": 2.7192985193390854e-05, "loss": 0.2714, "step": 30985 }, { "epoch": 1.45, "learning_rate": 2.7192517123554833e-05, "loss": 0.0879, "step": 30990 }, { "epoch": 1.45, "learning_rate": 2.7192049053718817e-05, "loss": 0.0799, "step": 30995 }, { "epoch": 1.45, "learning_rate": 2.7191580983882796e-05, "loss": 0.1103, "step": 31000 }, { "epoch": 1.45, "learning_rate": 2.7191112914046776e-05, "loss": 0.1332, "step": 31005 }, { "epoch": 1.45, "learning_rate": 2.7190644844210756e-05, "loss": 0.1342, "step": 31010 }, { "epoch": 1.45, "learning_rate": 2.719017677437474e-05, "loss": 0.1187, "step": 31015 }, { "epoch": 1.45, "learning_rate": 2.718970870453872e-05, "loss": 0.1804, "step": 31020 }, { "epoch": 1.45, "learning_rate": 2.71892406347027e-05, "loss": 0.2227, "step": 31025 }, { "epoch": 1.45, "learning_rate": 2.718877256486668e-05, "loss": 0.329, "step": 31030 }, { "epoch": 1.45, "learning_rate": 2.7188304495030662e-05, "loss": 0.2625, "step": 31035 }, { "epoch": 1.45, "learning_rate": 2.7187836425194638e-05, "loss": 0.0191, "step": 31040 }, { "epoch": 1.45, "learning_rate": 2.7187368355358618e-05, "loss": 0.0709, "step": 31045 }, { "epoch": 1.45, "learning_rate": 2.71869002855226e-05, "loss": 0.078, "step": 31050 }, { "epoch": 1.45, "learning_rate": 2.718643221568658e-05, "loss": 0.1497, "step": 31055 }, { "epoch": 1.45, "learning_rate": 2.718596414585056e-05, "loss": 0.0965, "step": 31060 }, { "epoch": 1.45, "learning_rate": 2.718549607601454e-05, "loss": 0.1385, "step": 31065 }, { "epoch": 1.45, "learning_rate": 2.7185028006178524e-05, "loss": 0.197, "step": 31070 }, { "epoch": 1.45, "learning_rate": 2.7184559936342504e-05, "loss": 0.2058, "step": 31075 }, { "epoch": 1.45, "learning_rate": 2.7184091866506484e-05, "loss": 0.3558, "step": 31080 }, { "epoch": 1.45, "learning_rate": 2.7183623796670463e-05, "loss": 0.1702, "step": 31085 }, { "epoch": 1.45, "learning_rate": 2.7183155726834447e-05, "loss": 0.062, "step": 31090 }, { "epoch": 1.45, "learning_rate": 2.7182687656998426e-05, "loss": 0.0479, "step": 31095 }, { "epoch": 1.45, "learning_rate": 2.7182219587162406e-05, "loss": 0.0988, "step": 31100 }, { "epoch": 1.45, "learning_rate": 2.7181751517326383e-05, "loss": 0.1179, "step": 31105 }, { "epoch": 1.45, "learning_rate": 2.7181283447490366e-05, "loss": 0.1696, "step": 31110 }, { "epoch": 1.45, "learning_rate": 2.7180815377654346e-05, "loss": 0.176, "step": 31115 }, { "epoch": 1.45, "learning_rate": 2.7180347307818326e-05, "loss": 0.1533, "step": 31120 }, { "epoch": 1.45, "learning_rate": 2.717987923798231e-05, "loss": 0.1897, "step": 31125 }, { "epoch": 1.45, "learning_rate": 2.717941116814629e-05, "loss": 0.3079, "step": 31130 }, { "epoch": 1.45, "learning_rate": 2.717894309831027e-05, "loss": 0.2082, "step": 31135 }, { "epoch": 1.45, "learning_rate": 2.7178475028474248e-05, "loss": 0.0371, "step": 31140 }, { "epoch": 1.45, "learning_rate": 2.717800695863823e-05, "loss": 0.0378, "step": 31145 }, { "epoch": 1.45, "learning_rate": 2.717753888880221e-05, "loss": 0.063, "step": 31150 }, { "epoch": 1.45, "learning_rate": 2.717707081896619e-05, "loss": 0.1161, "step": 31155 }, { "epoch": 1.45, "learning_rate": 2.717660274913017e-05, "loss": 0.2054, "step": 31160 }, { "epoch": 1.45, "learning_rate": 2.717613467929415e-05, "loss": 0.2005, "step": 31165 }, { "epoch": 1.45, "learning_rate": 2.717566660945813e-05, "loss": 0.1581, "step": 31170 }, { "epoch": 1.45, "learning_rate": 2.717519853962211e-05, "loss": 0.144, "step": 31175 }, { "epoch": 1.45, "learning_rate": 2.7174730469786094e-05, "loss": 0.2286, "step": 31180 }, { "epoch": 1.46, "learning_rate": 2.7174262399950073e-05, "loss": 0.3269, "step": 31185 }, { "epoch": 1.46, "learning_rate": 2.7173794330114053e-05, "loss": 0.1078, "step": 31190 }, { "epoch": 1.46, "learning_rate": 2.7173326260278033e-05, "loss": 0.0554, "step": 31195 }, { "epoch": 1.46, "learning_rate": 2.7172858190442016e-05, "loss": 0.0881, "step": 31200 }, { "epoch": 1.46, "learning_rate": 2.7172390120605996e-05, "loss": 0.1346, "step": 31205 }, { "epoch": 1.46, "learning_rate": 2.7171922050769976e-05, "loss": 0.1582, "step": 31210 }, { "epoch": 1.46, "learning_rate": 2.7171453980933956e-05, "loss": 0.2103, "step": 31215 }, { "epoch": 1.46, "learning_rate": 2.717098591109794e-05, "loss": 0.1397, "step": 31220 }, { "epoch": 1.46, "learning_rate": 2.717051784126192e-05, "loss": 0.3329, "step": 31225 }, { "epoch": 1.46, "learning_rate": 2.7170049771425895e-05, "loss": 0.2672, "step": 31230 }, { "epoch": 1.46, "learning_rate": 2.7169581701589878e-05, "loss": 0.2304, "step": 31235 }, { "epoch": 1.46, "learning_rate": 2.7169113631753858e-05, "loss": 0.072, "step": 31240 }, { "epoch": 1.46, "learning_rate": 2.7168645561917838e-05, "loss": 0.0674, "step": 31245 }, { "epoch": 1.46, "learning_rate": 2.7168177492081818e-05, "loss": 0.0143, "step": 31250 }, { "epoch": 1.46, "learning_rate": 2.71677094222458e-05, "loss": 0.078, "step": 31255 }, { "epoch": 1.46, "learning_rate": 2.716724135240978e-05, "loss": 0.1475, "step": 31260 }, { "epoch": 1.46, "learning_rate": 2.716677328257376e-05, "loss": 0.1014, "step": 31265 }, { "epoch": 1.46, "learning_rate": 2.716630521273774e-05, "loss": 0.1701, "step": 31270 }, { "epoch": 1.46, "learning_rate": 2.7165837142901724e-05, "loss": 0.2333, "step": 31275 }, { "epoch": 1.46, "learning_rate": 2.7165369073065703e-05, "loss": 0.3492, "step": 31280 }, { "epoch": 1.46, "learning_rate": 2.7164901003229683e-05, "loss": 0.2331, "step": 31285 }, { "epoch": 1.46, "learning_rate": 2.7164432933393663e-05, "loss": 0.0265, "step": 31290 }, { "epoch": 1.46, "learning_rate": 2.7163964863557643e-05, "loss": 0.0504, "step": 31295 }, { "epoch": 1.46, "learning_rate": 2.7163496793721623e-05, "loss": 0.0923, "step": 31300 }, { "epoch": 1.46, "learning_rate": 2.7163028723885603e-05, "loss": 0.0899, "step": 31305 }, { "epoch": 1.46, "learning_rate": 2.7162560654049586e-05, "loss": 0.1123, "step": 31310 }, { "epoch": 1.46, "learning_rate": 2.7162092584213566e-05, "loss": 0.1174, "step": 31315 }, { "epoch": 1.46, "learning_rate": 2.7161624514377545e-05, "loss": 0.1602, "step": 31320 }, { "epoch": 1.46, "learning_rate": 2.7161156444541525e-05, "loss": 0.3227, "step": 31325 }, { "epoch": 1.46, "learning_rate": 2.716068837470551e-05, "loss": 0.2324, "step": 31330 }, { "epoch": 1.46, "learning_rate": 2.7160220304869488e-05, "loss": 0.3153, "step": 31335 }, { "epoch": 1.46, "learning_rate": 2.7159752235033468e-05, "loss": 0.0325, "step": 31340 }, { "epoch": 1.46, "learning_rate": 2.7159284165197448e-05, "loss": 0.0524, "step": 31345 }, { "epoch": 1.46, "learning_rate": 2.715881609536143e-05, "loss": 0.0863, "step": 31350 }, { "epoch": 1.46, "learning_rate": 2.7158348025525407e-05, "loss": 0.0851, "step": 31355 }, { "epoch": 1.46, "learning_rate": 2.7157879955689387e-05, "loss": 0.1045, "step": 31360 }, { "epoch": 1.46, "learning_rate": 2.715741188585337e-05, "loss": 0.1, "step": 31365 }, { "epoch": 1.46, "learning_rate": 2.715694381601735e-05, "loss": 0.1667, "step": 31370 }, { "epoch": 1.46, "learning_rate": 2.715647574618133e-05, "loss": 0.1546, "step": 31375 }, { "epoch": 1.46, "learning_rate": 2.715600767634531e-05, "loss": 0.3749, "step": 31380 }, { "epoch": 1.46, "learning_rate": 2.7155539606509293e-05, "loss": 0.3595, "step": 31385 }, { "epoch": 1.46, "learning_rate": 2.7155071536673273e-05, "loss": 0.0289, "step": 31390 }, { "epoch": 1.46, "learning_rate": 2.7154603466837253e-05, "loss": 0.0314, "step": 31395 }, { "epoch": 1.47, "learning_rate": 2.7154135397001233e-05, "loss": 0.0526, "step": 31400 }, { "epoch": 1.47, "learning_rate": 2.7153667327165216e-05, "loss": 0.0726, "step": 31405 }, { "epoch": 1.47, "learning_rate": 2.7153199257329196e-05, "loss": 0.1525, "step": 31410 }, { "epoch": 1.47, "learning_rate": 2.7152731187493175e-05, "loss": 0.0999, "step": 31415 }, { "epoch": 1.47, "learning_rate": 2.7152263117657155e-05, "loss": 0.1569, "step": 31420 }, { "epoch": 1.47, "learning_rate": 2.7151795047821135e-05, "loss": 0.2005, "step": 31425 }, { "epoch": 1.47, "learning_rate": 2.7151326977985115e-05, "loss": 0.3146, "step": 31430 }, { "epoch": 1.47, "learning_rate": 2.7150858908149095e-05, "loss": 0.2056, "step": 31435 }, { "epoch": 1.47, "learning_rate": 2.7150390838313078e-05, "loss": 0.0459, "step": 31440 }, { "epoch": 1.47, "learning_rate": 2.7149922768477058e-05, "loss": 0.0732, "step": 31445 }, { "epoch": 1.47, "learning_rate": 2.7149454698641038e-05, "loss": 0.1148, "step": 31450 }, { "epoch": 1.47, "learning_rate": 2.7148986628805017e-05, "loss": 0.089, "step": 31455 }, { "epoch": 1.47, "learning_rate": 2.7148518558969e-05, "loss": 0.1015, "step": 31460 }, { "epoch": 1.47, "learning_rate": 2.714805048913298e-05, "loss": 0.1189, "step": 31465 }, { "epoch": 1.47, "learning_rate": 2.714758241929696e-05, "loss": 0.1489, "step": 31470 }, { "epoch": 1.47, "learning_rate": 2.714711434946094e-05, "loss": 0.2937, "step": 31475 }, { "epoch": 1.47, "learning_rate": 2.714664627962492e-05, "loss": 0.4018, "step": 31480 }, { "epoch": 1.47, "learning_rate": 2.71461782097889e-05, "loss": 0.2134, "step": 31485 }, { "epoch": 1.47, "learning_rate": 2.714571013995288e-05, "loss": 0.0386, "step": 31490 }, { "epoch": 1.47, "learning_rate": 2.7145242070116863e-05, "loss": 0.0726, "step": 31495 }, { "epoch": 1.47, "learning_rate": 2.7144774000280843e-05, "loss": 0.0902, "step": 31500 }, { "epoch": 1.47, "learning_rate": 2.7144305930444822e-05, "loss": 0.0905, "step": 31505 }, { "epoch": 1.47, "learning_rate": 2.7143837860608802e-05, "loss": 0.128, "step": 31510 }, { "epoch": 1.47, "learning_rate": 2.7143369790772785e-05, "loss": 0.1366, "step": 31515 }, { "epoch": 1.47, "learning_rate": 2.7142901720936765e-05, "loss": 0.1131, "step": 31520 }, { "epoch": 1.47, "learning_rate": 2.7142433651100745e-05, "loss": 0.1289, "step": 31525 }, { "epoch": 1.47, "learning_rate": 2.7141965581264725e-05, "loss": 0.2813, "step": 31530 }, { "epoch": 1.47, "learning_rate": 2.7141497511428708e-05, "loss": 0.2638, "step": 31535 }, { "epoch": 1.47, "learning_rate": 2.7141029441592688e-05, "loss": 0.06, "step": 31540 }, { "epoch": 1.47, "learning_rate": 2.7140561371756664e-05, "loss": 0.0568, "step": 31545 }, { "epoch": 1.47, "learning_rate": 2.7140093301920647e-05, "loss": 0.0778, "step": 31550 }, { "epoch": 1.47, "learning_rate": 2.7139625232084627e-05, "loss": 0.0678, "step": 31555 }, { "epoch": 1.47, "learning_rate": 2.7139157162248607e-05, "loss": 0.0792, "step": 31560 }, { "epoch": 1.47, "learning_rate": 2.7138689092412587e-05, "loss": 0.1343, "step": 31565 }, { "epoch": 1.47, "learning_rate": 2.713822102257657e-05, "loss": 0.158, "step": 31570 }, { "epoch": 1.47, "learning_rate": 2.713775295274055e-05, "loss": 0.2188, "step": 31575 }, { "epoch": 1.47, "learning_rate": 2.713728488290453e-05, "loss": 0.2547, "step": 31580 }, { "epoch": 1.47, "learning_rate": 2.713681681306851e-05, "loss": 0.3107, "step": 31585 }, { "epoch": 1.47, "learning_rate": 2.7136348743232493e-05, "loss": 0.0808, "step": 31590 }, { "epoch": 1.47, "learning_rate": 2.7135880673396473e-05, "loss": 0.1067, "step": 31595 }, { "epoch": 1.47, "learning_rate": 2.7135412603560452e-05, "loss": 0.0573, "step": 31600 }, { "epoch": 1.47, "learning_rate": 2.7134944533724436e-05, "loss": 0.0894, "step": 31605 }, { "epoch": 1.47, "learning_rate": 2.7134476463888412e-05, "loss": 0.096, "step": 31610 }, { "epoch": 1.48, "learning_rate": 2.7134008394052392e-05, "loss": 0.1275, "step": 31615 }, { "epoch": 1.48, "learning_rate": 2.713354032421637e-05, "loss": 0.2443, "step": 31620 }, { "epoch": 1.48, "learning_rate": 2.7133072254380355e-05, "loss": 0.1851, "step": 31625 }, { "epoch": 1.48, "learning_rate": 2.7132604184544335e-05, "loss": 0.3059, "step": 31630 }, { "epoch": 1.48, "learning_rate": 2.7132136114708315e-05, "loss": 0.2894, "step": 31635 }, { "epoch": 1.48, "learning_rate": 2.7131668044872294e-05, "loss": 0.0236, "step": 31640 }, { "epoch": 1.48, "learning_rate": 2.7131199975036278e-05, "loss": 0.044, "step": 31645 }, { "epoch": 1.48, "learning_rate": 2.7130731905200257e-05, "loss": 0.1022, "step": 31650 }, { "epoch": 1.48, "learning_rate": 2.7130263835364237e-05, "loss": 0.0992, "step": 31655 }, { "epoch": 1.48, "learning_rate": 2.7129795765528217e-05, "loss": 0.135, "step": 31660 }, { "epoch": 1.48, "learning_rate": 2.71293276956922e-05, "loss": 0.2547, "step": 31665 }, { "epoch": 1.48, "learning_rate": 2.7128859625856177e-05, "loss": 0.1903, "step": 31670 }, { "epoch": 1.48, "learning_rate": 2.7128391556020156e-05, "loss": 0.3072, "step": 31675 }, { "epoch": 1.48, "learning_rate": 2.712792348618414e-05, "loss": 0.461, "step": 31680 }, { "epoch": 1.48, "learning_rate": 2.712745541634812e-05, "loss": 0.3273, "step": 31685 }, { "epoch": 1.48, "learning_rate": 2.71269873465121e-05, "loss": 0.0519, "step": 31690 }, { "epoch": 1.48, "learning_rate": 2.712651927667608e-05, "loss": 0.0846, "step": 31695 }, { "epoch": 1.48, "learning_rate": 2.7126051206840062e-05, "loss": 0.12, "step": 31700 }, { "epoch": 1.48, "learning_rate": 2.7125583137004042e-05, "loss": 0.0657, "step": 31705 }, { "epoch": 1.48, "learning_rate": 2.7125115067168022e-05, "loss": 0.0977, "step": 31710 }, { "epoch": 1.48, "learning_rate": 2.7124646997332002e-05, "loss": 0.1374, "step": 31715 }, { "epoch": 1.48, "learning_rate": 2.7124178927495985e-05, "loss": 0.1501, "step": 31720 }, { "epoch": 1.48, "learning_rate": 2.7123710857659965e-05, "loss": 0.2465, "step": 31725 }, { "epoch": 1.48, "learning_rate": 2.7123242787823945e-05, "loss": 0.1671, "step": 31730 }, { "epoch": 1.48, "learning_rate": 2.7122774717987924e-05, "loss": 0.2262, "step": 31735 }, { "epoch": 1.48, "learning_rate": 2.7122306648151904e-05, "loss": 0.0625, "step": 31740 }, { "epoch": 1.48, "learning_rate": 2.7121838578315884e-05, "loss": 0.0483, "step": 31745 }, { "epoch": 1.48, "learning_rate": 2.7121370508479864e-05, "loss": 0.0927, "step": 31750 }, { "epoch": 1.48, "learning_rate": 2.7120902438643847e-05, "loss": 0.079, "step": 31755 }, { "epoch": 1.48, "learning_rate": 2.7120434368807827e-05, "loss": 0.1691, "step": 31760 }, { "epoch": 1.48, "learning_rate": 2.7119966298971807e-05, "loss": 0.1397, "step": 31765 }, { "epoch": 1.48, "learning_rate": 2.7119498229135787e-05, "loss": 0.1657, "step": 31770 }, { "epoch": 1.48, "learning_rate": 2.711903015929977e-05, "loss": 0.2364, "step": 31775 }, { "epoch": 1.48, "learning_rate": 2.711856208946375e-05, "loss": 0.2493, "step": 31780 }, { "epoch": 1.48, "learning_rate": 2.711809401962773e-05, "loss": 0.2651, "step": 31785 }, { "epoch": 1.48, "learning_rate": 2.7117625949791713e-05, "loss": 0.0424, "step": 31790 }, { "epoch": 1.48, "learning_rate": 2.7117157879955692e-05, "loss": 0.0796, "step": 31795 }, { "epoch": 1.48, "learning_rate": 2.711668981011967e-05, "loss": 0.1075, "step": 31800 }, { "epoch": 1.48, "learning_rate": 2.711622174028365e-05, "loss": 0.1335, "step": 31805 }, { "epoch": 1.48, "learning_rate": 2.7115753670447632e-05, "loss": 0.4187, "step": 31810 }, { "epoch": 1.48, "learning_rate": 2.711528560061161e-05, "loss": 0.1783, "step": 31815 }, { "epoch": 1.48, "learning_rate": 2.711481753077559e-05, "loss": 0.1089, "step": 31820 }, { "epoch": 1.48, "learning_rate": 2.711434946093957e-05, "loss": 0.2229, "step": 31825 }, { "epoch": 1.49, "learning_rate": 2.7113881391103555e-05, "loss": 0.3265, "step": 31830 }, { "epoch": 1.49, "learning_rate": 2.7113413321267534e-05, "loss": 0.241, "step": 31835 }, { "epoch": 1.49, "learning_rate": 2.7112945251431514e-05, "loss": 0.0489, "step": 31840 }, { "epoch": 1.49, "learning_rate": 2.7112477181595497e-05, "loss": 0.0377, "step": 31845 }, { "epoch": 1.49, "learning_rate": 2.7112009111759477e-05, "loss": 0.0665, "step": 31850 }, { "epoch": 1.49, "learning_rate": 2.7111541041923457e-05, "loss": 0.0568, "step": 31855 }, { "epoch": 1.49, "learning_rate": 2.7111072972087433e-05, "loss": 0.1248, "step": 31860 }, { "epoch": 1.49, "learning_rate": 2.7110604902251417e-05, "loss": 0.109, "step": 31865 }, { "epoch": 1.49, "learning_rate": 2.7110136832415396e-05, "loss": 0.217, "step": 31870 }, { "epoch": 1.49, "learning_rate": 2.7109668762579376e-05, "loss": 0.2257, "step": 31875 }, { "epoch": 1.49, "learning_rate": 2.7109200692743356e-05, "loss": 0.3889, "step": 31880 }, { "epoch": 1.49, "learning_rate": 2.710873262290734e-05, "loss": 0.3604, "step": 31885 }, { "epoch": 1.49, "learning_rate": 2.710826455307132e-05, "loss": 0.0339, "step": 31890 }, { "epoch": 1.49, "learning_rate": 2.71077964832353e-05, "loss": 0.0946, "step": 31895 }, { "epoch": 1.49, "learning_rate": 2.710732841339928e-05, "loss": 0.1173, "step": 31900 }, { "epoch": 1.49, "learning_rate": 2.7106860343563262e-05, "loss": 0.091, "step": 31905 }, { "epoch": 1.49, "learning_rate": 2.7106392273727242e-05, "loss": 0.1903, "step": 31910 }, { "epoch": 1.49, "learning_rate": 2.710592420389122e-05, "loss": 0.1138, "step": 31915 }, { "epoch": 1.49, "learning_rate": 2.7105456134055205e-05, "loss": 0.1817, "step": 31920 }, { "epoch": 1.49, "learning_rate": 2.710498806421918e-05, "loss": 0.1822, "step": 31925 }, { "epoch": 1.49, "learning_rate": 2.710451999438316e-05, "loss": 0.2536, "step": 31930 }, { "epoch": 1.49, "learning_rate": 2.710405192454714e-05, "loss": 0.2957, "step": 31935 }, { "epoch": 1.49, "learning_rate": 2.7103583854711124e-05, "loss": 0.0512, "step": 31940 }, { "epoch": 1.49, "learning_rate": 2.7103115784875104e-05, "loss": 0.0712, "step": 31945 }, { "epoch": 1.49, "learning_rate": 2.7102647715039084e-05, "loss": 0.0949, "step": 31950 }, { "epoch": 1.49, "learning_rate": 2.7102179645203064e-05, "loss": 0.1251, "step": 31955 }, { "epoch": 1.49, "learning_rate": 2.7101711575367047e-05, "loss": 0.0888, "step": 31960 }, { "epoch": 1.49, "learning_rate": 2.7101243505531027e-05, "loss": 0.0918, "step": 31965 }, { "epoch": 1.49, "learning_rate": 2.7100775435695006e-05, "loss": 0.1947, "step": 31970 }, { "epoch": 1.49, "learning_rate": 2.710030736585899e-05, "loss": 0.293, "step": 31975 }, { "epoch": 1.49, "learning_rate": 2.709983929602297e-05, "loss": 0.2794, "step": 31980 }, { "epoch": 1.49, "learning_rate": 2.709937122618695e-05, "loss": 0.2422, "step": 31985 }, { "epoch": 1.49, "learning_rate": 2.7098903156350926e-05, "loss": 0.0326, "step": 31990 }, { "epoch": 1.49, "learning_rate": 2.709843508651491e-05, "loss": 0.0456, "step": 31995 }, { "epoch": 1.49, "learning_rate": 2.709796701667889e-05, "loss": 0.124, "step": 32000 }, { "epoch": 1.49, "learning_rate": 2.709749894684287e-05, "loss": 0.1717, "step": 32005 }, { "epoch": 1.49, "learning_rate": 2.7097030877006848e-05, "loss": 0.0744, "step": 32010 }, { "epoch": 1.49, "learning_rate": 2.709656280717083e-05, "loss": 0.154, "step": 32015 }, { "epoch": 1.49, "learning_rate": 2.709609473733481e-05, "loss": 0.0861, "step": 32020 }, { "epoch": 1.49, "learning_rate": 2.709562666749879e-05, "loss": 0.2772, "step": 32025 }, { "epoch": 1.49, "learning_rate": 2.7095158597662774e-05, "loss": 0.3768, "step": 32030 }, { "epoch": 1.49, "learning_rate": 2.7094690527826754e-05, "loss": 0.2509, "step": 32035 }, { "epoch": 1.5, "learning_rate": 2.7094222457990734e-05, "loss": 0.0797, "step": 32040 }, { "epoch": 1.5, "learning_rate": 2.7093754388154714e-05, "loss": 0.0646, "step": 32045 }, { "epoch": 1.5, "learning_rate": 2.7093286318318694e-05, "loss": 0.1134, "step": 32050 }, { "epoch": 1.5, "learning_rate": 2.7092818248482673e-05, "loss": 0.0643, "step": 32055 }, { "epoch": 1.5, "learning_rate": 2.7092350178646653e-05, "loss": 0.111, "step": 32060 }, { "epoch": 1.5, "learning_rate": 2.7091882108810633e-05, "loss": 0.1633, "step": 32065 }, { "epoch": 1.5, "learning_rate": 2.7091414038974616e-05, "loss": 0.1374, "step": 32070 }, { "epoch": 1.5, "learning_rate": 2.7090945969138596e-05, "loss": 0.2281, "step": 32075 }, { "epoch": 1.5, "learning_rate": 2.7090477899302576e-05, "loss": 0.3332, "step": 32080 }, { "epoch": 1.5, "learning_rate": 2.7090009829466556e-05, "loss": 0.2668, "step": 32085 }, { "epoch": 1.5, "learning_rate": 2.708954175963054e-05, "loss": 0.0363, "step": 32090 }, { "epoch": 1.5, "learning_rate": 2.708907368979452e-05, "loss": 0.0522, "step": 32095 }, { "epoch": 1.5, "learning_rate": 2.70886056199585e-05, "loss": 0.1357, "step": 32100 }, { "epoch": 1.5, "learning_rate": 2.7088137550122482e-05, "loss": 0.1301, "step": 32105 }, { "epoch": 1.5, "learning_rate": 2.708766948028646e-05, "loss": 0.1362, "step": 32110 }, { "epoch": 1.5, "learning_rate": 2.7087201410450438e-05, "loss": 0.0767, "step": 32115 }, { "epoch": 1.5, "learning_rate": 2.7086733340614418e-05, "loss": 0.171, "step": 32120 }, { "epoch": 1.5, "learning_rate": 2.70862652707784e-05, "loss": 0.1663, "step": 32125 }, { "epoch": 1.5, "learning_rate": 2.708579720094238e-05, "loss": 0.3131, "step": 32130 }, { "epoch": 1.5, "learning_rate": 2.708532913110636e-05, "loss": 0.2447, "step": 32135 }, { "epoch": 1.5, "learning_rate": 2.708486106127034e-05, "loss": 0.0599, "step": 32140 }, { "epoch": 1.5, "learning_rate": 2.7084392991434324e-05, "loss": 0.0719, "step": 32145 }, { "epoch": 1.5, "learning_rate": 2.7083924921598304e-05, "loss": 0.0617, "step": 32150 }, { "epoch": 1.5, "learning_rate": 2.7083456851762283e-05, "loss": 0.1061, "step": 32155 }, { "epoch": 1.5, "learning_rate": 2.7082988781926267e-05, "loss": 0.1063, "step": 32160 }, { "epoch": 1.5, "learning_rate": 2.7082520712090246e-05, "loss": 0.1159, "step": 32165 }, { "epoch": 1.5, "learning_rate": 2.7082052642254226e-05, "loss": 0.2318, "step": 32170 }, { "epoch": 1.5, "learning_rate": 2.7081584572418206e-05, "loss": 0.2681, "step": 32175 }, { "epoch": 1.5, "learning_rate": 2.7081116502582186e-05, "loss": 0.3604, "step": 32180 }, { "epoch": 1.5, "learning_rate": 2.7080648432746166e-05, "loss": 0.2109, "step": 32185 }, { "epoch": 1.5, "learning_rate": 2.7080180362910145e-05, "loss": 0.0237, "step": 32190 }, { "epoch": 1.5, "learning_rate": 2.7079712293074125e-05, "loss": 0.0279, "step": 32195 }, { "epoch": 1.5, "learning_rate": 2.707924422323811e-05, "loss": 0.1434, "step": 32200 }, { "epoch": 1.5, "learning_rate": 2.7078776153402088e-05, "loss": 0.062, "step": 32205 }, { "epoch": 1.5, "learning_rate": 2.7078308083566068e-05, "loss": 0.092, "step": 32210 }, { "epoch": 1.5, "learning_rate": 2.707784001373005e-05, "loss": 0.061, "step": 32215 }, { "epoch": 1.5, "learning_rate": 2.707737194389403e-05, "loss": 0.1411, "step": 32220 }, { "epoch": 1.5, "learning_rate": 2.707690387405801e-05, "loss": 0.1898, "step": 32225 }, { "epoch": 1.5, "learning_rate": 2.707643580422199e-05, "loss": 0.3524, "step": 32230 }, { "epoch": 1.5, "learning_rate": 2.7075967734385974e-05, "loss": 0.293, "step": 32235 }, { "epoch": 1.5, "learning_rate": 2.707549966454995e-05, "loss": 0.0641, "step": 32240 }, { "epoch": 1.5, "learning_rate": 2.707503159471393e-05, "loss": 0.0273, "step": 32245 }, { "epoch": 1.5, "learning_rate": 2.707456352487791e-05, "loss": 0.0713, "step": 32250 }, { "epoch": 1.51, "learning_rate": 2.7074095455041893e-05, "loss": 0.1019, "step": 32255 }, { "epoch": 1.51, "learning_rate": 2.7073627385205873e-05, "loss": 0.0929, "step": 32260 }, { "epoch": 1.51, "learning_rate": 2.7073159315369853e-05, "loss": 0.1088, "step": 32265 }, { "epoch": 1.51, "learning_rate": 2.7072691245533833e-05, "loss": 0.0874, "step": 32270 }, { "epoch": 1.51, "learning_rate": 2.7072223175697816e-05, "loss": 0.271, "step": 32275 }, { "epoch": 1.51, "learning_rate": 2.7071755105861796e-05, "loss": 0.3929, "step": 32280 }, { "epoch": 1.51, "learning_rate": 2.7071287036025776e-05, "loss": 0.351, "step": 32285 }, { "epoch": 1.51, "learning_rate": 2.707081896618976e-05, "loss": 0.0732, "step": 32290 }, { "epoch": 1.51, "learning_rate": 2.707035089635374e-05, "loss": 0.0978, "step": 32295 }, { "epoch": 1.51, "learning_rate": 2.706988282651772e-05, "loss": 0.0841, "step": 32300 }, { "epoch": 1.51, "learning_rate": 2.7069414756681695e-05, "loss": 0.104, "step": 32305 }, { "epoch": 1.51, "learning_rate": 2.7068946686845678e-05, "loss": 0.1508, "step": 32310 }, { "epoch": 1.51, "learning_rate": 2.7068478617009658e-05, "loss": 0.1151, "step": 32315 }, { "epoch": 1.51, "learning_rate": 2.7068010547173638e-05, "loss": 0.2606, "step": 32320 }, { "epoch": 1.51, "learning_rate": 2.7067542477337617e-05, "loss": 0.2159, "step": 32325 }, { "epoch": 1.51, "learning_rate": 2.70670744075016e-05, "loss": 0.2934, "step": 32330 }, { "epoch": 1.51, "learning_rate": 2.706660633766558e-05, "loss": 0.3062, "step": 32335 }, { "epoch": 1.51, "learning_rate": 2.706613826782956e-05, "loss": 0.0399, "step": 32340 }, { "epoch": 1.51, "learning_rate": 2.7065670197993543e-05, "loss": 0.0873, "step": 32345 }, { "epoch": 1.51, "learning_rate": 2.7065202128157523e-05, "loss": 0.0339, "step": 32350 }, { "epoch": 1.51, "learning_rate": 2.7064734058321503e-05, "loss": 0.077, "step": 32355 }, { "epoch": 1.51, "learning_rate": 2.7064265988485483e-05, "loss": 0.261, "step": 32360 }, { "epoch": 1.51, "learning_rate": 2.7063797918649463e-05, "loss": 0.0711, "step": 32365 }, { "epoch": 1.51, "learning_rate": 2.7063329848813443e-05, "loss": 0.1113, "step": 32370 }, { "epoch": 1.51, "learning_rate": 2.7062861778977422e-05, "loss": 0.2722, "step": 32375 }, { "epoch": 1.51, "learning_rate": 2.7062393709141402e-05, "loss": 0.3051, "step": 32380 }, { "epoch": 1.51, "learning_rate": 2.7061925639305385e-05, "loss": 0.2774, "step": 32385 }, { "epoch": 1.51, "learning_rate": 2.7061457569469365e-05, "loss": 0.0502, "step": 32390 }, { "epoch": 1.51, "learning_rate": 2.7060989499633345e-05, "loss": 0.0761, "step": 32395 }, { "epoch": 1.51, "learning_rate": 2.7060521429797328e-05, "loss": 0.0651, "step": 32400 }, { "epoch": 1.51, "learning_rate": 2.7060053359961308e-05, "loss": 0.1633, "step": 32405 }, { "epoch": 1.51, "learning_rate": 2.7059585290125288e-05, "loss": 0.0858, "step": 32410 }, { "epoch": 1.51, "learning_rate": 2.7059117220289268e-05, "loss": 0.073, "step": 32415 }, { "epoch": 1.51, "learning_rate": 2.705864915045325e-05, "loss": 0.1283, "step": 32420 }, { "epoch": 1.51, "learning_rate": 2.705818108061723e-05, "loss": 0.2236, "step": 32425 }, { "epoch": 1.51, "learning_rate": 2.7057713010781207e-05, "loss": 0.1609, "step": 32430 }, { "epoch": 1.51, "learning_rate": 2.7057244940945187e-05, "loss": 0.3077, "step": 32435 }, { "epoch": 1.51, "learning_rate": 2.705677687110917e-05, "loss": 0.0376, "step": 32440 }, { "epoch": 1.51, "learning_rate": 2.705630880127315e-05, "loss": 0.0517, "step": 32445 }, { "epoch": 1.51, "learning_rate": 2.705584073143713e-05, "loss": 0.073, "step": 32450 }, { "epoch": 1.51, "learning_rate": 2.7055372661601113e-05, "loss": 0.1018, "step": 32455 }, { "epoch": 1.51, "learning_rate": 2.7054904591765093e-05, "loss": 0.1511, "step": 32460 }, { "epoch": 1.51, "learning_rate": 2.7054436521929073e-05, "loss": 0.1133, "step": 32465 }, { "epoch": 1.52, "learning_rate": 2.7053968452093052e-05, "loss": 0.1803, "step": 32470 }, { "epoch": 1.52, "learning_rate": 2.7053500382257036e-05, "loss": 0.1564, "step": 32475 }, { "epoch": 1.52, "learning_rate": 2.7053032312421016e-05, "loss": 0.2363, "step": 32480 }, { "epoch": 1.52, "learning_rate": 2.7052564242584995e-05, "loss": 0.2663, "step": 32485 }, { "epoch": 1.52, "learning_rate": 2.7052096172748975e-05, "loss": 0.0317, "step": 32490 }, { "epoch": 1.52, "learning_rate": 2.7051628102912955e-05, "loss": 0.0785, "step": 32495 }, { "epoch": 1.52, "learning_rate": 2.7051160033076935e-05, "loss": 0.117, "step": 32500 }, { "epoch": 1.52, "learning_rate": 2.7050691963240915e-05, "loss": 0.1065, "step": 32505 }, { "epoch": 1.52, "learning_rate": 2.7050223893404894e-05, "loss": 0.1299, "step": 32510 }, { "epoch": 1.52, "learning_rate": 2.7049755823568878e-05, "loss": 0.097, "step": 32515 }, { "epoch": 1.52, "learning_rate": 2.7049287753732857e-05, "loss": 0.2074, "step": 32520 }, { "epoch": 1.52, "learning_rate": 2.7048819683896837e-05, "loss": 0.2138, "step": 32525 }, { "epoch": 1.52, "learning_rate": 2.704835161406082e-05, "loss": 0.2639, "step": 32530 }, { "epoch": 1.52, "learning_rate": 2.70478835442248e-05, "loss": 0.3218, "step": 32535 }, { "epoch": 1.52, "learning_rate": 2.704741547438878e-05, "loss": 0.0558, "step": 32540 }, { "epoch": 1.52, "learning_rate": 2.704694740455276e-05, "loss": 0.0935, "step": 32545 }, { "epoch": 1.52, "learning_rate": 2.7046479334716743e-05, "loss": 0.0543, "step": 32550 }, { "epoch": 1.52, "learning_rate": 2.704601126488072e-05, "loss": 0.0953, "step": 32555 }, { "epoch": 1.52, "learning_rate": 2.70455431950447e-05, "loss": 0.0852, "step": 32560 }, { "epoch": 1.52, "learning_rate": 2.704507512520868e-05, "loss": 0.179, "step": 32565 }, { "epoch": 1.52, "learning_rate": 2.7044607055372662e-05, "loss": 0.2043, "step": 32570 }, { "epoch": 1.52, "learning_rate": 2.7044138985536642e-05, "loss": 0.2253, "step": 32575 }, { "epoch": 1.52, "learning_rate": 2.7043670915700622e-05, "loss": 0.3692, "step": 32580 }, { "epoch": 1.52, "learning_rate": 2.7043202845864605e-05, "loss": 0.1366, "step": 32585 }, { "epoch": 1.52, "learning_rate": 2.7042734776028585e-05, "loss": 0.0413, "step": 32590 }, { "epoch": 1.52, "learning_rate": 2.7042266706192565e-05, "loss": 0.0573, "step": 32595 }, { "epoch": 1.52, "learning_rate": 2.7041798636356545e-05, "loss": 0.0395, "step": 32600 }, { "epoch": 1.52, "learning_rate": 2.7041330566520528e-05, "loss": 0.1138, "step": 32605 }, { "epoch": 1.52, "learning_rate": 2.7040862496684508e-05, "loss": 0.0869, "step": 32610 }, { "epoch": 1.52, "learning_rate": 2.7040394426848488e-05, "loss": 0.1003, "step": 32615 }, { "epoch": 1.52, "learning_rate": 2.7039926357012464e-05, "loss": 0.0849, "step": 32620 }, { "epoch": 1.52, "learning_rate": 2.7039458287176447e-05, "loss": 0.2475, "step": 32625 }, { "epoch": 1.52, "learning_rate": 2.7038990217340427e-05, "loss": 0.329, "step": 32630 }, { "epoch": 1.52, "learning_rate": 2.7038522147504407e-05, "loss": 0.2789, "step": 32635 }, { "epoch": 1.52, "learning_rate": 2.703805407766839e-05, "loss": 0.0185, "step": 32640 }, { "epoch": 1.52, "learning_rate": 2.703758600783237e-05, "loss": 0.1266, "step": 32645 }, { "epoch": 1.52, "learning_rate": 2.703711793799635e-05, "loss": 0.0976, "step": 32650 }, { "epoch": 1.52, "learning_rate": 2.703664986816033e-05, "loss": 0.1195, "step": 32655 }, { "epoch": 1.52, "learning_rate": 2.7036181798324313e-05, "loss": 0.0895, "step": 32660 }, { "epoch": 1.52, "learning_rate": 2.7035713728488292e-05, "loss": 0.0975, "step": 32665 }, { "epoch": 1.52, "learning_rate": 2.7035245658652272e-05, "loss": 0.1516, "step": 32670 }, { "epoch": 1.52, "learning_rate": 2.7034777588816252e-05, "loss": 0.2423, "step": 32675 }, { "epoch": 1.52, "learning_rate": 2.7034309518980232e-05, "loss": 0.4485, "step": 32680 }, { "epoch": 1.53, "learning_rate": 2.7033841449144212e-05, "loss": 0.2426, "step": 32685 }, { "epoch": 1.53, "learning_rate": 2.703337337930819e-05, "loss": 0.0359, "step": 32690 }, { "epoch": 1.53, "learning_rate": 2.703290530947217e-05, "loss": 0.0842, "step": 32695 }, { "epoch": 1.53, "learning_rate": 2.7032437239636155e-05, "loss": 0.185, "step": 32700 }, { "epoch": 1.53, "learning_rate": 2.7031969169800134e-05, "loss": 0.1004, "step": 32705 }, { "epoch": 1.53, "learning_rate": 2.7031501099964114e-05, "loss": 0.1709, "step": 32710 }, { "epoch": 1.53, "learning_rate": 2.7031033030128097e-05, "loss": 0.1358, "step": 32715 }, { "epoch": 1.53, "learning_rate": 2.7030564960292077e-05, "loss": 0.2217, "step": 32720 }, { "epoch": 1.53, "learning_rate": 2.7030096890456057e-05, "loss": 0.173, "step": 32725 }, { "epoch": 1.53, "learning_rate": 2.7029628820620037e-05, "loss": 0.2953, "step": 32730 }, { "epoch": 1.53, "learning_rate": 2.702916075078402e-05, "loss": 0.2254, "step": 32735 }, { "epoch": 1.53, "learning_rate": 2.7028692680948e-05, "loss": 0.0873, "step": 32740 }, { "epoch": 1.53, "learning_rate": 2.7028224611111976e-05, "loss": 0.0436, "step": 32745 }, { "epoch": 1.53, "learning_rate": 2.7027756541275956e-05, "loss": 0.0812, "step": 32750 }, { "epoch": 1.53, "learning_rate": 2.702728847143994e-05, "loss": 0.1044, "step": 32755 }, { "epoch": 1.53, "learning_rate": 2.702682040160392e-05, "loss": 0.0775, "step": 32760 }, { "epoch": 1.53, "learning_rate": 2.70263523317679e-05, "loss": 0.1807, "step": 32765 }, { "epoch": 1.53, "learning_rate": 2.7025884261931882e-05, "loss": 0.159, "step": 32770 }, { "epoch": 1.53, "learning_rate": 2.7025416192095862e-05, "loss": 0.2465, "step": 32775 }, { "epoch": 1.53, "learning_rate": 2.7024948122259842e-05, "loss": 0.4615, "step": 32780 }, { "epoch": 1.53, "learning_rate": 2.702448005242382e-05, "loss": 0.2719, "step": 32785 }, { "epoch": 1.53, "learning_rate": 2.7024011982587805e-05, "loss": 0.0534, "step": 32790 }, { "epoch": 1.53, "learning_rate": 2.7023543912751785e-05, "loss": 0.0485, "step": 32795 }, { "epoch": 1.53, "learning_rate": 2.7023075842915765e-05, "loss": 0.0531, "step": 32800 }, { "epoch": 1.53, "learning_rate": 2.7022607773079744e-05, "loss": 0.1008, "step": 32805 }, { "epoch": 1.53, "learning_rate": 2.7022139703243724e-05, "loss": 0.1294, "step": 32810 }, { "epoch": 1.53, "learning_rate": 2.7021671633407704e-05, "loss": 0.1395, "step": 32815 }, { "epoch": 1.53, "learning_rate": 2.7021203563571684e-05, "loss": 0.2705, "step": 32820 }, { "epoch": 1.53, "learning_rate": 2.7020735493735667e-05, "loss": 0.1477, "step": 32825 }, { "epoch": 1.53, "learning_rate": 2.7020267423899647e-05, "loss": 0.3652, "step": 32830 }, { "epoch": 1.53, "learning_rate": 2.7019799354063627e-05, "loss": 0.2274, "step": 32835 }, { "epoch": 1.53, "learning_rate": 2.7019331284227606e-05, "loss": 0.0386, "step": 32840 }, { "epoch": 1.53, "learning_rate": 2.701886321439159e-05, "loss": 0.0185, "step": 32845 }, { "epoch": 1.53, "learning_rate": 2.701839514455557e-05, "loss": 0.0817, "step": 32850 }, { "epoch": 1.53, "learning_rate": 2.701792707471955e-05, "loss": 0.0923, "step": 32855 }, { "epoch": 1.53, "learning_rate": 2.701745900488353e-05, "loss": 0.1459, "step": 32860 }, { "epoch": 1.53, "learning_rate": 2.7016990935047512e-05, "loss": 0.1658, "step": 32865 }, { "epoch": 1.53, "learning_rate": 2.701652286521149e-05, "loss": 0.15, "step": 32870 }, { "epoch": 1.53, "learning_rate": 2.701605479537547e-05, "loss": 0.1887, "step": 32875 }, { "epoch": 1.53, "learning_rate": 2.701558672553945e-05, "loss": 0.3236, "step": 32880 }, { "epoch": 1.53, "learning_rate": 2.701511865570343e-05, "loss": 0.331, "step": 32885 }, { "epoch": 1.53, "learning_rate": 2.701465058586741e-05, "loss": 0.0743, "step": 32890 }, { "epoch": 1.53, "learning_rate": 2.701418251603139e-05, "loss": 0.0395, "step": 32895 }, { "epoch": 1.54, "learning_rate": 2.7013714446195374e-05, "loss": 0.0998, "step": 32900 }, { "epoch": 1.54, "learning_rate": 2.7013246376359354e-05, "loss": 0.0713, "step": 32905 }, { "epoch": 1.54, "learning_rate": 2.7012778306523334e-05, "loss": 0.1306, "step": 32910 }, { "epoch": 1.54, "learning_rate": 2.7012310236687314e-05, "loss": 0.1828, "step": 32915 }, { "epoch": 1.54, "learning_rate": 2.7011842166851297e-05, "loss": 0.1994, "step": 32920 }, { "epoch": 1.54, "learning_rate": 2.7011374097015277e-05, "loss": 0.2601, "step": 32925 }, { "epoch": 1.54, "learning_rate": 2.7010906027179257e-05, "loss": 0.4279, "step": 32930 }, { "epoch": 1.54, "learning_rate": 2.7010437957343233e-05, "loss": 0.3181, "step": 32935 }, { "epoch": 1.54, "learning_rate": 2.7009969887507216e-05, "loss": 0.0657, "step": 32940 }, { "epoch": 1.54, "learning_rate": 2.7009501817671196e-05, "loss": 0.0556, "step": 32945 }, { "epoch": 1.54, "learning_rate": 2.7009033747835176e-05, "loss": 0.1159, "step": 32950 }, { "epoch": 1.54, "learning_rate": 2.700856567799916e-05, "loss": 0.1832, "step": 32955 }, { "epoch": 1.54, "learning_rate": 2.700809760816314e-05, "loss": 0.1578, "step": 32960 }, { "epoch": 1.54, "learning_rate": 2.700762953832712e-05, "loss": 0.0559, "step": 32965 }, { "epoch": 1.54, "learning_rate": 2.70071614684911e-05, "loss": 0.2554, "step": 32970 }, { "epoch": 1.54, "learning_rate": 2.7006693398655082e-05, "loss": 0.2261, "step": 32975 }, { "epoch": 1.54, "learning_rate": 2.700622532881906e-05, "loss": 0.4121, "step": 32980 }, { "epoch": 1.54, "learning_rate": 2.700575725898304e-05, "loss": 0.2396, "step": 32985 }, { "epoch": 1.54, "learning_rate": 2.700528918914702e-05, "loss": 0.1215, "step": 32990 }, { "epoch": 1.54, "learning_rate": 2.7004821119311004e-05, "loss": 0.0893, "step": 32995 }, { "epoch": 1.54, "learning_rate": 2.700435304947498e-05, "loss": 0.0738, "step": 33000 }, { "epoch": 1.54, "learning_rate": 2.700388497963896e-05, "loss": 0.068, "step": 33005 }, { "epoch": 1.54, "learning_rate": 2.7003416909802944e-05, "loss": 0.1042, "step": 33010 }, { "epoch": 1.54, "learning_rate": 2.7002948839966924e-05, "loss": 0.2321, "step": 33015 }, { "epoch": 1.54, "learning_rate": 2.7002480770130904e-05, "loss": 0.1587, "step": 33020 }, { "epoch": 1.54, "learning_rate": 2.7002012700294883e-05, "loss": 0.2503, "step": 33025 }, { "epoch": 1.54, "learning_rate": 2.7001544630458867e-05, "loss": 0.3625, "step": 33030 }, { "epoch": 1.54, "learning_rate": 2.7001076560622846e-05, "loss": 0.3232, "step": 33035 }, { "epoch": 1.54, "learning_rate": 2.7000608490786826e-05, "loss": 0.0347, "step": 33040 }, { "epoch": 1.54, "learning_rate": 2.7000140420950806e-05, "loss": 0.0626, "step": 33045 }, { "epoch": 1.54, "learning_rate": 2.699967235111479e-05, "loss": 0.0422, "step": 33050 }, { "epoch": 1.54, "learning_rate": 2.699920428127877e-05, "loss": 0.1606, "step": 33055 }, { "epoch": 1.54, "learning_rate": 2.6998736211442746e-05, "loss": 0.0907, "step": 33060 }, { "epoch": 1.54, "learning_rate": 2.699826814160673e-05, "loss": 0.0803, "step": 33065 }, { "epoch": 1.54, "learning_rate": 2.699780007177071e-05, "loss": 0.2084, "step": 33070 }, { "epoch": 1.54, "learning_rate": 2.699733200193469e-05, "loss": 0.2523, "step": 33075 }, { "epoch": 1.54, "learning_rate": 2.6996863932098668e-05, "loss": 0.2984, "step": 33080 }, { "epoch": 1.54, "learning_rate": 2.699639586226265e-05, "loss": 0.2349, "step": 33085 }, { "epoch": 1.54, "learning_rate": 2.699592779242663e-05, "loss": 0.0274, "step": 33090 }, { "epoch": 1.54, "learning_rate": 2.699545972259061e-05, "loss": 0.0587, "step": 33095 }, { "epoch": 1.54, "learning_rate": 2.699499165275459e-05, "loss": 0.0998, "step": 33100 }, { "epoch": 1.54, "learning_rate": 2.6994523582918574e-05, "loss": 0.0799, "step": 33105 }, { "epoch": 1.54, "learning_rate": 2.6994055513082554e-05, "loss": 0.1107, "step": 33110 }, { "epoch": 1.55, "learning_rate": 2.6993587443246534e-05, "loss": 0.1558, "step": 33115 }, { "epoch": 1.55, "learning_rate": 2.6993119373410513e-05, "loss": 0.1899, "step": 33120 }, { "epoch": 1.55, "learning_rate": 2.6992651303574493e-05, "loss": 0.104, "step": 33125 }, { "epoch": 1.55, "learning_rate": 2.6992183233738473e-05, "loss": 0.3399, "step": 33130 }, { "epoch": 1.55, "learning_rate": 2.6991715163902453e-05, "loss": 0.2708, "step": 33135 }, { "epoch": 1.55, "learning_rate": 2.6991247094066436e-05, "loss": 0.0323, "step": 33140 }, { "epoch": 1.55, "learning_rate": 2.6990779024230416e-05, "loss": 0.0878, "step": 33145 }, { "epoch": 1.55, "learning_rate": 2.6990310954394396e-05, "loss": 0.0769, "step": 33150 }, { "epoch": 1.55, "learning_rate": 2.6989842884558376e-05, "loss": 0.1322, "step": 33155 }, { "epoch": 1.55, "learning_rate": 2.698937481472236e-05, "loss": 0.1286, "step": 33160 }, { "epoch": 1.55, "learning_rate": 2.698890674488634e-05, "loss": 0.1078, "step": 33165 }, { "epoch": 1.55, "learning_rate": 2.698843867505032e-05, "loss": 0.1352, "step": 33170 }, { "epoch": 1.55, "learning_rate": 2.6987970605214298e-05, "loss": 0.2099, "step": 33175 }, { "epoch": 1.55, "learning_rate": 2.698750253537828e-05, "loss": 0.3099, "step": 33180 }, { "epoch": 1.55, "learning_rate": 2.698703446554226e-05, "loss": 0.2091, "step": 33185 }, { "epoch": 1.55, "learning_rate": 2.6986566395706238e-05, "loss": 0.0674, "step": 33190 }, { "epoch": 1.55, "learning_rate": 2.698609832587022e-05, "loss": 0.0518, "step": 33195 }, { "epoch": 1.55, "learning_rate": 2.69856302560342e-05, "loss": 0.0735, "step": 33200 }, { "epoch": 1.55, "learning_rate": 2.698516218619818e-05, "loss": 0.0593, "step": 33205 }, { "epoch": 1.55, "learning_rate": 2.698469411636216e-05, "loss": 0.1139, "step": 33210 }, { "epoch": 1.55, "learning_rate": 2.6984226046526144e-05, "loss": 0.1457, "step": 33215 }, { "epoch": 1.55, "learning_rate": 2.6983757976690123e-05, "loss": 0.2006, "step": 33220 }, { "epoch": 1.55, "learning_rate": 2.6983289906854103e-05, "loss": 0.2604, "step": 33225 }, { "epoch": 1.55, "learning_rate": 2.6982821837018083e-05, "loss": 0.2674, "step": 33230 }, { "epoch": 1.55, "learning_rate": 2.6982353767182066e-05, "loss": 0.2838, "step": 33235 }, { "epoch": 1.55, "learning_rate": 2.6981885697346046e-05, "loss": 0.0742, "step": 33240 }, { "epoch": 1.55, "learning_rate": 2.6981417627510026e-05, "loss": 0.0737, "step": 33245 }, { "epoch": 1.55, "learning_rate": 2.6980949557674006e-05, "loss": 0.0998, "step": 33250 }, { "epoch": 1.55, "learning_rate": 2.6980481487837986e-05, "loss": 0.0874, "step": 33255 }, { "epoch": 1.55, "learning_rate": 2.6980013418001965e-05, "loss": 0.0858, "step": 33260 }, { "epoch": 1.55, "learning_rate": 2.6979545348165945e-05, "loss": 0.1706, "step": 33265 }, { "epoch": 1.55, "learning_rate": 2.697907727832993e-05, "loss": 0.2193, "step": 33270 }, { "epoch": 1.55, "learning_rate": 2.6978609208493908e-05, "loss": 0.1598, "step": 33275 }, { "epoch": 1.55, "learning_rate": 2.6978141138657888e-05, "loss": 0.3171, "step": 33280 }, { "epoch": 1.55, "learning_rate": 2.6977673068821868e-05, "loss": 0.4434, "step": 33285 }, { "epoch": 1.55, "learning_rate": 2.697720499898585e-05, "loss": 0.0919, "step": 33290 }, { "epoch": 1.55, "learning_rate": 2.697673692914983e-05, "loss": 0.0747, "step": 33295 }, { "epoch": 1.55, "learning_rate": 2.697626885931381e-05, "loss": 0.0845, "step": 33300 }, { "epoch": 1.55, "learning_rate": 2.697580078947779e-05, "loss": 0.0738, "step": 33305 }, { "epoch": 1.55, "learning_rate": 2.6975332719641774e-05, "loss": 0.0936, "step": 33310 }, { "epoch": 1.55, "learning_rate": 2.697486464980575e-05, "loss": 0.1804, "step": 33315 }, { "epoch": 1.55, "learning_rate": 2.697439657996973e-05, "loss": 0.1244, "step": 33320 }, { "epoch": 1.55, "learning_rate": 2.6973928510133713e-05, "loss": 0.2265, "step": 33325 }, { "epoch": 1.56, "learning_rate": 2.6973460440297693e-05, "loss": 0.2955, "step": 33330 }, { "epoch": 1.56, "learning_rate": 2.6972992370461673e-05, "loss": 0.2679, "step": 33335 }, { "epoch": 1.56, "learning_rate": 2.6972524300625653e-05, "loss": 0.0258, "step": 33340 }, { "epoch": 1.56, "learning_rate": 2.6972056230789636e-05, "loss": 0.0695, "step": 33345 }, { "epoch": 1.56, "learning_rate": 2.6971588160953616e-05, "loss": 0.0812, "step": 33350 }, { "epoch": 1.56, "learning_rate": 2.6971120091117595e-05, "loss": 0.0363, "step": 33355 }, { "epoch": 1.56, "learning_rate": 2.6970652021281575e-05, "loss": 0.1397, "step": 33360 }, { "epoch": 1.56, "learning_rate": 2.697018395144556e-05, "loss": 0.1164, "step": 33365 }, { "epoch": 1.56, "learning_rate": 2.6969715881609538e-05, "loss": 0.1975, "step": 33370 }, { "epoch": 1.56, "learning_rate": 2.6969247811773518e-05, "loss": 0.2331, "step": 33375 }, { "epoch": 1.56, "learning_rate": 2.6968779741937498e-05, "loss": 0.3827, "step": 33380 }, { "epoch": 1.56, "learning_rate": 2.6968311672101478e-05, "loss": 0.2749, "step": 33385 }, { "epoch": 1.56, "learning_rate": 2.6967843602265458e-05, "loss": 0.0281, "step": 33390 }, { "epoch": 1.56, "learning_rate": 2.6967375532429437e-05, "loss": 0.1315, "step": 33395 }, { "epoch": 1.56, "learning_rate": 2.696690746259342e-05, "loss": 0.0946, "step": 33400 }, { "epoch": 1.56, "learning_rate": 2.69664393927574e-05, "loss": 0.1075, "step": 33405 }, { "epoch": 1.56, "learning_rate": 2.696597132292138e-05, "loss": 0.0648, "step": 33410 }, { "epoch": 1.56, "learning_rate": 2.696550325308536e-05, "loss": 0.1504, "step": 33415 }, { "epoch": 1.56, "learning_rate": 2.6965035183249343e-05, "loss": 0.1815, "step": 33420 }, { "epoch": 1.56, "learning_rate": 2.6964567113413323e-05, "loss": 0.246, "step": 33425 }, { "epoch": 1.56, "learning_rate": 2.6964099043577303e-05, "loss": 0.347, "step": 33430 }, { "epoch": 1.56, "learning_rate": 2.6963630973741286e-05, "loss": 0.3222, "step": 33435 }, { "epoch": 1.56, "learning_rate": 2.6963162903905262e-05, "loss": 0.0387, "step": 33440 }, { "epoch": 1.56, "learning_rate": 2.6962694834069242e-05, "loss": 0.0481, "step": 33445 }, { "epoch": 1.56, "learning_rate": 2.6962226764233222e-05, "loss": 0.0647, "step": 33450 }, { "epoch": 1.56, "learning_rate": 2.6961758694397205e-05, "loss": 0.0807, "step": 33455 }, { "epoch": 1.56, "learning_rate": 2.6961290624561185e-05, "loss": 0.1542, "step": 33460 }, { "epoch": 1.56, "learning_rate": 2.6960822554725165e-05, "loss": 0.1546, "step": 33465 }, { "epoch": 1.56, "learning_rate": 2.6960354484889145e-05, "loss": 0.1595, "step": 33470 }, { "epoch": 1.56, "learning_rate": 2.6959886415053128e-05, "loss": 0.149, "step": 33475 }, { "epoch": 1.56, "learning_rate": 2.6959418345217108e-05, "loss": 0.2892, "step": 33480 }, { "epoch": 1.56, "learning_rate": 2.6958950275381088e-05, "loss": 0.1681, "step": 33485 }, { "epoch": 1.56, "learning_rate": 2.6958482205545067e-05, "loss": 0.0971, "step": 33490 }, { "epoch": 1.56, "learning_rate": 2.695801413570905e-05, "loss": 0.0487, "step": 33495 }, { "epoch": 1.56, "learning_rate": 2.695754606587303e-05, "loss": 0.0572, "step": 33500 }, { "epoch": 1.56, "learning_rate": 2.6957077996037007e-05, "loss": 0.1197, "step": 33505 }, { "epoch": 1.56, "learning_rate": 2.695660992620099e-05, "loss": 0.1154, "step": 33510 }, { "epoch": 1.56, "learning_rate": 2.695614185636497e-05, "loss": 0.1573, "step": 33515 }, { "epoch": 1.56, "learning_rate": 2.695567378652895e-05, "loss": 0.1595, "step": 33520 }, { "epoch": 1.56, "learning_rate": 2.695520571669293e-05, "loss": 0.1694, "step": 33525 }, { "epoch": 1.56, "learning_rate": 2.6954737646856913e-05, "loss": 0.4786, "step": 33530 }, { "epoch": 1.56, "learning_rate": 2.6954269577020893e-05, "loss": 0.3407, "step": 33535 }, { "epoch": 1.57, "learning_rate": 2.6953801507184872e-05, "loss": 0.0328, "step": 33540 }, { "epoch": 1.57, "learning_rate": 2.6953333437348852e-05, "loss": 0.0409, "step": 33545 }, { "epoch": 1.57, "learning_rate": 2.6952865367512835e-05, "loss": 0.0669, "step": 33550 }, { "epoch": 1.57, "learning_rate": 2.6952397297676815e-05, "loss": 0.1159, "step": 33555 }, { "epoch": 1.57, "learning_rate": 2.6951929227840795e-05, "loss": 0.0854, "step": 33560 }, { "epoch": 1.57, "learning_rate": 2.6951461158004775e-05, "loss": 0.1544, "step": 33565 }, { "epoch": 1.57, "learning_rate": 2.6950993088168755e-05, "loss": 0.239, "step": 33570 }, { "epoch": 1.57, "learning_rate": 2.6950525018332734e-05, "loss": 0.2662, "step": 33575 }, { "epoch": 1.57, "learning_rate": 2.6950056948496714e-05, "loss": 0.4318, "step": 33580 }, { "epoch": 1.57, "learning_rate": 2.6949588878660698e-05, "loss": 0.2809, "step": 33585 }, { "epoch": 1.57, "learning_rate": 2.6949120808824677e-05, "loss": 0.0523, "step": 33590 }, { "epoch": 1.57, "learning_rate": 2.6948652738988657e-05, "loss": 0.0609, "step": 33595 }, { "epoch": 1.57, "learning_rate": 2.6948184669152637e-05, "loss": 0.1092, "step": 33600 }, { "epoch": 1.57, "learning_rate": 2.694771659931662e-05, "loss": 0.101, "step": 33605 }, { "epoch": 1.57, "learning_rate": 2.69472485294806e-05, "loss": 0.1491, "step": 33610 }, { "epoch": 1.57, "learning_rate": 2.694678045964458e-05, "loss": 0.1125, "step": 33615 }, { "epoch": 1.57, "learning_rate": 2.6946312389808563e-05, "loss": 0.1055, "step": 33620 }, { "epoch": 1.57, "learning_rate": 2.6945844319972543e-05, "loss": 0.2823, "step": 33625 }, { "epoch": 1.57, "learning_rate": 2.694537625013652e-05, "loss": 0.2148, "step": 33630 }, { "epoch": 1.57, "learning_rate": 2.69449081803005e-05, "loss": 0.1385, "step": 33635 }, { "epoch": 1.57, "learning_rate": 2.6944440110464482e-05, "loss": 0.0313, "step": 33640 }, { "epoch": 1.57, "learning_rate": 2.6943972040628462e-05, "loss": 0.1056, "step": 33645 }, { "epoch": 1.57, "learning_rate": 2.6943503970792442e-05, "loss": 0.0715, "step": 33650 }, { "epoch": 1.57, "learning_rate": 2.6943035900956422e-05, "loss": 0.0989, "step": 33655 }, { "epoch": 1.57, "learning_rate": 2.6942567831120405e-05, "loss": 0.1732, "step": 33660 }, { "epoch": 1.57, "learning_rate": 2.6942099761284385e-05, "loss": 0.1354, "step": 33665 }, { "epoch": 1.57, "learning_rate": 2.6941631691448365e-05, "loss": 0.1368, "step": 33670 }, { "epoch": 1.57, "learning_rate": 2.6941163621612344e-05, "loss": 0.236, "step": 33675 }, { "epoch": 1.57, "learning_rate": 2.6940695551776328e-05, "loss": 0.2753, "step": 33680 }, { "epoch": 1.57, "learning_rate": 2.6940227481940307e-05, "loss": 0.2297, "step": 33685 }, { "epoch": 1.57, "learning_rate": 2.6939759412104287e-05, "loss": 0.0475, "step": 33690 }, { "epoch": 1.57, "learning_rate": 2.6939291342268267e-05, "loss": 0.0418, "step": 33695 }, { "epoch": 1.57, "learning_rate": 2.6938823272432247e-05, "loss": 0.0751, "step": 33700 }, { "epoch": 1.57, "learning_rate": 2.6938355202596227e-05, "loss": 0.0975, "step": 33705 }, { "epoch": 1.57, "learning_rate": 2.6937887132760207e-05, "loss": 0.0945, "step": 33710 }, { "epoch": 1.57, "learning_rate": 2.693741906292419e-05, "loss": 0.1498, "step": 33715 }, { "epoch": 1.57, "learning_rate": 2.693695099308817e-05, "loss": 0.1429, "step": 33720 }, { "epoch": 1.57, "learning_rate": 2.693648292325215e-05, "loss": 0.1884, "step": 33725 }, { "epoch": 1.57, "learning_rate": 2.693601485341613e-05, "loss": 0.3491, "step": 33730 }, { "epoch": 1.57, "learning_rate": 2.6935546783580112e-05, "loss": 0.2117, "step": 33735 }, { "epoch": 1.57, "learning_rate": 2.6935078713744092e-05, "loss": 0.0323, "step": 33740 }, { "epoch": 1.57, "learning_rate": 2.6934610643908072e-05, "loss": 0.0391, "step": 33745 }, { "epoch": 1.57, "learning_rate": 2.6934142574072055e-05, "loss": 0.1505, "step": 33750 }, { "epoch": 1.58, "learning_rate": 2.693367450423603e-05, "loss": 0.0932, "step": 33755 }, { "epoch": 1.58, "learning_rate": 2.693320643440001e-05, "loss": 0.1488, "step": 33760 }, { "epoch": 1.58, "learning_rate": 2.693273836456399e-05, "loss": 0.1667, "step": 33765 }, { "epoch": 1.58, "learning_rate": 2.6932270294727974e-05, "loss": 0.2161, "step": 33770 }, { "epoch": 1.58, "learning_rate": 2.6931802224891954e-05, "loss": 0.3211, "step": 33775 }, { "epoch": 1.58, "learning_rate": 2.6931334155055934e-05, "loss": 0.5273, "step": 33780 }, { "epoch": 1.58, "learning_rate": 2.6930866085219914e-05, "loss": 0.228, "step": 33785 }, { "epoch": 1.58, "learning_rate": 2.6930398015383897e-05, "loss": 0.039, "step": 33790 }, { "epoch": 1.58, "learning_rate": 2.6929929945547877e-05, "loss": 0.0235, "step": 33795 }, { "epoch": 1.58, "learning_rate": 2.6929461875711857e-05, "loss": 0.0688, "step": 33800 }, { "epoch": 1.58, "learning_rate": 2.692899380587584e-05, "loss": 0.1279, "step": 33805 }, { "epoch": 1.58, "learning_rate": 2.692852573603982e-05, "loss": 0.1052, "step": 33810 }, { "epoch": 1.58, "learning_rate": 2.69280576662038e-05, "loss": 0.2008, "step": 33815 }, { "epoch": 1.58, "learning_rate": 2.6927589596367776e-05, "loss": 0.1396, "step": 33820 }, { "epoch": 1.58, "learning_rate": 2.692712152653176e-05, "loss": 0.2285, "step": 33825 }, { "epoch": 1.58, "learning_rate": 2.692665345669574e-05, "loss": 0.3607, "step": 33830 }, { "epoch": 1.58, "learning_rate": 2.692618538685972e-05, "loss": 0.2623, "step": 33835 }, { "epoch": 1.58, "learning_rate": 2.69257173170237e-05, "loss": 0.0463, "step": 33840 }, { "epoch": 1.58, "learning_rate": 2.6925249247187682e-05, "loss": 0.1067, "step": 33845 }, { "epoch": 1.58, "learning_rate": 2.6924781177351662e-05, "loss": 0.1125, "step": 33850 }, { "epoch": 1.58, "learning_rate": 2.692431310751564e-05, "loss": 0.1145, "step": 33855 }, { "epoch": 1.58, "learning_rate": 2.6923845037679625e-05, "loss": 0.0639, "step": 33860 }, { "epoch": 1.58, "learning_rate": 2.6923376967843605e-05, "loss": 0.0958, "step": 33865 }, { "epoch": 1.58, "learning_rate": 2.6922908898007584e-05, "loss": 0.1917, "step": 33870 }, { "epoch": 1.58, "learning_rate": 2.6922440828171564e-05, "loss": 0.2605, "step": 33875 }, { "epoch": 1.58, "learning_rate": 2.6921972758335544e-05, "loss": 0.3061, "step": 33880 }, { "epoch": 1.58, "learning_rate": 2.6921504688499524e-05, "loss": 0.2658, "step": 33885 }, { "epoch": 1.58, "learning_rate": 2.6921036618663504e-05, "loss": 0.1025, "step": 33890 }, { "epoch": 1.58, "learning_rate": 2.6920568548827483e-05, "loss": 0.0978, "step": 33895 }, { "epoch": 1.58, "learning_rate": 2.6920100478991467e-05, "loss": 0.0522, "step": 33900 }, { "epoch": 1.58, "learning_rate": 2.6919632409155447e-05, "loss": 0.1059, "step": 33905 }, { "epoch": 1.58, "learning_rate": 2.6919164339319426e-05, "loss": 0.0923, "step": 33910 }, { "epoch": 1.58, "learning_rate": 2.6918696269483406e-05, "loss": 0.136, "step": 33915 }, { "epoch": 1.58, "learning_rate": 2.691822819964739e-05, "loss": 0.1848, "step": 33920 }, { "epoch": 1.58, "learning_rate": 2.691776012981137e-05, "loss": 0.2659, "step": 33925 }, { "epoch": 1.58, "learning_rate": 2.691729205997535e-05, "loss": 0.3446, "step": 33930 }, { "epoch": 1.58, "learning_rate": 2.6916823990139332e-05, "loss": 0.2859, "step": 33935 }, { "epoch": 1.58, "learning_rate": 2.6916355920303312e-05, "loss": 0.0622, "step": 33940 }, { "epoch": 1.58, "learning_rate": 2.691588785046729e-05, "loss": 0.0741, "step": 33945 }, { "epoch": 1.58, "learning_rate": 2.6915419780631268e-05, "loss": 0.0572, "step": 33950 }, { "epoch": 1.58, "learning_rate": 2.691495171079525e-05, "loss": 0.0683, "step": 33955 }, { "epoch": 1.58, "learning_rate": 2.691448364095923e-05, "loss": 0.1142, "step": 33960 }, { "epoch": 1.58, "learning_rate": 2.691401557112321e-05, "loss": 0.0827, "step": 33965 }, { "epoch": 1.59, "learning_rate": 2.691354750128719e-05, "loss": 0.2321, "step": 33970 }, { "epoch": 1.59, "learning_rate": 2.6913079431451174e-05, "loss": 0.2662, "step": 33975 }, { "epoch": 1.59, "learning_rate": 2.6912611361615154e-05, "loss": 0.2821, "step": 33980 }, { "epoch": 1.59, "learning_rate": 2.6912143291779134e-05, "loss": 0.2323, "step": 33985 }, { "epoch": 1.59, "learning_rate": 2.6911675221943117e-05, "loss": 0.0372, "step": 33990 }, { "epoch": 1.59, "learning_rate": 2.6911207152107097e-05, "loss": 0.0707, "step": 33995 }, { "epoch": 1.59, "learning_rate": 2.6910739082271077e-05, "loss": 0.0431, "step": 34000 }, { "epoch": 1.59, "learning_rate": 2.6910271012435056e-05, "loss": 0.0836, "step": 34005 }, { "epoch": 1.59, "learning_rate": 2.6909802942599036e-05, "loss": 0.1059, "step": 34010 }, { "epoch": 1.59, "learning_rate": 2.6909334872763016e-05, "loss": 0.1898, "step": 34015 }, { "epoch": 1.59, "learning_rate": 2.6908866802926996e-05, "loss": 0.2238, "step": 34020 }, { "epoch": 1.59, "learning_rate": 2.6908398733090976e-05, "loss": 0.3044, "step": 34025 }, { "epoch": 1.59, "learning_rate": 2.690793066325496e-05, "loss": 0.2505, "step": 34030 }, { "epoch": 1.59, "learning_rate": 2.690746259341894e-05, "loss": 0.2176, "step": 34035 }, { "epoch": 1.59, "learning_rate": 2.690699452358292e-05, "loss": 0.055, "step": 34040 }, { "epoch": 1.59, "learning_rate": 2.6906526453746902e-05, "loss": 0.0949, "step": 34045 }, { "epoch": 1.59, "learning_rate": 2.690605838391088e-05, "loss": 0.1021, "step": 34050 }, { "epoch": 1.59, "learning_rate": 2.690559031407486e-05, "loss": 0.0625, "step": 34055 }, { "epoch": 1.59, "learning_rate": 2.690512224423884e-05, "loss": 0.132, "step": 34060 }, { "epoch": 1.59, "learning_rate": 2.6904654174402824e-05, "loss": 0.187, "step": 34065 }, { "epoch": 1.59, "learning_rate": 2.69041861045668e-05, "loss": 0.1742, "step": 34070 }, { "epoch": 1.59, "learning_rate": 2.690371803473078e-05, "loss": 0.2368, "step": 34075 }, { "epoch": 1.59, "learning_rate": 2.690324996489476e-05, "loss": 0.4076, "step": 34080 }, { "epoch": 1.59, "learning_rate": 2.6902781895058744e-05, "loss": 0.2802, "step": 34085 }, { "epoch": 1.59, "learning_rate": 2.6902313825222723e-05, "loss": 0.0225, "step": 34090 }, { "epoch": 1.59, "learning_rate": 2.6901845755386703e-05, "loss": 0.0578, "step": 34095 }, { "epoch": 1.59, "learning_rate": 2.6901377685550683e-05, "loss": 0.0631, "step": 34100 }, { "epoch": 1.59, "learning_rate": 2.6900909615714666e-05, "loss": 0.0355, "step": 34105 }, { "epoch": 1.59, "learning_rate": 2.6900441545878646e-05, "loss": 0.1352, "step": 34110 }, { "epoch": 1.59, "learning_rate": 2.6899973476042626e-05, "loss": 0.1168, "step": 34115 }, { "epoch": 1.59, "learning_rate": 2.689950540620661e-05, "loss": 0.2445, "step": 34120 }, { "epoch": 1.59, "learning_rate": 2.689903733637059e-05, "loss": 0.1885, "step": 34125 }, { "epoch": 1.59, "learning_rate": 2.689856926653457e-05, "loss": 0.2987, "step": 34130 }, { "epoch": 1.59, "learning_rate": 2.6898101196698545e-05, "loss": 0.2409, "step": 34135 }, { "epoch": 1.59, "learning_rate": 2.689763312686253e-05, "loss": 0.0281, "step": 34140 }, { "epoch": 1.59, "learning_rate": 2.6897165057026508e-05, "loss": 0.0748, "step": 34145 }, { "epoch": 1.59, "learning_rate": 2.6896696987190488e-05, "loss": 0.0213, "step": 34150 }, { "epoch": 1.59, "learning_rate": 2.6896228917354468e-05, "loss": 0.1206, "step": 34155 }, { "epoch": 1.59, "learning_rate": 2.689576084751845e-05, "loss": 0.1155, "step": 34160 }, { "epoch": 1.59, "learning_rate": 2.689529277768243e-05, "loss": 0.1549, "step": 34165 }, { "epoch": 1.59, "learning_rate": 2.689482470784641e-05, "loss": 0.1066, "step": 34170 }, { "epoch": 1.59, "learning_rate": 2.6894356638010394e-05, "loss": 0.2163, "step": 34175 }, { "epoch": 1.59, "learning_rate": 2.6893888568174374e-05, "loss": 0.3527, "step": 34180 }, { "epoch": 1.6, "learning_rate": 2.6893420498338354e-05, "loss": 0.2397, "step": 34185 }, { "epoch": 1.6, "learning_rate": 2.6892952428502333e-05, "loss": 0.0529, "step": 34190 }, { "epoch": 1.6, "learning_rate": 2.6892484358666317e-05, "loss": 0.0579, "step": 34195 }, { "epoch": 1.6, "learning_rate": 2.6892016288830293e-05, "loss": 0.0869, "step": 34200 }, { "epoch": 1.6, "learning_rate": 2.6891548218994273e-05, "loss": 0.0736, "step": 34205 }, { "epoch": 1.6, "learning_rate": 2.6891080149158253e-05, "loss": 0.118, "step": 34210 }, { "epoch": 1.6, "learning_rate": 2.6890612079322236e-05, "loss": 0.1502, "step": 34215 }, { "epoch": 1.6, "learning_rate": 2.6890144009486216e-05, "loss": 0.222, "step": 34220 }, { "epoch": 1.6, "learning_rate": 2.6889675939650195e-05, "loss": 0.1775, "step": 34225 }, { "epoch": 1.6, "learning_rate": 2.688920786981418e-05, "loss": 0.274, "step": 34230 }, { "epoch": 1.6, "learning_rate": 2.688873979997816e-05, "loss": 0.2308, "step": 34235 }, { "epoch": 1.6, "learning_rate": 2.688827173014214e-05, "loss": 0.0443, "step": 34240 }, { "epoch": 1.6, "learning_rate": 2.6887803660306118e-05, "loss": 0.0793, "step": 34245 }, { "epoch": 1.6, "learning_rate": 2.68873355904701e-05, "loss": 0.0679, "step": 34250 }, { "epoch": 1.6, "learning_rate": 2.688686752063408e-05, "loss": 0.1434, "step": 34255 }, { "epoch": 1.6, "learning_rate": 2.6886399450798058e-05, "loss": 0.0738, "step": 34260 }, { "epoch": 1.6, "learning_rate": 2.6885931380962037e-05, "loss": 0.1883, "step": 34265 }, { "epoch": 1.6, "learning_rate": 2.688546331112602e-05, "loss": 0.2448, "step": 34270 }, { "epoch": 1.6, "learning_rate": 2.688499524129e-05, "loss": 0.3142, "step": 34275 }, { "epoch": 1.6, "learning_rate": 2.688452717145398e-05, "loss": 0.2646, "step": 34280 }, { "epoch": 1.6, "learning_rate": 2.688405910161796e-05, "loss": 0.1814, "step": 34285 }, { "epoch": 1.6, "learning_rate": 2.6883591031781943e-05, "loss": 0.0626, "step": 34290 }, { "epoch": 1.6, "learning_rate": 2.6883122961945923e-05, "loss": 0.0643, "step": 34295 }, { "epoch": 1.6, "learning_rate": 2.6882654892109903e-05, "loss": 0.062, "step": 34300 }, { "epoch": 1.6, "learning_rate": 2.6882186822273886e-05, "loss": 0.0597, "step": 34305 }, { "epoch": 1.6, "learning_rate": 2.6881718752437866e-05, "loss": 0.0672, "step": 34310 }, { "epoch": 1.6, "learning_rate": 2.6881250682601846e-05, "loss": 0.1581, "step": 34315 }, { "epoch": 1.6, "learning_rate": 2.6880782612765826e-05, "loss": 0.1584, "step": 34320 }, { "epoch": 1.6, "learning_rate": 2.6880314542929805e-05, "loss": 0.2152, "step": 34325 }, { "epoch": 1.6, "learning_rate": 2.6879846473093785e-05, "loss": 0.2151, "step": 34330 }, { "epoch": 1.6, "learning_rate": 2.6879378403257765e-05, "loss": 0.2441, "step": 34335 }, { "epoch": 1.6, "learning_rate": 2.6878910333421745e-05, "loss": 0.0613, "step": 34340 }, { "epoch": 1.6, "learning_rate": 2.6878442263585728e-05, "loss": 0.1019, "step": 34345 }, { "epoch": 1.6, "learning_rate": 2.6877974193749708e-05, "loss": 0.0949, "step": 34350 }, { "epoch": 1.6, "learning_rate": 2.6877506123913688e-05, "loss": 0.1059, "step": 34355 }, { "epoch": 1.6, "learning_rate": 2.687703805407767e-05, "loss": 0.1085, "step": 34360 }, { "epoch": 1.6, "learning_rate": 2.687656998424165e-05, "loss": 0.085, "step": 34365 }, { "epoch": 1.6, "learning_rate": 2.687610191440563e-05, "loss": 0.162, "step": 34370 }, { "epoch": 1.6, "learning_rate": 2.687563384456961e-05, "loss": 0.2523, "step": 34375 }, { "epoch": 1.6, "learning_rate": 2.6875165774733594e-05, "loss": 0.3305, "step": 34380 }, { "epoch": 1.6, "learning_rate": 2.6874697704897573e-05, "loss": 0.1923, "step": 34385 }, { "epoch": 1.6, "learning_rate": 2.687422963506155e-05, "loss": 0.0241, "step": 34390 }, { "epoch": 1.6, "learning_rate": 2.687376156522553e-05, "loss": 0.0576, "step": 34395 }, { "epoch": 1.61, "learning_rate": 2.6873293495389513e-05, "loss": 0.0547, "step": 34400 }, { "epoch": 1.61, "learning_rate": 2.6872825425553493e-05, "loss": 0.0452, "step": 34405 }, { "epoch": 1.61, "learning_rate": 2.6872357355717472e-05, "loss": 0.0912, "step": 34410 }, { "epoch": 1.61, "learning_rate": 2.6871889285881456e-05, "loss": 0.0769, "step": 34415 }, { "epoch": 1.61, "learning_rate": 2.6871421216045435e-05, "loss": 0.1924, "step": 34420 }, { "epoch": 1.61, "learning_rate": 2.6870953146209415e-05, "loss": 0.193, "step": 34425 }, { "epoch": 1.61, "learning_rate": 2.6870485076373395e-05, "loss": 0.3018, "step": 34430 }, { "epoch": 1.61, "learning_rate": 2.687001700653738e-05, "loss": 0.237, "step": 34435 }, { "epoch": 1.61, "learning_rate": 2.6869548936701358e-05, "loss": 0.0542, "step": 34440 }, { "epoch": 1.61, "learning_rate": 2.6869080866865338e-05, "loss": 0.0745, "step": 34445 }, { "epoch": 1.61, "learning_rate": 2.6868612797029314e-05, "loss": 0.1468, "step": 34450 }, { "epoch": 1.61, "learning_rate": 2.6868144727193298e-05, "loss": 0.0688, "step": 34455 }, { "epoch": 1.61, "learning_rate": 2.6867676657357277e-05, "loss": 0.1102, "step": 34460 }, { "epoch": 1.61, "learning_rate": 2.6867208587521257e-05, "loss": 0.1132, "step": 34465 }, { "epoch": 1.61, "learning_rate": 2.686674051768524e-05, "loss": 0.124, "step": 34470 }, { "epoch": 1.61, "learning_rate": 2.686627244784922e-05, "loss": 0.2497, "step": 34475 }, { "epoch": 1.61, "learning_rate": 2.68658043780132e-05, "loss": 0.453, "step": 34480 }, { "epoch": 1.61, "learning_rate": 2.686533630817718e-05, "loss": 0.3008, "step": 34485 }, { "epoch": 1.61, "learning_rate": 2.6864868238341163e-05, "loss": 0.0518, "step": 34490 }, { "epoch": 1.61, "learning_rate": 2.6864400168505143e-05, "loss": 0.0615, "step": 34495 }, { "epoch": 1.61, "learning_rate": 2.6863932098669123e-05, "loss": 0.0578, "step": 34500 }, { "epoch": 1.61, "learning_rate": 2.6863464028833103e-05, "loss": 0.0557, "step": 34505 }, { "epoch": 1.61, "learning_rate": 2.6862995958997086e-05, "loss": 0.112, "step": 34510 }, { "epoch": 1.61, "learning_rate": 2.6862527889161062e-05, "loss": 0.2484, "step": 34515 }, { "epoch": 1.61, "learning_rate": 2.6862059819325042e-05, "loss": 0.0795, "step": 34520 }, { "epoch": 1.61, "learning_rate": 2.6861591749489022e-05, "loss": 0.1779, "step": 34525 }, { "epoch": 1.61, "learning_rate": 2.6861123679653005e-05, "loss": 0.2807, "step": 34530 }, { "epoch": 1.61, "learning_rate": 2.6860655609816985e-05, "loss": 0.3128, "step": 34535 }, { "epoch": 1.61, "learning_rate": 2.6860187539980965e-05, "loss": 0.0468, "step": 34540 }, { "epoch": 1.61, "learning_rate": 2.6859719470144948e-05, "loss": 0.0491, "step": 34545 }, { "epoch": 1.61, "learning_rate": 2.6859251400308928e-05, "loss": 0.0459, "step": 34550 }, { "epoch": 1.61, "learning_rate": 2.6858783330472908e-05, "loss": 0.0644, "step": 34555 }, { "epoch": 1.61, "learning_rate": 2.6858315260636887e-05, "loss": 0.1558, "step": 34560 }, { "epoch": 1.61, "learning_rate": 2.685784719080087e-05, "loss": 0.111, "step": 34565 }, { "epoch": 1.61, "learning_rate": 2.685737912096485e-05, "loss": 0.1514, "step": 34570 }, { "epoch": 1.61, "learning_rate": 2.685691105112883e-05, "loss": 0.1856, "step": 34575 }, { "epoch": 1.61, "learning_rate": 2.6856442981292807e-05, "loss": 0.2387, "step": 34580 }, { "epoch": 1.61, "learning_rate": 2.685597491145679e-05, "loss": 0.3527, "step": 34585 }, { "epoch": 1.61, "learning_rate": 2.685550684162077e-05, "loss": 0.076, "step": 34590 }, { "epoch": 1.61, "learning_rate": 2.685503877178475e-05, "loss": 0.0482, "step": 34595 }, { "epoch": 1.61, "learning_rate": 2.6854570701948733e-05, "loss": 0.0494, "step": 34600 }, { "epoch": 1.61, "learning_rate": 2.6854102632112712e-05, "loss": 0.0806, "step": 34605 }, { "epoch": 1.61, "learning_rate": 2.6853634562276692e-05, "loss": 0.1098, "step": 34610 }, { "epoch": 1.62, "learning_rate": 2.6853166492440672e-05, "loss": 0.1137, "step": 34615 }, { "epoch": 1.62, "learning_rate": 2.6852698422604655e-05, "loss": 0.1498, "step": 34620 }, { "epoch": 1.62, "learning_rate": 2.6852230352768635e-05, "loss": 0.2765, "step": 34625 }, { "epoch": 1.62, "learning_rate": 2.6851762282932615e-05, "loss": 0.2931, "step": 34630 }, { "epoch": 1.62, "learning_rate": 2.6851294213096595e-05, "loss": 0.3579, "step": 34635 }, { "epoch": 1.62, "learning_rate": 2.6850826143260575e-05, "loss": 0.0235, "step": 34640 }, { "epoch": 1.62, "learning_rate": 2.6850358073424554e-05, "loss": 0.0594, "step": 34645 }, { "epoch": 1.62, "learning_rate": 2.6849890003588534e-05, "loss": 0.1259, "step": 34650 }, { "epoch": 1.62, "learning_rate": 2.6849421933752517e-05, "loss": 0.0955, "step": 34655 }, { "epoch": 1.62, "learning_rate": 2.6848953863916497e-05, "loss": 0.0988, "step": 34660 }, { "epoch": 1.62, "learning_rate": 2.6848485794080477e-05, "loss": 0.0903, "step": 34665 }, { "epoch": 1.62, "learning_rate": 2.6848017724244457e-05, "loss": 0.1835, "step": 34670 }, { "epoch": 1.62, "learning_rate": 2.684754965440844e-05, "loss": 0.1849, "step": 34675 }, { "epoch": 1.62, "learning_rate": 2.684708158457242e-05, "loss": 0.3924, "step": 34680 }, { "epoch": 1.62, "learning_rate": 2.68466135147364e-05, "loss": 0.2054, "step": 34685 }, { "epoch": 1.62, "learning_rate": 2.684614544490038e-05, "loss": 0.0244, "step": 34690 }, { "epoch": 1.62, "learning_rate": 2.6845677375064363e-05, "loss": 0.0388, "step": 34695 }, { "epoch": 1.62, "learning_rate": 2.6845209305228343e-05, "loss": 0.0666, "step": 34700 }, { "epoch": 1.62, "learning_rate": 2.684474123539232e-05, "loss": 0.0782, "step": 34705 }, { "epoch": 1.62, "learning_rate": 2.68442731655563e-05, "loss": 0.1094, "step": 34710 }, { "epoch": 1.62, "learning_rate": 2.6843805095720282e-05, "loss": 0.1747, "step": 34715 }, { "epoch": 1.62, "learning_rate": 2.6843337025884262e-05, "loss": 0.1708, "step": 34720 }, { "epoch": 1.62, "learning_rate": 2.684286895604824e-05, "loss": 0.2518, "step": 34725 }, { "epoch": 1.62, "learning_rate": 2.6842400886212225e-05, "loss": 0.1943, "step": 34730 }, { "epoch": 1.62, "learning_rate": 2.6841932816376205e-05, "loss": 0.3573, "step": 34735 }, { "epoch": 1.62, "learning_rate": 2.6841464746540184e-05, "loss": 0.0451, "step": 34740 }, { "epoch": 1.62, "learning_rate": 2.6840996676704164e-05, "loss": 0.0434, "step": 34745 }, { "epoch": 1.62, "learning_rate": 2.6840528606868148e-05, "loss": 0.0879, "step": 34750 }, { "epoch": 1.62, "learning_rate": 2.6840060537032127e-05, "loss": 0.0797, "step": 34755 }, { "epoch": 1.62, "learning_rate": 2.6839592467196107e-05, "loss": 0.1144, "step": 34760 }, { "epoch": 1.62, "learning_rate": 2.6839124397360087e-05, "loss": 0.1451, "step": 34765 }, { "epoch": 1.62, "learning_rate": 2.6838656327524067e-05, "loss": 0.2889, "step": 34770 }, { "epoch": 1.62, "learning_rate": 2.6838188257688047e-05, "loss": 0.1306, "step": 34775 }, { "epoch": 1.62, "learning_rate": 2.6837720187852026e-05, "loss": 0.3733, "step": 34780 }, { "epoch": 1.62, "learning_rate": 2.683725211801601e-05, "loss": 0.2779, "step": 34785 }, { "epoch": 1.62, "learning_rate": 2.683678404817999e-05, "loss": 0.0484, "step": 34790 }, { "epoch": 1.62, "learning_rate": 2.683631597834397e-05, "loss": 0.0482, "step": 34795 }, { "epoch": 1.62, "learning_rate": 2.683584790850795e-05, "loss": 0.0655, "step": 34800 }, { "epoch": 1.62, "learning_rate": 2.6835379838671932e-05, "loss": 0.0945, "step": 34805 }, { "epoch": 1.62, "learning_rate": 2.6834911768835912e-05, "loss": 0.1258, "step": 34810 }, { "epoch": 1.62, "learning_rate": 2.6834443698999892e-05, "loss": 0.1134, "step": 34815 }, { "epoch": 1.62, "learning_rate": 2.6833975629163872e-05, "loss": 0.1888, "step": 34820 }, { "epoch": 1.62, "learning_rate": 2.6833507559327855e-05, "loss": 0.2423, "step": 34825 }, { "epoch": 1.63, "learning_rate": 2.683303948949183e-05, "loss": 0.3884, "step": 34830 }, { "epoch": 1.63, "learning_rate": 2.683257141965581e-05, "loss": 0.3771, "step": 34835 }, { "epoch": 1.63, "learning_rate": 2.6832103349819794e-05, "loss": 0.0532, "step": 34840 }, { "epoch": 1.63, "learning_rate": 2.6831635279983774e-05, "loss": 0.0419, "step": 34845 }, { "epoch": 1.63, "learning_rate": 2.6831167210147754e-05, "loss": 0.056, "step": 34850 }, { "epoch": 1.63, "learning_rate": 2.6830699140311734e-05, "loss": 0.0864, "step": 34855 }, { "epoch": 1.63, "learning_rate": 2.6830231070475717e-05, "loss": 0.0579, "step": 34860 }, { "epoch": 1.63, "learning_rate": 2.6829763000639697e-05, "loss": 0.1614, "step": 34865 }, { "epoch": 1.63, "learning_rate": 2.6829294930803677e-05, "loss": 0.1465, "step": 34870 }, { "epoch": 1.63, "learning_rate": 2.6828826860967656e-05, "loss": 0.2053, "step": 34875 }, { "epoch": 1.63, "learning_rate": 2.682835879113164e-05, "loss": 0.3873, "step": 34880 }, { "epoch": 1.63, "learning_rate": 2.682789072129562e-05, "loss": 0.2562, "step": 34885 }, { "epoch": 1.63, "learning_rate": 2.68274226514596e-05, "loss": 0.016, "step": 34890 }, { "epoch": 1.63, "learning_rate": 2.6826954581623576e-05, "loss": 0.0762, "step": 34895 }, { "epoch": 1.63, "learning_rate": 2.682648651178756e-05, "loss": 0.1148, "step": 34900 }, { "epoch": 1.63, "learning_rate": 2.682601844195154e-05, "loss": 0.0941, "step": 34905 }, { "epoch": 1.63, "learning_rate": 2.682555037211552e-05, "loss": 0.1167, "step": 34910 }, { "epoch": 1.63, "learning_rate": 2.6825082302279502e-05, "loss": 0.2013, "step": 34915 }, { "epoch": 1.63, "learning_rate": 2.682461423244348e-05, "loss": 0.1354, "step": 34920 }, { "epoch": 1.63, "learning_rate": 2.682414616260746e-05, "loss": 0.1162, "step": 34925 }, { "epoch": 1.63, "learning_rate": 2.682367809277144e-05, "loss": 0.2893, "step": 34930 }, { "epoch": 1.63, "learning_rate": 2.6823210022935424e-05, "loss": 0.3477, "step": 34935 }, { "epoch": 1.63, "learning_rate": 2.6822741953099404e-05, "loss": 0.0413, "step": 34940 }, { "epoch": 1.63, "learning_rate": 2.6822273883263384e-05, "loss": 0.0796, "step": 34945 }, { "epoch": 1.63, "learning_rate": 2.6821805813427364e-05, "loss": 0.0605, "step": 34950 }, { "epoch": 1.63, "learning_rate": 2.6821337743591344e-05, "loss": 0.134, "step": 34955 }, { "epoch": 1.63, "learning_rate": 2.6820869673755324e-05, "loss": 0.0784, "step": 34960 }, { "epoch": 1.63, "learning_rate": 2.6820401603919303e-05, "loss": 0.1225, "step": 34965 }, { "epoch": 1.63, "learning_rate": 2.6819933534083287e-05, "loss": 0.1522, "step": 34970 }, { "epoch": 1.63, "learning_rate": 2.6819465464247266e-05, "loss": 0.2148, "step": 34975 }, { "epoch": 1.63, "learning_rate": 2.6818997394411246e-05, "loss": 0.3193, "step": 34980 }, { "epoch": 1.63, "learning_rate": 2.6818529324575226e-05, "loss": 0.2492, "step": 34985 }, { "epoch": 1.63, "learning_rate": 2.681806125473921e-05, "loss": 0.0347, "step": 34990 }, { "epoch": 1.63, "learning_rate": 2.681759318490319e-05, "loss": 0.0383, "step": 34995 }, { "epoch": 1.63, "learning_rate": 2.681712511506717e-05, "loss": 0.12, "step": 35000 }, { "epoch": 1.63, "learning_rate": 2.681665704523115e-05, "loss": 0.0772, "step": 35005 }, { "epoch": 1.63, "learning_rate": 2.6816188975395132e-05, "loss": 0.1024, "step": 35010 }, { "epoch": 1.63, "learning_rate": 2.6815720905559112e-05, "loss": 0.146, "step": 35015 }, { "epoch": 1.63, "learning_rate": 2.6815252835723088e-05, "loss": 0.0903, "step": 35020 }, { "epoch": 1.63, "learning_rate": 2.681478476588707e-05, "loss": 0.1709, "step": 35025 }, { "epoch": 1.63, "learning_rate": 2.681431669605105e-05, "loss": 0.4136, "step": 35030 }, { "epoch": 1.63, "learning_rate": 2.681384862621503e-05, "loss": 0.2824, "step": 35035 }, { "epoch": 1.64, "learning_rate": 2.681338055637901e-05, "loss": 0.0363, "step": 35040 }, { "epoch": 1.64, "learning_rate": 2.6812912486542994e-05, "loss": 0.0652, "step": 35045 }, { "epoch": 1.64, "learning_rate": 2.6812444416706974e-05, "loss": 0.0782, "step": 35050 }, { "epoch": 1.64, "learning_rate": 2.6811976346870954e-05, "loss": 0.0762, "step": 35055 }, { "epoch": 1.64, "learning_rate": 2.6811508277034933e-05, "loss": 0.0833, "step": 35060 }, { "epoch": 1.64, "learning_rate": 2.6811040207198917e-05, "loss": 0.206, "step": 35065 }, { "epoch": 1.64, "learning_rate": 2.6810572137362896e-05, "loss": 0.2424, "step": 35070 }, { "epoch": 1.64, "learning_rate": 2.6810104067526876e-05, "loss": 0.2071, "step": 35075 }, { "epoch": 1.64, "learning_rate": 2.6809635997690856e-05, "loss": 0.3054, "step": 35080 }, { "epoch": 1.64, "learning_rate": 2.6809167927854836e-05, "loss": 0.2174, "step": 35085 }, { "epoch": 1.64, "learning_rate": 2.6808699858018816e-05, "loss": 0.067, "step": 35090 }, { "epoch": 1.64, "learning_rate": 2.6808231788182796e-05, "loss": 0.1032, "step": 35095 }, { "epoch": 1.64, "learning_rate": 2.680776371834678e-05, "loss": 0.1334, "step": 35100 }, { "epoch": 1.64, "learning_rate": 2.680729564851076e-05, "loss": 0.0693, "step": 35105 }, { "epoch": 1.64, "learning_rate": 2.680682757867474e-05, "loss": 0.1273, "step": 35110 }, { "epoch": 1.64, "learning_rate": 2.6806359508838718e-05, "loss": 0.1383, "step": 35115 }, { "epoch": 1.64, "learning_rate": 2.68058914390027e-05, "loss": 0.1119, "step": 35120 }, { "epoch": 1.64, "learning_rate": 2.680542336916668e-05, "loss": 0.205, "step": 35125 }, { "epoch": 1.64, "learning_rate": 2.680495529933066e-05, "loss": 0.3333, "step": 35130 }, { "epoch": 1.64, "learning_rate": 2.680448722949464e-05, "loss": 0.2637, "step": 35135 }, { "epoch": 1.64, "learning_rate": 2.6804019159658624e-05, "loss": 0.0679, "step": 35140 }, { "epoch": 1.64, "learning_rate": 2.68035510898226e-05, "loss": 0.0672, "step": 35145 }, { "epoch": 1.64, "learning_rate": 2.680308301998658e-05, "loss": 0.0977, "step": 35150 }, { "epoch": 1.64, "learning_rate": 2.6802614950150564e-05, "loss": 0.0903, "step": 35155 }, { "epoch": 1.64, "learning_rate": 2.6802146880314543e-05, "loss": 0.1231, "step": 35160 }, { "epoch": 1.64, "learning_rate": 2.6801678810478523e-05, "loss": 0.1169, "step": 35165 }, { "epoch": 1.64, "learning_rate": 2.6801210740642503e-05, "loss": 0.1604, "step": 35170 }, { "epoch": 1.64, "learning_rate": 2.6800742670806486e-05, "loss": 0.2335, "step": 35175 }, { "epoch": 1.64, "learning_rate": 2.6800274600970466e-05, "loss": 0.3201, "step": 35180 }, { "epoch": 1.64, "learning_rate": 2.6799806531134446e-05, "loss": 0.2803, "step": 35185 }, { "epoch": 1.64, "learning_rate": 2.6799338461298426e-05, "loss": 0.0642, "step": 35190 }, { "epoch": 1.64, "learning_rate": 2.679887039146241e-05, "loss": 0.0155, "step": 35195 }, { "epoch": 1.64, "learning_rate": 2.679840232162639e-05, "loss": 0.0646, "step": 35200 }, { "epoch": 1.64, "learning_rate": 2.679793425179037e-05, "loss": 0.1013, "step": 35205 }, { "epoch": 1.64, "learning_rate": 2.679746618195435e-05, "loss": 0.1002, "step": 35210 }, { "epoch": 1.64, "learning_rate": 2.6796998112118328e-05, "loss": 0.1269, "step": 35215 }, { "epoch": 1.64, "learning_rate": 2.6796530042282308e-05, "loss": 0.0967, "step": 35220 }, { "epoch": 1.64, "learning_rate": 2.6796061972446288e-05, "loss": 0.2312, "step": 35225 }, { "epoch": 1.64, "learning_rate": 2.679559390261027e-05, "loss": 0.2334, "step": 35230 }, { "epoch": 1.64, "learning_rate": 2.679512583277425e-05, "loss": 0.319, "step": 35235 }, { "epoch": 1.64, "learning_rate": 2.679465776293823e-05, "loss": 0.0588, "step": 35240 }, { "epoch": 1.64, "learning_rate": 2.679418969310221e-05, "loss": 0.0514, "step": 35245 }, { "epoch": 1.64, "learning_rate": 2.6793721623266194e-05, "loss": 0.1554, "step": 35250 }, { "epoch": 1.65, "learning_rate": 2.6793253553430173e-05, "loss": 0.1115, "step": 35255 }, { "epoch": 1.65, "learning_rate": 2.6792785483594153e-05, "loss": 0.1398, "step": 35260 }, { "epoch": 1.65, "learning_rate": 2.6792317413758136e-05, "loss": 0.1965, "step": 35265 }, { "epoch": 1.65, "learning_rate": 2.6791849343922113e-05, "loss": 0.1584, "step": 35270 }, { "epoch": 1.65, "learning_rate": 2.6791381274086093e-05, "loss": 0.2064, "step": 35275 }, { "epoch": 1.65, "learning_rate": 2.6790913204250073e-05, "loss": 0.3626, "step": 35280 }, { "epoch": 1.65, "learning_rate": 2.6790445134414056e-05, "loss": 0.3062, "step": 35285 }, { "epoch": 1.65, "learning_rate": 2.6789977064578036e-05, "loss": 0.0683, "step": 35290 }, { "epoch": 1.65, "learning_rate": 2.6789508994742015e-05, "loss": 0.0324, "step": 35295 }, { "epoch": 1.65, "learning_rate": 2.6789040924905995e-05, "loss": 0.051, "step": 35300 }, { "epoch": 1.65, "learning_rate": 2.678857285506998e-05, "loss": 0.093, "step": 35305 }, { "epoch": 1.65, "learning_rate": 2.6788104785233958e-05, "loss": 0.1299, "step": 35310 }, { "epoch": 1.65, "learning_rate": 2.6787636715397938e-05, "loss": 0.1499, "step": 35315 }, { "epoch": 1.65, "learning_rate": 2.6787168645561918e-05, "loss": 0.2239, "step": 35320 }, { "epoch": 1.65, "learning_rate": 2.67867005757259e-05, "loss": 0.215, "step": 35325 }, { "epoch": 1.65, "learning_rate": 2.678623250588988e-05, "loss": 0.2326, "step": 35330 }, { "epoch": 1.65, "learning_rate": 2.6785764436053857e-05, "loss": 0.1818, "step": 35335 }, { "epoch": 1.65, "learning_rate": 2.678529636621784e-05, "loss": 0.0709, "step": 35340 }, { "epoch": 1.65, "learning_rate": 2.678482829638182e-05, "loss": 0.0507, "step": 35345 }, { "epoch": 1.65, "learning_rate": 2.67843602265458e-05, "loss": 0.0713, "step": 35350 }, { "epoch": 1.65, "learning_rate": 2.678389215670978e-05, "loss": 0.1011, "step": 35355 }, { "epoch": 1.65, "learning_rate": 2.6783424086873763e-05, "loss": 0.0856, "step": 35360 }, { "epoch": 1.65, "learning_rate": 2.6782956017037743e-05, "loss": 0.1326, "step": 35365 }, { "epoch": 1.65, "learning_rate": 2.6782487947201723e-05, "loss": 0.139, "step": 35370 }, { "epoch": 1.65, "learning_rate": 2.6782019877365703e-05, "loss": 0.2801, "step": 35375 }, { "epoch": 1.65, "learning_rate": 2.6781551807529686e-05, "loss": 0.3569, "step": 35380 }, { "epoch": 1.65, "learning_rate": 2.6781083737693666e-05, "loss": 0.2894, "step": 35385 }, { "epoch": 1.65, "learning_rate": 2.6780615667857645e-05, "loss": 0.0866, "step": 35390 }, { "epoch": 1.65, "learning_rate": 2.678014759802163e-05, "loss": 0.0207, "step": 35395 }, { "epoch": 1.65, "learning_rate": 2.6779679528185605e-05, "loss": 0.0807, "step": 35400 }, { "epoch": 1.65, "learning_rate": 2.6779211458349585e-05, "loss": 0.0606, "step": 35405 }, { "epoch": 1.65, "learning_rate": 2.6778743388513565e-05, "loss": 0.1061, "step": 35410 }, { "epoch": 1.65, "learning_rate": 2.6778275318677548e-05, "loss": 0.1483, "step": 35415 }, { "epoch": 1.65, "learning_rate": 2.6777807248841528e-05, "loss": 0.0769, "step": 35420 }, { "epoch": 1.65, "learning_rate": 2.6777339179005508e-05, "loss": 0.1539, "step": 35425 }, { "epoch": 1.65, "learning_rate": 2.6776871109169487e-05, "loss": 0.3495, "step": 35430 }, { "epoch": 1.65, "learning_rate": 2.677640303933347e-05, "loss": 0.2758, "step": 35435 }, { "epoch": 1.65, "learning_rate": 2.677593496949745e-05, "loss": 0.0222, "step": 35440 }, { "epoch": 1.65, "learning_rate": 2.677546689966143e-05, "loss": 0.0391, "step": 35445 }, { "epoch": 1.65, "learning_rate": 2.6774998829825413e-05, "loss": 0.0527, "step": 35450 }, { "epoch": 1.65, "learning_rate": 2.6774530759989393e-05, "loss": 0.1213, "step": 35455 }, { "epoch": 1.65, "learning_rate": 2.677406269015337e-05, "loss": 0.1702, "step": 35460 }, { "epoch": 1.65, "learning_rate": 2.677359462031735e-05, "loss": 0.0992, "step": 35465 }, { "epoch": 1.66, "learning_rate": 2.6773126550481333e-05, "loss": 0.113, "step": 35470 }, { "epoch": 1.66, "learning_rate": 2.6772658480645313e-05, "loss": 0.2729, "step": 35475 }, { "epoch": 1.66, "learning_rate": 2.6772190410809292e-05, "loss": 0.3645, "step": 35480 }, { "epoch": 1.66, "learning_rate": 2.6771722340973272e-05, "loss": 0.1929, "step": 35485 }, { "epoch": 1.66, "learning_rate": 2.6771254271137255e-05, "loss": 0.0392, "step": 35490 }, { "epoch": 1.66, "learning_rate": 2.6770786201301235e-05, "loss": 0.0357, "step": 35495 }, { "epoch": 1.66, "learning_rate": 2.6770318131465215e-05, "loss": 0.1291, "step": 35500 }, { "epoch": 1.66, "learning_rate": 2.6769850061629195e-05, "loss": 0.0994, "step": 35505 }, { "epoch": 1.66, "learning_rate": 2.6769381991793178e-05, "loss": 0.1167, "step": 35510 }, { "epoch": 1.66, "learning_rate": 2.6768913921957158e-05, "loss": 0.1407, "step": 35515 }, { "epoch": 1.66, "learning_rate": 2.6768445852121138e-05, "loss": 0.1327, "step": 35520 }, { "epoch": 1.66, "learning_rate": 2.6767977782285117e-05, "loss": 0.1641, "step": 35525 }, { "epoch": 1.66, "learning_rate": 2.6767509712449097e-05, "loss": 0.3573, "step": 35530 }, { "epoch": 1.66, "learning_rate": 2.6767041642613077e-05, "loss": 0.2123, "step": 35535 }, { "epoch": 1.66, "learning_rate": 2.6766573572777057e-05, "loss": 0.0258, "step": 35540 }, { "epoch": 1.66, "learning_rate": 2.676610550294104e-05, "loss": 0.0469, "step": 35545 }, { "epoch": 1.66, "learning_rate": 2.676563743310502e-05, "loss": 0.103, "step": 35550 }, { "epoch": 1.66, "learning_rate": 2.6765169363269e-05, "loss": 0.0814, "step": 35555 }, { "epoch": 1.66, "learning_rate": 2.676470129343298e-05, "loss": 0.0724, "step": 35560 }, { "epoch": 1.66, "learning_rate": 2.6764233223596963e-05, "loss": 0.1174, "step": 35565 }, { "epoch": 1.66, "learning_rate": 2.6763765153760943e-05, "loss": 0.1216, "step": 35570 }, { "epoch": 1.66, "learning_rate": 2.6763297083924922e-05, "loss": 0.2656, "step": 35575 }, { "epoch": 1.66, "learning_rate": 2.6762829014088906e-05, "loss": 0.3507, "step": 35580 }, { "epoch": 1.66, "learning_rate": 2.6762360944252885e-05, "loss": 0.2021, "step": 35585 }, { "epoch": 1.66, "learning_rate": 2.6761892874416862e-05, "loss": 0.0546, "step": 35590 }, { "epoch": 1.66, "learning_rate": 2.6761424804580842e-05, "loss": 0.0881, "step": 35595 }, { "epoch": 1.66, "learning_rate": 2.6760956734744825e-05, "loss": 0.0544, "step": 35600 }, { "epoch": 1.66, "learning_rate": 2.6760488664908805e-05, "loss": 0.1474, "step": 35605 }, { "epoch": 1.66, "learning_rate": 2.6760020595072785e-05, "loss": 0.1081, "step": 35610 }, { "epoch": 1.66, "learning_rate": 2.6759552525236764e-05, "loss": 0.1159, "step": 35615 }, { "epoch": 1.66, "learning_rate": 2.6759084455400748e-05, "loss": 0.0846, "step": 35620 }, { "epoch": 1.66, "learning_rate": 2.6758616385564727e-05, "loss": 0.2357, "step": 35625 }, { "epoch": 1.66, "learning_rate": 2.6758148315728707e-05, "loss": 0.3101, "step": 35630 }, { "epoch": 1.66, "learning_rate": 2.675768024589269e-05, "loss": 0.2465, "step": 35635 }, { "epoch": 1.66, "learning_rate": 2.675721217605667e-05, "loss": 0.0758, "step": 35640 }, { "epoch": 1.66, "learning_rate": 2.675674410622065e-05, "loss": 0.0579, "step": 35645 }, { "epoch": 1.66, "learning_rate": 2.6756276036384626e-05, "loss": 0.0876, "step": 35650 }, { "epoch": 1.66, "learning_rate": 2.675580796654861e-05, "loss": 0.082, "step": 35655 }, { "epoch": 1.66, "learning_rate": 2.675533989671259e-05, "loss": 0.1288, "step": 35660 }, { "epoch": 1.66, "learning_rate": 2.675487182687657e-05, "loss": 0.1343, "step": 35665 }, { "epoch": 1.66, "learning_rate": 2.675440375704055e-05, "loss": 0.1094, "step": 35670 }, { "epoch": 1.66, "learning_rate": 2.6753935687204532e-05, "loss": 0.1585, "step": 35675 }, { "epoch": 1.66, "learning_rate": 2.6753467617368512e-05, "loss": 0.3234, "step": 35680 }, { "epoch": 1.67, "learning_rate": 2.6752999547532492e-05, "loss": 0.2626, "step": 35685 }, { "epoch": 1.67, "learning_rate": 2.6752531477696472e-05, "loss": 0.0925, "step": 35690 }, { "epoch": 1.67, "learning_rate": 2.6752063407860455e-05, "loss": 0.0814, "step": 35695 }, { "epoch": 1.67, "learning_rate": 2.6751595338024435e-05, "loss": 0.0954, "step": 35700 }, { "epoch": 1.67, "learning_rate": 2.6751127268188415e-05, "loss": 0.0994, "step": 35705 }, { "epoch": 1.67, "learning_rate": 2.6750659198352398e-05, "loss": 0.0767, "step": 35710 }, { "epoch": 1.67, "learning_rate": 2.6750191128516374e-05, "loss": 0.138, "step": 35715 }, { "epoch": 1.67, "learning_rate": 2.6749723058680354e-05, "loss": 0.1812, "step": 35720 }, { "epoch": 1.67, "learning_rate": 2.6749254988844334e-05, "loss": 0.2075, "step": 35725 }, { "epoch": 1.67, "learning_rate": 2.6748786919008317e-05, "loss": 0.2661, "step": 35730 }, { "epoch": 1.67, "learning_rate": 2.6748318849172297e-05, "loss": 0.2096, "step": 35735 }, { "epoch": 1.67, "learning_rate": 2.6747850779336277e-05, "loss": 0.046, "step": 35740 }, { "epoch": 1.67, "learning_rate": 2.6747382709500257e-05, "loss": 0.091, "step": 35745 }, { "epoch": 1.67, "learning_rate": 2.674691463966424e-05, "loss": 0.0816, "step": 35750 }, { "epoch": 1.67, "learning_rate": 2.674644656982822e-05, "loss": 0.0655, "step": 35755 }, { "epoch": 1.67, "learning_rate": 2.67459784999922e-05, "loss": 0.1439, "step": 35760 }, { "epoch": 1.67, "learning_rate": 2.6745510430156183e-05, "loss": 0.1144, "step": 35765 }, { "epoch": 1.67, "learning_rate": 2.6745042360320162e-05, "loss": 0.1321, "step": 35770 }, { "epoch": 1.67, "learning_rate": 2.6744574290484142e-05, "loss": 0.2693, "step": 35775 }, { "epoch": 1.67, "learning_rate": 2.674410622064812e-05, "loss": 0.2617, "step": 35780 }, { "epoch": 1.67, "learning_rate": 2.6743638150812102e-05, "loss": 0.2595, "step": 35785 }, { "epoch": 1.67, "learning_rate": 2.6743170080976082e-05, "loss": 0.0289, "step": 35790 }, { "epoch": 1.67, "learning_rate": 2.674270201114006e-05, "loss": 0.0667, "step": 35795 }, { "epoch": 1.67, "learning_rate": 2.674223394130404e-05, "loss": 0.1146, "step": 35800 }, { "epoch": 1.67, "learning_rate": 2.6741765871468025e-05, "loss": 0.1248, "step": 35805 }, { "epoch": 1.67, "learning_rate": 2.6741297801632004e-05, "loss": 0.0679, "step": 35810 }, { "epoch": 1.67, "learning_rate": 2.6740829731795984e-05, "loss": 0.1903, "step": 35815 }, { "epoch": 1.67, "learning_rate": 2.6740361661959967e-05, "loss": 0.2129, "step": 35820 }, { "epoch": 1.67, "learning_rate": 2.6739893592123947e-05, "loss": 0.1297, "step": 35825 }, { "epoch": 1.67, "learning_rate": 2.6739425522287927e-05, "loss": 0.2295, "step": 35830 }, { "epoch": 1.67, "learning_rate": 2.6738957452451907e-05, "loss": 0.2495, "step": 35835 }, { "epoch": 1.67, "learning_rate": 2.6738489382615887e-05, "loss": 0.0683, "step": 35840 }, { "epoch": 1.67, "learning_rate": 2.6738021312779866e-05, "loss": 0.0677, "step": 35845 }, { "epoch": 1.67, "learning_rate": 2.6737553242943846e-05, "loss": 0.0445, "step": 35850 }, { "epoch": 1.67, "learning_rate": 2.6737085173107826e-05, "loss": 0.0884, "step": 35855 }, { "epoch": 1.67, "learning_rate": 2.673661710327181e-05, "loss": 0.1165, "step": 35860 }, { "epoch": 1.67, "learning_rate": 2.673614903343579e-05, "loss": 0.2289, "step": 35865 }, { "epoch": 1.67, "learning_rate": 2.673568096359977e-05, "loss": 0.148, "step": 35870 }, { "epoch": 1.67, "learning_rate": 2.6735212893763752e-05, "loss": 0.3205, "step": 35875 }, { "epoch": 1.67, "learning_rate": 2.6734744823927732e-05, "loss": 0.3924, "step": 35880 }, { "epoch": 1.67, "learning_rate": 2.6734276754091712e-05, "loss": 0.2781, "step": 35885 }, { "epoch": 1.67, "learning_rate": 2.673380868425569e-05, "loss": 0.0914, "step": 35890 }, { "epoch": 1.67, "learning_rate": 2.6733340614419675e-05, "loss": 0.1206, "step": 35895 }, { "epoch": 1.68, "learning_rate": 2.6732872544583655e-05, "loss": 0.107, "step": 35900 }, { "epoch": 1.68, "learning_rate": 2.673240447474763e-05, "loss": 0.196, "step": 35905 }, { "epoch": 1.68, "learning_rate": 2.673193640491161e-05, "loss": 0.0616, "step": 35910 }, { "epoch": 1.68, "learning_rate": 2.6731468335075594e-05, "loss": 0.1961, "step": 35915 }, { "epoch": 1.68, "learning_rate": 2.6731000265239574e-05, "loss": 0.2083, "step": 35920 }, { "epoch": 1.68, "learning_rate": 2.6730532195403554e-05, "loss": 0.2404, "step": 35925 }, { "epoch": 1.68, "learning_rate": 2.6730064125567534e-05, "loss": 0.4073, "step": 35930 }, { "epoch": 1.68, "learning_rate": 2.6729596055731517e-05, "loss": 0.2989, "step": 35935 }, { "epoch": 1.68, "learning_rate": 2.6729127985895497e-05, "loss": 0.0604, "step": 35940 }, { "epoch": 1.68, "learning_rate": 2.6728659916059476e-05, "loss": 0.1098, "step": 35945 }, { "epoch": 1.68, "learning_rate": 2.672819184622346e-05, "loss": 0.1052, "step": 35950 }, { "epoch": 1.68, "learning_rate": 2.672772377638744e-05, "loss": 0.0487, "step": 35955 }, { "epoch": 1.68, "learning_rate": 2.672725570655142e-05, "loss": 0.0529, "step": 35960 }, { "epoch": 1.68, "learning_rate": 2.67267876367154e-05, "loss": 0.1622, "step": 35965 }, { "epoch": 1.68, "learning_rate": 2.672631956687938e-05, "loss": 0.1384, "step": 35970 }, { "epoch": 1.68, "learning_rate": 2.672585149704336e-05, "loss": 0.2306, "step": 35975 }, { "epoch": 1.68, "learning_rate": 2.672538342720734e-05, "loss": 0.2583, "step": 35980 }, { "epoch": 1.68, "learning_rate": 2.672491535737132e-05, "loss": 0.205, "step": 35985 }, { "epoch": 1.68, "learning_rate": 2.67244472875353e-05, "loss": 0.0341, "step": 35990 }, { "epoch": 1.68, "learning_rate": 2.672397921769928e-05, "loss": 0.0919, "step": 35995 }, { "epoch": 1.68, "learning_rate": 2.672351114786326e-05, "loss": 0.076, "step": 36000 }, { "epoch": 1.68, "learning_rate": 2.6723043078027244e-05, "loss": 0.0696, "step": 36005 }, { "epoch": 1.68, "learning_rate": 2.6722575008191224e-05, "loss": 0.0602, "step": 36010 }, { "epoch": 1.68, "learning_rate": 2.6722106938355204e-05, "loss": 0.1402, "step": 36015 }, { "epoch": 1.68, "learning_rate": 2.6721638868519184e-05, "loss": 0.1106, "step": 36020 }, { "epoch": 1.68, "learning_rate": 2.6721170798683167e-05, "loss": 0.3324, "step": 36025 }, { "epoch": 1.68, "learning_rate": 2.6720702728847143e-05, "loss": 0.2599, "step": 36030 }, { "epoch": 1.68, "learning_rate": 2.6720234659011123e-05, "loss": 0.3429, "step": 36035 }, { "epoch": 1.68, "learning_rate": 2.6719766589175103e-05, "loss": 0.0509, "step": 36040 }, { "epoch": 1.68, "learning_rate": 2.6719298519339086e-05, "loss": 0.0872, "step": 36045 }, { "epoch": 1.68, "learning_rate": 2.6718830449503066e-05, "loss": 0.063, "step": 36050 }, { "epoch": 1.68, "learning_rate": 2.6718362379667046e-05, "loss": 0.0953, "step": 36055 }, { "epoch": 1.68, "learning_rate": 2.671789430983103e-05, "loss": 0.1015, "step": 36060 }, { "epoch": 1.68, "learning_rate": 2.671742623999501e-05, "loss": 0.0843, "step": 36065 }, { "epoch": 1.68, "learning_rate": 2.671695817015899e-05, "loss": 0.1827, "step": 36070 }, { "epoch": 1.68, "learning_rate": 2.671649010032297e-05, "loss": 0.1748, "step": 36075 }, { "epoch": 1.68, "learning_rate": 2.6716022030486952e-05, "loss": 0.3294, "step": 36080 }, { "epoch": 1.68, "learning_rate": 2.671555396065093e-05, "loss": 0.4892, "step": 36085 }, { "epoch": 1.68, "learning_rate": 2.671508589081491e-05, "loss": 0.0427, "step": 36090 }, { "epoch": 1.68, "learning_rate": 2.6714617820978888e-05, "loss": 0.1603, "step": 36095 }, { "epoch": 1.68, "learning_rate": 2.671414975114287e-05, "loss": 0.0572, "step": 36100 }, { "epoch": 1.68, "learning_rate": 2.671368168130685e-05, "loss": 0.1353, "step": 36105 }, { "epoch": 1.68, "learning_rate": 2.671321361147083e-05, "loss": 0.0941, "step": 36110 }, { "epoch": 1.69, "learning_rate": 2.671274554163481e-05, "loss": 0.1012, "step": 36115 }, { "epoch": 1.69, "learning_rate": 2.6712277471798794e-05, "loss": 0.2129, "step": 36120 }, { "epoch": 1.69, "learning_rate": 2.6711809401962774e-05, "loss": 0.1594, "step": 36125 }, { "epoch": 1.69, "learning_rate": 2.6711341332126753e-05, "loss": 0.2319, "step": 36130 }, { "epoch": 1.69, "learning_rate": 2.6710873262290737e-05, "loss": 0.2331, "step": 36135 }, { "epoch": 1.69, "learning_rate": 2.6710405192454716e-05, "loss": 0.0648, "step": 36140 }, { "epoch": 1.69, "learning_rate": 2.6709937122618696e-05, "loss": 0.06, "step": 36145 }, { "epoch": 1.69, "learning_rate": 2.6709469052782676e-05, "loss": 0.1357, "step": 36150 }, { "epoch": 1.69, "learning_rate": 2.6709000982946656e-05, "loss": 0.1054, "step": 36155 }, { "epoch": 1.69, "learning_rate": 2.6708532913110636e-05, "loss": 0.0946, "step": 36160 }, { "epoch": 1.69, "learning_rate": 2.6708064843274615e-05, "loss": 0.1108, "step": 36165 }, { "epoch": 1.69, "learning_rate": 2.6707596773438595e-05, "loss": 0.1554, "step": 36170 }, { "epoch": 1.69, "learning_rate": 2.670712870360258e-05, "loss": 0.2176, "step": 36175 }, { "epoch": 1.69, "learning_rate": 2.670666063376656e-05, "loss": 0.3576, "step": 36180 }, { "epoch": 1.69, "learning_rate": 2.6706192563930538e-05, "loss": 0.224, "step": 36185 }, { "epoch": 1.69, "learning_rate": 2.670572449409452e-05, "loss": 0.0446, "step": 36190 }, { "epoch": 1.69, "learning_rate": 2.67052564242585e-05, "loss": 0.0795, "step": 36195 }, { "epoch": 1.69, "learning_rate": 2.670478835442248e-05, "loss": 0.138, "step": 36200 }, { "epoch": 1.69, "learning_rate": 2.670432028458646e-05, "loss": 0.0851, "step": 36205 }, { "epoch": 1.69, "learning_rate": 2.6703852214750444e-05, "loss": 0.1084, "step": 36210 }, { "epoch": 1.69, "learning_rate": 2.6703384144914424e-05, "loss": 0.0979, "step": 36215 }, { "epoch": 1.69, "learning_rate": 2.67029160750784e-05, "loss": 0.2009, "step": 36220 }, { "epoch": 1.69, "learning_rate": 2.670244800524238e-05, "loss": 0.1659, "step": 36225 }, { "epoch": 1.69, "learning_rate": 2.6701979935406363e-05, "loss": 0.3288, "step": 36230 }, { "epoch": 1.69, "learning_rate": 2.6701511865570343e-05, "loss": 0.2036, "step": 36235 }, { "epoch": 1.69, "learning_rate": 2.6701043795734323e-05, "loss": 0.0384, "step": 36240 }, { "epoch": 1.69, "learning_rate": 2.6700575725898306e-05, "loss": 0.0857, "step": 36245 }, { "epoch": 1.69, "learning_rate": 2.6700107656062286e-05, "loss": 0.1182, "step": 36250 }, { "epoch": 1.69, "learning_rate": 2.6699639586226266e-05, "loss": 0.0459, "step": 36255 }, { "epoch": 1.69, "learning_rate": 2.6699171516390246e-05, "loss": 0.0812, "step": 36260 }, { "epoch": 1.69, "learning_rate": 2.669870344655423e-05, "loss": 0.1471, "step": 36265 }, { "epoch": 1.69, "learning_rate": 2.669823537671821e-05, "loss": 0.1491, "step": 36270 }, { "epoch": 1.69, "learning_rate": 2.669776730688219e-05, "loss": 0.1411, "step": 36275 }, { "epoch": 1.69, "learning_rate": 2.6697299237046168e-05, "loss": 0.2339, "step": 36280 }, { "epoch": 1.69, "learning_rate": 2.6696831167210148e-05, "loss": 0.2651, "step": 36285 }, { "epoch": 1.69, "learning_rate": 2.6696363097374128e-05, "loss": 0.0353, "step": 36290 }, { "epoch": 1.69, "learning_rate": 2.6695895027538108e-05, "loss": 0.0558, "step": 36295 }, { "epoch": 1.69, "learning_rate": 2.6695426957702087e-05, "loss": 0.063, "step": 36300 }, { "epoch": 1.69, "learning_rate": 2.669495888786607e-05, "loss": 0.1057, "step": 36305 }, { "epoch": 1.69, "learning_rate": 2.669449081803005e-05, "loss": 0.0718, "step": 36310 }, { "epoch": 1.69, "learning_rate": 2.669402274819403e-05, "loss": 0.1395, "step": 36315 }, { "epoch": 1.69, "learning_rate": 2.6693554678358014e-05, "loss": 0.2643, "step": 36320 }, { "epoch": 1.69, "learning_rate": 2.6693086608521993e-05, "loss": 0.2175, "step": 36325 }, { "epoch": 1.7, "learning_rate": 2.6692618538685973e-05, "loss": 0.3338, "step": 36330 }, { "epoch": 1.7, "learning_rate": 2.6692150468849953e-05, "loss": 0.1787, "step": 36335 }, { "epoch": 1.7, "learning_rate": 2.6691682399013936e-05, "loss": 0.0506, "step": 36340 }, { "epoch": 1.7, "learning_rate": 2.6691214329177913e-05, "loss": 0.0777, "step": 36345 }, { "epoch": 1.7, "learning_rate": 2.6690746259341892e-05, "loss": 0.0732, "step": 36350 }, { "epoch": 1.7, "learning_rate": 2.6690278189505872e-05, "loss": 0.0563, "step": 36355 }, { "epoch": 1.7, "learning_rate": 2.6689810119669855e-05, "loss": 0.0894, "step": 36360 }, { "epoch": 1.7, "learning_rate": 2.6689342049833835e-05, "loss": 0.1885, "step": 36365 }, { "epoch": 1.7, "learning_rate": 2.6688873979997815e-05, "loss": 0.1141, "step": 36370 }, { "epoch": 1.7, "learning_rate": 2.66884059101618e-05, "loss": 0.1828, "step": 36375 }, { "epoch": 1.7, "learning_rate": 2.6687937840325778e-05, "loss": 0.3184, "step": 36380 }, { "epoch": 1.7, "learning_rate": 2.6687469770489758e-05, "loss": 0.4257, "step": 36385 }, { "epoch": 1.7, "learning_rate": 2.6687001700653738e-05, "loss": 0.0557, "step": 36390 }, { "epoch": 1.7, "learning_rate": 2.668653363081772e-05, "loss": 0.1363, "step": 36395 }, { "epoch": 1.7, "learning_rate": 2.66860655609817e-05, "loss": 0.0601, "step": 36400 }, { "epoch": 1.7, "learning_rate": 2.668559749114568e-05, "loss": 0.0535, "step": 36405 }, { "epoch": 1.7, "learning_rate": 2.6685129421309657e-05, "loss": 0.1276, "step": 36410 }, { "epoch": 1.7, "learning_rate": 2.668466135147364e-05, "loss": 0.0647, "step": 36415 }, { "epoch": 1.7, "learning_rate": 2.668419328163762e-05, "loss": 0.1669, "step": 36420 }, { "epoch": 1.7, "learning_rate": 2.66837252118016e-05, "loss": 0.2521, "step": 36425 }, { "epoch": 1.7, "learning_rate": 2.6683257141965583e-05, "loss": 0.2862, "step": 36430 }, { "epoch": 1.7, "learning_rate": 2.6682789072129563e-05, "loss": 0.2764, "step": 36435 }, { "epoch": 1.7, "learning_rate": 2.6682321002293543e-05, "loss": 0.0828, "step": 36440 }, { "epoch": 1.7, "learning_rate": 2.6681852932457523e-05, "loss": 0.0915, "step": 36445 }, { "epoch": 1.7, "learning_rate": 2.6681384862621506e-05, "loss": 0.0906, "step": 36450 }, { "epoch": 1.7, "learning_rate": 2.6680916792785486e-05, "loss": 0.0871, "step": 36455 }, { "epoch": 1.7, "learning_rate": 2.6680448722949465e-05, "loss": 0.1012, "step": 36460 }, { "epoch": 1.7, "learning_rate": 2.6679980653113445e-05, "loss": 0.1285, "step": 36465 }, { "epoch": 1.7, "learning_rate": 2.667951258327743e-05, "loss": 0.1791, "step": 36470 }, { "epoch": 1.7, "learning_rate": 2.6679044513441405e-05, "loss": 0.178, "step": 36475 }, { "epoch": 1.7, "learning_rate": 2.6678576443605385e-05, "loss": 0.4208, "step": 36480 }, { "epoch": 1.7, "learning_rate": 2.6678108373769364e-05, "loss": 0.2376, "step": 36485 }, { "epoch": 1.7, "learning_rate": 2.6677640303933348e-05, "loss": 0.0551, "step": 36490 }, { "epoch": 1.7, "learning_rate": 2.6677172234097327e-05, "loss": 0.04, "step": 36495 }, { "epoch": 1.7, "learning_rate": 2.6676704164261307e-05, "loss": 0.0675, "step": 36500 }, { "epoch": 1.7, "learning_rate": 2.667623609442529e-05, "loss": 0.0488, "step": 36505 }, { "epoch": 1.7, "learning_rate": 2.667576802458927e-05, "loss": 0.094, "step": 36510 }, { "epoch": 1.7, "learning_rate": 2.667529995475325e-05, "loss": 0.1092, "step": 36515 }, { "epoch": 1.7, "learning_rate": 2.667483188491723e-05, "loss": 0.1357, "step": 36520 }, { "epoch": 1.7, "learning_rate": 2.6674363815081213e-05, "loss": 0.0793, "step": 36525 }, { "epoch": 1.7, "learning_rate": 2.6673895745245193e-05, "loss": 0.3235, "step": 36530 }, { "epoch": 1.7, "learning_rate": 2.667342767540917e-05, "loss": 0.2925, "step": 36535 }, { "epoch": 1.71, "learning_rate": 2.667295960557315e-05, "loss": 0.1054, "step": 36540 }, { "epoch": 1.71, "learning_rate": 2.6672491535737132e-05, "loss": 0.0482, "step": 36545 }, { "epoch": 1.71, "learning_rate": 2.6672023465901112e-05, "loss": 0.0982, "step": 36550 }, { "epoch": 1.71, "learning_rate": 2.6671555396065092e-05, "loss": 0.1188, "step": 36555 }, { "epoch": 1.71, "learning_rate": 2.6671087326229075e-05, "loss": 0.0915, "step": 36560 }, { "epoch": 1.71, "learning_rate": 2.6670619256393055e-05, "loss": 0.1817, "step": 36565 }, { "epoch": 1.71, "learning_rate": 2.6670151186557035e-05, "loss": 0.2731, "step": 36570 }, { "epoch": 1.71, "learning_rate": 2.6669683116721015e-05, "loss": 0.1788, "step": 36575 }, { "epoch": 1.71, "learning_rate": 2.6669215046884998e-05, "loss": 0.2102, "step": 36580 }, { "epoch": 1.71, "learning_rate": 2.6668746977048978e-05, "loss": 0.2778, "step": 36585 }, { "epoch": 1.71, "learning_rate": 2.6668278907212958e-05, "loss": 0.0297, "step": 36590 }, { "epoch": 1.71, "learning_rate": 2.6667810837376937e-05, "loss": 0.0796, "step": 36595 }, { "epoch": 1.71, "learning_rate": 2.6667342767540917e-05, "loss": 0.0984, "step": 36600 }, { "epoch": 1.71, "learning_rate": 2.6666874697704897e-05, "loss": 0.0883, "step": 36605 }, { "epoch": 1.71, "learning_rate": 2.6666406627868877e-05, "loss": 0.1426, "step": 36610 }, { "epoch": 1.71, "learning_rate": 2.666593855803286e-05, "loss": 0.1302, "step": 36615 }, { "epoch": 1.71, "learning_rate": 2.666547048819684e-05, "loss": 0.1131, "step": 36620 }, { "epoch": 1.71, "learning_rate": 2.666500241836082e-05, "loss": 0.1583, "step": 36625 }, { "epoch": 1.71, "learning_rate": 2.66645343485248e-05, "loss": 0.2502, "step": 36630 }, { "epoch": 1.71, "learning_rate": 2.6664066278688783e-05, "loss": 0.1735, "step": 36635 }, { "epoch": 1.71, "learning_rate": 2.6663598208852763e-05, "loss": 0.0247, "step": 36640 }, { "epoch": 1.71, "learning_rate": 2.6663130139016742e-05, "loss": 0.0666, "step": 36645 }, { "epoch": 1.71, "learning_rate": 2.6662662069180722e-05, "loss": 0.0861, "step": 36650 }, { "epoch": 1.71, "learning_rate": 2.6662193999344705e-05, "loss": 0.1031, "step": 36655 }, { "epoch": 1.71, "learning_rate": 2.6661725929508682e-05, "loss": 0.2127, "step": 36660 }, { "epoch": 1.71, "learning_rate": 2.666125785967266e-05, "loss": 0.1631, "step": 36665 }, { "epoch": 1.71, "learning_rate": 2.6660789789836645e-05, "loss": 0.1586, "step": 36670 }, { "epoch": 1.71, "learning_rate": 2.6660321720000625e-05, "loss": 0.1788, "step": 36675 }, { "epoch": 1.71, "learning_rate": 2.6659853650164604e-05, "loss": 0.3141, "step": 36680 }, { "epoch": 1.71, "learning_rate": 2.6659385580328584e-05, "loss": 0.174, "step": 36685 }, { "epoch": 1.71, "learning_rate": 2.6658917510492567e-05, "loss": 0.054, "step": 36690 }, { "epoch": 1.71, "learning_rate": 2.6658449440656547e-05, "loss": 0.0253, "step": 36695 }, { "epoch": 1.71, "learning_rate": 2.6657981370820527e-05, "loss": 0.1346, "step": 36700 }, { "epoch": 1.71, "learning_rate": 2.6657513300984507e-05, "loss": 0.099, "step": 36705 }, { "epoch": 1.71, "learning_rate": 2.665704523114849e-05, "loss": 0.2015, "step": 36710 }, { "epoch": 1.71, "learning_rate": 2.665657716131247e-05, "loss": 0.166, "step": 36715 }, { "epoch": 1.71, "learning_rate": 2.665610909147645e-05, "loss": 0.1954, "step": 36720 }, { "epoch": 1.71, "learning_rate": 2.6655641021640426e-05, "loss": 0.2481, "step": 36725 }, { "epoch": 1.71, "learning_rate": 2.665517295180441e-05, "loss": 0.3079, "step": 36730 }, { "epoch": 1.71, "learning_rate": 2.665470488196839e-05, "loss": 0.196, "step": 36735 }, { "epoch": 1.71, "learning_rate": 2.665423681213237e-05, "loss": 0.0433, "step": 36740 }, { "epoch": 1.71, "learning_rate": 2.6653768742296352e-05, "loss": 0.0548, "step": 36745 }, { "epoch": 1.71, "learning_rate": 2.6653300672460332e-05, "loss": 0.0999, "step": 36750 }, { "epoch": 1.72, "learning_rate": 2.6652832602624312e-05, "loss": 0.1155, "step": 36755 }, { "epoch": 1.72, "learning_rate": 2.6652364532788292e-05, "loss": 0.1976, "step": 36760 }, { "epoch": 1.72, "learning_rate": 2.6651896462952275e-05, "loss": 0.1469, "step": 36765 }, { "epoch": 1.72, "learning_rate": 2.6651428393116255e-05, "loss": 0.1902, "step": 36770 }, { "epoch": 1.72, "learning_rate": 2.6650960323280235e-05, "loss": 0.2093, "step": 36775 }, { "epoch": 1.72, "learning_rate": 2.6650492253444214e-05, "loss": 0.232, "step": 36780 }, { "epoch": 1.72, "learning_rate": 2.6650024183608198e-05, "loss": 0.1965, "step": 36785 }, { "epoch": 1.72, "learning_rate": 2.6649556113772174e-05, "loss": 0.0518, "step": 36790 }, { "epoch": 1.72, "learning_rate": 2.6649088043936154e-05, "loss": 0.0649, "step": 36795 }, { "epoch": 1.72, "learning_rate": 2.6648619974100137e-05, "loss": 0.0456, "step": 36800 }, { "epoch": 1.72, "learning_rate": 2.6648151904264117e-05, "loss": 0.1178, "step": 36805 }, { "epoch": 1.72, "learning_rate": 2.6647683834428097e-05, "loss": 0.0895, "step": 36810 }, { "epoch": 1.72, "learning_rate": 2.6647215764592076e-05, "loss": 0.1355, "step": 36815 }, { "epoch": 1.72, "learning_rate": 2.664674769475606e-05, "loss": 0.1733, "step": 36820 }, { "epoch": 1.72, "learning_rate": 2.664627962492004e-05, "loss": 0.3833, "step": 36825 }, { "epoch": 1.72, "learning_rate": 2.664581155508402e-05, "loss": 0.1653, "step": 36830 }, { "epoch": 1.72, "learning_rate": 2.6645343485248e-05, "loss": 0.2345, "step": 36835 }, { "epoch": 1.72, "learning_rate": 2.6644875415411982e-05, "loss": 0.0921, "step": 36840 }, { "epoch": 1.72, "learning_rate": 2.6644407345575962e-05, "loss": 0.0572, "step": 36845 }, { "epoch": 1.72, "learning_rate": 2.664393927573994e-05, "loss": 0.1026, "step": 36850 }, { "epoch": 1.72, "learning_rate": 2.6643471205903922e-05, "loss": 0.077, "step": 36855 }, { "epoch": 1.72, "learning_rate": 2.66430031360679e-05, "loss": 0.1494, "step": 36860 }, { "epoch": 1.72, "learning_rate": 2.664253506623188e-05, "loss": 0.2267, "step": 36865 }, { "epoch": 1.72, "learning_rate": 2.664206699639586e-05, "loss": 0.1093, "step": 36870 }, { "epoch": 1.72, "learning_rate": 2.6641598926559844e-05, "loss": 0.134, "step": 36875 }, { "epoch": 1.72, "learning_rate": 2.6641130856723824e-05, "loss": 0.3072, "step": 36880 }, { "epoch": 1.72, "learning_rate": 2.6640662786887804e-05, "loss": 0.3128, "step": 36885 }, { "epoch": 1.72, "learning_rate": 2.6640194717051784e-05, "loss": 0.0741, "step": 36890 }, { "epoch": 1.72, "learning_rate": 2.6639726647215767e-05, "loss": 0.0445, "step": 36895 }, { "epoch": 1.72, "learning_rate": 2.6639258577379747e-05, "loss": 0.1302, "step": 36900 }, { "epoch": 1.72, "learning_rate": 2.6638790507543727e-05, "loss": 0.058, "step": 36905 }, { "epoch": 1.72, "learning_rate": 2.6638322437707707e-05, "loss": 0.0709, "step": 36910 }, { "epoch": 1.72, "learning_rate": 2.6637854367871686e-05, "loss": 0.1321, "step": 36915 }, { "epoch": 1.72, "learning_rate": 2.6637386298035666e-05, "loss": 0.1702, "step": 36920 }, { "epoch": 1.72, "learning_rate": 2.6636918228199646e-05, "loss": 0.3435, "step": 36925 }, { "epoch": 1.72, "learning_rate": 2.663645015836363e-05, "loss": 0.2582, "step": 36930 }, { "epoch": 1.72, "learning_rate": 2.663598208852761e-05, "loss": 0.3827, "step": 36935 }, { "epoch": 1.72, "learning_rate": 2.663551401869159e-05, "loss": 0.0782, "step": 36940 }, { "epoch": 1.72, "learning_rate": 2.663504594885557e-05, "loss": 0.0933, "step": 36945 }, { "epoch": 1.72, "learning_rate": 2.6634577879019552e-05, "loss": 0.0798, "step": 36950 }, { "epoch": 1.72, "learning_rate": 2.6634109809183532e-05, "loss": 0.1146, "step": 36955 }, { "epoch": 1.72, "learning_rate": 2.663364173934751e-05, "loss": 0.104, "step": 36960 }, { "epoch": 1.72, "learning_rate": 2.663317366951149e-05, "loss": 0.1681, "step": 36965 }, { "epoch": 1.73, "learning_rate": 2.6632705599675475e-05, "loss": 0.1804, "step": 36970 }, { "epoch": 1.73, "learning_rate": 2.6632237529839454e-05, "loss": 0.233, "step": 36975 }, { "epoch": 1.73, "learning_rate": 2.663176946000343e-05, "loss": 0.2976, "step": 36980 }, { "epoch": 1.73, "learning_rate": 2.6631301390167414e-05, "loss": 0.1425, "step": 36985 }, { "epoch": 1.73, "learning_rate": 2.6630833320331394e-05, "loss": 0.0479, "step": 36990 }, { "epoch": 1.73, "learning_rate": 2.6630365250495374e-05, "loss": 0.0603, "step": 36995 }, { "epoch": 1.73, "learning_rate": 2.6629897180659353e-05, "loss": 0.0688, "step": 37000 }, { "epoch": 1.73, "learning_rate": 2.6629429110823337e-05, "loss": 0.0407, "step": 37005 }, { "epoch": 1.73, "learning_rate": 2.6628961040987316e-05, "loss": 0.077, "step": 37010 }, { "epoch": 1.73, "learning_rate": 2.6628492971151296e-05, "loss": 0.1341, "step": 37015 }, { "epoch": 1.73, "learning_rate": 2.6628024901315276e-05, "loss": 0.1999, "step": 37020 }, { "epoch": 1.73, "learning_rate": 2.662755683147926e-05, "loss": 0.1489, "step": 37025 }, { "epoch": 1.73, "learning_rate": 2.662708876164324e-05, "loss": 0.3498, "step": 37030 }, { "epoch": 1.73, "learning_rate": 2.662662069180722e-05, "loss": 0.2524, "step": 37035 }, { "epoch": 1.73, "learning_rate": 2.66261526219712e-05, "loss": 0.0547, "step": 37040 }, { "epoch": 1.73, "learning_rate": 2.662568455213518e-05, "loss": 0.0783, "step": 37045 }, { "epoch": 1.73, "learning_rate": 2.662521648229916e-05, "loss": 0.0745, "step": 37050 }, { "epoch": 1.73, "learning_rate": 2.6624748412463138e-05, "loss": 0.1096, "step": 37055 }, { "epoch": 1.73, "learning_rate": 2.662428034262712e-05, "loss": 0.0486, "step": 37060 }, { "epoch": 1.73, "learning_rate": 2.66238122727911e-05, "loss": 0.1119, "step": 37065 }, { "epoch": 1.73, "learning_rate": 2.662334420295508e-05, "loss": 0.2016, "step": 37070 }, { "epoch": 1.73, "learning_rate": 2.662287613311906e-05, "loss": 0.2012, "step": 37075 }, { "epoch": 1.73, "learning_rate": 2.6622408063283044e-05, "loss": 0.4105, "step": 37080 }, { "epoch": 1.73, "learning_rate": 2.6621939993447024e-05, "loss": 0.2848, "step": 37085 }, { "epoch": 1.73, "learning_rate": 2.6621471923611004e-05, "loss": 0.0369, "step": 37090 }, { "epoch": 1.73, "learning_rate": 2.6621003853774984e-05, "loss": 0.0729, "step": 37095 }, { "epoch": 1.73, "learning_rate": 2.6620535783938967e-05, "loss": 0.0477, "step": 37100 }, { "epoch": 1.73, "learning_rate": 2.6620067714102943e-05, "loss": 0.0941, "step": 37105 }, { "epoch": 1.73, "learning_rate": 2.6619599644266923e-05, "loss": 0.1189, "step": 37110 }, { "epoch": 1.73, "learning_rate": 2.6619131574430906e-05, "loss": 0.1189, "step": 37115 }, { "epoch": 1.73, "learning_rate": 2.6618663504594886e-05, "loss": 0.1714, "step": 37120 }, { "epoch": 1.73, "learning_rate": 2.6618195434758866e-05, "loss": 0.2479, "step": 37125 }, { "epoch": 1.73, "learning_rate": 2.6617727364922846e-05, "loss": 0.319, "step": 37130 }, { "epoch": 1.73, "learning_rate": 2.661725929508683e-05, "loss": 0.2928, "step": 37135 }, { "epoch": 1.73, "learning_rate": 2.661679122525081e-05, "loss": 0.0276, "step": 37140 }, { "epoch": 1.73, "learning_rate": 2.661632315541479e-05, "loss": 0.0419, "step": 37145 }, { "epoch": 1.73, "learning_rate": 2.6615855085578768e-05, "loss": 0.0781, "step": 37150 }, { "epoch": 1.73, "learning_rate": 2.661538701574275e-05, "loss": 0.0888, "step": 37155 }, { "epoch": 1.73, "learning_rate": 2.661491894590673e-05, "loss": 0.0736, "step": 37160 }, { "epoch": 1.73, "learning_rate": 2.661445087607071e-05, "loss": 0.1355, "step": 37165 }, { "epoch": 1.73, "learning_rate": 2.661398280623469e-05, "loss": 0.1683, "step": 37170 }, { "epoch": 1.73, "learning_rate": 2.661351473639867e-05, "loss": 0.227, "step": 37175 }, { "epoch": 1.73, "learning_rate": 2.661304666656265e-05, "loss": 0.388, "step": 37180 }, { "epoch": 1.74, "learning_rate": 2.661257859672663e-05, "loss": 0.2326, "step": 37185 }, { "epoch": 1.74, "learning_rate": 2.6612110526890614e-05, "loss": 0.0476, "step": 37190 }, { "epoch": 1.74, "learning_rate": 2.6611642457054593e-05, "loss": 0.053, "step": 37195 }, { "epoch": 1.74, "learning_rate": 2.6611174387218573e-05, "loss": 0.0602, "step": 37200 }, { "epoch": 1.74, "learning_rate": 2.6610706317382553e-05, "loss": 0.0839, "step": 37205 }, { "epoch": 1.74, "learning_rate": 2.6610238247546536e-05, "loss": 0.1323, "step": 37210 }, { "epoch": 1.74, "learning_rate": 2.6609770177710516e-05, "loss": 0.2045, "step": 37215 }, { "epoch": 1.74, "learning_rate": 2.6609302107874496e-05, "loss": 0.1737, "step": 37220 }, { "epoch": 1.74, "learning_rate": 2.660883403803848e-05, "loss": 0.2023, "step": 37225 }, { "epoch": 1.74, "learning_rate": 2.6608365968202456e-05, "loss": 0.1926, "step": 37230 }, { "epoch": 1.74, "learning_rate": 2.6607897898366435e-05, "loss": 0.2431, "step": 37235 }, { "epoch": 1.74, "learning_rate": 2.6607429828530415e-05, "loss": 0.0595, "step": 37240 }, { "epoch": 1.74, "learning_rate": 2.66069617586944e-05, "loss": 0.1053, "step": 37245 }, { "epoch": 1.74, "learning_rate": 2.6606493688858378e-05, "loss": 0.2247, "step": 37250 }, { "epoch": 1.74, "learning_rate": 2.6606025619022358e-05, "loss": 0.1125, "step": 37255 }, { "epoch": 1.74, "learning_rate": 2.6605557549186338e-05, "loss": 0.0894, "step": 37260 }, { "epoch": 1.74, "learning_rate": 2.660508947935032e-05, "loss": 0.1214, "step": 37265 }, { "epoch": 1.74, "learning_rate": 2.66046214095143e-05, "loss": 0.1771, "step": 37270 }, { "epoch": 1.74, "learning_rate": 2.660415333967828e-05, "loss": 0.1925, "step": 37275 }, { "epoch": 1.74, "learning_rate": 2.6603685269842264e-05, "loss": 0.3131, "step": 37280 }, { "epoch": 1.74, "learning_rate": 2.6603217200006244e-05, "loss": 0.2135, "step": 37285 }, { "epoch": 1.74, "learning_rate": 2.6602749130170224e-05, "loss": 0.0122, "step": 37290 }, { "epoch": 1.74, "learning_rate": 2.66022810603342e-05, "loss": 0.0769, "step": 37295 }, { "epoch": 1.74, "learning_rate": 2.6601812990498183e-05, "loss": 0.1122, "step": 37300 }, { "epoch": 1.74, "learning_rate": 2.6601344920662163e-05, "loss": 0.0907, "step": 37305 }, { "epoch": 1.74, "learning_rate": 2.6600876850826143e-05, "loss": 0.1191, "step": 37310 }, { "epoch": 1.74, "learning_rate": 2.6600408780990123e-05, "loss": 0.104, "step": 37315 }, { "epoch": 1.74, "learning_rate": 2.6599940711154106e-05, "loss": 0.0915, "step": 37320 }, { "epoch": 1.74, "learning_rate": 2.6599472641318086e-05, "loss": 0.2602, "step": 37325 }, { "epoch": 1.74, "learning_rate": 2.6599004571482065e-05, "loss": 0.2612, "step": 37330 }, { "epoch": 1.74, "learning_rate": 2.6598536501646045e-05, "loss": 0.3259, "step": 37335 }, { "epoch": 1.74, "learning_rate": 2.659806843181003e-05, "loss": 0.0526, "step": 37340 }, { "epoch": 1.74, "learning_rate": 2.6597600361974008e-05, "loss": 0.0308, "step": 37345 }, { "epoch": 1.74, "learning_rate": 2.6597132292137988e-05, "loss": 0.064, "step": 37350 }, { "epoch": 1.74, "learning_rate": 2.6596664222301968e-05, "loss": 0.1317, "step": 37355 }, { "epoch": 1.74, "learning_rate": 2.6596196152465948e-05, "loss": 0.1705, "step": 37360 }, { "epoch": 1.74, "learning_rate": 2.6595728082629928e-05, "loss": 0.1577, "step": 37365 }, { "epoch": 1.74, "learning_rate": 2.6595260012793907e-05, "loss": 0.2225, "step": 37370 }, { "epoch": 1.74, "learning_rate": 2.659479194295789e-05, "loss": 0.2118, "step": 37375 }, { "epoch": 1.74, "learning_rate": 2.659432387312187e-05, "loss": 0.2221, "step": 37380 }, { "epoch": 1.74, "learning_rate": 2.659385580328585e-05, "loss": 0.2258, "step": 37385 }, { "epoch": 1.74, "learning_rate": 2.659338773344983e-05, "loss": 0.0274, "step": 37390 }, { "epoch": 1.74, "learning_rate": 2.6592919663613813e-05, "loss": 0.0474, "step": 37395 }, { "epoch": 1.75, "learning_rate": 2.6592451593777793e-05, "loss": 0.0394, "step": 37400 }, { "epoch": 1.75, "learning_rate": 2.6591983523941773e-05, "loss": 0.0899, "step": 37405 }, { "epoch": 1.75, "learning_rate": 2.6591515454105756e-05, "loss": 0.1538, "step": 37410 }, { "epoch": 1.75, "learning_rate": 2.6591047384269736e-05, "loss": 0.1078, "step": 37415 }, { "epoch": 1.75, "learning_rate": 2.6590579314433712e-05, "loss": 0.1219, "step": 37420 }, { "epoch": 1.75, "learning_rate": 2.6590111244597692e-05, "loss": 0.2225, "step": 37425 }, { "epoch": 1.75, "learning_rate": 2.6589643174761675e-05, "loss": 0.4069, "step": 37430 }, { "epoch": 1.75, "learning_rate": 2.6589175104925655e-05, "loss": 0.2965, "step": 37435 }, { "epoch": 1.75, "learning_rate": 2.6588707035089635e-05, "loss": 0.1096, "step": 37440 }, { "epoch": 1.75, "learning_rate": 2.6588238965253615e-05, "loss": 0.0941, "step": 37445 }, { "epoch": 1.75, "learning_rate": 2.6587770895417598e-05, "loss": 0.0937, "step": 37450 }, { "epoch": 1.75, "learning_rate": 2.6587302825581578e-05, "loss": 0.0944, "step": 37455 }, { "epoch": 1.75, "learning_rate": 2.6586834755745558e-05, "loss": 0.1426, "step": 37460 }, { "epoch": 1.75, "learning_rate": 2.658636668590954e-05, "loss": 0.1898, "step": 37465 }, { "epoch": 1.75, "learning_rate": 2.658589861607352e-05, "loss": 0.178, "step": 37470 }, { "epoch": 1.75, "learning_rate": 2.65854305462375e-05, "loss": 0.2354, "step": 37475 }, { "epoch": 1.75, "learning_rate": 2.658496247640148e-05, "loss": 0.2504, "step": 37480 }, { "epoch": 1.75, "learning_rate": 2.658449440656546e-05, "loss": 0.279, "step": 37485 }, { "epoch": 1.75, "learning_rate": 2.658402633672944e-05, "loss": 0.0669, "step": 37490 }, { "epoch": 1.75, "learning_rate": 2.658355826689342e-05, "loss": 0.0254, "step": 37495 }, { "epoch": 1.75, "learning_rate": 2.65830901970574e-05, "loss": 0.054, "step": 37500 }, { "epoch": 1.75, "learning_rate": 2.6582622127221383e-05, "loss": 0.135, "step": 37505 }, { "epoch": 1.75, "learning_rate": 2.6582154057385363e-05, "loss": 0.1475, "step": 37510 }, { "epoch": 1.75, "learning_rate": 2.6581685987549342e-05, "loss": 0.1666, "step": 37515 }, { "epoch": 1.75, "learning_rate": 2.6581217917713322e-05, "loss": 0.1303, "step": 37520 }, { "epoch": 1.75, "learning_rate": 2.6580749847877305e-05, "loss": 0.2202, "step": 37525 }, { "epoch": 1.75, "learning_rate": 2.6580281778041285e-05, "loss": 0.2511, "step": 37530 }, { "epoch": 1.75, "learning_rate": 2.6579813708205265e-05, "loss": 0.1981, "step": 37535 }, { "epoch": 1.75, "learning_rate": 2.6579345638369248e-05, "loss": 0.0422, "step": 37540 }, { "epoch": 1.75, "learning_rate": 2.6578877568533225e-05, "loss": 0.0778, "step": 37545 }, { "epoch": 1.75, "learning_rate": 2.6578409498697205e-05, "loss": 0.0912, "step": 37550 }, { "epoch": 1.75, "learning_rate": 2.6577941428861184e-05, "loss": 0.081, "step": 37555 }, { "epoch": 1.75, "learning_rate": 2.6577473359025168e-05, "loss": 0.1607, "step": 37560 }, { "epoch": 1.75, "learning_rate": 2.6577005289189147e-05, "loss": 0.1126, "step": 37565 }, { "epoch": 1.75, "learning_rate": 2.6576537219353127e-05, "loss": 0.1883, "step": 37570 }, { "epoch": 1.75, "learning_rate": 2.6576069149517107e-05, "loss": 0.1592, "step": 37575 }, { "epoch": 1.75, "learning_rate": 2.657560107968109e-05, "loss": 0.2201, "step": 37580 }, { "epoch": 1.75, "learning_rate": 2.657513300984507e-05, "loss": 0.2566, "step": 37585 }, { "epoch": 1.75, "learning_rate": 2.657466494000905e-05, "loss": 0.0542, "step": 37590 }, { "epoch": 1.75, "learning_rate": 2.6574196870173033e-05, "loss": 0.0666, "step": 37595 }, { "epoch": 1.75, "learning_rate": 2.6573728800337013e-05, "loss": 0.0645, "step": 37600 }, { "epoch": 1.75, "learning_rate": 2.6573260730500993e-05, "loss": 0.0866, "step": 37605 }, { "epoch": 1.75, "learning_rate": 2.657279266066497e-05, "loss": 0.1118, "step": 37610 }, { "epoch": 1.76, "learning_rate": 2.6572324590828952e-05, "loss": 0.1532, "step": 37615 }, { "epoch": 1.76, "learning_rate": 2.6571856520992932e-05, "loss": 0.1636, "step": 37620 }, { "epoch": 1.76, "learning_rate": 2.6571388451156912e-05, "loss": 0.263, "step": 37625 }, { "epoch": 1.76, "learning_rate": 2.6570920381320892e-05, "loss": 0.2552, "step": 37630 }, { "epoch": 1.76, "learning_rate": 2.6570452311484875e-05, "loss": 0.2765, "step": 37635 }, { "epoch": 1.76, "learning_rate": 2.6569984241648855e-05, "loss": 0.0426, "step": 37640 }, { "epoch": 1.76, "learning_rate": 2.6569516171812835e-05, "loss": 0.0402, "step": 37645 }, { "epoch": 1.76, "learning_rate": 2.6569048101976818e-05, "loss": 0.0363, "step": 37650 }, { "epoch": 1.76, "learning_rate": 2.6568580032140798e-05, "loss": 0.105, "step": 37655 }, { "epoch": 1.76, "learning_rate": 2.6568111962304777e-05, "loss": 0.1412, "step": 37660 }, { "epoch": 1.76, "learning_rate": 2.6567643892468757e-05, "loss": 0.1898, "step": 37665 }, { "epoch": 1.76, "learning_rate": 2.656717582263274e-05, "loss": 0.1251, "step": 37670 }, { "epoch": 1.76, "learning_rate": 2.6566707752796717e-05, "loss": 0.1053, "step": 37675 }, { "epoch": 1.76, "learning_rate": 2.6566239682960697e-05, "loss": 0.3299, "step": 37680 }, { "epoch": 1.76, "learning_rate": 2.6565771613124677e-05, "loss": 0.3651, "step": 37685 }, { "epoch": 1.76, "learning_rate": 2.656530354328866e-05, "loss": 0.0508, "step": 37690 }, { "epoch": 1.76, "learning_rate": 2.656483547345264e-05, "loss": 0.0858, "step": 37695 }, { "epoch": 1.76, "learning_rate": 2.656436740361662e-05, "loss": 0.0825, "step": 37700 }, { "epoch": 1.76, "learning_rate": 2.65638993337806e-05, "loss": 0.1217, "step": 37705 }, { "epoch": 1.76, "learning_rate": 2.6563431263944582e-05, "loss": 0.1133, "step": 37710 }, { "epoch": 1.76, "learning_rate": 2.6562963194108562e-05, "loss": 0.1573, "step": 37715 }, { "epoch": 1.76, "learning_rate": 2.6562495124272542e-05, "loss": 0.1862, "step": 37720 }, { "epoch": 1.76, "learning_rate": 2.6562027054436525e-05, "loss": 0.212, "step": 37725 }, { "epoch": 1.76, "learning_rate": 2.6561558984600505e-05, "loss": 0.4626, "step": 37730 }, { "epoch": 1.76, "learning_rate": 2.656109091476448e-05, "loss": 0.2542, "step": 37735 }, { "epoch": 1.76, "learning_rate": 2.656062284492846e-05, "loss": 0.0593, "step": 37740 }, { "epoch": 1.76, "learning_rate": 2.6560154775092445e-05, "loss": 0.0434, "step": 37745 }, { "epoch": 1.76, "learning_rate": 2.6559686705256424e-05, "loss": 0.1627, "step": 37750 }, { "epoch": 1.76, "learning_rate": 2.6559218635420404e-05, "loss": 0.0383, "step": 37755 }, { "epoch": 1.76, "learning_rate": 2.6558750565584384e-05, "loss": 0.1075, "step": 37760 }, { "epoch": 1.76, "learning_rate": 2.6558282495748367e-05, "loss": 0.092, "step": 37765 }, { "epoch": 1.76, "learning_rate": 2.6557814425912347e-05, "loss": 0.1523, "step": 37770 }, { "epoch": 1.76, "learning_rate": 2.6557346356076327e-05, "loss": 0.189, "step": 37775 }, { "epoch": 1.76, "learning_rate": 2.655687828624031e-05, "loss": 0.2829, "step": 37780 }, { "epoch": 1.76, "learning_rate": 2.655641021640429e-05, "loss": 0.2666, "step": 37785 }, { "epoch": 1.76, "learning_rate": 2.655594214656827e-05, "loss": 0.0286, "step": 37790 }, { "epoch": 1.76, "learning_rate": 2.655547407673225e-05, "loss": 0.0591, "step": 37795 }, { "epoch": 1.76, "learning_rate": 2.655500600689623e-05, "loss": 0.0541, "step": 37800 }, { "epoch": 1.76, "learning_rate": 2.655453793706021e-05, "loss": 0.1371, "step": 37805 }, { "epoch": 1.76, "learning_rate": 2.655406986722419e-05, "loss": 0.1488, "step": 37810 }, { "epoch": 1.76, "learning_rate": 2.655360179738817e-05, "loss": 0.1617, "step": 37815 }, { "epoch": 1.76, "learning_rate": 2.6553133727552152e-05, "loss": 0.1725, "step": 37820 }, { "epoch": 1.76, "learning_rate": 2.6552665657716132e-05, "loss": 0.2266, "step": 37825 }, { "epoch": 1.77, "learning_rate": 2.655219758788011e-05, "loss": 0.3259, "step": 37830 }, { "epoch": 1.77, "learning_rate": 2.6551729518044095e-05, "loss": 0.2705, "step": 37835 }, { "epoch": 1.77, "learning_rate": 2.6551261448208075e-05, "loss": 0.0615, "step": 37840 }, { "epoch": 1.77, "learning_rate": 2.6550793378372054e-05, "loss": 0.0913, "step": 37845 }, { "epoch": 1.77, "learning_rate": 2.6550325308536034e-05, "loss": 0.0386, "step": 37850 }, { "epoch": 1.77, "learning_rate": 2.6549857238700017e-05, "loss": 0.0523, "step": 37855 }, { "epoch": 1.77, "learning_rate": 2.6549389168863994e-05, "loss": 0.1444, "step": 37860 }, { "epoch": 1.77, "learning_rate": 2.6548921099027974e-05, "loss": 0.1165, "step": 37865 }, { "epoch": 1.77, "learning_rate": 2.6548453029191954e-05, "loss": 0.1582, "step": 37870 }, { "epoch": 1.77, "learning_rate": 2.6547984959355937e-05, "loss": 0.224, "step": 37875 }, { "epoch": 1.77, "learning_rate": 2.6547516889519917e-05, "loss": 0.2586, "step": 37880 }, { "epoch": 1.77, "learning_rate": 2.6547048819683896e-05, "loss": 0.2661, "step": 37885 }, { "epoch": 1.77, "learning_rate": 2.654658074984788e-05, "loss": 0.0212, "step": 37890 }, { "epoch": 1.77, "learning_rate": 2.654611268001186e-05, "loss": 0.0424, "step": 37895 }, { "epoch": 1.77, "learning_rate": 2.654564461017584e-05, "loss": 0.0965, "step": 37900 }, { "epoch": 1.77, "learning_rate": 2.654517654033982e-05, "loss": 0.1297, "step": 37905 }, { "epoch": 1.77, "learning_rate": 2.6544708470503802e-05, "loss": 0.1428, "step": 37910 }, { "epoch": 1.77, "learning_rate": 2.6544240400667782e-05, "loss": 0.1064, "step": 37915 }, { "epoch": 1.77, "learning_rate": 2.6543772330831762e-05, "loss": 0.2083, "step": 37920 }, { "epoch": 1.77, "learning_rate": 2.6543304260995738e-05, "loss": 0.2806, "step": 37925 }, { "epoch": 1.77, "learning_rate": 2.654283619115972e-05, "loss": 0.2333, "step": 37930 }, { "epoch": 1.77, "learning_rate": 2.65423681213237e-05, "loss": 0.208, "step": 37935 }, { "epoch": 1.77, "learning_rate": 2.654190005148768e-05, "loss": 0.0392, "step": 37940 }, { "epoch": 1.77, "learning_rate": 2.654143198165166e-05, "loss": 0.0543, "step": 37945 }, { "epoch": 1.77, "learning_rate": 2.6540963911815644e-05, "loss": 0.0375, "step": 37950 }, { "epoch": 1.77, "learning_rate": 2.6540495841979624e-05, "loss": 0.0513, "step": 37955 }, { "epoch": 1.77, "learning_rate": 2.6540027772143604e-05, "loss": 0.0978, "step": 37960 }, { "epoch": 1.77, "learning_rate": 2.6539559702307587e-05, "loss": 0.1591, "step": 37965 }, { "epoch": 1.77, "learning_rate": 2.6539091632471567e-05, "loss": 0.186, "step": 37970 }, { "epoch": 1.77, "learning_rate": 2.6538623562635547e-05, "loss": 0.2742, "step": 37975 }, { "epoch": 1.77, "learning_rate": 2.6538155492799526e-05, "loss": 0.2661, "step": 37980 }, { "epoch": 1.77, "learning_rate": 2.653768742296351e-05, "loss": 0.3483, "step": 37985 }, { "epoch": 1.77, "learning_rate": 2.6537219353127486e-05, "loss": 0.0242, "step": 37990 }, { "epoch": 1.77, "learning_rate": 2.6536751283291466e-05, "loss": 0.0226, "step": 37995 }, { "epoch": 1.77, "learning_rate": 2.6536283213455446e-05, "loss": 0.0236, "step": 38000 }, { "epoch": 1.77, "learning_rate": 2.653581514361943e-05, "loss": 0.0715, "step": 38005 }, { "epoch": 1.77, "learning_rate": 2.653534707378341e-05, "loss": 0.1452, "step": 38010 }, { "epoch": 1.77, "learning_rate": 2.653487900394739e-05, "loss": 0.0935, "step": 38015 }, { "epoch": 1.77, "learning_rate": 2.6534410934111372e-05, "loss": 0.1286, "step": 38020 }, { "epoch": 1.77, "learning_rate": 2.653394286427535e-05, "loss": 0.2023, "step": 38025 }, { "epoch": 1.77, "learning_rate": 2.653347479443933e-05, "loss": 0.2038, "step": 38030 }, { "epoch": 1.77, "learning_rate": 2.653300672460331e-05, "loss": 0.3278, "step": 38035 }, { "epoch": 1.77, "learning_rate": 2.6532538654767294e-05, "loss": 0.0397, "step": 38040 }, { "epoch": 1.78, "learning_rate": 2.6532070584931274e-05, "loss": 0.1115, "step": 38045 }, { "epoch": 1.78, "learning_rate": 2.653160251509525e-05, "loss": 0.0542, "step": 38050 }, { "epoch": 1.78, "learning_rate": 2.653113444525923e-05, "loss": 0.0861, "step": 38055 }, { "epoch": 1.78, "learning_rate": 2.6530666375423214e-05, "loss": 0.1127, "step": 38060 }, { "epoch": 1.78, "learning_rate": 2.6530198305587194e-05, "loss": 0.1366, "step": 38065 }, { "epoch": 1.78, "learning_rate": 2.6529730235751173e-05, "loss": 0.1672, "step": 38070 }, { "epoch": 1.78, "learning_rate": 2.6529262165915157e-05, "loss": 0.2479, "step": 38075 }, { "epoch": 1.78, "learning_rate": 2.6528794096079136e-05, "loss": 0.3074, "step": 38080 }, { "epoch": 1.78, "learning_rate": 2.6528326026243116e-05, "loss": 0.2058, "step": 38085 }, { "epoch": 1.78, "learning_rate": 2.6527857956407096e-05, "loss": 0.0266, "step": 38090 }, { "epoch": 1.78, "learning_rate": 2.652738988657108e-05, "loss": 0.0687, "step": 38095 }, { "epoch": 1.78, "learning_rate": 2.652692181673506e-05, "loss": 0.0764, "step": 38100 }, { "epoch": 1.78, "learning_rate": 2.652645374689904e-05, "loss": 0.0698, "step": 38105 }, { "epoch": 1.78, "learning_rate": 2.652598567706302e-05, "loss": 0.1162, "step": 38110 }, { "epoch": 1.78, "learning_rate": 2.6525517607227e-05, "loss": 0.1294, "step": 38115 }, { "epoch": 1.78, "learning_rate": 2.6525049537390978e-05, "loss": 0.1742, "step": 38120 }, { "epoch": 1.78, "learning_rate": 2.6524581467554958e-05, "loss": 0.125, "step": 38125 }, { "epoch": 1.78, "learning_rate": 2.6524113397718938e-05, "loss": 0.3642, "step": 38130 }, { "epoch": 1.78, "learning_rate": 2.652364532788292e-05, "loss": 0.2685, "step": 38135 }, { "epoch": 1.78, "learning_rate": 2.65231772580469e-05, "loss": 0.0532, "step": 38140 }, { "epoch": 1.78, "learning_rate": 2.652270918821088e-05, "loss": 0.0493, "step": 38145 }, { "epoch": 1.78, "learning_rate": 2.6522241118374864e-05, "loss": 0.0884, "step": 38150 }, { "epoch": 1.78, "learning_rate": 2.6521773048538844e-05, "loss": 0.0669, "step": 38155 }, { "epoch": 1.78, "learning_rate": 2.6521304978702824e-05, "loss": 0.12, "step": 38160 }, { "epoch": 1.78, "learning_rate": 2.6520836908866803e-05, "loss": 0.0843, "step": 38165 }, { "epoch": 1.78, "learning_rate": 2.6520368839030787e-05, "loss": 0.2058, "step": 38170 }, { "epoch": 1.78, "learning_rate": 2.6519900769194766e-05, "loss": 0.1966, "step": 38175 }, { "epoch": 1.78, "learning_rate": 2.6519432699358743e-05, "loss": 0.3597, "step": 38180 }, { "epoch": 1.78, "learning_rate": 2.6518964629522723e-05, "loss": 0.3126, "step": 38185 }, { "epoch": 1.78, "learning_rate": 2.6518496559686706e-05, "loss": 0.0241, "step": 38190 }, { "epoch": 1.78, "learning_rate": 2.6518028489850686e-05, "loss": 0.0725, "step": 38195 }, { "epoch": 1.78, "learning_rate": 2.6517560420014666e-05, "loss": 0.044, "step": 38200 }, { "epoch": 1.78, "learning_rate": 2.651709235017865e-05, "loss": 0.1043, "step": 38205 }, { "epoch": 1.78, "learning_rate": 2.651662428034263e-05, "loss": 0.0978, "step": 38210 }, { "epoch": 1.78, "learning_rate": 2.651615621050661e-05, "loss": 0.2056, "step": 38215 }, { "epoch": 1.78, "learning_rate": 2.6515688140670588e-05, "loss": 0.1408, "step": 38220 }, { "epoch": 1.78, "learning_rate": 2.651522007083457e-05, "loss": 0.1687, "step": 38225 }, { "epoch": 1.78, "learning_rate": 2.651475200099855e-05, "loss": 0.2897, "step": 38230 }, { "epoch": 1.78, "learning_rate": 2.651428393116253e-05, "loss": 0.1992, "step": 38235 }, { "epoch": 1.78, "learning_rate": 2.6513815861326507e-05, "loss": 0.0673, "step": 38240 }, { "epoch": 1.78, "learning_rate": 2.651334779149049e-05, "loss": 0.058, "step": 38245 }, { "epoch": 1.78, "learning_rate": 2.651287972165447e-05, "loss": 0.1164, "step": 38250 }, { "epoch": 1.79, "learning_rate": 2.651241165181845e-05, "loss": 0.1625, "step": 38255 }, { "epoch": 1.79, "learning_rate": 2.6511943581982434e-05, "loss": 0.1882, "step": 38260 }, { "epoch": 1.79, "learning_rate": 2.6511475512146413e-05, "loss": 0.1015, "step": 38265 }, { "epoch": 1.79, "learning_rate": 2.6511007442310393e-05, "loss": 0.2987, "step": 38270 }, { "epoch": 1.79, "learning_rate": 2.6510539372474373e-05, "loss": 0.1598, "step": 38275 }, { "epoch": 1.79, "learning_rate": 2.6510071302638356e-05, "loss": 0.2135, "step": 38280 }, { "epoch": 1.79, "learning_rate": 2.6509603232802336e-05, "loss": 0.2599, "step": 38285 }, { "epoch": 1.79, "learning_rate": 2.6509135162966316e-05, "loss": 0.0309, "step": 38290 }, { "epoch": 1.79, "learning_rate": 2.6508667093130296e-05, "loss": 0.0275, "step": 38295 }, { "epoch": 1.79, "learning_rate": 2.650819902329428e-05, "loss": 0.0937, "step": 38300 }, { "epoch": 1.79, "learning_rate": 2.6507730953458255e-05, "loss": 0.0716, "step": 38305 }, { "epoch": 1.79, "learning_rate": 2.6507262883622235e-05, "loss": 0.158, "step": 38310 }, { "epoch": 1.79, "learning_rate": 2.6506794813786215e-05, "loss": 0.1843, "step": 38315 }, { "epoch": 1.79, "learning_rate": 2.6506326743950198e-05, "loss": 0.2241, "step": 38320 }, { "epoch": 1.79, "learning_rate": 2.6505858674114178e-05, "loss": 0.1921, "step": 38325 }, { "epoch": 1.79, "learning_rate": 2.6505390604278158e-05, "loss": 0.3494, "step": 38330 }, { "epoch": 1.79, "learning_rate": 2.650492253444214e-05, "loss": 0.2617, "step": 38335 }, { "epoch": 1.79, "learning_rate": 2.650445446460612e-05, "loss": 0.0337, "step": 38340 }, { "epoch": 1.79, "learning_rate": 2.65039863947701e-05, "loss": 0.0641, "step": 38345 }, { "epoch": 1.79, "learning_rate": 2.650351832493408e-05, "loss": 0.104, "step": 38350 }, { "epoch": 1.79, "learning_rate": 2.6503050255098064e-05, "loss": 0.0843, "step": 38355 }, { "epoch": 1.79, "learning_rate": 2.6502582185262043e-05, "loss": 0.0992, "step": 38360 }, { "epoch": 1.79, "learning_rate": 2.6502114115426023e-05, "loss": 0.1044, "step": 38365 }, { "epoch": 1.79, "learning_rate": 2.650164604559e-05, "loss": 0.1495, "step": 38370 }, { "epoch": 1.79, "learning_rate": 2.6501177975753983e-05, "loss": 0.1757, "step": 38375 }, { "epoch": 1.79, "learning_rate": 2.6500709905917963e-05, "loss": 0.3565, "step": 38380 }, { "epoch": 1.79, "learning_rate": 2.6500241836081943e-05, "loss": 0.28, "step": 38385 }, { "epoch": 1.79, "learning_rate": 2.6499773766245926e-05, "loss": 0.0916, "step": 38390 }, { "epoch": 1.79, "learning_rate": 2.6499305696409906e-05, "loss": 0.054, "step": 38395 }, { "epoch": 1.79, "learning_rate": 2.6498837626573885e-05, "loss": 0.0463, "step": 38400 }, { "epoch": 1.79, "learning_rate": 2.6498369556737865e-05, "loss": 0.0573, "step": 38405 }, { "epoch": 1.79, "learning_rate": 2.649790148690185e-05, "loss": 0.0691, "step": 38410 }, { "epoch": 1.79, "learning_rate": 2.6497433417065828e-05, "loss": 0.0652, "step": 38415 }, { "epoch": 1.79, "learning_rate": 2.6496965347229808e-05, "loss": 0.1039, "step": 38420 }, { "epoch": 1.79, "learning_rate": 2.6496497277393788e-05, "loss": 0.2298, "step": 38425 }, { "epoch": 1.79, "learning_rate": 2.6496029207557768e-05, "loss": 0.2337, "step": 38430 }, { "epoch": 1.79, "learning_rate": 2.6495561137721747e-05, "loss": 0.3419, "step": 38435 }, { "epoch": 1.79, "learning_rate": 2.6495093067885727e-05, "loss": 0.0615, "step": 38440 }, { "epoch": 1.79, "learning_rate": 2.649462499804971e-05, "loss": 0.048, "step": 38445 }, { "epoch": 1.79, "learning_rate": 2.649415692821369e-05, "loss": 0.0772, "step": 38450 }, { "epoch": 1.79, "learning_rate": 2.649368885837767e-05, "loss": 0.0492, "step": 38455 }, { "epoch": 1.79, "learning_rate": 2.649322078854165e-05, "loss": 0.1067, "step": 38460 }, { "epoch": 1.79, "learning_rate": 2.6492752718705633e-05, "loss": 0.1087, "step": 38465 }, { "epoch": 1.8, "learning_rate": 2.6492284648869613e-05, "loss": 0.0963, "step": 38470 }, { "epoch": 1.8, "learning_rate": 2.6491816579033593e-05, "loss": 0.1953, "step": 38475 }, { "epoch": 1.8, "learning_rate": 2.6491348509197573e-05, "loss": 0.2136, "step": 38480 }, { "epoch": 1.8, "learning_rate": 2.6490880439361556e-05, "loss": 0.2515, "step": 38485 }, { "epoch": 1.8, "learning_rate": 2.6490412369525536e-05, "loss": 0.0383, "step": 38490 }, { "epoch": 1.8, "learning_rate": 2.6489944299689512e-05, "loss": 0.0402, "step": 38495 }, { "epoch": 1.8, "learning_rate": 2.6489476229853492e-05, "loss": 0.125, "step": 38500 }, { "epoch": 1.8, "learning_rate": 2.6489008160017475e-05, "loss": 0.0805, "step": 38505 }, { "epoch": 1.8, "learning_rate": 2.6488540090181455e-05, "loss": 0.0915, "step": 38510 }, { "epoch": 1.8, "learning_rate": 2.6488072020345435e-05, "loss": 0.1007, "step": 38515 }, { "epoch": 1.8, "learning_rate": 2.6487603950509418e-05, "loss": 0.1853, "step": 38520 }, { "epoch": 1.8, "learning_rate": 2.6487135880673398e-05, "loss": 0.214, "step": 38525 }, { "epoch": 1.8, "learning_rate": 2.6486667810837378e-05, "loss": 0.2567, "step": 38530 }, { "epoch": 1.8, "learning_rate": 2.6486199741001357e-05, "loss": 0.2431, "step": 38535 }, { "epoch": 1.8, "learning_rate": 2.648573167116534e-05, "loss": 0.1163, "step": 38540 }, { "epoch": 1.8, "learning_rate": 2.648526360132932e-05, "loss": 0.0315, "step": 38545 }, { "epoch": 1.8, "learning_rate": 2.64847955314933e-05, "loss": 0.0545, "step": 38550 }, { "epoch": 1.8, "learning_rate": 2.648432746165728e-05, "loss": 0.1156, "step": 38555 }, { "epoch": 1.8, "learning_rate": 2.648385939182126e-05, "loss": 0.0526, "step": 38560 }, { "epoch": 1.8, "learning_rate": 2.648339132198524e-05, "loss": 0.0689, "step": 38565 }, { "epoch": 1.8, "learning_rate": 2.648292325214922e-05, "loss": 0.1696, "step": 38570 }, { "epoch": 1.8, "learning_rate": 2.6482455182313203e-05, "loss": 0.2157, "step": 38575 }, { "epoch": 1.8, "learning_rate": 2.6481987112477183e-05, "loss": 0.4338, "step": 38580 }, { "epoch": 1.8, "learning_rate": 2.6481519042641162e-05, "loss": 0.2748, "step": 38585 }, { "epoch": 1.8, "learning_rate": 2.6481050972805142e-05, "loss": 0.0525, "step": 38590 }, { "epoch": 1.8, "learning_rate": 2.6480582902969125e-05, "loss": 0.0672, "step": 38595 }, { "epoch": 1.8, "learning_rate": 2.6480114833133105e-05, "loss": 0.0741, "step": 38600 }, { "epoch": 1.8, "learning_rate": 2.6479646763297085e-05, "loss": 0.1044, "step": 38605 }, { "epoch": 1.8, "learning_rate": 2.6479178693461065e-05, "loss": 0.1173, "step": 38610 }, { "epoch": 1.8, "learning_rate": 2.6478710623625048e-05, "loss": 0.0755, "step": 38615 }, { "epoch": 1.8, "learning_rate": 2.6478242553789024e-05, "loss": 0.1642, "step": 38620 }, { "epoch": 1.8, "learning_rate": 2.6477774483953004e-05, "loss": 0.1546, "step": 38625 }, { "epoch": 1.8, "learning_rate": 2.6477306414116987e-05, "loss": 0.3669, "step": 38630 }, { "epoch": 1.8, "learning_rate": 2.6476838344280967e-05, "loss": 0.4163, "step": 38635 }, { "epoch": 1.8, "learning_rate": 2.6476370274444947e-05, "loss": 0.0366, "step": 38640 }, { "epoch": 1.8, "learning_rate": 2.6475902204608927e-05, "loss": 0.0352, "step": 38645 }, { "epoch": 1.8, "learning_rate": 2.647543413477291e-05, "loss": 0.0581, "step": 38650 }, { "epoch": 1.8, "learning_rate": 2.647496606493689e-05, "loss": 0.0955, "step": 38655 }, { "epoch": 1.8, "learning_rate": 2.647449799510087e-05, "loss": 0.1382, "step": 38660 }, { "epoch": 1.8, "learning_rate": 2.647402992526485e-05, "loss": 0.1084, "step": 38665 }, { "epoch": 1.8, "learning_rate": 2.6473561855428833e-05, "loss": 0.143, "step": 38670 }, { "epoch": 1.8, "learning_rate": 2.6473093785592813e-05, "loss": 0.0941, "step": 38675 }, { "epoch": 1.8, "learning_rate": 2.6472625715756792e-05, "loss": 0.3088, "step": 38680 }, { "epoch": 1.81, "learning_rate": 2.6472157645920772e-05, "loss": 0.2556, "step": 38685 }, { "epoch": 1.81, "learning_rate": 2.6471689576084752e-05, "loss": 0.0559, "step": 38690 }, { "epoch": 1.81, "learning_rate": 2.6471221506248732e-05, "loss": 0.0947, "step": 38695 }, { "epoch": 1.81, "learning_rate": 2.647075343641271e-05, "loss": 0.1005, "step": 38700 }, { "epoch": 1.81, "learning_rate": 2.6470285366576695e-05, "loss": 0.1112, "step": 38705 }, { "epoch": 1.81, "learning_rate": 2.6469817296740675e-05, "loss": 0.2133, "step": 38710 }, { "epoch": 1.81, "learning_rate": 2.6469349226904655e-05, "loss": 0.123, "step": 38715 }, { "epoch": 1.81, "learning_rate": 2.6468881157068634e-05, "loss": 0.1556, "step": 38720 }, { "epoch": 1.81, "learning_rate": 2.6468413087232618e-05, "loss": 0.2376, "step": 38725 }, { "epoch": 1.81, "learning_rate": 2.6467945017396597e-05, "loss": 0.3147, "step": 38730 }, { "epoch": 1.81, "learning_rate": 2.6467476947560577e-05, "loss": 0.2535, "step": 38735 }, { "epoch": 1.81, "learning_rate": 2.6467008877724557e-05, "loss": 0.0479, "step": 38740 }, { "epoch": 1.81, "learning_rate": 2.6466540807888537e-05, "loss": 0.0924, "step": 38745 }, { "epoch": 1.81, "learning_rate": 2.6466072738052517e-05, "loss": 0.0957, "step": 38750 }, { "epoch": 1.81, "learning_rate": 2.6465604668216496e-05, "loss": 0.1482, "step": 38755 }, { "epoch": 1.81, "learning_rate": 2.646513659838048e-05, "loss": 0.1378, "step": 38760 }, { "epoch": 1.81, "learning_rate": 2.646466852854446e-05, "loss": 0.1682, "step": 38765 }, { "epoch": 1.81, "learning_rate": 2.646420045870844e-05, "loss": 0.2081, "step": 38770 }, { "epoch": 1.81, "learning_rate": 2.646373238887242e-05, "loss": 0.2407, "step": 38775 }, { "epoch": 1.81, "learning_rate": 2.6463264319036402e-05, "loss": 0.2936, "step": 38780 }, { "epoch": 1.81, "learning_rate": 2.6462796249200382e-05, "loss": 0.2447, "step": 38785 }, { "epoch": 1.81, "learning_rate": 2.6462328179364362e-05, "loss": 0.0566, "step": 38790 }, { "epoch": 1.81, "learning_rate": 2.6461860109528342e-05, "loss": 0.0584, "step": 38795 }, { "epoch": 1.81, "learning_rate": 2.6461392039692325e-05, "loss": 0.1093, "step": 38800 }, { "epoch": 1.81, "learning_rate": 2.6460923969856305e-05, "loss": 0.0997, "step": 38805 }, { "epoch": 1.81, "learning_rate": 2.646045590002028e-05, "loss": 0.0793, "step": 38810 }, { "epoch": 1.81, "learning_rate": 2.6459987830184264e-05, "loss": 0.1457, "step": 38815 }, { "epoch": 1.81, "learning_rate": 2.6459519760348244e-05, "loss": 0.1768, "step": 38820 }, { "epoch": 1.81, "learning_rate": 2.6459051690512224e-05, "loss": 0.2911, "step": 38825 }, { "epoch": 1.81, "learning_rate": 2.6458583620676204e-05, "loss": 0.2443, "step": 38830 }, { "epoch": 1.81, "learning_rate": 2.6458115550840187e-05, "loss": 0.3313, "step": 38835 }, { "epoch": 1.81, "learning_rate": 2.6457647481004167e-05, "loss": 0.0332, "step": 38840 }, { "epoch": 1.81, "learning_rate": 2.6457179411168147e-05, "loss": 0.1396, "step": 38845 }, { "epoch": 1.81, "learning_rate": 2.6456711341332127e-05, "loss": 0.0749, "step": 38850 }, { "epoch": 1.81, "learning_rate": 2.645624327149611e-05, "loss": 0.1388, "step": 38855 }, { "epoch": 1.81, "learning_rate": 2.645577520166009e-05, "loss": 0.0646, "step": 38860 }, { "epoch": 1.81, "learning_rate": 2.645530713182407e-05, "loss": 0.1281, "step": 38865 }, { "epoch": 1.81, "learning_rate": 2.6454839061988053e-05, "loss": 0.1946, "step": 38870 }, { "epoch": 1.81, "learning_rate": 2.645437099215203e-05, "loss": 0.2856, "step": 38875 }, { "epoch": 1.81, "learning_rate": 2.645390292231601e-05, "loss": 0.2936, "step": 38880 }, { "epoch": 1.81, "learning_rate": 2.645343485247999e-05, "loss": 0.2996, "step": 38885 }, { "epoch": 1.81, "learning_rate": 2.6452966782643972e-05, "loss": 0.0434, "step": 38890 }, { "epoch": 1.81, "learning_rate": 2.645249871280795e-05, "loss": 0.0587, "step": 38895 }, { "epoch": 1.82, "learning_rate": 2.645203064297193e-05, "loss": 0.0594, "step": 38900 }, { "epoch": 1.82, "learning_rate": 2.645156257313591e-05, "loss": 0.0412, "step": 38905 }, { "epoch": 1.82, "learning_rate": 2.6451094503299895e-05, "loss": 0.08, "step": 38910 }, { "epoch": 1.82, "learning_rate": 2.6450626433463874e-05, "loss": 0.1877, "step": 38915 }, { "epoch": 1.82, "learning_rate": 2.6450158363627854e-05, "loss": 0.1803, "step": 38920 }, { "epoch": 1.82, "learning_rate": 2.6449690293791834e-05, "loss": 0.1956, "step": 38925 }, { "epoch": 1.82, "learning_rate": 2.6449222223955817e-05, "loss": 0.1952, "step": 38930 }, { "epoch": 1.82, "learning_rate": 2.6448754154119794e-05, "loss": 0.2757, "step": 38935 }, { "epoch": 1.82, "learning_rate": 2.6448286084283773e-05, "loss": 0.0558, "step": 38940 }, { "epoch": 1.82, "learning_rate": 2.6447818014447757e-05, "loss": 0.0174, "step": 38945 }, { "epoch": 1.82, "learning_rate": 2.6447349944611736e-05, "loss": 0.0503, "step": 38950 }, { "epoch": 1.82, "learning_rate": 2.6446881874775716e-05, "loss": 0.1026, "step": 38955 }, { "epoch": 1.82, "learning_rate": 2.6446413804939696e-05, "loss": 0.0998, "step": 38960 }, { "epoch": 1.82, "learning_rate": 2.644594573510368e-05, "loss": 0.1039, "step": 38965 }, { "epoch": 1.82, "learning_rate": 2.644547766526766e-05, "loss": 0.1503, "step": 38970 }, { "epoch": 1.82, "learning_rate": 2.644500959543164e-05, "loss": 0.2892, "step": 38975 }, { "epoch": 1.82, "learning_rate": 2.644454152559562e-05, "loss": 0.3429, "step": 38980 }, { "epoch": 1.82, "learning_rate": 2.6444073455759602e-05, "loss": 0.2999, "step": 38985 }, { "epoch": 1.82, "learning_rate": 2.6443605385923582e-05, "loss": 0.075, "step": 38990 }, { "epoch": 1.82, "learning_rate": 2.644313731608756e-05, "loss": 0.0762, "step": 38995 }, { "epoch": 1.82, "learning_rate": 2.644266924625154e-05, "loss": 0.0375, "step": 39000 }, { "epoch": 1.82, "learning_rate": 2.644220117641552e-05, "loss": 0.0747, "step": 39005 }, { "epoch": 1.82, "learning_rate": 2.64417331065795e-05, "loss": 0.1126, "step": 39010 }, { "epoch": 1.82, "learning_rate": 2.644126503674348e-05, "loss": 0.1672, "step": 39015 }, { "epoch": 1.82, "learning_rate": 2.6440796966907464e-05, "loss": 0.093, "step": 39020 }, { "epoch": 1.82, "learning_rate": 2.6440328897071444e-05, "loss": 0.2025, "step": 39025 }, { "epoch": 1.82, "learning_rate": 2.6439860827235424e-05, "loss": 0.3027, "step": 39030 }, { "epoch": 1.82, "learning_rate": 2.6439392757399404e-05, "loss": 0.1986, "step": 39035 }, { "epoch": 1.82, "learning_rate": 2.6438924687563387e-05, "loss": 0.0359, "step": 39040 }, { "epoch": 1.82, "learning_rate": 2.6438456617727367e-05, "loss": 0.0378, "step": 39045 }, { "epoch": 1.82, "learning_rate": 2.6437988547891346e-05, "loss": 0.1124, "step": 39050 }, { "epoch": 1.82, "learning_rate": 2.643752047805533e-05, "loss": 0.1104, "step": 39055 }, { "epoch": 1.82, "learning_rate": 2.643705240821931e-05, "loss": 0.1282, "step": 39060 }, { "epoch": 1.82, "learning_rate": 2.6436584338383286e-05, "loss": 0.166, "step": 39065 }, { "epoch": 1.82, "learning_rate": 2.6436116268547266e-05, "loss": 0.2484, "step": 39070 }, { "epoch": 1.82, "learning_rate": 2.643564819871125e-05, "loss": 0.1008, "step": 39075 }, { "epoch": 1.82, "learning_rate": 2.643518012887523e-05, "loss": 0.3377, "step": 39080 }, { "epoch": 1.82, "learning_rate": 2.643471205903921e-05, "loss": 0.2496, "step": 39085 }, { "epoch": 1.82, "learning_rate": 2.6434243989203188e-05, "loss": 0.0415, "step": 39090 }, { "epoch": 1.82, "learning_rate": 2.643377591936717e-05, "loss": 0.0437, "step": 39095 }, { "epoch": 1.82, "learning_rate": 2.643330784953115e-05, "loss": 0.0635, "step": 39100 }, { "epoch": 1.82, "learning_rate": 2.643283977969513e-05, "loss": 0.0702, "step": 39105 }, { "epoch": 1.82, "learning_rate": 2.643237170985911e-05, "loss": 0.152, "step": 39110 }, { "epoch": 1.83, "learning_rate": 2.6431903640023094e-05, "loss": 0.1106, "step": 39115 }, { "epoch": 1.83, "learning_rate": 2.6431435570187074e-05, "loss": 0.1144, "step": 39120 }, { "epoch": 1.83, "learning_rate": 2.643096750035105e-05, "loss": 0.1665, "step": 39125 }, { "epoch": 1.83, "learning_rate": 2.6430499430515034e-05, "loss": 0.4198, "step": 39130 }, { "epoch": 1.83, "learning_rate": 2.6430031360679013e-05, "loss": 0.1532, "step": 39135 }, { "epoch": 1.83, "learning_rate": 2.6429563290842993e-05, "loss": 0.0502, "step": 39140 }, { "epoch": 1.83, "learning_rate": 2.6429095221006973e-05, "loss": 0.042, "step": 39145 }, { "epoch": 1.83, "learning_rate": 2.6428627151170956e-05, "loss": 0.0995, "step": 39150 }, { "epoch": 1.83, "learning_rate": 2.6428159081334936e-05, "loss": 0.0965, "step": 39155 }, { "epoch": 1.83, "learning_rate": 2.6427691011498916e-05, "loss": 0.1431, "step": 39160 }, { "epoch": 1.83, "learning_rate": 2.6427222941662896e-05, "loss": 0.2606, "step": 39165 }, { "epoch": 1.83, "learning_rate": 2.642675487182688e-05, "loss": 0.2053, "step": 39170 }, { "epoch": 1.83, "learning_rate": 2.642628680199086e-05, "loss": 0.2393, "step": 39175 }, { "epoch": 1.83, "learning_rate": 2.642581873215484e-05, "loss": 0.3359, "step": 39180 }, { "epoch": 1.83, "learning_rate": 2.6425350662318822e-05, "loss": 0.2666, "step": 39185 }, { "epoch": 1.83, "learning_rate": 2.6424882592482798e-05, "loss": 0.0378, "step": 39190 }, { "epoch": 1.83, "learning_rate": 2.6424414522646778e-05, "loss": 0.125, "step": 39195 }, { "epoch": 1.83, "learning_rate": 2.6423946452810758e-05, "loss": 0.0993, "step": 39200 }, { "epoch": 1.83, "learning_rate": 2.642347838297474e-05, "loss": 0.0649, "step": 39205 }, { "epoch": 1.83, "learning_rate": 2.642301031313872e-05, "loss": 0.1498, "step": 39210 }, { "epoch": 1.83, "learning_rate": 2.64225422433027e-05, "loss": 0.1484, "step": 39215 }, { "epoch": 1.83, "learning_rate": 2.642207417346668e-05, "loss": 0.1065, "step": 39220 }, { "epoch": 1.83, "learning_rate": 2.6421606103630664e-05, "loss": 0.1668, "step": 39225 }, { "epoch": 1.83, "learning_rate": 2.6421138033794644e-05, "loss": 0.4266, "step": 39230 }, { "epoch": 1.83, "learning_rate": 2.6420669963958623e-05, "loss": 0.218, "step": 39235 }, { "epoch": 1.83, "learning_rate": 2.6420201894122607e-05, "loss": 0.0613, "step": 39240 }, { "epoch": 1.83, "learning_rate": 2.6419733824286586e-05, "loss": 0.0745, "step": 39245 }, { "epoch": 1.83, "learning_rate": 2.6419265754450563e-05, "loss": 0.0922, "step": 39250 }, { "epoch": 1.83, "learning_rate": 2.6418797684614543e-05, "loss": 0.0721, "step": 39255 }, { "epoch": 1.83, "learning_rate": 2.6418329614778526e-05, "loss": 0.1169, "step": 39260 }, { "epoch": 1.83, "learning_rate": 2.6417861544942506e-05, "loss": 0.0791, "step": 39265 }, { "epoch": 1.83, "learning_rate": 2.6417393475106485e-05, "loss": 0.1684, "step": 39270 }, { "epoch": 1.83, "learning_rate": 2.6416925405270465e-05, "loss": 0.216, "step": 39275 }, { "epoch": 1.83, "learning_rate": 2.641645733543445e-05, "loss": 0.1538, "step": 39280 }, { "epoch": 1.83, "learning_rate": 2.6415989265598428e-05, "loss": 0.2589, "step": 39285 }, { "epoch": 1.83, "learning_rate": 2.6415521195762408e-05, "loss": 0.0595, "step": 39290 }, { "epoch": 1.83, "learning_rate": 2.641505312592639e-05, "loss": 0.0859, "step": 39295 }, { "epoch": 1.83, "learning_rate": 2.641458505609037e-05, "loss": 0.0406, "step": 39300 }, { "epoch": 1.83, "learning_rate": 2.641411698625435e-05, "loss": 0.0789, "step": 39305 }, { "epoch": 1.83, "learning_rate": 2.641364891641833e-05, "loss": 0.0925, "step": 39310 }, { "epoch": 1.83, "learning_rate": 2.641318084658231e-05, "loss": 0.0874, "step": 39315 }, { "epoch": 1.83, "learning_rate": 2.641271277674629e-05, "loss": 0.1556, "step": 39320 }, { "epoch": 1.83, "learning_rate": 2.641224470691027e-05, "loss": 0.2507, "step": 39325 }, { "epoch": 1.84, "learning_rate": 2.641177663707425e-05, "loss": 0.37, "step": 39330 }, { "epoch": 1.84, "learning_rate": 2.6411308567238233e-05, "loss": 0.1776, "step": 39335 }, { "epoch": 1.84, "learning_rate": 2.6410840497402213e-05, "loss": 0.028, "step": 39340 }, { "epoch": 1.84, "learning_rate": 2.6410372427566193e-05, "loss": 0.069, "step": 39345 }, { "epoch": 1.84, "learning_rate": 2.6409904357730173e-05, "loss": 0.032, "step": 39350 }, { "epoch": 1.84, "learning_rate": 2.6409436287894156e-05, "loss": 0.0548, "step": 39355 }, { "epoch": 1.84, "learning_rate": 2.6408968218058136e-05, "loss": 0.1884, "step": 39360 }, { "epoch": 1.84, "learning_rate": 2.6408500148222116e-05, "loss": 0.1181, "step": 39365 }, { "epoch": 1.84, "learning_rate": 2.64080320783861e-05, "loss": 0.182, "step": 39370 }, { "epoch": 1.84, "learning_rate": 2.640756400855008e-05, "loss": 0.1324, "step": 39375 }, { "epoch": 1.84, "learning_rate": 2.6407095938714055e-05, "loss": 0.2175, "step": 39380 }, { "epoch": 1.84, "learning_rate": 2.6406627868878035e-05, "loss": 0.2448, "step": 39385 }, { "epoch": 1.84, "learning_rate": 2.6406159799042018e-05, "loss": 0.0444, "step": 39390 }, { "epoch": 1.84, "learning_rate": 2.6405691729205998e-05, "loss": 0.0595, "step": 39395 }, { "epoch": 1.84, "learning_rate": 2.6405223659369978e-05, "loss": 0.0644, "step": 39400 }, { "epoch": 1.84, "learning_rate": 2.6404755589533957e-05, "loss": 0.0736, "step": 39405 }, { "epoch": 1.84, "learning_rate": 2.640428751969794e-05, "loss": 0.0735, "step": 39410 }, { "epoch": 1.84, "learning_rate": 2.640381944986192e-05, "loss": 0.0852, "step": 39415 }, { "epoch": 1.84, "learning_rate": 2.64033513800259e-05, "loss": 0.1139, "step": 39420 }, { "epoch": 1.84, "learning_rate": 2.6402883310189884e-05, "loss": 0.1785, "step": 39425 }, { "epoch": 1.84, "learning_rate": 2.6402415240353863e-05, "loss": 0.268, "step": 39430 }, { "epoch": 1.84, "learning_rate": 2.6401947170517843e-05, "loss": 0.1806, "step": 39435 }, { "epoch": 1.84, "learning_rate": 2.640147910068182e-05, "loss": 0.03, "step": 39440 }, { "epoch": 1.84, "learning_rate": 2.6401011030845803e-05, "loss": 0.0586, "step": 39445 }, { "epoch": 1.84, "learning_rate": 2.6400542961009783e-05, "loss": 0.0785, "step": 39450 }, { "epoch": 1.84, "learning_rate": 2.6400074891173762e-05, "loss": 0.0861, "step": 39455 }, { "epoch": 1.84, "learning_rate": 2.6399606821337742e-05, "loss": 0.1614, "step": 39460 }, { "epoch": 1.84, "learning_rate": 2.6399138751501725e-05, "loss": 0.1647, "step": 39465 }, { "epoch": 1.84, "learning_rate": 2.6398670681665705e-05, "loss": 0.1461, "step": 39470 }, { "epoch": 1.84, "learning_rate": 2.6398202611829685e-05, "loss": 0.1881, "step": 39475 }, { "epoch": 1.84, "learning_rate": 2.6397734541993668e-05, "loss": 0.3431, "step": 39480 }, { "epoch": 1.84, "learning_rate": 2.6397266472157648e-05, "loss": 0.2543, "step": 39485 }, { "epoch": 1.84, "learning_rate": 2.6396798402321628e-05, "loss": 0.05, "step": 39490 }, { "epoch": 1.84, "learning_rate": 2.6396330332485608e-05, "loss": 0.0729, "step": 39495 }, { "epoch": 1.84, "learning_rate": 2.639586226264959e-05, "loss": 0.0622, "step": 39500 }, { "epoch": 1.84, "learning_rate": 2.6395394192813567e-05, "loss": 0.1133, "step": 39505 }, { "epoch": 1.84, "learning_rate": 2.6394926122977547e-05, "loss": 0.1117, "step": 39510 }, { "epoch": 1.84, "learning_rate": 2.6394458053141527e-05, "loss": 0.0936, "step": 39515 }, { "epoch": 1.84, "learning_rate": 2.639398998330551e-05, "loss": 0.1008, "step": 39520 }, { "epoch": 1.84, "learning_rate": 2.639352191346949e-05, "loss": 0.212, "step": 39525 }, { "epoch": 1.84, "learning_rate": 2.639305384363347e-05, "loss": 0.4356, "step": 39530 }, { "epoch": 1.84, "learning_rate": 2.639258577379745e-05, "loss": 0.2185, "step": 39535 }, { "epoch": 1.84, "learning_rate": 2.6392117703961433e-05, "loss": 0.0723, "step": 39540 }, { "epoch": 1.85, "learning_rate": 2.6391649634125413e-05, "loss": 0.0484, "step": 39545 }, { "epoch": 1.85, "learning_rate": 2.6391181564289392e-05, "loss": 0.0843, "step": 39550 }, { "epoch": 1.85, "learning_rate": 2.6390713494453376e-05, "loss": 0.0646, "step": 39555 }, { "epoch": 1.85, "learning_rate": 2.6390245424617356e-05, "loss": 0.0454, "step": 39560 }, { "epoch": 1.85, "learning_rate": 2.6389777354781335e-05, "loss": 0.1635, "step": 39565 }, { "epoch": 1.85, "learning_rate": 2.6389309284945312e-05, "loss": 0.1328, "step": 39570 }, { "epoch": 1.85, "learning_rate": 2.6388841215109295e-05, "loss": 0.1878, "step": 39575 }, { "epoch": 1.85, "learning_rate": 2.6388373145273275e-05, "loss": 0.223, "step": 39580 }, { "epoch": 1.85, "learning_rate": 2.6387905075437255e-05, "loss": 0.1876, "step": 39585 }, { "epoch": 1.85, "learning_rate": 2.6387437005601234e-05, "loss": 0.0429, "step": 39590 }, { "epoch": 1.85, "learning_rate": 2.6386968935765218e-05, "loss": 0.0444, "step": 39595 }, { "epoch": 1.85, "learning_rate": 2.6386500865929197e-05, "loss": 0.0619, "step": 39600 }, { "epoch": 1.85, "learning_rate": 2.6386032796093177e-05, "loss": 0.1302, "step": 39605 }, { "epoch": 1.85, "learning_rate": 2.638556472625716e-05, "loss": 0.1599, "step": 39610 }, { "epoch": 1.85, "learning_rate": 2.638509665642114e-05, "loss": 0.0831, "step": 39615 }, { "epoch": 1.85, "learning_rate": 2.638462858658512e-05, "loss": 0.1128, "step": 39620 }, { "epoch": 1.85, "learning_rate": 2.63841605167491e-05, "loss": 0.1849, "step": 39625 }, { "epoch": 1.85, "learning_rate": 2.638369244691308e-05, "loss": 0.4699, "step": 39630 }, { "epoch": 1.85, "learning_rate": 2.638322437707706e-05, "loss": 0.2351, "step": 39635 }, { "epoch": 1.85, "learning_rate": 2.638275630724104e-05, "loss": 0.0361, "step": 39640 }, { "epoch": 1.85, "learning_rate": 2.638228823740502e-05, "loss": 0.0308, "step": 39645 }, { "epoch": 1.85, "learning_rate": 2.6381820167569002e-05, "loss": 0.0579, "step": 39650 }, { "epoch": 1.85, "learning_rate": 2.6381352097732982e-05, "loss": 0.0801, "step": 39655 }, { "epoch": 1.85, "learning_rate": 2.6380884027896962e-05, "loss": 0.1749, "step": 39660 }, { "epoch": 1.85, "learning_rate": 2.6380415958060945e-05, "loss": 0.1459, "step": 39665 }, { "epoch": 1.85, "learning_rate": 2.6379947888224925e-05, "loss": 0.1231, "step": 39670 }, { "epoch": 1.85, "learning_rate": 2.6379479818388905e-05, "loss": 0.1966, "step": 39675 }, { "epoch": 1.85, "learning_rate": 2.6379011748552885e-05, "loss": 0.2489, "step": 39680 }, { "epoch": 1.85, "learning_rate": 2.6378543678716868e-05, "loss": 0.3014, "step": 39685 }, { "epoch": 1.85, "learning_rate": 2.6378075608880848e-05, "loss": 0.0527, "step": 39690 }, { "epoch": 1.85, "learning_rate": 2.6377607539044824e-05, "loss": 0.0413, "step": 39695 }, { "epoch": 1.85, "learning_rate": 2.6377139469208804e-05, "loss": 0.078, "step": 39700 }, { "epoch": 1.85, "learning_rate": 2.6376671399372787e-05, "loss": 0.0978, "step": 39705 }, { "epoch": 1.85, "learning_rate": 2.6376203329536767e-05, "loss": 0.1308, "step": 39710 }, { "epoch": 1.85, "learning_rate": 2.6375735259700747e-05, "loss": 0.1082, "step": 39715 }, { "epoch": 1.85, "learning_rate": 2.6375267189864727e-05, "loss": 0.1716, "step": 39720 }, { "epoch": 1.85, "learning_rate": 2.637479912002871e-05, "loss": 0.2019, "step": 39725 }, { "epoch": 1.85, "learning_rate": 2.637433105019269e-05, "loss": 0.3588, "step": 39730 }, { "epoch": 1.85, "learning_rate": 2.637386298035667e-05, "loss": 0.2441, "step": 39735 }, { "epoch": 1.85, "learning_rate": 2.6373394910520653e-05, "loss": 0.0206, "step": 39740 }, { "epoch": 1.85, "learning_rate": 2.6372926840684632e-05, "loss": 0.1103, "step": 39745 }, { "epoch": 1.85, "learning_rate": 2.6372458770848612e-05, "loss": 0.0423, "step": 39750 }, { "epoch": 1.86, "learning_rate": 2.6371990701012592e-05, "loss": 0.0699, "step": 39755 }, { "epoch": 1.86, "learning_rate": 2.6371522631176572e-05, "loss": 0.1116, "step": 39760 }, { "epoch": 1.86, "learning_rate": 2.6371054561340552e-05, "loss": 0.1634, "step": 39765 }, { "epoch": 1.86, "learning_rate": 2.637058649150453e-05, "loss": 0.122, "step": 39770 }, { "epoch": 1.86, "learning_rate": 2.637011842166851e-05, "loss": 0.1931, "step": 39775 }, { "epoch": 1.86, "learning_rate": 2.6369650351832495e-05, "loss": 0.3544, "step": 39780 }, { "epoch": 1.86, "learning_rate": 2.6369182281996474e-05, "loss": 0.2152, "step": 39785 }, { "epoch": 1.86, "learning_rate": 2.6368714212160454e-05, "loss": 0.0587, "step": 39790 }, { "epoch": 1.86, "learning_rate": 2.6368246142324437e-05, "loss": 0.0413, "step": 39795 }, { "epoch": 1.86, "learning_rate": 2.6367778072488417e-05, "loss": 0.0856, "step": 39800 }, { "epoch": 1.86, "learning_rate": 2.6367310002652397e-05, "loss": 0.0619, "step": 39805 }, { "epoch": 1.86, "learning_rate": 2.6366841932816377e-05, "loss": 0.0829, "step": 39810 }, { "epoch": 1.86, "learning_rate": 2.636637386298036e-05, "loss": 0.1344, "step": 39815 }, { "epoch": 1.86, "learning_rate": 2.6365905793144337e-05, "loss": 0.1833, "step": 39820 }, { "epoch": 1.86, "learning_rate": 2.6365437723308316e-05, "loss": 0.267, "step": 39825 }, { "epoch": 1.86, "learning_rate": 2.6364969653472296e-05, "loss": 0.2902, "step": 39830 }, { "epoch": 1.86, "learning_rate": 2.636450158363628e-05, "loss": 0.2734, "step": 39835 }, { "epoch": 1.86, "learning_rate": 2.636403351380026e-05, "loss": 0.0288, "step": 39840 }, { "epoch": 1.86, "learning_rate": 2.636356544396424e-05, "loss": 0.0335, "step": 39845 }, { "epoch": 1.86, "learning_rate": 2.6363097374128222e-05, "loss": 0.0425, "step": 39850 }, { "epoch": 1.86, "learning_rate": 2.6362629304292202e-05, "loss": 0.0949, "step": 39855 }, { "epoch": 1.86, "learning_rate": 2.6362161234456182e-05, "loss": 0.1479, "step": 39860 }, { "epoch": 1.86, "learning_rate": 2.636169316462016e-05, "loss": 0.1293, "step": 39865 }, { "epoch": 1.86, "learning_rate": 2.6361225094784145e-05, "loss": 0.1361, "step": 39870 }, { "epoch": 1.86, "learning_rate": 2.6360757024948125e-05, "loss": 0.1473, "step": 39875 }, { "epoch": 1.86, "learning_rate": 2.6360288955112105e-05, "loss": 0.3929, "step": 39880 }, { "epoch": 1.86, "learning_rate": 2.635982088527608e-05, "loss": 0.1703, "step": 39885 }, { "epoch": 1.86, "learning_rate": 2.6359352815440064e-05, "loss": 0.0499, "step": 39890 }, { "epoch": 1.86, "learning_rate": 2.6358884745604044e-05, "loss": 0.0729, "step": 39895 }, { "epoch": 1.86, "learning_rate": 2.6358416675768024e-05, "loss": 0.0829, "step": 39900 }, { "epoch": 1.86, "learning_rate": 2.6357948605932004e-05, "loss": 0.0523, "step": 39905 }, { "epoch": 1.86, "learning_rate": 2.6357480536095987e-05, "loss": 0.132, "step": 39910 }, { "epoch": 1.86, "learning_rate": 2.6357012466259967e-05, "loss": 0.0819, "step": 39915 }, { "epoch": 1.86, "learning_rate": 2.6356544396423946e-05, "loss": 0.2226, "step": 39920 }, { "epoch": 1.86, "learning_rate": 2.635607632658793e-05, "loss": 0.138, "step": 39925 }, { "epoch": 1.86, "learning_rate": 2.635560825675191e-05, "loss": 0.2037, "step": 39930 }, { "epoch": 1.86, "learning_rate": 2.635514018691589e-05, "loss": 0.1822, "step": 39935 }, { "epoch": 1.86, "learning_rate": 2.635467211707987e-05, "loss": 0.0095, "step": 39940 }, { "epoch": 1.86, "learning_rate": 2.635420404724385e-05, "loss": 0.0325, "step": 39945 }, { "epoch": 1.86, "learning_rate": 2.635373597740783e-05, "loss": 0.0915, "step": 39950 }, { "epoch": 1.86, "learning_rate": 2.635326790757181e-05, "loss": 0.0729, "step": 39955 }, { "epoch": 1.86, "learning_rate": 2.635279983773579e-05, "loss": 0.1122, "step": 39960 }, { "epoch": 1.86, "learning_rate": 2.635233176789977e-05, "loss": 0.1503, "step": 39965 }, { "epoch": 1.87, "learning_rate": 2.635186369806375e-05, "loss": 0.2107, "step": 39970 }, { "epoch": 1.87, "learning_rate": 2.635139562822773e-05, "loss": 0.1089, "step": 39975 }, { "epoch": 1.87, "learning_rate": 2.6350927558391714e-05, "loss": 0.3286, "step": 39980 }, { "epoch": 1.87, "learning_rate": 2.6350459488555694e-05, "loss": 0.2829, "step": 39985 }, { "epoch": 1.87, "learning_rate": 2.6349991418719674e-05, "loss": 0.0567, "step": 39990 }, { "epoch": 1.87, "learning_rate": 2.6349523348883654e-05, "loss": 0.0397, "step": 39995 }, { "epoch": 1.87, "learning_rate": 2.6349055279047637e-05, "loss": 0.0723, "step": 40000 }, { "epoch": 1.87, "learning_rate": 2.6348587209211617e-05, "loss": 0.1345, "step": 40005 }, { "epoch": 1.87, "learning_rate": 2.6348119139375593e-05, "loss": 0.0691, "step": 40010 }, { "epoch": 1.87, "learning_rate": 2.6347651069539573e-05, "loss": 0.1061, "step": 40015 }, { "epoch": 1.87, "learning_rate": 2.6347182999703556e-05, "loss": 0.081, "step": 40020 }, { "epoch": 1.87, "learning_rate": 2.6346714929867536e-05, "loss": 0.2578, "step": 40025 }, { "epoch": 1.87, "learning_rate": 2.6346246860031516e-05, "loss": 0.222, "step": 40030 }, { "epoch": 1.87, "learning_rate": 2.63457787901955e-05, "loss": 0.2264, "step": 40035 }, { "epoch": 1.87, "learning_rate": 2.634531072035948e-05, "loss": 0.0269, "step": 40040 }, { "epoch": 1.87, "learning_rate": 2.634484265052346e-05, "loss": 0.0937, "step": 40045 }, { "epoch": 1.87, "learning_rate": 2.634437458068744e-05, "loss": 0.0581, "step": 40050 }, { "epoch": 1.87, "learning_rate": 2.6343906510851422e-05, "loss": 0.0891, "step": 40055 }, { "epoch": 1.87, "learning_rate": 2.63434384410154e-05, "loss": 0.0834, "step": 40060 }, { "epoch": 1.87, "learning_rate": 2.634297037117938e-05, "loss": 0.0933, "step": 40065 }, { "epoch": 1.87, "learning_rate": 2.634250230134336e-05, "loss": 0.0986, "step": 40070 }, { "epoch": 1.87, "learning_rate": 2.634203423150734e-05, "loss": 0.2179, "step": 40075 }, { "epoch": 1.87, "learning_rate": 2.634156616167132e-05, "loss": 0.4146, "step": 40080 }, { "epoch": 1.87, "learning_rate": 2.63410980918353e-05, "loss": 0.2141, "step": 40085 }, { "epoch": 1.87, "learning_rate": 2.6340630021999284e-05, "loss": 0.0474, "step": 40090 }, { "epoch": 1.87, "learning_rate": 2.6340161952163264e-05, "loss": 0.0566, "step": 40095 }, { "epoch": 1.87, "learning_rate": 2.6339693882327244e-05, "loss": 0.0735, "step": 40100 }, { "epoch": 1.87, "learning_rate": 2.6339225812491223e-05, "loss": 0.126, "step": 40105 }, { "epoch": 1.87, "learning_rate": 2.6338757742655207e-05, "loss": 0.1018, "step": 40110 }, { "epoch": 1.87, "learning_rate": 2.6338289672819186e-05, "loss": 0.1289, "step": 40115 }, { "epoch": 1.87, "learning_rate": 2.6337821602983166e-05, "loss": 0.1872, "step": 40120 }, { "epoch": 1.87, "learning_rate": 2.6337353533147146e-05, "loss": 0.2074, "step": 40125 }, { "epoch": 1.87, "learning_rate": 2.633688546331113e-05, "loss": 0.3795, "step": 40130 }, { "epoch": 1.87, "learning_rate": 2.6336417393475106e-05, "loss": 0.3749, "step": 40135 }, { "epoch": 1.87, "learning_rate": 2.6335949323639086e-05, "loss": 0.0988, "step": 40140 }, { "epoch": 1.87, "learning_rate": 2.6335481253803065e-05, "loss": 0.0861, "step": 40145 }, { "epoch": 1.87, "learning_rate": 2.633501318396705e-05, "loss": 0.0576, "step": 40150 }, { "epoch": 1.87, "learning_rate": 2.633454511413103e-05, "loss": 0.0771, "step": 40155 }, { "epoch": 1.87, "learning_rate": 2.6334077044295008e-05, "loss": 0.1084, "step": 40160 }, { "epoch": 1.87, "learning_rate": 2.633360897445899e-05, "loss": 0.1072, "step": 40165 }, { "epoch": 1.87, "learning_rate": 2.633314090462297e-05, "loss": 0.157, "step": 40170 }, { "epoch": 1.87, "learning_rate": 2.633267283478695e-05, "loss": 0.2229, "step": 40175 }, { "epoch": 1.87, "learning_rate": 2.633220476495093e-05, "loss": 0.2369, "step": 40180 }, { "epoch": 1.88, "learning_rate": 2.6331736695114914e-05, "loss": 0.321, "step": 40185 }, { "epoch": 1.88, "learning_rate": 2.6331268625278894e-05, "loss": 0.0187, "step": 40190 }, { "epoch": 1.88, "learning_rate": 2.6330800555442874e-05, "loss": 0.0409, "step": 40195 }, { "epoch": 1.88, "learning_rate": 2.633033248560685e-05, "loss": 0.0531, "step": 40200 }, { "epoch": 1.88, "learning_rate": 2.6329864415770833e-05, "loss": 0.1098, "step": 40205 }, { "epoch": 1.88, "learning_rate": 2.6329396345934813e-05, "loss": 0.1696, "step": 40210 }, { "epoch": 1.88, "learning_rate": 2.6328928276098793e-05, "loss": 0.1012, "step": 40215 }, { "epoch": 1.88, "learning_rate": 2.6328460206262776e-05, "loss": 0.0892, "step": 40220 }, { "epoch": 1.88, "learning_rate": 2.6327992136426756e-05, "loss": 0.1553, "step": 40225 }, { "epoch": 1.88, "learning_rate": 2.6327524066590736e-05, "loss": 0.2304, "step": 40230 }, { "epoch": 1.88, "learning_rate": 2.6327055996754716e-05, "loss": 0.1584, "step": 40235 }, { "epoch": 1.88, "learning_rate": 2.63265879269187e-05, "loss": 0.0443, "step": 40240 }, { "epoch": 1.88, "learning_rate": 2.632611985708268e-05, "loss": 0.0389, "step": 40245 }, { "epoch": 1.88, "learning_rate": 2.632565178724666e-05, "loss": 0.0536, "step": 40250 }, { "epoch": 1.88, "learning_rate": 2.6325183717410638e-05, "loss": 0.1238, "step": 40255 }, { "epoch": 1.88, "learning_rate": 2.632471564757462e-05, "loss": 0.1084, "step": 40260 }, { "epoch": 1.88, "learning_rate": 2.6324247577738598e-05, "loss": 0.1341, "step": 40265 }, { "epoch": 1.88, "learning_rate": 2.6323779507902578e-05, "loss": 0.1369, "step": 40270 }, { "epoch": 1.88, "learning_rate": 2.632331143806656e-05, "loss": 0.1298, "step": 40275 }, { "epoch": 1.88, "learning_rate": 2.632284336823054e-05, "loss": 0.2002, "step": 40280 }, { "epoch": 1.88, "learning_rate": 2.632237529839452e-05, "loss": 0.209, "step": 40285 }, { "epoch": 1.88, "learning_rate": 2.63219072285585e-05, "loss": 0.036, "step": 40290 }, { "epoch": 1.88, "learning_rate": 2.6321439158722484e-05, "loss": 0.0855, "step": 40295 }, { "epoch": 1.88, "learning_rate": 2.6320971088886463e-05, "loss": 0.0532, "step": 40300 }, { "epoch": 1.88, "learning_rate": 2.6320503019050443e-05, "loss": 0.1029, "step": 40305 }, { "epoch": 1.88, "learning_rate": 2.6320034949214423e-05, "loss": 0.0991, "step": 40310 }, { "epoch": 1.88, "learning_rate": 2.6319566879378406e-05, "loss": 0.1283, "step": 40315 }, { "epoch": 1.88, "learning_rate": 2.6319098809542386e-05, "loss": 0.1572, "step": 40320 }, { "epoch": 1.88, "learning_rate": 2.6318630739706362e-05, "loss": 0.1945, "step": 40325 }, { "epoch": 1.88, "learning_rate": 2.6318162669870342e-05, "loss": 0.3183, "step": 40330 }, { "epoch": 1.88, "learning_rate": 2.6317694600034326e-05, "loss": 0.4122, "step": 40335 }, { "epoch": 1.88, "learning_rate": 2.6317226530198305e-05, "loss": 0.0527, "step": 40340 }, { "epoch": 1.88, "learning_rate": 2.6316758460362285e-05, "loss": 0.0604, "step": 40345 }, { "epoch": 1.88, "learning_rate": 2.631629039052627e-05, "loss": 0.0646, "step": 40350 }, { "epoch": 1.88, "learning_rate": 2.6315822320690248e-05, "loss": 0.1079, "step": 40355 }, { "epoch": 1.88, "learning_rate": 2.6315354250854228e-05, "loss": 0.1001, "step": 40360 }, { "epoch": 1.88, "learning_rate": 2.6314886181018208e-05, "loss": 0.1328, "step": 40365 }, { "epoch": 1.88, "learning_rate": 2.631441811118219e-05, "loss": 0.155, "step": 40370 }, { "epoch": 1.88, "learning_rate": 2.631395004134617e-05, "loss": 0.1359, "step": 40375 }, { "epoch": 1.88, "learning_rate": 2.631348197151015e-05, "loss": 0.2913, "step": 40380 }, { "epoch": 1.88, "learning_rate": 2.631301390167413e-05, "loss": 0.2886, "step": 40385 }, { "epoch": 1.88, "learning_rate": 2.631254583183811e-05, "loss": 0.062, "step": 40390 }, { "epoch": 1.88, "learning_rate": 2.631207776200209e-05, "loss": 0.0582, "step": 40395 }, { "epoch": 1.89, "learning_rate": 2.631160969216607e-05, "loss": 0.0464, "step": 40400 }, { "epoch": 1.89, "learning_rate": 2.6311141622330053e-05, "loss": 0.1078, "step": 40405 }, { "epoch": 1.89, "learning_rate": 2.6310673552494033e-05, "loss": 0.0807, "step": 40410 }, { "epoch": 1.89, "learning_rate": 2.6310205482658013e-05, "loss": 0.1149, "step": 40415 }, { "epoch": 1.89, "learning_rate": 2.6309737412821993e-05, "loss": 0.1297, "step": 40420 }, { "epoch": 1.89, "learning_rate": 2.6309269342985976e-05, "loss": 0.1764, "step": 40425 }, { "epoch": 1.89, "learning_rate": 2.6308801273149956e-05, "loss": 0.1905, "step": 40430 }, { "epoch": 1.89, "learning_rate": 2.6308333203313935e-05, "loss": 0.2088, "step": 40435 }, { "epoch": 1.89, "learning_rate": 2.6307865133477915e-05, "loss": 0.0662, "step": 40440 }, { "epoch": 1.89, "learning_rate": 2.63073970636419e-05, "loss": 0.0339, "step": 40445 }, { "epoch": 1.89, "learning_rate": 2.6306928993805878e-05, "loss": 0.0661, "step": 40450 }, { "epoch": 1.89, "learning_rate": 2.6306460923969855e-05, "loss": 0.1632, "step": 40455 }, { "epoch": 1.89, "learning_rate": 2.6305992854133838e-05, "loss": 0.1307, "step": 40460 }, { "epoch": 1.89, "learning_rate": 2.6305524784297818e-05, "loss": 0.1238, "step": 40465 }, { "epoch": 1.89, "learning_rate": 2.6305056714461798e-05, "loss": 0.13, "step": 40470 }, { "epoch": 1.89, "learning_rate": 2.6304588644625777e-05, "loss": 0.2832, "step": 40475 }, { "epoch": 1.89, "learning_rate": 2.630412057478976e-05, "loss": 0.3345, "step": 40480 }, { "epoch": 1.89, "learning_rate": 2.630365250495374e-05, "loss": 0.2461, "step": 40485 }, { "epoch": 1.89, "learning_rate": 2.630318443511772e-05, "loss": 0.0824, "step": 40490 }, { "epoch": 1.89, "learning_rate": 2.63027163652817e-05, "loss": 0.0598, "step": 40495 }, { "epoch": 1.89, "learning_rate": 2.6302248295445683e-05, "loss": 0.0739, "step": 40500 }, { "epoch": 1.89, "learning_rate": 2.6301780225609663e-05, "loss": 0.0547, "step": 40505 }, { "epoch": 1.89, "learning_rate": 2.6301312155773643e-05, "loss": 0.1084, "step": 40510 }, { "epoch": 1.89, "learning_rate": 2.630084408593762e-05, "loss": 0.0908, "step": 40515 }, { "epoch": 1.89, "learning_rate": 2.6300376016101602e-05, "loss": 0.1632, "step": 40520 }, { "epoch": 1.89, "learning_rate": 2.6299907946265582e-05, "loss": 0.1422, "step": 40525 }, { "epoch": 1.89, "learning_rate": 2.6299439876429562e-05, "loss": 0.5254, "step": 40530 }, { "epoch": 1.89, "learning_rate": 2.6298971806593545e-05, "loss": 0.2415, "step": 40535 }, { "epoch": 1.89, "learning_rate": 2.6298503736757525e-05, "loss": 0.023, "step": 40540 }, { "epoch": 1.89, "learning_rate": 2.6298035666921505e-05, "loss": 0.0293, "step": 40545 }, { "epoch": 1.89, "learning_rate": 2.6297567597085485e-05, "loss": 0.0752, "step": 40550 }, { "epoch": 1.89, "learning_rate": 2.6297099527249468e-05, "loss": 0.0975, "step": 40555 }, { "epoch": 1.89, "learning_rate": 2.6296631457413448e-05, "loss": 0.1287, "step": 40560 }, { "epoch": 1.89, "learning_rate": 2.6296163387577428e-05, "loss": 0.1435, "step": 40565 }, { "epoch": 1.89, "learning_rate": 2.6295695317741407e-05, "loss": 0.1247, "step": 40570 }, { "epoch": 1.89, "learning_rate": 2.629522724790539e-05, "loss": 0.1938, "step": 40575 }, { "epoch": 1.89, "learning_rate": 2.6294759178069367e-05, "loss": 0.2045, "step": 40580 }, { "epoch": 1.89, "learning_rate": 2.6294291108233347e-05, "loss": 0.3442, "step": 40585 }, { "epoch": 1.89, "learning_rate": 2.629382303839733e-05, "loss": 0.0306, "step": 40590 }, { "epoch": 1.89, "learning_rate": 2.629335496856131e-05, "loss": 0.0456, "step": 40595 }, { "epoch": 1.89, "learning_rate": 2.629288689872529e-05, "loss": 0.0721, "step": 40600 }, { "epoch": 1.89, "learning_rate": 2.629241882888927e-05, "loss": 0.1394, "step": 40605 }, { "epoch": 1.89, "learning_rate": 2.6291950759053253e-05, "loss": 0.1354, "step": 40610 }, { "epoch": 1.9, "learning_rate": 2.6291482689217233e-05, "loss": 0.1383, "step": 40615 }, { "epoch": 1.9, "learning_rate": 2.6291014619381212e-05, "loss": 0.0947, "step": 40620 }, { "epoch": 1.9, "learning_rate": 2.6290546549545192e-05, "loss": 0.2224, "step": 40625 }, { "epoch": 1.9, "learning_rate": 2.6290078479709175e-05, "loss": 0.2473, "step": 40630 }, { "epoch": 1.9, "learning_rate": 2.6289610409873155e-05, "loss": 0.2824, "step": 40635 }, { "epoch": 1.9, "learning_rate": 2.628914234003713e-05, "loss": 0.0375, "step": 40640 }, { "epoch": 1.9, "learning_rate": 2.6288674270201115e-05, "loss": 0.0308, "step": 40645 }, { "epoch": 1.9, "learning_rate": 2.6288206200365095e-05, "loss": 0.0811, "step": 40650 }, { "epoch": 1.9, "learning_rate": 2.6287738130529075e-05, "loss": 0.0656, "step": 40655 }, { "epoch": 1.9, "learning_rate": 2.6287270060693054e-05, "loss": 0.126, "step": 40660 }, { "epoch": 1.9, "learning_rate": 2.6286801990857038e-05, "loss": 0.104, "step": 40665 }, { "epoch": 1.9, "learning_rate": 2.6286333921021017e-05, "loss": 0.1067, "step": 40670 }, { "epoch": 1.9, "learning_rate": 2.6285865851184997e-05, "loss": 0.1963, "step": 40675 }, { "epoch": 1.9, "learning_rate": 2.6285397781348977e-05, "loss": 0.2602, "step": 40680 }, { "epoch": 1.9, "learning_rate": 2.628492971151296e-05, "loss": 0.2171, "step": 40685 }, { "epoch": 1.9, "learning_rate": 2.628446164167694e-05, "loss": 0.056, "step": 40690 }, { "epoch": 1.9, "learning_rate": 2.628399357184092e-05, "loss": 0.034, "step": 40695 }, { "epoch": 1.9, "learning_rate": 2.6283525502004903e-05, "loss": 0.0761, "step": 40700 }, { "epoch": 1.9, "learning_rate": 2.628305743216888e-05, "loss": 0.1378, "step": 40705 }, { "epoch": 1.9, "learning_rate": 2.628258936233286e-05, "loss": 0.1445, "step": 40710 }, { "epoch": 1.9, "learning_rate": 2.628212129249684e-05, "loss": 0.1106, "step": 40715 }, { "epoch": 1.9, "learning_rate": 2.6281653222660822e-05, "loss": 0.1074, "step": 40720 }, { "epoch": 1.9, "learning_rate": 2.6281185152824802e-05, "loss": 0.1821, "step": 40725 }, { "epoch": 1.9, "learning_rate": 2.6280717082988782e-05, "loss": 0.4791, "step": 40730 }, { "epoch": 1.9, "learning_rate": 2.6280249013152762e-05, "loss": 0.2698, "step": 40735 }, { "epoch": 1.9, "learning_rate": 2.6279780943316745e-05, "loss": 0.0433, "step": 40740 }, { "epoch": 1.9, "learning_rate": 2.6279312873480725e-05, "loss": 0.0372, "step": 40745 }, { "epoch": 1.9, "learning_rate": 2.6278844803644705e-05, "loss": 0.0911, "step": 40750 }, { "epoch": 1.9, "learning_rate": 2.6278376733808684e-05, "loss": 0.0713, "step": 40755 }, { "epoch": 1.9, "learning_rate": 2.6277908663972668e-05, "loss": 0.1668, "step": 40760 }, { "epoch": 1.9, "learning_rate": 2.6277440594136647e-05, "loss": 0.1463, "step": 40765 }, { "epoch": 1.9, "learning_rate": 2.6276972524300624e-05, "loss": 0.1883, "step": 40770 }, { "epoch": 1.9, "learning_rate": 2.6276504454464607e-05, "loss": 0.2241, "step": 40775 }, { "epoch": 1.9, "learning_rate": 2.6276036384628587e-05, "loss": 0.3225, "step": 40780 }, { "epoch": 1.9, "learning_rate": 2.6275568314792567e-05, "loss": 0.2225, "step": 40785 }, { "epoch": 1.9, "learning_rate": 2.6275100244956547e-05, "loss": 0.055, "step": 40790 }, { "epoch": 1.9, "learning_rate": 2.627463217512053e-05, "loss": 0.0597, "step": 40795 }, { "epoch": 1.9, "learning_rate": 2.627416410528451e-05, "loss": 0.0441, "step": 40800 }, { "epoch": 1.9, "learning_rate": 2.627369603544849e-05, "loss": 0.1143, "step": 40805 }, { "epoch": 1.9, "learning_rate": 2.627322796561247e-05, "loss": 0.0817, "step": 40810 }, { "epoch": 1.9, "learning_rate": 2.6272759895776452e-05, "loss": 0.1266, "step": 40815 }, { "epoch": 1.9, "learning_rate": 2.6272291825940432e-05, "loss": 0.1432, "step": 40820 }, { "epoch": 1.9, "learning_rate": 2.6271823756104412e-05, "loss": 0.2385, "step": 40825 }, { "epoch": 1.91, "learning_rate": 2.6271355686268392e-05, "loss": 0.3393, "step": 40830 }, { "epoch": 1.91, "learning_rate": 2.627088761643237e-05, "loss": 0.1935, "step": 40835 }, { "epoch": 1.91, "learning_rate": 2.627041954659635e-05, "loss": 0.0436, "step": 40840 }, { "epoch": 1.91, "learning_rate": 2.626995147676033e-05, "loss": 0.0911, "step": 40845 }, { "epoch": 1.91, "learning_rate": 2.6269483406924314e-05, "loss": 0.0434, "step": 40850 }, { "epoch": 1.91, "learning_rate": 2.6269015337088294e-05, "loss": 0.0843, "step": 40855 }, { "epoch": 1.91, "learning_rate": 2.6268547267252274e-05, "loss": 0.0878, "step": 40860 }, { "epoch": 1.91, "learning_rate": 2.6268079197416254e-05, "loss": 0.1078, "step": 40865 }, { "epoch": 1.91, "learning_rate": 2.6267611127580237e-05, "loss": 0.1954, "step": 40870 }, { "epoch": 1.91, "learning_rate": 2.6267143057744217e-05, "loss": 0.1516, "step": 40875 }, { "epoch": 1.91, "learning_rate": 2.6266674987908197e-05, "loss": 0.205, "step": 40880 }, { "epoch": 1.91, "learning_rate": 2.626620691807218e-05, "loss": 0.4242, "step": 40885 }, { "epoch": 1.91, "learning_rate": 2.626573884823616e-05, "loss": 0.0665, "step": 40890 }, { "epoch": 1.91, "learning_rate": 2.6265270778400136e-05, "loss": 0.0406, "step": 40895 }, { "epoch": 1.91, "learning_rate": 2.6264802708564116e-05, "loss": 0.1, "step": 40900 }, { "epoch": 1.91, "learning_rate": 2.62643346387281e-05, "loss": 0.0459, "step": 40905 }, { "epoch": 1.91, "learning_rate": 2.626386656889208e-05, "loss": 0.0963, "step": 40910 }, { "epoch": 1.91, "learning_rate": 2.626339849905606e-05, "loss": 0.1903, "step": 40915 }, { "epoch": 1.91, "learning_rate": 2.626293042922004e-05, "loss": 0.1762, "step": 40920 }, { "epoch": 1.91, "learning_rate": 2.6262462359384022e-05, "loss": 0.1476, "step": 40925 }, { "epoch": 1.91, "learning_rate": 2.6261994289548002e-05, "loss": 0.2332, "step": 40930 }, { "epoch": 1.91, "learning_rate": 2.626152621971198e-05, "loss": 0.2199, "step": 40935 }, { "epoch": 1.91, "learning_rate": 2.626105814987596e-05, "loss": 0.0099, "step": 40940 }, { "epoch": 1.91, "learning_rate": 2.6260590080039945e-05, "loss": 0.0458, "step": 40945 }, { "epoch": 1.91, "learning_rate": 2.6260122010203924e-05, "loss": 0.083, "step": 40950 }, { "epoch": 1.91, "learning_rate": 2.6259653940367904e-05, "loss": 0.0522, "step": 40955 }, { "epoch": 1.91, "learning_rate": 2.6259185870531884e-05, "loss": 0.0786, "step": 40960 }, { "epoch": 1.91, "learning_rate": 2.6258717800695864e-05, "loss": 0.0891, "step": 40965 }, { "epoch": 1.91, "learning_rate": 2.6258249730859844e-05, "loss": 0.1781, "step": 40970 }, { "epoch": 1.91, "learning_rate": 2.6257781661023823e-05, "loss": 0.2083, "step": 40975 }, { "epoch": 1.91, "learning_rate": 2.6257313591187807e-05, "loss": 0.3915, "step": 40980 }, { "epoch": 1.91, "learning_rate": 2.6256845521351787e-05, "loss": 0.331, "step": 40985 }, { "epoch": 1.91, "learning_rate": 2.6256377451515766e-05, "loss": 0.0316, "step": 40990 }, { "epoch": 1.91, "learning_rate": 2.6255909381679746e-05, "loss": 0.0784, "step": 40995 }, { "epoch": 1.91, "learning_rate": 2.625544131184373e-05, "loss": 0.0979, "step": 41000 }, { "epoch": 1.91, "learning_rate": 2.625497324200771e-05, "loss": 0.0366, "step": 41005 }, { "epoch": 1.91, "learning_rate": 2.625450517217169e-05, "loss": 0.1194, "step": 41010 }, { "epoch": 1.91, "learning_rate": 2.6254037102335672e-05, "loss": 0.1278, "step": 41015 }, { "epoch": 1.91, "learning_rate": 2.625356903249965e-05, "loss": 0.1545, "step": 41020 }, { "epoch": 1.91, "learning_rate": 2.625310096266363e-05, "loss": 0.2239, "step": 41025 }, { "epoch": 1.91, "learning_rate": 2.6252632892827608e-05, "loss": 0.2885, "step": 41030 }, { "epoch": 1.91, "learning_rate": 2.625216482299159e-05, "loss": 0.2535, "step": 41035 }, { "epoch": 1.91, "learning_rate": 2.625169675315557e-05, "loss": 0.0614, "step": 41040 }, { "epoch": 1.92, "learning_rate": 2.625122868331955e-05, "loss": 0.077, "step": 41045 }, { "epoch": 1.92, "learning_rate": 2.625076061348353e-05, "loss": 0.0631, "step": 41050 }, { "epoch": 1.92, "learning_rate": 2.6250292543647514e-05, "loss": 0.0667, "step": 41055 }, { "epoch": 1.92, "learning_rate": 2.6249824473811494e-05, "loss": 0.0974, "step": 41060 }, { "epoch": 1.92, "learning_rate": 2.6249356403975474e-05, "loss": 0.1243, "step": 41065 }, { "epoch": 1.92, "learning_rate": 2.6248888334139457e-05, "loss": 0.0933, "step": 41070 }, { "epoch": 1.92, "learning_rate": 2.6248420264303437e-05, "loss": 0.2899, "step": 41075 }, { "epoch": 1.92, "learning_rate": 2.6247952194467417e-05, "loss": 0.2763, "step": 41080 }, { "epoch": 1.92, "learning_rate": 2.6247484124631393e-05, "loss": 0.2076, "step": 41085 }, { "epoch": 1.92, "learning_rate": 2.6247016054795376e-05, "loss": 0.0124, "step": 41090 }, { "epoch": 1.92, "learning_rate": 2.6246547984959356e-05, "loss": 0.0327, "step": 41095 }, { "epoch": 1.92, "learning_rate": 2.6246079915123336e-05, "loss": 0.1161, "step": 41100 }, { "epoch": 1.92, "learning_rate": 2.6245611845287316e-05, "loss": 0.07, "step": 41105 }, { "epoch": 1.92, "learning_rate": 2.62451437754513e-05, "loss": 0.0587, "step": 41110 }, { "epoch": 1.92, "learning_rate": 2.624467570561528e-05, "loss": 0.1299, "step": 41115 }, { "epoch": 1.92, "learning_rate": 2.624420763577926e-05, "loss": 0.1363, "step": 41120 }, { "epoch": 1.92, "learning_rate": 2.624373956594324e-05, "loss": 0.1922, "step": 41125 }, { "epoch": 1.92, "learning_rate": 2.624327149610722e-05, "loss": 0.271, "step": 41130 }, { "epoch": 1.92, "learning_rate": 2.62428034262712e-05, "loss": 0.2567, "step": 41135 }, { "epoch": 1.92, "learning_rate": 2.624233535643518e-05, "loss": 0.0299, "step": 41140 }, { "epoch": 1.92, "learning_rate": 2.624186728659916e-05, "loss": 0.0921, "step": 41145 }, { "epoch": 1.92, "learning_rate": 2.624139921676314e-05, "loss": 0.0731, "step": 41150 }, { "epoch": 1.92, "learning_rate": 2.624093114692712e-05, "loss": 0.0317, "step": 41155 }, { "epoch": 1.92, "learning_rate": 2.62404630770911e-05, "loss": 0.1856, "step": 41160 }, { "epoch": 1.92, "learning_rate": 2.6239995007255084e-05, "loss": 0.1505, "step": 41165 }, { "epoch": 1.92, "learning_rate": 2.6239526937419063e-05, "loss": 0.0735, "step": 41170 }, { "epoch": 1.92, "learning_rate": 2.6239058867583043e-05, "loss": 0.1157, "step": 41175 }, { "epoch": 1.92, "learning_rate": 2.6238590797747023e-05, "loss": 0.2787, "step": 41180 }, { "epoch": 1.92, "learning_rate": 2.6238122727911006e-05, "loss": 0.156, "step": 41185 }, { "epoch": 1.92, "learning_rate": 2.6237654658074986e-05, "loss": 0.0521, "step": 41190 }, { "epoch": 1.92, "learning_rate": 2.6237186588238966e-05, "loss": 0.038, "step": 41195 }, { "epoch": 1.92, "learning_rate": 2.623671851840295e-05, "loss": 0.1511, "step": 41200 }, { "epoch": 1.92, "learning_rate": 2.623625044856693e-05, "loss": 0.0708, "step": 41205 }, { "epoch": 1.92, "learning_rate": 2.6235782378730905e-05, "loss": 0.0415, "step": 41210 }, { "epoch": 1.92, "learning_rate": 2.6235314308894885e-05, "loss": 0.1097, "step": 41215 }, { "epoch": 1.92, "learning_rate": 2.623484623905887e-05, "loss": 0.1184, "step": 41220 }, { "epoch": 1.92, "learning_rate": 2.6234378169222848e-05, "loss": 0.2045, "step": 41225 }, { "epoch": 1.92, "learning_rate": 2.6233910099386828e-05, "loss": 0.7751, "step": 41230 }, { "epoch": 1.92, "learning_rate": 2.6233442029550808e-05, "loss": 0.3461, "step": 41235 }, { "epoch": 1.92, "learning_rate": 2.623297395971479e-05, "loss": 0.0319, "step": 41240 }, { "epoch": 1.92, "learning_rate": 2.623250588987877e-05, "loss": 0.0489, "step": 41245 }, { "epoch": 1.92, "learning_rate": 2.623203782004275e-05, "loss": 0.0812, "step": 41250 }, { "epoch": 1.93, "learning_rate": 2.6231569750206734e-05, "loss": 0.1004, "step": 41255 }, { "epoch": 1.93, "learning_rate": 2.6231101680370714e-05, "loss": 0.111, "step": 41260 }, { "epoch": 1.93, "learning_rate": 2.6230633610534694e-05, "loss": 0.096, "step": 41265 }, { "epoch": 1.93, "learning_rate": 2.6230165540698673e-05, "loss": 0.1388, "step": 41270 }, { "epoch": 1.93, "learning_rate": 2.6229697470862653e-05, "loss": 0.1167, "step": 41275 }, { "epoch": 1.93, "learning_rate": 2.6229229401026633e-05, "loss": 0.3065, "step": 41280 }, { "epoch": 1.93, "learning_rate": 2.6228761331190613e-05, "loss": 0.1655, "step": 41285 }, { "epoch": 1.93, "learning_rate": 2.6228293261354593e-05, "loss": 0.0553, "step": 41290 }, { "epoch": 1.93, "learning_rate": 2.6227825191518576e-05, "loss": 0.0778, "step": 41295 }, { "epoch": 1.93, "learning_rate": 2.6227357121682556e-05, "loss": 0.0503, "step": 41300 }, { "epoch": 1.93, "learning_rate": 2.6226889051846536e-05, "loss": 0.0601, "step": 41305 }, { "epoch": 1.93, "learning_rate": 2.6226420982010515e-05, "loss": 0.123, "step": 41310 }, { "epoch": 1.93, "learning_rate": 2.62259529121745e-05, "loss": 0.2035, "step": 41315 }, { "epoch": 1.93, "learning_rate": 2.622548484233848e-05, "loss": 0.1594, "step": 41320 }, { "epoch": 1.93, "learning_rate": 2.6225016772502458e-05, "loss": 0.2528, "step": 41325 }, { "epoch": 1.93, "learning_rate": 2.622454870266644e-05, "loss": 0.5483, "step": 41330 }, { "epoch": 1.93, "learning_rate": 2.6224080632830418e-05, "loss": 0.2355, "step": 41335 }, { "epoch": 1.93, "learning_rate": 2.6223612562994398e-05, "loss": 0.0509, "step": 41340 }, { "epoch": 1.93, "learning_rate": 2.6223144493158377e-05, "loss": 0.0884, "step": 41345 }, { "epoch": 1.93, "learning_rate": 2.622267642332236e-05, "loss": 0.0321, "step": 41350 }, { "epoch": 1.93, "learning_rate": 2.622220835348634e-05, "loss": 0.1214, "step": 41355 }, { "epoch": 1.93, "learning_rate": 2.622174028365032e-05, "loss": 0.0687, "step": 41360 }, { "epoch": 1.93, "learning_rate": 2.62212722138143e-05, "loss": 0.0954, "step": 41365 }, { "epoch": 1.93, "learning_rate": 2.6220804143978283e-05, "loss": 0.152, "step": 41370 }, { "epoch": 1.93, "learning_rate": 2.6220336074142263e-05, "loss": 0.1892, "step": 41375 }, { "epoch": 1.93, "learning_rate": 2.6219868004306243e-05, "loss": 0.3289, "step": 41380 }, { "epoch": 1.93, "learning_rate": 2.6219399934470226e-05, "loss": 0.2048, "step": 41385 }, { "epoch": 1.93, "learning_rate": 2.6218931864634206e-05, "loss": 0.0957, "step": 41390 }, { "epoch": 1.93, "learning_rate": 2.6218463794798186e-05, "loss": 0.0559, "step": 41395 }, { "epoch": 1.93, "learning_rate": 2.6217995724962162e-05, "loss": 0.068, "step": 41400 }, { "epoch": 1.93, "learning_rate": 2.6217527655126145e-05, "loss": 0.1379, "step": 41405 }, { "epoch": 1.93, "learning_rate": 2.6217059585290125e-05, "loss": 0.1487, "step": 41410 }, { "epoch": 1.93, "learning_rate": 2.6216591515454105e-05, "loss": 0.0974, "step": 41415 }, { "epoch": 1.93, "learning_rate": 2.6216123445618085e-05, "loss": 0.1483, "step": 41420 }, { "epoch": 1.93, "learning_rate": 2.6215655375782068e-05, "loss": 0.1493, "step": 41425 }, { "epoch": 1.93, "learning_rate": 2.6215187305946048e-05, "loss": 0.2837, "step": 41430 }, { "epoch": 1.93, "learning_rate": 2.6214719236110028e-05, "loss": 0.2083, "step": 41435 }, { "epoch": 1.93, "learning_rate": 2.621425116627401e-05, "loss": 0.077, "step": 41440 }, { "epoch": 1.93, "learning_rate": 2.621378309643799e-05, "loss": 0.0595, "step": 41445 }, { "epoch": 1.93, "learning_rate": 2.621331502660197e-05, "loss": 0.0702, "step": 41450 }, { "epoch": 1.93, "learning_rate": 2.621284695676595e-05, "loss": 0.0853, "step": 41455 }, { "epoch": 1.93, "learning_rate": 2.6212378886929934e-05, "loss": 0.0832, "step": 41460 }, { "epoch": 1.93, "learning_rate": 2.621191081709391e-05, "loss": 0.0786, "step": 41465 }, { "epoch": 1.94, "learning_rate": 2.621144274725789e-05, "loss": 0.1633, "step": 41470 }, { "epoch": 1.94, "learning_rate": 2.621097467742187e-05, "loss": 0.2989, "step": 41475 }, { "epoch": 1.94, "learning_rate": 2.6210506607585853e-05, "loss": 0.2603, "step": 41480 }, { "epoch": 1.94, "learning_rate": 2.6210038537749833e-05, "loss": 0.2763, "step": 41485 }, { "epoch": 1.94, "learning_rate": 2.6209570467913812e-05, "loss": 0.0806, "step": 41490 }, { "epoch": 1.94, "learning_rate": 2.6209102398077796e-05, "loss": 0.1053, "step": 41495 }, { "epoch": 1.94, "learning_rate": 2.6208634328241775e-05, "loss": 0.0842, "step": 41500 }, { "epoch": 1.94, "learning_rate": 2.6208166258405755e-05, "loss": 0.0804, "step": 41505 }, { "epoch": 1.94, "learning_rate": 2.6207698188569735e-05, "loss": 0.1078, "step": 41510 }, { "epoch": 1.94, "learning_rate": 2.620723011873372e-05, "loss": 0.1337, "step": 41515 }, { "epoch": 1.94, "learning_rate": 2.6206762048897698e-05, "loss": 0.1703, "step": 41520 }, { "epoch": 1.94, "learning_rate": 2.6206293979061675e-05, "loss": 0.2391, "step": 41525 }, { "epoch": 1.94, "learning_rate": 2.6205825909225654e-05, "loss": 0.3403, "step": 41530 }, { "epoch": 1.94, "learning_rate": 2.6205357839389638e-05, "loss": 0.2625, "step": 41535 }, { "epoch": 1.94, "learning_rate": 2.6204889769553617e-05, "loss": 0.1023, "step": 41540 }, { "epoch": 1.94, "learning_rate": 2.6204421699717597e-05, "loss": 0.0718, "step": 41545 }, { "epoch": 1.94, "learning_rate": 2.6203953629881577e-05, "loss": 0.0991, "step": 41550 }, { "epoch": 1.94, "learning_rate": 2.620348556004556e-05, "loss": 0.0729, "step": 41555 }, { "epoch": 1.94, "learning_rate": 2.620301749020954e-05, "loss": 0.1279, "step": 41560 }, { "epoch": 1.94, "learning_rate": 2.620254942037352e-05, "loss": 0.0683, "step": 41565 }, { "epoch": 1.94, "learning_rate": 2.6202081350537503e-05, "loss": 0.0898, "step": 41570 }, { "epoch": 1.94, "learning_rate": 2.6201613280701483e-05, "loss": 0.2204, "step": 41575 }, { "epoch": 1.94, "learning_rate": 2.6201145210865463e-05, "loss": 0.3399, "step": 41580 }, { "epoch": 1.94, "learning_rate": 2.6200677141029443e-05, "loss": 0.244, "step": 41585 }, { "epoch": 1.94, "learning_rate": 2.6200209071193422e-05, "loss": 0.0466, "step": 41590 }, { "epoch": 1.94, "learning_rate": 2.6199741001357402e-05, "loss": 0.0319, "step": 41595 }, { "epoch": 1.94, "learning_rate": 2.6199272931521382e-05, "loss": 0.0508, "step": 41600 }, { "epoch": 1.94, "learning_rate": 2.6198804861685362e-05, "loss": 0.0871, "step": 41605 }, { "epoch": 1.94, "learning_rate": 2.6198336791849345e-05, "loss": 0.0987, "step": 41610 }, { "epoch": 1.94, "learning_rate": 2.6197868722013325e-05, "loss": 0.1401, "step": 41615 }, { "epoch": 1.94, "learning_rate": 2.6197400652177305e-05, "loss": 0.0965, "step": 41620 }, { "epoch": 1.94, "learning_rate": 2.6196932582341288e-05, "loss": 0.1949, "step": 41625 }, { "epoch": 1.94, "learning_rate": 2.6196464512505268e-05, "loss": 0.2828, "step": 41630 }, { "epoch": 1.94, "learning_rate": 2.6195996442669248e-05, "loss": 0.2189, "step": 41635 }, { "epoch": 1.94, "learning_rate": 2.6195528372833227e-05, "loss": 0.0161, "step": 41640 }, { "epoch": 1.94, "learning_rate": 2.619506030299721e-05, "loss": 0.0631, "step": 41645 }, { "epoch": 1.94, "learning_rate": 2.619459223316119e-05, "loss": 0.0599, "step": 41650 }, { "epoch": 1.94, "learning_rate": 2.6194124163325167e-05, "loss": 0.103, "step": 41655 }, { "epoch": 1.94, "learning_rate": 2.6193656093489147e-05, "loss": 0.0704, "step": 41660 }, { "epoch": 1.94, "learning_rate": 2.619318802365313e-05, "loss": 0.1196, "step": 41665 }, { "epoch": 1.94, "learning_rate": 2.619271995381711e-05, "loss": 0.1629, "step": 41670 }, { "epoch": 1.94, "learning_rate": 2.619225188398109e-05, "loss": 0.1673, "step": 41675 }, { "epoch": 1.94, "learning_rate": 2.6191783814145073e-05, "loss": 0.284, "step": 41680 }, { "epoch": 1.95, "learning_rate": 2.6191315744309052e-05, "loss": 0.2082, "step": 41685 }, { "epoch": 1.95, "learning_rate": 2.6190847674473032e-05, "loss": 0.0637, "step": 41690 }, { "epoch": 1.95, "learning_rate": 2.6190379604637012e-05, "loss": 0.0804, "step": 41695 }, { "epoch": 1.95, "learning_rate": 2.6189911534800995e-05, "loss": 0.0269, "step": 41700 }, { "epoch": 1.95, "learning_rate": 2.6189443464964975e-05, "loss": 0.0865, "step": 41705 }, { "epoch": 1.95, "learning_rate": 2.6188975395128955e-05, "loss": 0.0704, "step": 41710 }, { "epoch": 1.95, "learning_rate": 2.618850732529293e-05, "loss": 0.1167, "step": 41715 }, { "epoch": 1.95, "learning_rate": 2.6188039255456915e-05, "loss": 0.2401, "step": 41720 }, { "epoch": 1.95, "learning_rate": 2.6187571185620894e-05, "loss": 0.0924, "step": 41725 }, { "epoch": 1.95, "learning_rate": 2.6187103115784874e-05, "loss": 0.222, "step": 41730 }, { "epoch": 1.95, "learning_rate": 2.6186635045948854e-05, "loss": 0.2267, "step": 41735 }, { "epoch": 1.95, "learning_rate": 2.6186166976112837e-05, "loss": 0.0438, "step": 41740 }, { "epoch": 1.95, "learning_rate": 2.6185698906276817e-05, "loss": 0.0657, "step": 41745 }, { "epoch": 1.95, "learning_rate": 2.6185230836440797e-05, "loss": 0.0662, "step": 41750 }, { "epoch": 1.95, "learning_rate": 2.618476276660478e-05, "loss": 0.0745, "step": 41755 }, { "epoch": 1.95, "learning_rate": 2.618429469676876e-05, "loss": 0.1449, "step": 41760 }, { "epoch": 1.95, "learning_rate": 2.618382662693274e-05, "loss": 0.0906, "step": 41765 }, { "epoch": 1.95, "learning_rate": 2.618335855709672e-05, "loss": 0.1437, "step": 41770 }, { "epoch": 1.95, "learning_rate": 2.6182890487260703e-05, "loss": 0.1986, "step": 41775 }, { "epoch": 1.95, "learning_rate": 2.618242241742468e-05, "loss": 0.28, "step": 41780 }, { "epoch": 1.95, "learning_rate": 2.618195434758866e-05, "loss": 0.2747, "step": 41785 }, { "epoch": 1.95, "learning_rate": 2.618148627775264e-05, "loss": 0.0362, "step": 41790 }, { "epoch": 1.95, "learning_rate": 2.6181018207916622e-05, "loss": 0.0989, "step": 41795 }, { "epoch": 1.95, "learning_rate": 2.6180550138080602e-05, "loss": 0.0329, "step": 41800 }, { "epoch": 1.95, "learning_rate": 2.618008206824458e-05, "loss": 0.0739, "step": 41805 }, { "epoch": 1.95, "learning_rate": 2.6179613998408565e-05, "loss": 0.1304, "step": 41810 }, { "epoch": 1.95, "learning_rate": 2.6179145928572545e-05, "loss": 0.0963, "step": 41815 }, { "epoch": 1.95, "learning_rate": 2.6178677858736524e-05, "loss": 0.1589, "step": 41820 }, { "epoch": 1.95, "learning_rate": 2.6178209788900504e-05, "loss": 0.1445, "step": 41825 }, { "epoch": 1.95, "learning_rate": 2.6177741719064488e-05, "loss": 0.2208, "step": 41830 }, { "epoch": 1.95, "learning_rate": 2.6177273649228467e-05, "loss": 0.3257, "step": 41835 }, { "epoch": 1.95, "learning_rate": 2.6176805579392447e-05, "loss": 0.0379, "step": 41840 }, { "epoch": 1.95, "learning_rate": 2.6176337509556424e-05, "loss": 0.0409, "step": 41845 }, { "epoch": 1.95, "learning_rate": 2.6175869439720407e-05, "loss": 0.0238, "step": 41850 }, { "epoch": 1.95, "learning_rate": 2.6175401369884387e-05, "loss": 0.1279, "step": 41855 }, { "epoch": 1.95, "learning_rate": 2.6174933300048366e-05, "loss": 0.122, "step": 41860 }, { "epoch": 1.95, "learning_rate": 2.617446523021235e-05, "loss": 0.1906, "step": 41865 }, { "epoch": 1.95, "learning_rate": 2.617399716037633e-05, "loss": 0.1503, "step": 41870 }, { "epoch": 1.95, "learning_rate": 2.617352909054031e-05, "loss": 0.1955, "step": 41875 }, { "epoch": 1.95, "learning_rate": 2.617306102070429e-05, "loss": 0.3087, "step": 41880 }, { "epoch": 1.95, "learning_rate": 2.6172592950868272e-05, "loss": 0.4449, "step": 41885 }, { "epoch": 1.95, "learning_rate": 2.6172124881032252e-05, "loss": 0.0521, "step": 41890 }, { "epoch": 1.95, "learning_rate": 2.6171656811196232e-05, "loss": 0.0868, "step": 41895 }, { "epoch": 1.96, "learning_rate": 2.6171188741360212e-05, "loss": 0.0675, "step": 41900 }, { "epoch": 1.96, "learning_rate": 2.617072067152419e-05, "loss": 0.0765, "step": 41905 }, { "epoch": 1.96, "learning_rate": 2.617025260168817e-05, "loss": 0.0803, "step": 41910 }, { "epoch": 1.96, "learning_rate": 2.616978453185215e-05, "loss": 0.0727, "step": 41915 }, { "epoch": 1.96, "learning_rate": 2.616931646201613e-05, "loss": 0.1943, "step": 41920 }, { "epoch": 1.96, "learning_rate": 2.6168848392180114e-05, "loss": 0.2615, "step": 41925 }, { "epoch": 1.96, "learning_rate": 2.6168380322344094e-05, "loss": 0.2774, "step": 41930 }, { "epoch": 1.96, "learning_rate": 2.6167912252508074e-05, "loss": 0.2885, "step": 41935 }, { "epoch": 1.96, "learning_rate": 2.6167444182672057e-05, "loss": 0.0393, "step": 41940 }, { "epoch": 1.96, "learning_rate": 2.6166976112836037e-05, "loss": 0.052, "step": 41945 }, { "epoch": 1.96, "learning_rate": 2.6166508043000017e-05, "loss": 0.0562, "step": 41950 }, { "epoch": 1.96, "learning_rate": 2.6166039973163997e-05, "loss": 0.0304, "step": 41955 }, { "epoch": 1.96, "learning_rate": 2.616557190332798e-05, "loss": 0.0789, "step": 41960 }, { "epoch": 1.96, "learning_rate": 2.616510383349196e-05, "loss": 0.1289, "step": 41965 }, { "epoch": 1.96, "learning_rate": 2.6164635763655936e-05, "loss": 0.2104, "step": 41970 }, { "epoch": 1.96, "learning_rate": 2.6164167693819916e-05, "loss": 0.3554, "step": 41975 }, { "epoch": 1.96, "learning_rate": 2.61636996239839e-05, "loss": 0.2247, "step": 41980 }, { "epoch": 1.96, "learning_rate": 2.616323155414788e-05, "loss": 0.281, "step": 41985 }, { "epoch": 1.96, "learning_rate": 2.616276348431186e-05, "loss": 0.0692, "step": 41990 }, { "epoch": 1.96, "learning_rate": 2.6162295414475842e-05, "loss": 0.0418, "step": 41995 }, { "epoch": 1.96, "learning_rate": 2.616182734463982e-05, "loss": 0.1275, "step": 42000 }, { "epoch": 1.96, "learning_rate": 2.61613592748038e-05, "loss": 0.1626, "step": 42005 }, { "epoch": 1.96, "learning_rate": 2.616089120496778e-05, "loss": 0.0906, "step": 42010 }, { "epoch": 1.96, "learning_rate": 2.6160423135131764e-05, "loss": 0.1578, "step": 42015 }, { "epoch": 1.96, "learning_rate": 2.6159955065295744e-05, "loss": 0.1298, "step": 42020 }, { "epoch": 1.96, "learning_rate": 2.6159486995459724e-05, "loss": 0.1097, "step": 42025 }, { "epoch": 1.96, "learning_rate": 2.61590189256237e-05, "loss": 0.3597, "step": 42030 }, { "epoch": 1.96, "learning_rate": 2.6158550855787684e-05, "loss": 0.2528, "step": 42035 }, { "epoch": 1.96, "learning_rate": 2.6158082785951664e-05, "loss": 0.0207, "step": 42040 }, { "epoch": 1.96, "learning_rate": 2.6157614716115643e-05, "loss": 0.0425, "step": 42045 }, { "epoch": 1.96, "learning_rate": 2.6157146646279627e-05, "loss": 0.0486, "step": 42050 }, { "epoch": 1.96, "learning_rate": 2.6156678576443606e-05, "loss": 0.0863, "step": 42055 }, { "epoch": 1.96, "learning_rate": 2.6156210506607586e-05, "loss": 0.1083, "step": 42060 }, { "epoch": 1.96, "learning_rate": 2.6155742436771566e-05, "loss": 0.1881, "step": 42065 }, { "epoch": 1.96, "learning_rate": 2.615527436693555e-05, "loss": 0.2123, "step": 42070 }, { "epoch": 1.96, "learning_rate": 2.615480629709953e-05, "loss": 0.2059, "step": 42075 }, { "epoch": 1.96, "learning_rate": 2.615433822726351e-05, "loss": 0.2769, "step": 42080 }, { "epoch": 1.96, "learning_rate": 2.615387015742749e-05, "loss": 0.2228, "step": 42085 }, { "epoch": 1.96, "learning_rate": 2.6153402087591472e-05, "loss": 0.0218, "step": 42090 }, { "epoch": 1.96, "learning_rate": 2.615293401775545e-05, "loss": 0.3974, "step": 42095 }, { "epoch": 1.96, "learning_rate": 2.6152465947919428e-05, "loss": 0.0647, "step": 42100 }, { "epoch": 1.96, "learning_rate": 2.615199787808341e-05, "loss": 0.0883, "step": 42105 }, { "epoch": 1.96, "learning_rate": 2.615152980824739e-05, "loss": 0.0682, "step": 42110 }, { "epoch": 1.97, "learning_rate": 2.615106173841137e-05, "loss": 0.1631, "step": 42115 }, { "epoch": 1.97, "learning_rate": 2.615059366857535e-05, "loss": 0.1816, "step": 42120 }, { "epoch": 1.97, "learning_rate": 2.6150125598739334e-05, "loss": 0.2507, "step": 42125 }, { "epoch": 1.97, "learning_rate": 2.6149657528903314e-05, "loss": 0.2652, "step": 42130 }, { "epoch": 1.97, "learning_rate": 2.6149189459067294e-05, "loss": 0.2137, "step": 42135 }, { "epoch": 1.97, "learning_rate": 2.6148721389231273e-05, "loss": 0.0332, "step": 42140 }, { "epoch": 1.97, "learning_rate": 2.6148253319395257e-05, "loss": 0.0513, "step": 42145 }, { "epoch": 1.97, "learning_rate": 2.6147785249559236e-05, "loss": 0.0946, "step": 42150 }, { "epoch": 1.97, "learning_rate": 2.6147317179723216e-05, "loss": 0.089, "step": 42155 }, { "epoch": 1.97, "learning_rate": 2.6146849109887193e-05, "loss": 0.0767, "step": 42160 }, { "epoch": 1.97, "learning_rate": 2.6146381040051176e-05, "loss": 0.0689, "step": 42165 }, { "epoch": 1.97, "learning_rate": 2.6145912970215156e-05, "loss": 0.1849, "step": 42170 }, { "epoch": 1.97, "learning_rate": 2.6145444900379136e-05, "loss": 0.2643, "step": 42175 }, { "epoch": 1.97, "learning_rate": 2.614497683054312e-05, "loss": 0.387, "step": 42180 }, { "epoch": 1.97, "learning_rate": 2.61445087607071e-05, "loss": 0.2659, "step": 42185 }, { "epoch": 1.97, "learning_rate": 2.614404069087108e-05, "loss": 0.0683, "step": 42190 }, { "epoch": 1.97, "learning_rate": 2.6143572621035058e-05, "loss": 0.0357, "step": 42195 }, { "epoch": 1.97, "learning_rate": 2.614310455119904e-05, "loss": 0.0824, "step": 42200 }, { "epoch": 1.97, "learning_rate": 2.614263648136302e-05, "loss": 0.0815, "step": 42205 }, { "epoch": 1.97, "learning_rate": 2.6142168411527e-05, "loss": 0.1082, "step": 42210 }, { "epoch": 1.97, "learning_rate": 2.614170034169098e-05, "loss": 0.1158, "step": 42215 }, { "epoch": 1.97, "learning_rate": 2.614123227185496e-05, "loss": 0.1754, "step": 42220 }, { "epoch": 1.97, "learning_rate": 2.614076420201894e-05, "loss": 0.1782, "step": 42225 }, { "epoch": 1.97, "learning_rate": 2.614029613218292e-05, "loss": 0.2233, "step": 42230 }, { "epoch": 1.97, "learning_rate": 2.6139828062346904e-05, "loss": 0.2004, "step": 42235 }, { "epoch": 1.97, "learning_rate": 2.6139359992510883e-05, "loss": 0.0311, "step": 42240 }, { "epoch": 1.97, "learning_rate": 2.6138891922674863e-05, "loss": 0.0693, "step": 42245 }, { "epoch": 1.97, "learning_rate": 2.6138423852838843e-05, "loss": 0.0721, "step": 42250 }, { "epoch": 1.97, "learning_rate": 2.6137955783002826e-05, "loss": 0.072, "step": 42255 }, { "epoch": 1.97, "learning_rate": 2.6137487713166806e-05, "loss": 0.1183, "step": 42260 }, { "epoch": 1.97, "learning_rate": 2.6137019643330786e-05, "loss": 0.1451, "step": 42265 }, { "epoch": 1.97, "learning_rate": 2.6136551573494766e-05, "loss": 0.1398, "step": 42270 }, { "epoch": 1.97, "learning_rate": 2.613608350365875e-05, "loss": 0.1883, "step": 42275 }, { "epoch": 1.97, "learning_rate": 2.613561543382273e-05, "loss": 0.2863, "step": 42280 }, { "epoch": 1.97, "learning_rate": 2.6135147363986705e-05, "loss": 0.2521, "step": 42285 }, { "epoch": 1.97, "learning_rate": 2.613467929415069e-05, "loss": 0.0794, "step": 42290 }, { "epoch": 1.97, "learning_rate": 2.6134211224314668e-05, "loss": 0.0666, "step": 42295 }, { "epoch": 1.97, "learning_rate": 2.6133743154478648e-05, "loss": 0.0405, "step": 42300 }, { "epoch": 1.97, "learning_rate": 2.6133275084642628e-05, "loss": 0.0851, "step": 42305 }, { "epoch": 1.97, "learning_rate": 2.613280701480661e-05, "loss": 0.0478, "step": 42310 }, { "epoch": 1.97, "learning_rate": 2.613233894497059e-05, "loss": 0.0851, "step": 42315 }, { "epoch": 1.97, "learning_rate": 2.613187087513457e-05, "loss": 0.1475, "step": 42320 }, { "epoch": 1.97, "learning_rate": 2.613140280529855e-05, "loss": 0.1359, "step": 42325 }, { "epoch": 1.98, "learning_rate": 2.6130934735462534e-05, "loss": 0.3473, "step": 42330 }, { "epoch": 1.98, "learning_rate": 2.6130466665626513e-05, "loss": 0.3202, "step": 42335 }, { "epoch": 1.98, "learning_rate": 2.6129998595790493e-05, "loss": 0.037, "step": 42340 }, { "epoch": 1.98, "learning_rate": 2.6129530525954473e-05, "loss": 0.0496, "step": 42345 }, { "epoch": 1.98, "learning_rate": 2.6129062456118453e-05, "loss": 0.0428, "step": 42350 }, { "epoch": 1.98, "learning_rate": 2.6128594386282433e-05, "loss": 0.0577, "step": 42355 }, { "epoch": 1.98, "learning_rate": 2.6128126316446413e-05, "loss": 0.102, "step": 42360 }, { "epoch": 1.98, "learning_rate": 2.6127658246610396e-05, "loss": 0.1314, "step": 42365 }, { "epoch": 1.98, "learning_rate": 2.6127190176774376e-05, "loss": 0.1403, "step": 42370 }, { "epoch": 1.98, "learning_rate": 2.6126722106938355e-05, "loss": 0.1798, "step": 42375 }, { "epoch": 1.98, "learning_rate": 2.6126254037102335e-05, "loss": 0.3162, "step": 42380 }, { "epoch": 1.98, "learning_rate": 2.612578596726632e-05, "loss": 0.1984, "step": 42385 }, { "epoch": 1.98, "learning_rate": 2.6125317897430298e-05, "loss": 0.0688, "step": 42390 }, { "epoch": 1.98, "learning_rate": 2.6124849827594278e-05, "loss": 0.1072, "step": 42395 }, { "epoch": 1.98, "learning_rate": 2.6124381757758258e-05, "loss": 0.0962, "step": 42400 }, { "epoch": 1.98, "learning_rate": 2.612391368792224e-05, "loss": 0.0709, "step": 42405 }, { "epoch": 1.98, "learning_rate": 2.6123445618086218e-05, "loss": 0.1281, "step": 42410 }, { "epoch": 1.98, "learning_rate": 2.6122977548250197e-05, "loss": 0.1265, "step": 42415 }, { "epoch": 1.98, "learning_rate": 2.612250947841418e-05, "loss": 0.2309, "step": 42420 }, { "epoch": 1.98, "learning_rate": 2.612204140857816e-05, "loss": 0.1706, "step": 42425 }, { "epoch": 1.98, "learning_rate": 2.612157333874214e-05, "loss": 0.3254, "step": 42430 }, { "epoch": 1.98, "learning_rate": 2.612110526890612e-05, "loss": 0.2413, "step": 42435 }, { "epoch": 1.98, "learning_rate": 2.6120637199070103e-05, "loss": 0.0151, "step": 42440 }, { "epoch": 1.98, "learning_rate": 2.6120169129234083e-05, "loss": 0.0524, "step": 42445 }, { "epoch": 1.98, "learning_rate": 2.6119701059398063e-05, "loss": 0.0802, "step": 42450 }, { "epoch": 1.98, "learning_rate": 2.6119232989562043e-05, "loss": 0.0699, "step": 42455 }, { "epoch": 1.98, "learning_rate": 2.6118764919726026e-05, "loss": 0.1366, "step": 42460 }, { "epoch": 1.98, "learning_rate": 2.6118296849890006e-05, "loss": 0.1103, "step": 42465 }, { "epoch": 1.98, "learning_rate": 2.6117828780053985e-05, "loss": 0.199, "step": 42470 }, { "epoch": 1.98, "learning_rate": 2.6117360710217965e-05, "loss": 0.1311, "step": 42475 }, { "epoch": 1.98, "learning_rate": 2.6116892640381945e-05, "loss": 0.2063, "step": 42480 }, { "epoch": 1.98, "learning_rate": 2.6116424570545925e-05, "loss": 0.1962, "step": 42485 }, { "epoch": 1.98, "learning_rate": 2.6115956500709905e-05, "loss": 0.0497, "step": 42490 }, { "epoch": 1.98, "learning_rate": 2.6115488430873888e-05, "loss": 0.0615, "step": 42495 }, { "epoch": 1.98, "learning_rate": 2.6115020361037868e-05, "loss": 0.1056, "step": 42500 }, { "epoch": 1.98, "learning_rate": 2.6114552291201848e-05, "loss": 0.0626, "step": 42505 }, { "epoch": 1.98, "learning_rate": 2.6114084221365827e-05, "loss": 0.0998, "step": 42510 }, { "epoch": 1.98, "learning_rate": 2.611361615152981e-05, "loss": 0.1034, "step": 42515 }, { "epoch": 1.98, "learning_rate": 2.611314808169379e-05, "loss": 0.2097, "step": 42520 }, { "epoch": 1.98, "learning_rate": 2.611268001185777e-05, "loss": 0.273, "step": 42525 }, { "epoch": 1.98, "learning_rate": 2.611221194202175e-05, "loss": 0.2922, "step": 42530 }, { "epoch": 1.98, "learning_rate": 2.611174387218573e-05, "loss": 0.3163, "step": 42535 }, { "epoch": 1.98, "learning_rate": 2.611127580234971e-05, "loss": 0.0387, "step": 42540 }, { "epoch": 1.99, "learning_rate": 2.611080773251369e-05, "loss": 0.0681, "step": 42545 }, { "epoch": 1.99, "learning_rate": 2.6110339662677673e-05, "loss": 0.0488, "step": 42550 }, { "epoch": 1.99, "learning_rate": 2.6109871592841653e-05, "loss": 0.0721, "step": 42555 }, { "epoch": 1.99, "learning_rate": 2.6109403523005632e-05, "loss": 0.0597, "step": 42560 }, { "epoch": 1.99, "learning_rate": 2.6108935453169612e-05, "loss": 0.1119, "step": 42565 }, { "epoch": 1.99, "learning_rate": 2.6108467383333595e-05, "loss": 0.149, "step": 42570 }, { "epoch": 1.99, "learning_rate": 2.6107999313497575e-05, "loss": 0.0987, "step": 42575 }, { "epoch": 1.99, "learning_rate": 2.6107531243661555e-05, "loss": 0.3054, "step": 42580 }, { "epoch": 1.99, "learning_rate": 2.6107063173825535e-05, "loss": 0.2318, "step": 42585 }, { "epoch": 1.99, "learning_rate": 2.6106595103989518e-05, "loss": 0.0366, "step": 42590 }, { "epoch": 1.99, "learning_rate": 2.6106127034153498e-05, "loss": 0.1054, "step": 42595 }, { "epoch": 1.99, "learning_rate": 2.6105658964317474e-05, "loss": 0.0642, "step": 42600 }, { "epoch": 1.99, "learning_rate": 2.6105190894481458e-05, "loss": 0.1019, "step": 42605 }, { "epoch": 1.99, "learning_rate": 2.6104722824645437e-05, "loss": 0.0302, "step": 42610 }, { "epoch": 1.99, "learning_rate": 2.6104254754809417e-05, "loss": 0.0899, "step": 42615 }, { "epoch": 1.99, "learning_rate": 2.6103786684973397e-05, "loss": 0.1203, "step": 42620 }, { "epoch": 1.99, "learning_rate": 2.610331861513738e-05, "loss": 0.2097, "step": 42625 }, { "epoch": 1.99, "learning_rate": 2.610285054530136e-05, "loss": 0.2266, "step": 42630 }, { "epoch": 1.99, "learning_rate": 2.610238247546534e-05, "loss": 0.2678, "step": 42635 }, { "epoch": 1.99, "learning_rate": 2.610191440562932e-05, "loss": 0.0429, "step": 42640 }, { "epoch": 1.99, "learning_rate": 2.6101446335793303e-05, "loss": 0.0258, "step": 42645 }, { "epoch": 1.99, "learning_rate": 2.6100978265957283e-05, "loss": 0.099, "step": 42650 }, { "epoch": 1.99, "learning_rate": 2.6100510196121262e-05, "loss": 0.0877, "step": 42655 }, { "epoch": 1.99, "learning_rate": 2.6100042126285246e-05, "loss": 0.1255, "step": 42660 }, { "epoch": 1.99, "learning_rate": 2.6099574056449222e-05, "loss": 0.1109, "step": 42665 }, { "epoch": 1.99, "learning_rate": 2.6099105986613202e-05, "loss": 0.1523, "step": 42670 }, { "epoch": 1.99, "learning_rate": 2.6098637916777182e-05, "loss": 0.1558, "step": 42675 }, { "epoch": 1.99, "learning_rate": 2.6098169846941165e-05, "loss": 0.2726, "step": 42680 }, { "epoch": 1.99, "learning_rate": 2.6097701777105145e-05, "loss": 0.3082, "step": 42685 }, { "epoch": 1.99, "learning_rate": 2.6097233707269125e-05, "loss": 0.0495, "step": 42690 }, { "epoch": 1.99, "learning_rate": 2.6096765637433104e-05, "loss": 0.0402, "step": 42695 }, { "epoch": 1.99, "learning_rate": 2.6096297567597088e-05, "loss": 0.0397, "step": 42700 }, { "epoch": 1.99, "learning_rate": 2.6095829497761067e-05, "loss": 0.0766, "step": 42705 }, { "epoch": 1.99, "learning_rate": 2.6095361427925047e-05, "loss": 0.104, "step": 42710 }, { "epoch": 1.99, "learning_rate": 2.609489335808903e-05, "loss": 0.1302, "step": 42715 }, { "epoch": 1.99, "learning_rate": 2.609442528825301e-05, "loss": 0.1071, "step": 42720 }, { "epoch": 1.99, "learning_rate": 2.6093957218416987e-05, "loss": 0.2285, "step": 42725 }, { "epoch": 1.99, "learning_rate": 2.6093489148580966e-05, "loss": 0.4066, "step": 42730 }, { "epoch": 1.99, "learning_rate": 2.609302107874495e-05, "loss": 0.1984, "step": 42735 }, { "epoch": 1.99, "learning_rate": 2.609255300890893e-05, "loss": 0.0729, "step": 42740 }, { "epoch": 1.99, "learning_rate": 2.609208493907291e-05, "loss": 0.0228, "step": 42745 }, { "epoch": 1.99, "learning_rate": 2.609161686923689e-05, "loss": 0.0371, "step": 42750 }, { "epoch": 2.0, "learning_rate": 2.6091148799400872e-05, "loss": 0.0475, "step": 42755 }, { "epoch": 2.0, "learning_rate": 2.6090680729564852e-05, "loss": 0.0637, "step": 42760 }, { "epoch": 2.0, "learning_rate": 2.6090212659728832e-05, "loss": 0.122, "step": 42765 }, { "epoch": 2.0, "learning_rate": 2.6089744589892812e-05, "loss": 0.1042, "step": 42770 }, { "epoch": 2.0, "learning_rate": 2.6089276520056795e-05, "loss": 0.1345, "step": 42775 }, { "epoch": 2.0, "learning_rate": 2.6088808450220775e-05, "loss": 0.3056, "step": 42780 }, { "epoch": 2.0, "learning_rate": 2.6088340380384755e-05, "loss": 0.2785, "step": 42785 }, { "epoch": 2.0, "learning_rate": 2.6087872310548734e-05, "loss": 0.0451, "step": 42790 }, { "epoch": 2.0, "learning_rate": 2.6087404240712714e-05, "loss": 0.0623, "step": 42795 }, { "epoch": 2.0, "learning_rate": 2.6086936170876694e-05, "loss": 0.074, "step": 42800 }, { "epoch": 2.0, "learning_rate": 2.6086468101040674e-05, "loss": 0.0478, "step": 42805 }, { "epoch": 2.0, "learning_rate": 2.6086000031204657e-05, "loss": 0.0585, "step": 42810 }, { "epoch": 2.0, "learning_rate": 2.6085531961368637e-05, "loss": 0.0827, "step": 42815 }, { "epoch": 2.0, "learning_rate": 2.6085063891532617e-05, "loss": 0.1835, "step": 42820 }, { "epoch": 2.0, "learning_rate": 2.6084595821696597e-05, "loss": 0.2746, "step": 42825 }, { "epoch": 2.0, "learning_rate": 2.608412775186058e-05, "loss": 0.3944, "step": 42830 }, { "epoch": 2.0, "learning_rate": 2.608365968202456e-05, "loss": 0.221, "step": 42835 }, { "epoch": 2.0, "learning_rate": 2.608319161218854e-05, "loss": 0.0433, "step": 42840 }, { "epoch": 2.0, "learning_rate": 2.6082723542352523e-05, "loss": 0.0439, "step": 42845 }, { "epoch": 2.0, "learning_rate": 2.6082255472516502e-05, "loss": 0.1034, "step": 42850 }, { "epoch": 2.0, "learning_rate": 2.608178740268048e-05, "loss": 0.1014, "step": 42855 }, { "epoch": 2.0, "learning_rate": 2.608131933284446e-05, "loss": 0.2214, "step": 42860 }, { "epoch": 2.0, "eval_cer": 0.012627477835976975, "eval_loss": 0.026034235954284668, "eval_runtime": 396.3831, "eval_samples_per_second": 48.06, "eval_steps_per_second": 12.016, "eval_wer": 0.10848767794275219, "step": 42862 }, { "epoch": 2.0, "learning_rate": 2.6080851263008442e-05, "loss": 0.3494, "step": 42865 }, { "epoch": 2.0, "learning_rate": 2.6080383193172422e-05, "loss": 0.1003, "step": 42870 }, { "epoch": 2.0, "learning_rate": 2.60799151233364e-05, "loss": 0.0811, "step": 42875 }, { "epoch": 2.0, "learning_rate": 2.607944705350038e-05, "loss": 0.0584, "step": 42880 }, { "epoch": 2.0, "learning_rate": 2.6078978983664365e-05, "loss": 0.1147, "step": 42885 }, { "epoch": 2.0, "learning_rate": 2.6078510913828344e-05, "loss": 0.1141, "step": 42890 }, { "epoch": 2.0, "learning_rate": 2.6078042843992324e-05, "loss": 0.1389, "step": 42895 }, { "epoch": 2.0, "learning_rate": 2.6077574774156307e-05, "loss": 0.0897, "step": 42900 }, { "epoch": 2.0, "learning_rate": 2.6077106704320287e-05, "loss": 0.1826, "step": 42905 }, { "epoch": 2.0, "learning_rate": 2.6076638634484267e-05, "loss": 0.2396, "step": 42910 }, { "epoch": 2.0, "learning_rate": 2.6076170564648243e-05, "loss": 0.2953, "step": 42915 }, { "epoch": 2.0, "learning_rate": 2.6075702494812227e-05, "loss": 0.0186, "step": 42920 }, { "epoch": 2.0, "learning_rate": 2.6075234424976206e-05, "loss": 0.0479, "step": 42925 }, { "epoch": 2.0, "learning_rate": 2.6074766355140186e-05, "loss": 0.0722, "step": 42930 }, { "epoch": 2.0, "learning_rate": 2.6074298285304166e-05, "loss": 0.1272, "step": 42935 }, { "epoch": 2.0, "learning_rate": 2.607383021546815e-05, "loss": 0.1276, "step": 42940 }, { "epoch": 2.0, "learning_rate": 2.607336214563213e-05, "loss": 0.0871, "step": 42945 }, { "epoch": 2.0, "learning_rate": 2.607289407579611e-05, "loss": 0.1311, "step": 42950 }, { "epoch": 2.0, "learning_rate": 2.607242600596009e-05, "loss": 0.2117, "step": 42955 }, { "epoch": 2.0, "learning_rate": 2.6071957936124072e-05, "loss": 0.2268, "step": 42960 }, { "epoch": 2.0, "learning_rate": 2.6071489866288052e-05, "loss": 0.3229, "step": 42965 }, { "epoch": 2.01, "learning_rate": 2.607102179645203e-05, "loss": 0.0631, "step": 42970 }, { "epoch": 2.01, "learning_rate": 2.6070553726616015e-05, "loss": 0.0875, "step": 42975 }, { "epoch": 2.01, "learning_rate": 2.607008565677999e-05, "loss": 0.0383, "step": 42980 }, { "epoch": 2.01, "learning_rate": 2.606961758694397e-05, "loss": 0.0947, "step": 42985 }, { "epoch": 2.01, "learning_rate": 2.606914951710795e-05, "loss": 0.0477, "step": 42990 }, { "epoch": 2.01, "learning_rate": 2.6068681447271934e-05, "loss": 0.1214, "step": 42995 }, { "epoch": 2.01, "learning_rate": 2.6068213377435914e-05, "loss": 0.1402, "step": 43000 }, { "epoch": 2.01, "learning_rate": 2.6067745307599894e-05, "loss": 0.1877, "step": 43005 }, { "epoch": 2.01, "learning_rate": 2.6067277237763874e-05, "loss": 0.2252, "step": 43010 }, { "epoch": 2.01, "learning_rate": 2.6066809167927857e-05, "loss": 0.3702, "step": 43015 }, { "epoch": 2.01, "learning_rate": 2.6066341098091837e-05, "loss": 0.0727, "step": 43020 }, { "epoch": 2.01, "learning_rate": 2.6065873028255816e-05, "loss": 0.0472, "step": 43025 }, { "epoch": 2.01, "learning_rate": 2.60654049584198e-05, "loss": 0.0736, "step": 43030 }, { "epoch": 2.01, "learning_rate": 2.606493688858378e-05, "loss": 0.073, "step": 43035 }, { "epoch": 2.01, "learning_rate": 2.606446881874776e-05, "loss": 0.1267, "step": 43040 }, { "epoch": 2.01, "learning_rate": 2.6064000748911736e-05, "loss": 0.0941, "step": 43045 }, { "epoch": 2.01, "learning_rate": 2.606353267907572e-05, "loss": 0.1539, "step": 43050 }, { "epoch": 2.01, "learning_rate": 2.60630646092397e-05, "loss": 0.1133, "step": 43055 }, { "epoch": 2.01, "learning_rate": 2.606259653940368e-05, "loss": 0.1859, "step": 43060 }, { "epoch": 2.01, "learning_rate": 2.606212846956766e-05, "loss": 0.4176, "step": 43065 }, { "epoch": 2.01, "learning_rate": 2.606166039973164e-05, "loss": 0.0344, "step": 43070 }, { "epoch": 2.01, "learning_rate": 2.606119232989562e-05, "loss": 0.0579, "step": 43075 }, { "epoch": 2.01, "learning_rate": 2.60607242600596e-05, "loss": 0.0737, "step": 43080 }, { "epoch": 2.01, "learning_rate": 2.6060256190223584e-05, "loss": 0.0629, "step": 43085 }, { "epoch": 2.01, "learning_rate": 2.6059788120387564e-05, "loss": 0.1527, "step": 43090 }, { "epoch": 2.01, "learning_rate": 2.6059320050551544e-05, "loss": 0.0979, "step": 43095 }, { "epoch": 2.01, "learning_rate": 2.6058851980715524e-05, "loss": 0.1139, "step": 43100 }, { "epoch": 2.01, "learning_rate": 2.6058383910879504e-05, "loss": 0.1651, "step": 43105 }, { "epoch": 2.01, "learning_rate": 2.6057915841043483e-05, "loss": 0.24, "step": 43110 }, { "epoch": 2.01, "learning_rate": 2.6057447771207463e-05, "loss": 0.3547, "step": 43115 }, { "epoch": 2.01, "learning_rate": 2.6056979701371443e-05, "loss": 0.0932, "step": 43120 }, { "epoch": 2.01, "learning_rate": 2.6056511631535426e-05, "loss": 0.0421, "step": 43125 }, { "epoch": 2.01, "learning_rate": 2.6056043561699406e-05, "loss": 0.0528, "step": 43130 }, { "epoch": 2.01, "learning_rate": 2.6055575491863386e-05, "loss": 0.0339, "step": 43135 }, { "epoch": 2.01, "learning_rate": 2.6055107422027366e-05, "loss": 0.1307, "step": 43140 }, { "epoch": 2.01, "learning_rate": 2.605463935219135e-05, "loss": 0.0994, "step": 43145 }, { "epoch": 2.01, "learning_rate": 2.605417128235533e-05, "loss": 0.1795, "step": 43150 }, { "epoch": 2.01, "learning_rate": 2.605370321251931e-05, "loss": 0.1295, "step": 43155 }, { "epoch": 2.01, "learning_rate": 2.6053235142683292e-05, "loss": 0.2791, "step": 43160 }, { "epoch": 2.01, "learning_rate": 2.605276707284727e-05, "loss": 0.3165, "step": 43165 }, { "epoch": 2.01, "learning_rate": 2.6052299003011248e-05, "loss": 0.0509, "step": 43170 }, { "epoch": 2.01, "learning_rate": 2.6051830933175228e-05, "loss": 0.0227, "step": 43175 }, { "epoch": 2.01, "learning_rate": 2.605136286333921e-05, "loss": 0.0606, "step": 43180 }, { "epoch": 2.02, "learning_rate": 2.605089479350319e-05, "loss": 0.0552, "step": 43185 }, { "epoch": 2.02, "learning_rate": 2.605042672366717e-05, "loss": 0.1295, "step": 43190 }, { "epoch": 2.02, "learning_rate": 2.604995865383115e-05, "loss": 0.0784, "step": 43195 }, { "epoch": 2.02, "learning_rate": 2.6049490583995134e-05, "loss": 0.1391, "step": 43200 }, { "epoch": 2.02, "learning_rate": 2.6049022514159114e-05, "loss": 0.1368, "step": 43205 }, { "epoch": 2.02, "learning_rate": 2.6048554444323093e-05, "loss": 0.2506, "step": 43210 }, { "epoch": 2.02, "learning_rate": 2.6048086374487077e-05, "loss": 0.2382, "step": 43215 }, { "epoch": 2.02, "learning_rate": 2.6047618304651056e-05, "loss": 0.0312, "step": 43220 }, { "epoch": 2.02, "learning_rate": 2.6047150234815036e-05, "loss": 0.066, "step": 43225 }, { "epoch": 2.02, "learning_rate": 2.6046682164979016e-05, "loss": 0.0797, "step": 43230 }, { "epoch": 2.02, "learning_rate": 2.6046214095142996e-05, "loss": 0.0473, "step": 43235 }, { "epoch": 2.02, "learning_rate": 2.6045746025306976e-05, "loss": 0.1001, "step": 43240 }, { "epoch": 2.02, "learning_rate": 2.6045277955470955e-05, "loss": 0.0724, "step": 43245 }, { "epoch": 2.02, "learning_rate": 2.6044809885634935e-05, "loss": 0.1473, "step": 43250 }, { "epoch": 2.02, "learning_rate": 2.604434181579892e-05, "loss": 0.1675, "step": 43255 }, { "epoch": 2.02, "learning_rate": 2.60438737459629e-05, "loss": 0.1769, "step": 43260 }, { "epoch": 2.02, "learning_rate": 2.6043405676126878e-05, "loss": 0.262, "step": 43265 }, { "epoch": 2.02, "learning_rate": 2.604293760629086e-05, "loss": 0.036, "step": 43270 }, { "epoch": 2.02, "learning_rate": 2.604246953645484e-05, "loss": 0.0645, "step": 43275 }, { "epoch": 2.02, "learning_rate": 2.604200146661882e-05, "loss": 0.0808, "step": 43280 }, { "epoch": 2.02, "learning_rate": 2.60415333967828e-05, "loss": 0.0532, "step": 43285 }, { "epoch": 2.02, "learning_rate": 2.6041065326946784e-05, "loss": 0.063, "step": 43290 }, { "epoch": 2.02, "learning_rate": 2.604059725711076e-05, "loss": 0.1389, "step": 43295 }, { "epoch": 2.02, "learning_rate": 2.604012918727474e-05, "loss": 0.1972, "step": 43300 }, { "epoch": 2.02, "learning_rate": 2.603966111743872e-05, "loss": 0.1335, "step": 43305 }, { "epoch": 2.02, "learning_rate": 2.6039193047602703e-05, "loss": 0.3182, "step": 43310 }, { "epoch": 2.02, "learning_rate": 2.6038724977766683e-05, "loss": 0.2989, "step": 43315 }, { "epoch": 2.02, "learning_rate": 2.6038256907930663e-05, "loss": 0.0236, "step": 43320 }, { "epoch": 2.02, "learning_rate": 2.6037788838094643e-05, "loss": 0.0543, "step": 43325 }, { "epoch": 2.02, "learning_rate": 2.6037320768258626e-05, "loss": 0.0413, "step": 43330 }, { "epoch": 2.02, "learning_rate": 2.6036852698422606e-05, "loss": 0.0411, "step": 43335 }, { "epoch": 2.02, "learning_rate": 2.6036384628586586e-05, "loss": 0.0893, "step": 43340 }, { "epoch": 2.02, "learning_rate": 2.603591655875057e-05, "loss": 0.0652, "step": 43345 }, { "epoch": 2.02, "learning_rate": 2.603544848891455e-05, "loss": 0.102, "step": 43350 }, { "epoch": 2.02, "learning_rate": 2.603498041907853e-05, "loss": 0.1159, "step": 43355 }, { "epoch": 2.02, "learning_rate": 2.6034512349242505e-05, "loss": 0.1669, "step": 43360 }, { "epoch": 2.02, "learning_rate": 2.6034044279406488e-05, "loss": 0.3293, "step": 43365 }, { "epoch": 2.02, "learning_rate": 2.6033576209570468e-05, "loss": 0.0544, "step": 43370 }, { "epoch": 2.02, "learning_rate": 2.6033108139734448e-05, "loss": 0.0811, "step": 43375 }, { "epoch": 2.02, "learning_rate": 2.6032640069898427e-05, "loss": 0.0443, "step": 43380 }, { "epoch": 2.02, "learning_rate": 2.603217200006241e-05, "loss": 0.0427, "step": 43385 }, { "epoch": 2.02, "learning_rate": 2.603170393022639e-05, "loss": 0.1606, "step": 43390 }, { "epoch": 2.02, "learning_rate": 2.603123586039037e-05, "loss": 0.0655, "step": 43395 }, { "epoch": 2.03, "learning_rate": 2.6030767790554354e-05, "loss": 0.1189, "step": 43400 }, { "epoch": 2.03, "learning_rate": 2.6030299720718333e-05, "loss": 0.2558, "step": 43405 }, { "epoch": 2.03, "learning_rate": 2.6029831650882313e-05, "loss": 0.2996, "step": 43410 }, { "epoch": 2.03, "learning_rate": 2.6029363581046293e-05, "loss": 0.3012, "step": 43415 }, { "epoch": 2.03, "learning_rate": 2.6028895511210273e-05, "loss": 0.0208, "step": 43420 }, { "epoch": 2.03, "learning_rate": 2.6028427441374253e-05, "loss": 0.0588, "step": 43425 }, { "epoch": 2.03, "learning_rate": 2.6027959371538232e-05, "loss": 0.0582, "step": 43430 }, { "epoch": 2.03, "learning_rate": 2.6027491301702212e-05, "loss": 0.1172, "step": 43435 }, { "epoch": 2.03, "learning_rate": 2.6027023231866195e-05, "loss": 0.0628, "step": 43440 }, { "epoch": 2.03, "learning_rate": 2.6026555162030175e-05, "loss": 0.0249, "step": 43445 }, { "epoch": 2.03, "learning_rate": 2.6026087092194155e-05, "loss": 0.1262, "step": 43450 }, { "epoch": 2.03, "learning_rate": 2.602561902235814e-05, "loss": 0.2652, "step": 43455 }, { "epoch": 2.03, "learning_rate": 2.6025150952522118e-05, "loss": 0.1766, "step": 43460 }, { "epoch": 2.03, "learning_rate": 2.6024682882686098e-05, "loss": 0.2257, "step": 43465 }, { "epoch": 2.03, "learning_rate": 2.6024214812850078e-05, "loss": 0.0303, "step": 43470 }, { "epoch": 2.03, "learning_rate": 2.602374674301406e-05, "loss": 0.0402, "step": 43475 }, { "epoch": 2.03, "learning_rate": 2.602327867317804e-05, "loss": 0.0562, "step": 43480 }, { "epoch": 2.03, "learning_rate": 2.6022810603342017e-05, "loss": 0.1124, "step": 43485 }, { "epoch": 2.03, "learning_rate": 2.6022342533505997e-05, "loss": 0.0864, "step": 43490 }, { "epoch": 2.03, "learning_rate": 2.602187446366998e-05, "loss": 0.0988, "step": 43495 }, { "epoch": 2.03, "learning_rate": 2.602140639383396e-05, "loss": 0.1198, "step": 43500 }, { "epoch": 2.03, "learning_rate": 2.602093832399794e-05, "loss": 0.2089, "step": 43505 }, { "epoch": 2.03, "learning_rate": 2.6020470254161923e-05, "loss": 0.2975, "step": 43510 }, { "epoch": 2.03, "learning_rate": 2.6020002184325903e-05, "loss": 0.3667, "step": 43515 }, { "epoch": 2.03, "learning_rate": 2.6019534114489883e-05, "loss": 0.0904, "step": 43520 }, { "epoch": 2.03, "learning_rate": 2.6019066044653863e-05, "loss": 0.0354, "step": 43525 }, { "epoch": 2.03, "learning_rate": 2.6018597974817846e-05, "loss": 0.0737, "step": 43530 }, { "epoch": 2.03, "learning_rate": 2.6018129904981826e-05, "loss": 0.0782, "step": 43535 }, { "epoch": 2.03, "learning_rate": 2.6017661835145805e-05, "loss": 0.0542, "step": 43540 }, { "epoch": 2.03, "learning_rate": 2.6017193765309785e-05, "loss": 0.1266, "step": 43545 }, { "epoch": 2.03, "learning_rate": 2.6016725695473765e-05, "loss": 0.1402, "step": 43550 }, { "epoch": 2.03, "learning_rate": 2.6016257625637745e-05, "loss": 0.1939, "step": 43555 }, { "epoch": 2.03, "learning_rate": 2.6015789555801725e-05, "loss": 0.2771, "step": 43560 }, { "epoch": 2.03, "learning_rate": 2.6015321485965704e-05, "loss": 0.2739, "step": 43565 }, { "epoch": 2.03, "learning_rate": 2.6014853416129688e-05, "loss": 0.0758, "step": 43570 }, { "epoch": 2.03, "learning_rate": 2.6014385346293667e-05, "loss": 0.0429, "step": 43575 }, { "epoch": 2.03, "learning_rate": 2.6013917276457647e-05, "loss": 0.0594, "step": 43580 }, { "epoch": 2.03, "learning_rate": 2.601344920662163e-05, "loss": 0.0997, "step": 43585 }, { "epoch": 2.03, "learning_rate": 2.601298113678561e-05, "loss": 0.1139, "step": 43590 }, { "epoch": 2.03, "learning_rate": 2.601251306694959e-05, "loss": 0.1391, "step": 43595 }, { "epoch": 2.03, "learning_rate": 2.601204499711357e-05, "loss": 0.1522, "step": 43600 }, { "epoch": 2.03, "learning_rate": 2.6011576927277553e-05, "loss": 0.1734, "step": 43605 }, { "epoch": 2.03, "learning_rate": 2.601110885744153e-05, "loss": 0.3617, "step": 43610 }, { "epoch": 2.04, "learning_rate": 2.601064078760551e-05, "loss": 0.292, "step": 43615 }, { "epoch": 2.04, "learning_rate": 2.601017271776949e-05, "loss": 0.0504, "step": 43620 }, { "epoch": 2.04, "learning_rate": 2.6009704647933472e-05, "loss": 0.0315, "step": 43625 }, { "epoch": 2.04, "learning_rate": 2.6009236578097452e-05, "loss": 0.0434, "step": 43630 }, { "epoch": 2.04, "learning_rate": 2.6008768508261432e-05, "loss": 0.0796, "step": 43635 }, { "epoch": 2.04, "learning_rate": 2.6008300438425415e-05, "loss": 0.0695, "step": 43640 }, { "epoch": 2.04, "learning_rate": 2.6007832368589395e-05, "loss": 0.0673, "step": 43645 }, { "epoch": 2.04, "learning_rate": 2.6007364298753375e-05, "loss": 0.1279, "step": 43650 }, { "epoch": 2.04, "learning_rate": 2.6006896228917355e-05, "loss": 0.203, "step": 43655 }, { "epoch": 2.04, "learning_rate": 2.6006428159081338e-05, "loss": 0.2057, "step": 43660 }, { "epoch": 2.04, "learning_rate": 2.6005960089245318e-05, "loss": 0.2759, "step": 43665 }, { "epoch": 2.04, "learning_rate": 2.6005492019409298e-05, "loss": 0.0515, "step": 43670 }, { "epoch": 2.04, "learning_rate": 2.6005023949573274e-05, "loss": 0.0193, "step": 43675 }, { "epoch": 2.04, "learning_rate": 2.6004555879737257e-05, "loss": 0.0598, "step": 43680 }, { "epoch": 2.04, "learning_rate": 2.6004087809901237e-05, "loss": 0.0771, "step": 43685 }, { "epoch": 2.04, "learning_rate": 2.6003619740065217e-05, "loss": 0.0433, "step": 43690 }, { "epoch": 2.04, "learning_rate": 2.60031516702292e-05, "loss": 0.1658, "step": 43695 }, { "epoch": 2.04, "learning_rate": 2.600268360039318e-05, "loss": 0.0976, "step": 43700 }, { "epoch": 2.04, "learning_rate": 2.600221553055716e-05, "loss": 0.222, "step": 43705 }, { "epoch": 2.04, "learning_rate": 2.600174746072114e-05, "loss": 0.2616, "step": 43710 }, { "epoch": 2.04, "learning_rate": 2.6001279390885123e-05, "loss": 0.2958, "step": 43715 }, { "epoch": 2.04, "learning_rate": 2.6000811321049103e-05, "loss": 0.0172, "step": 43720 }, { "epoch": 2.04, "learning_rate": 2.6000343251213082e-05, "loss": 0.0272, "step": 43725 }, { "epoch": 2.04, "learning_rate": 2.5999875181377062e-05, "loss": 0.0986, "step": 43730 }, { "epoch": 2.04, "learning_rate": 2.5999407111541042e-05, "loss": 0.0818, "step": 43735 }, { "epoch": 2.04, "learning_rate": 2.5998939041705022e-05, "loss": 0.099, "step": 43740 }, { "epoch": 2.04, "learning_rate": 2.5998470971869e-05, "loss": 0.0866, "step": 43745 }, { "epoch": 2.04, "learning_rate": 2.599800290203298e-05, "loss": 0.1462, "step": 43750 }, { "epoch": 2.04, "learning_rate": 2.5997534832196965e-05, "loss": 0.1062, "step": 43755 }, { "epoch": 2.04, "learning_rate": 2.5997066762360944e-05, "loss": 0.1921, "step": 43760 }, { "epoch": 2.04, "learning_rate": 2.5996598692524924e-05, "loss": 0.3237, "step": 43765 }, { "epoch": 2.04, "learning_rate": 2.5996130622688907e-05, "loss": 0.0672, "step": 43770 }, { "epoch": 2.04, "learning_rate": 2.5995662552852887e-05, "loss": 0.0397, "step": 43775 }, { "epoch": 2.04, "learning_rate": 2.5995194483016867e-05, "loss": 0.048, "step": 43780 }, { "epoch": 2.04, "learning_rate": 2.5994726413180847e-05, "loss": 0.0685, "step": 43785 }, { "epoch": 2.04, "learning_rate": 2.599425834334483e-05, "loss": 0.0745, "step": 43790 }, { "epoch": 2.04, "learning_rate": 2.599379027350881e-05, "loss": 0.0742, "step": 43795 }, { "epoch": 2.04, "learning_rate": 2.5993322203672786e-05, "loss": 0.1082, "step": 43800 }, { "epoch": 2.04, "learning_rate": 2.5992854133836766e-05, "loss": 0.2451, "step": 43805 }, { "epoch": 2.04, "learning_rate": 2.599238606400075e-05, "loss": 0.2064, "step": 43810 }, { "epoch": 2.04, "learning_rate": 2.599191799416473e-05, "loss": 0.3373, "step": 43815 }, { "epoch": 2.04, "learning_rate": 2.599144992432871e-05, "loss": 0.0609, "step": 43820 }, { "epoch": 2.04, "learning_rate": 2.5990981854492692e-05, "loss": 0.0868, "step": 43825 }, { "epoch": 2.05, "learning_rate": 2.5990513784656672e-05, "loss": 0.0606, "step": 43830 }, { "epoch": 2.05, "learning_rate": 2.5990045714820652e-05, "loss": 0.1133, "step": 43835 }, { "epoch": 2.05, "learning_rate": 2.5989577644984632e-05, "loss": 0.0708, "step": 43840 }, { "epoch": 2.05, "learning_rate": 2.5989109575148615e-05, "loss": 0.1147, "step": 43845 }, { "epoch": 2.05, "learning_rate": 2.5988641505312595e-05, "loss": 0.1057, "step": 43850 }, { "epoch": 2.05, "learning_rate": 2.5988173435476575e-05, "loss": 0.1761, "step": 43855 }, { "epoch": 2.05, "learning_rate": 2.5987705365640554e-05, "loss": 0.2206, "step": 43860 }, { "epoch": 2.05, "learning_rate": 2.5987237295804534e-05, "loss": 0.4006, "step": 43865 }, { "epoch": 2.05, "learning_rate": 2.5986769225968514e-05, "loss": 0.0626, "step": 43870 }, { "epoch": 2.05, "learning_rate": 2.5986301156132494e-05, "loss": 0.0403, "step": 43875 }, { "epoch": 2.05, "learning_rate": 2.5985833086296477e-05, "loss": 0.0535, "step": 43880 }, { "epoch": 2.05, "learning_rate": 2.5985365016460457e-05, "loss": 0.0547, "step": 43885 }, { "epoch": 2.05, "learning_rate": 2.5984896946624437e-05, "loss": 0.1163, "step": 43890 }, { "epoch": 2.05, "learning_rate": 2.5984428876788416e-05, "loss": 0.1051, "step": 43895 }, { "epoch": 2.05, "learning_rate": 2.59839608069524e-05, "loss": 0.1141, "step": 43900 }, { "epoch": 2.05, "learning_rate": 2.598349273711638e-05, "loss": 0.2341, "step": 43905 }, { "epoch": 2.05, "learning_rate": 2.598302466728036e-05, "loss": 0.2875, "step": 43910 }, { "epoch": 2.05, "learning_rate": 2.598255659744434e-05, "loss": 0.212, "step": 43915 }, { "epoch": 2.05, "learning_rate": 2.5982088527608322e-05, "loss": 0.0225, "step": 43920 }, { "epoch": 2.05, "learning_rate": 2.59816204577723e-05, "loss": 0.0303, "step": 43925 }, { "epoch": 2.05, "learning_rate": 2.598115238793628e-05, "loss": 0.0887, "step": 43930 }, { "epoch": 2.05, "learning_rate": 2.598068431810026e-05, "loss": 0.0739, "step": 43935 }, { "epoch": 2.05, "learning_rate": 2.598021624826424e-05, "loss": 0.0731, "step": 43940 }, { "epoch": 2.05, "learning_rate": 2.597974817842822e-05, "loss": 0.1504, "step": 43945 }, { "epoch": 2.05, "learning_rate": 2.59792801085922e-05, "loss": 0.1396, "step": 43950 }, { "epoch": 2.05, "learning_rate": 2.5978812038756184e-05, "loss": 0.1312, "step": 43955 }, { "epoch": 2.05, "learning_rate": 2.5978343968920164e-05, "loss": 0.3022, "step": 43960 }, { "epoch": 2.05, "learning_rate": 2.5977875899084144e-05, "loss": 0.4683, "step": 43965 }, { "epoch": 2.05, "learning_rate": 2.5977407829248124e-05, "loss": 0.081, "step": 43970 }, { "epoch": 2.05, "learning_rate": 2.5976939759412107e-05, "loss": 0.0326, "step": 43975 }, { "epoch": 2.05, "learning_rate": 2.5976471689576087e-05, "loss": 0.0402, "step": 43980 }, { "epoch": 2.05, "learning_rate": 2.5976003619740067e-05, "loss": 0.0648, "step": 43985 }, { "epoch": 2.05, "learning_rate": 2.5975535549904043e-05, "loss": 0.1982, "step": 43990 }, { "epoch": 2.05, "learning_rate": 2.5975067480068026e-05, "loss": 0.1179, "step": 43995 }, { "epoch": 2.05, "learning_rate": 2.5974599410232006e-05, "loss": 0.127, "step": 44000 }, { "epoch": 2.05, "learning_rate": 2.5974131340395986e-05, "loss": 0.1602, "step": 44005 }, { "epoch": 2.05, "learning_rate": 2.597366327055997e-05, "loss": 0.2638, "step": 44010 }, { "epoch": 2.05, "learning_rate": 2.597319520072395e-05, "loss": 0.2213, "step": 44015 }, { "epoch": 2.05, "learning_rate": 2.597272713088793e-05, "loss": 0.0252, "step": 44020 }, { "epoch": 2.05, "learning_rate": 2.597225906105191e-05, "loss": 0.0487, "step": 44025 }, { "epoch": 2.05, "learning_rate": 2.5971790991215892e-05, "loss": 0.0386, "step": 44030 }, { "epoch": 2.05, "learning_rate": 2.5971322921379872e-05, "loss": 0.0931, "step": 44035 }, { "epoch": 2.05, "learning_rate": 2.597085485154385e-05, "loss": 0.1834, "step": 44040 }, { "epoch": 2.06, "learning_rate": 2.597038678170783e-05, "loss": 0.2191, "step": 44045 }, { "epoch": 2.06, "learning_rate": 2.5969918711871815e-05, "loss": 0.1914, "step": 44050 }, { "epoch": 2.06, "learning_rate": 2.596945064203579e-05, "loss": 0.1308, "step": 44055 }, { "epoch": 2.06, "learning_rate": 2.596898257219977e-05, "loss": 0.3082, "step": 44060 }, { "epoch": 2.06, "learning_rate": 2.5968514502363754e-05, "loss": 0.2636, "step": 44065 }, { "epoch": 2.06, "learning_rate": 2.5968046432527734e-05, "loss": 0.1079, "step": 44070 }, { "epoch": 2.06, "learning_rate": 2.5967578362691714e-05, "loss": 0.0352, "step": 44075 }, { "epoch": 2.06, "learning_rate": 2.5967110292855693e-05, "loss": 0.057, "step": 44080 }, { "epoch": 2.06, "learning_rate": 2.5966642223019677e-05, "loss": 0.0525, "step": 44085 }, { "epoch": 2.06, "learning_rate": 2.5966174153183656e-05, "loss": 0.1108, "step": 44090 }, { "epoch": 2.06, "learning_rate": 2.5965706083347636e-05, "loss": 0.1278, "step": 44095 }, { "epoch": 2.06, "learning_rate": 2.5965238013511616e-05, "loss": 0.1354, "step": 44100 }, { "epoch": 2.06, "learning_rate": 2.59647699436756e-05, "loss": 0.159, "step": 44105 }, { "epoch": 2.06, "learning_rate": 2.596430187383958e-05, "loss": 0.2614, "step": 44110 }, { "epoch": 2.06, "learning_rate": 2.5963833804003556e-05, "loss": 0.2367, "step": 44115 }, { "epoch": 2.06, "learning_rate": 2.596336573416754e-05, "loss": 0.053, "step": 44120 }, { "epoch": 2.06, "learning_rate": 2.596289766433152e-05, "loss": 0.0542, "step": 44125 }, { "epoch": 2.06, "learning_rate": 2.59624295944955e-05, "loss": 0.0561, "step": 44130 }, { "epoch": 2.06, "learning_rate": 2.5961961524659478e-05, "loss": 0.0369, "step": 44135 }, { "epoch": 2.06, "learning_rate": 2.596149345482346e-05, "loss": 0.1458, "step": 44140 }, { "epoch": 2.06, "learning_rate": 2.596102538498744e-05, "loss": 0.0595, "step": 44145 }, { "epoch": 2.06, "learning_rate": 2.596055731515142e-05, "loss": 0.2433, "step": 44150 }, { "epoch": 2.06, "learning_rate": 2.59600892453154e-05, "loss": 0.1633, "step": 44155 }, { "epoch": 2.06, "learning_rate": 2.5959621175479384e-05, "loss": 0.2184, "step": 44160 }, { "epoch": 2.06, "learning_rate": 2.5959153105643364e-05, "loss": 0.3487, "step": 44165 }, { "epoch": 2.06, "learning_rate": 2.5958685035807344e-05, "loss": 0.0224, "step": 44170 }, { "epoch": 2.06, "learning_rate": 2.5958216965971324e-05, "loss": 0.0847, "step": 44175 }, { "epoch": 2.06, "learning_rate": 2.5957748896135303e-05, "loss": 0.0655, "step": 44180 }, { "epoch": 2.06, "learning_rate": 2.5957280826299283e-05, "loss": 0.0575, "step": 44185 }, { "epoch": 2.06, "learning_rate": 2.5956812756463263e-05, "loss": 0.0914, "step": 44190 }, { "epoch": 2.06, "learning_rate": 2.5956344686627246e-05, "loss": 0.1687, "step": 44195 }, { "epoch": 2.06, "learning_rate": 2.5955876616791226e-05, "loss": 0.1551, "step": 44200 }, { "epoch": 2.06, "learning_rate": 2.5955408546955206e-05, "loss": 0.1556, "step": 44205 }, { "epoch": 2.06, "learning_rate": 2.5954940477119186e-05, "loss": 0.234, "step": 44210 }, { "epoch": 2.06, "learning_rate": 2.595447240728317e-05, "loss": 0.2907, "step": 44215 }, { "epoch": 2.06, "learning_rate": 2.595400433744715e-05, "loss": 0.0634, "step": 44220 }, { "epoch": 2.06, "learning_rate": 2.595353626761113e-05, "loss": 0.057, "step": 44225 }, { "epoch": 2.06, "learning_rate": 2.595306819777511e-05, "loss": 0.0898, "step": 44230 }, { "epoch": 2.06, "learning_rate": 2.595260012793909e-05, "loss": 0.0623, "step": 44235 }, { "epoch": 2.06, "learning_rate": 2.595213205810307e-05, "loss": 0.0834, "step": 44240 }, { "epoch": 2.06, "learning_rate": 2.5951663988267048e-05, "loss": 0.1182, "step": 44245 }, { "epoch": 2.06, "learning_rate": 2.595119591843103e-05, "loss": 0.1866, "step": 44250 }, { "epoch": 2.06, "learning_rate": 2.595072784859501e-05, "loss": 0.1783, "step": 44255 }, { "epoch": 2.07, "learning_rate": 2.595025977875899e-05, "loss": 0.2101, "step": 44260 }, { "epoch": 2.07, "learning_rate": 2.594979170892297e-05, "loss": 0.3375, "step": 44265 }, { "epoch": 2.07, "learning_rate": 2.5949323639086954e-05, "loss": 0.0616, "step": 44270 }, { "epoch": 2.07, "learning_rate": 2.5948855569250933e-05, "loss": 0.0168, "step": 44275 }, { "epoch": 2.07, "learning_rate": 2.5948387499414913e-05, "loss": 0.0396, "step": 44280 }, { "epoch": 2.07, "learning_rate": 2.5947919429578893e-05, "loss": 0.0761, "step": 44285 }, { "epoch": 2.07, "learning_rate": 2.5947451359742876e-05, "loss": 0.0863, "step": 44290 }, { "epoch": 2.07, "learning_rate": 2.5946983289906856e-05, "loss": 0.1398, "step": 44295 }, { "epoch": 2.07, "learning_rate": 2.5946515220070836e-05, "loss": 0.2071, "step": 44300 }, { "epoch": 2.07, "learning_rate": 2.5946047150234816e-05, "loss": 0.1878, "step": 44305 }, { "epoch": 2.07, "learning_rate": 2.5945579080398796e-05, "loss": 0.2988, "step": 44310 }, { "epoch": 2.07, "learning_rate": 2.5945111010562775e-05, "loss": 0.2748, "step": 44315 }, { "epoch": 2.07, "learning_rate": 2.5944642940726755e-05, "loss": 0.0561, "step": 44320 }, { "epoch": 2.07, "learning_rate": 2.594417487089074e-05, "loss": 0.0897, "step": 44325 }, { "epoch": 2.07, "learning_rate": 2.5943706801054718e-05, "loss": 0.025, "step": 44330 }, { "epoch": 2.07, "learning_rate": 2.5943238731218698e-05, "loss": 0.0864, "step": 44335 }, { "epoch": 2.07, "learning_rate": 2.5942770661382678e-05, "loss": 0.08, "step": 44340 }, { "epoch": 2.07, "learning_rate": 2.594230259154666e-05, "loss": 0.0743, "step": 44345 }, { "epoch": 2.07, "learning_rate": 2.594183452171064e-05, "loss": 0.1179, "step": 44350 }, { "epoch": 2.07, "learning_rate": 2.594136645187462e-05, "loss": 0.1469, "step": 44355 }, { "epoch": 2.07, "learning_rate": 2.59408983820386e-05, "loss": 0.2232, "step": 44360 }, { "epoch": 2.07, "learning_rate": 2.5940430312202584e-05, "loss": 0.223, "step": 44365 }, { "epoch": 2.07, "learning_rate": 2.593996224236656e-05, "loss": 0.0095, "step": 44370 }, { "epoch": 2.07, "learning_rate": 2.593949417253054e-05, "loss": 0.0184, "step": 44375 }, { "epoch": 2.07, "learning_rate": 2.5939026102694523e-05, "loss": 0.1017, "step": 44380 }, { "epoch": 2.07, "learning_rate": 2.5938558032858503e-05, "loss": 0.0862, "step": 44385 }, { "epoch": 2.07, "learning_rate": 2.5938089963022483e-05, "loss": 0.154, "step": 44390 }, { "epoch": 2.07, "learning_rate": 2.5937621893186463e-05, "loss": 0.1636, "step": 44395 }, { "epoch": 2.07, "learning_rate": 2.5937153823350446e-05, "loss": 0.1058, "step": 44400 }, { "epoch": 2.07, "learning_rate": 2.5936685753514426e-05, "loss": 0.234, "step": 44405 }, { "epoch": 2.07, "learning_rate": 2.5936217683678405e-05, "loss": 0.2028, "step": 44410 }, { "epoch": 2.07, "learning_rate": 2.5935749613842385e-05, "loss": 0.3016, "step": 44415 }, { "epoch": 2.07, "learning_rate": 2.593528154400637e-05, "loss": 0.0569, "step": 44420 }, { "epoch": 2.07, "learning_rate": 2.5934813474170348e-05, "loss": 0.0465, "step": 44425 }, { "epoch": 2.07, "learning_rate": 2.5934345404334328e-05, "loss": 0.0423, "step": 44430 }, { "epoch": 2.07, "learning_rate": 2.5933877334498308e-05, "loss": 0.0806, "step": 44435 }, { "epoch": 2.07, "learning_rate": 2.5933409264662288e-05, "loss": 0.1182, "step": 44440 }, { "epoch": 2.07, "learning_rate": 2.5932941194826268e-05, "loss": 0.1466, "step": 44445 }, { "epoch": 2.07, "learning_rate": 2.5932473124990247e-05, "loss": 0.2372, "step": 44450 }, { "epoch": 2.07, "learning_rate": 2.593200505515423e-05, "loss": 0.2064, "step": 44455 }, { "epoch": 2.07, "learning_rate": 2.593153698531821e-05, "loss": 0.3178, "step": 44460 }, { "epoch": 2.07, "learning_rate": 2.593106891548219e-05, "loss": 0.3023, "step": 44465 }, { "epoch": 2.08, "learning_rate": 2.593060084564617e-05, "loss": 0.0771, "step": 44470 }, { "epoch": 2.08, "learning_rate": 2.5930132775810153e-05, "loss": 0.029, "step": 44475 }, { "epoch": 2.08, "learning_rate": 2.5929664705974133e-05, "loss": 0.0605, "step": 44480 }, { "epoch": 2.08, "learning_rate": 2.5929196636138113e-05, "loss": 0.0319, "step": 44485 }, { "epoch": 2.08, "learning_rate": 2.5928728566302096e-05, "loss": 0.065, "step": 44490 }, { "epoch": 2.08, "learning_rate": 2.5928260496466073e-05, "loss": 0.0691, "step": 44495 }, { "epoch": 2.08, "learning_rate": 2.5927792426630052e-05, "loss": 0.1597, "step": 44500 }, { "epoch": 2.08, "learning_rate": 2.5927324356794032e-05, "loss": 0.1046, "step": 44505 }, { "epoch": 2.08, "learning_rate": 2.5926856286958015e-05, "loss": 0.1856, "step": 44510 }, { "epoch": 2.08, "learning_rate": 2.5926388217121995e-05, "loss": 0.3088, "step": 44515 }, { "epoch": 2.08, "learning_rate": 2.5925920147285975e-05, "loss": 0.0857, "step": 44520 }, { "epoch": 2.08, "learning_rate": 2.5925452077449955e-05, "loss": 0.0453, "step": 44525 }, { "epoch": 2.08, "learning_rate": 2.5924984007613938e-05, "loss": 0.0773, "step": 44530 }, { "epoch": 2.08, "learning_rate": 2.5924515937777918e-05, "loss": 0.0542, "step": 44535 }, { "epoch": 2.08, "learning_rate": 2.5924047867941898e-05, "loss": 0.0752, "step": 44540 }, { "epoch": 2.08, "learning_rate": 2.5923579798105877e-05, "loss": 0.119, "step": 44545 }, { "epoch": 2.08, "learning_rate": 2.592311172826986e-05, "loss": 0.1554, "step": 44550 }, { "epoch": 2.08, "learning_rate": 2.592264365843384e-05, "loss": 0.1366, "step": 44555 }, { "epoch": 2.08, "learning_rate": 2.5922175588597817e-05, "loss": 0.2023, "step": 44560 }, { "epoch": 2.08, "learning_rate": 2.59217075187618e-05, "loss": 0.3999, "step": 44565 }, { "epoch": 2.08, "learning_rate": 2.592123944892578e-05, "loss": 0.0511, "step": 44570 }, { "epoch": 2.08, "learning_rate": 2.592077137908976e-05, "loss": 0.0905, "step": 44575 }, { "epoch": 2.08, "learning_rate": 2.592030330925374e-05, "loss": 0.0634, "step": 44580 }, { "epoch": 2.08, "learning_rate": 2.5919835239417723e-05, "loss": 0.1139, "step": 44585 }, { "epoch": 2.08, "learning_rate": 2.5919367169581703e-05, "loss": 0.1004, "step": 44590 }, { "epoch": 2.08, "learning_rate": 2.5918899099745682e-05, "loss": 0.1277, "step": 44595 }, { "epoch": 2.08, "learning_rate": 2.5918431029909662e-05, "loss": 0.1798, "step": 44600 }, { "epoch": 2.08, "learning_rate": 2.5917962960073645e-05, "loss": 0.2032, "step": 44605 }, { "epoch": 2.08, "learning_rate": 2.5917494890237625e-05, "loss": 0.2548, "step": 44610 }, { "epoch": 2.08, "learning_rate": 2.5917026820401605e-05, "loss": 0.4322, "step": 44615 }, { "epoch": 2.08, "learning_rate": 2.5916558750565585e-05, "loss": 0.0485, "step": 44620 }, { "epoch": 2.08, "learning_rate": 2.5916090680729565e-05, "loss": 0.063, "step": 44625 }, { "epoch": 2.08, "learning_rate": 2.5915622610893545e-05, "loss": 0.0712, "step": 44630 }, { "epoch": 2.08, "learning_rate": 2.5915154541057524e-05, "loss": 0.0526, "step": 44635 }, { "epoch": 2.08, "learning_rate": 2.5914686471221508e-05, "loss": 0.1514, "step": 44640 }, { "epoch": 2.08, "learning_rate": 2.5914218401385487e-05, "loss": 0.058, "step": 44645 }, { "epoch": 2.08, "learning_rate": 2.5913750331549467e-05, "loss": 0.0908, "step": 44650 }, { "epoch": 2.08, "learning_rate": 2.5913282261713447e-05, "loss": 0.1926, "step": 44655 }, { "epoch": 2.08, "learning_rate": 2.591281419187743e-05, "loss": 0.3495, "step": 44660 }, { "epoch": 2.08, "learning_rate": 2.591234612204141e-05, "loss": 0.4036, "step": 44665 }, { "epoch": 2.08, "learning_rate": 2.591187805220539e-05, "loss": 0.0753, "step": 44670 }, { "epoch": 2.08, "learning_rate": 2.5911409982369373e-05, "loss": 0.0733, "step": 44675 }, { "epoch": 2.08, "learning_rate": 2.5910941912533353e-05, "loss": 0.0703, "step": 44680 }, { "epoch": 2.09, "learning_rate": 2.591047384269733e-05, "loss": 0.142, "step": 44685 }, { "epoch": 2.09, "learning_rate": 2.591000577286131e-05, "loss": 0.1062, "step": 44690 }, { "epoch": 2.09, "learning_rate": 2.5909537703025292e-05, "loss": 0.0661, "step": 44695 }, { "epoch": 2.09, "learning_rate": 2.5909069633189272e-05, "loss": 0.1498, "step": 44700 }, { "epoch": 2.09, "learning_rate": 2.5908601563353252e-05, "loss": 0.1408, "step": 44705 }, { "epoch": 2.09, "learning_rate": 2.5908133493517232e-05, "loss": 0.2573, "step": 44710 }, { "epoch": 2.09, "learning_rate": 2.5907665423681215e-05, "loss": 0.2723, "step": 44715 }, { "epoch": 2.09, "learning_rate": 2.5907197353845195e-05, "loss": 0.0442, "step": 44720 }, { "epoch": 2.09, "learning_rate": 2.5906729284009175e-05, "loss": 0.0766, "step": 44725 }, { "epoch": 2.09, "learning_rate": 2.5906261214173154e-05, "loss": 0.0274, "step": 44730 }, { "epoch": 2.09, "learning_rate": 2.5905793144337138e-05, "loss": 0.1058, "step": 44735 }, { "epoch": 2.09, "learning_rate": 2.5905325074501117e-05, "loss": 0.1708, "step": 44740 }, { "epoch": 2.09, "learning_rate": 2.5904857004665097e-05, "loss": 0.2116, "step": 44745 }, { "epoch": 2.09, "learning_rate": 2.5904388934829077e-05, "loss": 0.1723, "step": 44750 }, { "epoch": 2.09, "learning_rate": 2.5903920864993057e-05, "loss": 0.1797, "step": 44755 }, { "epoch": 2.09, "learning_rate": 2.5903452795157037e-05, "loss": 0.2608, "step": 44760 }, { "epoch": 2.09, "learning_rate": 2.5902984725321017e-05, "loss": 0.2864, "step": 44765 }, { "epoch": 2.09, "learning_rate": 2.5902516655485e-05, "loss": 0.1087, "step": 44770 }, { "epoch": 2.09, "learning_rate": 2.590204858564898e-05, "loss": 0.0488, "step": 44775 }, { "epoch": 2.09, "learning_rate": 2.590158051581296e-05, "loss": 0.0883, "step": 44780 }, { "epoch": 2.09, "learning_rate": 2.590111244597694e-05, "loss": 0.0909, "step": 44785 }, { "epoch": 2.09, "learning_rate": 2.5900644376140922e-05, "loss": 0.0817, "step": 44790 }, { "epoch": 2.09, "learning_rate": 2.5900176306304902e-05, "loss": 0.1462, "step": 44795 }, { "epoch": 2.09, "learning_rate": 2.5899708236468882e-05, "loss": 0.1889, "step": 44800 }, { "epoch": 2.09, "learning_rate": 2.5899240166632865e-05, "loss": 0.1847, "step": 44805 }, { "epoch": 2.09, "learning_rate": 2.5898772096796842e-05, "loss": 0.2511, "step": 44810 }, { "epoch": 2.09, "learning_rate": 2.589830402696082e-05, "loss": 0.2548, "step": 44815 }, { "epoch": 2.09, "learning_rate": 2.58978359571248e-05, "loss": 0.0699, "step": 44820 }, { "epoch": 2.09, "learning_rate": 2.5897367887288785e-05, "loss": 0.0488, "step": 44825 }, { "epoch": 2.09, "learning_rate": 2.5896899817452764e-05, "loss": 0.0716, "step": 44830 }, { "epoch": 2.09, "learning_rate": 2.5896431747616744e-05, "loss": 0.0266, "step": 44835 }, { "epoch": 2.09, "learning_rate": 2.5895963677780724e-05, "loss": 0.1222, "step": 44840 }, { "epoch": 2.09, "learning_rate": 2.5895495607944707e-05, "loss": 0.1125, "step": 44845 }, { "epoch": 2.09, "learning_rate": 2.5895027538108687e-05, "loss": 0.1399, "step": 44850 }, { "epoch": 2.09, "learning_rate": 2.5894559468272667e-05, "loss": 0.2239, "step": 44855 }, { "epoch": 2.09, "learning_rate": 2.589409139843665e-05, "loss": 0.3015, "step": 44860 }, { "epoch": 2.09, "learning_rate": 2.589362332860063e-05, "loss": 0.2039, "step": 44865 }, { "epoch": 2.09, "learning_rate": 2.589315525876461e-05, "loss": 0.0715, "step": 44870 }, { "epoch": 2.09, "learning_rate": 2.5892687188928586e-05, "loss": 0.094, "step": 44875 }, { "epoch": 2.09, "learning_rate": 2.589221911909257e-05, "loss": 0.0492, "step": 44880 }, { "epoch": 2.09, "learning_rate": 2.589175104925655e-05, "loss": 0.0639, "step": 44885 }, { "epoch": 2.09, "learning_rate": 2.589128297942053e-05, "loss": 0.0756, "step": 44890 }, { "epoch": 2.09, "learning_rate": 2.589081490958451e-05, "loss": 0.1301, "step": 44895 }, { "epoch": 2.1, "learning_rate": 2.5890346839748492e-05, "loss": 0.1054, "step": 44900 }, { "epoch": 2.1, "learning_rate": 2.5889878769912472e-05, "loss": 0.2198, "step": 44905 }, { "epoch": 2.1, "learning_rate": 2.588941070007645e-05, "loss": 0.3217, "step": 44910 }, { "epoch": 2.1, "learning_rate": 2.5888942630240435e-05, "loss": 0.1799, "step": 44915 }, { "epoch": 2.1, "learning_rate": 2.5888474560404415e-05, "loss": 0.0618, "step": 44920 }, { "epoch": 2.1, "learning_rate": 2.5888006490568394e-05, "loss": 0.0614, "step": 44925 }, { "epoch": 2.1, "learning_rate": 2.5887538420732374e-05, "loss": 0.0542, "step": 44930 }, { "epoch": 2.1, "learning_rate": 2.5887070350896354e-05, "loss": 0.0665, "step": 44935 }, { "epoch": 2.1, "learning_rate": 2.5886602281060334e-05, "loss": 0.0419, "step": 44940 }, { "epoch": 2.1, "learning_rate": 2.5886134211224314e-05, "loss": 0.0984, "step": 44945 }, { "epoch": 2.1, "learning_rate": 2.5885666141388294e-05, "loss": 0.0959, "step": 44950 }, { "epoch": 2.1, "learning_rate": 2.5885198071552277e-05, "loss": 0.1719, "step": 44955 }, { "epoch": 2.1, "learning_rate": 2.5884730001716257e-05, "loss": 0.1955, "step": 44960 }, { "epoch": 2.1, "learning_rate": 2.5884261931880236e-05, "loss": 0.199, "step": 44965 }, { "epoch": 2.1, "learning_rate": 2.5883793862044216e-05, "loss": 0.0628, "step": 44970 }, { "epoch": 2.1, "learning_rate": 2.58833257922082e-05, "loss": 0.0414, "step": 44975 }, { "epoch": 2.1, "learning_rate": 2.588285772237218e-05, "loss": 0.0432, "step": 44980 }, { "epoch": 2.1, "learning_rate": 2.588238965253616e-05, "loss": 0.1006, "step": 44985 }, { "epoch": 2.1, "learning_rate": 2.5881921582700142e-05, "loss": 0.1676, "step": 44990 }, { "epoch": 2.1, "learning_rate": 2.5881453512864122e-05, "loss": 0.1348, "step": 44995 }, { "epoch": 2.1, "learning_rate": 2.58809854430281e-05, "loss": 0.1127, "step": 45000 }, { "epoch": 2.1, "learning_rate": 2.5880517373192078e-05, "loss": 0.1484, "step": 45005 }, { "epoch": 2.1, "learning_rate": 2.588004930335606e-05, "loss": 0.2177, "step": 45010 }, { "epoch": 2.1, "learning_rate": 2.587958123352004e-05, "loss": 0.3085, "step": 45015 }, { "epoch": 2.1, "learning_rate": 2.587911316368402e-05, "loss": 0.06, "step": 45020 }, { "epoch": 2.1, "learning_rate": 2.5878645093848e-05, "loss": 0.064, "step": 45025 }, { "epoch": 2.1, "learning_rate": 2.5878177024011984e-05, "loss": 0.0564, "step": 45030 }, { "epoch": 2.1, "learning_rate": 2.5877708954175964e-05, "loss": 0.0696, "step": 45035 }, { "epoch": 2.1, "learning_rate": 2.5877240884339944e-05, "loss": 0.0959, "step": 45040 }, { "epoch": 2.1, "learning_rate": 2.5876772814503927e-05, "loss": 0.1166, "step": 45045 }, { "epoch": 2.1, "learning_rate": 2.5876304744667907e-05, "loss": 0.0576, "step": 45050 }, { "epoch": 2.1, "learning_rate": 2.5875836674831887e-05, "loss": 0.1842, "step": 45055 }, { "epoch": 2.1, "learning_rate": 2.5875368604995866e-05, "loss": 0.148, "step": 45060 }, { "epoch": 2.1, "learning_rate": 2.5874900535159846e-05, "loss": 0.2105, "step": 45065 }, { "epoch": 2.1, "learning_rate": 2.5874432465323826e-05, "loss": 0.1142, "step": 45070 }, { "epoch": 2.1, "learning_rate": 2.5873964395487806e-05, "loss": 0.0621, "step": 45075 }, { "epoch": 2.1, "learning_rate": 2.5873496325651786e-05, "loss": 0.1044, "step": 45080 }, { "epoch": 2.1, "learning_rate": 2.587302825581577e-05, "loss": 0.0424, "step": 45085 }, { "epoch": 2.1, "learning_rate": 2.587256018597975e-05, "loss": 0.0724, "step": 45090 }, { "epoch": 2.1, "learning_rate": 2.587209211614373e-05, "loss": 0.0786, "step": 45095 }, { "epoch": 2.1, "learning_rate": 2.5871624046307712e-05, "loss": 0.2606, "step": 45100 }, { "epoch": 2.1, "learning_rate": 2.587115597647169e-05, "loss": 0.2202, "step": 45105 }, { "epoch": 2.1, "learning_rate": 2.587068790663567e-05, "loss": 0.2612, "step": 45110 }, { "epoch": 2.11, "learning_rate": 2.587021983679965e-05, "loss": 0.3177, "step": 45115 }, { "epoch": 2.11, "learning_rate": 2.5869751766963634e-05, "loss": 0.0566, "step": 45120 }, { "epoch": 2.11, "learning_rate": 2.586928369712761e-05, "loss": 0.017, "step": 45125 }, { "epoch": 2.11, "learning_rate": 2.586881562729159e-05, "loss": 0.0516, "step": 45130 }, { "epoch": 2.11, "learning_rate": 2.586834755745557e-05, "loss": 0.0864, "step": 45135 }, { "epoch": 2.11, "learning_rate": 2.5867879487619554e-05, "loss": 0.0942, "step": 45140 }, { "epoch": 2.11, "learning_rate": 2.5867411417783534e-05, "loss": 0.1463, "step": 45145 }, { "epoch": 2.11, "learning_rate": 2.5866943347947513e-05, "loss": 0.1791, "step": 45150 }, { "epoch": 2.11, "learning_rate": 2.5866475278111493e-05, "loss": 0.1478, "step": 45155 }, { "epoch": 2.11, "learning_rate": 2.5866007208275476e-05, "loss": 0.139, "step": 45160 }, { "epoch": 2.11, "learning_rate": 2.5865539138439456e-05, "loss": 0.2919, "step": 45165 }, { "epoch": 2.11, "learning_rate": 2.5865071068603436e-05, "loss": 0.0715, "step": 45170 }, { "epoch": 2.11, "learning_rate": 2.586460299876742e-05, "loss": 0.0356, "step": 45175 }, { "epoch": 2.11, "learning_rate": 2.58641349289314e-05, "loss": 0.086, "step": 45180 }, { "epoch": 2.11, "learning_rate": 2.586366685909538e-05, "loss": 0.0643, "step": 45185 }, { "epoch": 2.11, "learning_rate": 2.5863198789259355e-05, "loss": 0.079, "step": 45190 }, { "epoch": 2.11, "learning_rate": 2.586273071942334e-05, "loss": 0.089, "step": 45195 }, { "epoch": 2.11, "learning_rate": 2.5862262649587318e-05, "loss": 0.1203, "step": 45200 }, { "epoch": 2.11, "learning_rate": 2.5861794579751298e-05, "loss": 0.2201, "step": 45205 }, { "epoch": 2.11, "learning_rate": 2.5861326509915278e-05, "loss": 0.2835, "step": 45210 }, { "epoch": 2.11, "learning_rate": 2.586085844007926e-05, "loss": 0.2557, "step": 45215 }, { "epoch": 2.11, "learning_rate": 2.586039037024324e-05, "loss": 0.0483, "step": 45220 }, { "epoch": 2.11, "learning_rate": 2.585992230040722e-05, "loss": 0.028, "step": 45225 }, { "epoch": 2.11, "learning_rate": 2.5859454230571204e-05, "loss": 0.0995, "step": 45230 }, { "epoch": 2.11, "learning_rate": 2.5858986160735184e-05, "loss": 0.125, "step": 45235 }, { "epoch": 2.11, "learning_rate": 2.5858518090899164e-05, "loss": 0.0796, "step": 45240 }, { "epoch": 2.11, "learning_rate": 2.5858050021063143e-05, "loss": 0.079, "step": 45245 }, { "epoch": 2.11, "learning_rate": 2.5857581951227127e-05, "loss": 0.1018, "step": 45250 }, { "epoch": 2.11, "learning_rate": 2.5857113881391103e-05, "loss": 0.1585, "step": 45255 }, { "epoch": 2.11, "learning_rate": 2.5856645811555083e-05, "loss": 0.2384, "step": 45260 }, { "epoch": 2.11, "learning_rate": 2.5856177741719063e-05, "loss": 0.3055, "step": 45265 }, { "epoch": 2.11, "learning_rate": 2.5855709671883046e-05, "loss": 0.0666, "step": 45270 }, { "epoch": 2.11, "learning_rate": 2.5855241602047026e-05, "loss": 0.1191, "step": 45275 }, { "epoch": 2.11, "learning_rate": 2.5854773532211006e-05, "loss": 0.083, "step": 45280 }, { "epoch": 2.11, "learning_rate": 2.585430546237499e-05, "loss": 0.0824, "step": 45285 }, { "epoch": 2.11, "learning_rate": 2.585383739253897e-05, "loss": 0.1073, "step": 45290 }, { "epoch": 2.11, "learning_rate": 2.585336932270295e-05, "loss": 0.1634, "step": 45295 }, { "epoch": 2.11, "learning_rate": 2.5852901252866928e-05, "loss": 0.1676, "step": 45300 }, { "epoch": 2.11, "learning_rate": 2.585243318303091e-05, "loss": 0.2054, "step": 45305 }, { "epoch": 2.11, "learning_rate": 2.585196511319489e-05, "loss": 0.1974, "step": 45310 }, { "epoch": 2.11, "learning_rate": 2.5851497043358868e-05, "loss": 0.3973, "step": 45315 }, { "epoch": 2.11, "learning_rate": 2.5851028973522847e-05, "loss": 0.0979, "step": 45320 }, { "epoch": 2.11, "learning_rate": 2.585056090368683e-05, "loss": 0.0769, "step": 45325 }, { "epoch": 2.12, "learning_rate": 2.585009283385081e-05, "loss": 0.0697, "step": 45330 }, { "epoch": 2.12, "learning_rate": 2.584962476401479e-05, "loss": 0.0937, "step": 45335 }, { "epoch": 2.12, "learning_rate": 2.584915669417877e-05, "loss": 0.0941, "step": 45340 }, { "epoch": 2.12, "learning_rate": 2.5848688624342753e-05, "loss": 0.075, "step": 45345 }, { "epoch": 2.12, "learning_rate": 2.5848220554506733e-05, "loss": 0.0991, "step": 45350 }, { "epoch": 2.12, "learning_rate": 2.5847752484670713e-05, "loss": 0.2013, "step": 45355 }, { "epoch": 2.12, "learning_rate": 2.5847284414834696e-05, "loss": 0.2069, "step": 45360 }, { "epoch": 2.12, "learning_rate": 2.5846816344998676e-05, "loss": 0.223, "step": 45365 }, { "epoch": 2.12, "learning_rate": 2.5846348275162656e-05, "loss": 0.0948, "step": 45370 }, { "epoch": 2.12, "learning_rate": 2.5845880205326636e-05, "loss": 0.0327, "step": 45375 }, { "epoch": 2.12, "learning_rate": 2.5845412135490615e-05, "loss": 0.0618, "step": 45380 }, { "epoch": 2.12, "learning_rate": 2.5844944065654595e-05, "loss": 0.2013, "step": 45385 }, { "epoch": 2.12, "learning_rate": 2.5844475995818575e-05, "loss": 0.0632, "step": 45390 }, { "epoch": 2.12, "learning_rate": 2.5844007925982555e-05, "loss": 0.1223, "step": 45395 }, { "epoch": 2.12, "learning_rate": 2.5843539856146538e-05, "loss": 0.0794, "step": 45400 }, { "epoch": 2.12, "learning_rate": 2.5843071786310518e-05, "loss": 0.1949, "step": 45405 }, { "epoch": 2.12, "learning_rate": 2.5842603716474498e-05, "loss": 0.2036, "step": 45410 }, { "epoch": 2.12, "learning_rate": 2.584213564663848e-05, "loss": 0.1645, "step": 45415 }, { "epoch": 2.12, "learning_rate": 2.584166757680246e-05, "loss": 0.0458, "step": 45420 }, { "epoch": 2.12, "learning_rate": 2.584119950696644e-05, "loss": 0.041, "step": 45425 }, { "epoch": 2.12, "learning_rate": 2.584073143713042e-05, "loss": 0.0514, "step": 45430 }, { "epoch": 2.12, "learning_rate": 2.5840263367294404e-05, "loss": 0.0592, "step": 45435 }, { "epoch": 2.12, "learning_rate": 2.5839795297458383e-05, "loss": 0.1285, "step": 45440 }, { "epoch": 2.12, "learning_rate": 2.583932722762236e-05, "loss": 0.089, "step": 45445 }, { "epoch": 2.12, "learning_rate": 2.583885915778634e-05, "loss": 0.1468, "step": 45450 }, { "epoch": 2.12, "learning_rate": 2.5838391087950323e-05, "loss": 0.2113, "step": 45455 }, { "epoch": 2.12, "learning_rate": 2.5837923018114303e-05, "loss": 0.292, "step": 45460 }, { "epoch": 2.12, "learning_rate": 2.5837454948278283e-05, "loss": 0.3104, "step": 45465 }, { "epoch": 2.12, "learning_rate": 2.5836986878442266e-05, "loss": 0.115, "step": 45470 }, { "epoch": 2.12, "learning_rate": 2.5836518808606246e-05, "loss": 0.0699, "step": 45475 }, { "epoch": 2.12, "learning_rate": 2.5836050738770225e-05, "loss": 0.0519, "step": 45480 }, { "epoch": 2.12, "learning_rate": 2.5835582668934205e-05, "loss": 0.0238, "step": 45485 }, { "epoch": 2.12, "learning_rate": 2.583511459909819e-05, "loss": 0.1441, "step": 45490 }, { "epoch": 2.12, "learning_rate": 2.5834646529262168e-05, "loss": 0.0614, "step": 45495 }, { "epoch": 2.12, "learning_rate": 2.5834178459426148e-05, "loss": 0.1103, "step": 45500 }, { "epoch": 2.12, "learning_rate": 2.5833710389590124e-05, "loss": 0.1489, "step": 45505 }, { "epoch": 2.12, "learning_rate": 2.5833242319754108e-05, "loss": 0.3097, "step": 45510 }, { "epoch": 2.12, "learning_rate": 2.5832774249918087e-05, "loss": 0.365, "step": 45515 }, { "epoch": 2.12, "learning_rate": 2.5832306180082067e-05, "loss": 0.0619, "step": 45520 }, { "epoch": 2.12, "learning_rate": 2.583183811024605e-05, "loss": 0.0363, "step": 45525 }, { "epoch": 2.12, "learning_rate": 2.583137004041003e-05, "loss": 0.0892, "step": 45530 }, { "epoch": 2.12, "learning_rate": 2.583090197057401e-05, "loss": 0.108, "step": 45535 }, { "epoch": 2.12, "learning_rate": 2.583043390073799e-05, "loss": 0.0585, "step": 45540 }, { "epoch": 2.13, "learning_rate": 2.5829965830901973e-05, "loss": 0.0997, "step": 45545 }, { "epoch": 2.13, "learning_rate": 2.5829497761065953e-05, "loss": 0.1027, "step": 45550 }, { "epoch": 2.13, "learning_rate": 2.5829029691229933e-05, "loss": 0.1712, "step": 45555 }, { "epoch": 2.13, "learning_rate": 2.5828561621393913e-05, "loss": 0.2349, "step": 45560 }, { "epoch": 2.13, "learning_rate": 2.5828093551557896e-05, "loss": 0.2188, "step": 45565 }, { "epoch": 2.13, "learning_rate": 2.5827625481721872e-05, "loss": 0.0574, "step": 45570 }, { "epoch": 2.13, "learning_rate": 2.5827157411885852e-05, "loss": 0.0386, "step": 45575 }, { "epoch": 2.13, "learning_rate": 2.5826689342049832e-05, "loss": 0.0655, "step": 45580 }, { "epoch": 2.13, "learning_rate": 2.5826221272213815e-05, "loss": 0.0543, "step": 45585 }, { "epoch": 2.13, "learning_rate": 2.5825753202377795e-05, "loss": 0.123, "step": 45590 }, { "epoch": 2.13, "learning_rate": 2.5825285132541775e-05, "loss": 0.1047, "step": 45595 }, { "epoch": 2.13, "learning_rate": 2.5824817062705758e-05, "loss": 0.1185, "step": 45600 }, { "epoch": 2.13, "learning_rate": 2.5824348992869738e-05, "loss": 0.2249, "step": 45605 }, { "epoch": 2.13, "learning_rate": 2.5823880923033718e-05, "loss": 0.2311, "step": 45610 }, { "epoch": 2.13, "learning_rate": 2.5823412853197697e-05, "loss": 0.2489, "step": 45615 }, { "epoch": 2.13, "learning_rate": 2.582294478336168e-05, "loss": 0.0881, "step": 45620 }, { "epoch": 2.13, "learning_rate": 2.582247671352566e-05, "loss": 0.0905, "step": 45625 }, { "epoch": 2.13, "learning_rate": 2.582200864368964e-05, "loss": 0.0707, "step": 45630 }, { "epoch": 2.13, "learning_rate": 2.5821540573853617e-05, "loss": 0.051, "step": 45635 }, { "epoch": 2.13, "learning_rate": 2.58210725040176e-05, "loss": 0.0836, "step": 45640 }, { "epoch": 2.13, "learning_rate": 2.582060443418158e-05, "loss": 0.1078, "step": 45645 }, { "epoch": 2.13, "learning_rate": 2.582013636434556e-05, "loss": 0.1175, "step": 45650 }, { "epoch": 2.13, "learning_rate": 2.5819668294509543e-05, "loss": 0.1433, "step": 45655 }, { "epoch": 2.13, "learning_rate": 2.5819200224673523e-05, "loss": 0.2411, "step": 45660 }, { "epoch": 2.13, "learning_rate": 2.5818732154837502e-05, "loss": 0.3004, "step": 45665 }, { "epoch": 2.13, "learning_rate": 2.5818264085001482e-05, "loss": 0.1811, "step": 45670 }, { "epoch": 2.13, "learning_rate": 2.5817796015165465e-05, "loss": 0.0373, "step": 45675 }, { "epoch": 2.13, "learning_rate": 2.5817327945329445e-05, "loss": 0.0428, "step": 45680 }, { "epoch": 2.13, "learning_rate": 2.5816859875493425e-05, "loss": 0.0723, "step": 45685 }, { "epoch": 2.13, "learning_rate": 2.5816391805657405e-05, "loss": 0.1206, "step": 45690 }, { "epoch": 2.13, "learning_rate": 2.5815923735821385e-05, "loss": 0.082, "step": 45695 }, { "epoch": 2.13, "learning_rate": 2.5815455665985364e-05, "loss": 0.0632, "step": 45700 }, { "epoch": 2.13, "learning_rate": 2.5814987596149344e-05, "loss": 0.2105, "step": 45705 }, { "epoch": 2.13, "learning_rate": 2.5814519526313327e-05, "loss": 0.1711, "step": 45710 }, { "epoch": 2.13, "learning_rate": 2.5814051456477307e-05, "loss": 0.2524, "step": 45715 }, { "epoch": 2.13, "learning_rate": 2.5813583386641287e-05, "loss": 0.0903, "step": 45720 }, { "epoch": 2.13, "learning_rate": 2.5813115316805267e-05, "loss": 0.0578, "step": 45725 }, { "epoch": 2.13, "learning_rate": 2.581264724696925e-05, "loss": 0.0888, "step": 45730 }, { "epoch": 2.13, "learning_rate": 2.581217917713323e-05, "loss": 0.0723, "step": 45735 }, { "epoch": 2.13, "learning_rate": 2.581171110729721e-05, "loss": 0.1233, "step": 45740 }, { "epoch": 2.13, "learning_rate": 2.581124303746119e-05, "loss": 0.0671, "step": 45745 }, { "epoch": 2.13, "learning_rate": 2.5810774967625173e-05, "loss": 0.1057, "step": 45750 }, { "epoch": 2.13, "learning_rate": 2.5810306897789153e-05, "loss": 0.2059, "step": 45755 }, { "epoch": 2.14, "learning_rate": 2.580983882795313e-05, "loss": 0.2111, "step": 45760 }, { "epoch": 2.14, "learning_rate": 2.580937075811711e-05, "loss": 0.3121, "step": 45765 }, { "epoch": 2.14, "learning_rate": 2.5808902688281092e-05, "loss": 0.097, "step": 45770 }, { "epoch": 2.14, "learning_rate": 2.5808434618445072e-05, "loss": 0.0592, "step": 45775 }, { "epoch": 2.14, "learning_rate": 2.580796654860905e-05, "loss": 0.0329, "step": 45780 }, { "epoch": 2.14, "learning_rate": 2.5807498478773035e-05, "loss": 0.0813, "step": 45785 }, { "epoch": 2.14, "learning_rate": 2.5807030408937015e-05, "loss": 0.1641, "step": 45790 }, { "epoch": 2.14, "learning_rate": 2.5806562339100995e-05, "loss": 0.0567, "step": 45795 }, { "epoch": 2.14, "learning_rate": 2.5806094269264974e-05, "loss": 0.0832, "step": 45800 }, { "epoch": 2.14, "learning_rate": 2.5805626199428958e-05, "loss": 0.1632, "step": 45805 }, { "epoch": 2.14, "learning_rate": 2.5805158129592937e-05, "loss": 0.2241, "step": 45810 }, { "epoch": 2.14, "learning_rate": 2.5804690059756917e-05, "loss": 0.3609, "step": 45815 }, { "epoch": 2.14, "learning_rate": 2.5804221989920897e-05, "loss": 0.1113, "step": 45820 }, { "epoch": 2.14, "learning_rate": 2.5803753920084877e-05, "loss": 0.054, "step": 45825 }, { "epoch": 2.14, "learning_rate": 2.5803285850248857e-05, "loss": 0.0465, "step": 45830 }, { "epoch": 2.14, "learning_rate": 2.5802817780412836e-05, "loss": 0.0529, "step": 45835 }, { "epoch": 2.14, "learning_rate": 2.580234971057682e-05, "loss": 0.1233, "step": 45840 }, { "epoch": 2.14, "learning_rate": 2.58018816407408e-05, "loss": 0.2243, "step": 45845 }, { "epoch": 2.14, "learning_rate": 2.580141357090478e-05, "loss": 0.2473, "step": 45850 }, { "epoch": 2.14, "learning_rate": 2.580094550106876e-05, "loss": 0.2115, "step": 45855 }, { "epoch": 2.14, "learning_rate": 2.5800477431232742e-05, "loss": 0.1574, "step": 45860 }, { "epoch": 2.14, "learning_rate": 2.5800009361396722e-05, "loss": 0.273, "step": 45865 }, { "epoch": 2.14, "learning_rate": 2.5799541291560702e-05, "loss": 0.0301, "step": 45870 }, { "epoch": 2.14, "learning_rate": 2.5799073221724682e-05, "loss": 0.0551, "step": 45875 }, { "epoch": 2.14, "learning_rate": 2.5798605151888665e-05, "loss": 0.0194, "step": 45880 }, { "epoch": 2.14, "learning_rate": 2.579813708205264e-05, "loss": 0.1256, "step": 45885 }, { "epoch": 2.14, "learning_rate": 2.579766901221662e-05, "loss": 0.1096, "step": 45890 }, { "epoch": 2.14, "learning_rate": 2.5797200942380604e-05, "loss": 0.1196, "step": 45895 }, { "epoch": 2.14, "learning_rate": 2.5796732872544584e-05, "loss": 0.1433, "step": 45900 }, { "epoch": 2.14, "learning_rate": 2.5796264802708564e-05, "loss": 0.1346, "step": 45905 }, { "epoch": 2.14, "learning_rate": 2.5795796732872544e-05, "loss": 0.2153, "step": 45910 }, { "epoch": 2.14, "learning_rate": 2.5795328663036527e-05, "loss": 0.2737, "step": 45915 }, { "epoch": 2.14, "learning_rate": 2.5794860593200507e-05, "loss": 0.0833, "step": 45920 }, { "epoch": 2.14, "learning_rate": 2.5794392523364487e-05, "loss": 0.0502, "step": 45925 }, { "epoch": 2.14, "learning_rate": 2.5793924453528467e-05, "loss": 0.0853, "step": 45930 }, { "epoch": 2.14, "learning_rate": 2.579345638369245e-05, "loss": 0.0362, "step": 45935 }, { "epoch": 2.14, "learning_rate": 2.579298831385643e-05, "loss": 0.0989, "step": 45940 }, { "epoch": 2.14, "learning_rate": 2.579252024402041e-05, "loss": 0.1137, "step": 45945 }, { "epoch": 2.14, "learning_rate": 2.5792052174184386e-05, "loss": 0.0982, "step": 45950 }, { "epoch": 2.14, "learning_rate": 2.579158410434837e-05, "loss": 0.2569, "step": 45955 }, { "epoch": 2.14, "learning_rate": 2.579111603451235e-05, "loss": 0.2473, "step": 45960 }, { "epoch": 2.14, "learning_rate": 2.579064796467633e-05, "loss": 0.2652, "step": 45965 }, { "epoch": 2.15, "learning_rate": 2.5790179894840312e-05, "loss": 0.0475, "step": 45970 }, { "epoch": 2.15, "learning_rate": 2.578971182500429e-05, "loss": 0.0636, "step": 45975 }, { "epoch": 2.15, "learning_rate": 2.578924375516827e-05, "loss": 0.085, "step": 45980 }, { "epoch": 2.15, "learning_rate": 2.578877568533225e-05, "loss": 0.0648, "step": 45985 }, { "epoch": 2.15, "learning_rate": 2.5788307615496235e-05, "loss": 0.1436, "step": 45990 }, { "epoch": 2.15, "learning_rate": 2.5787839545660214e-05, "loss": 0.1249, "step": 45995 }, { "epoch": 2.15, "learning_rate": 2.5787371475824194e-05, "loss": 0.2969, "step": 46000 }, { "epoch": 2.15, "learning_rate": 2.5786903405988174e-05, "loss": 0.1492, "step": 46005 }, { "epoch": 2.15, "learning_rate": 2.5786435336152154e-05, "loss": 0.3294, "step": 46010 }, { "epoch": 2.15, "learning_rate": 2.5785967266316134e-05, "loss": 0.3349, "step": 46015 }, { "epoch": 2.15, "learning_rate": 2.5785499196480113e-05, "loss": 0.0752, "step": 46020 }, { "epoch": 2.15, "learning_rate": 2.5785031126644097e-05, "loss": 0.023, "step": 46025 }, { "epoch": 2.15, "learning_rate": 2.5784563056808076e-05, "loss": 0.0877, "step": 46030 }, { "epoch": 2.15, "learning_rate": 2.5784094986972056e-05, "loss": 0.0426, "step": 46035 }, { "epoch": 2.15, "learning_rate": 2.5783626917136036e-05, "loss": 0.0787, "step": 46040 }, { "epoch": 2.15, "learning_rate": 2.578315884730002e-05, "loss": 0.1063, "step": 46045 }, { "epoch": 2.15, "learning_rate": 2.5782690777464e-05, "loss": 0.1187, "step": 46050 }, { "epoch": 2.15, "learning_rate": 2.578222270762798e-05, "loss": 0.2699, "step": 46055 }, { "epoch": 2.15, "learning_rate": 2.578175463779196e-05, "loss": 0.3069, "step": 46060 }, { "epoch": 2.15, "learning_rate": 2.5781286567955942e-05, "loss": 0.2714, "step": 46065 }, { "epoch": 2.15, "learning_rate": 2.5780818498119922e-05, "loss": 0.0248, "step": 46070 }, { "epoch": 2.15, "learning_rate": 2.5780350428283898e-05, "loss": 0.0766, "step": 46075 }, { "epoch": 2.15, "learning_rate": 2.577988235844788e-05, "loss": 0.046, "step": 46080 }, { "epoch": 2.15, "learning_rate": 2.577941428861186e-05, "loss": 0.0724, "step": 46085 }, { "epoch": 2.15, "learning_rate": 2.577894621877584e-05, "loss": 0.0817, "step": 46090 }, { "epoch": 2.15, "learning_rate": 2.577847814893982e-05, "loss": 0.0572, "step": 46095 }, { "epoch": 2.15, "learning_rate": 2.5778010079103804e-05, "loss": 0.1032, "step": 46100 }, { "epoch": 2.15, "learning_rate": 2.5777542009267784e-05, "loss": 0.2213, "step": 46105 }, { "epoch": 2.15, "learning_rate": 2.5777073939431764e-05, "loss": 0.2276, "step": 46110 }, { "epoch": 2.15, "learning_rate": 2.5776605869595744e-05, "loss": 0.2599, "step": 46115 }, { "epoch": 2.15, "learning_rate": 2.5776137799759727e-05, "loss": 0.056, "step": 46120 }, { "epoch": 2.15, "learning_rate": 2.5775669729923707e-05, "loss": 0.0643, "step": 46125 }, { "epoch": 2.15, "learning_rate": 2.5775201660087686e-05, "loss": 0.1097, "step": 46130 }, { "epoch": 2.15, "learning_rate": 2.5774733590251666e-05, "loss": 0.0898, "step": 46135 }, { "epoch": 2.15, "learning_rate": 2.5774265520415646e-05, "loss": 0.0432, "step": 46140 }, { "epoch": 2.15, "learning_rate": 2.5773797450579626e-05, "loss": 0.1249, "step": 46145 }, { "epoch": 2.15, "learning_rate": 2.5773329380743606e-05, "loss": 0.1728, "step": 46150 }, { "epoch": 2.15, "learning_rate": 2.577286131090759e-05, "loss": 0.1268, "step": 46155 }, { "epoch": 2.15, "learning_rate": 2.577239324107157e-05, "loss": 0.1939, "step": 46160 }, { "epoch": 2.15, "learning_rate": 2.577192517123555e-05, "loss": 0.2519, "step": 46165 }, { "epoch": 2.15, "learning_rate": 2.5771457101399528e-05, "loss": 0.4268, "step": 46170 }, { "epoch": 2.15, "learning_rate": 2.577098903156351e-05, "loss": 0.0604, "step": 46175 }, { "epoch": 2.15, "learning_rate": 2.577052096172749e-05, "loss": 0.0232, "step": 46180 }, { "epoch": 2.16, "learning_rate": 2.577005289189147e-05, "loss": 0.0735, "step": 46185 }, { "epoch": 2.16, "learning_rate": 2.576958482205545e-05, "loss": 0.0903, "step": 46190 }, { "epoch": 2.16, "learning_rate": 2.5769116752219434e-05, "loss": 0.0929, "step": 46195 }, { "epoch": 2.16, "learning_rate": 2.576864868238341e-05, "loss": 0.1613, "step": 46200 }, { "epoch": 2.16, "learning_rate": 2.576818061254739e-05, "loss": 0.145, "step": 46205 }, { "epoch": 2.16, "learning_rate": 2.5767712542711374e-05, "loss": 0.3726, "step": 46210 }, { "epoch": 2.16, "learning_rate": 2.5767244472875353e-05, "loss": 0.3217, "step": 46215 }, { "epoch": 2.16, "learning_rate": 2.5766776403039333e-05, "loss": 0.0123, "step": 46220 }, { "epoch": 2.16, "learning_rate": 2.5766308333203313e-05, "loss": 0.0311, "step": 46225 }, { "epoch": 2.16, "learning_rate": 2.5765840263367296e-05, "loss": 0.0328, "step": 46230 }, { "epoch": 2.16, "learning_rate": 2.5765372193531276e-05, "loss": 0.0435, "step": 46235 }, { "epoch": 2.16, "learning_rate": 2.5764904123695256e-05, "loss": 0.1408, "step": 46240 }, { "epoch": 2.16, "learning_rate": 2.5764436053859236e-05, "loss": 0.1077, "step": 46245 }, { "epoch": 2.16, "learning_rate": 2.576396798402322e-05, "loss": 0.0843, "step": 46250 }, { "epoch": 2.16, "learning_rate": 2.57634999141872e-05, "loss": 0.1736, "step": 46255 }, { "epoch": 2.16, "learning_rate": 2.576303184435118e-05, "loss": 0.2352, "step": 46260 }, { "epoch": 2.16, "learning_rate": 2.576256377451516e-05, "loss": 0.246, "step": 46265 }, { "epoch": 2.16, "learning_rate": 2.5762095704679138e-05, "loss": 0.0334, "step": 46270 }, { "epoch": 2.16, "learning_rate": 2.5761627634843118e-05, "loss": 0.0407, "step": 46275 }, { "epoch": 2.16, "learning_rate": 2.5761159565007098e-05, "loss": 0.0385, "step": 46280 }, { "epoch": 2.16, "learning_rate": 2.576069149517108e-05, "loss": 0.0832, "step": 46285 }, { "epoch": 2.16, "learning_rate": 2.576022342533506e-05, "loss": 0.0967, "step": 46290 }, { "epoch": 2.16, "learning_rate": 2.575975535549904e-05, "loss": 0.1262, "step": 46295 }, { "epoch": 2.16, "learning_rate": 2.575928728566302e-05, "loss": 0.0887, "step": 46300 }, { "epoch": 2.16, "learning_rate": 2.5758819215827004e-05, "loss": 0.1202, "step": 46305 }, { "epoch": 2.16, "learning_rate": 2.5758351145990984e-05, "loss": 0.3248, "step": 46310 }, { "epoch": 2.16, "learning_rate": 2.5757883076154963e-05, "loss": 0.1538, "step": 46315 }, { "epoch": 2.16, "learning_rate": 2.5757415006318947e-05, "loss": 0.0732, "step": 46320 }, { "epoch": 2.16, "learning_rate": 2.5756946936482923e-05, "loss": 0.0257, "step": 46325 }, { "epoch": 2.16, "learning_rate": 2.5756478866646903e-05, "loss": 0.0388, "step": 46330 }, { "epoch": 2.16, "learning_rate": 2.5756010796810883e-05, "loss": 0.0968, "step": 46335 }, { "epoch": 2.16, "learning_rate": 2.5755542726974866e-05, "loss": 0.0325, "step": 46340 }, { "epoch": 2.16, "learning_rate": 2.5755074657138846e-05, "loss": 0.0649, "step": 46345 }, { "epoch": 2.16, "learning_rate": 2.5754606587302825e-05, "loss": 0.1796, "step": 46350 }, { "epoch": 2.16, "learning_rate": 2.5754138517466805e-05, "loss": 0.1551, "step": 46355 }, { "epoch": 2.16, "learning_rate": 2.575367044763079e-05, "loss": 0.3733, "step": 46360 }, { "epoch": 2.16, "learning_rate": 2.5753202377794768e-05, "loss": 0.3383, "step": 46365 }, { "epoch": 2.16, "learning_rate": 2.5752734307958748e-05, "loss": 0.045, "step": 46370 }, { "epoch": 2.16, "learning_rate": 2.5752266238122728e-05, "loss": 0.0682, "step": 46375 }, { "epoch": 2.16, "learning_rate": 2.575179816828671e-05, "loss": 0.0732, "step": 46380 }, { "epoch": 2.16, "learning_rate": 2.575133009845069e-05, "loss": 0.0674, "step": 46385 }, { "epoch": 2.16, "learning_rate": 2.5750862028614667e-05, "loss": 0.0772, "step": 46390 }, { "epoch": 2.16, "learning_rate": 2.575039395877865e-05, "loss": 0.0877, "step": 46395 }, { "epoch": 2.17, "learning_rate": 2.574992588894263e-05, "loss": 0.1246, "step": 46400 }, { "epoch": 2.17, "learning_rate": 2.574945781910661e-05, "loss": 0.1309, "step": 46405 }, { "epoch": 2.17, "learning_rate": 2.574898974927059e-05, "loss": 0.1698, "step": 46410 }, { "epoch": 2.17, "learning_rate": 2.5748521679434573e-05, "loss": 0.2194, "step": 46415 }, { "epoch": 2.17, "learning_rate": 2.5748053609598553e-05, "loss": 0.038, "step": 46420 }, { "epoch": 2.17, "learning_rate": 2.5747585539762533e-05, "loss": 0.0315, "step": 46425 }, { "epoch": 2.17, "learning_rate": 2.5747117469926513e-05, "loss": 0.0625, "step": 46430 }, { "epoch": 2.17, "learning_rate": 2.5746649400090496e-05, "loss": 0.0699, "step": 46435 }, { "epoch": 2.17, "learning_rate": 2.5746181330254476e-05, "loss": 0.1504, "step": 46440 }, { "epoch": 2.17, "learning_rate": 2.5745713260418456e-05, "loss": 0.0954, "step": 46445 }, { "epoch": 2.17, "learning_rate": 2.574524519058244e-05, "loss": 0.1145, "step": 46450 }, { "epoch": 2.17, "learning_rate": 2.5744777120746415e-05, "loss": 0.1848, "step": 46455 }, { "epoch": 2.17, "learning_rate": 2.5744309050910395e-05, "loss": 0.169, "step": 46460 }, { "epoch": 2.17, "learning_rate": 2.5743840981074375e-05, "loss": 0.2501, "step": 46465 }, { "epoch": 2.17, "learning_rate": 2.5743372911238358e-05, "loss": 0.0285, "step": 46470 }, { "epoch": 2.17, "learning_rate": 2.5742904841402338e-05, "loss": 0.0967, "step": 46475 }, { "epoch": 2.17, "learning_rate": 2.5742436771566318e-05, "loss": 0.0184, "step": 46480 }, { "epoch": 2.17, "learning_rate": 2.5741968701730297e-05, "loss": 0.0831, "step": 46485 }, { "epoch": 2.17, "learning_rate": 2.574150063189428e-05, "loss": 0.0888, "step": 46490 }, { "epoch": 2.17, "learning_rate": 2.574103256205826e-05, "loss": 0.072, "step": 46495 }, { "epoch": 2.17, "learning_rate": 2.574056449222224e-05, "loss": 0.2972, "step": 46500 }, { "epoch": 2.17, "learning_rate": 2.5740096422386224e-05, "loss": 0.136, "step": 46505 }, { "epoch": 2.17, "learning_rate": 2.5739628352550203e-05, "loss": 0.2372, "step": 46510 }, { "epoch": 2.17, "learning_rate": 2.573916028271418e-05, "loss": 0.2938, "step": 46515 }, { "epoch": 2.17, "learning_rate": 2.573869221287816e-05, "loss": 0.1196, "step": 46520 }, { "epoch": 2.17, "learning_rate": 2.5738224143042143e-05, "loss": 0.0345, "step": 46525 }, { "epoch": 2.17, "learning_rate": 2.5737756073206123e-05, "loss": 0.0601, "step": 46530 }, { "epoch": 2.17, "learning_rate": 2.5737288003370102e-05, "loss": 0.0258, "step": 46535 }, { "epoch": 2.17, "learning_rate": 2.5736819933534082e-05, "loss": 0.1341, "step": 46540 }, { "epoch": 2.17, "learning_rate": 2.5736351863698065e-05, "loss": 0.1137, "step": 46545 }, { "epoch": 2.17, "learning_rate": 2.5735883793862045e-05, "loss": 0.1336, "step": 46550 }, { "epoch": 2.17, "learning_rate": 2.5735415724026025e-05, "loss": 0.2547, "step": 46555 }, { "epoch": 2.17, "learning_rate": 2.5734947654190005e-05, "loss": 0.3075, "step": 46560 }, { "epoch": 2.17, "learning_rate": 2.5734479584353988e-05, "loss": 0.3183, "step": 46565 }, { "epoch": 2.17, "learning_rate": 2.5734011514517968e-05, "loss": 0.0344, "step": 46570 }, { "epoch": 2.17, "learning_rate": 2.5733543444681948e-05, "loss": 0.0464, "step": 46575 }, { "epoch": 2.17, "learning_rate": 2.5733075374845928e-05, "loss": 0.0964, "step": 46580 }, { "epoch": 2.17, "learning_rate": 2.5732607305009907e-05, "loss": 0.1135, "step": 46585 }, { "epoch": 2.17, "learning_rate": 2.5732139235173887e-05, "loss": 0.0418, "step": 46590 }, { "epoch": 2.17, "learning_rate": 2.5731671165337867e-05, "loss": 0.1447, "step": 46595 }, { "epoch": 2.17, "learning_rate": 2.573120309550185e-05, "loss": 0.1324, "step": 46600 }, { "epoch": 2.17, "learning_rate": 2.573073502566583e-05, "loss": 0.1428, "step": 46605 }, { "epoch": 2.17, "learning_rate": 2.573026695582981e-05, "loss": 0.1874, "step": 46610 }, { "epoch": 2.18, "learning_rate": 2.572979888599379e-05, "loss": 0.3622, "step": 46615 }, { "epoch": 2.18, "learning_rate": 2.5729330816157773e-05, "loss": 0.0439, "step": 46620 }, { "epoch": 2.18, "learning_rate": 2.5728862746321753e-05, "loss": 0.0388, "step": 46625 }, { "epoch": 2.18, "learning_rate": 2.5728394676485733e-05, "loss": 0.0778, "step": 46630 }, { "epoch": 2.18, "learning_rate": 2.5727926606649716e-05, "loss": 0.0736, "step": 46635 }, { "epoch": 2.18, "learning_rate": 2.5727458536813696e-05, "loss": 0.0724, "step": 46640 }, { "epoch": 2.18, "learning_rate": 2.5726990466977672e-05, "loss": 0.0939, "step": 46645 }, { "epoch": 2.18, "learning_rate": 2.5726522397141652e-05, "loss": 0.2064, "step": 46650 }, { "epoch": 2.18, "learning_rate": 2.5726054327305635e-05, "loss": 0.2022, "step": 46655 }, { "epoch": 2.18, "learning_rate": 2.5725586257469615e-05, "loss": 0.201, "step": 46660 }, { "epoch": 2.18, "learning_rate": 2.5725118187633595e-05, "loss": 0.3758, "step": 46665 }, { "epoch": 2.18, "learning_rate": 2.5724650117797574e-05, "loss": 0.0136, "step": 46670 }, { "epoch": 2.18, "learning_rate": 2.5724182047961558e-05, "loss": 0.0379, "step": 46675 }, { "epoch": 2.18, "learning_rate": 2.5723713978125537e-05, "loss": 0.0547, "step": 46680 }, { "epoch": 2.18, "learning_rate": 2.5723245908289517e-05, "loss": 0.0536, "step": 46685 }, { "epoch": 2.18, "learning_rate": 2.57227778384535e-05, "loss": 0.119, "step": 46690 }, { "epoch": 2.18, "learning_rate": 2.572230976861748e-05, "loss": 0.1019, "step": 46695 }, { "epoch": 2.18, "learning_rate": 2.572184169878146e-05, "loss": 0.23, "step": 46700 }, { "epoch": 2.18, "learning_rate": 2.5721373628945437e-05, "loss": 0.2845, "step": 46705 }, { "epoch": 2.18, "learning_rate": 2.572090555910942e-05, "loss": 0.3395, "step": 46710 }, { "epoch": 2.18, "learning_rate": 2.57204374892734e-05, "loss": 0.462, "step": 46715 }, { "epoch": 2.18, "learning_rate": 2.571996941943738e-05, "loss": 0.0359, "step": 46720 }, { "epoch": 2.18, "learning_rate": 2.571950134960136e-05, "loss": 0.0462, "step": 46725 }, { "epoch": 2.18, "learning_rate": 2.5719033279765342e-05, "loss": 0.0381, "step": 46730 }, { "epoch": 2.18, "learning_rate": 2.5718565209929322e-05, "loss": 0.0372, "step": 46735 }, { "epoch": 2.18, "learning_rate": 2.5718097140093302e-05, "loss": 0.0884, "step": 46740 }, { "epoch": 2.18, "learning_rate": 2.5717629070257282e-05, "loss": 0.1161, "step": 46745 }, { "epoch": 2.18, "learning_rate": 2.5717161000421265e-05, "loss": 0.1198, "step": 46750 }, { "epoch": 2.18, "learning_rate": 2.5716692930585245e-05, "loss": 0.1816, "step": 46755 }, { "epoch": 2.18, "learning_rate": 2.5716224860749225e-05, "loss": 0.2607, "step": 46760 }, { "epoch": 2.18, "learning_rate": 2.5715756790913208e-05, "loss": 0.2049, "step": 46765 }, { "epoch": 2.18, "learning_rate": 2.5715288721077184e-05, "loss": 0.0805, "step": 46770 }, { "epoch": 2.18, "learning_rate": 2.5714820651241164e-05, "loss": 0.0338, "step": 46775 }, { "epoch": 2.18, "learning_rate": 2.5714352581405144e-05, "loss": 0.0759, "step": 46780 }, { "epoch": 2.18, "learning_rate": 2.5713884511569127e-05, "loss": 0.0604, "step": 46785 }, { "epoch": 2.18, "learning_rate": 2.5713416441733107e-05, "loss": 0.1115, "step": 46790 }, { "epoch": 2.18, "learning_rate": 2.5712948371897087e-05, "loss": 0.129, "step": 46795 }, { "epoch": 2.18, "learning_rate": 2.5712480302061067e-05, "loss": 0.1771, "step": 46800 }, { "epoch": 2.18, "learning_rate": 2.571201223222505e-05, "loss": 0.2118, "step": 46805 }, { "epoch": 2.18, "learning_rate": 2.571154416238903e-05, "loss": 0.3582, "step": 46810 }, { "epoch": 2.18, "learning_rate": 2.571107609255301e-05, "loss": 0.1941, "step": 46815 }, { "epoch": 2.18, "learning_rate": 2.5710608022716993e-05, "loss": 0.0445, "step": 46820 }, { "epoch": 2.18, "learning_rate": 2.5710139952880972e-05, "loss": 0.0163, "step": 46825 }, { "epoch": 2.19, "learning_rate": 2.5709671883044952e-05, "loss": 0.082, "step": 46830 }, { "epoch": 2.19, "learning_rate": 2.570920381320893e-05, "loss": 0.0987, "step": 46835 }, { "epoch": 2.19, "learning_rate": 2.5708735743372912e-05, "loss": 0.0951, "step": 46840 }, { "epoch": 2.19, "learning_rate": 2.5708267673536892e-05, "loss": 0.1261, "step": 46845 }, { "epoch": 2.19, "learning_rate": 2.570779960370087e-05, "loss": 0.081, "step": 46850 }, { "epoch": 2.19, "learning_rate": 2.570733153386485e-05, "loss": 0.1243, "step": 46855 }, { "epoch": 2.19, "learning_rate": 2.5706863464028835e-05, "loss": 0.2111, "step": 46860 }, { "epoch": 2.19, "learning_rate": 2.5706395394192814e-05, "loss": 0.3578, "step": 46865 }, { "epoch": 2.19, "learning_rate": 2.5705927324356794e-05, "loss": 0.0699, "step": 46870 }, { "epoch": 2.19, "learning_rate": 2.5705459254520777e-05, "loss": 0.0366, "step": 46875 }, { "epoch": 2.19, "learning_rate": 2.5704991184684757e-05, "loss": 0.0836, "step": 46880 }, { "epoch": 2.19, "learning_rate": 2.5704523114848737e-05, "loss": 0.1054, "step": 46885 }, { "epoch": 2.19, "learning_rate": 2.5704055045012717e-05, "loss": 0.1063, "step": 46890 }, { "epoch": 2.19, "learning_rate": 2.5703586975176697e-05, "loss": 0.1112, "step": 46895 }, { "epoch": 2.19, "learning_rate": 2.5703118905340677e-05, "loss": 0.1248, "step": 46900 }, { "epoch": 2.19, "learning_rate": 2.5702650835504656e-05, "loss": 0.1293, "step": 46905 }, { "epoch": 2.19, "learning_rate": 2.5702182765668636e-05, "loss": 0.2318, "step": 46910 }, { "epoch": 2.19, "learning_rate": 2.570171469583262e-05, "loss": 0.3429, "step": 46915 }, { "epoch": 2.19, "learning_rate": 2.57012466259966e-05, "loss": 0.0605, "step": 46920 }, { "epoch": 2.19, "learning_rate": 2.570077855616058e-05, "loss": 0.0198, "step": 46925 }, { "epoch": 2.19, "learning_rate": 2.5700310486324562e-05, "loss": 0.0185, "step": 46930 }, { "epoch": 2.19, "learning_rate": 2.5699842416488542e-05, "loss": 0.0859, "step": 46935 }, { "epoch": 2.19, "learning_rate": 2.5699374346652522e-05, "loss": 0.0596, "step": 46940 }, { "epoch": 2.19, "learning_rate": 2.56989062768165e-05, "loss": 0.1072, "step": 46945 }, { "epoch": 2.19, "learning_rate": 2.5698438206980485e-05, "loss": 0.2226, "step": 46950 }, { "epoch": 2.19, "learning_rate": 2.5697970137144465e-05, "loss": 0.0692, "step": 46955 }, { "epoch": 2.19, "learning_rate": 2.569750206730844e-05, "loss": 0.2887, "step": 46960 }, { "epoch": 2.19, "learning_rate": 2.569703399747242e-05, "loss": 0.2476, "step": 46965 }, { "epoch": 2.19, "learning_rate": 2.5696565927636404e-05, "loss": 0.0231, "step": 46970 }, { "epoch": 2.19, "learning_rate": 2.5696097857800384e-05, "loss": 0.0317, "step": 46975 }, { "epoch": 2.19, "learning_rate": 2.5695629787964364e-05, "loss": 0.1597, "step": 46980 }, { "epoch": 2.19, "learning_rate": 2.5695161718128344e-05, "loss": 0.0719, "step": 46985 }, { "epoch": 2.19, "learning_rate": 2.5694693648292327e-05, "loss": 0.1164, "step": 46990 }, { "epoch": 2.19, "learning_rate": 2.5694225578456307e-05, "loss": 0.0615, "step": 46995 }, { "epoch": 2.19, "learning_rate": 2.5693757508620286e-05, "loss": 0.1833, "step": 47000 }, { "epoch": 2.19, "learning_rate": 2.569328943878427e-05, "loss": 0.2171, "step": 47005 }, { "epoch": 2.19, "learning_rate": 2.569282136894825e-05, "loss": 0.1349, "step": 47010 }, { "epoch": 2.19, "learning_rate": 2.569235329911223e-05, "loss": 0.3012, "step": 47015 }, { "epoch": 2.19, "learning_rate": 2.569188522927621e-05, "loss": 0.0959, "step": 47020 }, { "epoch": 2.19, "learning_rate": 2.569141715944019e-05, "loss": 0.0605, "step": 47025 }, { "epoch": 2.19, "learning_rate": 2.569094908960417e-05, "loss": 0.0373, "step": 47030 }, { "epoch": 2.19, "learning_rate": 2.569048101976815e-05, "loss": 0.048, "step": 47035 }, { "epoch": 2.19, "learning_rate": 2.569001294993213e-05, "loss": 0.1739, "step": 47040 }, { "epoch": 2.2, "learning_rate": 2.568954488009611e-05, "loss": 0.0755, "step": 47045 }, { "epoch": 2.2, "learning_rate": 2.568907681026009e-05, "loss": 0.1085, "step": 47050 }, { "epoch": 2.2, "learning_rate": 2.568860874042407e-05, "loss": 0.0949, "step": 47055 }, { "epoch": 2.2, "learning_rate": 2.5688140670588054e-05, "loss": 0.2284, "step": 47060 }, { "epoch": 2.2, "learning_rate": 2.5687672600752034e-05, "loss": 0.2739, "step": 47065 }, { "epoch": 2.2, "learning_rate": 2.5687204530916014e-05, "loss": 0.0467, "step": 47070 }, { "epoch": 2.2, "learning_rate": 2.5686736461079994e-05, "loss": 0.0681, "step": 47075 }, { "epoch": 2.2, "learning_rate": 2.5686268391243977e-05, "loss": 0.0443, "step": 47080 }, { "epoch": 2.2, "learning_rate": 2.5685800321407954e-05, "loss": 0.0386, "step": 47085 }, { "epoch": 2.2, "learning_rate": 2.5685332251571933e-05, "loss": 0.0573, "step": 47090 }, { "epoch": 2.2, "learning_rate": 2.5684864181735913e-05, "loss": 0.0854, "step": 47095 }, { "epoch": 2.2, "learning_rate": 2.5684396111899896e-05, "loss": 0.1618, "step": 47100 }, { "epoch": 2.2, "learning_rate": 2.5683928042063876e-05, "loss": 0.2264, "step": 47105 }, { "epoch": 2.2, "learning_rate": 2.5683459972227856e-05, "loss": 0.2645, "step": 47110 }, { "epoch": 2.2, "learning_rate": 2.568299190239184e-05, "loss": 0.2872, "step": 47115 }, { "epoch": 2.2, "learning_rate": 2.568252383255582e-05, "loss": 0.0214, "step": 47120 }, { "epoch": 2.2, "learning_rate": 2.56820557627198e-05, "loss": 0.0342, "step": 47125 }, { "epoch": 2.2, "learning_rate": 2.568158769288378e-05, "loss": 0.1035, "step": 47130 }, { "epoch": 2.2, "learning_rate": 2.5681119623047762e-05, "loss": 0.0843, "step": 47135 }, { "epoch": 2.2, "learning_rate": 2.568065155321174e-05, "loss": 0.1107, "step": 47140 }, { "epoch": 2.2, "learning_rate": 2.568018348337572e-05, "loss": 0.1401, "step": 47145 }, { "epoch": 2.2, "learning_rate": 2.5679715413539698e-05, "loss": 0.1624, "step": 47150 }, { "epoch": 2.2, "learning_rate": 2.567924734370368e-05, "loss": 0.1962, "step": 47155 }, { "epoch": 2.2, "learning_rate": 2.567877927386766e-05, "loss": 0.3318, "step": 47160 }, { "epoch": 2.2, "learning_rate": 2.567831120403164e-05, "loss": 0.1882, "step": 47165 }, { "epoch": 2.2, "learning_rate": 2.567784313419562e-05, "loss": 0.1179, "step": 47170 }, { "epoch": 2.2, "learning_rate": 2.5677375064359604e-05, "loss": 0.0591, "step": 47175 }, { "epoch": 2.2, "learning_rate": 2.5676906994523584e-05, "loss": 0.0799, "step": 47180 }, { "epoch": 2.2, "learning_rate": 2.5676438924687563e-05, "loss": 0.1135, "step": 47185 }, { "epoch": 2.2, "learning_rate": 2.5675970854851547e-05, "loss": 0.0873, "step": 47190 }, { "epoch": 2.2, "learning_rate": 2.5675502785015526e-05, "loss": 0.0543, "step": 47195 }, { "epoch": 2.2, "learning_rate": 2.5675034715179506e-05, "loss": 0.1805, "step": 47200 }, { "epoch": 2.2, "learning_rate": 2.5674566645343486e-05, "loss": 0.0892, "step": 47205 }, { "epoch": 2.2, "learning_rate": 2.5674098575507466e-05, "loss": 0.2918, "step": 47210 }, { "epoch": 2.2, "learning_rate": 2.5673630505671446e-05, "loss": 0.221, "step": 47215 }, { "epoch": 2.2, "learning_rate": 2.5673162435835426e-05, "loss": 0.1082, "step": 47220 }, { "epoch": 2.2, "learning_rate": 2.5672694365999405e-05, "loss": 0.0489, "step": 47225 }, { "epoch": 2.2, "learning_rate": 2.567222629616339e-05, "loss": 0.0402, "step": 47230 }, { "epoch": 2.2, "learning_rate": 2.567175822632737e-05, "loss": 0.048, "step": 47235 }, { "epoch": 2.2, "learning_rate": 2.5671290156491348e-05, "loss": 0.0809, "step": 47240 }, { "epoch": 2.2, "learning_rate": 2.567082208665533e-05, "loss": 0.1109, "step": 47245 }, { "epoch": 2.2, "learning_rate": 2.567035401681931e-05, "loss": 0.0536, "step": 47250 }, { "epoch": 2.2, "learning_rate": 2.566988594698329e-05, "loss": 0.1372, "step": 47255 }, { "epoch": 2.21, "learning_rate": 2.566941787714727e-05, "loss": 0.2469, "step": 47260 }, { "epoch": 2.21, "learning_rate": 2.5668949807311254e-05, "loss": 0.2443, "step": 47265 }, { "epoch": 2.21, "learning_rate": 2.5668481737475234e-05, "loss": 0.045, "step": 47270 }, { "epoch": 2.21, "learning_rate": 2.566801366763921e-05, "loss": 0.0592, "step": 47275 }, { "epoch": 2.21, "learning_rate": 2.566754559780319e-05, "loss": 0.0674, "step": 47280 }, { "epoch": 2.21, "learning_rate": 2.5667077527967173e-05, "loss": 0.0417, "step": 47285 }, { "epoch": 2.21, "learning_rate": 2.5666609458131153e-05, "loss": 0.1509, "step": 47290 }, { "epoch": 2.21, "learning_rate": 2.5666141388295133e-05, "loss": 0.1222, "step": 47295 }, { "epoch": 2.21, "learning_rate": 2.5665673318459116e-05, "loss": 0.0784, "step": 47300 }, { "epoch": 2.21, "learning_rate": 2.5665205248623096e-05, "loss": 0.0995, "step": 47305 }, { "epoch": 2.21, "learning_rate": 2.5664737178787076e-05, "loss": 0.2392, "step": 47310 }, { "epoch": 2.21, "learning_rate": 2.5664269108951056e-05, "loss": 0.2065, "step": 47315 }, { "epoch": 2.21, "learning_rate": 2.566380103911504e-05, "loss": 0.0554, "step": 47320 }, { "epoch": 2.21, "learning_rate": 2.566333296927902e-05, "loss": 0.1014, "step": 47325 }, { "epoch": 2.21, "learning_rate": 2.5662864899443e-05, "loss": 0.0283, "step": 47330 }, { "epoch": 2.21, "learning_rate": 2.5662396829606978e-05, "loss": 0.0521, "step": 47335 }, { "epoch": 2.21, "learning_rate": 2.5661928759770958e-05, "loss": 0.0955, "step": 47340 }, { "epoch": 2.21, "learning_rate": 2.5661460689934938e-05, "loss": 0.074, "step": 47345 }, { "epoch": 2.21, "learning_rate": 2.5660992620098918e-05, "loss": 0.1441, "step": 47350 }, { "epoch": 2.21, "learning_rate": 2.5660524550262898e-05, "loss": 0.1411, "step": 47355 }, { "epoch": 2.21, "learning_rate": 2.566005648042688e-05, "loss": 0.2362, "step": 47360 }, { "epoch": 2.21, "learning_rate": 2.565958841059086e-05, "loss": 0.2412, "step": 47365 }, { "epoch": 2.21, "learning_rate": 2.565912034075484e-05, "loss": 0.0503, "step": 47370 }, { "epoch": 2.21, "learning_rate": 2.5658652270918824e-05, "loss": 0.0417, "step": 47375 }, { "epoch": 2.21, "learning_rate": 2.5658184201082803e-05, "loss": 0.0409, "step": 47380 }, { "epoch": 2.21, "learning_rate": 2.5657716131246783e-05, "loss": 0.0817, "step": 47385 }, { "epoch": 2.21, "learning_rate": 2.5657248061410763e-05, "loss": 0.0825, "step": 47390 }, { "epoch": 2.21, "learning_rate": 2.5656779991574746e-05, "loss": 0.1566, "step": 47395 }, { "epoch": 2.21, "learning_rate": 2.5656311921738723e-05, "loss": 0.0994, "step": 47400 }, { "epoch": 2.21, "learning_rate": 2.5655843851902702e-05, "loss": 0.1846, "step": 47405 }, { "epoch": 2.21, "learning_rate": 2.5655375782066682e-05, "loss": 0.2791, "step": 47410 }, { "epoch": 2.21, "learning_rate": 2.5654907712230666e-05, "loss": 0.3791, "step": 47415 }, { "epoch": 2.21, "learning_rate": 2.5654439642394645e-05, "loss": 0.0784, "step": 47420 }, { "epoch": 2.21, "learning_rate": 2.5653971572558625e-05, "loss": 0.0635, "step": 47425 }, { "epoch": 2.21, "learning_rate": 2.565350350272261e-05, "loss": 0.1326, "step": 47430 }, { "epoch": 2.21, "learning_rate": 2.5653035432886588e-05, "loss": 0.1016, "step": 47435 }, { "epoch": 2.21, "learning_rate": 2.5652567363050568e-05, "loss": 0.0786, "step": 47440 }, { "epoch": 2.21, "learning_rate": 2.5652099293214548e-05, "loss": 0.1149, "step": 47445 }, { "epoch": 2.21, "learning_rate": 2.565163122337853e-05, "loss": 0.1575, "step": 47450 }, { "epoch": 2.21, "learning_rate": 2.565116315354251e-05, "loss": 0.2128, "step": 47455 }, { "epoch": 2.21, "learning_rate": 2.565069508370649e-05, "loss": 0.2799, "step": 47460 }, { "epoch": 2.21, "learning_rate": 2.5650227013870467e-05, "loss": 0.3339, "step": 47465 }, { "epoch": 2.22, "learning_rate": 2.564975894403445e-05, "loss": 0.0249, "step": 47470 }, { "epoch": 2.22, "learning_rate": 2.564929087419843e-05, "loss": 0.0835, "step": 47475 }, { "epoch": 2.22, "learning_rate": 2.564882280436241e-05, "loss": 0.032, "step": 47480 }, { "epoch": 2.22, "learning_rate": 2.5648354734526393e-05, "loss": 0.0497, "step": 47485 }, { "epoch": 2.22, "learning_rate": 2.5647886664690373e-05, "loss": 0.0919, "step": 47490 }, { "epoch": 2.22, "learning_rate": 2.5647418594854353e-05, "loss": 0.061, "step": 47495 }, { "epoch": 2.22, "learning_rate": 2.5646950525018333e-05, "loss": 0.1598, "step": 47500 }, { "epoch": 2.22, "learning_rate": 2.5646482455182316e-05, "loss": 0.1173, "step": 47505 }, { "epoch": 2.22, "learning_rate": 2.5646014385346296e-05, "loss": 0.3099, "step": 47510 }, { "epoch": 2.22, "learning_rate": 2.5645546315510275e-05, "loss": 0.3127, "step": 47515 }, { "epoch": 2.22, "learning_rate": 2.5645078245674255e-05, "loss": 0.0806, "step": 47520 }, { "epoch": 2.22, "learning_rate": 2.5644610175838235e-05, "loss": 0.0177, "step": 47525 }, { "epoch": 2.22, "learning_rate": 2.5644142106002215e-05, "loss": 0.0359, "step": 47530 }, { "epoch": 2.22, "learning_rate": 2.5643674036166195e-05, "loss": 0.0904, "step": 47535 }, { "epoch": 2.22, "learning_rate": 2.5643205966330178e-05, "loss": 0.1223, "step": 47540 }, { "epoch": 2.22, "learning_rate": 2.5642737896494158e-05, "loss": 0.1329, "step": 47545 }, { "epoch": 2.22, "learning_rate": 2.5642269826658138e-05, "loss": 0.1225, "step": 47550 }, { "epoch": 2.22, "learning_rate": 2.5641801756822117e-05, "loss": 0.196, "step": 47555 }, { "epoch": 2.22, "learning_rate": 2.56413336869861e-05, "loss": 0.2821, "step": 47560 }, { "epoch": 2.22, "learning_rate": 2.564086561715008e-05, "loss": 0.3675, "step": 47565 }, { "epoch": 2.22, "learning_rate": 2.564039754731406e-05, "loss": 0.1089, "step": 47570 }, { "epoch": 2.22, "learning_rate": 2.563992947747804e-05, "loss": 0.0502, "step": 47575 }, { "epoch": 2.22, "learning_rate": 2.5639461407642023e-05, "loss": 0.04, "step": 47580 }, { "epoch": 2.22, "learning_rate": 2.5638993337806003e-05, "loss": 0.1266, "step": 47585 }, { "epoch": 2.22, "learning_rate": 2.563852526796998e-05, "loss": 0.069, "step": 47590 }, { "epoch": 2.22, "learning_rate": 2.563805719813396e-05, "loss": 0.0625, "step": 47595 }, { "epoch": 2.22, "learning_rate": 2.5637589128297942e-05, "loss": 0.1356, "step": 47600 }, { "epoch": 2.22, "learning_rate": 2.5637121058461922e-05, "loss": 0.1755, "step": 47605 }, { "epoch": 2.22, "learning_rate": 2.5636652988625902e-05, "loss": 0.2815, "step": 47610 }, { "epoch": 2.22, "learning_rate": 2.5636184918789885e-05, "loss": 0.3773, "step": 47615 }, { "epoch": 2.22, "learning_rate": 2.5635716848953865e-05, "loss": 0.0562, "step": 47620 }, { "epoch": 2.22, "learning_rate": 2.5635248779117845e-05, "loss": 0.0617, "step": 47625 }, { "epoch": 2.22, "learning_rate": 2.5634780709281825e-05, "loss": 0.0379, "step": 47630 }, { "epoch": 2.22, "learning_rate": 2.5634312639445808e-05, "loss": 0.0643, "step": 47635 }, { "epoch": 2.22, "learning_rate": 2.5633844569609788e-05, "loss": 0.0903, "step": 47640 }, { "epoch": 2.22, "learning_rate": 2.5633376499773768e-05, "loss": 0.1456, "step": 47645 }, { "epoch": 2.22, "learning_rate": 2.5632908429937747e-05, "loss": 0.165, "step": 47650 }, { "epoch": 2.22, "learning_rate": 2.5632440360101727e-05, "loss": 0.133, "step": 47655 }, { "epoch": 2.22, "learning_rate": 2.5631972290265707e-05, "loss": 0.1501, "step": 47660 }, { "epoch": 2.22, "learning_rate": 2.5631504220429687e-05, "loss": 0.1967, "step": 47665 }, { "epoch": 2.22, "learning_rate": 2.563103615059367e-05, "loss": 0.0856, "step": 47670 }, { "epoch": 2.22, "learning_rate": 2.563056808075765e-05, "loss": 0.0795, "step": 47675 }, { "epoch": 2.22, "learning_rate": 2.563010001092163e-05, "loss": 0.0539, "step": 47680 }, { "epoch": 2.23, "learning_rate": 2.562963194108561e-05, "loss": 0.0669, "step": 47685 }, { "epoch": 2.23, "learning_rate": 2.5629163871249593e-05, "loss": 0.0773, "step": 47690 }, { "epoch": 2.23, "learning_rate": 2.5628695801413573e-05, "loss": 0.0621, "step": 47695 }, { "epoch": 2.23, "learning_rate": 2.5628227731577552e-05, "loss": 0.1681, "step": 47700 }, { "epoch": 2.23, "learning_rate": 2.5627759661741532e-05, "loss": 0.1648, "step": 47705 }, { "epoch": 2.23, "learning_rate": 2.5627291591905515e-05, "loss": 0.374, "step": 47710 }, { "epoch": 2.23, "learning_rate": 2.5626823522069492e-05, "loss": 0.2538, "step": 47715 }, { "epoch": 2.23, "learning_rate": 2.562635545223347e-05, "loss": 0.0902, "step": 47720 }, { "epoch": 2.23, "learning_rate": 2.5625887382397455e-05, "loss": 0.045, "step": 47725 }, { "epoch": 2.23, "learning_rate": 2.5625419312561435e-05, "loss": 0.1089, "step": 47730 }, { "epoch": 2.23, "learning_rate": 2.5624951242725415e-05, "loss": 0.0462, "step": 47735 }, { "epoch": 2.23, "learning_rate": 2.5624483172889394e-05, "loss": 0.1399, "step": 47740 }, { "epoch": 2.23, "learning_rate": 2.5624015103053378e-05, "loss": 0.0925, "step": 47745 }, { "epoch": 2.23, "learning_rate": 2.5623547033217357e-05, "loss": 0.094, "step": 47750 }, { "epoch": 2.23, "learning_rate": 2.5623078963381337e-05, "loss": 0.2146, "step": 47755 }, { "epoch": 2.23, "learning_rate": 2.5622610893545317e-05, "loss": 0.2848, "step": 47760 }, { "epoch": 2.23, "learning_rate": 2.56221428237093e-05, "loss": 0.3323, "step": 47765 }, { "epoch": 2.23, "learning_rate": 2.562167475387328e-05, "loss": 0.0185, "step": 47770 }, { "epoch": 2.23, "learning_rate": 2.562120668403726e-05, "loss": 0.0359, "step": 47775 }, { "epoch": 2.23, "learning_rate": 2.5620738614201236e-05, "loss": 0.0438, "step": 47780 }, { "epoch": 2.23, "learning_rate": 2.562027054436522e-05, "loss": 0.103, "step": 47785 }, { "epoch": 2.23, "learning_rate": 2.56198024745292e-05, "loss": 0.1085, "step": 47790 }, { "epoch": 2.23, "learning_rate": 2.561933440469318e-05, "loss": 0.1381, "step": 47795 }, { "epoch": 2.23, "learning_rate": 2.5618866334857162e-05, "loss": 0.1167, "step": 47800 }, { "epoch": 2.23, "learning_rate": 2.5618398265021142e-05, "loss": 0.1696, "step": 47805 }, { "epoch": 2.23, "learning_rate": 2.5617930195185122e-05, "loss": 0.2581, "step": 47810 }, { "epoch": 2.23, "learning_rate": 2.5617462125349102e-05, "loss": 0.2795, "step": 47815 }, { "epoch": 2.23, "learning_rate": 2.5616994055513085e-05, "loss": 0.0395, "step": 47820 }, { "epoch": 2.23, "learning_rate": 2.5616525985677065e-05, "loss": 0.0924, "step": 47825 }, { "epoch": 2.23, "learning_rate": 2.5616057915841045e-05, "loss": 0.117, "step": 47830 }, { "epoch": 2.23, "learning_rate": 2.5615589846005024e-05, "loss": 0.0537, "step": 47835 }, { "epoch": 2.23, "learning_rate": 2.5615121776169008e-05, "loss": 0.0896, "step": 47840 }, { "epoch": 2.23, "learning_rate": 2.5614653706332984e-05, "loss": 0.1674, "step": 47845 }, { "epoch": 2.23, "learning_rate": 2.5614185636496964e-05, "loss": 0.1446, "step": 47850 }, { "epoch": 2.23, "learning_rate": 2.5613717566660947e-05, "loss": 0.1778, "step": 47855 }, { "epoch": 2.23, "learning_rate": 2.5613249496824927e-05, "loss": 0.2172, "step": 47860 }, { "epoch": 2.23, "learning_rate": 2.5612781426988907e-05, "loss": 0.2915, "step": 47865 }, { "epoch": 2.23, "learning_rate": 2.5612313357152887e-05, "loss": 0.042, "step": 47870 }, { "epoch": 2.23, "learning_rate": 2.561184528731687e-05, "loss": 0.0512, "step": 47875 }, { "epoch": 2.23, "learning_rate": 2.561137721748085e-05, "loss": 0.0285, "step": 47880 }, { "epoch": 2.23, "learning_rate": 2.561090914764483e-05, "loss": 0.0876, "step": 47885 }, { "epoch": 2.23, "learning_rate": 2.561044107780881e-05, "loss": 0.0687, "step": 47890 }, { "epoch": 2.23, "learning_rate": 2.5609973007972792e-05, "loss": 0.0614, "step": 47895 }, { "epoch": 2.24, "learning_rate": 2.5609504938136772e-05, "loss": 0.1214, "step": 47900 }, { "epoch": 2.24, "learning_rate": 2.560903686830075e-05, "loss": 0.162, "step": 47905 }, { "epoch": 2.24, "learning_rate": 2.5608568798464732e-05, "loss": 0.2478, "step": 47910 }, { "epoch": 2.24, "learning_rate": 2.560810072862871e-05, "loss": 0.3183, "step": 47915 }, { "epoch": 2.24, "learning_rate": 2.560763265879269e-05, "loss": 0.0465, "step": 47920 }, { "epoch": 2.24, "learning_rate": 2.560716458895667e-05, "loss": 0.0585, "step": 47925 }, { "epoch": 2.24, "learning_rate": 2.5606696519120655e-05, "loss": 0.0294, "step": 47930 }, { "epoch": 2.24, "learning_rate": 2.5606228449284634e-05, "loss": 0.0236, "step": 47935 }, { "epoch": 2.24, "learning_rate": 2.5605760379448614e-05, "loss": 0.1196, "step": 47940 }, { "epoch": 2.24, "learning_rate": 2.5605292309612594e-05, "loss": 0.1423, "step": 47945 }, { "epoch": 2.24, "learning_rate": 2.5604824239776577e-05, "loss": 0.1468, "step": 47950 }, { "epoch": 2.24, "learning_rate": 2.5604356169940557e-05, "loss": 0.2614, "step": 47955 }, { "epoch": 2.24, "learning_rate": 2.5603888100104537e-05, "loss": 0.1917, "step": 47960 }, { "epoch": 2.24, "learning_rate": 2.5603420030268517e-05, "loss": 0.244, "step": 47965 }, { "epoch": 2.24, "learning_rate": 2.5602951960432496e-05, "loss": 0.0856, "step": 47970 }, { "epoch": 2.24, "learning_rate": 2.5602483890596476e-05, "loss": 0.0494, "step": 47975 }, { "epoch": 2.24, "learning_rate": 2.5602015820760456e-05, "loss": 0.0748, "step": 47980 }, { "epoch": 2.24, "learning_rate": 2.560154775092444e-05, "loss": 0.0276, "step": 47985 }, { "epoch": 2.24, "learning_rate": 2.560107968108842e-05, "loss": 0.1039, "step": 47990 }, { "epoch": 2.24, "learning_rate": 2.56006116112524e-05, "loss": 0.113, "step": 47995 }, { "epoch": 2.24, "learning_rate": 2.560014354141638e-05, "loss": 0.1198, "step": 48000 }, { "epoch": 2.24, "learning_rate": 2.5599675471580362e-05, "loss": 0.1373, "step": 48005 }, { "epoch": 2.24, "learning_rate": 2.5599207401744342e-05, "loss": 0.3318, "step": 48010 }, { "epoch": 2.24, "learning_rate": 2.559873933190832e-05, "loss": 0.2747, "step": 48015 }, { "epoch": 2.24, "learning_rate": 2.55982712620723e-05, "loss": 0.0327, "step": 48020 }, { "epoch": 2.24, "learning_rate": 2.5597803192236285e-05, "loss": 0.0317, "step": 48025 }, { "epoch": 2.24, "learning_rate": 2.5597335122400264e-05, "loss": 0.0481, "step": 48030 }, { "epoch": 2.24, "learning_rate": 2.559686705256424e-05, "loss": 0.0458, "step": 48035 }, { "epoch": 2.24, "learning_rate": 2.5596398982728224e-05, "loss": 0.0291, "step": 48040 }, { "epoch": 2.24, "learning_rate": 2.5595930912892204e-05, "loss": 0.1611, "step": 48045 }, { "epoch": 2.24, "learning_rate": 2.5595462843056184e-05, "loss": 0.1481, "step": 48050 }, { "epoch": 2.24, "learning_rate": 2.5594994773220163e-05, "loss": 0.1921, "step": 48055 }, { "epoch": 2.24, "learning_rate": 2.5594526703384147e-05, "loss": 0.2832, "step": 48060 }, { "epoch": 2.24, "learning_rate": 2.5594058633548127e-05, "loss": 0.38, "step": 48065 }, { "epoch": 2.24, "learning_rate": 2.5593590563712106e-05, "loss": 0.0514, "step": 48070 }, { "epoch": 2.24, "learning_rate": 2.5593122493876086e-05, "loss": 0.0453, "step": 48075 }, { "epoch": 2.24, "learning_rate": 2.559265442404007e-05, "loss": 0.074, "step": 48080 }, { "epoch": 2.24, "learning_rate": 2.559218635420405e-05, "loss": 0.055, "step": 48085 }, { "epoch": 2.24, "learning_rate": 2.559171828436803e-05, "loss": 0.0755, "step": 48090 }, { "epoch": 2.24, "learning_rate": 2.559125021453201e-05, "loss": 0.0403, "step": 48095 }, { "epoch": 2.24, "learning_rate": 2.559078214469599e-05, "loss": 0.2185, "step": 48100 }, { "epoch": 2.24, "learning_rate": 2.559031407485997e-05, "loss": 0.2096, "step": 48105 }, { "epoch": 2.24, "learning_rate": 2.5589846005023948e-05, "loss": 0.3046, "step": 48110 }, { "epoch": 2.25, "learning_rate": 2.558937793518793e-05, "loss": 0.281, "step": 48115 }, { "epoch": 2.25, "learning_rate": 2.558890986535191e-05, "loss": 0.0864, "step": 48120 }, { "epoch": 2.25, "learning_rate": 2.558844179551589e-05, "loss": 0.0291, "step": 48125 }, { "epoch": 2.25, "learning_rate": 2.558797372567987e-05, "loss": 0.0645, "step": 48130 }, { "epoch": 2.25, "learning_rate": 2.5587505655843854e-05, "loss": 0.0875, "step": 48135 }, { "epoch": 2.25, "learning_rate": 2.5587037586007834e-05, "loss": 0.1075, "step": 48140 }, { "epoch": 2.25, "learning_rate": 2.5586569516171814e-05, "loss": 0.0819, "step": 48145 }, { "epoch": 2.25, "learning_rate": 2.5586101446335794e-05, "loss": 0.1165, "step": 48150 }, { "epoch": 2.25, "learning_rate": 2.5585633376499777e-05, "loss": 0.1718, "step": 48155 }, { "epoch": 2.25, "learning_rate": 2.5585165306663753e-05, "loss": 0.127, "step": 48160 }, { "epoch": 2.25, "learning_rate": 2.5584697236827733e-05, "loss": 0.3313, "step": 48165 }, { "epoch": 2.25, "learning_rate": 2.5584229166991716e-05, "loss": 0.0502, "step": 48170 }, { "epoch": 2.25, "learning_rate": 2.5583761097155696e-05, "loss": 0.0626, "step": 48175 }, { "epoch": 2.25, "learning_rate": 2.5583293027319676e-05, "loss": 0.0683, "step": 48180 }, { "epoch": 2.25, "learning_rate": 2.5582824957483656e-05, "loss": 0.0778, "step": 48185 }, { "epoch": 2.25, "learning_rate": 2.558235688764764e-05, "loss": 0.1569, "step": 48190 }, { "epoch": 2.25, "learning_rate": 2.558188881781162e-05, "loss": 0.1026, "step": 48195 }, { "epoch": 2.25, "learning_rate": 2.55814207479756e-05, "loss": 0.1475, "step": 48200 }, { "epoch": 2.25, "learning_rate": 2.558095267813958e-05, "loss": 0.0808, "step": 48205 }, { "epoch": 2.25, "learning_rate": 2.558048460830356e-05, "loss": 0.3092, "step": 48210 }, { "epoch": 2.25, "learning_rate": 2.558001653846754e-05, "loss": 0.32, "step": 48215 }, { "epoch": 2.25, "learning_rate": 2.557954846863152e-05, "loss": 0.0459, "step": 48220 }, { "epoch": 2.25, "learning_rate": 2.55790803987955e-05, "loss": 0.0482, "step": 48225 }, { "epoch": 2.25, "learning_rate": 2.557861232895948e-05, "loss": 0.1041, "step": 48230 }, { "epoch": 2.25, "learning_rate": 2.557814425912346e-05, "loss": 0.0626, "step": 48235 }, { "epoch": 2.25, "learning_rate": 2.557767618928744e-05, "loss": 0.1226, "step": 48240 }, { "epoch": 2.25, "learning_rate": 2.5577208119451424e-05, "loss": 0.1336, "step": 48245 }, { "epoch": 2.25, "learning_rate": 2.5576740049615403e-05, "loss": 0.1735, "step": 48250 }, { "epoch": 2.25, "learning_rate": 2.5576271979779383e-05, "loss": 0.1934, "step": 48255 }, { "epoch": 2.25, "learning_rate": 2.5575803909943363e-05, "loss": 0.3538, "step": 48260 }, { "epoch": 2.25, "learning_rate": 2.5575335840107346e-05, "loss": 0.223, "step": 48265 }, { "epoch": 2.25, "learning_rate": 2.5574867770271326e-05, "loss": 0.0087, "step": 48270 }, { "epoch": 2.25, "learning_rate": 2.5574399700435306e-05, "loss": 0.1056, "step": 48275 }, { "epoch": 2.25, "learning_rate": 2.557393163059929e-05, "loss": 0.0458, "step": 48280 }, { "epoch": 2.25, "learning_rate": 2.5573463560763266e-05, "loss": 0.0967, "step": 48285 }, { "epoch": 2.25, "learning_rate": 2.5572995490927245e-05, "loss": 0.0967, "step": 48290 }, { "epoch": 2.25, "learning_rate": 2.5572527421091225e-05, "loss": 0.071, "step": 48295 }, { "epoch": 2.25, "learning_rate": 2.557205935125521e-05, "loss": 0.1769, "step": 48300 }, { "epoch": 2.25, "learning_rate": 2.5571591281419188e-05, "loss": 0.1909, "step": 48305 }, { "epoch": 2.25, "learning_rate": 2.5571123211583168e-05, "loss": 0.1933, "step": 48310 }, { "epoch": 2.25, "learning_rate": 2.5570655141747148e-05, "loss": 0.2301, "step": 48315 }, { "epoch": 2.25, "learning_rate": 2.557018707191113e-05, "loss": 0.0581, "step": 48320 }, { "epoch": 2.25, "learning_rate": 2.556971900207511e-05, "loss": 0.0651, "step": 48325 }, { "epoch": 2.26, "learning_rate": 2.556925093223909e-05, "loss": 0.066, "step": 48330 }, { "epoch": 2.26, "learning_rate": 2.5568782862403074e-05, "loss": 0.0422, "step": 48335 }, { "epoch": 2.26, "learning_rate": 2.5568314792567054e-05, "loss": 0.098, "step": 48340 }, { "epoch": 2.26, "learning_rate": 2.5567846722731034e-05, "loss": 0.1011, "step": 48345 }, { "epoch": 2.26, "learning_rate": 2.556737865289501e-05, "loss": 0.1688, "step": 48350 }, { "epoch": 2.26, "learning_rate": 2.5566910583058993e-05, "loss": 0.1351, "step": 48355 }, { "epoch": 2.26, "learning_rate": 2.5566442513222973e-05, "loss": 0.1699, "step": 48360 }, { "epoch": 2.26, "learning_rate": 2.5565974443386953e-05, "loss": 0.3377, "step": 48365 }, { "epoch": 2.26, "learning_rate": 2.5565506373550933e-05, "loss": 0.0536, "step": 48370 }, { "epoch": 2.26, "learning_rate": 2.5565038303714916e-05, "loss": 0.0161, "step": 48375 }, { "epoch": 2.26, "learning_rate": 2.5564570233878896e-05, "loss": 0.0467, "step": 48380 }, { "epoch": 2.26, "learning_rate": 2.5564102164042876e-05, "loss": 0.0788, "step": 48385 }, { "epoch": 2.26, "learning_rate": 2.5563634094206855e-05, "loss": 0.1354, "step": 48390 }, { "epoch": 2.26, "learning_rate": 2.556316602437084e-05, "loss": 0.088, "step": 48395 }, { "epoch": 2.26, "learning_rate": 2.556269795453482e-05, "loss": 0.1352, "step": 48400 }, { "epoch": 2.26, "learning_rate": 2.5562229884698798e-05, "loss": 0.1948, "step": 48405 }, { "epoch": 2.26, "learning_rate": 2.5561761814862778e-05, "loss": 0.2968, "step": 48410 }, { "epoch": 2.26, "learning_rate": 2.5561293745026758e-05, "loss": 0.2943, "step": 48415 }, { "epoch": 2.26, "learning_rate": 2.5560825675190738e-05, "loss": 0.0413, "step": 48420 }, { "epoch": 2.26, "learning_rate": 2.5560357605354717e-05, "loss": 0.0396, "step": 48425 }, { "epoch": 2.26, "learning_rate": 2.55598895355187e-05, "loss": 0.0293, "step": 48430 }, { "epoch": 2.26, "learning_rate": 2.555942146568268e-05, "loss": 0.0459, "step": 48435 }, { "epoch": 2.26, "learning_rate": 2.555895339584666e-05, "loss": 0.1337, "step": 48440 }, { "epoch": 2.26, "learning_rate": 2.555848532601064e-05, "loss": 0.0624, "step": 48445 }, { "epoch": 2.26, "learning_rate": 2.5558017256174623e-05, "loss": 0.1602, "step": 48450 }, { "epoch": 2.26, "learning_rate": 2.5557549186338603e-05, "loss": 0.1678, "step": 48455 }, { "epoch": 2.26, "learning_rate": 2.5557081116502583e-05, "loss": 0.3125, "step": 48460 }, { "epoch": 2.26, "learning_rate": 2.5556613046666566e-05, "loss": 0.1925, "step": 48465 }, { "epoch": 2.26, "learning_rate": 2.5556144976830546e-05, "loss": 0.0865, "step": 48470 }, { "epoch": 2.26, "learning_rate": 2.5555676906994522e-05, "loss": 0.0542, "step": 48475 }, { "epoch": 2.26, "learning_rate": 2.5555208837158502e-05, "loss": 0.0358, "step": 48480 }, { "epoch": 2.26, "learning_rate": 2.5554740767322485e-05, "loss": 0.0612, "step": 48485 }, { "epoch": 2.26, "learning_rate": 2.5554272697486465e-05, "loss": 0.0953, "step": 48490 }, { "epoch": 2.26, "learning_rate": 2.5553804627650445e-05, "loss": 0.071, "step": 48495 }, { "epoch": 2.26, "learning_rate": 2.5553336557814425e-05, "loss": 0.1023, "step": 48500 }, { "epoch": 2.26, "learning_rate": 2.5552868487978408e-05, "loss": 0.1225, "step": 48505 }, { "epoch": 2.26, "learning_rate": 2.5552400418142388e-05, "loss": 0.185, "step": 48510 }, { "epoch": 2.26, "learning_rate": 2.5551932348306368e-05, "loss": 0.3224, "step": 48515 }, { "epoch": 2.26, "learning_rate": 2.555146427847035e-05, "loss": 0.0792, "step": 48520 }, { "epoch": 2.26, "learning_rate": 2.555099620863433e-05, "loss": 0.0224, "step": 48525 }, { "epoch": 2.26, "learning_rate": 2.555052813879831e-05, "loss": 0.0744, "step": 48530 }, { "epoch": 2.26, "learning_rate": 2.555006006896229e-05, "loss": 0.0628, "step": 48535 }, { "epoch": 2.26, "learning_rate": 2.554959199912627e-05, "loss": 0.079, "step": 48540 }, { "epoch": 2.27, "learning_rate": 2.554912392929025e-05, "loss": 0.1051, "step": 48545 }, { "epoch": 2.27, "learning_rate": 2.554865585945423e-05, "loss": 0.1271, "step": 48550 }, { "epoch": 2.27, "learning_rate": 2.554818778961821e-05, "loss": 0.1691, "step": 48555 }, { "epoch": 2.27, "learning_rate": 2.5547719719782193e-05, "loss": 0.2027, "step": 48560 }, { "epoch": 2.27, "learning_rate": 2.5547251649946173e-05, "loss": 0.2881, "step": 48565 }, { "epoch": 2.27, "learning_rate": 2.5546783580110152e-05, "loss": 0.0433, "step": 48570 }, { "epoch": 2.27, "learning_rate": 2.5546315510274132e-05, "loss": 0.0822, "step": 48575 }, { "epoch": 2.27, "learning_rate": 2.5545847440438116e-05, "loss": 0.039, "step": 48580 }, { "epoch": 2.27, "learning_rate": 2.5545379370602095e-05, "loss": 0.0476, "step": 48585 }, { "epoch": 2.27, "learning_rate": 2.5544911300766075e-05, "loss": 0.1121, "step": 48590 }, { "epoch": 2.27, "learning_rate": 2.554444323093006e-05, "loss": 0.0562, "step": 48595 }, { "epoch": 2.27, "learning_rate": 2.5543975161094035e-05, "loss": 0.1988, "step": 48600 }, { "epoch": 2.27, "learning_rate": 2.5543507091258015e-05, "loss": 0.1998, "step": 48605 }, { "epoch": 2.27, "learning_rate": 2.5543039021421994e-05, "loss": 0.1557, "step": 48610 }, { "epoch": 2.27, "learning_rate": 2.5542570951585978e-05, "loss": 0.2661, "step": 48615 }, { "epoch": 2.27, "learning_rate": 2.5542102881749957e-05, "loss": 0.0776, "step": 48620 }, { "epoch": 2.27, "learning_rate": 2.5541634811913937e-05, "loss": 0.0704, "step": 48625 }, { "epoch": 2.27, "learning_rate": 2.5541166742077917e-05, "loss": 0.0453, "step": 48630 }, { "epoch": 2.27, "learning_rate": 2.55406986722419e-05, "loss": 0.0816, "step": 48635 }, { "epoch": 2.27, "learning_rate": 2.554023060240588e-05, "loss": 0.0485, "step": 48640 }, { "epoch": 2.27, "learning_rate": 2.553976253256986e-05, "loss": 0.14, "step": 48645 }, { "epoch": 2.27, "learning_rate": 2.5539294462733843e-05, "loss": 0.1704, "step": 48650 }, { "epoch": 2.27, "learning_rate": 2.5538826392897823e-05, "loss": 0.0825, "step": 48655 }, { "epoch": 2.27, "learning_rate": 2.5538358323061803e-05, "loss": 0.2581, "step": 48660 }, { "epoch": 2.27, "learning_rate": 2.553789025322578e-05, "loss": 0.2511, "step": 48665 }, { "epoch": 2.27, "learning_rate": 2.5537422183389762e-05, "loss": 0.0491, "step": 48670 }, { "epoch": 2.27, "learning_rate": 2.5536954113553742e-05, "loss": 0.0602, "step": 48675 }, { "epoch": 2.27, "learning_rate": 2.5536486043717722e-05, "loss": 0.0596, "step": 48680 }, { "epoch": 2.27, "learning_rate": 2.5536017973881702e-05, "loss": 0.0686, "step": 48685 }, { "epoch": 2.27, "learning_rate": 2.5535549904045685e-05, "loss": 0.0731, "step": 48690 }, { "epoch": 2.27, "learning_rate": 2.5535081834209665e-05, "loss": 0.1049, "step": 48695 }, { "epoch": 2.27, "learning_rate": 2.5534613764373645e-05, "loss": 0.1095, "step": 48700 }, { "epoch": 2.27, "learning_rate": 2.5534145694537628e-05, "loss": 0.1277, "step": 48705 }, { "epoch": 2.27, "learning_rate": 2.5533677624701608e-05, "loss": 0.3545, "step": 48710 }, { "epoch": 2.27, "learning_rate": 2.5533209554865588e-05, "loss": 0.3105, "step": 48715 }, { "epoch": 2.27, "learning_rate": 2.5532741485029567e-05, "loss": 0.0461, "step": 48720 }, { "epoch": 2.27, "learning_rate": 2.553227341519355e-05, "loss": 0.0943, "step": 48725 }, { "epoch": 2.27, "learning_rate": 2.5531805345357527e-05, "loss": 0.0464, "step": 48730 }, { "epoch": 2.27, "learning_rate": 2.5531337275521507e-05, "loss": 0.0793, "step": 48735 }, { "epoch": 2.27, "learning_rate": 2.5530869205685487e-05, "loss": 0.1217, "step": 48740 }, { "epoch": 2.27, "learning_rate": 2.553040113584947e-05, "loss": 0.1033, "step": 48745 }, { "epoch": 2.27, "learning_rate": 2.552993306601345e-05, "loss": 0.1486, "step": 48750 }, { "epoch": 2.27, "learning_rate": 2.552946499617743e-05, "loss": 0.185, "step": 48755 }, { "epoch": 2.28, "learning_rate": 2.552899692634141e-05, "loss": 0.2401, "step": 48760 }, { "epoch": 2.28, "learning_rate": 2.5528528856505392e-05, "loss": 0.2602, "step": 48765 }, { "epoch": 2.28, "learning_rate": 2.5528060786669372e-05, "loss": 0.0412, "step": 48770 }, { "epoch": 2.28, "learning_rate": 2.5527592716833352e-05, "loss": 0.0727, "step": 48775 }, { "epoch": 2.28, "learning_rate": 2.5527124646997335e-05, "loss": 0.061, "step": 48780 }, { "epoch": 2.28, "learning_rate": 2.5526656577161315e-05, "loss": 0.0596, "step": 48785 }, { "epoch": 2.28, "learning_rate": 2.552618850732529e-05, "loss": 0.0897, "step": 48790 }, { "epoch": 2.28, "learning_rate": 2.552572043748927e-05, "loss": 0.1228, "step": 48795 }, { "epoch": 2.28, "learning_rate": 2.5525252367653255e-05, "loss": 0.1323, "step": 48800 }, { "epoch": 2.28, "learning_rate": 2.5524784297817234e-05, "loss": 0.1526, "step": 48805 }, { "epoch": 2.28, "learning_rate": 2.5524316227981214e-05, "loss": 0.1709, "step": 48810 }, { "epoch": 2.28, "learning_rate": 2.5523848158145194e-05, "loss": 0.2357, "step": 48815 }, { "epoch": 2.28, "learning_rate": 2.5523380088309177e-05, "loss": 0.0479, "step": 48820 }, { "epoch": 2.28, "learning_rate": 2.5522912018473157e-05, "loss": 0.1354, "step": 48825 }, { "epoch": 2.28, "learning_rate": 2.5522443948637137e-05, "loss": 0.0579, "step": 48830 }, { "epoch": 2.28, "learning_rate": 2.552197587880112e-05, "loss": 0.0505, "step": 48835 }, { "epoch": 2.28, "learning_rate": 2.55215078089651e-05, "loss": 0.1908, "step": 48840 }, { "epoch": 2.28, "learning_rate": 2.552103973912908e-05, "loss": 0.0595, "step": 48845 }, { "epoch": 2.28, "learning_rate": 2.552057166929306e-05, "loss": 0.2049, "step": 48850 }, { "epoch": 2.28, "learning_rate": 2.552010359945704e-05, "loss": 0.1395, "step": 48855 }, { "epoch": 2.28, "learning_rate": 2.551963552962102e-05, "loss": 0.162, "step": 48860 }, { "epoch": 2.28, "learning_rate": 2.5519167459785e-05, "loss": 0.1792, "step": 48865 }, { "epoch": 2.28, "learning_rate": 2.551869938994898e-05, "loss": 0.0393, "step": 48870 }, { "epoch": 2.28, "learning_rate": 2.5518231320112962e-05, "loss": 0.0813, "step": 48875 }, { "epoch": 2.28, "learning_rate": 2.5517763250276942e-05, "loss": 0.019, "step": 48880 }, { "epoch": 2.28, "learning_rate": 2.551729518044092e-05, "loss": 0.0596, "step": 48885 }, { "epoch": 2.28, "learning_rate": 2.5516827110604905e-05, "loss": 0.1122, "step": 48890 }, { "epoch": 2.28, "learning_rate": 2.5516359040768885e-05, "loss": 0.084, "step": 48895 }, { "epoch": 2.28, "learning_rate": 2.5515890970932864e-05, "loss": 0.1043, "step": 48900 }, { "epoch": 2.28, "learning_rate": 2.5515422901096844e-05, "loss": 0.1767, "step": 48905 }, { "epoch": 2.28, "learning_rate": 2.5514954831260828e-05, "loss": 0.265, "step": 48910 }, { "epoch": 2.28, "learning_rate": 2.5514486761424804e-05, "loss": 0.2737, "step": 48915 }, { "epoch": 2.28, "learning_rate": 2.5514018691588784e-05, "loss": 0.0207, "step": 48920 }, { "epoch": 2.28, "learning_rate": 2.5513550621752764e-05, "loss": 0.0435, "step": 48925 }, { "epoch": 2.28, "learning_rate": 2.5513082551916747e-05, "loss": 0.0996, "step": 48930 }, { "epoch": 2.28, "learning_rate": 2.5512614482080727e-05, "loss": 0.0856, "step": 48935 }, { "epoch": 2.28, "learning_rate": 2.5512146412244706e-05, "loss": 0.085, "step": 48940 }, { "epoch": 2.28, "learning_rate": 2.551167834240869e-05, "loss": 0.0692, "step": 48945 }, { "epoch": 2.28, "learning_rate": 2.551121027257267e-05, "loss": 0.1815, "step": 48950 }, { "epoch": 2.28, "learning_rate": 2.551074220273665e-05, "loss": 0.2231, "step": 48955 }, { "epoch": 2.28, "learning_rate": 2.551027413290063e-05, "loss": 0.1885, "step": 48960 }, { "epoch": 2.28, "learning_rate": 2.5509806063064612e-05, "loss": 0.2617, "step": 48965 }, { "epoch": 2.29, "learning_rate": 2.5509337993228592e-05, "loss": 0.0374, "step": 48970 }, { "epoch": 2.29, "learning_rate": 2.5508869923392572e-05, "loss": 0.0236, "step": 48975 }, { "epoch": 2.29, "learning_rate": 2.550840185355655e-05, "loss": 0.0527, "step": 48980 }, { "epoch": 2.29, "learning_rate": 2.550793378372053e-05, "loss": 0.0911, "step": 48985 }, { "epoch": 2.29, "learning_rate": 2.550746571388451e-05, "loss": 0.1677, "step": 48990 }, { "epoch": 2.29, "learning_rate": 2.550699764404849e-05, "loss": 0.0754, "step": 48995 }, { "epoch": 2.29, "learning_rate": 2.550652957421247e-05, "loss": 0.1093, "step": 49000 }, { "epoch": 2.29, "learning_rate": 2.5506061504376454e-05, "loss": 0.1782, "step": 49005 }, { "epoch": 2.29, "learning_rate": 2.5505593434540434e-05, "loss": 0.3286, "step": 49010 }, { "epoch": 2.29, "learning_rate": 2.5505125364704414e-05, "loss": 0.3655, "step": 49015 }, { "epoch": 2.29, "learning_rate": 2.5504657294868397e-05, "loss": 0.0308, "step": 49020 }, { "epoch": 2.29, "learning_rate": 2.5504189225032377e-05, "loss": 0.0628, "step": 49025 }, { "epoch": 2.29, "learning_rate": 2.5503721155196357e-05, "loss": 0.0821, "step": 49030 }, { "epoch": 2.29, "learning_rate": 2.5503253085360337e-05, "loss": 0.1082, "step": 49035 }, { "epoch": 2.29, "learning_rate": 2.550278501552432e-05, "loss": 0.0796, "step": 49040 }, { "epoch": 2.29, "learning_rate": 2.5502316945688296e-05, "loss": 0.0901, "step": 49045 }, { "epoch": 2.29, "learning_rate": 2.5501848875852276e-05, "loss": 0.1315, "step": 49050 }, { "epoch": 2.29, "learning_rate": 2.5501380806016256e-05, "loss": 0.174, "step": 49055 }, { "epoch": 2.29, "learning_rate": 2.550091273618024e-05, "loss": 0.2425, "step": 49060 }, { "epoch": 2.29, "learning_rate": 2.550044466634422e-05, "loss": 0.3346, "step": 49065 }, { "epoch": 2.29, "learning_rate": 2.54999765965082e-05, "loss": 0.0782, "step": 49070 }, { "epoch": 2.29, "learning_rate": 2.5499508526672182e-05, "loss": 0.0316, "step": 49075 }, { "epoch": 2.29, "learning_rate": 2.549904045683616e-05, "loss": 0.1175, "step": 49080 }, { "epoch": 2.29, "learning_rate": 2.549857238700014e-05, "loss": 0.0944, "step": 49085 }, { "epoch": 2.29, "learning_rate": 2.549810431716412e-05, "loss": 0.0893, "step": 49090 }, { "epoch": 2.29, "learning_rate": 2.5497636247328104e-05, "loss": 0.0829, "step": 49095 }, { "epoch": 2.29, "learning_rate": 2.5497168177492084e-05, "loss": 0.1089, "step": 49100 }, { "epoch": 2.29, "learning_rate": 2.549670010765606e-05, "loss": 0.2078, "step": 49105 }, { "epoch": 2.29, "learning_rate": 2.549623203782004e-05, "loss": 0.3748, "step": 49110 }, { "epoch": 2.29, "learning_rate": 2.5495763967984024e-05, "loss": 0.2284, "step": 49115 }, { "epoch": 2.29, "learning_rate": 2.5495295898148004e-05, "loss": 0.0441, "step": 49120 }, { "epoch": 2.29, "learning_rate": 2.5494827828311983e-05, "loss": 0.058, "step": 49125 }, { "epoch": 2.29, "learning_rate": 2.5494359758475967e-05, "loss": 0.0295, "step": 49130 }, { "epoch": 2.29, "learning_rate": 2.5493891688639946e-05, "loss": 0.0434, "step": 49135 }, { "epoch": 2.29, "learning_rate": 2.5493423618803926e-05, "loss": 0.1006, "step": 49140 }, { "epoch": 2.29, "learning_rate": 2.5492955548967906e-05, "loss": 0.1513, "step": 49145 }, { "epoch": 2.29, "learning_rate": 2.549248747913189e-05, "loss": 0.0728, "step": 49150 }, { "epoch": 2.29, "learning_rate": 2.549201940929587e-05, "loss": 0.1477, "step": 49155 }, { "epoch": 2.29, "learning_rate": 2.549155133945985e-05, "loss": 0.2181, "step": 49160 }, { "epoch": 2.29, "learning_rate": 2.549108326962383e-05, "loss": 0.2554, "step": 49165 }, { "epoch": 2.29, "learning_rate": 2.549061519978781e-05, "loss": 0.0275, "step": 49170 }, { "epoch": 2.29, "learning_rate": 2.549014712995179e-05, "loss": 0.0503, "step": 49175 }, { "epoch": 2.29, "learning_rate": 2.5489679060115768e-05, "loss": 0.074, "step": 49180 }, { "epoch": 2.3, "learning_rate": 2.5489210990279748e-05, "loss": 0.0854, "step": 49185 }, { "epoch": 2.3, "learning_rate": 2.548874292044373e-05, "loss": 0.1153, "step": 49190 }, { "epoch": 2.3, "learning_rate": 2.548827485060771e-05, "loss": 0.1457, "step": 49195 }, { "epoch": 2.3, "learning_rate": 2.548780678077169e-05, "loss": 0.1199, "step": 49200 }, { "epoch": 2.3, "learning_rate": 2.5487338710935674e-05, "loss": 0.1506, "step": 49205 }, { "epoch": 2.3, "learning_rate": 2.5486870641099654e-05, "loss": 0.2608, "step": 49210 }, { "epoch": 2.3, "learning_rate": 2.5486402571263634e-05, "loss": 0.2974, "step": 49215 }, { "epoch": 2.3, "learning_rate": 2.5485934501427613e-05, "loss": 0.04, "step": 49220 }, { "epoch": 2.3, "learning_rate": 2.5485466431591597e-05, "loss": 0.0455, "step": 49225 }, { "epoch": 2.3, "learning_rate": 2.5484998361755577e-05, "loss": 0.0973, "step": 49230 }, { "epoch": 2.3, "learning_rate": 2.5484530291919553e-05, "loss": 0.0707, "step": 49235 }, { "epoch": 2.3, "learning_rate": 2.5484062222083533e-05, "loss": 0.0664, "step": 49240 }, { "epoch": 2.3, "learning_rate": 2.5483594152247516e-05, "loss": 0.1586, "step": 49245 }, { "epoch": 2.3, "learning_rate": 2.5483126082411496e-05, "loss": 0.1245, "step": 49250 }, { "epoch": 2.3, "learning_rate": 2.5482658012575476e-05, "loss": 0.1546, "step": 49255 }, { "epoch": 2.3, "learning_rate": 2.548218994273946e-05, "loss": 0.2112, "step": 49260 }, { "epoch": 2.3, "learning_rate": 2.548172187290344e-05, "loss": 0.2794, "step": 49265 }, { "epoch": 2.3, "learning_rate": 2.548125380306742e-05, "loss": 0.0569, "step": 49270 }, { "epoch": 2.3, "learning_rate": 2.5480785733231398e-05, "loss": 0.0636, "step": 49275 }, { "epoch": 2.3, "learning_rate": 2.548031766339538e-05, "loss": 0.0337, "step": 49280 }, { "epoch": 2.3, "learning_rate": 2.547984959355936e-05, "loss": 0.1443, "step": 49285 }, { "epoch": 2.3, "learning_rate": 2.547938152372334e-05, "loss": 0.0839, "step": 49290 }, { "epoch": 2.3, "learning_rate": 2.5478913453887318e-05, "loss": 0.0944, "step": 49295 }, { "epoch": 2.3, "learning_rate": 2.54784453840513e-05, "loss": 0.1429, "step": 49300 }, { "epoch": 2.3, "learning_rate": 2.547797731421528e-05, "loss": 0.132, "step": 49305 }, { "epoch": 2.3, "learning_rate": 2.547750924437926e-05, "loss": 0.2637, "step": 49310 }, { "epoch": 2.3, "learning_rate": 2.5477041174543244e-05, "loss": 0.4144, "step": 49315 }, { "epoch": 2.3, "learning_rate": 2.5476573104707223e-05, "loss": 0.076, "step": 49320 }, { "epoch": 2.3, "learning_rate": 2.5476105034871203e-05, "loss": 0.0528, "step": 49325 }, { "epoch": 2.3, "learning_rate": 2.5475636965035183e-05, "loss": 0.0759, "step": 49330 }, { "epoch": 2.3, "learning_rate": 2.5475168895199166e-05, "loss": 0.0648, "step": 49335 }, { "epoch": 2.3, "learning_rate": 2.5474700825363146e-05, "loss": 0.0784, "step": 49340 }, { "epoch": 2.3, "learning_rate": 2.5474232755527126e-05, "loss": 0.1186, "step": 49345 }, { "epoch": 2.3, "learning_rate": 2.5473764685691106e-05, "loss": 0.0815, "step": 49350 }, { "epoch": 2.3, "learning_rate": 2.547329661585509e-05, "loss": 0.1659, "step": 49355 }, { "epoch": 2.3, "learning_rate": 2.5472828546019065e-05, "loss": 0.3107, "step": 49360 }, { "epoch": 2.3, "learning_rate": 2.5472360476183045e-05, "loss": 0.3417, "step": 49365 }, { "epoch": 2.3, "learning_rate": 2.5471892406347025e-05, "loss": 0.0458, "step": 49370 }, { "epoch": 2.3, "learning_rate": 2.5471424336511008e-05, "loss": 0.0509, "step": 49375 }, { "epoch": 2.3, "learning_rate": 2.5470956266674988e-05, "loss": 0.0771, "step": 49380 }, { "epoch": 2.3, "learning_rate": 2.5470488196838968e-05, "loss": 0.1004, "step": 49385 }, { "epoch": 2.3, "learning_rate": 2.547002012700295e-05, "loss": 0.1161, "step": 49390 }, { "epoch": 2.3, "learning_rate": 2.546955205716693e-05, "loss": 0.2007, "step": 49395 }, { "epoch": 2.31, "learning_rate": 2.546908398733091e-05, "loss": 0.1251, "step": 49400 }, { "epoch": 2.31, "learning_rate": 2.546861591749489e-05, "loss": 0.1663, "step": 49405 }, { "epoch": 2.31, "learning_rate": 2.5468147847658874e-05, "loss": 0.2334, "step": 49410 }, { "epoch": 2.31, "learning_rate": 2.5467679777822853e-05, "loss": 0.3047, "step": 49415 }, { "epoch": 2.31, "learning_rate": 2.5467211707986833e-05, "loss": 0.0639, "step": 49420 }, { "epoch": 2.31, "learning_rate": 2.546674363815081e-05, "loss": 0.0752, "step": 49425 }, { "epoch": 2.31, "learning_rate": 2.5466275568314793e-05, "loss": 0.0783, "step": 49430 }, { "epoch": 2.31, "learning_rate": 2.5465807498478773e-05, "loss": 0.0544, "step": 49435 }, { "epoch": 2.31, "learning_rate": 2.5465339428642753e-05, "loss": 0.034, "step": 49440 }, { "epoch": 2.31, "learning_rate": 2.5464871358806736e-05, "loss": 0.1552, "step": 49445 }, { "epoch": 2.31, "learning_rate": 2.5464403288970716e-05, "loss": 0.1129, "step": 49450 }, { "epoch": 2.31, "learning_rate": 2.5463935219134695e-05, "loss": 0.2289, "step": 49455 }, { "epoch": 2.31, "learning_rate": 2.5463467149298675e-05, "loss": 0.2528, "step": 49460 }, { "epoch": 2.31, "learning_rate": 2.546299907946266e-05, "loss": 0.3527, "step": 49465 }, { "epoch": 2.31, "learning_rate": 2.5462531009626638e-05, "loss": 0.0871, "step": 49470 }, { "epoch": 2.31, "learning_rate": 2.5462062939790618e-05, "loss": 0.0753, "step": 49475 }, { "epoch": 2.31, "learning_rate": 2.5461594869954598e-05, "loss": 0.0513, "step": 49480 }, { "epoch": 2.31, "learning_rate": 2.5461126800118578e-05, "loss": 0.1227, "step": 49485 }, { "epoch": 2.31, "learning_rate": 2.5460658730282558e-05, "loss": 0.1106, "step": 49490 }, { "epoch": 2.31, "learning_rate": 2.5460190660446537e-05, "loss": 0.0785, "step": 49495 }, { "epoch": 2.31, "learning_rate": 2.545972259061052e-05, "loss": 0.0876, "step": 49500 }, { "epoch": 2.31, "learning_rate": 2.54592545207745e-05, "loss": 0.2105, "step": 49505 }, { "epoch": 2.31, "learning_rate": 2.545878645093848e-05, "loss": 0.1714, "step": 49510 }, { "epoch": 2.31, "learning_rate": 2.545831838110246e-05, "loss": 0.4019, "step": 49515 }, { "epoch": 2.31, "learning_rate": 2.5457850311266443e-05, "loss": 0.0654, "step": 49520 }, { "epoch": 2.31, "learning_rate": 2.5457382241430423e-05, "loss": 0.0716, "step": 49525 }, { "epoch": 2.31, "learning_rate": 2.5456914171594403e-05, "loss": 0.06, "step": 49530 }, { "epoch": 2.31, "learning_rate": 2.5456446101758383e-05, "loss": 0.052, "step": 49535 }, { "epoch": 2.31, "learning_rate": 2.5455978031922366e-05, "loss": 0.0881, "step": 49540 }, { "epoch": 2.31, "learning_rate": 2.5455509962086346e-05, "loss": 0.0607, "step": 49545 }, { "epoch": 2.31, "learning_rate": 2.5455041892250322e-05, "loss": 0.0982, "step": 49550 }, { "epoch": 2.31, "learning_rate": 2.5454573822414302e-05, "loss": 0.1946, "step": 49555 }, { "epoch": 2.31, "learning_rate": 2.5454105752578285e-05, "loss": 0.2121, "step": 49560 }, { "epoch": 2.31, "learning_rate": 2.5453637682742265e-05, "loss": 0.3449, "step": 49565 }, { "epoch": 2.31, "learning_rate": 2.5453169612906245e-05, "loss": 0.1156, "step": 49570 }, { "epoch": 2.31, "learning_rate": 2.5452701543070228e-05, "loss": 0.047, "step": 49575 }, { "epoch": 2.31, "learning_rate": 2.5452233473234208e-05, "loss": 0.06, "step": 49580 }, { "epoch": 2.31, "learning_rate": 2.5451765403398188e-05, "loss": 0.0634, "step": 49585 }, { "epoch": 2.31, "learning_rate": 2.5451297333562167e-05, "loss": 0.0328, "step": 49590 }, { "epoch": 2.31, "learning_rate": 2.545082926372615e-05, "loss": 0.0912, "step": 49595 }, { "epoch": 2.31, "learning_rate": 2.545036119389013e-05, "loss": 0.1614, "step": 49600 }, { "epoch": 2.31, "learning_rate": 2.544989312405411e-05, "loss": 0.1363, "step": 49605 }, { "epoch": 2.31, "learning_rate": 2.544942505421809e-05, "loss": 0.1532, "step": 49610 }, { "epoch": 2.32, "learning_rate": 2.544895698438207e-05, "loss": 0.3348, "step": 49615 }, { "epoch": 2.32, "learning_rate": 2.544848891454605e-05, "loss": 0.0651, "step": 49620 }, { "epoch": 2.32, "learning_rate": 2.544802084471003e-05, "loss": 0.0531, "step": 49625 }, { "epoch": 2.32, "learning_rate": 2.5447552774874013e-05, "loss": 0.0337, "step": 49630 }, { "epoch": 2.32, "learning_rate": 2.5447084705037993e-05, "loss": 0.1521, "step": 49635 }, { "epoch": 2.32, "learning_rate": 2.5446616635201972e-05, "loss": 0.1037, "step": 49640 }, { "epoch": 2.32, "learning_rate": 2.5446148565365952e-05, "loss": 0.0829, "step": 49645 }, { "epoch": 2.32, "learning_rate": 2.5445680495529935e-05, "loss": 0.1731, "step": 49650 }, { "epoch": 2.32, "learning_rate": 2.5445212425693915e-05, "loss": 0.1272, "step": 49655 }, { "epoch": 2.32, "learning_rate": 2.5444744355857895e-05, "loss": 0.3034, "step": 49660 }, { "epoch": 2.32, "learning_rate": 2.5444276286021875e-05, "loss": 0.2454, "step": 49665 }, { "epoch": 2.32, "learning_rate": 2.5443808216185858e-05, "loss": 0.0299, "step": 49670 }, { "epoch": 2.32, "learning_rate": 2.5443340146349834e-05, "loss": 0.0292, "step": 49675 }, { "epoch": 2.32, "learning_rate": 2.5442872076513814e-05, "loss": 0.0221, "step": 49680 }, { "epoch": 2.32, "learning_rate": 2.5442404006677798e-05, "loss": 0.0613, "step": 49685 }, { "epoch": 2.32, "learning_rate": 2.5441935936841777e-05, "loss": 0.1138, "step": 49690 }, { "epoch": 2.32, "learning_rate": 2.5441467867005757e-05, "loss": 0.1598, "step": 49695 }, { "epoch": 2.32, "learning_rate": 2.5440999797169737e-05, "loss": 0.1373, "step": 49700 }, { "epoch": 2.32, "learning_rate": 2.544053172733372e-05, "loss": 0.1314, "step": 49705 }, { "epoch": 2.32, "learning_rate": 2.54400636574977e-05, "loss": 0.1474, "step": 49710 }, { "epoch": 2.32, "learning_rate": 2.543959558766168e-05, "loss": 0.2503, "step": 49715 }, { "epoch": 2.32, "learning_rate": 2.543912751782566e-05, "loss": 0.0241, "step": 49720 }, { "epoch": 2.32, "learning_rate": 2.5438659447989643e-05, "loss": 0.0413, "step": 49725 }, { "epoch": 2.32, "learning_rate": 2.5438191378153623e-05, "loss": 0.0351, "step": 49730 }, { "epoch": 2.32, "learning_rate": 2.5437723308317602e-05, "loss": 0.0802, "step": 49735 }, { "epoch": 2.32, "learning_rate": 2.5437255238481582e-05, "loss": 0.115, "step": 49740 }, { "epoch": 2.32, "learning_rate": 2.5436787168645562e-05, "loss": 0.0864, "step": 49745 }, { "epoch": 2.32, "learning_rate": 2.5436319098809542e-05, "loss": 0.0862, "step": 49750 }, { "epoch": 2.32, "learning_rate": 2.5435851028973522e-05, "loss": 0.1779, "step": 49755 }, { "epoch": 2.32, "learning_rate": 2.5435382959137505e-05, "loss": 0.2722, "step": 49760 }, { "epoch": 2.32, "learning_rate": 2.5434914889301485e-05, "loss": 0.2308, "step": 49765 }, { "epoch": 2.32, "learning_rate": 2.5434446819465465e-05, "loss": 0.0952, "step": 49770 }, { "epoch": 2.32, "learning_rate": 2.5433978749629444e-05, "loss": 0.0569, "step": 49775 }, { "epoch": 2.32, "learning_rate": 2.5433510679793428e-05, "loss": 0.0189, "step": 49780 }, { "epoch": 2.32, "learning_rate": 2.5433042609957407e-05, "loss": 0.0628, "step": 49785 }, { "epoch": 2.32, "learning_rate": 2.5432574540121387e-05, "loss": 0.0487, "step": 49790 }, { "epoch": 2.32, "learning_rate": 2.5432106470285367e-05, "loss": 0.0649, "step": 49795 }, { "epoch": 2.32, "learning_rate": 2.5431638400449347e-05, "loss": 0.1791, "step": 49800 }, { "epoch": 2.32, "learning_rate": 2.5431170330613327e-05, "loss": 0.1248, "step": 49805 }, { "epoch": 2.32, "learning_rate": 2.5430702260777307e-05, "loss": 0.1678, "step": 49810 }, { "epoch": 2.32, "learning_rate": 2.543023419094129e-05, "loss": 0.2428, "step": 49815 }, { "epoch": 2.32, "learning_rate": 2.542976612110527e-05, "loss": 0.0325, "step": 49820 }, { "epoch": 2.32, "learning_rate": 2.542929805126925e-05, "loss": 0.0578, "step": 49825 }, { "epoch": 2.33, "learning_rate": 2.542882998143323e-05, "loss": 0.0784, "step": 49830 }, { "epoch": 2.33, "learning_rate": 2.5428361911597212e-05, "loss": 0.0647, "step": 49835 }, { "epoch": 2.33, "learning_rate": 2.5427893841761192e-05, "loss": 0.1084, "step": 49840 }, { "epoch": 2.33, "learning_rate": 2.5427425771925172e-05, "loss": 0.1054, "step": 49845 }, { "epoch": 2.33, "learning_rate": 2.5426957702089152e-05, "loss": 0.1604, "step": 49850 }, { "epoch": 2.33, "learning_rate": 2.5426489632253135e-05, "loss": 0.1043, "step": 49855 }, { "epoch": 2.33, "learning_rate": 2.5426021562417115e-05, "loss": 0.2644, "step": 49860 }, { "epoch": 2.33, "learning_rate": 2.542555349258109e-05, "loss": 0.2072, "step": 49865 }, { "epoch": 2.33, "learning_rate": 2.5425085422745074e-05, "loss": 0.0799, "step": 49870 }, { "epoch": 2.33, "learning_rate": 2.5424617352909054e-05, "loss": 0.1019, "step": 49875 }, { "epoch": 2.33, "learning_rate": 2.5424149283073034e-05, "loss": 0.0481, "step": 49880 }, { "epoch": 2.33, "learning_rate": 2.5423681213237014e-05, "loss": 0.0452, "step": 49885 }, { "epoch": 2.33, "learning_rate": 2.5423213143400997e-05, "loss": 0.0779, "step": 49890 }, { "epoch": 2.33, "learning_rate": 2.5422745073564977e-05, "loss": 0.0588, "step": 49895 }, { "epoch": 2.33, "learning_rate": 2.5422277003728957e-05, "loss": 0.1759, "step": 49900 }, { "epoch": 2.33, "learning_rate": 2.5421808933892937e-05, "loss": 0.1676, "step": 49905 }, { "epoch": 2.33, "learning_rate": 2.542134086405692e-05, "loss": 0.2049, "step": 49910 }, { "epoch": 2.33, "learning_rate": 2.54208727942209e-05, "loss": 0.2557, "step": 49915 }, { "epoch": 2.33, "learning_rate": 2.542040472438488e-05, "loss": 0.0453, "step": 49920 }, { "epoch": 2.33, "learning_rate": 2.5419936654548863e-05, "loss": 0.0367, "step": 49925 }, { "epoch": 2.33, "learning_rate": 2.541946858471284e-05, "loss": 0.0789, "step": 49930 }, { "epoch": 2.33, "learning_rate": 2.541900051487682e-05, "loss": 0.0582, "step": 49935 }, { "epoch": 2.33, "learning_rate": 2.54185324450408e-05, "loss": 0.168, "step": 49940 }, { "epoch": 2.33, "learning_rate": 2.5418064375204782e-05, "loss": 0.0575, "step": 49945 }, { "epoch": 2.33, "learning_rate": 2.5417596305368762e-05, "loss": 0.1285, "step": 49950 }, { "epoch": 2.33, "learning_rate": 2.541712823553274e-05, "loss": 0.1581, "step": 49955 }, { "epoch": 2.33, "learning_rate": 2.541666016569672e-05, "loss": 0.1517, "step": 49960 }, { "epoch": 2.33, "learning_rate": 2.5416192095860705e-05, "loss": 0.3418, "step": 49965 }, { "epoch": 2.33, "learning_rate": 2.5415724026024684e-05, "loss": 0.0279, "step": 49970 }, { "epoch": 2.33, "learning_rate": 2.5415255956188664e-05, "loss": 0.0606, "step": 49975 }, { "epoch": 2.33, "learning_rate": 2.5414787886352644e-05, "loss": 0.0848, "step": 49980 }, { "epoch": 2.33, "learning_rate": 2.5414319816516627e-05, "loss": 0.0611, "step": 49985 }, { "epoch": 2.33, "learning_rate": 2.5413851746680604e-05, "loss": 0.078, "step": 49990 }, { "epoch": 2.33, "learning_rate": 2.5413383676844583e-05, "loss": 0.1221, "step": 49995 }, { "epoch": 2.33, "learning_rate": 2.5412915607008567e-05, "loss": 0.3131, "step": 50000 }, { "epoch": 2.33, "learning_rate": 2.5412447537172546e-05, "loss": 0.1315, "step": 50005 }, { "epoch": 2.33, "learning_rate": 2.5411979467336526e-05, "loss": 0.1607, "step": 50010 }, { "epoch": 2.33, "learning_rate": 2.5411511397500506e-05, "loss": 0.2701, "step": 50015 }, { "epoch": 2.33, "learning_rate": 2.541104332766449e-05, "loss": 0.0967, "step": 50020 }, { "epoch": 2.33, "learning_rate": 2.541057525782847e-05, "loss": 0.0494, "step": 50025 }, { "epoch": 2.33, "learning_rate": 2.541010718799245e-05, "loss": 0.1049, "step": 50030 }, { "epoch": 2.33, "learning_rate": 2.540963911815643e-05, "loss": 0.0495, "step": 50035 }, { "epoch": 2.33, "learning_rate": 2.5409171048320412e-05, "loss": 0.1027, "step": 50040 }, { "epoch": 2.34, "learning_rate": 2.5408702978484392e-05, "loss": 0.0669, "step": 50045 }, { "epoch": 2.34, "learning_rate": 2.540823490864837e-05, "loss": 0.1393, "step": 50050 }, { "epoch": 2.34, "learning_rate": 2.540776683881235e-05, "loss": 0.2202, "step": 50055 }, { "epoch": 2.34, "learning_rate": 2.540729876897633e-05, "loss": 0.2238, "step": 50060 }, { "epoch": 2.34, "learning_rate": 2.540683069914031e-05, "loss": 0.182, "step": 50065 }, { "epoch": 2.34, "learning_rate": 2.540636262930429e-05, "loss": 0.0626, "step": 50070 }, { "epoch": 2.34, "learning_rate": 2.5405894559468274e-05, "loss": 0.0237, "step": 50075 }, { "epoch": 2.34, "learning_rate": 2.5405426489632254e-05, "loss": 0.066, "step": 50080 }, { "epoch": 2.34, "learning_rate": 2.5404958419796234e-05, "loss": 0.0602, "step": 50085 }, { "epoch": 2.34, "learning_rate": 2.5404490349960214e-05, "loss": 0.0647, "step": 50090 }, { "epoch": 2.34, "learning_rate": 2.5404022280124197e-05, "loss": 0.0931, "step": 50095 }, { "epoch": 2.34, "learning_rate": 2.5403554210288177e-05, "loss": 0.1107, "step": 50100 }, { "epoch": 2.34, "learning_rate": 2.5403086140452156e-05, "loss": 0.1569, "step": 50105 }, { "epoch": 2.34, "learning_rate": 2.540261807061614e-05, "loss": 0.1128, "step": 50110 }, { "epoch": 2.34, "learning_rate": 2.540215000078012e-05, "loss": 0.3365, "step": 50115 }, { "epoch": 2.34, "learning_rate": 2.5401681930944096e-05, "loss": 0.0456, "step": 50120 }, { "epoch": 2.34, "learning_rate": 2.5401213861108076e-05, "loss": 0.0748, "step": 50125 }, { "epoch": 2.34, "learning_rate": 2.540074579127206e-05, "loss": 0.0557, "step": 50130 }, { "epoch": 2.34, "learning_rate": 2.540027772143604e-05, "loss": 0.0999, "step": 50135 }, { "epoch": 2.34, "learning_rate": 2.539980965160002e-05, "loss": 0.0658, "step": 50140 }, { "epoch": 2.34, "learning_rate": 2.5399341581764e-05, "loss": 0.1675, "step": 50145 }, { "epoch": 2.34, "learning_rate": 2.539887351192798e-05, "loss": 0.2124, "step": 50150 }, { "epoch": 2.34, "learning_rate": 2.539840544209196e-05, "loss": 0.1499, "step": 50155 }, { "epoch": 2.34, "learning_rate": 2.539793737225594e-05, "loss": 0.3174, "step": 50160 }, { "epoch": 2.34, "learning_rate": 2.539746930241992e-05, "loss": 0.2783, "step": 50165 }, { "epoch": 2.34, "learning_rate": 2.5397001232583904e-05, "loss": 0.0268, "step": 50170 }, { "epoch": 2.34, "learning_rate": 2.5396533162747884e-05, "loss": 0.0272, "step": 50175 }, { "epoch": 2.34, "learning_rate": 2.539606509291186e-05, "loss": 0.0436, "step": 50180 }, { "epoch": 2.34, "learning_rate": 2.5395597023075844e-05, "loss": 0.0458, "step": 50185 }, { "epoch": 2.34, "learning_rate": 2.5395128953239823e-05, "loss": 0.1333, "step": 50190 }, { "epoch": 2.34, "learning_rate": 2.5394660883403803e-05, "loss": 0.068, "step": 50195 }, { "epoch": 2.34, "learning_rate": 2.5394192813567783e-05, "loss": 0.0794, "step": 50200 }, { "epoch": 2.34, "learning_rate": 2.5393724743731766e-05, "loss": 0.1917, "step": 50205 }, { "epoch": 2.34, "learning_rate": 2.5393256673895746e-05, "loss": 0.4107, "step": 50210 }, { "epoch": 2.34, "learning_rate": 2.5392788604059726e-05, "loss": 0.2252, "step": 50215 }, { "epoch": 2.34, "learning_rate": 2.5392320534223706e-05, "loss": 0.0406, "step": 50220 }, { "epoch": 2.34, "learning_rate": 2.539185246438769e-05, "loss": 0.0432, "step": 50225 }, { "epoch": 2.34, "learning_rate": 2.539138439455167e-05, "loss": 0.0594, "step": 50230 }, { "epoch": 2.34, "learning_rate": 2.539091632471565e-05, "loss": 0.1008, "step": 50235 }, { "epoch": 2.34, "learning_rate": 2.5390448254879632e-05, "loss": 0.0651, "step": 50240 }, { "epoch": 2.34, "learning_rate": 2.5389980185043608e-05, "loss": 0.1193, "step": 50245 }, { "epoch": 2.34, "learning_rate": 2.5389512115207588e-05, "loss": 0.1278, "step": 50250 }, { "epoch": 2.34, "learning_rate": 2.5389044045371568e-05, "loss": 0.1976, "step": 50255 }, { "epoch": 2.35, "learning_rate": 2.538857597553555e-05, "loss": 0.1723, "step": 50260 }, { "epoch": 2.35, "learning_rate": 2.538810790569953e-05, "loss": 0.2009, "step": 50265 }, { "epoch": 2.35, "learning_rate": 2.538763983586351e-05, "loss": 0.0442, "step": 50270 }, { "epoch": 2.35, "learning_rate": 2.538717176602749e-05, "loss": 0.0261, "step": 50275 }, { "epoch": 2.35, "learning_rate": 2.5386703696191474e-05, "loss": 0.0849, "step": 50280 }, { "epoch": 2.35, "learning_rate": 2.5386235626355454e-05, "loss": 0.0594, "step": 50285 }, { "epoch": 2.35, "learning_rate": 2.5385767556519433e-05, "loss": 0.0961, "step": 50290 }, { "epoch": 2.35, "learning_rate": 2.5385299486683417e-05, "loss": 0.0355, "step": 50295 }, { "epoch": 2.35, "learning_rate": 2.5384831416847396e-05, "loss": 0.089, "step": 50300 }, { "epoch": 2.35, "learning_rate": 2.5384363347011373e-05, "loss": 0.1026, "step": 50305 }, { "epoch": 2.35, "learning_rate": 2.5383895277175353e-05, "loss": 0.1942, "step": 50310 }, { "epoch": 2.35, "learning_rate": 2.5383427207339336e-05, "loss": 0.3554, "step": 50315 }, { "epoch": 2.35, "learning_rate": 2.5382959137503316e-05, "loss": 0.0324, "step": 50320 }, { "epoch": 2.35, "learning_rate": 2.5382491067667295e-05, "loss": 0.0389, "step": 50325 }, { "epoch": 2.35, "learning_rate": 2.5382022997831275e-05, "loss": 0.0259, "step": 50330 }, { "epoch": 2.35, "learning_rate": 2.538155492799526e-05, "loss": 0.0538, "step": 50335 }, { "epoch": 2.35, "learning_rate": 2.538108685815924e-05, "loss": 0.0534, "step": 50340 }, { "epoch": 2.35, "learning_rate": 2.5380618788323218e-05, "loss": 0.1144, "step": 50345 }, { "epoch": 2.35, "learning_rate": 2.53801507184872e-05, "loss": 0.175, "step": 50350 }, { "epoch": 2.35, "learning_rate": 2.537968264865118e-05, "loss": 0.1183, "step": 50355 }, { "epoch": 2.35, "learning_rate": 2.537921457881516e-05, "loss": 0.2255, "step": 50360 }, { "epoch": 2.35, "learning_rate": 2.537874650897914e-05, "loss": 0.1964, "step": 50365 }, { "epoch": 2.35, "learning_rate": 2.537827843914312e-05, "loss": 0.052, "step": 50370 }, { "epoch": 2.35, "learning_rate": 2.53778103693071e-05, "loss": 0.0467, "step": 50375 }, { "epoch": 2.35, "learning_rate": 2.537734229947108e-05, "loss": 0.0439, "step": 50380 }, { "epoch": 2.35, "learning_rate": 2.537687422963506e-05, "loss": 0.0687, "step": 50385 }, { "epoch": 2.35, "learning_rate": 2.5376406159799043e-05, "loss": 0.1015, "step": 50390 }, { "epoch": 2.35, "learning_rate": 2.5375938089963023e-05, "loss": 0.0946, "step": 50395 }, { "epoch": 2.35, "learning_rate": 2.5375470020127003e-05, "loss": 0.1287, "step": 50400 }, { "epoch": 2.35, "learning_rate": 2.5375001950290983e-05, "loss": 0.1982, "step": 50405 }, { "epoch": 2.35, "learning_rate": 2.5374533880454966e-05, "loss": 0.2237, "step": 50410 }, { "epoch": 2.35, "learning_rate": 2.5374065810618946e-05, "loss": 0.2411, "step": 50415 }, { "epoch": 2.35, "learning_rate": 2.5373597740782926e-05, "loss": 0.0149, "step": 50420 }, { "epoch": 2.35, "learning_rate": 2.537312967094691e-05, "loss": 0.0509, "step": 50425 }, { "epoch": 2.35, "learning_rate": 2.537266160111089e-05, "loss": 0.0873, "step": 50430 }, { "epoch": 2.35, "learning_rate": 2.5372193531274865e-05, "loss": 0.092, "step": 50435 }, { "epoch": 2.35, "learning_rate": 2.5371725461438845e-05, "loss": 0.0939, "step": 50440 }, { "epoch": 2.35, "learning_rate": 2.5371257391602828e-05, "loss": 0.1211, "step": 50445 }, { "epoch": 2.35, "learning_rate": 2.5370789321766808e-05, "loss": 0.191, "step": 50450 }, { "epoch": 2.35, "learning_rate": 2.5370321251930788e-05, "loss": 0.1858, "step": 50455 }, { "epoch": 2.35, "learning_rate": 2.5369853182094768e-05, "loss": 0.3049, "step": 50460 }, { "epoch": 2.35, "learning_rate": 2.536938511225875e-05, "loss": 0.2958, "step": 50465 }, { "epoch": 2.35, "learning_rate": 2.536891704242273e-05, "loss": 0.0345, "step": 50470 }, { "epoch": 2.36, "learning_rate": 2.536844897258671e-05, "loss": 0.0568, "step": 50475 }, { "epoch": 2.36, "learning_rate": 2.5367980902750694e-05, "loss": 0.0802, "step": 50480 }, { "epoch": 2.36, "learning_rate": 2.5367512832914673e-05, "loss": 0.0865, "step": 50485 }, { "epoch": 2.36, "learning_rate": 2.5367044763078653e-05, "loss": 0.0524, "step": 50490 }, { "epoch": 2.36, "learning_rate": 2.536657669324263e-05, "loss": 0.0844, "step": 50495 }, { "epoch": 2.36, "learning_rate": 2.5366108623406613e-05, "loss": 0.0929, "step": 50500 }, { "epoch": 2.36, "learning_rate": 2.5365640553570593e-05, "loss": 0.1287, "step": 50505 }, { "epoch": 2.36, "learning_rate": 2.5365172483734572e-05, "loss": 0.1441, "step": 50510 }, { "epoch": 2.36, "learning_rate": 2.5364704413898552e-05, "loss": 0.2264, "step": 50515 }, { "epoch": 2.36, "learning_rate": 2.5364236344062535e-05, "loss": 0.0316, "step": 50520 }, { "epoch": 2.36, "learning_rate": 2.5363768274226515e-05, "loss": 0.0592, "step": 50525 }, { "epoch": 2.36, "learning_rate": 2.5363300204390495e-05, "loss": 0.0379, "step": 50530 }, { "epoch": 2.36, "learning_rate": 2.536283213455448e-05, "loss": 0.0985, "step": 50535 }, { "epoch": 2.36, "learning_rate": 2.5362364064718458e-05, "loss": 0.1048, "step": 50540 }, { "epoch": 2.36, "learning_rate": 2.5361895994882438e-05, "loss": 0.1704, "step": 50545 }, { "epoch": 2.36, "learning_rate": 2.5361427925046418e-05, "loss": 0.1961, "step": 50550 }, { "epoch": 2.36, "learning_rate": 2.53609598552104e-05, "loss": 0.1636, "step": 50555 }, { "epoch": 2.36, "learning_rate": 2.5360491785374377e-05, "loss": 0.3128, "step": 50560 }, { "epoch": 2.36, "learning_rate": 2.5360023715538357e-05, "loss": 0.2527, "step": 50565 }, { "epoch": 2.36, "learning_rate": 2.5359555645702337e-05, "loss": 0.0768, "step": 50570 }, { "epoch": 2.36, "learning_rate": 2.535908757586632e-05, "loss": 0.0512, "step": 50575 }, { "epoch": 2.36, "learning_rate": 2.53586195060303e-05, "loss": 0.0392, "step": 50580 }, { "epoch": 2.36, "learning_rate": 2.535815143619428e-05, "loss": 0.0605, "step": 50585 }, { "epoch": 2.36, "learning_rate": 2.535768336635826e-05, "loss": 0.1337, "step": 50590 }, { "epoch": 2.36, "learning_rate": 2.5357215296522243e-05, "loss": 0.1132, "step": 50595 }, { "epoch": 2.36, "learning_rate": 2.5356747226686223e-05, "loss": 0.0603, "step": 50600 }, { "epoch": 2.36, "learning_rate": 2.5356279156850203e-05, "loss": 0.1236, "step": 50605 }, { "epoch": 2.36, "learning_rate": 2.5355811087014186e-05, "loss": 0.1894, "step": 50610 }, { "epoch": 2.36, "learning_rate": 2.5355343017178166e-05, "loss": 0.31, "step": 50615 }, { "epoch": 2.36, "learning_rate": 2.5354874947342145e-05, "loss": 0.0427, "step": 50620 }, { "epoch": 2.36, "learning_rate": 2.5354406877506122e-05, "loss": 0.0668, "step": 50625 }, { "epoch": 2.36, "learning_rate": 2.5353938807670105e-05, "loss": 0.0399, "step": 50630 }, { "epoch": 2.36, "learning_rate": 2.5353470737834085e-05, "loss": 0.016, "step": 50635 }, { "epoch": 2.36, "learning_rate": 2.5353002667998065e-05, "loss": 0.1485, "step": 50640 }, { "epoch": 2.36, "learning_rate": 2.5352534598162044e-05, "loss": 0.0868, "step": 50645 }, { "epoch": 2.36, "learning_rate": 2.5352066528326028e-05, "loss": 0.065, "step": 50650 }, { "epoch": 2.36, "learning_rate": 2.5351598458490007e-05, "loss": 0.1806, "step": 50655 }, { "epoch": 2.36, "learning_rate": 2.5351130388653987e-05, "loss": 0.1838, "step": 50660 }, { "epoch": 2.36, "learning_rate": 2.535066231881797e-05, "loss": 0.4168, "step": 50665 }, { "epoch": 2.36, "learning_rate": 2.535019424898195e-05, "loss": 0.0335, "step": 50670 }, { "epoch": 2.36, "learning_rate": 2.534972617914593e-05, "loss": 0.0152, "step": 50675 }, { "epoch": 2.36, "learning_rate": 2.534925810930991e-05, "loss": 0.0493, "step": 50680 }, { "epoch": 2.37, "learning_rate": 2.534879003947389e-05, "loss": 0.0674, "step": 50685 }, { "epoch": 2.37, "learning_rate": 2.534832196963787e-05, "loss": 0.0883, "step": 50690 }, { "epoch": 2.37, "learning_rate": 2.534785389980185e-05, "loss": 0.1207, "step": 50695 }, { "epoch": 2.37, "learning_rate": 2.534738582996583e-05, "loss": 0.1926, "step": 50700 }, { "epoch": 2.37, "learning_rate": 2.5346917760129812e-05, "loss": 0.1596, "step": 50705 }, { "epoch": 2.37, "learning_rate": 2.5346449690293792e-05, "loss": 0.1482, "step": 50710 }, { "epoch": 2.37, "learning_rate": 2.5345981620457772e-05, "loss": 0.3368, "step": 50715 }, { "epoch": 2.37, "learning_rate": 2.5345513550621755e-05, "loss": 0.0387, "step": 50720 }, { "epoch": 2.37, "learning_rate": 2.5345045480785735e-05, "loss": 0.0467, "step": 50725 }, { "epoch": 2.37, "learning_rate": 2.5344577410949715e-05, "loss": 0.0813, "step": 50730 }, { "epoch": 2.37, "learning_rate": 2.5344109341113695e-05, "loss": 0.0819, "step": 50735 }, { "epoch": 2.37, "learning_rate": 2.5343641271277678e-05, "loss": 0.1093, "step": 50740 }, { "epoch": 2.37, "learning_rate": 2.5343173201441658e-05, "loss": 0.097, "step": 50745 }, { "epoch": 2.37, "learning_rate": 2.5342705131605634e-05, "loss": 0.302, "step": 50750 }, { "epoch": 2.37, "learning_rate": 2.5342237061769614e-05, "loss": 0.1494, "step": 50755 }, { "epoch": 2.37, "learning_rate": 2.5341768991933597e-05, "loss": 0.1882, "step": 50760 }, { "epoch": 2.37, "learning_rate": 2.5341300922097577e-05, "loss": 0.2207, "step": 50765 }, { "epoch": 2.37, "learning_rate": 2.5340832852261557e-05, "loss": 0.0713, "step": 50770 }, { "epoch": 2.37, "learning_rate": 2.5340364782425537e-05, "loss": 0.0471, "step": 50775 }, { "epoch": 2.37, "learning_rate": 2.533989671258952e-05, "loss": 0.0565, "step": 50780 }, { "epoch": 2.37, "learning_rate": 2.53394286427535e-05, "loss": 0.0745, "step": 50785 }, { "epoch": 2.37, "learning_rate": 2.533896057291748e-05, "loss": 0.1431, "step": 50790 }, { "epoch": 2.37, "learning_rate": 2.5338492503081463e-05, "loss": 0.177, "step": 50795 }, { "epoch": 2.37, "learning_rate": 2.5338024433245443e-05, "loss": 0.2157, "step": 50800 }, { "epoch": 2.37, "learning_rate": 2.5337556363409422e-05, "loss": 0.1742, "step": 50805 }, { "epoch": 2.37, "learning_rate": 2.5337088293573402e-05, "loss": 0.3254, "step": 50810 }, { "epoch": 2.37, "learning_rate": 2.5336620223737382e-05, "loss": 0.3474, "step": 50815 }, { "epoch": 2.37, "learning_rate": 2.5336152153901362e-05, "loss": 0.103, "step": 50820 }, { "epoch": 2.37, "learning_rate": 2.533568408406534e-05, "loss": 0.082, "step": 50825 }, { "epoch": 2.37, "learning_rate": 2.533521601422932e-05, "loss": 0.0972, "step": 50830 }, { "epoch": 2.37, "learning_rate": 2.5334747944393305e-05, "loss": 0.0583, "step": 50835 }, { "epoch": 2.37, "learning_rate": 2.5334279874557284e-05, "loss": 0.1032, "step": 50840 }, { "epoch": 2.37, "learning_rate": 2.5333811804721264e-05, "loss": 0.1137, "step": 50845 }, { "epoch": 2.37, "learning_rate": 2.5333343734885247e-05, "loss": 0.1596, "step": 50850 }, { "epoch": 2.37, "learning_rate": 2.5332875665049227e-05, "loss": 0.2884, "step": 50855 }, { "epoch": 2.37, "learning_rate": 2.5332407595213207e-05, "loss": 0.249, "step": 50860 }, { "epoch": 2.37, "learning_rate": 2.5331939525377187e-05, "loss": 0.3164, "step": 50865 }, { "epoch": 2.37, "learning_rate": 2.533147145554117e-05, "loss": 0.0429, "step": 50870 }, { "epoch": 2.37, "learning_rate": 2.5331003385705147e-05, "loss": 0.0177, "step": 50875 }, { "epoch": 2.37, "learning_rate": 2.5330535315869126e-05, "loss": 0.0341, "step": 50880 }, { "epoch": 2.37, "learning_rate": 2.5330067246033106e-05, "loss": 0.0251, "step": 50885 }, { "epoch": 2.37, "learning_rate": 2.532959917619709e-05, "loss": 0.1125, "step": 50890 }, { "epoch": 2.37, "learning_rate": 2.532913110636107e-05, "loss": 0.0925, "step": 50895 }, { "epoch": 2.38, "learning_rate": 2.532866303652505e-05, "loss": 0.1246, "step": 50900 }, { "epoch": 2.38, "learning_rate": 2.5328194966689032e-05, "loss": 0.2222, "step": 50905 }, { "epoch": 2.38, "learning_rate": 2.5327726896853012e-05, "loss": 0.3999, "step": 50910 }, { "epoch": 2.38, "learning_rate": 2.5327258827016992e-05, "loss": 0.2956, "step": 50915 }, { "epoch": 2.38, "learning_rate": 2.5326790757180972e-05, "loss": 0.0303, "step": 50920 }, { "epoch": 2.38, "learning_rate": 2.5326322687344955e-05, "loss": 0.0555, "step": 50925 }, { "epoch": 2.38, "learning_rate": 2.5325854617508935e-05, "loss": 0.0436, "step": 50930 }, { "epoch": 2.38, "learning_rate": 2.5325386547672915e-05, "loss": 0.0464, "step": 50935 }, { "epoch": 2.38, "learning_rate": 2.532491847783689e-05, "loss": 0.1225, "step": 50940 }, { "epoch": 2.38, "learning_rate": 2.5324450408000874e-05, "loss": 0.0913, "step": 50945 }, { "epoch": 2.38, "learning_rate": 2.5323982338164854e-05, "loss": 0.1366, "step": 50950 }, { "epoch": 2.38, "learning_rate": 2.5323514268328834e-05, "loss": 0.1804, "step": 50955 }, { "epoch": 2.38, "learning_rate": 2.5323046198492814e-05, "loss": 0.1268, "step": 50960 }, { "epoch": 2.38, "learning_rate": 2.5322578128656797e-05, "loss": 0.1835, "step": 50965 }, { "epoch": 2.38, "learning_rate": 2.5322110058820777e-05, "loss": 0.0251, "step": 50970 }, { "epoch": 2.38, "learning_rate": 2.5321641988984756e-05, "loss": 0.0554, "step": 50975 }, { "epoch": 2.38, "learning_rate": 2.532117391914874e-05, "loss": 0.0826, "step": 50980 }, { "epoch": 2.38, "learning_rate": 2.532070584931272e-05, "loss": 0.1398, "step": 50985 }, { "epoch": 2.38, "learning_rate": 2.53202377794767e-05, "loss": 0.0734, "step": 50990 }, { "epoch": 2.38, "learning_rate": 2.531976970964068e-05, "loss": 0.0862, "step": 50995 }, { "epoch": 2.38, "learning_rate": 2.531930163980466e-05, "loss": 0.1195, "step": 51000 }, { "epoch": 2.38, "learning_rate": 2.531883356996864e-05, "loss": 0.2509, "step": 51005 }, { "epoch": 2.38, "learning_rate": 2.531836550013262e-05, "loss": 0.1815, "step": 51010 }, { "epoch": 2.38, "learning_rate": 2.53178974302966e-05, "loss": 0.1515, "step": 51015 }, { "epoch": 2.38, "learning_rate": 2.531742936046058e-05, "loss": 0.0767, "step": 51020 }, { "epoch": 2.38, "learning_rate": 2.531696129062456e-05, "loss": 0.0138, "step": 51025 }, { "epoch": 2.38, "learning_rate": 2.531649322078854e-05, "loss": 0.0405, "step": 51030 }, { "epoch": 2.38, "learning_rate": 2.5316025150952524e-05, "loss": 0.0609, "step": 51035 }, { "epoch": 2.38, "learning_rate": 2.5315557081116504e-05, "loss": 0.1235, "step": 51040 }, { "epoch": 2.38, "learning_rate": 2.5315089011280484e-05, "loss": 0.114, "step": 51045 }, { "epoch": 2.38, "learning_rate": 2.5314620941444464e-05, "loss": 0.111, "step": 51050 }, { "epoch": 2.38, "learning_rate": 2.5314152871608447e-05, "loss": 0.1615, "step": 51055 }, { "epoch": 2.38, "learning_rate": 2.5313684801772427e-05, "loss": 0.3308, "step": 51060 }, { "epoch": 2.38, "learning_rate": 2.5313216731936403e-05, "loss": 0.3028, "step": 51065 }, { "epoch": 2.38, "learning_rate": 2.5312748662100383e-05, "loss": 0.0886, "step": 51070 }, { "epoch": 2.38, "learning_rate": 2.5312280592264366e-05, "loss": 0.0574, "step": 51075 }, { "epoch": 2.38, "learning_rate": 2.5311812522428346e-05, "loss": 0.0315, "step": 51080 }, { "epoch": 2.38, "learning_rate": 2.5311344452592326e-05, "loss": 0.086, "step": 51085 }, { "epoch": 2.38, "learning_rate": 2.531087638275631e-05, "loss": 0.0702, "step": 51090 }, { "epoch": 2.38, "learning_rate": 2.531040831292029e-05, "loss": 0.1331, "step": 51095 }, { "epoch": 2.38, "learning_rate": 2.530994024308427e-05, "loss": 0.0748, "step": 51100 }, { "epoch": 2.38, "learning_rate": 2.530947217324825e-05, "loss": 0.1176, "step": 51105 }, { "epoch": 2.38, "learning_rate": 2.5309004103412232e-05, "loss": 0.3016, "step": 51110 }, { "epoch": 2.39, "learning_rate": 2.5308536033576212e-05, "loss": 0.2768, "step": 51115 }, { "epoch": 2.39, "learning_rate": 2.530806796374019e-05, "loss": 0.0692, "step": 51120 }, { "epoch": 2.39, "learning_rate": 2.530759989390417e-05, "loss": 0.053, "step": 51125 }, { "epoch": 2.39, "learning_rate": 2.530713182406815e-05, "loss": 0.0446, "step": 51130 }, { "epoch": 2.39, "learning_rate": 2.530666375423213e-05, "loss": 0.09, "step": 51135 }, { "epoch": 2.39, "learning_rate": 2.530619568439611e-05, "loss": 0.0447, "step": 51140 }, { "epoch": 2.39, "learning_rate": 2.5305727614560094e-05, "loss": 0.0908, "step": 51145 }, { "epoch": 2.39, "learning_rate": 2.5305259544724074e-05, "loss": 0.1345, "step": 51150 }, { "epoch": 2.39, "learning_rate": 2.5304791474888054e-05, "loss": 0.1828, "step": 51155 }, { "epoch": 2.39, "learning_rate": 2.5304323405052033e-05, "loss": 0.3409, "step": 51160 }, { "epoch": 2.39, "learning_rate": 2.5303855335216017e-05, "loss": 0.1768, "step": 51165 }, { "epoch": 2.39, "learning_rate": 2.5303387265379996e-05, "loss": 0.0451, "step": 51170 }, { "epoch": 2.39, "learning_rate": 2.5302919195543976e-05, "loss": 0.0485, "step": 51175 }, { "epoch": 2.39, "learning_rate": 2.5302451125707956e-05, "loss": 0.0633, "step": 51180 }, { "epoch": 2.39, "learning_rate": 2.530198305587194e-05, "loss": 0.0743, "step": 51185 }, { "epoch": 2.39, "learning_rate": 2.5301514986035916e-05, "loss": 0.0756, "step": 51190 }, { "epoch": 2.39, "learning_rate": 2.5301046916199896e-05, "loss": 0.1049, "step": 51195 }, { "epoch": 2.39, "learning_rate": 2.5300578846363875e-05, "loss": 0.1493, "step": 51200 }, { "epoch": 2.39, "learning_rate": 2.530011077652786e-05, "loss": 0.1933, "step": 51205 }, { "epoch": 2.39, "learning_rate": 2.529964270669184e-05, "loss": 0.3533, "step": 51210 }, { "epoch": 2.39, "learning_rate": 2.5299174636855818e-05, "loss": 0.4012, "step": 51215 }, { "epoch": 2.39, "learning_rate": 2.52987065670198e-05, "loss": 0.0426, "step": 51220 }, { "epoch": 2.39, "learning_rate": 2.529823849718378e-05, "loss": 0.0591, "step": 51225 }, { "epoch": 2.39, "learning_rate": 2.529777042734776e-05, "loss": 0.0662, "step": 51230 }, { "epoch": 2.39, "learning_rate": 2.529730235751174e-05, "loss": 0.0435, "step": 51235 }, { "epoch": 2.39, "learning_rate": 2.5296834287675724e-05, "loss": 0.0817, "step": 51240 }, { "epoch": 2.39, "learning_rate": 2.5296366217839704e-05, "loss": 0.0782, "step": 51245 }, { "epoch": 2.39, "learning_rate": 2.5295898148003684e-05, "loss": 0.0707, "step": 51250 }, { "epoch": 2.39, "learning_rate": 2.529543007816766e-05, "loss": 0.156, "step": 51255 }, { "epoch": 2.39, "learning_rate": 2.5294962008331643e-05, "loss": 0.2861, "step": 51260 }, { "epoch": 2.39, "learning_rate": 2.5294493938495623e-05, "loss": 0.19, "step": 51265 }, { "epoch": 2.39, "learning_rate": 2.5294025868659603e-05, "loss": 0.0505, "step": 51270 }, { "epoch": 2.39, "learning_rate": 2.5293557798823586e-05, "loss": 0.0584, "step": 51275 }, { "epoch": 2.39, "learning_rate": 2.5293089728987566e-05, "loss": 0.0735, "step": 51280 }, { "epoch": 2.39, "learning_rate": 2.5292621659151546e-05, "loss": 0.0549, "step": 51285 }, { "epoch": 2.39, "learning_rate": 2.5292153589315526e-05, "loss": 0.1312, "step": 51290 }, { "epoch": 2.39, "learning_rate": 2.529168551947951e-05, "loss": 0.0389, "step": 51295 }, { "epoch": 2.39, "learning_rate": 2.529121744964349e-05, "loss": 0.1786, "step": 51300 }, { "epoch": 2.39, "learning_rate": 2.529074937980747e-05, "loss": 0.1736, "step": 51305 }, { "epoch": 2.39, "learning_rate": 2.529028130997145e-05, "loss": 0.138, "step": 51310 }, { "epoch": 2.39, "learning_rate": 2.528981324013543e-05, "loss": 0.2678, "step": 51315 }, { "epoch": 2.39, "learning_rate": 2.5289345170299408e-05, "loss": 0.0584, "step": 51320 }, { "epoch": 2.39, "learning_rate": 2.5288877100463388e-05, "loss": 0.0534, "step": 51325 }, { "epoch": 2.4, "learning_rate": 2.528840903062737e-05, "loss": 0.0327, "step": 51330 }, { "epoch": 2.4, "learning_rate": 2.528794096079135e-05, "loss": 0.0561, "step": 51335 }, { "epoch": 2.4, "learning_rate": 2.528747289095533e-05, "loss": 0.1063, "step": 51340 }, { "epoch": 2.4, "learning_rate": 2.528700482111931e-05, "loss": 0.1915, "step": 51345 }, { "epoch": 2.4, "learning_rate": 2.5286536751283294e-05, "loss": 0.1177, "step": 51350 }, { "epoch": 2.4, "learning_rate": 2.5286068681447273e-05, "loss": 0.1729, "step": 51355 }, { "epoch": 2.4, "learning_rate": 2.5285600611611253e-05, "loss": 0.1774, "step": 51360 }, { "epoch": 2.4, "learning_rate": 2.5285132541775233e-05, "loss": 0.2218, "step": 51365 }, { "epoch": 2.4, "learning_rate": 2.5284664471939216e-05, "loss": 0.063, "step": 51370 }, { "epoch": 2.4, "learning_rate": 2.5284196402103196e-05, "loss": 0.0544, "step": 51375 }, { "epoch": 2.4, "learning_rate": 2.5283728332267173e-05, "loss": 0.0402, "step": 51380 }, { "epoch": 2.4, "learning_rate": 2.5283260262431152e-05, "loss": 0.0721, "step": 51385 }, { "epoch": 2.4, "learning_rate": 2.5282792192595136e-05, "loss": 0.0583, "step": 51390 }, { "epoch": 2.4, "learning_rate": 2.5282324122759115e-05, "loss": 0.1504, "step": 51395 }, { "epoch": 2.4, "learning_rate": 2.5281856052923095e-05, "loss": 0.1562, "step": 51400 }, { "epoch": 2.4, "learning_rate": 2.528138798308708e-05, "loss": 0.12, "step": 51405 }, { "epoch": 2.4, "learning_rate": 2.5280919913251058e-05, "loss": 0.3152, "step": 51410 }, { "epoch": 2.4, "learning_rate": 2.5280451843415038e-05, "loss": 0.1733, "step": 51415 }, { "epoch": 2.4, "learning_rate": 2.5279983773579018e-05, "loss": 0.0389, "step": 51420 }, { "epoch": 2.4, "learning_rate": 2.5279515703743e-05, "loss": 0.0556, "step": 51425 }, { "epoch": 2.4, "learning_rate": 2.527904763390698e-05, "loss": 0.0541, "step": 51430 }, { "epoch": 2.4, "learning_rate": 2.527857956407096e-05, "loss": 0.0539, "step": 51435 }, { "epoch": 2.4, "learning_rate": 2.527811149423494e-05, "loss": 0.0457, "step": 51440 }, { "epoch": 2.4, "learning_rate": 2.527764342439892e-05, "loss": 0.1715, "step": 51445 }, { "epoch": 2.4, "learning_rate": 2.52771753545629e-05, "loss": 0.1252, "step": 51450 }, { "epoch": 2.4, "learning_rate": 2.527670728472688e-05, "loss": 0.1371, "step": 51455 }, { "epoch": 2.4, "learning_rate": 2.5276239214890863e-05, "loss": 0.1839, "step": 51460 }, { "epoch": 2.4, "learning_rate": 2.5275771145054843e-05, "loss": 0.2957, "step": 51465 }, { "epoch": 2.4, "learning_rate": 2.5275303075218823e-05, "loss": 0.0435, "step": 51470 }, { "epoch": 2.4, "learning_rate": 2.5274835005382803e-05, "loss": 0.0607, "step": 51475 }, { "epoch": 2.4, "learning_rate": 2.5274366935546786e-05, "loss": 0.1073, "step": 51480 }, { "epoch": 2.4, "learning_rate": 2.5273898865710766e-05, "loss": 0.0364, "step": 51485 }, { "epoch": 2.4, "learning_rate": 2.5273430795874745e-05, "loss": 0.0529, "step": 51490 }, { "epoch": 2.4, "learning_rate": 2.5272962726038725e-05, "loss": 0.1174, "step": 51495 }, { "epoch": 2.4, "learning_rate": 2.527249465620271e-05, "loss": 0.1237, "step": 51500 }, { "epoch": 2.4, "learning_rate": 2.5272026586366685e-05, "loss": 0.1694, "step": 51505 }, { "epoch": 2.4, "learning_rate": 2.5271558516530665e-05, "loss": 0.2309, "step": 51510 }, { "epoch": 2.4, "learning_rate": 2.5271090446694648e-05, "loss": 0.2679, "step": 51515 }, { "epoch": 2.4, "learning_rate": 2.5270622376858628e-05, "loss": 0.0264, "step": 51520 }, { "epoch": 2.4, "learning_rate": 2.5270154307022608e-05, "loss": 0.0459, "step": 51525 }, { "epoch": 2.4, "learning_rate": 2.5269686237186587e-05, "loss": 0.0516, "step": 51530 }, { "epoch": 2.4, "learning_rate": 2.526921816735057e-05, "loss": 0.0762, "step": 51535 }, { "epoch": 2.4, "learning_rate": 2.526875009751455e-05, "loss": 0.1188, "step": 51540 }, { "epoch": 2.41, "learning_rate": 2.526828202767853e-05, "loss": 0.1007, "step": 51545 }, { "epoch": 2.41, "learning_rate": 2.526781395784251e-05, "loss": 0.11, "step": 51550 }, { "epoch": 2.41, "learning_rate": 2.5267345888006493e-05, "loss": 0.1335, "step": 51555 }, { "epoch": 2.41, "learning_rate": 2.5266877818170473e-05, "loss": 0.1747, "step": 51560 }, { "epoch": 2.41, "learning_rate": 2.5266409748334453e-05, "loss": 0.4072, "step": 51565 }, { "epoch": 2.41, "learning_rate": 2.526594167849843e-05, "loss": 0.0689, "step": 51570 }, { "epoch": 2.41, "learning_rate": 2.5265473608662413e-05, "loss": 0.0277, "step": 51575 }, { "epoch": 2.41, "learning_rate": 2.5265005538826392e-05, "loss": 0.0869, "step": 51580 }, { "epoch": 2.41, "learning_rate": 2.5264537468990372e-05, "loss": 0.0741, "step": 51585 }, { "epoch": 2.41, "learning_rate": 2.5264069399154355e-05, "loss": 0.1226, "step": 51590 }, { "epoch": 2.41, "learning_rate": 2.5263601329318335e-05, "loss": 0.1205, "step": 51595 }, { "epoch": 2.41, "learning_rate": 2.5263133259482315e-05, "loss": 0.0689, "step": 51600 }, { "epoch": 2.41, "learning_rate": 2.5262665189646295e-05, "loss": 0.1892, "step": 51605 }, { "epoch": 2.41, "learning_rate": 2.5262197119810278e-05, "loss": 0.329, "step": 51610 }, { "epoch": 2.41, "learning_rate": 2.5261729049974258e-05, "loss": 0.2043, "step": 51615 }, { "epoch": 2.41, "learning_rate": 2.5261260980138238e-05, "loss": 0.034, "step": 51620 }, { "epoch": 2.41, "learning_rate": 2.5260792910302217e-05, "loss": 0.0212, "step": 51625 }, { "epoch": 2.41, "learning_rate": 2.52603248404662e-05, "loss": 0.0976, "step": 51630 }, { "epoch": 2.41, "learning_rate": 2.5259856770630177e-05, "loss": 0.134, "step": 51635 }, { "epoch": 2.41, "learning_rate": 2.5259388700794157e-05, "loss": 0.1299, "step": 51640 }, { "epoch": 2.41, "learning_rate": 2.525892063095814e-05, "loss": 0.1233, "step": 51645 }, { "epoch": 2.41, "learning_rate": 2.525845256112212e-05, "loss": 0.047, "step": 51650 }, { "epoch": 2.41, "learning_rate": 2.52579844912861e-05, "loss": 0.1237, "step": 51655 }, { "epoch": 2.41, "learning_rate": 2.525751642145008e-05, "loss": 0.2555, "step": 51660 }, { "epoch": 2.41, "learning_rate": 2.5257048351614063e-05, "loss": 0.3221, "step": 51665 }, { "epoch": 2.41, "learning_rate": 2.5256580281778043e-05, "loss": 0.0669, "step": 51670 }, { "epoch": 2.41, "learning_rate": 2.5256112211942022e-05, "loss": 0.0397, "step": 51675 }, { "epoch": 2.41, "learning_rate": 2.5255644142106002e-05, "loss": 0.1037, "step": 51680 }, { "epoch": 2.41, "learning_rate": 2.5255176072269985e-05, "loss": 0.049, "step": 51685 }, { "epoch": 2.41, "learning_rate": 2.5254708002433965e-05, "loss": 0.045, "step": 51690 }, { "epoch": 2.41, "learning_rate": 2.5254239932597942e-05, "loss": 0.0432, "step": 51695 }, { "epoch": 2.41, "learning_rate": 2.5253771862761925e-05, "loss": 0.0717, "step": 51700 }, { "epoch": 2.41, "learning_rate": 2.5253303792925905e-05, "loss": 0.2429, "step": 51705 }, { "epoch": 2.41, "learning_rate": 2.5252835723089885e-05, "loss": 0.2468, "step": 51710 }, { "epoch": 2.41, "learning_rate": 2.5252367653253864e-05, "loss": 0.3373, "step": 51715 }, { "epoch": 2.41, "learning_rate": 2.5251899583417848e-05, "loss": 0.0418, "step": 51720 }, { "epoch": 2.41, "learning_rate": 2.5251431513581827e-05, "loss": 0.0174, "step": 51725 }, { "epoch": 2.41, "learning_rate": 2.5250963443745807e-05, "loss": 0.0545, "step": 51730 }, { "epoch": 2.41, "learning_rate": 2.5250495373909787e-05, "loss": 0.0718, "step": 51735 }, { "epoch": 2.41, "learning_rate": 2.525002730407377e-05, "loss": 0.1581, "step": 51740 }, { "epoch": 2.41, "learning_rate": 2.524955923423775e-05, "loss": 0.101, "step": 51745 }, { "epoch": 2.41, "learning_rate": 2.524909116440173e-05, "loss": 0.1302, "step": 51750 }, { "epoch": 2.41, "learning_rate": 2.5248623094565713e-05, "loss": 0.1135, "step": 51755 }, { "epoch": 2.42, "learning_rate": 2.524815502472969e-05, "loss": 0.2549, "step": 51760 }, { "epoch": 2.42, "learning_rate": 2.524768695489367e-05, "loss": 0.4319, "step": 51765 }, { "epoch": 2.42, "learning_rate": 2.524721888505765e-05, "loss": 0.055, "step": 51770 }, { "epoch": 2.42, "learning_rate": 2.5246750815221632e-05, "loss": 0.0365, "step": 51775 }, { "epoch": 2.42, "learning_rate": 2.5246282745385612e-05, "loss": 0.0658, "step": 51780 }, { "epoch": 2.42, "learning_rate": 2.5245814675549592e-05, "loss": 0.084, "step": 51785 }, { "epoch": 2.42, "learning_rate": 2.5245346605713572e-05, "loss": 0.1265, "step": 51790 }, { "epoch": 2.42, "learning_rate": 2.5244878535877555e-05, "loss": 0.0726, "step": 51795 }, { "epoch": 2.42, "learning_rate": 2.5244410466041535e-05, "loss": 0.1055, "step": 51800 }, { "epoch": 2.42, "learning_rate": 2.5243942396205515e-05, "loss": 0.1952, "step": 51805 }, { "epoch": 2.42, "learning_rate": 2.5243474326369494e-05, "loss": 0.1421, "step": 51810 }, { "epoch": 2.42, "learning_rate": 2.5243006256533478e-05, "loss": 0.3485, "step": 51815 }, { "epoch": 2.42, "learning_rate": 2.5242538186697457e-05, "loss": 0.0908, "step": 51820 }, { "epoch": 2.42, "learning_rate": 2.5242070116861434e-05, "loss": 0.0349, "step": 51825 }, { "epoch": 2.42, "learning_rate": 2.5241602047025417e-05, "loss": 0.0974, "step": 51830 }, { "epoch": 2.42, "learning_rate": 2.5241133977189397e-05, "loss": 0.0189, "step": 51835 }, { "epoch": 2.42, "learning_rate": 2.5240665907353377e-05, "loss": 0.0704, "step": 51840 }, { "epoch": 2.42, "learning_rate": 2.5240197837517357e-05, "loss": 0.1367, "step": 51845 }, { "epoch": 2.42, "learning_rate": 2.523972976768134e-05, "loss": 0.1309, "step": 51850 }, { "epoch": 2.42, "learning_rate": 2.523926169784532e-05, "loss": 0.2016, "step": 51855 }, { "epoch": 2.42, "learning_rate": 2.52387936280093e-05, "loss": 0.1476, "step": 51860 }, { "epoch": 2.42, "learning_rate": 2.523832555817328e-05, "loss": 0.3493, "step": 51865 }, { "epoch": 2.42, "learning_rate": 2.5237857488337262e-05, "loss": 0.0545, "step": 51870 }, { "epoch": 2.42, "learning_rate": 2.5237389418501242e-05, "loss": 0.0496, "step": 51875 }, { "epoch": 2.42, "learning_rate": 2.5236921348665222e-05, "loss": 0.0513, "step": 51880 }, { "epoch": 2.42, "learning_rate": 2.5236453278829202e-05, "loss": 0.0263, "step": 51885 }, { "epoch": 2.42, "learning_rate": 2.5235985208993182e-05, "loss": 0.0704, "step": 51890 }, { "epoch": 2.42, "learning_rate": 2.523551713915716e-05, "loss": 0.1454, "step": 51895 }, { "epoch": 2.42, "learning_rate": 2.523504906932114e-05, "loss": 0.201, "step": 51900 }, { "epoch": 2.42, "learning_rate": 2.5234580999485125e-05, "loss": 0.2355, "step": 51905 }, { "epoch": 2.42, "learning_rate": 2.5234112929649104e-05, "loss": 0.3176, "step": 51910 }, { "epoch": 2.42, "learning_rate": 2.5233644859813084e-05, "loss": 0.2759, "step": 51915 }, { "epoch": 2.42, "learning_rate": 2.5233176789977064e-05, "loss": 0.0285, "step": 51920 }, { "epoch": 2.42, "learning_rate": 2.5232708720141047e-05, "loss": 0.0275, "step": 51925 }, { "epoch": 2.42, "learning_rate": 2.5232240650305027e-05, "loss": 0.0391, "step": 51930 }, { "epoch": 2.42, "learning_rate": 2.5231772580469007e-05, "loss": 0.0544, "step": 51935 }, { "epoch": 2.42, "learning_rate": 2.523130451063299e-05, "loss": 0.084, "step": 51940 }, { "epoch": 2.42, "learning_rate": 2.523083644079697e-05, "loss": 0.0853, "step": 51945 }, { "epoch": 2.42, "learning_rate": 2.5230368370960946e-05, "loss": 0.1162, "step": 51950 }, { "epoch": 2.42, "learning_rate": 2.5229900301124926e-05, "loss": 0.2122, "step": 51955 }, { "epoch": 2.42, "learning_rate": 2.522943223128891e-05, "loss": 0.26, "step": 51960 }, { "epoch": 2.42, "learning_rate": 2.522896416145289e-05, "loss": 0.2532, "step": 51965 }, { "epoch": 2.42, "learning_rate": 2.522849609161687e-05, "loss": 0.0949, "step": 51970 }, { "epoch": 2.43, "learning_rate": 2.522802802178085e-05, "loss": 0.0662, "step": 51975 }, { "epoch": 2.43, "learning_rate": 2.5227559951944832e-05, "loss": 0.0292, "step": 51980 }, { "epoch": 2.43, "learning_rate": 2.5227091882108812e-05, "loss": 0.0273, "step": 51985 }, { "epoch": 2.43, "learning_rate": 2.522662381227279e-05, "loss": 0.0494, "step": 51990 }, { "epoch": 2.43, "learning_rate": 2.522615574243677e-05, "loss": 0.0851, "step": 51995 }, { "epoch": 2.43, "learning_rate": 2.5225687672600755e-05, "loss": 0.1209, "step": 52000 }, { "epoch": 2.43, "learning_rate": 2.5225219602764734e-05, "loss": 0.2379, "step": 52005 }, { "epoch": 2.43, "learning_rate": 2.5224751532928714e-05, "loss": 0.1028, "step": 52010 }, { "epoch": 2.43, "learning_rate": 2.5224283463092694e-05, "loss": 0.2222, "step": 52015 }, { "epoch": 2.43, "learning_rate": 2.5223815393256674e-05, "loss": 0.0303, "step": 52020 }, { "epoch": 2.43, "learning_rate": 2.5223347323420654e-05, "loss": 0.0438, "step": 52025 }, { "epoch": 2.43, "learning_rate": 2.5222879253584634e-05, "loss": 0.0611, "step": 52030 }, { "epoch": 2.43, "learning_rate": 2.5222411183748617e-05, "loss": 0.0848, "step": 52035 }, { "epoch": 2.43, "learning_rate": 2.5221943113912597e-05, "loss": 0.0753, "step": 52040 }, { "epoch": 2.43, "learning_rate": 2.5221475044076576e-05, "loss": 0.1874, "step": 52045 }, { "epoch": 2.43, "learning_rate": 2.5221006974240556e-05, "loss": 0.1245, "step": 52050 }, { "epoch": 2.43, "learning_rate": 2.522053890440454e-05, "loss": 0.1719, "step": 52055 }, { "epoch": 2.43, "learning_rate": 2.522007083456852e-05, "loss": 0.3701, "step": 52060 }, { "epoch": 2.43, "learning_rate": 2.52196027647325e-05, "loss": 0.201, "step": 52065 }, { "epoch": 2.43, "learning_rate": 2.5219134694896482e-05, "loss": 0.0651, "step": 52070 }, { "epoch": 2.43, "learning_rate": 2.521866662506046e-05, "loss": 0.0635, "step": 52075 }, { "epoch": 2.43, "learning_rate": 2.521819855522444e-05, "loss": 0.0555, "step": 52080 }, { "epoch": 2.43, "learning_rate": 2.521773048538842e-05, "loss": 0.0525, "step": 52085 }, { "epoch": 2.43, "learning_rate": 2.52172624155524e-05, "loss": 0.0648, "step": 52090 }, { "epoch": 2.43, "learning_rate": 2.521679434571638e-05, "loss": 0.1424, "step": 52095 }, { "epoch": 2.43, "learning_rate": 2.521632627588036e-05, "loss": 0.1147, "step": 52100 }, { "epoch": 2.43, "learning_rate": 2.521585820604434e-05, "loss": 0.1584, "step": 52105 }, { "epoch": 2.43, "learning_rate": 2.5215390136208324e-05, "loss": 0.2292, "step": 52110 }, { "epoch": 2.43, "learning_rate": 2.5214922066372304e-05, "loss": 0.3829, "step": 52115 }, { "epoch": 2.43, "learning_rate": 2.5214453996536284e-05, "loss": 0.047, "step": 52120 }, { "epoch": 2.43, "learning_rate": 2.5213985926700267e-05, "loss": 0.0609, "step": 52125 }, { "epoch": 2.43, "learning_rate": 2.5213517856864247e-05, "loss": 0.066, "step": 52130 }, { "epoch": 2.43, "learning_rate": 2.5213049787028227e-05, "loss": 0.0748, "step": 52135 }, { "epoch": 2.43, "learning_rate": 2.5212581717192203e-05, "loss": 0.1025, "step": 52140 }, { "epoch": 2.43, "learning_rate": 2.5212113647356186e-05, "loss": 0.0543, "step": 52145 }, { "epoch": 2.43, "learning_rate": 2.5211645577520166e-05, "loss": 0.0698, "step": 52150 }, { "epoch": 2.43, "learning_rate": 2.5211177507684146e-05, "loss": 0.2142, "step": 52155 }, { "epoch": 2.43, "learning_rate": 2.5210709437848126e-05, "loss": 0.2254, "step": 52160 }, { "epoch": 2.43, "learning_rate": 2.521024136801211e-05, "loss": 0.4205, "step": 52165 }, { "epoch": 2.43, "learning_rate": 2.520977329817609e-05, "loss": 0.1196, "step": 52170 }, { "epoch": 2.43, "learning_rate": 2.520930522834007e-05, "loss": 0.0188, "step": 52175 }, { "epoch": 2.43, "learning_rate": 2.520883715850405e-05, "loss": 0.0982, "step": 52180 }, { "epoch": 2.44, "learning_rate": 2.520836908866803e-05, "loss": 0.1506, "step": 52185 }, { "epoch": 2.44, "learning_rate": 2.520790101883201e-05, "loss": 0.0921, "step": 52190 }, { "epoch": 2.44, "learning_rate": 2.520743294899599e-05, "loss": 0.1336, "step": 52195 }, { "epoch": 2.44, "learning_rate": 2.520696487915997e-05, "loss": 0.1143, "step": 52200 }, { "epoch": 2.44, "learning_rate": 2.520649680932395e-05, "loss": 0.214, "step": 52205 }, { "epoch": 2.44, "learning_rate": 2.520602873948793e-05, "loss": 0.2243, "step": 52210 }, { "epoch": 2.44, "learning_rate": 2.520556066965191e-05, "loss": 0.4304, "step": 52215 }, { "epoch": 2.44, "learning_rate": 2.5205092599815894e-05, "loss": 0.0319, "step": 52220 }, { "epoch": 2.44, "learning_rate": 2.5204624529979874e-05, "loss": 0.0231, "step": 52225 }, { "epoch": 2.44, "learning_rate": 2.5204156460143853e-05, "loss": 0.0737, "step": 52230 }, { "epoch": 2.44, "learning_rate": 2.5203688390307833e-05, "loss": 0.0324, "step": 52235 }, { "epoch": 2.44, "learning_rate": 2.5203220320471816e-05, "loss": 0.1592, "step": 52240 }, { "epoch": 2.44, "learning_rate": 2.5202752250635796e-05, "loss": 0.0736, "step": 52245 }, { "epoch": 2.44, "learning_rate": 2.5202284180799776e-05, "loss": 0.1332, "step": 52250 }, { "epoch": 2.44, "learning_rate": 2.520181611096376e-05, "loss": 0.1342, "step": 52255 }, { "epoch": 2.44, "learning_rate": 2.520134804112774e-05, "loss": 0.1728, "step": 52260 }, { "epoch": 2.44, "learning_rate": 2.5200879971291715e-05, "loss": 0.278, "step": 52265 }, { "epoch": 2.44, "learning_rate": 2.5200411901455695e-05, "loss": 0.0266, "step": 52270 }, { "epoch": 2.44, "learning_rate": 2.519994383161968e-05, "loss": 0.046, "step": 52275 }, { "epoch": 2.44, "learning_rate": 2.5199475761783658e-05, "loss": 0.0686, "step": 52280 }, { "epoch": 2.44, "learning_rate": 2.5199007691947638e-05, "loss": 0.0621, "step": 52285 }, { "epoch": 2.44, "learning_rate": 2.5198539622111618e-05, "loss": 0.0907, "step": 52290 }, { "epoch": 2.44, "learning_rate": 2.51980715522756e-05, "loss": 0.0781, "step": 52295 }, { "epoch": 2.44, "learning_rate": 2.519760348243958e-05, "loss": 0.0904, "step": 52300 }, { "epoch": 2.44, "learning_rate": 2.519713541260356e-05, "loss": 0.1418, "step": 52305 }, { "epoch": 2.44, "learning_rate": 2.5196667342767544e-05, "loss": 0.1473, "step": 52310 }, { "epoch": 2.44, "learning_rate": 2.5196199272931524e-05, "loss": 0.2129, "step": 52315 }, { "epoch": 2.44, "learning_rate": 2.5195731203095504e-05, "loss": 0.0451, "step": 52320 }, { "epoch": 2.44, "learning_rate": 2.5195263133259483e-05, "loss": 0.0179, "step": 52325 }, { "epoch": 2.44, "learning_rate": 2.5194795063423463e-05, "loss": 0.0259, "step": 52330 }, { "epoch": 2.44, "learning_rate": 2.5194326993587443e-05, "loss": 0.0715, "step": 52335 }, { "epoch": 2.44, "learning_rate": 2.5193858923751423e-05, "loss": 0.1341, "step": 52340 }, { "epoch": 2.44, "learning_rate": 2.5193390853915403e-05, "loss": 0.054, "step": 52345 }, { "epoch": 2.44, "learning_rate": 2.5192922784079386e-05, "loss": 0.0753, "step": 52350 }, { "epoch": 2.44, "learning_rate": 2.5192454714243366e-05, "loss": 0.1833, "step": 52355 }, { "epoch": 2.44, "learning_rate": 2.5191986644407346e-05, "loss": 0.6172, "step": 52360 }, { "epoch": 2.44, "learning_rate": 2.519151857457133e-05, "loss": 0.2876, "step": 52365 }, { "epoch": 2.44, "learning_rate": 2.519105050473531e-05, "loss": 0.0317, "step": 52370 }, { "epoch": 2.44, "learning_rate": 2.519058243489929e-05, "loss": 0.0545, "step": 52375 }, { "epoch": 2.44, "learning_rate": 2.5190114365063268e-05, "loss": 0.0487, "step": 52380 }, { "epoch": 2.44, "learning_rate": 2.518964629522725e-05, "loss": 0.0414, "step": 52385 }, { "epoch": 2.44, "learning_rate": 2.5189178225391228e-05, "loss": 0.0762, "step": 52390 }, { "epoch": 2.44, "learning_rate": 2.5188710155555208e-05, "loss": 0.1468, "step": 52395 }, { "epoch": 2.45, "learning_rate": 2.5188242085719187e-05, "loss": 0.0978, "step": 52400 }, { "epoch": 2.45, "learning_rate": 2.518777401588317e-05, "loss": 0.1532, "step": 52405 }, { "epoch": 2.45, "learning_rate": 2.518730594604715e-05, "loss": 0.3888, "step": 52410 }, { "epoch": 2.45, "learning_rate": 2.518683787621113e-05, "loss": 0.2808, "step": 52415 }, { "epoch": 2.45, "learning_rate": 2.518636980637511e-05, "loss": 0.043, "step": 52420 }, { "epoch": 2.45, "learning_rate": 2.5185901736539093e-05, "loss": 0.0324, "step": 52425 }, { "epoch": 2.45, "learning_rate": 2.5185433666703073e-05, "loss": 0.0765, "step": 52430 }, { "epoch": 2.45, "learning_rate": 2.5184965596867053e-05, "loss": 0.069, "step": 52435 }, { "epoch": 2.45, "learning_rate": 2.5184497527031036e-05, "loss": 0.0406, "step": 52440 }, { "epoch": 2.45, "learning_rate": 2.5184029457195016e-05, "loss": 0.0692, "step": 52445 }, { "epoch": 2.45, "learning_rate": 2.5183561387358996e-05, "loss": 0.2368, "step": 52450 }, { "epoch": 2.45, "learning_rate": 2.5183093317522972e-05, "loss": 0.131, "step": 52455 }, { "epoch": 2.45, "learning_rate": 2.5182625247686955e-05, "loss": 0.2531, "step": 52460 }, { "epoch": 2.45, "learning_rate": 2.5182157177850935e-05, "loss": 0.2602, "step": 52465 }, { "epoch": 2.45, "learning_rate": 2.5181689108014915e-05, "loss": 0.0813, "step": 52470 }, { "epoch": 2.45, "learning_rate": 2.5181221038178895e-05, "loss": 0.0428, "step": 52475 }, { "epoch": 2.45, "learning_rate": 2.5180752968342878e-05, "loss": 0.0207, "step": 52480 }, { "epoch": 2.45, "learning_rate": 2.5180284898506858e-05, "loss": 0.0731, "step": 52485 }, { "epoch": 2.45, "learning_rate": 2.5179816828670838e-05, "loss": 0.1143, "step": 52490 }, { "epoch": 2.45, "learning_rate": 2.517934875883482e-05, "loss": 0.1158, "step": 52495 }, { "epoch": 2.45, "learning_rate": 2.51788806889988e-05, "loss": 0.1362, "step": 52500 }, { "epoch": 2.45, "learning_rate": 2.517841261916278e-05, "loss": 0.0893, "step": 52505 }, { "epoch": 2.45, "learning_rate": 2.517794454932676e-05, "loss": 0.2606, "step": 52510 }, { "epoch": 2.45, "learning_rate": 2.5177476479490744e-05, "loss": 0.3078, "step": 52515 }, { "epoch": 2.45, "learning_rate": 2.517700840965472e-05, "loss": 0.0162, "step": 52520 }, { "epoch": 2.45, "learning_rate": 2.51765403398187e-05, "loss": 0.0139, "step": 52525 }, { "epoch": 2.45, "learning_rate": 2.517607226998268e-05, "loss": 0.0568, "step": 52530 }, { "epoch": 2.45, "learning_rate": 2.5175604200146663e-05, "loss": 0.056, "step": 52535 }, { "epoch": 2.45, "learning_rate": 2.5175136130310643e-05, "loss": 0.1181, "step": 52540 }, { "epoch": 2.45, "learning_rate": 2.5174668060474623e-05, "loss": 0.0799, "step": 52545 }, { "epoch": 2.45, "learning_rate": 2.5174199990638606e-05, "loss": 0.1234, "step": 52550 }, { "epoch": 2.45, "learning_rate": 2.5173731920802586e-05, "loss": 0.1662, "step": 52555 }, { "epoch": 2.45, "learning_rate": 2.5173263850966565e-05, "loss": 0.1641, "step": 52560 }, { "epoch": 2.45, "learning_rate": 2.5172795781130545e-05, "loss": 0.4078, "step": 52565 }, { "epoch": 2.45, "learning_rate": 2.517232771129453e-05, "loss": 0.0641, "step": 52570 }, { "epoch": 2.45, "learning_rate": 2.5171859641458508e-05, "loss": 0.0375, "step": 52575 }, { "epoch": 2.45, "learning_rate": 2.5171391571622485e-05, "loss": 0.0733, "step": 52580 }, { "epoch": 2.45, "learning_rate": 2.5170923501786464e-05, "loss": 0.0778, "step": 52585 }, { "epoch": 2.45, "learning_rate": 2.5170455431950448e-05, "loss": 0.0773, "step": 52590 }, { "epoch": 2.45, "learning_rate": 2.5169987362114427e-05, "loss": 0.117, "step": 52595 }, { "epoch": 2.45, "learning_rate": 2.5169519292278407e-05, "loss": 0.226, "step": 52600 }, { "epoch": 2.45, "learning_rate": 2.5169051222442387e-05, "loss": 0.184, "step": 52605 }, { "epoch": 2.45, "learning_rate": 2.516858315260637e-05, "loss": 0.2767, "step": 52610 }, { "epoch": 2.46, "learning_rate": 2.516811508277035e-05, "loss": 0.3545, "step": 52615 }, { "epoch": 2.46, "learning_rate": 2.516764701293433e-05, "loss": 0.046, "step": 52620 }, { "epoch": 2.46, "learning_rate": 2.5167178943098313e-05, "loss": 0.0502, "step": 52625 }, { "epoch": 2.46, "learning_rate": 2.5166710873262293e-05, "loss": 0.1021, "step": 52630 }, { "epoch": 2.46, "learning_rate": 2.5166242803426273e-05, "loss": 0.0652, "step": 52635 }, { "epoch": 2.46, "learning_rate": 2.5165774733590253e-05, "loss": 0.1092, "step": 52640 }, { "epoch": 2.46, "learning_rate": 2.5165306663754232e-05, "loss": 0.0526, "step": 52645 }, { "epoch": 2.46, "learning_rate": 2.5164838593918212e-05, "loss": 0.0951, "step": 52650 }, { "epoch": 2.46, "learning_rate": 2.5164370524082192e-05, "loss": 0.1798, "step": 52655 }, { "epoch": 2.46, "learning_rate": 2.5163902454246172e-05, "loss": 0.2423, "step": 52660 }, { "epoch": 2.46, "learning_rate": 2.5163434384410155e-05, "loss": 0.2527, "step": 52665 }, { "epoch": 2.46, "learning_rate": 2.5162966314574135e-05, "loss": 0.0154, "step": 52670 }, { "epoch": 2.46, "learning_rate": 2.5162498244738115e-05, "loss": 0.0181, "step": 52675 }, { "epoch": 2.46, "learning_rate": 2.5162030174902098e-05, "loss": 0.0505, "step": 52680 }, { "epoch": 2.46, "learning_rate": 2.5161562105066078e-05, "loss": 0.0972, "step": 52685 }, { "epoch": 2.46, "learning_rate": 2.5161094035230058e-05, "loss": 0.0295, "step": 52690 }, { "epoch": 2.46, "learning_rate": 2.5160625965394037e-05, "loss": 0.0877, "step": 52695 }, { "epoch": 2.46, "learning_rate": 2.516015789555802e-05, "loss": 0.1176, "step": 52700 }, { "epoch": 2.46, "learning_rate": 2.5159689825722e-05, "loss": 0.109, "step": 52705 }, { "epoch": 2.46, "learning_rate": 2.5159221755885977e-05, "loss": 0.316, "step": 52710 }, { "epoch": 2.46, "learning_rate": 2.5158753686049957e-05, "loss": 0.3343, "step": 52715 }, { "epoch": 2.46, "learning_rate": 2.515828561621394e-05, "loss": 0.0537, "step": 52720 }, { "epoch": 2.46, "learning_rate": 2.515781754637792e-05, "loss": 0.0435, "step": 52725 }, { "epoch": 2.46, "learning_rate": 2.51573494765419e-05, "loss": 0.0566, "step": 52730 }, { "epoch": 2.46, "learning_rate": 2.5156881406705883e-05, "loss": 0.0749, "step": 52735 }, { "epoch": 2.46, "learning_rate": 2.5156413336869863e-05, "loss": 0.0451, "step": 52740 }, { "epoch": 2.46, "learning_rate": 2.5155945267033842e-05, "loss": 0.1212, "step": 52745 }, { "epoch": 2.46, "learning_rate": 2.5155477197197822e-05, "loss": 0.1203, "step": 52750 }, { "epoch": 2.46, "learning_rate": 2.5155009127361805e-05, "loss": 0.2059, "step": 52755 }, { "epoch": 2.46, "learning_rate": 2.5154541057525785e-05, "loss": 0.2683, "step": 52760 }, { "epoch": 2.46, "learning_rate": 2.5154072987689765e-05, "loss": 0.2852, "step": 52765 }, { "epoch": 2.46, "learning_rate": 2.515360491785374e-05, "loss": 0.0461, "step": 52770 }, { "epoch": 2.46, "learning_rate": 2.5153136848017725e-05, "loss": 0.0449, "step": 52775 }, { "epoch": 2.46, "learning_rate": 2.5152668778181704e-05, "loss": 0.047, "step": 52780 }, { "epoch": 2.46, "learning_rate": 2.5152200708345684e-05, "loss": 0.0799, "step": 52785 }, { "epoch": 2.46, "learning_rate": 2.5151732638509664e-05, "loss": 0.055, "step": 52790 }, { "epoch": 2.46, "learning_rate": 2.5151264568673647e-05, "loss": 0.1364, "step": 52795 }, { "epoch": 2.46, "learning_rate": 2.5150796498837627e-05, "loss": 0.0631, "step": 52800 }, { "epoch": 2.46, "learning_rate": 2.5150328429001607e-05, "loss": 0.1585, "step": 52805 }, { "epoch": 2.46, "learning_rate": 2.514986035916559e-05, "loss": 0.3361, "step": 52810 }, { "epoch": 2.46, "learning_rate": 2.514939228932957e-05, "loss": 0.2617, "step": 52815 }, { "epoch": 2.46, "learning_rate": 2.514892421949355e-05, "loss": 0.1086, "step": 52820 }, { "epoch": 2.46, "learning_rate": 2.514845614965753e-05, "loss": 0.0886, "step": 52825 }, { "epoch": 2.47, "learning_rate": 2.5147988079821513e-05, "loss": 0.0402, "step": 52830 }, { "epoch": 2.47, "learning_rate": 2.514752000998549e-05, "loss": 0.0471, "step": 52835 }, { "epoch": 2.47, "learning_rate": 2.514705194014947e-05, "loss": 0.1097, "step": 52840 }, { "epoch": 2.47, "learning_rate": 2.514658387031345e-05, "loss": 0.0571, "step": 52845 }, { "epoch": 2.47, "learning_rate": 2.5146115800477432e-05, "loss": 0.1192, "step": 52850 }, { "epoch": 2.47, "learning_rate": 2.5145647730641412e-05, "loss": 0.1675, "step": 52855 }, { "epoch": 2.47, "learning_rate": 2.514517966080539e-05, "loss": 0.1556, "step": 52860 }, { "epoch": 2.47, "learning_rate": 2.5144711590969375e-05, "loss": 0.2968, "step": 52865 }, { "epoch": 2.47, "learning_rate": 2.5144243521133355e-05, "loss": 0.0888, "step": 52870 }, { "epoch": 2.47, "learning_rate": 2.5143775451297335e-05, "loss": 0.0484, "step": 52875 }, { "epoch": 2.47, "learning_rate": 2.5143307381461314e-05, "loss": 0.0456, "step": 52880 }, { "epoch": 2.47, "learning_rate": 2.5142839311625298e-05, "loss": 0.0619, "step": 52885 }, { "epoch": 2.47, "learning_rate": 2.5142371241789277e-05, "loss": 0.0961, "step": 52890 }, { "epoch": 2.47, "learning_rate": 2.5141903171953254e-05, "loss": 0.1127, "step": 52895 }, { "epoch": 2.47, "learning_rate": 2.5141435102117234e-05, "loss": 0.1211, "step": 52900 }, { "epoch": 2.47, "learning_rate": 2.5140967032281217e-05, "loss": 0.2063, "step": 52905 }, { "epoch": 2.47, "learning_rate": 2.5140498962445197e-05, "loss": 0.2051, "step": 52910 }, { "epoch": 2.47, "learning_rate": 2.5140030892609176e-05, "loss": 0.4252, "step": 52915 }, { "epoch": 2.47, "learning_rate": 2.513956282277316e-05, "loss": 0.0802, "step": 52920 }, { "epoch": 2.47, "learning_rate": 2.513909475293714e-05, "loss": 0.0148, "step": 52925 }, { "epoch": 2.47, "learning_rate": 2.513862668310112e-05, "loss": 0.1018, "step": 52930 }, { "epoch": 2.47, "learning_rate": 2.51381586132651e-05, "loss": 0.0707, "step": 52935 }, { "epoch": 2.47, "learning_rate": 2.5137690543429082e-05, "loss": 0.0244, "step": 52940 }, { "epoch": 2.47, "learning_rate": 2.5137222473593062e-05, "loss": 0.1077, "step": 52945 }, { "epoch": 2.47, "learning_rate": 2.5136754403757042e-05, "loss": 0.1026, "step": 52950 }, { "epoch": 2.47, "learning_rate": 2.5136286333921022e-05, "loss": 0.1085, "step": 52955 }, { "epoch": 2.47, "learning_rate": 2.5135818264085e-05, "loss": 0.2029, "step": 52960 }, { "epoch": 2.47, "learning_rate": 2.513535019424898e-05, "loss": 0.3476, "step": 52965 }, { "epoch": 2.47, "learning_rate": 2.513488212441296e-05, "loss": 0.0672, "step": 52970 }, { "epoch": 2.47, "learning_rate": 2.513441405457694e-05, "loss": 0.038, "step": 52975 }, { "epoch": 2.47, "learning_rate": 2.5133945984740924e-05, "loss": 0.0318, "step": 52980 }, { "epoch": 2.47, "learning_rate": 2.5133477914904904e-05, "loss": 0.1066, "step": 52985 }, { "epoch": 2.47, "learning_rate": 2.5133009845068884e-05, "loss": 0.0741, "step": 52990 }, { "epoch": 2.47, "learning_rate": 2.5132541775232867e-05, "loss": 0.2065, "step": 52995 }, { "epoch": 2.47, "learning_rate": 2.5132073705396847e-05, "loss": 0.2066, "step": 53000 }, { "epoch": 2.47, "learning_rate": 2.5131605635560827e-05, "loss": 0.124, "step": 53005 }, { "epoch": 2.47, "learning_rate": 2.5131137565724807e-05, "loss": 0.1881, "step": 53010 }, { "epoch": 2.47, "learning_rate": 2.513066949588879e-05, "loss": 0.225, "step": 53015 }, { "epoch": 2.47, "learning_rate": 2.513020142605277e-05, "loss": 0.0445, "step": 53020 }, { "epoch": 2.47, "learning_rate": 2.5129733356216746e-05, "loss": 0.0175, "step": 53025 }, { "epoch": 2.47, "learning_rate": 2.5129265286380726e-05, "loss": 0.0474, "step": 53030 }, { "epoch": 2.47, "learning_rate": 2.512879721654471e-05, "loss": 0.0526, "step": 53035 }, { "epoch": 2.47, "learning_rate": 2.512832914670869e-05, "loss": 0.4421, "step": 53040 }, { "epoch": 2.48, "learning_rate": 2.512786107687267e-05, "loss": 0.0497, "step": 53045 }, { "epoch": 2.48, "learning_rate": 2.5127393007036652e-05, "loss": 0.1382, "step": 53050 }, { "epoch": 2.48, "learning_rate": 2.512692493720063e-05, "loss": 0.1484, "step": 53055 }, { "epoch": 2.48, "learning_rate": 2.512645686736461e-05, "loss": 0.4425, "step": 53060 }, { "epoch": 2.48, "learning_rate": 2.512598879752859e-05, "loss": 0.3004, "step": 53065 }, { "epoch": 2.48, "learning_rate": 2.5125520727692575e-05, "loss": 0.0968, "step": 53070 }, { "epoch": 2.48, "learning_rate": 2.5125052657856554e-05, "loss": 0.0778, "step": 53075 }, { "epoch": 2.48, "learning_rate": 2.5124584588020534e-05, "loss": 0.0768, "step": 53080 }, { "epoch": 2.48, "learning_rate": 2.512411651818451e-05, "loss": 0.0855, "step": 53085 }, { "epoch": 2.48, "learning_rate": 2.5123648448348494e-05, "loss": 0.0752, "step": 53090 }, { "epoch": 2.48, "learning_rate": 2.5123180378512474e-05, "loss": 0.0801, "step": 53095 }, { "epoch": 2.48, "learning_rate": 2.5122712308676453e-05, "loss": 0.1066, "step": 53100 }, { "epoch": 2.48, "learning_rate": 2.5122244238840437e-05, "loss": 0.1512, "step": 53105 }, { "epoch": 2.48, "learning_rate": 2.5121776169004416e-05, "loss": 0.1996, "step": 53110 }, { "epoch": 2.48, "learning_rate": 2.5121308099168396e-05, "loss": 0.3174, "step": 53115 }, { "epoch": 2.48, "learning_rate": 2.5120840029332376e-05, "loss": 0.0307, "step": 53120 }, { "epoch": 2.48, "learning_rate": 2.512037195949636e-05, "loss": 0.0242, "step": 53125 }, { "epoch": 2.48, "learning_rate": 2.511990388966034e-05, "loss": 0.1071, "step": 53130 }, { "epoch": 2.48, "learning_rate": 2.511943581982432e-05, "loss": 0.0978, "step": 53135 }, { "epoch": 2.48, "learning_rate": 2.51189677499883e-05, "loss": 0.0576, "step": 53140 }, { "epoch": 2.48, "learning_rate": 2.5118499680152282e-05, "loss": 0.0713, "step": 53145 }, { "epoch": 2.48, "learning_rate": 2.511803161031626e-05, "loss": 0.038, "step": 53150 }, { "epoch": 2.48, "learning_rate": 2.5117563540480238e-05, "loss": 0.2546, "step": 53155 }, { "epoch": 2.48, "learning_rate": 2.511709547064422e-05, "loss": 0.2199, "step": 53160 }, { "epoch": 2.48, "learning_rate": 2.51166274008082e-05, "loss": 0.3676, "step": 53165 }, { "epoch": 2.48, "learning_rate": 2.511615933097218e-05, "loss": 0.0242, "step": 53170 }, { "epoch": 2.48, "learning_rate": 2.511569126113616e-05, "loss": 0.0589, "step": 53175 }, { "epoch": 2.48, "learning_rate": 2.5115223191300144e-05, "loss": 0.0429, "step": 53180 }, { "epoch": 2.48, "learning_rate": 2.5114755121464124e-05, "loss": 0.0971, "step": 53185 }, { "epoch": 2.48, "learning_rate": 2.5114287051628104e-05, "loss": 0.0915, "step": 53190 }, { "epoch": 2.48, "learning_rate": 2.5113818981792084e-05, "loss": 0.0838, "step": 53195 }, { "epoch": 2.48, "learning_rate": 2.5113350911956067e-05, "loss": 0.1195, "step": 53200 }, { "epoch": 2.48, "learning_rate": 2.5112882842120047e-05, "loss": 0.1195, "step": 53205 }, { "epoch": 2.48, "learning_rate": 2.5112414772284026e-05, "loss": 0.1937, "step": 53210 }, { "epoch": 2.48, "learning_rate": 2.5111946702448003e-05, "loss": 0.2583, "step": 53215 }, { "epoch": 2.48, "learning_rate": 2.5111478632611986e-05, "loss": 0.052, "step": 53220 }, { "epoch": 2.48, "learning_rate": 2.5111010562775966e-05, "loss": 0.0525, "step": 53225 }, { "epoch": 2.48, "learning_rate": 2.5110542492939946e-05, "loss": 0.0723, "step": 53230 }, { "epoch": 2.48, "learning_rate": 2.511007442310393e-05, "loss": 0.0695, "step": 53235 }, { "epoch": 2.48, "learning_rate": 2.510960635326791e-05, "loss": 0.0692, "step": 53240 }, { "epoch": 2.48, "learning_rate": 2.510913828343189e-05, "loss": 0.0966, "step": 53245 }, { "epoch": 2.48, "learning_rate": 2.5108670213595868e-05, "loss": 0.2415, "step": 53250 }, { "epoch": 2.48, "learning_rate": 2.510820214375985e-05, "loss": 0.2034, "step": 53255 }, { "epoch": 2.49, "learning_rate": 2.510773407392383e-05, "loss": 0.222, "step": 53260 }, { "epoch": 2.49, "learning_rate": 2.510726600408781e-05, "loss": 0.296, "step": 53265 }, { "epoch": 2.49, "learning_rate": 2.510679793425179e-05, "loss": 0.0594, "step": 53270 }, { "epoch": 2.49, "learning_rate": 2.510632986441577e-05, "loss": 0.0711, "step": 53275 }, { "epoch": 2.49, "learning_rate": 2.510586179457975e-05, "loss": 0.0367, "step": 53280 }, { "epoch": 2.49, "learning_rate": 2.510539372474373e-05, "loss": 0.0363, "step": 53285 }, { "epoch": 2.49, "learning_rate": 2.5104925654907714e-05, "loss": 0.0593, "step": 53290 }, { "epoch": 2.49, "learning_rate": 2.5104457585071693e-05, "loss": 0.1109, "step": 53295 }, { "epoch": 2.49, "learning_rate": 2.5103989515235673e-05, "loss": 0.1238, "step": 53300 }, { "epoch": 2.49, "learning_rate": 2.5103521445399653e-05, "loss": 0.1999, "step": 53305 }, { "epoch": 2.49, "learning_rate": 2.5103053375563636e-05, "loss": 0.2999, "step": 53310 }, { "epoch": 2.49, "learning_rate": 2.5102585305727616e-05, "loss": 0.1659, "step": 53315 }, { "epoch": 2.49, "learning_rate": 2.5102117235891596e-05, "loss": 0.0613, "step": 53320 }, { "epoch": 2.49, "learning_rate": 2.5101649166055576e-05, "loss": 0.0335, "step": 53325 }, { "epoch": 2.49, "learning_rate": 2.510118109621956e-05, "loss": 0.0715, "step": 53330 }, { "epoch": 2.49, "learning_rate": 2.510071302638354e-05, "loss": 0.0905, "step": 53335 }, { "epoch": 2.49, "learning_rate": 2.5100244956547515e-05, "loss": 0.0503, "step": 53340 }, { "epoch": 2.49, "learning_rate": 2.50997768867115e-05, "loss": 0.1676, "step": 53345 }, { "epoch": 2.49, "learning_rate": 2.5099308816875478e-05, "loss": 0.1139, "step": 53350 }, { "epoch": 2.49, "learning_rate": 2.5098840747039458e-05, "loss": 0.2108, "step": 53355 }, { "epoch": 2.49, "learning_rate": 2.5098372677203438e-05, "loss": 0.2596, "step": 53360 }, { "epoch": 2.49, "learning_rate": 2.509790460736742e-05, "loss": 0.2595, "step": 53365 }, { "epoch": 2.49, "learning_rate": 2.50974365375314e-05, "loss": 0.0527, "step": 53370 }, { "epoch": 2.49, "learning_rate": 2.509696846769538e-05, "loss": 0.022, "step": 53375 }, { "epoch": 2.49, "learning_rate": 2.509650039785936e-05, "loss": 0.0557, "step": 53380 }, { "epoch": 2.49, "learning_rate": 2.5096032328023344e-05, "loss": 0.0506, "step": 53385 }, { "epoch": 2.49, "learning_rate": 2.5095564258187324e-05, "loss": 0.0867, "step": 53390 }, { "epoch": 2.49, "learning_rate": 2.5095096188351303e-05, "loss": 0.0508, "step": 53395 }, { "epoch": 2.49, "learning_rate": 2.5094628118515283e-05, "loss": 0.1314, "step": 53400 }, { "epoch": 2.49, "learning_rate": 2.5094160048679263e-05, "loss": 0.2294, "step": 53405 }, { "epoch": 2.49, "learning_rate": 2.5093691978843243e-05, "loss": 0.2391, "step": 53410 }, { "epoch": 2.49, "learning_rate": 2.5093223909007223e-05, "loss": 0.2303, "step": 53415 }, { "epoch": 2.49, "learning_rate": 2.5092755839171206e-05, "loss": 0.0052, "step": 53420 }, { "epoch": 2.49, "learning_rate": 2.5092287769335186e-05, "loss": 0.0195, "step": 53425 }, { "epoch": 2.49, "learning_rate": 2.5091819699499165e-05, "loss": 0.0534, "step": 53430 }, { "epoch": 2.49, "learning_rate": 2.5091351629663145e-05, "loss": 0.0563, "step": 53435 }, { "epoch": 2.49, "learning_rate": 2.509088355982713e-05, "loss": 0.1052, "step": 53440 }, { "epoch": 2.49, "learning_rate": 2.5090415489991108e-05, "loss": 0.0889, "step": 53445 }, { "epoch": 2.49, "learning_rate": 2.5089947420155088e-05, "loss": 0.0898, "step": 53450 }, { "epoch": 2.49, "learning_rate": 2.5089479350319068e-05, "loss": 0.1483, "step": 53455 }, { "epoch": 2.49, "learning_rate": 2.508901128048305e-05, "loss": 0.2292, "step": 53460 }, { "epoch": 2.49, "learning_rate": 2.5088543210647028e-05, "loss": 0.3382, "step": 53465 }, { "epoch": 2.49, "learning_rate": 2.5088075140811007e-05, "loss": 0.0341, "step": 53470 }, { "epoch": 2.5, "learning_rate": 2.508760707097499e-05, "loss": 0.0266, "step": 53475 }, { "epoch": 2.5, "learning_rate": 2.508713900113897e-05, "loss": 0.0568, "step": 53480 }, { "epoch": 2.5, "learning_rate": 2.508667093130295e-05, "loss": 0.0772, "step": 53485 }, { "epoch": 2.5, "learning_rate": 2.508620286146693e-05, "loss": 0.0588, "step": 53490 }, { "epoch": 2.5, "learning_rate": 2.5085734791630913e-05, "loss": 0.1131, "step": 53495 }, { "epoch": 2.5, "learning_rate": 2.5085266721794893e-05, "loss": 0.1028, "step": 53500 }, { "epoch": 2.5, "learning_rate": 2.5084798651958873e-05, "loss": 0.0818, "step": 53505 }, { "epoch": 2.5, "learning_rate": 2.5084330582122853e-05, "loss": 0.127, "step": 53510 }, { "epoch": 2.5, "learning_rate": 2.5083862512286836e-05, "loss": 0.28, "step": 53515 }, { "epoch": 2.5, "learning_rate": 2.5083394442450816e-05, "loss": 0.0727, "step": 53520 }, { "epoch": 2.5, "learning_rate": 2.5082926372614796e-05, "loss": 0.0303, "step": 53525 }, { "epoch": 2.5, "learning_rate": 2.5082458302778775e-05, "loss": 0.0236, "step": 53530 }, { "epoch": 2.5, "learning_rate": 2.5081990232942755e-05, "loss": 0.0735, "step": 53535 }, { "epoch": 2.5, "learning_rate": 2.5081522163106735e-05, "loss": 0.061, "step": 53540 }, { "epoch": 2.5, "learning_rate": 2.5081054093270715e-05, "loss": 0.1206, "step": 53545 }, { "epoch": 2.5, "learning_rate": 2.5080586023434698e-05, "loss": 0.0839, "step": 53550 }, { "epoch": 2.5, "learning_rate": 2.5080117953598678e-05, "loss": 0.178, "step": 53555 }, { "epoch": 2.5, "learning_rate": 2.5079649883762658e-05, "loss": 0.1644, "step": 53560 }, { "epoch": 2.5, "learning_rate": 2.5079181813926637e-05, "loss": 0.2305, "step": 53565 }, { "epoch": 2.5, "learning_rate": 2.507871374409062e-05, "loss": 0.0892, "step": 53570 }, { "epoch": 2.5, "learning_rate": 2.50782456742546e-05, "loss": 0.0783, "step": 53575 }, { "epoch": 2.5, "learning_rate": 2.507777760441858e-05, "loss": 0.065, "step": 53580 }, { "epoch": 2.5, "learning_rate": 2.507730953458256e-05, "loss": 0.0319, "step": 53585 }, { "epoch": 2.5, "learning_rate": 2.507684146474654e-05, "loss": 0.0967, "step": 53590 }, { "epoch": 2.5, "learning_rate": 2.507637339491052e-05, "loss": 0.0123, "step": 53595 }, { "epoch": 2.5, "learning_rate": 2.50759053250745e-05, "loss": 0.0947, "step": 53600 }, { "epoch": 2.5, "learning_rate": 2.5075437255238483e-05, "loss": 0.3027, "step": 53605 }, { "epoch": 2.5, "learning_rate": 2.5074969185402463e-05, "loss": 0.2883, "step": 53610 }, { "epoch": 2.5, "learning_rate": 2.5074501115566442e-05, "loss": 0.2533, "step": 53615 }, { "epoch": 2.5, "learning_rate": 2.5074033045730422e-05, "loss": 0.0162, "step": 53620 }, { "epoch": 2.5, "learning_rate": 2.5073564975894405e-05, "loss": 0.0637, "step": 53625 }, { "epoch": 2.5, "learning_rate": 2.5073096906058385e-05, "loss": 0.1104, "step": 53630 }, { "epoch": 2.5, "learning_rate": 2.5072628836222365e-05, "loss": 0.0611, "step": 53635 }, { "epoch": 2.5, "learning_rate": 2.5072160766386345e-05, "loss": 0.0894, "step": 53640 }, { "epoch": 2.5, "learning_rate": 2.5071692696550328e-05, "loss": 0.093, "step": 53645 }, { "epoch": 2.5, "learning_rate": 2.5071224626714308e-05, "loss": 0.2569, "step": 53650 }, { "epoch": 2.5, "learning_rate": 2.5070756556878284e-05, "loss": 0.2252, "step": 53655 }, { "epoch": 2.5, "learning_rate": 2.5070288487042268e-05, "loss": 0.2512, "step": 53660 }, { "epoch": 2.5, "learning_rate": 2.5069820417206247e-05, "loss": 0.2363, "step": 53665 }, { "epoch": 2.5, "learning_rate": 2.5069352347370227e-05, "loss": 0.1054, "step": 53670 }, { "epoch": 2.5, "learning_rate": 2.5068884277534207e-05, "loss": 0.0431, "step": 53675 }, { "epoch": 2.5, "learning_rate": 2.506841620769819e-05, "loss": 0.0788, "step": 53680 }, { "epoch": 2.51, "learning_rate": 2.506794813786217e-05, "loss": 0.0512, "step": 53685 }, { "epoch": 2.51, "learning_rate": 2.506748006802615e-05, "loss": 0.0916, "step": 53690 }, { "epoch": 2.51, "learning_rate": 2.506701199819013e-05, "loss": 0.0755, "step": 53695 }, { "epoch": 2.51, "learning_rate": 2.5066543928354113e-05, "loss": 0.1341, "step": 53700 }, { "epoch": 2.51, "learning_rate": 2.5066075858518093e-05, "loss": 0.1662, "step": 53705 }, { "epoch": 2.51, "learning_rate": 2.5065607788682073e-05, "loss": 0.1018, "step": 53710 }, { "epoch": 2.51, "learning_rate": 2.5065139718846056e-05, "loss": 0.4115, "step": 53715 }, { "epoch": 2.51, "learning_rate": 2.5064671649010032e-05, "loss": 0.0496, "step": 53720 }, { "epoch": 2.51, "learning_rate": 2.5064203579174012e-05, "loss": 0.0623, "step": 53725 }, { "epoch": 2.51, "learning_rate": 2.5063735509337992e-05, "loss": 0.034, "step": 53730 }, { "epoch": 2.51, "learning_rate": 2.5063267439501975e-05, "loss": 0.0329, "step": 53735 }, { "epoch": 2.51, "learning_rate": 2.5062799369665955e-05, "loss": 0.0754, "step": 53740 }, { "epoch": 2.51, "learning_rate": 2.5062331299829935e-05, "loss": 0.0921, "step": 53745 }, { "epoch": 2.51, "learning_rate": 2.5061863229993914e-05, "loss": 0.1461, "step": 53750 }, { "epoch": 2.51, "learning_rate": 2.5061395160157898e-05, "loss": 0.1212, "step": 53755 }, { "epoch": 2.51, "learning_rate": 2.5060927090321877e-05, "loss": 0.23, "step": 53760 }, { "epoch": 2.51, "learning_rate": 2.5060459020485857e-05, "loss": 0.257, "step": 53765 }, { "epoch": 2.51, "learning_rate": 2.505999095064984e-05, "loss": 0.0439, "step": 53770 }, { "epoch": 2.51, "learning_rate": 2.505952288081382e-05, "loss": 0.0298, "step": 53775 }, { "epoch": 2.51, "learning_rate": 2.5059054810977797e-05, "loss": 0.0593, "step": 53780 }, { "epoch": 2.51, "learning_rate": 2.5058586741141777e-05, "loss": 0.0885, "step": 53785 }, { "epoch": 2.51, "learning_rate": 2.505811867130576e-05, "loss": 0.0743, "step": 53790 }, { "epoch": 2.51, "learning_rate": 2.505765060146974e-05, "loss": 0.1249, "step": 53795 }, { "epoch": 2.51, "learning_rate": 2.505718253163372e-05, "loss": 0.0854, "step": 53800 }, { "epoch": 2.51, "learning_rate": 2.50567144617977e-05, "loss": 0.1943, "step": 53805 }, { "epoch": 2.51, "learning_rate": 2.5056246391961682e-05, "loss": 0.1837, "step": 53810 }, { "epoch": 2.51, "learning_rate": 2.5055778322125662e-05, "loss": 0.2483, "step": 53815 }, { "epoch": 2.51, "learning_rate": 2.5055310252289642e-05, "loss": 0.0288, "step": 53820 }, { "epoch": 2.51, "learning_rate": 2.5054842182453622e-05, "loss": 0.0789, "step": 53825 }, { "epoch": 2.51, "learning_rate": 2.5054374112617605e-05, "loss": 0.079, "step": 53830 }, { "epoch": 2.51, "learning_rate": 2.5053906042781585e-05, "loss": 0.0805, "step": 53835 }, { "epoch": 2.51, "learning_rate": 2.5053437972945565e-05, "loss": 0.1404, "step": 53840 }, { "epoch": 2.51, "learning_rate": 2.5052969903109545e-05, "loss": 0.1767, "step": 53845 }, { "epoch": 2.51, "learning_rate": 2.5052501833273524e-05, "loss": 0.0998, "step": 53850 }, { "epoch": 2.51, "learning_rate": 2.5052033763437504e-05, "loss": 0.1828, "step": 53855 }, { "epoch": 2.51, "learning_rate": 2.5051565693601484e-05, "loss": 0.32, "step": 53860 }, { "epoch": 2.51, "learning_rate": 2.5051097623765467e-05, "loss": 0.3163, "step": 53865 }, { "epoch": 2.51, "learning_rate": 2.5050629553929447e-05, "loss": 0.1114, "step": 53870 }, { "epoch": 2.51, "learning_rate": 2.5050161484093427e-05, "loss": 0.0232, "step": 53875 }, { "epoch": 2.51, "learning_rate": 2.5049693414257407e-05, "loss": 0.0577, "step": 53880 }, { "epoch": 2.51, "learning_rate": 2.504922534442139e-05, "loss": 0.0678, "step": 53885 }, { "epoch": 2.51, "learning_rate": 2.504875727458537e-05, "loss": 0.0459, "step": 53890 }, { "epoch": 2.51, "learning_rate": 2.504828920474935e-05, "loss": 0.1759, "step": 53895 }, { "epoch": 2.52, "learning_rate": 2.5047821134913333e-05, "loss": 0.1275, "step": 53900 }, { "epoch": 2.52, "learning_rate": 2.5047353065077313e-05, "loss": 0.1476, "step": 53905 }, { "epoch": 2.52, "learning_rate": 2.504688499524129e-05, "loss": 0.2772, "step": 53910 }, { "epoch": 2.52, "learning_rate": 2.504641692540527e-05, "loss": 0.2799, "step": 53915 }, { "epoch": 2.52, "learning_rate": 2.5045948855569252e-05, "loss": 0.0564, "step": 53920 }, { "epoch": 2.52, "learning_rate": 2.5045480785733232e-05, "loss": 0.0162, "step": 53925 }, { "epoch": 2.52, "learning_rate": 2.504501271589721e-05, "loss": 0.0571, "step": 53930 }, { "epoch": 2.52, "learning_rate": 2.504454464606119e-05, "loss": 0.1392, "step": 53935 }, { "epoch": 2.52, "learning_rate": 2.5044076576225175e-05, "loss": 0.0791, "step": 53940 }, { "epoch": 2.52, "learning_rate": 2.5043608506389154e-05, "loss": 0.0467, "step": 53945 }, { "epoch": 2.52, "learning_rate": 2.5043140436553134e-05, "loss": 0.1503, "step": 53950 }, { "epoch": 2.52, "learning_rate": 2.5042672366717117e-05, "loss": 0.1876, "step": 53955 }, { "epoch": 2.52, "learning_rate": 2.5042204296881097e-05, "loss": 0.2658, "step": 53960 }, { "epoch": 2.52, "learning_rate": 2.5041736227045077e-05, "loss": 0.2373, "step": 53965 }, { "epoch": 2.52, "learning_rate": 2.5041268157209054e-05, "loss": 0.0701, "step": 53970 }, { "epoch": 2.52, "learning_rate": 2.5040800087373037e-05, "loss": 0.0436, "step": 53975 }, { "epoch": 2.52, "learning_rate": 2.5040332017537017e-05, "loss": 0.0613, "step": 53980 }, { "epoch": 2.52, "learning_rate": 2.5039863947700996e-05, "loss": 0.04, "step": 53985 }, { "epoch": 2.52, "learning_rate": 2.5039395877864976e-05, "loss": 0.2007, "step": 53990 }, { "epoch": 2.52, "learning_rate": 2.503892780802896e-05, "loss": 0.0659, "step": 53995 }, { "epoch": 2.52, "learning_rate": 2.503845973819294e-05, "loss": 0.1564, "step": 54000 }, { "epoch": 2.52, "learning_rate": 2.503799166835692e-05, "loss": 0.0791, "step": 54005 }, { "epoch": 2.52, "learning_rate": 2.50375235985209e-05, "loss": 0.1994, "step": 54010 }, { "epoch": 2.52, "learning_rate": 2.5037055528684882e-05, "loss": 0.255, "step": 54015 }, { "epoch": 2.52, "learning_rate": 2.5036587458848862e-05, "loss": 0.045, "step": 54020 }, { "epoch": 2.52, "learning_rate": 2.503611938901284e-05, "loss": 0.0832, "step": 54025 }, { "epoch": 2.52, "learning_rate": 2.5035651319176825e-05, "loss": 0.0536, "step": 54030 }, { "epoch": 2.52, "learning_rate": 2.50351832493408e-05, "loss": 0.0861, "step": 54035 }, { "epoch": 2.52, "learning_rate": 2.503471517950478e-05, "loss": 0.0867, "step": 54040 }, { "epoch": 2.52, "learning_rate": 2.503424710966876e-05, "loss": 0.0945, "step": 54045 }, { "epoch": 2.52, "learning_rate": 2.5033779039832744e-05, "loss": 0.0858, "step": 54050 }, { "epoch": 2.52, "learning_rate": 2.5033310969996724e-05, "loss": 0.1485, "step": 54055 }, { "epoch": 2.52, "learning_rate": 2.5032842900160704e-05, "loss": 0.1836, "step": 54060 }, { "epoch": 2.52, "learning_rate": 2.5032374830324684e-05, "loss": 0.2405, "step": 54065 }, { "epoch": 2.52, "learning_rate": 2.5031906760488667e-05, "loss": 0.0285, "step": 54070 }, { "epoch": 2.52, "learning_rate": 2.5031438690652647e-05, "loss": 0.0372, "step": 54075 }, { "epoch": 2.52, "learning_rate": 2.5030970620816626e-05, "loss": 0.0403, "step": 54080 }, { "epoch": 2.52, "learning_rate": 2.503050255098061e-05, "loss": 0.0616, "step": 54085 }, { "epoch": 2.52, "learning_rate": 2.503003448114459e-05, "loss": 0.101, "step": 54090 }, { "epoch": 2.52, "learning_rate": 2.502956641130857e-05, "loss": 0.0591, "step": 54095 }, { "epoch": 2.52, "learning_rate": 2.5029098341472546e-05, "loss": 0.0808, "step": 54100 }, { "epoch": 2.52, "learning_rate": 2.502863027163653e-05, "loss": 0.1996, "step": 54105 }, { "epoch": 2.52, "learning_rate": 2.502816220180051e-05, "loss": 0.1673, "step": 54110 }, { "epoch": 2.53, "learning_rate": 2.502769413196449e-05, "loss": 0.2462, "step": 54115 }, { "epoch": 2.53, "learning_rate": 2.502722606212847e-05, "loss": 0.0194, "step": 54120 }, { "epoch": 2.53, "learning_rate": 2.502675799229245e-05, "loss": 0.0507, "step": 54125 }, { "epoch": 2.53, "learning_rate": 2.502628992245643e-05, "loss": 0.1076, "step": 54130 }, { "epoch": 2.53, "learning_rate": 2.502582185262041e-05, "loss": 0.0711, "step": 54135 }, { "epoch": 2.53, "learning_rate": 2.5025353782784394e-05, "loss": 0.0609, "step": 54140 }, { "epoch": 2.53, "learning_rate": 2.5024885712948374e-05, "loss": 0.1024, "step": 54145 }, { "epoch": 2.53, "learning_rate": 2.5024417643112354e-05, "loss": 0.0734, "step": 54150 }, { "epoch": 2.53, "learning_rate": 2.5023949573276334e-05, "loss": 0.1106, "step": 54155 }, { "epoch": 2.53, "learning_rate": 2.5023481503440314e-05, "loss": 0.1954, "step": 54160 }, { "epoch": 2.53, "learning_rate": 2.5023013433604294e-05, "loss": 0.3085, "step": 54165 }, { "epoch": 2.53, "learning_rate": 2.5022545363768273e-05, "loss": 0.0695, "step": 54170 }, { "epoch": 2.53, "learning_rate": 2.5022077293932253e-05, "loss": 0.0673, "step": 54175 }, { "epoch": 2.53, "learning_rate": 2.5021609224096236e-05, "loss": 0.0393, "step": 54180 }, { "epoch": 2.53, "learning_rate": 2.5021141154260216e-05, "loss": 0.0529, "step": 54185 }, { "epoch": 2.53, "learning_rate": 2.5020673084424196e-05, "loss": 0.1031, "step": 54190 }, { "epoch": 2.53, "learning_rate": 2.5020205014588176e-05, "loss": 0.0893, "step": 54195 }, { "epoch": 2.53, "learning_rate": 2.501973694475216e-05, "loss": 0.246, "step": 54200 }, { "epoch": 2.53, "learning_rate": 2.501926887491614e-05, "loss": 0.2143, "step": 54205 }, { "epoch": 2.53, "learning_rate": 2.501880080508012e-05, "loss": 0.2443, "step": 54210 }, { "epoch": 2.53, "learning_rate": 2.5018332735244102e-05, "loss": 0.4272, "step": 54215 }, { "epoch": 2.53, "learning_rate": 2.501786466540808e-05, "loss": 0.0986, "step": 54220 }, { "epoch": 2.53, "learning_rate": 2.5017396595572058e-05, "loss": 0.0217, "step": 54225 }, { "epoch": 2.53, "learning_rate": 2.5016928525736038e-05, "loss": 0.0281, "step": 54230 }, { "epoch": 2.53, "learning_rate": 2.501646045590002e-05, "loss": 0.0641, "step": 54235 }, { "epoch": 2.53, "learning_rate": 2.5015992386064e-05, "loss": 0.0824, "step": 54240 }, { "epoch": 2.53, "learning_rate": 2.501552431622798e-05, "loss": 0.0702, "step": 54245 }, { "epoch": 2.53, "learning_rate": 2.501505624639196e-05, "loss": 0.107, "step": 54250 }, { "epoch": 2.53, "learning_rate": 2.5014588176555944e-05, "loss": 0.0983, "step": 54255 }, { "epoch": 2.53, "learning_rate": 2.5014120106719924e-05, "loss": 0.1499, "step": 54260 }, { "epoch": 2.53, "learning_rate": 2.5013652036883903e-05, "loss": 0.2461, "step": 54265 }, { "epoch": 2.53, "learning_rate": 2.5013183967047887e-05, "loss": 0.0382, "step": 54270 }, { "epoch": 2.53, "learning_rate": 2.5012715897211866e-05, "loss": 0.0754, "step": 54275 }, { "epoch": 2.53, "learning_rate": 2.5012247827375846e-05, "loss": 0.0482, "step": 54280 }, { "epoch": 2.53, "learning_rate": 2.5011779757539823e-05, "loss": 0.0496, "step": 54285 }, { "epoch": 2.53, "learning_rate": 2.5011311687703806e-05, "loss": 0.0858, "step": 54290 }, { "epoch": 2.53, "learning_rate": 2.5010843617867786e-05, "loss": 0.0985, "step": 54295 }, { "epoch": 2.53, "learning_rate": 2.5010375548031766e-05, "loss": 0.1118, "step": 54300 }, { "epoch": 2.53, "learning_rate": 2.5009907478195745e-05, "loss": 0.1143, "step": 54305 }, { "epoch": 2.53, "learning_rate": 2.500943940835973e-05, "loss": 0.1826, "step": 54310 }, { "epoch": 2.53, "learning_rate": 2.500897133852371e-05, "loss": 0.2129, "step": 54315 }, { "epoch": 2.53, "learning_rate": 2.5008503268687688e-05, "loss": 0.0589, "step": 54320 }, { "epoch": 2.53, "learning_rate": 2.500803519885167e-05, "loss": 0.041, "step": 54325 }, { "epoch": 2.54, "learning_rate": 2.500756712901565e-05, "loss": 0.0422, "step": 54330 }, { "epoch": 2.54, "learning_rate": 2.500709905917963e-05, "loss": 0.0874, "step": 54335 }, { "epoch": 2.54, "learning_rate": 2.500663098934361e-05, "loss": 0.1185, "step": 54340 }, { "epoch": 2.54, "learning_rate": 2.5006162919507594e-05, "loss": 0.0856, "step": 54345 }, { "epoch": 2.54, "learning_rate": 2.500569484967157e-05, "loss": 0.0997, "step": 54350 }, { "epoch": 2.54, "learning_rate": 2.500522677983555e-05, "loss": 0.1972, "step": 54355 }, { "epoch": 2.54, "learning_rate": 2.500475870999953e-05, "loss": 0.1929, "step": 54360 }, { "epoch": 2.54, "learning_rate": 2.5004290640163513e-05, "loss": 0.3079, "step": 54365 }, { "epoch": 2.54, "learning_rate": 2.5003822570327493e-05, "loss": 0.0272, "step": 54370 }, { "epoch": 2.54, "learning_rate": 2.5003354500491473e-05, "loss": 0.0555, "step": 54375 }, { "epoch": 2.54, "learning_rate": 2.5002886430655453e-05, "loss": 0.071, "step": 54380 }, { "epoch": 2.54, "learning_rate": 2.5002418360819436e-05, "loss": 0.0784, "step": 54385 }, { "epoch": 2.54, "learning_rate": 2.5001950290983416e-05, "loss": 0.091, "step": 54390 }, { "epoch": 2.54, "learning_rate": 2.5001482221147396e-05, "loss": 0.0573, "step": 54395 }, { "epoch": 2.54, "learning_rate": 2.500101415131138e-05, "loss": 0.1737, "step": 54400 }, { "epoch": 2.54, "learning_rate": 2.500054608147536e-05, "loss": 0.148, "step": 54405 }, { "epoch": 2.54, "learning_rate": 2.500007801163934e-05, "loss": 0.2922, "step": 54410 }, { "epoch": 2.54, "learning_rate": 2.4999609941803315e-05, "loss": 0.2347, "step": 54415 }, { "epoch": 2.54, "learning_rate": 2.4999141871967298e-05, "loss": 0.044, "step": 54420 }, { "epoch": 2.54, "learning_rate": 2.4998673802131278e-05, "loss": 0.0217, "step": 54425 }, { "epoch": 2.54, "learning_rate": 2.4998205732295258e-05, "loss": 0.0272, "step": 54430 }, { "epoch": 2.54, "learning_rate": 2.4997737662459238e-05, "loss": 0.0377, "step": 54435 }, { "epoch": 2.54, "learning_rate": 2.499726959262322e-05, "loss": 0.0569, "step": 54440 }, { "epoch": 2.54, "learning_rate": 2.49968015227872e-05, "loss": 0.1419, "step": 54445 }, { "epoch": 2.54, "learning_rate": 2.499633345295118e-05, "loss": 0.1619, "step": 54450 }, { "epoch": 2.54, "learning_rate": 2.4995865383115164e-05, "loss": 0.1279, "step": 54455 }, { "epoch": 2.54, "learning_rate": 2.4995397313279143e-05, "loss": 0.1513, "step": 54460 }, { "epoch": 2.54, "learning_rate": 2.4994929243443123e-05, "loss": 0.3358, "step": 54465 }, { "epoch": 2.54, "learning_rate": 2.4994461173607103e-05, "loss": 0.0211, "step": 54470 }, { "epoch": 2.54, "learning_rate": 2.4993993103771083e-05, "loss": 0.0362, "step": 54475 }, { "epoch": 2.54, "learning_rate": 2.4993525033935063e-05, "loss": 0.0855, "step": 54480 }, { "epoch": 2.54, "learning_rate": 2.4993056964099043e-05, "loss": 0.0433, "step": 54485 }, { "epoch": 2.54, "learning_rate": 2.4992588894263022e-05, "loss": 0.0331, "step": 54490 }, { "epoch": 2.54, "learning_rate": 2.4992120824427006e-05, "loss": 0.1079, "step": 54495 }, { "epoch": 2.54, "learning_rate": 2.4991652754590985e-05, "loss": 0.2036, "step": 54500 }, { "epoch": 2.54, "learning_rate": 2.4991184684754965e-05, "loss": 0.1188, "step": 54505 }, { "epoch": 2.54, "learning_rate": 2.499071661491895e-05, "loss": 0.2893, "step": 54510 }, { "epoch": 2.54, "learning_rate": 2.4990248545082928e-05, "loss": 0.3575, "step": 54515 }, { "epoch": 2.54, "learning_rate": 2.4989780475246908e-05, "loss": 0.0312, "step": 54520 }, { "epoch": 2.54, "learning_rate": 2.4989312405410888e-05, "loss": 0.0433, "step": 54525 }, { "epoch": 2.54, "learning_rate": 2.498884433557487e-05, "loss": 0.0632, "step": 54530 }, { "epoch": 2.54, "learning_rate": 2.498837626573885e-05, "loss": 0.1006, "step": 54535 }, { "epoch": 2.54, "learning_rate": 2.4987908195902827e-05, "loss": 0.0779, "step": 54540 }, { "epoch": 2.55, "learning_rate": 2.4987440126066807e-05, "loss": 0.1326, "step": 54545 }, { "epoch": 2.55, "learning_rate": 2.498697205623079e-05, "loss": 0.1437, "step": 54550 }, { "epoch": 2.55, "learning_rate": 2.498650398639477e-05, "loss": 0.1954, "step": 54555 }, { "epoch": 2.55, "learning_rate": 2.498603591655875e-05, "loss": 0.3787, "step": 54560 }, { "epoch": 2.55, "learning_rate": 2.4985567846722733e-05, "loss": 0.1898, "step": 54565 }, { "epoch": 2.55, "learning_rate": 2.4985099776886713e-05, "loss": 0.0355, "step": 54570 }, { "epoch": 2.55, "learning_rate": 2.4984631707050693e-05, "loss": 0.0389, "step": 54575 }, { "epoch": 2.55, "learning_rate": 2.4984163637214673e-05, "loss": 0.1168, "step": 54580 }, { "epoch": 2.55, "learning_rate": 2.4983695567378656e-05, "loss": 0.0968, "step": 54585 }, { "epoch": 2.55, "learning_rate": 2.4983227497542636e-05, "loss": 0.0676, "step": 54590 }, { "epoch": 2.55, "learning_rate": 2.4982759427706615e-05, "loss": 0.0811, "step": 54595 }, { "epoch": 2.55, "learning_rate": 2.4982291357870595e-05, "loss": 0.1514, "step": 54600 }, { "epoch": 2.55, "learning_rate": 2.4981823288034575e-05, "loss": 0.1824, "step": 54605 }, { "epoch": 2.55, "learning_rate": 2.4981355218198555e-05, "loss": 0.2406, "step": 54610 }, { "epoch": 2.55, "learning_rate": 2.4980887148362535e-05, "loss": 0.3005, "step": 54615 }, { "epoch": 2.55, "learning_rate": 2.4980419078526515e-05, "loss": 0.0663, "step": 54620 }, { "epoch": 2.55, "learning_rate": 2.4979951008690498e-05, "loss": 0.036, "step": 54625 }, { "epoch": 2.55, "learning_rate": 2.4979482938854478e-05, "loss": 0.0223, "step": 54630 }, { "epoch": 2.55, "learning_rate": 2.4979014869018457e-05, "loss": 0.0473, "step": 54635 }, { "epoch": 2.55, "learning_rate": 2.497854679918244e-05, "loss": 0.0789, "step": 54640 }, { "epoch": 2.55, "learning_rate": 2.497807872934642e-05, "loss": 0.1912, "step": 54645 }, { "epoch": 2.55, "learning_rate": 2.49776106595104e-05, "loss": 0.1911, "step": 54650 }, { "epoch": 2.55, "learning_rate": 2.497714258967438e-05, "loss": 0.185, "step": 54655 }, { "epoch": 2.55, "learning_rate": 2.4976674519838363e-05, "loss": 0.2641, "step": 54660 }, { "epoch": 2.55, "learning_rate": 2.497620645000234e-05, "loss": 0.1693, "step": 54665 }, { "epoch": 2.55, "learning_rate": 2.497573838016632e-05, "loss": 0.034, "step": 54670 }, { "epoch": 2.55, "learning_rate": 2.49752703103303e-05, "loss": 0.0635, "step": 54675 }, { "epoch": 2.55, "learning_rate": 2.4974802240494282e-05, "loss": 0.0585, "step": 54680 }, { "epoch": 2.55, "learning_rate": 2.4974334170658262e-05, "loss": 0.0314, "step": 54685 }, { "epoch": 2.55, "learning_rate": 2.4973866100822242e-05, "loss": 0.0966, "step": 54690 }, { "epoch": 2.55, "learning_rate": 2.4973398030986225e-05, "loss": 0.0917, "step": 54695 }, { "epoch": 2.55, "learning_rate": 2.4972929961150205e-05, "loss": 0.1054, "step": 54700 }, { "epoch": 2.55, "learning_rate": 2.4972461891314185e-05, "loss": 0.166, "step": 54705 }, { "epoch": 2.55, "learning_rate": 2.4971993821478165e-05, "loss": 0.1697, "step": 54710 }, { "epoch": 2.55, "learning_rate": 2.4971525751642148e-05, "loss": 0.2646, "step": 54715 }, { "epoch": 2.55, "learning_rate": 2.4971057681806128e-05, "loss": 0.074, "step": 54720 }, { "epoch": 2.55, "learning_rate": 2.4970589611970108e-05, "loss": 0.0266, "step": 54725 }, { "epoch": 2.55, "learning_rate": 2.4970121542134084e-05, "loss": 0.0337, "step": 54730 }, { "epoch": 2.55, "learning_rate": 2.4969653472298067e-05, "loss": 0.0613, "step": 54735 }, { "epoch": 2.55, "learning_rate": 2.4969185402462047e-05, "loss": 0.1644, "step": 54740 }, { "epoch": 2.55, "learning_rate": 2.4968717332626027e-05, "loss": 0.075, "step": 54745 }, { "epoch": 2.55, "learning_rate": 2.496824926279001e-05, "loss": 0.0901, "step": 54750 }, { "epoch": 2.55, "learning_rate": 2.496778119295399e-05, "loss": 0.145, "step": 54755 }, { "epoch": 2.56, "learning_rate": 2.496731312311797e-05, "loss": 0.1835, "step": 54760 }, { "epoch": 2.56, "learning_rate": 2.496684505328195e-05, "loss": 0.3753, "step": 54765 }, { "epoch": 2.56, "learning_rate": 2.4966376983445933e-05, "loss": 0.0463, "step": 54770 }, { "epoch": 2.56, "learning_rate": 2.4965908913609913e-05, "loss": 0.022, "step": 54775 }, { "epoch": 2.56, "learning_rate": 2.4965440843773892e-05, "loss": 0.0587, "step": 54780 }, { "epoch": 2.56, "learning_rate": 2.4964972773937872e-05, "loss": 0.0567, "step": 54785 }, { "epoch": 2.56, "learning_rate": 2.4964504704101852e-05, "loss": 0.1246, "step": 54790 }, { "epoch": 2.56, "learning_rate": 2.4964036634265832e-05, "loss": 0.1212, "step": 54795 }, { "epoch": 2.56, "learning_rate": 2.496356856442981e-05, "loss": 0.2095, "step": 54800 }, { "epoch": 2.56, "learning_rate": 2.496310049459379e-05, "loss": 0.1394, "step": 54805 }, { "epoch": 2.56, "learning_rate": 2.4962632424757775e-05, "loss": 0.2123, "step": 54810 }, { "epoch": 2.56, "learning_rate": 2.4962164354921755e-05, "loss": 0.1837, "step": 54815 }, { "epoch": 2.56, "learning_rate": 2.4961696285085734e-05, "loss": 0.0484, "step": 54820 }, { "epoch": 2.56, "learning_rate": 2.4961228215249718e-05, "loss": 0.0366, "step": 54825 }, { "epoch": 2.56, "learning_rate": 2.4960760145413697e-05, "loss": 0.0379, "step": 54830 }, { "epoch": 2.56, "learning_rate": 2.4960292075577677e-05, "loss": 0.0851, "step": 54835 }, { "epoch": 2.56, "learning_rate": 2.4959824005741657e-05, "loss": 0.0592, "step": 54840 }, { "epoch": 2.56, "learning_rate": 2.495935593590564e-05, "loss": 0.1129, "step": 54845 }, { "epoch": 2.56, "learning_rate": 2.495888786606962e-05, "loss": 0.1563, "step": 54850 }, { "epoch": 2.56, "learning_rate": 2.4958419796233596e-05, "loss": 0.1945, "step": 54855 }, { "epoch": 2.56, "learning_rate": 2.4957951726397576e-05, "loss": 0.1793, "step": 54860 }, { "epoch": 2.56, "learning_rate": 2.495748365656156e-05, "loss": 0.2664, "step": 54865 }, { "epoch": 2.56, "learning_rate": 2.495701558672554e-05, "loss": 0.039, "step": 54870 }, { "epoch": 2.56, "learning_rate": 2.495654751688952e-05, "loss": 0.0367, "step": 54875 }, { "epoch": 2.56, "learning_rate": 2.4956079447053502e-05, "loss": 0.0369, "step": 54880 }, { "epoch": 2.56, "learning_rate": 2.4955611377217482e-05, "loss": 0.0672, "step": 54885 }, { "epoch": 2.56, "learning_rate": 2.4955143307381462e-05, "loss": 0.0556, "step": 54890 }, { "epoch": 2.56, "learning_rate": 2.4954675237545442e-05, "loss": 0.1584, "step": 54895 }, { "epoch": 2.56, "learning_rate": 2.4954207167709425e-05, "loss": 0.0898, "step": 54900 }, { "epoch": 2.56, "learning_rate": 2.4953739097873405e-05, "loss": 0.1182, "step": 54905 }, { "epoch": 2.56, "learning_rate": 2.4953271028037385e-05, "loss": 0.1661, "step": 54910 }, { "epoch": 2.56, "learning_rate": 2.4952802958201364e-05, "loss": 0.1882, "step": 54915 }, { "epoch": 2.56, "learning_rate": 2.4952334888365344e-05, "loss": 0.0429, "step": 54920 }, { "epoch": 2.56, "learning_rate": 2.4951866818529324e-05, "loss": 0.0296, "step": 54925 }, { "epoch": 2.56, "learning_rate": 2.4951398748693304e-05, "loss": 0.0698, "step": 54930 }, { "epoch": 2.56, "learning_rate": 2.4950930678857287e-05, "loss": 0.0282, "step": 54935 }, { "epoch": 2.56, "learning_rate": 2.4950462609021267e-05, "loss": 0.0385, "step": 54940 }, { "epoch": 2.56, "learning_rate": 2.4949994539185247e-05, "loss": 0.0697, "step": 54945 }, { "epoch": 2.56, "learning_rate": 2.4949526469349227e-05, "loss": 0.0864, "step": 54950 }, { "epoch": 2.56, "learning_rate": 2.494905839951321e-05, "loss": 0.2991, "step": 54955 }, { "epoch": 2.56, "learning_rate": 2.494859032967719e-05, "loss": 0.3339, "step": 54960 }, { "epoch": 2.56, "learning_rate": 2.494812225984117e-05, "loss": 0.3374, "step": 54965 }, { "epoch": 2.56, "learning_rate": 2.494765419000515e-05, "loss": 0.0754, "step": 54970 }, { "epoch": 2.57, "learning_rate": 2.4947186120169132e-05, "loss": 0.0798, "step": 54975 }, { "epoch": 2.57, "learning_rate": 2.494671805033311e-05, "loss": 0.0414, "step": 54980 }, { "epoch": 2.57, "learning_rate": 2.494624998049709e-05, "loss": 0.0265, "step": 54985 }, { "epoch": 2.57, "learning_rate": 2.494578191066107e-05, "loss": 0.0548, "step": 54990 }, { "epoch": 2.57, "learning_rate": 2.494531384082505e-05, "loss": 0.0758, "step": 54995 }, { "epoch": 2.57, "learning_rate": 2.494484577098903e-05, "loss": 0.0983, "step": 55000 }, { "epoch": 2.57, "learning_rate": 2.494437770115301e-05, "loss": 0.1379, "step": 55005 }, { "epoch": 2.57, "learning_rate": 2.4943909631316995e-05, "loss": 0.2312, "step": 55010 }, { "epoch": 2.57, "learning_rate": 2.4943441561480974e-05, "loss": 0.2212, "step": 55015 }, { "epoch": 2.57, "learning_rate": 2.4942973491644954e-05, "loss": 0.0245, "step": 55020 }, { "epoch": 2.57, "learning_rate": 2.4942505421808934e-05, "loss": 0.0547, "step": 55025 }, { "epoch": 2.57, "learning_rate": 2.4942037351972917e-05, "loss": 0.0427, "step": 55030 }, { "epoch": 2.57, "learning_rate": 2.4941569282136897e-05, "loss": 0.0377, "step": 55035 }, { "epoch": 2.57, "learning_rate": 2.4941101212300877e-05, "loss": 0.1138, "step": 55040 }, { "epoch": 2.57, "learning_rate": 2.4940633142464853e-05, "loss": 0.154, "step": 55045 }, { "epoch": 2.57, "learning_rate": 2.4940165072628836e-05, "loss": 0.1163, "step": 55050 }, { "epoch": 2.57, "learning_rate": 2.4939697002792816e-05, "loss": 0.2235, "step": 55055 }, { "epoch": 2.57, "learning_rate": 2.4939228932956796e-05, "loss": 0.2344, "step": 55060 }, { "epoch": 2.57, "learning_rate": 2.493876086312078e-05, "loss": 0.3719, "step": 55065 }, { "epoch": 2.57, "learning_rate": 2.493829279328476e-05, "loss": 0.0359, "step": 55070 }, { "epoch": 2.57, "learning_rate": 2.493782472344874e-05, "loss": 0.0035, "step": 55075 }, { "epoch": 2.57, "learning_rate": 2.493735665361272e-05, "loss": 0.0642, "step": 55080 }, { "epoch": 2.57, "learning_rate": 2.4936888583776702e-05, "loss": 0.0889, "step": 55085 }, { "epoch": 2.57, "learning_rate": 2.4936420513940682e-05, "loss": 0.1642, "step": 55090 }, { "epoch": 2.57, "learning_rate": 2.493595244410466e-05, "loss": 0.124, "step": 55095 }, { "epoch": 2.57, "learning_rate": 2.493548437426864e-05, "loss": 0.0738, "step": 55100 }, { "epoch": 2.57, "learning_rate": 2.4935016304432625e-05, "loss": 0.1238, "step": 55105 }, { "epoch": 2.57, "learning_rate": 2.49345482345966e-05, "loss": 0.3144, "step": 55110 }, { "epoch": 2.57, "learning_rate": 2.493408016476058e-05, "loss": 0.2044, "step": 55115 }, { "epoch": 2.57, "learning_rate": 2.4933612094924564e-05, "loss": 0.0599, "step": 55120 }, { "epoch": 2.57, "learning_rate": 2.4933144025088544e-05, "loss": 0.0402, "step": 55125 }, { "epoch": 2.57, "learning_rate": 2.4932675955252524e-05, "loss": 0.0926, "step": 55130 }, { "epoch": 2.57, "learning_rate": 2.4932207885416504e-05, "loss": 0.0623, "step": 55135 }, { "epoch": 2.57, "learning_rate": 2.4931739815580487e-05, "loss": 0.0852, "step": 55140 }, { "epoch": 2.57, "learning_rate": 2.4931271745744467e-05, "loss": 0.1275, "step": 55145 }, { "epoch": 2.57, "learning_rate": 2.4930803675908446e-05, "loss": 0.2069, "step": 55150 }, { "epoch": 2.57, "learning_rate": 2.4930335606072426e-05, "loss": 0.0938, "step": 55155 }, { "epoch": 2.57, "learning_rate": 2.492986753623641e-05, "loss": 0.2646, "step": 55160 }, { "epoch": 2.57, "learning_rate": 2.492939946640039e-05, "loss": 0.4099, "step": 55165 }, { "epoch": 2.57, "learning_rate": 2.4928931396564366e-05, "loss": 0.0682, "step": 55170 }, { "epoch": 2.57, "learning_rate": 2.492846332672835e-05, "loss": 0.017, "step": 55175 }, { "epoch": 2.57, "learning_rate": 2.492799525689233e-05, "loss": 0.0336, "step": 55180 }, { "epoch": 2.58, "learning_rate": 2.492752718705631e-05, "loss": 0.0702, "step": 55185 }, { "epoch": 2.58, "learning_rate": 2.4927059117220288e-05, "loss": 0.0996, "step": 55190 }, { "epoch": 2.58, "learning_rate": 2.492659104738427e-05, "loss": 0.0921, "step": 55195 }, { "epoch": 2.58, "learning_rate": 2.492612297754825e-05, "loss": 0.1747, "step": 55200 }, { "epoch": 2.58, "learning_rate": 2.492565490771223e-05, "loss": 0.0909, "step": 55205 }, { "epoch": 2.58, "learning_rate": 2.492518683787621e-05, "loss": 0.5925, "step": 55210 }, { "epoch": 2.58, "learning_rate": 2.4924718768040194e-05, "loss": 0.267, "step": 55215 }, { "epoch": 2.58, "learning_rate": 2.4924250698204174e-05, "loss": 0.0249, "step": 55220 }, { "epoch": 2.58, "learning_rate": 2.4923782628368154e-05, "loss": 0.0301, "step": 55225 }, { "epoch": 2.58, "learning_rate": 2.4923314558532134e-05, "loss": 0.0575, "step": 55230 }, { "epoch": 2.58, "learning_rate": 2.4922846488696113e-05, "loss": 0.0627, "step": 55235 }, { "epoch": 2.58, "learning_rate": 2.4922378418860093e-05, "loss": 0.0823, "step": 55240 }, { "epoch": 2.58, "learning_rate": 2.4921910349024073e-05, "loss": 0.0871, "step": 55245 }, { "epoch": 2.58, "learning_rate": 2.4921442279188056e-05, "loss": 0.1885, "step": 55250 }, { "epoch": 2.58, "learning_rate": 2.4920974209352036e-05, "loss": 0.1823, "step": 55255 }, { "epoch": 2.58, "learning_rate": 2.4920506139516016e-05, "loss": 0.1295, "step": 55260 }, { "epoch": 2.58, "learning_rate": 2.4920038069679996e-05, "loss": 0.2162, "step": 55265 }, { "epoch": 2.58, "learning_rate": 2.491956999984398e-05, "loss": 0.014, "step": 55270 }, { "epoch": 2.58, "learning_rate": 2.491910193000796e-05, "loss": 0.0814, "step": 55275 }, { "epoch": 2.58, "learning_rate": 2.491863386017194e-05, "loss": 0.0591, "step": 55280 }, { "epoch": 2.58, "learning_rate": 2.491816579033592e-05, "loss": 0.0482, "step": 55285 }, { "epoch": 2.58, "learning_rate": 2.49176977204999e-05, "loss": 0.0677, "step": 55290 }, { "epoch": 2.58, "learning_rate": 2.491722965066388e-05, "loss": 0.044, "step": 55295 }, { "epoch": 2.58, "learning_rate": 2.4916761580827858e-05, "loss": 0.0898, "step": 55300 }, { "epoch": 2.58, "learning_rate": 2.491629351099184e-05, "loss": 0.2449, "step": 55305 }, { "epoch": 2.58, "learning_rate": 2.491582544115582e-05, "loss": 0.3145, "step": 55310 }, { "epoch": 2.58, "learning_rate": 2.49153573713198e-05, "loss": 0.2186, "step": 55315 }, { "epoch": 2.58, "learning_rate": 2.491488930148378e-05, "loss": 0.0839, "step": 55320 }, { "epoch": 2.58, "learning_rate": 2.4914421231647764e-05, "loss": 0.0311, "step": 55325 }, { "epoch": 2.58, "learning_rate": 2.4913953161811743e-05, "loss": 0.0593, "step": 55330 }, { "epoch": 2.58, "learning_rate": 2.4913485091975723e-05, "loss": 0.044, "step": 55335 }, { "epoch": 2.58, "learning_rate": 2.4913017022139703e-05, "loss": 0.0976, "step": 55340 }, { "epoch": 2.58, "learning_rate": 2.4912548952303686e-05, "loss": 0.114, "step": 55345 }, { "epoch": 2.58, "learning_rate": 2.4912080882467666e-05, "loss": 0.1655, "step": 55350 }, { "epoch": 2.58, "learning_rate": 2.4911612812631646e-05, "loss": 0.2045, "step": 55355 }, { "epoch": 2.58, "learning_rate": 2.4911144742795626e-05, "loss": 0.1919, "step": 55360 }, { "epoch": 2.58, "learning_rate": 2.4910676672959606e-05, "loss": 0.24, "step": 55365 }, { "epoch": 2.58, "learning_rate": 2.4910208603123585e-05, "loss": 0.0574, "step": 55370 }, { "epoch": 2.58, "learning_rate": 2.4909740533287565e-05, "loss": 0.0477, "step": 55375 }, { "epoch": 2.58, "learning_rate": 2.490927246345155e-05, "loss": 0.0812, "step": 55380 }, { "epoch": 2.58, "learning_rate": 2.4908804393615528e-05, "loss": 0.0386, "step": 55385 }, { "epoch": 2.58, "learning_rate": 2.4908336323779508e-05, "loss": 0.1101, "step": 55390 }, { "epoch": 2.58, "learning_rate": 2.4907868253943488e-05, "loss": 0.096, "step": 55395 }, { "epoch": 2.59, "learning_rate": 2.490740018410747e-05, "loss": 0.0696, "step": 55400 }, { "epoch": 2.59, "learning_rate": 2.490693211427145e-05, "loss": 0.1879, "step": 55405 }, { "epoch": 2.59, "learning_rate": 2.490646404443543e-05, "loss": 0.4245, "step": 55410 }, { "epoch": 2.59, "learning_rate": 2.490599597459941e-05, "loss": 0.3266, "step": 55415 }, { "epoch": 2.59, "learning_rate": 2.4905527904763394e-05, "loss": 0.0176, "step": 55420 }, { "epoch": 2.59, "learning_rate": 2.490505983492737e-05, "loss": 0.0238, "step": 55425 }, { "epoch": 2.59, "learning_rate": 2.490459176509135e-05, "loss": 0.0475, "step": 55430 }, { "epoch": 2.59, "learning_rate": 2.4904123695255333e-05, "loss": 0.098, "step": 55435 }, { "epoch": 2.59, "learning_rate": 2.4903655625419313e-05, "loss": 0.1087, "step": 55440 }, { "epoch": 2.59, "learning_rate": 2.4903187555583293e-05, "loss": 0.0753, "step": 55445 }, { "epoch": 2.59, "learning_rate": 2.4902719485747273e-05, "loss": 0.1167, "step": 55450 }, { "epoch": 2.59, "learning_rate": 2.4902251415911256e-05, "loss": 0.0595, "step": 55455 }, { "epoch": 2.59, "learning_rate": 2.4901783346075236e-05, "loss": 0.1833, "step": 55460 }, { "epoch": 2.59, "learning_rate": 2.4901315276239216e-05, "loss": 0.3303, "step": 55465 }, { "epoch": 2.59, "learning_rate": 2.4900847206403195e-05, "loss": 0.0577, "step": 55470 }, { "epoch": 2.59, "learning_rate": 2.490037913656718e-05, "loss": 0.0685, "step": 55475 }, { "epoch": 2.59, "learning_rate": 2.489991106673116e-05, "loss": 0.0419, "step": 55480 }, { "epoch": 2.59, "learning_rate": 2.4899442996895138e-05, "loss": 0.0418, "step": 55485 }, { "epoch": 2.59, "learning_rate": 2.4898974927059118e-05, "loss": 0.1089, "step": 55490 }, { "epoch": 2.59, "learning_rate": 2.4898506857223098e-05, "loss": 0.1051, "step": 55495 }, { "epoch": 2.59, "learning_rate": 2.4898038787387078e-05, "loss": 0.1464, "step": 55500 }, { "epoch": 2.59, "learning_rate": 2.4897570717551057e-05, "loss": 0.1362, "step": 55505 }, { "epoch": 2.59, "learning_rate": 2.489710264771504e-05, "loss": 0.1921, "step": 55510 }, { "epoch": 2.59, "learning_rate": 2.489663457787902e-05, "loss": 0.1861, "step": 55515 }, { "epoch": 2.59, "learning_rate": 2.4896166508043e-05, "loss": 0.0843, "step": 55520 }, { "epoch": 2.59, "learning_rate": 2.489569843820698e-05, "loss": 0.0699, "step": 55525 }, { "epoch": 2.59, "learning_rate": 2.4895230368370963e-05, "loss": 0.0606, "step": 55530 }, { "epoch": 2.59, "learning_rate": 2.4894762298534943e-05, "loss": 0.0675, "step": 55535 }, { "epoch": 2.59, "learning_rate": 2.4894294228698923e-05, "loss": 0.1192, "step": 55540 }, { "epoch": 2.59, "learning_rate": 2.4893826158862906e-05, "loss": 0.0594, "step": 55545 }, { "epoch": 2.59, "learning_rate": 2.4893358089026883e-05, "loss": 0.1487, "step": 55550 }, { "epoch": 2.59, "learning_rate": 2.4892890019190862e-05, "loss": 0.1876, "step": 55555 }, { "epoch": 2.59, "learning_rate": 2.4892421949354842e-05, "loss": 0.2431, "step": 55560 }, { "epoch": 2.59, "learning_rate": 2.4891953879518825e-05, "loss": 0.1864, "step": 55565 }, { "epoch": 2.59, "learning_rate": 2.4891485809682805e-05, "loss": 0.0444, "step": 55570 }, { "epoch": 2.59, "learning_rate": 2.4891017739846785e-05, "loss": 0.0379, "step": 55575 }, { "epoch": 2.59, "learning_rate": 2.4890549670010765e-05, "loss": 0.0577, "step": 55580 }, { "epoch": 2.59, "learning_rate": 2.4890081600174748e-05, "loss": 0.0597, "step": 55585 }, { "epoch": 2.59, "learning_rate": 2.4889613530338728e-05, "loss": 0.0855, "step": 55590 }, { "epoch": 2.59, "learning_rate": 2.4889145460502708e-05, "loss": 0.0489, "step": 55595 }, { "epoch": 2.59, "learning_rate": 2.4888677390666688e-05, "loss": 0.0751, "step": 55600 }, { "epoch": 2.59, "learning_rate": 2.488820932083067e-05, "loss": 0.2856, "step": 55605 }, { "epoch": 2.59, "learning_rate": 2.488774125099465e-05, "loss": 0.236, "step": 55610 }, { "epoch": 2.6, "learning_rate": 2.4887273181158627e-05, "loss": 0.2413, "step": 55615 }, { "epoch": 2.6, "learning_rate": 2.488680511132261e-05, "loss": 0.0216, "step": 55620 }, { "epoch": 2.6, "learning_rate": 2.488633704148659e-05, "loss": 0.0681, "step": 55625 }, { "epoch": 2.6, "learning_rate": 2.488586897165057e-05, "loss": 0.0589, "step": 55630 }, { "epoch": 2.6, "learning_rate": 2.488540090181455e-05, "loss": 0.0702, "step": 55635 }, { "epoch": 2.6, "learning_rate": 2.4884932831978533e-05, "loss": 0.0506, "step": 55640 }, { "epoch": 2.6, "learning_rate": 2.4884464762142513e-05, "loss": 0.0524, "step": 55645 }, { "epoch": 2.6, "learning_rate": 2.4883996692306492e-05, "loss": 0.2176, "step": 55650 }, { "epoch": 2.6, "learning_rate": 2.4883528622470472e-05, "loss": 0.1403, "step": 55655 }, { "epoch": 2.6, "learning_rate": 2.4883060552634456e-05, "loss": 0.2125, "step": 55660 }, { "epoch": 2.6, "learning_rate": 2.4882592482798435e-05, "loss": 0.3621, "step": 55665 }, { "epoch": 2.6, "learning_rate": 2.4882124412962415e-05, "loss": 0.0241, "step": 55670 }, { "epoch": 2.6, "learning_rate": 2.4881656343126395e-05, "loss": 0.0337, "step": 55675 }, { "epoch": 2.6, "learning_rate": 2.4881188273290375e-05, "loss": 0.0556, "step": 55680 }, { "epoch": 2.6, "learning_rate": 2.4880720203454355e-05, "loss": 0.0622, "step": 55685 }, { "epoch": 2.6, "learning_rate": 2.4880252133618334e-05, "loss": 0.1056, "step": 55690 }, { "epoch": 2.6, "learning_rate": 2.4879784063782318e-05, "loss": 0.1868, "step": 55695 }, { "epoch": 2.6, "learning_rate": 2.4879315993946297e-05, "loss": 0.1624, "step": 55700 }, { "epoch": 2.6, "learning_rate": 2.4878847924110277e-05, "loss": 0.1596, "step": 55705 }, { "epoch": 2.6, "learning_rate": 2.4878379854274257e-05, "loss": 0.1789, "step": 55710 }, { "epoch": 2.6, "learning_rate": 2.487791178443824e-05, "loss": 0.2479, "step": 55715 }, { "epoch": 2.6, "learning_rate": 2.487744371460222e-05, "loss": 0.0446, "step": 55720 }, { "epoch": 2.6, "learning_rate": 2.48769756447662e-05, "loss": 0.0764, "step": 55725 }, { "epoch": 2.6, "learning_rate": 2.4876507574930183e-05, "loss": 0.0528, "step": 55730 }, { "epoch": 2.6, "learning_rate": 2.4876039505094163e-05, "loss": 0.0915, "step": 55735 }, { "epoch": 2.6, "learning_rate": 2.487557143525814e-05, "loss": 0.1073, "step": 55740 }, { "epoch": 2.6, "learning_rate": 2.487510336542212e-05, "loss": 0.0716, "step": 55745 }, { "epoch": 2.6, "learning_rate": 2.4874635295586102e-05, "loss": 0.1146, "step": 55750 }, { "epoch": 2.6, "learning_rate": 2.4874167225750082e-05, "loss": 0.1948, "step": 55755 }, { "epoch": 2.6, "learning_rate": 2.4873699155914062e-05, "loss": 0.2386, "step": 55760 }, { "epoch": 2.6, "learning_rate": 2.4873231086078042e-05, "loss": 0.2818, "step": 55765 }, { "epoch": 2.6, "learning_rate": 2.4872763016242025e-05, "loss": 0.0279, "step": 55770 }, { "epoch": 2.6, "learning_rate": 2.4872294946406005e-05, "loss": 0.039, "step": 55775 }, { "epoch": 2.6, "learning_rate": 2.4871826876569985e-05, "loss": 0.0803, "step": 55780 }, { "epoch": 2.6, "learning_rate": 2.4871358806733965e-05, "loss": 0.0976, "step": 55785 }, { "epoch": 2.6, "learning_rate": 2.4870890736897948e-05, "loss": 0.0641, "step": 55790 }, { "epoch": 2.6, "learning_rate": 2.4870422667061928e-05, "loss": 0.1088, "step": 55795 }, { "epoch": 2.6, "learning_rate": 2.4869954597225907e-05, "loss": 0.2088, "step": 55800 }, { "epoch": 2.6, "learning_rate": 2.4869486527389887e-05, "loss": 0.1582, "step": 55805 }, { "epoch": 2.6, "learning_rate": 2.4869018457553867e-05, "loss": 0.2437, "step": 55810 }, { "epoch": 2.6, "learning_rate": 2.4868550387717847e-05, "loss": 0.2285, "step": 55815 }, { "epoch": 2.6, "learning_rate": 2.4868082317881827e-05, "loss": 0.0515, "step": 55820 }, { "epoch": 2.6, "learning_rate": 2.486761424804581e-05, "loss": 0.0516, "step": 55825 }, { "epoch": 2.61, "learning_rate": 2.486714617820979e-05, "loss": 0.0385, "step": 55830 }, { "epoch": 2.61, "learning_rate": 2.486667810837377e-05, "loss": 0.0661, "step": 55835 }, { "epoch": 2.61, "learning_rate": 2.486621003853775e-05, "loss": 0.1022, "step": 55840 }, { "epoch": 2.61, "learning_rate": 2.4865741968701732e-05, "loss": 0.102, "step": 55845 }, { "epoch": 2.61, "learning_rate": 2.4865273898865712e-05, "loss": 0.1489, "step": 55850 }, { "epoch": 2.61, "learning_rate": 2.4864805829029692e-05, "loss": 0.2035, "step": 55855 }, { "epoch": 2.61, "learning_rate": 2.4864337759193675e-05, "loss": 0.2289, "step": 55860 }, { "epoch": 2.61, "learning_rate": 2.4863869689357652e-05, "loss": 0.2436, "step": 55865 }, { "epoch": 2.61, "learning_rate": 2.486340161952163e-05, "loss": 0.0206, "step": 55870 }, { "epoch": 2.61, "learning_rate": 2.486293354968561e-05, "loss": 0.0459, "step": 55875 }, { "epoch": 2.61, "learning_rate": 2.4862465479849595e-05, "loss": 0.1318, "step": 55880 }, { "epoch": 2.61, "learning_rate": 2.4861997410013574e-05, "loss": 0.0235, "step": 55885 }, { "epoch": 2.61, "learning_rate": 2.4861529340177554e-05, "loss": 0.0769, "step": 55890 }, { "epoch": 2.61, "learning_rate": 2.4861061270341534e-05, "loss": 0.1074, "step": 55895 }, { "epoch": 2.61, "learning_rate": 2.4860593200505517e-05, "loss": 0.1184, "step": 55900 }, { "epoch": 2.61, "learning_rate": 2.4860125130669497e-05, "loss": 0.1981, "step": 55905 }, { "epoch": 2.61, "learning_rate": 2.4859657060833477e-05, "loss": 0.2717, "step": 55910 }, { "epoch": 2.61, "learning_rate": 2.485918899099746e-05, "loss": 0.3014, "step": 55915 }, { "epoch": 2.61, "learning_rate": 2.485872092116144e-05, "loss": 0.0629, "step": 55920 }, { "epoch": 2.61, "learning_rate": 2.485825285132542e-05, "loss": 0.0372, "step": 55925 }, { "epoch": 2.61, "learning_rate": 2.4857784781489396e-05, "loss": 0.0103, "step": 55930 }, { "epoch": 2.61, "learning_rate": 2.485731671165338e-05, "loss": 0.0399, "step": 55935 }, { "epoch": 2.61, "learning_rate": 2.485684864181736e-05, "loss": 0.0618, "step": 55940 }, { "epoch": 2.61, "learning_rate": 2.485638057198134e-05, "loss": 0.0404, "step": 55945 }, { "epoch": 2.61, "learning_rate": 2.485591250214532e-05, "loss": 0.1291, "step": 55950 }, { "epoch": 2.61, "learning_rate": 2.4855444432309302e-05, "loss": 0.1403, "step": 55955 }, { "epoch": 2.61, "learning_rate": 2.4854976362473282e-05, "loss": 0.2289, "step": 55960 }, { "epoch": 2.61, "learning_rate": 2.485450829263726e-05, "loss": 0.3351, "step": 55965 }, { "epoch": 2.61, "learning_rate": 2.4854040222801245e-05, "loss": 0.0568, "step": 55970 }, { "epoch": 2.61, "learning_rate": 2.4853572152965225e-05, "loss": 0.0629, "step": 55975 }, { "epoch": 2.61, "learning_rate": 2.4853104083129204e-05, "loss": 0.0856, "step": 55980 }, { "epoch": 2.61, "learning_rate": 2.4852636013293184e-05, "loss": 0.1445, "step": 55985 }, { "epoch": 2.61, "learning_rate": 2.4852167943457164e-05, "loss": 0.132, "step": 55990 }, { "epoch": 2.61, "learning_rate": 2.4851699873621144e-05, "loss": 0.0855, "step": 55995 }, { "epoch": 2.61, "learning_rate": 2.4851231803785124e-05, "loss": 0.2332, "step": 56000 }, { "epoch": 2.61, "learning_rate": 2.4850763733949104e-05, "loss": 0.1916, "step": 56005 }, { "epoch": 2.61, "learning_rate": 2.4850295664113087e-05, "loss": 0.2721, "step": 56010 }, { "epoch": 2.61, "learning_rate": 2.4849827594277067e-05, "loss": 0.258, "step": 56015 }, { "epoch": 2.61, "learning_rate": 2.4849359524441046e-05, "loss": 0.0678, "step": 56020 }, { "epoch": 2.61, "learning_rate": 2.4848891454605026e-05, "loss": 0.0619, "step": 56025 }, { "epoch": 2.61, "learning_rate": 2.484842338476901e-05, "loss": 0.0764, "step": 56030 }, { "epoch": 2.61, "learning_rate": 2.484795531493299e-05, "loss": 0.0429, "step": 56035 }, { "epoch": 2.61, "learning_rate": 2.484748724509697e-05, "loss": 0.078, "step": 56040 }, { "epoch": 2.62, "learning_rate": 2.4847019175260952e-05, "loss": 0.0844, "step": 56045 }, { "epoch": 2.62, "learning_rate": 2.4846551105424932e-05, "loss": 0.0517, "step": 56050 }, { "epoch": 2.62, "learning_rate": 2.484608303558891e-05, "loss": 0.2129, "step": 56055 }, { "epoch": 2.62, "learning_rate": 2.484561496575289e-05, "loss": 0.2468, "step": 56060 }, { "epoch": 2.62, "learning_rate": 2.484514689591687e-05, "loss": 0.3149, "step": 56065 }, { "epoch": 2.62, "learning_rate": 2.484467882608085e-05, "loss": 0.061, "step": 56070 }, { "epoch": 2.62, "learning_rate": 2.484421075624483e-05, "loss": 0.0645, "step": 56075 }, { "epoch": 2.62, "learning_rate": 2.484374268640881e-05, "loss": 0.0696, "step": 56080 }, { "epoch": 2.62, "learning_rate": 2.4843274616572794e-05, "loss": 0.0416, "step": 56085 }, { "epoch": 2.62, "learning_rate": 2.4842806546736774e-05, "loss": 0.0714, "step": 56090 }, { "epoch": 2.62, "learning_rate": 2.4842338476900754e-05, "loss": 0.0705, "step": 56095 }, { "epoch": 2.62, "learning_rate": 2.4841870407064737e-05, "loss": 0.1065, "step": 56100 }, { "epoch": 2.62, "learning_rate": 2.4841402337228717e-05, "loss": 0.1444, "step": 56105 }, { "epoch": 2.62, "learning_rate": 2.4840934267392697e-05, "loss": 0.2391, "step": 56110 }, { "epoch": 2.62, "learning_rate": 2.4840466197556677e-05, "loss": 0.2695, "step": 56115 }, { "epoch": 2.62, "learning_rate": 2.4839998127720656e-05, "loss": 0.0378, "step": 56120 }, { "epoch": 2.62, "learning_rate": 2.4839530057884636e-05, "loss": 0.0255, "step": 56125 }, { "epoch": 2.62, "learning_rate": 2.4839061988048616e-05, "loss": 0.0216, "step": 56130 }, { "epoch": 2.62, "learning_rate": 2.4838593918212596e-05, "loss": 0.0649, "step": 56135 }, { "epoch": 2.62, "learning_rate": 2.483812584837658e-05, "loss": 0.0932, "step": 56140 }, { "epoch": 2.62, "learning_rate": 2.483765777854056e-05, "loss": 0.0792, "step": 56145 }, { "epoch": 2.62, "learning_rate": 2.483718970870454e-05, "loss": 0.2263, "step": 56150 }, { "epoch": 2.62, "learning_rate": 2.4836721638868522e-05, "loss": 0.2608, "step": 56155 }, { "epoch": 2.62, "learning_rate": 2.48362535690325e-05, "loss": 0.171, "step": 56160 }, { "epoch": 2.62, "learning_rate": 2.483578549919648e-05, "loss": 0.3316, "step": 56165 }, { "epoch": 2.62, "learning_rate": 2.483531742936046e-05, "loss": 0.0179, "step": 56170 }, { "epoch": 2.62, "learning_rate": 2.4834849359524444e-05, "loss": 0.0258, "step": 56175 }, { "epoch": 2.62, "learning_rate": 2.483438128968842e-05, "loss": 0.0419, "step": 56180 }, { "epoch": 2.62, "learning_rate": 2.48339132198524e-05, "loss": 0.0238, "step": 56185 }, { "epoch": 2.62, "learning_rate": 2.483344515001638e-05, "loss": 0.0839, "step": 56190 }, { "epoch": 2.62, "learning_rate": 2.4832977080180364e-05, "loss": 0.1399, "step": 56195 }, { "epoch": 2.62, "learning_rate": 2.4832509010344344e-05, "loss": 0.1505, "step": 56200 }, { "epoch": 2.62, "learning_rate": 2.4832040940508323e-05, "loss": 0.248, "step": 56205 }, { "epoch": 2.62, "learning_rate": 2.4831572870672303e-05, "loss": 0.2085, "step": 56210 }, { "epoch": 2.62, "learning_rate": 2.4831104800836286e-05, "loss": 0.2723, "step": 56215 }, { "epoch": 2.62, "learning_rate": 2.4830636731000266e-05, "loss": 0.0246, "step": 56220 }, { "epoch": 2.62, "learning_rate": 2.4830168661164246e-05, "loss": 0.0453, "step": 56225 }, { "epoch": 2.62, "learning_rate": 2.482970059132823e-05, "loss": 0.0159, "step": 56230 }, { "epoch": 2.62, "learning_rate": 2.482923252149221e-05, "loss": 0.0821, "step": 56235 }, { "epoch": 2.62, "learning_rate": 2.482876445165619e-05, "loss": 0.0611, "step": 56240 }, { "epoch": 2.62, "learning_rate": 2.4828296381820165e-05, "loss": 0.0454, "step": 56245 }, { "epoch": 2.62, "learning_rate": 2.482782831198415e-05, "loss": 0.1062, "step": 56250 }, { "epoch": 2.62, "learning_rate": 2.482736024214813e-05, "loss": 0.1669, "step": 56255 }, { "epoch": 2.63, "learning_rate": 2.4826892172312108e-05, "loss": 0.1415, "step": 56260 }, { "epoch": 2.63, "learning_rate": 2.4826424102476088e-05, "loss": 0.2721, "step": 56265 }, { "epoch": 2.63, "learning_rate": 2.482595603264007e-05, "loss": 0.0676, "step": 56270 }, { "epoch": 2.63, "learning_rate": 2.482548796280405e-05, "loss": 0.1631, "step": 56275 }, { "epoch": 2.63, "learning_rate": 2.482501989296803e-05, "loss": 0.0285, "step": 56280 }, { "epoch": 2.63, "learning_rate": 2.4824551823132014e-05, "loss": 0.0791, "step": 56285 }, { "epoch": 2.63, "learning_rate": 2.4824083753295994e-05, "loss": 0.1029, "step": 56290 }, { "epoch": 2.63, "learning_rate": 2.4823615683459974e-05, "loss": 0.1792, "step": 56295 }, { "epoch": 2.63, "learning_rate": 2.4823147613623953e-05, "loss": 0.1128, "step": 56300 }, { "epoch": 2.63, "learning_rate": 2.4822679543787937e-05, "loss": 0.1171, "step": 56305 }, { "epoch": 2.63, "learning_rate": 2.4822211473951913e-05, "loss": 0.293, "step": 56310 }, { "epoch": 2.63, "learning_rate": 2.4821743404115893e-05, "loss": 0.3136, "step": 56315 }, { "epoch": 2.63, "learning_rate": 2.4821275334279873e-05, "loss": 0.0717, "step": 56320 }, { "epoch": 2.63, "learning_rate": 2.4820807264443856e-05, "loss": 0.0332, "step": 56325 }, { "epoch": 2.63, "learning_rate": 2.4820339194607836e-05, "loss": 0.0609, "step": 56330 }, { "epoch": 2.63, "learning_rate": 2.4819871124771816e-05, "loss": 0.0853, "step": 56335 }, { "epoch": 2.63, "learning_rate": 2.48194030549358e-05, "loss": 0.0824, "step": 56340 }, { "epoch": 2.63, "learning_rate": 2.481893498509978e-05, "loss": 0.1004, "step": 56345 }, { "epoch": 2.63, "learning_rate": 2.481846691526376e-05, "loss": 0.1309, "step": 56350 }, { "epoch": 2.63, "learning_rate": 2.4817998845427738e-05, "loss": 0.1891, "step": 56355 }, { "epoch": 2.63, "learning_rate": 2.481753077559172e-05, "loss": 0.2519, "step": 56360 }, { "epoch": 2.63, "learning_rate": 2.48170627057557e-05, "loss": 0.1681, "step": 56365 }, { "epoch": 2.63, "learning_rate": 2.4816594635919678e-05, "loss": 0.0233, "step": 56370 }, { "epoch": 2.63, "learning_rate": 2.4816126566083658e-05, "loss": 0.0524, "step": 56375 }, { "epoch": 2.63, "learning_rate": 2.481565849624764e-05, "loss": 0.0128, "step": 56380 }, { "epoch": 2.63, "learning_rate": 2.481519042641162e-05, "loss": 0.0438, "step": 56385 }, { "epoch": 2.63, "learning_rate": 2.48147223565756e-05, "loss": 0.0893, "step": 56390 }, { "epoch": 2.63, "learning_rate": 2.481425428673958e-05, "loss": 0.1187, "step": 56395 }, { "epoch": 2.63, "learning_rate": 2.4813786216903563e-05, "loss": 0.1523, "step": 56400 }, { "epoch": 2.63, "learning_rate": 2.4813318147067543e-05, "loss": 0.1913, "step": 56405 }, { "epoch": 2.63, "learning_rate": 2.4812850077231523e-05, "loss": 0.2801, "step": 56410 }, { "epoch": 2.63, "learning_rate": 2.4812382007395506e-05, "loss": 0.4468, "step": 56415 }, { "epoch": 2.63, "learning_rate": 2.4811913937559486e-05, "loss": 0.0819, "step": 56420 }, { "epoch": 2.63, "learning_rate": 2.4811445867723466e-05, "loss": 0.0279, "step": 56425 }, { "epoch": 2.63, "learning_rate": 2.4810977797887446e-05, "loss": 0.1139, "step": 56430 }, { "epoch": 2.63, "learning_rate": 2.4810509728051426e-05, "loss": 0.0545, "step": 56435 }, { "epoch": 2.63, "learning_rate": 2.4810041658215405e-05, "loss": 0.0967, "step": 56440 }, { "epoch": 2.63, "learning_rate": 2.4809573588379385e-05, "loss": 0.1217, "step": 56445 }, { "epoch": 2.63, "learning_rate": 2.4809105518543365e-05, "loss": 0.1794, "step": 56450 }, { "epoch": 2.63, "learning_rate": 2.4808637448707348e-05, "loss": 0.1976, "step": 56455 }, { "epoch": 2.63, "learning_rate": 2.4808169378871328e-05, "loss": 0.3034, "step": 56460 }, { "epoch": 2.63, "learning_rate": 2.4807701309035308e-05, "loss": 0.205, "step": 56465 }, { "epoch": 2.63, "learning_rate": 2.480723323919929e-05, "loss": 0.069, "step": 56470 }, { "epoch": 2.64, "learning_rate": 2.480676516936327e-05, "loss": 0.0386, "step": 56475 }, { "epoch": 2.64, "learning_rate": 2.480629709952725e-05, "loss": 0.0713, "step": 56480 }, { "epoch": 2.64, "learning_rate": 2.480582902969123e-05, "loss": 0.0785, "step": 56485 }, { "epoch": 2.64, "learning_rate": 2.4805360959855214e-05, "loss": 0.0872, "step": 56490 }, { "epoch": 2.64, "learning_rate": 2.4804892890019193e-05, "loss": 0.1564, "step": 56495 }, { "epoch": 2.64, "learning_rate": 2.480442482018317e-05, "loss": 0.2005, "step": 56500 }, { "epoch": 2.64, "learning_rate": 2.480395675034715e-05, "loss": 0.0865, "step": 56505 }, { "epoch": 2.64, "learning_rate": 2.4803488680511133e-05, "loss": 0.213, "step": 56510 }, { "epoch": 2.64, "learning_rate": 2.4803020610675113e-05, "loss": 0.4198, "step": 56515 }, { "epoch": 2.64, "learning_rate": 2.4802552540839093e-05, "loss": 0.0398, "step": 56520 }, { "epoch": 2.64, "learning_rate": 2.4802084471003076e-05, "loss": 0.0896, "step": 56525 }, { "epoch": 2.64, "learning_rate": 2.4801616401167056e-05, "loss": 0.0395, "step": 56530 }, { "epoch": 2.64, "learning_rate": 2.4801148331331035e-05, "loss": 0.0644, "step": 56535 }, { "epoch": 2.64, "learning_rate": 2.4800680261495015e-05, "loss": 0.0664, "step": 56540 }, { "epoch": 2.64, "learning_rate": 2.4800212191659e-05, "loss": 0.1634, "step": 56545 }, { "epoch": 2.64, "learning_rate": 2.4799744121822978e-05, "loss": 0.0776, "step": 56550 }, { "epoch": 2.64, "learning_rate": 2.4799276051986958e-05, "loss": 0.1236, "step": 56555 }, { "epoch": 2.64, "learning_rate": 2.4798807982150934e-05, "loss": 0.2574, "step": 56560 }, { "epoch": 2.64, "learning_rate": 2.4798339912314918e-05, "loss": 0.3454, "step": 56565 }, { "epoch": 2.64, "learning_rate": 2.4797871842478898e-05, "loss": 0.0262, "step": 56570 }, { "epoch": 2.64, "learning_rate": 2.4797403772642877e-05, "loss": 0.1027, "step": 56575 }, { "epoch": 2.64, "learning_rate": 2.479693570280686e-05, "loss": 0.0402, "step": 56580 }, { "epoch": 2.64, "learning_rate": 2.479646763297084e-05, "loss": 0.0715, "step": 56585 }, { "epoch": 2.64, "learning_rate": 2.479599956313482e-05, "loss": 0.0632, "step": 56590 }, { "epoch": 2.64, "learning_rate": 2.47955314932988e-05, "loss": 0.1507, "step": 56595 }, { "epoch": 2.64, "learning_rate": 2.4795063423462783e-05, "loss": 0.1241, "step": 56600 }, { "epoch": 2.64, "learning_rate": 2.4794595353626763e-05, "loss": 0.1424, "step": 56605 }, { "epoch": 2.64, "learning_rate": 2.4794127283790743e-05, "loss": 0.2284, "step": 56610 }, { "epoch": 2.64, "learning_rate": 2.4793659213954723e-05, "loss": 0.2301, "step": 56615 }, { "epoch": 2.64, "learning_rate": 2.4793191144118706e-05, "loss": 0.0888, "step": 56620 }, { "epoch": 2.64, "learning_rate": 2.4792723074282682e-05, "loss": 0.0232, "step": 56625 }, { "epoch": 2.64, "learning_rate": 2.4792255004446662e-05, "loss": 0.0165, "step": 56630 }, { "epoch": 2.64, "learning_rate": 2.4791786934610642e-05, "loss": 0.0989, "step": 56635 }, { "epoch": 2.64, "learning_rate": 2.4791318864774625e-05, "loss": 0.0603, "step": 56640 }, { "epoch": 2.64, "learning_rate": 2.4790850794938605e-05, "loss": 0.09, "step": 56645 }, { "epoch": 2.64, "learning_rate": 2.4790382725102585e-05, "loss": 0.1327, "step": 56650 }, { "epoch": 2.64, "learning_rate": 2.4789914655266568e-05, "loss": 0.1227, "step": 56655 }, { "epoch": 2.64, "learning_rate": 2.4789446585430548e-05, "loss": 0.2768, "step": 56660 }, { "epoch": 2.64, "learning_rate": 2.4788978515594528e-05, "loss": 0.2845, "step": 56665 }, { "epoch": 2.64, "learning_rate": 2.4788510445758507e-05, "loss": 0.0072, "step": 56670 }, { "epoch": 2.64, "learning_rate": 2.478804237592249e-05, "loss": 0.0421, "step": 56675 }, { "epoch": 2.64, "learning_rate": 2.478757430608647e-05, "loss": 0.0871, "step": 56680 }, { "epoch": 2.65, "learning_rate": 2.478710623625045e-05, "loss": 0.134, "step": 56685 }, { "epoch": 2.65, "learning_rate": 2.4786638166414427e-05, "loss": 0.108, "step": 56690 }, { "epoch": 2.65, "learning_rate": 2.478617009657841e-05, "loss": 0.0634, "step": 56695 }, { "epoch": 2.65, "learning_rate": 2.478570202674239e-05, "loss": 0.0972, "step": 56700 }, { "epoch": 2.65, "learning_rate": 2.478523395690637e-05, "loss": 0.151, "step": 56705 }, { "epoch": 2.65, "learning_rate": 2.4784765887070353e-05, "loss": 0.1934, "step": 56710 }, { "epoch": 2.65, "learning_rate": 2.4784297817234333e-05, "loss": 0.2165, "step": 56715 }, { "epoch": 2.65, "learning_rate": 2.4783829747398312e-05, "loss": 0.0419, "step": 56720 }, { "epoch": 2.65, "learning_rate": 2.4783361677562292e-05, "loss": 0.0295, "step": 56725 }, { "epoch": 2.65, "learning_rate": 2.4782893607726275e-05, "loss": 0.0331, "step": 56730 }, { "epoch": 2.65, "learning_rate": 2.4782425537890255e-05, "loss": 0.0786, "step": 56735 }, { "epoch": 2.65, "learning_rate": 2.4781957468054235e-05, "loss": 0.0559, "step": 56740 }, { "epoch": 2.65, "learning_rate": 2.4781489398218215e-05, "loss": 0.0944, "step": 56745 }, { "epoch": 2.65, "learning_rate": 2.4781021328382195e-05, "loss": 0.153, "step": 56750 }, { "epoch": 2.65, "learning_rate": 2.4780553258546174e-05, "loss": 0.1432, "step": 56755 }, { "epoch": 2.65, "learning_rate": 2.4780085188710154e-05, "loss": 0.1883, "step": 56760 }, { "epoch": 2.65, "learning_rate": 2.4779617118874138e-05, "loss": 0.3125, "step": 56765 }, { "epoch": 2.65, "learning_rate": 2.4779149049038117e-05, "loss": 0.0633, "step": 56770 }, { "epoch": 2.65, "learning_rate": 2.4778680979202097e-05, "loss": 0.0644, "step": 56775 }, { "epoch": 2.65, "learning_rate": 2.4778212909366077e-05, "loss": 0.1129, "step": 56780 }, { "epoch": 2.65, "learning_rate": 2.477774483953006e-05, "loss": 0.0388, "step": 56785 }, { "epoch": 2.65, "learning_rate": 2.477727676969404e-05, "loss": 0.0727, "step": 56790 }, { "epoch": 2.65, "learning_rate": 2.477680869985802e-05, "loss": 0.1007, "step": 56795 }, { "epoch": 2.65, "learning_rate": 2.4776340630022e-05, "loss": 0.1425, "step": 56800 }, { "epoch": 2.65, "learning_rate": 2.4775872560185983e-05, "loss": 0.2667, "step": 56805 }, { "epoch": 2.65, "learning_rate": 2.4775404490349963e-05, "loss": 0.2314, "step": 56810 }, { "epoch": 2.65, "learning_rate": 2.477493642051394e-05, "loss": 0.2242, "step": 56815 }, { "epoch": 2.65, "learning_rate": 2.477446835067792e-05, "loss": 0.0418, "step": 56820 }, { "epoch": 2.65, "learning_rate": 2.4774000280841902e-05, "loss": 0.0443, "step": 56825 }, { "epoch": 2.65, "learning_rate": 2.4773532211005882e-05, "loss": 0.0666, "step": 56830 }, { "epoch": 2.65, "learning_rate": 2.4773064141169862e-05, "loss": 0.0267, "step": 56835 }, { "epoch": 2.65, "learning_rate": 2.4772596071333845e-05, "loss": 0.065, "step": 56840 }, { "epoch": 2.65, "learning_rate": 2.4772128001497825e-05, "loss": 0.0928, "step": 56845 }, { "epoch": 2.65, "learning_rate": 2.4771659931661805e-05, "loss": 0.1144, "step": 56850 }, { "epoch": 2.65, "learning_rate": 2.4771191861825784e-05, "loss": 0.1679, "step": 56855 }, { "epoch": 2.65, "learning_rate": 2.4770723791989768e-05, "loss": 0.2068, "step": 56860 }, { "epoch": 2.65, "learning_rate": 2.4770255722153747e-05, "loss": 0.1983, "step": 56865 }, { "epoch": 2.65, "learning_rate": 2.4769787652317727e-05, "loss": 0.0492, "step": 56870 }, { "epoch": 2.65, "learning_rate": 2.4769319582481704e-05, "loss": 0.0197, "step": 56875 }, { "epoch": 2.65, "learning_rate": 2.4768851512645687e-05, "loss": 0.0459, "step": 56880 }, { "epoch": 2.65, "learning_rate": 2.4768383442809667e-05, "loss": 0.0812, "step": 56885 }, { "epoch": 2.65, "learning_rate": 2.4767915372973647e-05, "loss": 0.0608, "step": 56890 }, { "epoch": 2.65, "learning_rate": 2.476744730313763e-05, "loss": 0.1539, "step": 56895 }, { "epoch": 2.66, "learning_rate": 2.476697923330161e-05, "loss": 0.1474, "step": 56900 }, { "epoch": 2.66, "learning_rate": 2.476651116346559e-05, "loss": 0.1491, "step": 56905 }, { "epoch": 2.66, "learning_rate": 2.476604309362957e-05, "loss": 0.1857, "step": 56910 }, { "epoch": 2.66, "learning_rate": 2.4765575023793552e-05, "loss": 0.2067, "step": 56915 }, { "epoch": 2.66, "learning_rate": 2.4765106953957532e-05, "loss": 0.0749, "step": 56920 }, { "epoch": 2.66, "learning_rate": 2.4764638884121512e-05, "loss": 0.0356, "step": 56925 }, { "epoch": 2.66, "learning_rate": 2.4764170814285492e-05, "loss": 0.0496, "step": 56930 }, { "epoch": 2.66, "learning_rate": 2.4763702744449475e-05, "loss": 0.0854, "step": 56935 }, { "epoch": 2.66, "learning_rate": 2.476323467461345e-05, "loss": 0.1459, "step": 56940 }, { "epoch": 2.66, "learning_rate": 2.476276660477743e-05, "loss": 0.0947, "step": 56945 }, { "epoch": 2.66, "learning_rate": 2.4762298534941414e-05, "loss": 0.1419, "step": 56950 }, { "epoch": 2.66, "learning_rate": 2.4761830465105394e-05, "loss": 0.088, "step": 56955 }, { "epoch": 2.66, "learning_rate": 2.4761362395269374e-05, "loss": 0.2764, "step": 56960 }, { "epoch": 2.66, "learning_rate": 2.4760894325433354e-05, "loss": 0.2743, "step": 56965 }, { "epoch": 2.66, "learning_rate": 2.4760426255597337e-05, "loss": 0.0423, "step": 56970 }, { "epoch": 2.66, "learning_rate": 2.4759958185761317e-05, "loss": 0.038, "step": 56975 }, { "epoch": 2.66, "learning_rate": 2.4759490115925297e-05, "loss": 0.0391, "step": 56980 }, { "epoch": 2.66, "learning_rate": 2.4759022046089277e-05, "loss": 0.0262, "step": 56985 }, { "epoch": 2.66, "learning_rate": 2.475855397625326e-05, "loss": 0.0708, "step": 56990 }, { "epoch": 2.66, "learning_rate": 2.475808590641724e-05, "loss": 0.1333, "step": 56995 }, { "epoch": 2.66, "learning_rate": 2.475761783658122e-05, "loss": 0.0835, "step": 57000 }, { "epoch": 2.66, "learning_rate": 2.4757149766745196e-05, "loss": 0.1772, "step": 57005 }, { "epoch": 2.66, "learning_rate": 2.475668169690918e-05, "loss": 0.206, "step": 57010 }, { "epoch": 2.66, "learning_rate": 2.475621362707316e-05, "loss": 0.2662, "step": 57015 }, { "epoch": 2.66, "learning_rate": 2.475574555723714e-05, "loss": 0.0727, "step": 57020 }, { "epoch": 2.66, "learning_rate": 2.4755277487401122e-05, "loss": 0.0496, "step": 57025 }, { "epoch": 2.66, "learning_rate": 2.4754809417565102e-05, "loss": 0.0757, "step": 57030 }, { "epoch": 2.66, "learning_rate": 2.475434134772908e-05, "loss": 0.0172, "step": 57035 }, { "epoch": 2.66, "learning_rate": 2.475387327789306e-05, "loss": 0.0645, "step": 57040 }, { "epoch": 2.66, "learning_rate": 2.4753405208057045e-05, "loss": 0.0947, "step": 57045 }, { "epoch": 2.66, "learning_rate": 2.4752937138221024e-05, "loss": 0.1176, "step": 57050 }, { "epoch": 2.66, "learning_rate": 2.4752469068385004e-05, "loss": 0.1293, "step": 57055 }, { "epoch": 2.66, "learning_rate": 2.4752000998548984e-05, "loss": 0.3489, "step": 57060 }, { "epoch": 2.66, "learning_rate": 2.4751532928712964e-05, "loss": 0.2541, "step": 57065 }, { "epoch": 2.66, "learning_rate": 2.4751064858876944e-05, "loss": 0.0181, "step": 57070 }, { "epoch": 2.66, "learning_rate": 2.4750596789040923e-05, "loss": 0.0612, "step": 57075 }, { "epoch": 2.66, "learning_rate": 2.4750128719204907e-05, "loss": 0.0179, "step": 57080 }, { "epoch": 2.66, "learning_rate": 2.4749660649368887e-05, "loss": 0.0416, "step": 57085 }, { "epoch": 2.66, "learning_rate": 2.4749192579532866e-05, "loss": 0.0869, "step": 57090 }, { "epoch": 2.66, "learning_rate": 2.4748724509696846e-05, "loss": 0.1073, "step": 57095 }, { "epoch": 2.66, "learning_rate": 2.474825643986083e-05, "loss": 0.1542, "step": 57100 }, { "epoch": 2.66, "learning_rate": 2.474778837002481e-05, "loss": 0.0929, "step": 57105 }, { "epoch": 2.66, "learning_rate": 2.474732030018879e-05, "loss": 0.2521, "step": 57110 }, { "epoch": 2.67, "learning_rate": 2.474685223035277e-05, "loss": 0.2317, "step": 57115 }, { "epoch": 2.67, "learning_rate": 2.4746384160516752e-05, "loss": 0.0269, "step": 57120 }, { "epoch": 2.67, "learning_rate": 2.4745916090680732e-05, "loss": 0.0535, "step": 57125 }, { "epoch": 2.67, "learning_rate": 2.4745448020844708e-05, "loss": 0.044, "step": 57130 }, { "epoch": 2.67, "learning_rate": 2.474497995100869e-05, "loss": 0.0247, "step": 57135 }, { "epoch": 2.67, "learning_rate": 2.474451188117267e-05, "loss": 0.1443, "step": 57140 }, { "epoch": 2.67, "learning_rate": 2.474404381133665e-05, "loss": 0.121, "step": 57145 }, { "epoch": 2.67, "learning_rate": 2.474357574150063e-05, "loss": 0.1178, "step": 57150 }, { "epoch": 2.67, "learning_rate": 2.4743107671664614e-05, "loss": 0.0933, "step": 57155 }, { "epoch": 2.67, "learning_rate": 2.4742639601828594e-05, "loss": 0.2813, "step": 57160 }, { "epoch": 2.67, "learning_rate": 2.4742171531992574e-05, "loss": 0.301, "step": 57165 }, { "epoch": 2.67, "learning_rate": 2.4741703462156554e-05, "loss": 0.1123, "step": 57170 }, { "epoch": 2.67, "learning_rate": 2.4741235392320537e-05, "loss": 0.0386, "step": 57175 }, { "epoch": 2.67, "learning_rate": 2.4740767322484517e-05, "loss": 0.0567, "step": 57180 }, { "epoch": 2.67, "learning_rate": 2.4740299252648496e-05, "loss": 0.0879, "step": 57185 }, { "epoch": 2.67, "learning_rate": 2.4739831182812476e-05, "loss": 0.0381, "step": 57190 }, { "epoch": 2.67, "learning_rate": 2.4739363112976456e-05, "loss": 0.1663, "step": 57195 }, { "epoch": 2.67, "learning_rate": 2.4738895043140436e-05, "loss": 0.1015, "step": 57200 }, { "epoch": 2.67, "learning_rate": 2.4738426973304416e-05, "loss": 0.0837, "step": 57205 }, { "epoch": 2.67, "learning_rate": 2.47379589034684e-05, "loss": 0.2533, "step": 57210 }, { "epoch": 2.67, "learning_rate": 2.473749083363238e-05, "loss": 0.3941, "step": 57215 }, { "epoch": 2.67, "learning_rate": 2.473702276379636e-05, "loss": 0.0044, "step": 57220 }, { "epoch": 2.67, "learning_rate": 2.473655469396034e-05, "loss": 0.0511, "step": 57225 }, { "epoch": 2.67, "learning_rate": 2.473608662412432e-05, "loss": 0.0827, "step": 57230 }, { "epoch": 2.67, "learning_rate": 2.47356185542883e-05, "loss": 0.0745, "step": 57235 }, { "epoch": 2.67, "learning_rate": 2.473515048445228e-05, "loss": 0.0707, "step": 57240 }, { "epoch": 2.67, "learning_rate": 2.473468241461626e-05, "loss": 0.1048, "step": 57245 }, { "epoch": 2.67, "learning_rate": 2.4734214344780244e-05, "loss": 0.1516, "step": 57250 }, { "epoch": 2.67, "learning_rate": 2.473374627494422e-05, "loss": 0.1173, "step": 57255 }, { "epoch": 2.67, "learning_rate": 2.47332782051082e-05, "loss": 0.1629, "step": 57260 }, { "epoch": 2.67, "learning_rate": 2.4732810135272184e-05, "loss": 0.2528, "step": 57265 }, { "epoch": 2.67, "learning_rate": 2.4732342065436163e-05, "loss": 0.0543, "step": 57270 }, { "epoch": 2.67, "learning_rate": 2.4731873995600143e-05, "loss": 0.0509, "step": 57275 }, { "epoch": 2.67, "learning_rate": 2.4731405925764123e-05, "loss": 0.1243, "step": 57280 }, { "epoch": 2.67, "learning_rate": 2.4730937855928106e-05, "loss": 0.0362, "step": 57285 }, { "epoch": 2.67, "learning_rate": 2.4730469786092086e-05, "loss": 0.0898, "step": 57290 }, { "epoch": 2.67, "learning_rate": 2.4730001716256066e-05, "loss": 0.0491, "step": 57295 }, { "epoch": 2.67, "learning_rate": 2.4729533646420046e-05, "loss": 0.075, "step": 57300 }, { "epoch": 2.67, "learning_rate": 2.472906557658403e-05, "loss": 0.1388, "step": 57305 }, { "epoch": 2.67, "learning_rate": 2.472859750674801e-05, "loss": 0.2626, "step": 57310 }, { "epoch": 2.67, "learning_rate": 2.472812943691199e-05, "loss": 0.3333, "step": 57315 }, { "epoch": 2.67, "learning_rate": 2.472766136707597e-05, "loss": 0.0611, "step": 57320 }, { "epoch": 2.67, "learning_rate": 2.4727193297239948e-05, "loss": 0.0921, "step": 57325 }, { "epoch": 2.68, "learning_rate": 2.4726725227403928e-05, "loss": 0.0896, "step": 57330 }, { "epoch": 2.68, "learning_rate": 2.4726257157567908e-05, "loss": 0.0765, "step": 57335 }, { "epoch": 2.68, "learning_rate": 2.472578908773189e-05, "loss": 0.0917, "step": 57340 }, { "epoch": 2.68, "learning_rate": 2.472532101789587e-05, "loss": 0.1287, "step": 57345 }, { "epoch": 2.68, "learning_rate": 2.472485294805985e-05, "loss": 0.0531, "step": 57350 }, { "epoch": 2.68, "learning_rate": 2.472438487822383e-05, "loss": 0.1417, "step": 57355 }, { "epoch": 2.68, "learning_rate": 2.4723916808387814e-05, "loss": 0.2587, "step": 57360 }, { "epoch": 2.68, "learning_rate": 2.4723448738551794e-05, "loss": 0.2192, "step": 57365 }, { "epoch": 2.68, "learning_rate": 2.4722980668715773e-05, "loss": 0.0221, "step": 57370 }, { "epoch": 2.68, "learning_rate": 2.4722512598879757e-05, "loss": 0.0919, "step": 57375 }, { "epoch": 2.68, "learning_rate": 2.4722044529043733e-05, "loss": 0.0742, "step": 57380 }, { "epoch": 2.68, "learning_rate": 2.4721576459207713e-05, "loss": 0.0954, "step": 57385 }, { "epoch": 2.68, "learning_rate": 2.4721108389371693e-05, "loss": 0.0399, "step": 57390 }, { "epoch": 2.68, "learning_rate": 2.4720640319535676e-05, "loss": 0.0851, "step": 57395 }, { "epoch": 2.68, "learning_rate": 2.4720172249699656e-05, "loss": 0.1236, "step": 57400 }, { "epoch": 2.68, "learning_rate": 2.4719704179863635e-05, "loss": 0.1798, "step": 57405 }, { "epoch": 2.68, "learning_rate": 2.4719236110027615e-05, "loss": 0.2454, "step": 57410 }, { "epoch": 2.68, "learning_rate": 2.47187680401916e-05, "loss": 0.3206, "step": 57415 }, { "epoch": 2.68, "learning_rate": 2.471829997035558e-05, "loss": 0.0538, "step": 57420 }, { "epoch": 2.68, "learning_rate": 2.4717831900519558e-05, "loss": 0.0657, "step": 57425 }, { "epoch": 2.68, "learning_rate": 2.4717363830683538e-05, "loss": 0.0539, "step": 57430 }, { "epoch": 2.68, "learning_rate": 2.471689576084752e-05, "loss": 0.0371, "step": 57435 }, { "epoch": 2.68, "learning_rate": 2.47164276910115e-05, "loss": 0.107, "step": 57440 }, { "epoch": 2.68, "learning_rate": 2.4715959621175477e-05, "loss": 0.2194, "step": 57445 }, { "epoch": 2.68, "learning_rate": 2.471549155133946e-05, "loss": 0.119, "step": 57450 }, { "epoch": 2.68, "learning_rate": 2.471502348150344e-05, "loss": 0.2139, "step": 57455 }, { "epoch": 2.68, "learning_rate": 2.471455541166742e-05, "loss": 0.1608, "step": 57460 }, { "epoch": 2.68, "learning_rate": 2.47140873418314e-05, "loss": 0.3079, "step": 57465 }, { "epoch": 2.68, "learning_rate": 2.4713619271995383e-05, "loss": 0.0415, "step": 57470 }, { "epoch": 2.68, "learning_rate": 2.4713151202159363e-05, "loss": 0.0355, "step": 57475 }, { "epoch": 2.68, "learning_rate": 2.4712683132323343e-05, "loss": 0.0756, "step": 57480 }, { "epoch": 2.68, "learning_rate": 2.4712215062487323e-05, "loss": 0.043, "step": 57485 }, { "epoch": 2.68, "learning_rate": 2.4711746992651306e-05, "loss": 0.0609, "step": 57490 }, { "epoch": 2.68, "learning_rate": 2.4711278922815286e-05, "loss": 0.0613, "step": 57495 }, { "epoch": 2.68, "learning_rate": 2.4710810852979266e-05, "loss": 0.1123, "step": 57500 }, { "epoch": 2.68, "learning_rate": 2.471034278314325e-05, "loss": 0.0651, "step": 57505 }, { "epoch": 2.68, "learning_rate": 2.4709874713307225e-05, "loss": 0.1312, "step": 57510 }, { "epoch": 2.68, "learning_rate": 2.4709406643471205e-05, "loss": 0.1825, "step": 57515 }, { "epoch": 2.68, "learning_rate": 2.4708938573635185e-05, "loss": 0.0397, "step": 57520 }, { "epoch": 2.68, "learning_rate": 2.4708470503799168e-05, "loss": 0.0311, "step": 57525 }, { "epoch": 2.68, "learning_rate": 2.4708002433963148e-05, "loss": 0.0295, "step": 57530 }, { "epoch": 2.68, "learning_rate": 2.4707534364127128e-05, "loss": 0.038, "step": 57535 }, { "epoch": 2.68, "learning_rate": 2.4707066294291108e-05, "loss": 0.0646, "step": 57540 }, { "epoch": 2.69, "learning_rate": 2.470659822445509e-05, "loss": 0.0442, "step": 57545 }, { "epoch": 2.69, "learning_rate": 2.470613015461907e-05, "loss": 0.1179, "step": 57550 }, { "epoch": 2.69, "learning_rate": 2.470566208478305e-05, "loss": 0.1921, "step": 57555 }, { "epoch": 2.69, "learning_rate": 2.4705194014947034e-05, "loss": 0.2513, "step": 57560 }, { "epoch": 2.69, "learning_rate": 2.4704725945111013e-05, "loss": 0.4523, "step": 57565 }, { "epoch": 2.69, "learning_rate": 2.470425787527499e-05, "loss": 0.0439, "step": 57570 }, { "epoch": 2.69, "learning_rate": 2.470378980543897e-05, "loss": 0.0282, "step": 57575 }, { "epoch": 2.69, "learning_rate": 2.4703321735602953e-05, "loss": 0.0716, "step": 57580 }, { "epoch": 2.69, "learning_rate": 2.4702853665766933e-05, "loss": 0.0731, "step": 57585 }, { "epoch": 2.69, "learning_rate": 2.4702385595930912e-05, "loss": 0.1536, "step": 57590 }, { "epoch": 2.69, "learning_rate": 2.4701917526094892e-05, "loss": 0.1643, "step": 57595 }, { "epoch": 2.69, "learning_rate": 2.4701449456258875e-05, "loss": 0.0649, "step": 57600 }, { "epoch": 2.69, "learning_rate": 2.4700981386422855e-05, "loss": 0.2279, "step": 57605 }, { "epoch": 2.69, "learning_rate": 2.4700513316586835e-05, "loss": 0.185, "step": 57610 }, { "epoch": 2.69, "learning_rate": 2.4700045246750815e-05, "loss": 0.1194, "step": 57615 }, { "epoch": 2.69, "learning_rate": 2.4699577176914798e-05, "loss": 0.044, "step": 57620 }, { "epoch": 2.69, "learning_rate": 2.4699109107078778e-05, "loss": 0.0205, "step": 57625 }, { "epoch": 2.69, "learning_rate": 2.4698641037242758e-05, "loss": 0.058, "step": 57630 }, { "epoch": 2.69, "learning_rate": 2.4698172967406738e-05, "loss": 0.0455, "step": 57635 }, { "epoch": 2.69, "learning_rate": 2.4697704897570717e-05, "loss": 0.0448, "step": 57640 }, { "epoch": 2.69, "learning_rate": 2.4697236827734697e-05, "loss": 0.1466, "step": 57645 }, { "epoch": 2.69, "learning_rate": 2.4696768757898677e-05, "loss": 0.1105, "step": 57650 }, { "epoch": 2.69, "learning_rate": 2.469630068806266e-05, "loss": 0.4656, "step": 57655 }, { "epoch": 2.69, "learning_rate": 2.469583261822664e-05, "loss": 0.24, "step": 57660 }, { "epoch": 2.69, "learning_rate": 2.469536454839062e-05, "loss": 0.2053, "step": 57665 }, { "epoch": 2.69, "learning_rate": 2.46948964785546e-05, "loss": 0.0614, "step": 57670 }, { "epoch": 2.69, "learning_rate": 2.4694428408718583e-05, "loss": 0.0198, "step": 57675 }, { "epoch": 2.69, "learning_rate": 2.4693960338882563e-05, "loss": 0.0258, "step": 57680 }, { "epoch": 2.69, "learning_rate": 2.4693492269046543e-05, "loss": 0.0339, "step": 57685 }, { "epoch": 2.69, "learning_rate": 2.4693024199210526e-05, "loss": 0.0708, "step": 57690 }, { "epoch": 2.69, "learning_rate": 2.4692556129374506e-05, "loss": 0.1115, "step": 57695 }, { "epoch": 2.69, "learning_rate": 2.4692088059538482e-05, "loss": 0.0644, "step": 57700 }, { "epoch": 2.69, "learning_rate": 2.4691619989702462e-05, "loss": 0.1303, "step": 57705 }, { "epoch": 2.69, "learning_rate": 2.4691151919866445e-05, "loss": 0.2525, "step": 57710 }, { "epoch": 2.69, "learning_rate": 2.4690683850030425e-05, "loss": 0.1985, "step": 57715 }, { "epoch": 2.69, "learning_rate": 2.4690215780194405e-05, "loss": 0.0156, "step": 57720 }, { "epoch": 2.69, "learning_rate": 2.4689747710358384e-05, "loss": 0.0468, "step": 57725 }, { "epoch": 2.69, "learning_rate": 2.4689279640522368e-05, "loss": 0.0986, "step": 57730 }, { "epoch": 2.69, "learning_rate": 2.4688811570686348e-05, "loss": 0.092, "step": 57735 }, { "epoch": 2.69, "learning_rate": 2.4688343500850327e-05, "loss": 0.1119, "step": 57740 }, { "epoch": 2.69, "learning_rate": 2.468787543101431e-05, "loss": 0.0664, "step": 57745 }, { "epoch": 2.69, "learning_rate": 2.468740736117829e-05, "loss": 0.14, "step": 57750 }, { "epoch": 2.69, "learning_rate": 2.468693929134227e-05, "loss": 0.1064, "step": 57755 }, { "epoch": 2.7, "learning_rate": 2.4686471221506247e-05, "loss": 0.2536, "step": 57760 }, { "epoch": 2.7, "learning_rate": 2.468600315167023e-05, "loss": 0.1373, "step": 57765 }, { "epoch": 2.7, "learning_rate": 2.468553508183421e-05, "loss": 0.0581, "step": 57770 }, { "epoch": 2.7, "learning_rate": 2.468506701199819e-05, "loss": 0.07, "step": 57775 }, { "epoch": 2.7, "learning_rate": 2.468459894216217e-05, "loss": 0.0596, "step": 57780 }, { "epoch": 2.7, "learning_rate": 2.4684130872326152e-05, "loss": 0.056, "step": 57785 }, { "epoch": 2.7, "learning_rate": 2.4683662802490132e-05, "loss": 0.1122, "step": 57790 }, { "epoch": 2.7, "learning_rate": 2.4683194732654112e-05, "loss": 0.0833, "step": 57795 }, { "epoch": 2.7, "learning_rate": 2.4682726662818092e-05, "loss": 0.0884, "step": 57800 }, { "epoch": 2.7, "learning_rate": 2.4682258592982075e-05, "loss": 0.1954, "step": 57805 }, { "epoch": 2.7, "learning_rate": 2.4681790523146055e-05, "loss": 0.2491, "step": 57810 }, { "epoch": 2.7, "learning_rate": 2.4681322453310035e-05, "loss": 0.2213, "step": 57815 }, { "epoch": 2.7, "learning_rate": 2.4680854383474018e-05, "loss": 0.0237, "step": 57820 }, { "epoch": 2.7, "learning_rate": 2.4680386313637994e-05, "loss": 0.0484, "step": 57825 }, { "epoch": 2.7, "learning_rate": 2.4679918243801974e-05, "loss": 0.0398, "step": 57830 }, { "epoch": 2.7, "learning_rate": 2.4679450173965954e-05, "loss": 0.0586, "step": 57835 }, { "epoch": 2.7, "learning_rate": 2.4678982104129937e-05, "loss": 0.0567, "step": 57840 }, { "epoch": 2.7, "learning_rate": 2.4678514034293917e-05, "loss": 0.0963, "step": 57845 }, { "epoch": 2.7, "learning_rate": 2.4678045964457897e-05, "loss": 0.0822, "step": 57850 }, { "epoch": 2.7, "learning_rate": 2.4677577894621877e-05, "loss": 0.1504, "step": 57855 }, { "epoch": 2.7, "learning_rate": 2.467710982478586e-05, "loss": 0.2156, "step": 57860 }, { "epoch": 2.7, "learning_rate": 2.467664175494984e-05, "loss": 0.2169, "step": 57865 }, { "epoch": 2.7, "learning_rate": 2.467617368511382e-05, "loss": 0.011, "step": 57870 }, { "epoch": 2.7, "learning_rate": 2.4675705615277803e-05, "loss": 0.0415, "step": 57875 }, { "epoch": 2.7, "learning_rate": 2.4675237545441783e-05, "loss": 0.0223, "step": 57880 }, { "epoch": 2.7, "learning_rate": 2.4674769475605762e-05, "loss": 0.083, "step": 57885 }, { "epoch": 2.7, "learning_rate": 2.467430140576974e-05, "loss": 0.0573, "step": 57890 }, { "epoch": 2.7, "learning_rate": 2.4673833335933722e-05, "loss": 0.0796, "step": 57895 }, { "epoch": 2.7, "learning_rate": 2.4673365266097702e-05, "loss": 0.1186, "step": 57900 }, { "epoch": 2.7, "learning_rate": 2.467289719626168e-05, "loss": 0.1458, "step": 57905 }, { "epoch": 2.7, "learning_rate": 2.467242912642566e-05, "loss": 0.1805, "step": 57910 }, { "epoch": 2.7, "learning_rate": 2.4671961056589645e-05, "loss": 0.4035, "step": 57915 }, { "epoch": 2.7, "learning_rate": 2.4671492986753624e-05, "loss": 0.0672, "step": 57920 }, { "epoch": 2.7, "learning_rate": 2.4671024916917604e-05, "loss": 0.058, "step": 57925 }, { "epoch": 2.7, "learning_rate": 2.4670556847081587e-05, "loss": 0.0408, "step": 57930 }, { "epoch": 2.7, "learning_rate": 2.4670088777245567e-05, "loss": 0.0546, "step": 57935 }, { "epoch": 2.7, "learning_rate": 2.4669620707409547e-05, "loss": 0.1206, "step": 57940 }, { "epoch": 2.7, "learning_rate": 2.4669152637573527e-05, "loss": 0.0743, "step": 57945 }, { "epoch": 2.7, "learning_rate": 2.4668684567737507e-05, "loss": 0.1345, "step": 57950 }, { "epoch": 2.7, "learning_rate": 2.4668216497901487e-05, "loss": 0.1159, "step": 57955 }, { "epoch": 2.7, "learning_rate": 2.4667748428065466e-05, "loss": 0.3535, "step": 57960 }, { "epoch": 2.7, "learning_rate": 2.4667280358229446e-05, "loss": 0.3006, "step": 57965 }, { "epoch": 2.7, "learning_rate": 2.466681228839343e-05, "loss": 0.0226, "step": 57970 }, { "epoch": 2.71, "learning_rate": 2.466634421855741e-05, "loss": 0.0352, "step": 57975 }, { "epoch": 2.71, "learning_rate": 2.466587614872139e-05, "loss": 0.0564, "step": 57980 }, { "epoch": 2.71, "learning_rate": 2.4665408078885372e-05, "loss": 0.0693, "step": 57985 }, { "epoch": 2.71, "learning_rate": 2.4664940009049352e-05, "loss": 0.0672, "step": 57990 }, { "epoch": 2.71, "learning_rate": 2.4664471939213332e-05, "loss": 0.1035, "step": 57995 }, { "epoch": 2.71, "learning_rate": 2.4664003869377312e-05, "loss": 0.0663, "step": 58000 }, { "epoch": 2.71, "learning_rate": 2.4663535799541295e-05, "loss": 0.2174, "step": 58005 }, { "epoch": 2.71, "learning_rate": 2.4663067729705275e-05, "loss": 0.2538, "step": 58010 }, { "epoch": 2.71, "learning_rate": 2.466259965986925e-05, "loss": 0.189, "step": 58015 }, { "epoch": 2.71, "learning_rate": 2.466213159003323e-05, "loss": 0.0508, "step": 58020 }, { "epoch": 2.71, "learning_rate": 2.4661663520197214e-05, "loss": 0.0506, "step": 58025 }, { "epoch": 2.71, "learning_rate": 2.4661195450361194e-05, "loss": 0.0819, "step": 58030 }, { "epoch": 2.71, "learning_rate": 2.4660727380525174e-05, "loss": 0.1452, "step": 58035 }, { "epoch": 2.71, "learning_rate": 2.4660259310689154e-05, "loss": 0.0802, "step": 58040 }, { "epoch": 2.71, "learning_rate": 2.4659791240853137e-05, "loss": 0.1489, "step": 58045 }, { "epoch": 2.71, "learning_rate": 2.4659323171017117e-05, "loss": 0.1678, "step": 58050 }, { "epoch": 2.71, "learning_rate": 2.4658855101181096e-05, "loss": 0.1887, "step": 58055 }, { "epoch": 2.71, "learning_rate": 2.465838703134508e-05, "loss": 0.266, "step": 58060 }, { "epoch": 2.71, "learning_rate": 2.465791896150906e-05, "loss": 0.2013, "step": 58065 }, { "epoch": 2.71, "learning_rate": 2.465745089167304e-05, "loss": 0.0661, "step": 58070 }, { "epoch": 2.71, "learning_rate": 2.465698282183702e-05, "loss": 0.0563, "step": 58075 }, { "epoch": 2.71, "learning_rate": 2.4656514752001e-05, "loss": 0.0246, "step": 58080 }, { "epoch": 2.71, "learning_rate": 2.465604668216498e-05, "loss": 0.0701, "step": 58085 }, { "epoch": 2.71, "learning_rate": 2.465557861232896e-05, "loss": 0.1198, "step": 58090 }, { "epoch": 2.71, "learning_rate": 2.465511054249294e-05, "loss": 0.0721, "step": 58095 }, { "epoch": 2.71, "learning_rate": 2.465464247265692e-05, "loss": 0.1164, "step": 58100 }, { "epoch": 2.71, "learning_rate": 2.46541744028209e-05, "loss": 0.1727, "step": 58105 }, { "epoch": 2.71, "learning_rate": 2.465370633298488e-05, "loss": 0.3224, "step": 58110 }, { "epoch": 2.71, "learning_rate": 2.4653238263148864e-05, "loss": 0.2044, "step": 58115 }, { "epoch": 2.71, "learning_rate": 2.4652770193312844e-05, "loss": 0.0322, "step": 58120 }, { "epoch": 2.71, "learning_rate": 2.4652302123476824e-05, "loss": 0.0549, "step": 58125 }, { "epoch": 2.71, "learning_rate": 2.4651834053640804e-05, "loss": 0.067, "step": 58130 }, { "epoch": 2.71, "learning_rate": 2.4651365983804787e-05, "loss": 0.017, "step": 58135 }, { "epoch": 2.71, "learning_rate": 2.4650897913968764e-05, "loss": 0.1286, "step": 58140 }, { "epoch": 2.71, "learning_rate": 2.4650429844132743e-05, "loss": 0.1283, "step": 58145 }, { "epoch": 2.71, "learning_rate": 2.4649961774296723e-05, "loss": 0.137, "step": 58150 }, { "epoch": 2.71, "learning_rate": 2.4649493704460706e-05, "loss": 0.1427, "step": 58155 }, { "epoch": 2.71, "learning_rate": 2.4649025634624686e-05, "loss": 0.1582, "step": 58160 }, { "epoch": 2.71, "learning_rate": 2.4648557564788666e-05, "loss": 0.29, "step": 58165 }, { "epoch": 2.71, "learning_rate": 2.464808949495265e-05, "loss": 0.0159, "step": 58170 }, { "epoch": 2.71, "learning_rate": 2.464762142511663e-05, "loss": 0.0291, "step": 58175 }, { "epoch": 2.71, "learning_rate": 2.464715335528061e-05, "loss": 0.1181, "step": 58180 }, { "epoch": 2.71, "learning_rate": 2.464668528544459e-05, "loss": 0.0576, "step": 58185 }, { "epoch": 2.72, "learning_rate": 2.4646217215608572e-05, "loss": 0.1022, "step": 58190 }, { "epoch": 2.72, "learning_rate": 2.4645749145772552e-05, "loss": 0.0577, "step": 58195 }, { "epoch": 2.72, "learning_rate": 2.464528107593653e-05, "loss": 0.0849, "step": 58200 }, { "epoch": 2.72, "learning_rate": 2.4644813006100508e-05, "loss": 0.0814, "step": 58205 }, { "epoch": 2.72, "learning_rate": 2.464434493626449e-05, "loss": 0.178, "step": 58210 }, { "epoch": 2.72, "learning_rate": 2.464387686642847e-05, "loss": 0.2908, "step": 58215 }, { "epoch": 2.72, "learning_rate": 2.464340879659245e-05, "loss": 0.0314, "step": 58220 }, { "epoch": 2.72, "learning_rate": 2.464294072675643e-05, "loss": 0.0343, "step": 58225 }, { "epoch": 2.72, "learning_rate": 2.4642472656920414e-05, "loss": 0.0463, "step": 58230 }, { "epoch": 2.72, "learning_rate": 2.4642004587084394e-05, "loss": 0.0404, "step": 58235 }, { "epoch": 2.72, "learning_rate": 2.4641536517248373e-05, "loss": 0.0787, "step": 58240 }, { "epoch": 2.72, "learning_rate": 2.4641068447412357e-05, "loss": 0.1323, "step": 58245 }, { "epoch": 2.72, "learning_rate": 2.4640600377576336e-05, "loss": 0.1375, "step": 58250 }, { "epoch": 2.72, "learning_rate": 2.4640132307740316e-05, "loss": 0.1497, "step": 58255 }, { "epoch": 2.72, "learning_rate": 2.4639664237904296e-05, "loss": 0.3273, "step": 58260 }, { "epoch": 2.72, "learning_rate": 2.4639196168068276e-05, "loss": 0.3801, "step": 58265 }, { "epoch": 2.72, "learning_rate": 2.4638728098232256e-05, "loss": 0.0657, "step": 58270 }, { "epoch": 2.72, "learning_rate": 2.4638260028396236e-05, "loss": 0.0782, "step": 58275 }, { "epoch": 2.72, "learning_rate": 2.4637791958560215e-05, "loss": 0.0649, "step": 58280 }, { "epoch": 2.72, "learning_rate": 2.46373238887242e-05, "loss": 0.0892, "step": 58285 }, { "epoch": 2.72, "learning_rate": 2.463685581888818e-05, "loss": 0.0592, "step": 58290 }, { "epoch": 2.72, "learning_rate": 2.4636387749052158e-05, "loss": 0.0859, "step": 58295 }, { "epoch": 2.72, "learning_rate": 2.463591967921614e-05, "loss": 0.0975, "step": 58300 }, { "epoch": 2.72, "learning_rate": 2.463545160938012e-05, "loss": 0.1094, "step": 58305 }, { "epoch": 2.72, "learning_rate": 2.46349835395441e-05, "loss": 0.1775, "step": 58310 }, { "epoch": 2.72, "learning_rate": 2.463451546970808e-05, "loss": 0.3567, "step": 58315 }, { "epoch": 2.72, "learning_rate": 2.4634047399872064e-05, "loss": 0.0263, "step": 58320 }, { "epoch": 2.72, "learning_rate": 2.4633579330036044e-05, "loss": 0.0487, "step": 58325 }, { "epoch": 2.72, "learning_rate": 2.463311126020002e-05, "loss": 0.0568, "step": 58330 }, { "epoch": 2.72, "learning_rate": 2.4632643190364e-05, "loss": 0.0598, "step": 58335 }, { "epoch": 2.72, "learning_rate": 2.4632175120527983e-05, "loss": 0.0464, "step": 58340 }, { "epoch": 2.72, "learning_rate": 2.4631707050691963e-05, "loss": 0.1119, "step": 58345 }, { "epoch": 2.72, "learning_rate": 2.4631238980855943e-05, "loss": 0.0869, "step": 58350 }, { "epoch": 2.72, "learning_rate": 2.4630770911019926e-05, "loss": 0.1877, "step": 58355 }, { "epoch": 2.72, "learning_rate": 2.4630302841183906e-05, "loss": 0.2785, "step": 58360 }, { "epoch": 2.72, "learning_rate": 2.4629834771347886e-05, "loss": 0.2328, "step": 58365 }, { "epoch": 2.72, "learning_rate": 2.4629366701511866e-05, "loss": 0.0259, "step": 58370 }, { "epoch": 2.72, "learning_rate": 2.462889863167585e-05, "loss": 0.0814, "step": 58375 }, { "epoch": 2.72, "learning_rate": 2.462843056183983e-05, "loss": 0.04, "step": 58380 }, { "epoch": 2.72, "learning_rate": 2.462796249200381e-05, "loss": 0.0433, "step": 58385 }, { "epoch": 2.72, "learning_rate": 2.462749442216779e-05, "loss": 0.0543, "step": 58390 }, { "epoch": 2.72, "learning_rate": 2.4627026352331768e-05, "loss": 0.129, "step": 58395 }, { "epoch": 2.73, "learning_rate": 2.4626558282495748e-05, "loss": 0.1634, "step": 58400 }, { "epoch": 2.73, "learning_rate": 2.4626090212659728e-05, "loss": 0.1017, "step": 58405 }, { "epoch": 2.73, "learning_rate": 2.4625622142823708e-05, "loss": 0.3119, "step": 58410 }, { "epoch": 2.73, "learning_rate": 2.462515407298769e-05, "loss": 0.2509, "step": 58415 }, { "epoch": 2.73, "learning_rate": 2.462468600315167e-05, "loss": 0.0798, "step": 58420 }, { "epoch": 2.73, "learning_rate": 2.462421793331565e-05, "loss": 0.0665, "step": 58425 }, { "epoch": 2.73, "learning_rate": 2.4623749863479634e-05, "loss": 0.031, "step": 58430 }, { "epoch": 2.73, "learning_rate": 2.4623281793643613e-05, "loss": 0.0664, "step": 58435 }, { "epoch": 2.73, "learning_rate": 2.4622813723807593e-05, "loss": 0.0758, "step": 58440 }, { "epoch": 2.73, "learning_rate": 2.4622345653971573e-05, "loss": 0.1216, "step": 58445 }, { "epoch": 2.73, "learning_rate": 2.4621877584135556e-05, "loss": 0.1148, "step": 58450 }, { "epoch": 2.73, "learning_rate": 2.4621409514299533e-05, "loss": 0.2464, "step": 58455 }, { "epoch": 2.73, "learning_rate": 2.4620941444463513e-05, "loss": 0.2974, "step": 58460 }, { "epoch": 2.73, "learning_rate": 2.4620473374627492e-05, "loss": 0.3838, "step": 58465 }, { "epoch": 2.73, "learning_rate": 2.4620005304791476e-05, "loss": 0.0361, "step": 58470 }, { "epoch": 2.73, "learning_rate": 2.4619537234955455e-05, "loss": 0.0284, "step": 58475 }, { "epoch": 2.73, "learning_rate": 2.4619069165119435e-05, "loss": 0.0614, "step": 58480 }, { "epoch": 2.73, "learning_rate": 2.461860109528342e-05, "loss": 0.0618, "step": 58485 }, { "epoch": 2.73, "learning_rate": 2.4618133025447398e-05, "loss": 0.0829, "step": 58490 }, { "epoch": 2.73, "learning_rate": 2.4617664955611378e-05, "loss": 0.0779, "step": 58495 }, { "epoch": 2.73, "learning_rate": 2.4617196885775358e-05, "loss": 0.111, "step": 58500 }, { "epoch": 2.73, "learning_rate": 2.461672881593934e-05, "loss": 0.0683, "step": 58505 }, { "epoch": 2.73, "learning_rate": 2.461626074610332e-05, "loss": 0.2439, "step": 58510 }, { "epoch": 2.73, "learning_rate": 2.46157926762673e-05, "loss": 0.2115, "step": 58515 }, { "epoch": 2.73, "learning_rate": 2.4615324606431277e-05, "loss": 0.0217, "step": 58520 }, { "epoch": 2.73, "learning_rate": 2.461485653659526e-05, "loss": 0.0406, "step": 58525 }, { "epoch": 2.73, "learning_rate": 2.461438846675924e-05, "loss": 0.0669, "step": 58530 }, { "epoch": 2.73, "learning_rate": 2.461392039692322e-05, "loss": 0.0582, "step": 58535 }, { "epoch": 2.73, "learning_rate": 2.4613452327087203e-05, "loss": 0.1054, "step": 58540 }, { "epoch": 2.73, "learning_rate": 2.4612984257251183e-05, "loss": 0.1727, "step": 58545 }, { "epoch": 2.73, "learning_rate": 2.4612516187415163e-05, "loss": 0.1682, "step": 58550 }, { "epoch": 2.73, "learning_rate": 2.4612048117579143e-05, "loss": 0.1315, "step": 58555 }, { "epoch": 2.73, "learning_rate": 2.4611580047743126e-05, "loss": 0.2927, "step": 58560 }, { "epoch": 2.73, "learning_rate": 2.4611111977907106e-05, "loss": 0.371, "step": 58565 }, { "epoch": 2.73, "learning_rate": 2.4610643908071085e-05, "loss": 0.0317, "step": 58570 }, { "epoch": 2.73, "learning_rate": 2.4610175838235065e-05, "loss": 0.0541, "step": 58575 }, { "epoch": 2.73, "learning_rate": 2.4609707768399045e-05, "loss": 0.0844, "step": 58580 }, { "epoch": 2.73, "learning_rate": 2.4609239698563025e-05, "loss": 0.1184, "step": 58585 }, { "epoch": 2.73, "learning_rate": 2.4608771628727005e-05, "loss": 0.0605, "step": 58590 }, { "epoch": 2.73, "learning_rate": 2.4608303558890988e-05, "loss": 0.0964, "step": 58595 }, { "epoch": 2.73, "learning_rate": 2.4607835489054968e-05, "loss": 0.1134, "step": 58600 }, { "epoch": 2.73, "learning_rate": 2.4607367419218948e-05, "loss": 0.0912, "step": 58605 }, { "epoch": 2.73, "learning_rate": 2.4606899349382927e-05, "loss": 0.2443, "step": 58610 }, { "epoch": 2.74, "learning_rate": 2.460643127954691e-05, "loss": 0.3277, "step": 58615 }, { "epoch": 2.74, "learning_rate": 2.460596320971089e-05, "loss": 0.0408, "step": 58620 }, { "epoch": 2.74, "learning_rate": 2.460549513987487e-05, "loss": 0.0378, "step": 58625 }, { "epoch": 2.74, "learning_rate": 2.460502707003885e-05, "loss": 0.0622, "step": 58630 }, { "epoch": 2.74, "learning_rate": 2.4604559000202833e-05, "loss": 0.0432, "step": 58635 }, { "epoch": 2.74, "learning_rate": 2.4604090930366813e-05, "loss": 0.0413, "step": 58640 }, { "epoch": 2.74, "learning_rate": 2.460362286053079e-05, "loss": 0.1034, "step": 58645 }, { "epoch": 2.74, "learning_rate": 2.460315479069477e-05, "loss": 0.0883, "step": 58650 }, { "epoch": 2.74, "learning_rate": 2.4602686720858753e-05, "loss": 0.1961, "step": 58655 }, { "epoch": 2.74, "learning_rate": 2.4602218651022732e-05, "loss": 0.2413, "step": 58660 }, { "epoch": 2.74, "learning_rate": 2.4601750581186712e-05, "loss": 0.3507, "step": 58665 }, { "epoch": 2.74, "learning_rate": 2.4601282511350695e-05, "loss": 0.0091, "step": 58670 }, { "epoch": 2.74, "learning_rate": 2.4600814441514675e-05, "loss": 0.0602, "step": 58675 }, { "epoch": 2.74, "learning_rate": 2.4600346371678655e-05, "loss": 0.0563, "step": 58680 }, { "epoch": 2.74, "learning_rate": 2.4599878301842635e-05, "loss": 0.0797, "step": 58685 }, { "epoch": 2.74, "learning_rate": 2.4599410232006618e-05, "loss": 0.0701, "step": 58690 }, { "epoch": 2.74, "learning_rate": 2.4598942162170598e-05, "loss": 0.1964, "step": 58695 }, { "epoch": 2.74, "learning_rate": 2.4598474092334578e-05, "loss": 0.1213, "step": 58700 }, { "epoch": 2.74, "learning_rate": 2.4598006022498557e-05, "loss": 0.0986, "step": 58705 }, { "epoch": 2.74, "learning_rate": 2.4597537952662537e-05, "loss": 0.208, "step": 58710 }, { "epoch": 2.74, "learning_rate": 2.4597069882826517e-05, "loss": 0.3041, "step": 58715 }, { "epoch": 2.74, "learning_rate": 2.4596601812990497e-05, "loss": 0.0336, "step": 58720 }, { "epoch": 2.74, "learning_rate": 2.459613374315448e-05, "loss": 0.0485, "step": 58725 }, { "epoch": 2.74, "learning_rate": 2.459566567331846e-05, "loss": 0.0615, "step": 58730 }, { "epoch": 2.74, "learning_rate": 2.459519760348244e-05, "loss": 0.0847, "step": 58735 }, { "epoch": 2.74, "learning_rate": 2.459472953364642e-05, "loss": 0.0681, "step": 58740 }, { "epoch": 2.74, "learning_rate": 2.4594261463810403e-05, "loss": 0.0926, "step": 58745 }, { "epoch": 2.74, "learning_rate": 2.4593793393974383e-05, "loss": 0.1255, "step": 58750 }, { "epoch": 2.74, "learning_rate": 2.4593325324138362e-05, "loss": 0.1154, "step": 58755 }, { "epoch": 2.74, "learning_rate": 2.4592857254302342e-05, "loss": 0.322, "step": 58760 }, { "epoch": 2.74, "learning_rate": 2.4592389184466325e-05, "loss": 0.2027, "step": 58765 }, { "epoch": 2.74, "learning_rate": 2.4591921114630302e-05, "loss": 0.0245, "step": 58770 }, { "epoch": 2.74, "learning_rate": 2.4591453044794282e-05, "loss": 0.0577, "step": 58775 }, { "epoch": 2.74, "learning_rate": 2.4590984974958265e-05, "loss": 0.0626, "step": 58780 }, { "epoch": 2.74, "learning_rate": 2.4590516905122245e-05, "loss": 0.1288, "step": 58785 }, { "epoch": 2.74, "learning_rate": 2.4590048835286225e-05, "loss": 0.1014, "step": 58790 }, { "epoch": 2.74, "learning_rate": 2.4589580765450204e-05, "loss": 0.186, "step": 58795 }, { "epoch": 2.74, "learning_rate": 2.4589112695614188e-05, "loss": 0.1054, "step": 58800 }, { "epoch": 2.74, "learning_rate": 2.4588644625778167e-05, "loss": 0.0748, "step": 58805 }, { "epoch": 2.74, "learning_rate": 2.4588176555942147e-05, "loss": 0.258, "step": 58810 }, { "epoch": 2.74, "learning_rate": 2.4587708486106127e-05, "loss": 0.231, "step": 58815 }, { "epoch": 2.74, "learning_rate": 2.458724041627011e-05, "loss": 0.0669, "step": 58820 }, { "epoch": 2.74, "learning_rate": 2.458677234643409e-05, "loss": 0.0449, "step": 58825 }, { "epoch": 2.75, "learning_rate": 2.458630427659807e-05, "loss": 0.0623, "step": 58830 }, { "epoch": 2.75, "learning_rate": 2.4585836206762046e-05, "loss": 0.0469, "step": 58835 }, { "epoch": 2.75, "learning_rate": 2.458536813692603e-05, "loss": 0.104, "step": 58840 }, { "epoch": 2.75, "learning_rate": 2.458490006709001e-05, "loss": 0.0931, "step": 58845 }, { "epoch": 2.75, "learning_rate": 2.458443199725399e-05, "loss": 0.1118, "step": 58850 }, { "epoch": 2.75, "learning_rate": 2.4583963927417972e-05, "loss": 0.2072, "step": 58855 }, { "epoch": 2.75, "learning_rate": 2.4583495857581952e-05, "loss": 0.1836, "step": 58860 }, { "epoch": 2.75, "learning_rate": 2.4583027787745932e-05, "loss": 0.3149, "step": 58865 }, { "epoch": 2.75, "learning_rate": 2.4582559717909912e-05, "loss": 0.0274, "step": 58870 }, { "epoch": 2.75, "learning_rate": 2.4582091648073895e-05, "loss": 0.0136, "step": 58875 }, { "epoch": 2.75, "learning_rate": 2.4581623578237875e-05, "loss": 0.1211, "step": 58880 }, { "epoch": 2.75, "learning_rate": 2.4581155508401855e-05, "loss": 0.0906, "step": 58885 }, { "epoch": 2.75, "learning_rate": 2.4580687438565834e-05, "loss": 0.0337, "step": 58890 }, { "epoch": 2.75, "learning_rate": 2.4580219368729818e-05, "loss": 0.0511, "step": 58895 }, { "epoch": 2.75, "learning_rate": 2.4579751298893794e-05, "loss": 0.1626, "step": 58900 }, { "epoch": 2.75, "learning_rate": 2.4579283229057774e-05, "loss": 0.0778, "step": 58905 }, { "epoch": 2.75, "learning_rate": 2.4578815159221757e-05, "loss": 0.1925, "step": 58910 }, { "epoch": 2.75, "learning_rate": 2.4578347089385737e-05, "loss": 0.3855, "step": 58915 }, { "epoch": 2.75, "learning_rate": 2.4577879019549717e-05, "loss": 0.0631, "step": 58920 }, { "epoch": 2.75, "learning_rate": 2.4577410949713697e-05, "loss": 0.0465, "step": 58925 }, { "epoch": 2.75, "learning_rate": 2.457694287987768e-05, "loss": 0.0557, "step": 58930 }, { "epoch": 2.75, "learning_rate": 2.457647481004166e-05, "loss": 0.0435, "step": 58935 }, { "epoch": 2.75, "learning_rate": 2.457600674020564e-05, "loss": 0.103, "step": 58940 }, { "epoch": 2.75, "learning_rate": 2.457553867036962e-05, "loss": 0.0684, "step": 58945 }, { "epoch": 2.75, "learning_rate": 2.4575070600533602e-05, "loss": 0.1101, "step": 58950 }, { "epoch": 2.75, "learning_rate": 2.4574602530697582e-05, "loss": 0.1637, "step": 58955 }, { "epoch": 2.75, "learning_rate": 2.457413446086156e-05, "loss": 0.1489, "step": 58960 }, { "epoch": 2.75, "learning_rate": 2.4573666391025542e-05, "loss": 0.2171, "step": 58965 }, { "epoch": 2.75, "learning_rate": 2.4573198321189522e-05, "loss": 0.0703, "step": 58970 }, { "epoch": 2.75, "learning_rate": 2.45727302513535e-05, "loss": 0.0522, "step": 58975 }, { "epoch": 2.75, "learning_rate": 2.457226218151748e-05, "loss": 0.0429, "step": 58980 }, { "epoch": 2.75, "learning_rate": 2.4571794111681465e-05, "loss": 0.1265, "step": 58985 }, { "epoch": 2.75, "learning_rate": 2.4571326041845444e-05, "loss": 0.1143, "step": 58990 }, { "epoch": 2.75, "learning_rate": 2.4570857972009424e-05, "loss": 0.0375, "step": 58995 }, { "epoch": 2.75, "learning_rate": 2.4570389902173404e-05, "loss": 0.0493, "step": 59000 }, { "epoch": 2.75, "learning_rate": 2.4569921832337387e-05, "loss": 0.1706, "step": 59005 }, { "epoch": 2.75, "learning_rate": 2.4569453762501367e-05, "loss": 0.2414, "step": 59010 }, { "epoch": 2.75, "learning_rate": 2.4568985692665347e-05, "loss": 0.2894, "step": 59015 }, { "epoch": 2.75, "learning_rate": 2.4568517622829327e-05, "loss": 0.0499, "step": 59020 }, { "epoch": 2.75, "learning_rate": 2.4568049552993306e-05, "loss": 0.0648, "step": 59025 }, { "epoch": 2.75, "learning_rate": 2.4567581483157286e-05, "loss": 0.0364, "step": 59030 }, { "epoch": 2.75, "learning_rate": 2.4567113413321266e-05, "loss": 0.078, "step": 59035 }, { "epoch": 2.75, "learning_rate": 2.456664534348525e-05, "loss": 0.0685, "step": 59040 }, { "epoch": 2.76, "learning_rate": 2.456617727364923e-05, "loss": 0.0632, "step": 59045 }, { "epoch": 2.76, "learning_rate": 2.456570920381321e-05, "loss": 0.0546, "step": 59050 }, { "epoch": 2.76, "learning_rate": 2.456524113397719e-05, "loss": 0.2355, "step": 59055 }, { "epoch": 2.76, "learning_rate": 2.4564773064141172e-05, "loss": 0.2846, "step": 59060 }, { "epoch": 2.76, "learning_rate": 2.4564304994305152e-05, "loss": 0.2363, "step": 59065 }, { "epoch": 2.76, "learning_rate": 2.456383692446913e-05, "loss": 0.055, "step": 59070 }, { "epoch": 2.76, "learning_rate": 2.456336885463311e-05, "loss": 0.0605, "step": 59075 }, { "epoch": 2.76, "learning_rate": 2.4562900784797095e-05, "loss": 0.0555, "step": 59080 }, { "epoch": 2.76, "learning_rate": 2.4562432714961074e-05, "loss": 0.0546, "step": 59085 }, { "epoch": 2.76, "learning_rate": 2.456196464512505e-05, "loss": 0.1494, "step": 59090 }, { "epoch": 2.76, "learning_rate": 2.4561496575289034e-05, "loss": 0.1066, "step": 59095 }, { "epoch": 2.76, "learning_rate": 2.4561028505453014e-05, "loss": 0.0528, "step": 59100 }, { "epoch": 2.76, "learning_rate": 2.4560560435616994e-05, "loss": 0.1629, "step": 59105 }, { "epoch": 2.76, "learning_rate": 2.4560092365780974e-05, "loss": 0.3085, "step": 59110 }, { "epoch": 2.76, "learning_rate": 2.4559624295944957e-05, "loss": 0.3952, "step": 59115 }, { "epoch": 2.76, "learning_rate": 2.4559156226108937e-05, "loss": 0.0439, "step": 59120 }, { "epoch": 2.76, "learning_rate": 2.4558688156272916e-05, "loss": 0.0105, "step": 59125 }, { "epoch": 2.76, "learning_rate": 2.4558220086436896e-05, "loss": 0.0264, "step": 59130 }, { "epoch": 2.76, "learning_rate": 2.455775201660088e-05, "loss": 0.0578, "step": 59135 }, { "epoch": 2.76, "learning_rate": 2.455728394676486e-05, "loss": 0.0735, "step": 59140 }, { "epoch": 2.76, "learning_rate": 2.455681587692884e-05, "loss": 0.0657, "step": 59145 }, { "epoch": 2.76, "learning_rate": 2.455634780709282e-05, "loss": 0.1978, "step": 59150 }, { "epoch": 2.76, "learning_rate": 2.45558797372568e-05, "loss": 0.0991, "step": 59155 }, { "epoch": 2.76, "learning_rate": 2.455541166742078e-05, "loss": 0.2722, "step": 59160 }, { "epoch": 2.76, "learning_rate": 2.455494359758476e-05, "loss": 0.2074, "step": 59165 }, { "epoch": 2.76, "learning_rate": 2.455447552774874e-05, "loss": 0.0808, "step": 59170 }, { "epoch": 2.76, "learning_rate": 2.455400745791272e-05, "loss": 0.0249, "step": 59175 }, { "epoch": 2.76, "learning_rate": 2.45535393880767e-05, "loss": 0.0193, "step": 59180 }, { "epoch": 2.76, "learning_rate": 2.455307131824068e-05, "loss": 0.1137, "step": 59185 }, { "epoch": 2.76, "learning_rate": 2.4552603248404664e-05, "loss": 0.062, "step": 59190 }, { "epoch": 2.76, "learning_rate": 2.4552135178568644e-05, "loss": 0.0972, "step": 59195 }, { "epoch": 2.76, "learning_rate": 2.4551667108732624e-05, "loss": 0.1854, "step": 59200 }, { "epoch": 2.76, "learning_rate": 2.4551199038896604e-05, "loss": 0.1432, "step": 59205 }, { "epoch": 2.76, "learning_rate": 2.4550730969060587e-05, "loss": 0.2352, "step": 59210 }, { "epoch": 2.76, "learning_rate": 2.4550262899224563e-05, "loss": 0.2308, "step": 59215 }, { "epoch": 2.76, "learning_rate": 2.4549794829388543e-05, "loss": 0.0405, "step": 59220 }, { "epoch": 2.76, "learning_rate": 2.4549326759552526e-05, "loss": 0.059, "step": 59225 }, { "epoch": 2.76, "learning_rate": 2.4548858689716506e-05, "loss": 0.0529, "step": 59230 }, { "epoch": 2.76, "learning_rate": 2.4548390619880486e-05, "loss": 0.0721, "step": 59235 }, { "epoch": 2.76, "learning_rate": 2.4547922550044466e-05, "loss": 0.1207, "step": 59240 }, { "epoch": 2.76, "learning_rate": 2.454745448020845e-05, "loss": 0.0641, "step": 59245 }, { "epoch": 2.76, "learning_rate": 2.454698641037243e-05, "loss": 0.0872, "step": 59250 }, { "epoch": 2.76, "learning_rate": 2.454651834053641e-05, "loss": 0.1695, "step": 59255 }, { "epoch": 2.77, "learning_rate": 2.454605027070039e-05, "loss": 0.2749, "step": 59260 }, { "epoch": 2.77, "learning_rate": 2.454558220086437e-05, "loss": 0.3331, "step": 59265 }, { "epoch": 2.77, "learning_rate": 2.454511413102835e-05, "loss": 0.0622, "step": 59270 }, { "epoch": 2.77, "learning_rate": 2.454464606119233e-05, "loss": 0.0229, "step": 59275 }, { "epoch": 2.77, "learning_rate": 2.454417799135631e-05, "loss": 0.0383, "step": 59280 }, { "epoch": 2.77, "learning_rate": 2.454370992152029e-05, "loss": 0.0859, "step": 59285 }, { "epoch": 2.77, "learning_rate": 2.454324185168427e-05, "loss": 0.0216, "step": 59290 }, { "epoch": 2.77, "learning_rate": 2.454277378184825e-05, "loss": 0.077, "step": 59295 }, { "epoch": 2.77, "learning_rate": 2.4542305712012234e-05, "loss": 0.1276, "step": 59300 }, { "epoch": 2.77, "learning_rate": 2.4541837642176214e-05, "loss": 0.1631, "step": 59305 }, { "epoch": 2.77, "learning_rate": 2.4541369572340193e-05, "loss": 0.2293, "step": 59310 }, { "epoch": 2.77, "learning_rate": 2.4540901502504173e-05, "loss": 0.2981, "step": 59315 }, { "epoch": 2.77, "learning_rate": 2.4540433432668156e-05, "loss": 0.0204, "step": 59320 }, { "epoch": 2.77, "learning_rate": 2.4539965362832136e-05, "loss": 0.0446, "step": 59325 }, { "epoch": 2.77, "learning_rate": 2.4539497292996116e-05, "loss": 0.1101, "step": 59330 }, { "epoch": 2.77, "learning_rate": 2.45390292231601e-05, "loss": 0.0811, "step": 59335 }, { "epoch": 2.77, "learning_rate": 2.4538561153324076e-05, "loss": 0.0517, "step": 59340 }, { "epoch": 2.77, "learning_rate": 2.4538093083488055e-05, "loss": 0.0766, "step": 59345 }, { "epoch": 2.77, "learning_rate": 2.4537625013652035e-05, "loss": 0.1947, "step": 59350 }, { "epoch": 2.77, "learning_rate": 2.453715694381602e-05, "loss": 0.1056, "step": 59355 }, { "epoch": 2.77, "learning_rate": 2.453668887398e-05, "loss": 0.23, "step": 59360 }, { "epoch": 2.77, "learning_rate": 2.4536220804143978e-05, "loss": 0.3537, "step": 59365 }, { "epoch": 2.77, "learning_rate": 2.4535752734307958e-05, "loss": 0.0398, "step": 59370 }, { "epoch": 2.77, "learning_rate": 2.453528466447194e-05, "loss": 0.0274, "step": 59375 }, { "epoch": 2.77, "learning_rate": 2.453481659463592e-05, "loss": 0.0728, "step": 59380 }, { "epoch": 2.77, "learning_rate": 2.45343485247999e-05, "loss": 0.0965, "step": 59385 }, { "epoch": 2.77, "learning_rate": 2.4533880454963884e-05, "loss": 0.0684, "step": 59390 }, { "epoch": 2.77, "learning_rate": 2.4533412385127864e-05, "loss": 0.0728, "step": 59395 }, { "epoch": 2.77, "learning_rate": 2.4532944315291844e-05, "loss": 0.1069, "step": 59400 }, { "epoch": 2.77, "learning_rate": 2.453247624545582e-05, "loss": 0.2129, "step": 59405 }, { "epoch": 2.77, "learning_rate": 2.4532008175619803e-05, "loss": 0.2177, "step": 59410 }, { "epoch": 2.77, "learning_rate": 2.4531540105783783e-05, "loss": 0.2914, "step": 59415 }, { "epoch": 2.77, "learning_rate": 2.4531072035947763e-05, "loss": 0.1037, "step": 59420 }, { "epoch": 2.77, "learning_rate": 2.4530603966111743e-05, "loss": 0.0971, "step": 59425 }, { "epoch": 2.77, "learning_rate": 2.4530135896275726e-05, "loss": 0.0731, "step": 59430 }, { "epoch": 2.77, "learning_rate": 2.4529667826439706e-05, "loss": 0.081, "step": 59435 }, { "epoch": 2.77, "learning_rate": 2.4529199756603686e-05, "loss": 0.1812, "step": 59440 }, { "epoch": 2.77, "learning_rate": 2.4528731686767665e-05, "loss": 0.0985, "step": 59445 }, { "epoch": 2.77, "learning_rate": 2.452826361693165e-05, "loss": 0.1389, "step": 59450 }, { "epoch": 2.77, "learning_rate": 2.452779554709563e-05, "loss": 0.2728, "step": 59455 }, { "epoch": 2.77, "learning_rate": 2.4527327477259608e-05, "loss": 0.1614, "step": 59460 }, { "epoch": 2.77, "learning_rate": 2.4526859407423588e-05, "loss": 0.3465, "step": 59465 }, { "epoch": 2.77, "learning_rate": 2.4526391337587568e-05, "loss": 0.0404, "step": 59470 }, { "epoch": 2.78, "learning_rate": 2.4525923267751548e-05, "loss": 0.0417, "step": 59475 }, { "epoch": 2.78, "learning_rate": 2.4525455197915527e-05, "loss": 0.1063, "step": 59480 }, { "epoch": 2.78, "learning_rate": 2.452498712807951e-05, "loss": 0.0506, "step": 59485 }, { "epoch": 2.78, "learning_rate": 2.452451905824349e-05, "loss": 0.0576, "step": 59490 }, { "epoch": 2.78, "learning_rate": 2.452405098840747e-05, "loss": 0.1029, "step": 59495 }, { "epoch": 2.78, "learning_rate": 2.452358291857145e-05, "loss": 0.1083, "step": 59500 }, { "epoch": 2.78, "learning_rate": 2.4523114848735433e-05, "loss": 0.1709, "step": 59505 }, { "epoch": 2.78, "learning_rate": 2.4522646778899413e-05, "loss": 0.3075, "step": 59510 }, { "epoch": 2.78, "learning_rate": 2.4522178709063393e-05, "loss": 0.2071, "step": 59515 }, { "epoch": 2.78, "learning_rate": 2.4521710639227376e-05, "loss": 0.0512, "step": 59520 }, { "epoch": 2.78, "learning_rate": 2.4521242569391356e-05, "loss": 0.0446, "step": 59525 }, { "epoch": 2.78, "learning_rate": 2.4520774499555332e-05, "loss": 0.0948, "step": 59530 }, { "epoch": 2.78, "learning_rate": 2.4520306429719312e-05, "loss": 0.0577, "step": 59535 }, { "epoch": 2.78, "learning_rate": 2.4519838359883295e-05, "loss": 0.0826, "step": 59540 }, { "epoch": 2.78, "learning_rate": 2.4519370290047275e-05, "loss": 0.0682, "step": 59545 }, { "epoch": 2.78, "learning_rate": 2.4518902220211255e-05, "loss": 0.1066, "step": 59550 }, { "epoch": 2.78, "learning_rate": 2.4518434150375235e-05, "loss": 0.1609, "step": 59555 }, { "epoch": 2.78, "learning_rate": 2.4517966080539218e-05, "loss": 0.1646, "step": 59560 }, { "epoch": 2.78, "learning_rate": 2.4517498010703198e-05, "loss": 0.2672, "step": 59565 }, { "epoch": 2.78, "learning_rate": 2.4517029940867178e-05, "loss": 0.1135, "step": 59570 }, { "epoch": 2.78, "learning_rate": 2.451656187103116e-05, "loss": 0.071, "step": 59575 }, { "epoch": 2.78, "learning_rate": 2.451609380119514e-05, "loss": 0.022, "step": 59580 }, { "epoch": 2.78, "learning_rate": 2.451562573135912e-05, "loss": 0.0553, "step": 59585 }, { "epoch": 2.78, "learning_rate": 2.45151576615231e-05, "loss": 0.1212, "step": 59590 }, { "epoch": 2.78, "learning_rate": 2.451468959168708e-05, "loss": 0.0703, "step": 59595 }, { "epoch": 2.78, "learning_rate": 2.451422152185106e-05, "loss": 0.0869, "step": 59600 }, { "epoch": 2.78, "learning_rate": 2.451375345201504e-05, "loss": 0.1151, "step": 59605 }, { "epoch": 2.78, "learning_rate": 2.451328538217902e-05, "loss": 0.2272, "step": 59610 }, { "epoch": 2.78, "learning_rate": 2.4512817312343003e-05, "loss": 0.3476, "step": 59615 }, { "epoch": 2.78, "learning_rate": 2.4512349242506983e-05, "loss": 0.0297, "step": 59620 }, { "epoch": 2.78, "learning_rate": 2.4511881172670963e-05, "loss": 0.0401, "step": 59625 }, { "epoch": 2.78, "learning_rate": 2.4511413102834942e-05, "loss": 0.0499, "step": 59630 }, { "epoch": 2.78, "learning_rate": 2.4510945032998926e-05, "loss": 0.0771, "step": 59635 }, { "epoch": 2.78, "learning_rate": 2.4510476963162905e-05, "loss": 0.0387, "step": 59640 }, { "epoch": 2.78, "learning_rate": 2.4510008893326885e-05, "loss": 0.1, "step": 59645 }, { "epoch": 2.78, "learning_rate": 2.450954082349087e-05, "loss": 0.115, "step": 59650 }, { "epoch": 2.78, "learning_rate": 2.4509072753654845e-05, "loss": 0.152, "step": 59655 }, { "epoch": 2.78, "learning_rate": 2.4508604683818825e-05, "loss": 0.2709, "step": 59660 }, { "epoch": 2.78, "learning_rate": 2.4508136613982804e-05, "loss": 0.2453, "step": 59665 }, { "epoch": 2.78, "learning_rate": 2.4507668544146788e-05, "loss": 0.0476, "step": 59670 }, { "epoch": 2.78, "learning_rate": 2.4507200474310767e-05, "loss": 0.0311, "step": 59675 }, { "epoch": 2.78, "learning_rate": 2.4506732404474747e-05, "loss": 0.0805, "step": 59680 }, { "epoch": 2.78, "learning_rate": 2.4506264334638727e-05, "loss": 0.1354, "step": 59685 }, { "epoch": 2.79, "learning_rate": 2.450579626480271e-05, "loss": 0.1196, "step": 59690 }, { "epoch": 2.79, "learning_rate": 2.450532819496669e-05, "loss": 0.0571, "step": 59695 }, { "epoch": 2.79, "learning_rate": 2.450486012513067e-05, "loss": 0.0936, "step": 59700 }, { "epoch": 2.79, "learning_rate": 2.4504392055294653e-05, "loss": 0.2136, "step": 59705 }, { "epoch": 2.79, "learning_rate": 2.4503923985458633e-05, "loss": 0.2878, "step": 59710 }, { "epoch": 2.79, "learning_rate": 2.4503455915622613e-05, "loss": 0.2839, "step": 59715 }, { "epoch": 2.79, "learning_rate": 2.450298784578659e-05, "loss": 0.026, "step": 59720 }, { "epoch": 2.79, "learning_rate": 2.4502519775950572e-05, "loss": 0.0408, "step": 59725 }, { "epoch": 2.79, "learning_rate": 2.4502051706114552e-05, "loss": 0.1123, "step": 59730 }, { "epoch": 2.79, "learning_rate": 2.4501583636278532e-05, "loss": 0.0279, "step": 59735 }, { "epoch": 2.79, "learning_rate": 2.4501115566442512e-05, "loss": 0.1024, "step": 59740 }, { "epoch": 2.79, "learning_rate": 2.4500647496606495e-05, "loss": 0.1186, "step": 59745 }, { "epoch": 2.79, "learning_rate": 2.4500179426770475e-05, "loss": 0.0859, "step": 59750 }, { "epoch": 2.79, "learning_rate": 2.4499711356934455e-05, "loss": 0.1142, "step": 59755 }, { "epoch": 2.79, "learning_rate": 2.4499243287098438e-05, "loss": 0.2424, "step": 59760 }, { "epoch": 2.79, "learning_rate": 2.4498775217262418e-05, "loss": 0.3073, "step": 59765 }, { "epoch": 2.79, "learning_rate": 2.4498307147426398e-05, "loss": 0.0441, "step": 59770 }, { "epoch": 2.79, "learning_rate": 2.4497839077590377e-05, "loss": 0.0324, "step": 59775 }, { "epoch": 2.79, "learning_rate": 2.4497371007754357e-05, "loss": 0.0323, "step": 59780 }, { "epoch": 2.79, "learning_rate": 2.4496902937918337e-05, "loss": 0.0365, "step": 59785 }, { "epoch": 2.79, "learning_rate": 2.4496434868082317e-05, "loss": 0.0499, "step": 59790 }, { "epoch": 2.79, "learning_rate": 2.4495966798246297e-05, "loss": 0.0849, "step": 59795 }, { "epoch": 2.79, "learning_rate": 2.449549872841028e-05, "loss": 0.1446, "step": 59800 }, { "epoch": 2.79, "learning_rate": 2.449503065857426e-05, "loss": 0.1163, "step": 59805 }, { "epoch": 2.79, "learning_rate": 2.449456258873824e-05, "loss": 0.1863, "step": 59810 }, { "epoch": 2.79, "learning_rate": 2.449409451890222e-05, "loss": 0.2137, "step": 59815 }, { "epoch": 2.79, "learning_rate": 2.4493626449066203e-05, "loss": 0.0261, "step": 59820 }, { "epoch": 2.79, "learning_rate": 2.4493158379230182e-05, "loss": 0.0784, "step": 59825 }, { "epoch": 2.79, "learning_rate": 2.4492690309394162e-05, "loss": 0.1117, "step": 59830 }, { "epoch": 2.79, "learning_rate": 2.4492222239558145e-05, "loss": 0.079, "step": 59835 }, { "epoch": 2.79, "learning_rate": 2.4491754169722125e-05, "loss": 0.071, "step": 59840 }, { "epoch": 2.79, "learning_rate": 2.44912860998861e-05, "loss": 0.1125, "step": 59845 }, { "epoch": 2.79, "learning_rate": 2.449081803005008e-05, "loss": 0.0955, "step": 59850 }, { "epoch": 2.79, "learning_rate": 2.4490349960214065e-05, "loss": 0.1655, "step": 59855 }, { "epoch": 2.79, "learning_rate": 2.4489881890378044e-05, "loss": 0.1266, "step": 59860 }, { "epoch": 2.79, "learning_rate": 2.4489413820542024e-05, "loss": 0.3275, "step": 59865 }, { "epoch": 2.79, "learning_rate": 2.4488945750706004e-05, "loss": 0.0658, "step": 59870 }, { "epoch": 2.79, "learning_rate": 2.4488477680869987e-05, "loss": 0.0367, "step": 59875 }, { "epoch": 2.79, "learning_rate": 2.4488009611033967e-05, "loss": 0.0492, "step": 59880 }, { "epoch": 2.79, "learning_rate": 2.4487541541197947e-05, "loss": 0.0986, "step": 59885 }, { "epoch": 2.79, "learning_rate": 2.448707347136193e-05, "loss": 0.0614, "step": 59890 }, { "epoch": 2.79, "learning_rate": 2.448660540152591e-05, "loss": 0.0977, "step": 59895 }, { "epoch": 2.8, "learning_rate": 2.448613733168989e-05, "loss": 0.1351, "step": 59900 }, { "epoch": 2.8, "learning_rate": 2.448566926185387e-05, "loss": 0.1106, "step": 59905 }, { "epoch": 2.8, "learning_rate": 2.448520119201785e-05, "loss": 0.248, "step": 59910 }, { "epoch": 2.8, "learning_rate": 2.448473312218183e-05, "loss": 0.269, "step": 59915 }, { "epoch": 2.8, "learning_rate": 2.448426505234581e-05, "loss": 0.0223, "step": 59920 }, { "epoch": 2.8, "learning_rate": 2.448379698250979e-05, "loss": 0.0459, "step": 59925 }, { "epoch": 2.8, "learning_rate": 2.4483328912673772e-05, "loss": 0.0505, "step": 59930 }, { "epoch": 2.8, "learning_rate": 2.4482860842837752e-05, "loss": 0.0455, "step": 59935 }, { "epoch": 2.8, "learning_rate": 2.4482392773001732e-05, "loss": 0.0739, "step": 59940 }, { "epoch": 2.8, "learning_rate": 2.4481924703165715e-05, "loss": 0.1352, "step": 59945 }, { "epoch": 2.8, "learning_rate": 2.4481456633329695e-05, "loss": 0.1433, "step": 59950 }, { "epoch": 2.8, "learning_rate": 2.4480988563493675e-05, "loss": 0.1034, "step": 59955 }, { "epoch": 2.8, "learning_rate": 2.4480520493657654e-05, "loss": 0.1249, "step": 59960 }, { "epoch": 2.8, "learning_rate": 2.4480052423821638e-05, "loss": 0.2903, "step": 59965 }, { "epoch": 2.8, "learning_rate": 2.4479584353985614e-05, "loss": 0.0758, "step": 59970 }, { "epoch": 2.8, "learning_rate": 2.4479116284149594e-05, "loss": 0.0193, "step": 59975 }, { "epoch": 2.8, "learning_rate": 2.4478648214313574e-05, "loss": 0.0632, "step": 59980 }, { "epoch": 2.8, "learning_rate": 2.4478180144477557e-05, "loss": 0.0705, "step": 59985 }, { "epoch": 2.8, "learning_rate": 2.4477712074641537e-05, "loss": 0.1063, "step": 59990 }, { "epoch": 2.8, "learning_rate": 2.4477244004805516e-05, "loss": 0.0351, "step": 59995 }, { "epoch": 2.8, "learning_rate": 2.44767759349695e-05, "loss": 0.1615, "step": 60000 }, { "epoch": 2.8, "learning_rate": 2.447630786513348e-05, "loss": 0.1302, "step": 60005 }, { "epoch": 2.8, "learning_rate": 2.447583979529746e-05, "loss": 0.2119, "step": 60010 }, { "epoch": 2.8, "learning_rate": 2.447537172546144e-05, "loss": 0.2506, "step": 60015 }, { "epoch": 2.8, "learning_rate": 2.4474903655625422e-05, "loss": 0.0472, "step": 60020 }, { "epoch": 2.8, "learning_rate": 2.4474435585789402e-05, "loss": 0.0203, "step": 60025 }, { "epoch": 2.8, "learning_rate": 2.4473967515953382e-05, "loss": 0.0179, "step": 60030 }, { "epoch": 2.8, "learning_rate": 2.447349944611736e-05, "loss": 0.0817, "step": 60035 }, { "epoch": 2.8, "learning_rate": 2.447303137628134e-05, "loss": 0.0724, "step": 60040 }, { "epoch": 2.8, "learning_rate": 2.447256330644532e-05, "loss": 0.0559, "step": 60045 }, { "epoch": 2.8, "learning_rate": 2.44720952366093e-05, "loss": 0.0986, "step": 60050 }, { "epoch": 2.8, "learning_rate": 2.447162716677328e-05, "loss": 0.0917, "step": 60055 }, { "epoch": 2.8, "learning_rate": 2.4471159096937264e-05, "loss": 0.3108, "step": 60060 }, { "epoch": 2.8, "learning_rate": 2.4470691027101244e-05, "loss": 0.2526, "step": 60065 }, { "epoch": 2.8, "learning_rate": 2.4470222957265224e-05, "loss": 0.0525, "step": 60070 }, { "epoch": 2.8, "learning_rate": 2.4469754887429207e-05, "loss": 0.0104, "step": 60075 }, { "epoch": 2.8, "learning_rate": 2.4469286817593187e-05, "loss": 0.0332, "step": 60080 }, { "epoch": 2.8, "learning_rate": 2.4468818747757167e-05, "loss": 0.0493, "step": 60085 }, { "epoch": 2.8, "learning_rate": 2.4468350677921147e-05, "loss": 0.0624, "step": 60090 }, { "epoch": 2.8, "learning_rate": 2.446788260808513e-05, "loss": 0.0975, "step": 60095 }, { "epoch": 2.8, "learning_rate": 2.4467414538249106e-05, "loss": 0.0622, "step": 60100 }, { "epoch": 2.8, "learning_rate": 2.4466946468413086e-05, "loss": 0.1787, "step": 60105 }, { "epoch": 2.8, "learning_rate": 2.4466478398577066e-05, "loss": 0.255, "step": 60110 }, { "epoch": 2.81, "learning_rate": 2.446601032874105e-05, "loss": 0.2408, "step": 60115 }, { "epoch": 2.81, "learning_rate": 2.446554225890503e-05, "loss": 0.0576, "step": 60120 }, { "epoch": 2.81, "learning_rate": 2.446507418906901e-05, "loss": 0.0079, "step": 60125 }, { "epoch": 2.81, "learning_rate": 2.4464606119232992e-05, "loss": 0.0536, "step": 60130 }, { "epoch": 2.81, "learning_rate": 2.446413804939697e-05, "loss": 0.0669, "step": 60135 }, { "epoch": 2.81, "learning_rate": 2.446366997956095e-05, "loss": 0.08, "step": 60140 }, { "epoch": 2.81, "learning_rate": 2.446320190972493e-05, "loss": 0.1084, "step": 60145 }, { "epoch": 2.81, "learning_rate": 2.4462733839888915e-05, "loss": 0.0946, "step": 60150 }, { "epoch": 2.81, "learning_rate": 2.4462265770052894e-05, "loss": 0.2113, "step": 60155 }, { "epoch": 2.81, "learning_rate": 2.446179770021687e-05, "loss": 0.1546, "step": 60160 }, { "epoch": 2.81, "learning_rate": 2.446132963038085e-05, "loss": 0.2933, "step": 60165 }, { "epoch": 2.81, "learning_rate": 2.4460861560544834e-05, "loss": 0.039, "step": 60170 }, { "epoch": 2.81, "learning_rate": 2.4460393490708814e-05, "loss": 0.0461, "step": 60175 }, { "epoch": 2.81, "learning_rate": 2.4459925420872793e-05, "loss": 0.0186, "step": 60180 }, { "epoch": 2.81, "learning_rate": 2.4459457351036777e-05, "loss": 0.0763, "step": 60185 }, { "epoch": 2.81, "learning_rate": 2.4458989281200756e-05, "loss": 0.0467, "step": 60190 }, { "epoch": 2.81, "learning_rate": 2.4458521211364736e-05, "loss": 0.0992, "step": 60195 }, { "epoch": 2.81, "learning_rate": 2.4458053141528716e-05, "loss": 0.0936, "step": 60200 }, { "epoch": 2.81, "learning_rate": 2.44575850716927e-05, "loss": 0.1019, "step": 60205 }, { "epoch": 2.81, "learning_rate": 2.445711700185668e-05, "loss": 0.1582, "step": 60210 }, { "epoch": 2.81, "learning_rate": 2.445664893202066e-05, "loss": 0.2476, "step": 60215 }, { "epoch": 2.81, "learning_rate": 2.445618086218464e-05, "loss": 0.0351, "step": 60220 }, { "epoch": 2.81, "learning_rate": 2.445571279234862e-05, "loss": 0.0333, "step": 60225 }, { "epoch": 2.81, "learning_rate": 2.44552447225126e-05, "loss": 0.0444, "step": 60230 }, { "epoch": 2.81, "learning_rate": 2.4454776652676578e-05, "loss": 0.0282, "step": 60235 }, { "epoch": 2.81, "learning_rate": 2.4454308582840558e-05, "loss": 0.0617, "step": 60240 }, { "epoch": 2.81, "learning_rate": 2.445384051300454e-05, "loss": 0.1827, "step": 60245 }, { "epoch": 2.81, "learning_rate": 2.445337244316852e-05, "loss": 0.2022, "step": 60250 }, { "epoch": 2.81, "learning_rate": 2.44529043733325e-05, "loss": 0.1955, "step": 60255 }, { "epoch": 2.81, "learning_rate": 2.4452436303496484e-05, "loss": 0.213, "step": 60260 }, { "epoch": 2.81, "learning_rate": 2.4451968233660464e-05, "loss": 0.2272, "step": 60265 }, { "epoch": 2.81, "learning_rate": 2.4451500163824444e-05, "loss": 0.0147, "step": 60270 }, { "epoch": 2.81, "learning_rate": 2.4451032093988424e-05, "loss": 0.0804, "step": 60275 }, { "epoch": 2.81, "learning_rate": 2.4450564024152407e-05, "loss": 0.0417, "step": 60280 }, { "epoch": 2.81, "learning_rate": 2.4450095954316387e-05, "loss": 0.0388, "step": 60285 }, { "epoch": 2.81, "learning_rate": 2.4449627884480363e-05, "loss": 0.1295, "step": 60290 }, { "epoch": 2.81, "learning_rate": 2.4449159814644343e-05, "loss": 0.173, "step": 60295 }, { "epoch": 2.81, "learning_rate": 2.4448691744808326e-05, "loss": 0.0887, "step": 60300 }, { "epoch": 2.81, "learning_rate": 2.4448223674972306e-05, "loss": 0.176, "step": 60305 }, { "epoch": 2.81, "learning_rate": 2.4447755605136286e-05, "loss": 0.1139, "step": 60310 }, { "epoch": 2.81, "learning_rate": 2.444728753530027e-05, "loss": 0.2864, "step": 60315 }, { "epoch": 2.81, "learning_rate": 2.444681946546425e-05, "loss": 0.0525, "step": 60320 }, { "epoch": 2.81, "learning_rate": 2.444635139562823e-05, "loss": 0.0649, "step": 60325 }, { "epoch": 2.82, "learning_rate": 2.4445883325792208e-05, "loss": 0.063, "step": 60330 }, { "epoch": 2.82, "learning_rate": 2.444541525595619e-05, "loss": 0.0791, "step": 60335 }, { "epoch": 2.82, "learning_rate": 2.444494718612017e-05, "loss": 0.1119, "step": 60340 }, { "epoch": 2.82, "learning_rate": 2.444447911628415e-05, "loss": 0.1171, "step": 60345 }, { "epoch": 2.82, "learning_rate": 2.4444011046448128e-05, "loss": 0.2339, "step": 60350 }, { "epoch": 2.82, "learning_rate": 2.444354297661211e-05, "loss": 0.2005, "step": 60355 }, { "epoch": 2.82, "learning_rate": 2.444307490677609e-05, "loss": 0.3308, "step": 60360 }, { "epoch": 2.82, "learning_rate": 2.444260683694007e-05, "loss": 0.3391, "step": 60365 }, { "epoch": 2.82, "learning_rate": 2.4442138767104054e-05, "loss": 0.031, "step": 60370 }, { "epoch": 2.82, "learning_rate": 2.4441670697268033e-05, "loss": 0.0791, "step": 60375 }, { "epoch": 2.82, "learning_rate": 2.4441202627432013e-05, "loss": 0.0219, "step": 60380 }, { "epoch": 2.82, "learning_rate": 2.4440734557595993e-05, "loss": 0.0218, "step": 60385 }, { "epoch": 2.82, "learning_rate": 2.4440266487759976e-05, "loss": 0.0515, "step": 60390 }, { "epoch": 2.82, "learning_rate": 2.4439798417923956e-05, "loss": 0.1438, "step": 60395 }, { "epoch": 2.82, "learning_rate": 2.4439330348087936e-05, "loss": 0.159, "step": 60400 }, { "epoch": 2.82, "learning_rate": 2.4438862278251916e-05, "loss": 0.1257, "step": 60405 }, { "epoch": 2.82, "learning_rate": 2.44383942084159e-05, "loss": 0.1489, "step": 60410 }, { "epoch": 2.82, "learning_rate": 2.4437926138579875e-05, "loss": 0.2007, "step": 60415 }, { "epoch": 2.82, "learning_rate": 2.4437458068743855e-05, "loss": 0.033, "step": 60420 }, { "epoch": 2.82, "learning_rate": 2.4436989998907835e-05, "loss": 0.1097, "step": 60425 }, { "epoch": 2.82, "learning_rate": 2.4436521929071818e-05, "loss": 0.1007, "step": 60430 }, { "epoch": 2.82, "learning_rate": 2.4436053859235798e-05, "loss": 0.0648, "step": 60435 }, { "epoch": 2.82, "learning_rate": 2.4435585789399778e-05, "loss": 0.1449, "step": 60440 }, { "epoch": 2.82, "learning_rate": 2.443511771956376e-05, "loss": 0.1316, "step": 60445 }, { "epoch": 2.82, "learning_rate": 2.443464964972774e-05, "loss": 0.0919, "step": 60450 }, { "epoch": 2.82, "learning_rate": 2.443418157989172e-05, "loss": 0.1348, "step": 60455 }, { "epoch": 2.82, "learning_rate": 2.44337135100557e-05, "loss": 0.2028, "step": 60460 }, { "epoch": 2.82, "learning_rate": 2.4433245440219684e-05, "loss": 0.3667, "step": 60465 }, { "epoch": 2.82, "learning_rate": 2.4432777370383664e-05, "loss": 0.0199, "step": 60470 }, { "epoch": 2.82, "learning_rate": 2.4432309300547643e-05, "loss": 0.0429, "step": 60475 }, { "epoch": 2.82, "learning_rate": 2.443184123071162e-05, "loss": 0.0243, "step": 60480 }, { "epoch": 2.82, "learning_rate": 2.4431373160875603e-05, "loss": 0.0567, "step": 60485 }, { "epoch": 2.82, "learning_rate": 2.4430905091039583e-05, "loss": 0.0627, "step": 60490 }, { "epoch": 2.82, "learning_rate": 2.4430437021203563e-05, "loss": 0.0771, "step": 60495 }, { "epoch": 2.82, "learning_rate": 2.4429968951367546e-05, "loss": 0.1636, "step": 60500 }, { "epoch": 2.82, "learning_rate": 2.4429500881531526e-05, "loss": 0.1397, "step": 60505 }, { "epoch": 2.82, "learning_rate": 2.4429032811695505e-05, "loss": 0.2548, "step": 60510 }, { "epoch": 2.82, "learning_rate": 2.4428564741859485e-05, "loss": 0.3335, "step": 60515 }, { "epoch": 2.82, "learning_rate": 2.442809667202347e-05, "loss": 0.0661, "step": 60520 }, { "epoch": 2.82, "learning_rate": 2.4427628602187448e-05, "loss": 0.0513, "step": 60525 }, { "epoch": 2.82, "learning_rate": 2.4427160532351428e-05, "loss": 0.0348, "step": 60530 }, { "epoch": 2.82, "learning_rate": 2.4426692462515408e-05, "loss": 0.1036, "step": 60535 }, { "epoch": 2.82, "learning_rate": 2.4426224392679388e-05, "loss": 0.0371, "step": 60540 }, { "epoch": 2.83, "learning_rate": 2.4425756322843368e-05, "loss": 0.0545, "step": 60545 }, { "epoch": 2.83, "learning_rate": 2.4425288253007347e-05, "loss": 0.0814, "step": 60550 }, { "epoch": 2.83, "learning_rate": 2.442482018317133e-05, "loss": 0.1206, "step": 60555 }, { "epoch": 2.83, "learning_rate": 2.442435211333531e-05, "loss": 0.2098, "step": 60560 }, { "epoch": 2.83, "learning_rate": 2.442388404349929e-05, "loss": 0.2092, "step": 60565 }, { "epoch": 2.83, "learning_rate": 2.442341597366327e-05, "loss": 0.0672, "step": 60570 }, { "epoch": 2.83, "learning_rate": 2.4422947903827253e-05, "loss": 0.0361, "step": 60575 }, { "epoch": 2.83, "learning_rate": 2.4422479833991233e-05, "loss": 0.0672, "step": 60580 }, { "epoch": 2.83, "learning_rate": 2.4422011764155213e-05, "loss": 0.0813, "step": 60585 }, { "epoch": 2.83, "learning_rate": 2.4421543694319193e-05, "loss": 0.1038, "step": 60590 }, { "epoch": 2.83, "learning_rate": 2.4421075624483176e-05, "loss": 0.085, "step": 60595 }, { "epoch": 2.83, "learning_rate": 2.4420607554647156e-05, "loss": 0.1322, "step": 60600 }, { "epoch": 2.83, "learning_rate": 2.4420139484811132e-05, "loss": 0.1983, "step": 60605 }, { "epoch": 2.83, "learning_rate": 2.4419671414975112e-05, "loss": 0.1474, "step": 60610 }, { "epoch": 2.83, "learning_rate": 2.4419203345139095e-05, "loss": 0.3845, "step": 60615 }, { "epoch": 2.83, "learning_rate": 2.4418735275303075e-05, "loss": 0.055, "step": 60620 }, { "epoch": 2.83, "learning_rate": 2.4418267205467055e-05, "loss": 0.0347, "step": 60625 }, { "epoch": 2.83, "learning_rate": 2.4417799135631038e-05, "loss": 0.0667, "step": 60630 }, { "epoch": 2.83, "learning_rate": 2.4417331065795018e-05, "loss": 0.0311, "step": 60635 }, { "epoch": 2.83, "learning_rate": 2.4416862995958998e-05, "loss": 0.0487, "step": 60640 }, { "epoch": 2.83, "learning_rate": 2.4416394926122977e-05, "loss": 0.0644, "step": 60645 }, { "epoch": 2.83, "learning_rate": 2.441592685628696e-05, "loss": 0.1066, "step": 60650 }, { "epoch": 2.83, "learning_rate": 2.441545878645094e-05, "loss": 0.1963, "step": 60655 }, { "epoch": 2.83, "learning_rate": 2.441499071661492e-05, "loss": 0.3125, "step": 60660 }, { "epoch": 2.83, "learning_rate": 2.44145226467789e-05, "loss": 0.2088, "step": 60665 }, { "epoch": 2.83, "learning_rate": 2.441405457694288e-05, "loss": 0.0164, "step": 60670 }, { "epoch": 2.83, "learning_rate": 2.441358650710686e-05, "loss": 0.0451, "step": 60675 }, { "epoch": 2.83, "learning_rate": 2.441311843727084e-05, "loss": 0.0368, "step": 60680 }, { "epoch": 2.83, "learning_rate": 2.4412650367434823e-05, "loss": 0.0716, "step": 60685 }, { "epoch": 2.83, "learning_rate": 2.4412182297598803e-05, "loss": 0.0277, "step": 60690 }, { "epoch": 2.83, "learning_rate": 2.4411714227762782e-05, "loss": 0.0362, "step": 60695 }, { "epoch": 2.83, "learning_rate": 2.4411246157926762e-05, "loss": 0.1361, "step": 60700 }, { "epoch": 2.83, "learning_rate": 2.4410778088090745e-05, "loss": 0.1438, "step": 60705 }, { "epoch": 2.83, "learning_rate": 2.4410310018254725e-05, "loss": 0.1406, "step": 60710 }, { "epoch": 2.83, "learning_rate": 2.4409841948418705e-05, "loss": 0.2761, "step": 60715 }, { "epoch": 2.83, "learning_rate": 2.4409373878582685e-05, "loss": 0.0729, "step": 60720 }, { "epoch": 2.83, "learning_rate": 2.4408905808746668e-05, "loss": 0.063, "step": 60725 }, { "epoch": 2.83, "learning_rate": 2.4408437738910645e-05, "loss": 0.0291, "step": 60730 }, { "epoch": 2.83, "learning_rate": 2.4407969669074624e-05, "loss": 0.0664, "step": 60735 }, { "epoch": 2.83, "learning_rate": 2.4407501599238608e-05, "loss": 0.0994, "step": 60740 }, { "epoch": 2.83, "learning_rate": 2.4407033529402587e-05, "loss": 0.0764, "step": 60745 }, { "epoch": 2.83, "learning_rate": 2.4406565459566567e-05, "loss": 0.094, "step": 60750 }, { "epoch": 2.83, "learning_rate": 2.4406097389730547e-05, "loss": 0.1506, "step": 60755 }, { "epoch": 2.84, "learning_rate": 2.440562931989453e-05, "loss": 0.149, "step": 60760 }, { "epoch": 2.84, "learning_rate": 2.440516125005851e-05, "loss": 0.2185, "step": 60765 }, { "epoch": 2.84, "learning_rate": 2.440469318022249e-05, "loss": 0.0627, "step": 60770 }, { "epoch": 2.84, "learning_rate": 2.440422511038647e-05, "loss": 0.0275, "step": 60775 }, { "epoch": 2.84, "learning_rate": 2.4403757040550453e-05, "loss": 0.0384, "step": 60780 }, { "epoch": 2.84, "learning_rate": 2.4403288970714433e-05, "loss": 0.096, "step": 60785 }, { "epoch": 2.84, "learning_rate": 2.4402820900878413e-05, "loss": 0.1149, "step": 60790 }, { "epoch": 2.84, "learning_rate": 2.4402352831042392e-05, "loss": 0.1999, "step": 60795 }, { "epoch": 2.84, "learning_rate": 2.4401884761206372e-05, "loss": 0.0822, "step": 60800 }, { "epoch": 2.84, "learning_rate": 2.4401416691370352e-05, "loss": 0.1295, "step": 60805 }, { "epoch": 2.84, "learning_rate": 2.4400948621534332e-05, "loss": 0.2997, "step": 60810 }, { "epoch": 2.84, "learning_rate": 2.4400480551698315e-05, "loss": 0.2911, "step": 60815 }, { "epoch": 2.84, "learning_rate": 2.4400012481862295e-05, "loss": 0.0212, "step": 60820 }, { "epoch": 2.84, "learning_rate": 2.4399544412026275e-05, "loss": 0.0702, "step": 60825 }, { "epoch": 2.84, "learning_rate": 2.4399076342190254e-05, "loss": 0.0685, "step": 60830 }, { "epoch": 2.84, "learning_rate": 2.4398608272354238e-05, "loss": 0.0828, "step": 60835 }, { "epoch": 2.84, "learning_rate": 2.4398140202518217e-05, "loss": 0.0848, "step": 60840 }, { "epoch": 2.84, "learning_rate": 2.4397672132682197e-05, "loss": 0.0776, "step": 60845 }, { "epoch": 2.84, "learning_rate": 2.4397204062846177e-05, "loss": 0.141, "step": 60850 }, { "epoch": 2.84, "learning_rate": 2.4396735993010157e-05, "loss": 0.0924, "step": 60855 }, { "epoch": 2.84, "learning_rate": 2.4396267923174137e-05, "loss": 0.2505, "step": 60860 }, { "epoch": 2.84, "learning_rate": 2.4395799853338117e-05, "loss": 0.3442, "step": 60865 }, { "epoch": 2.84, "learning_rate": 2.43953317835021e-05, "loss": 0.024, "step": 60870 }, { "epoch": 2.84, "learning_rate": 2.439486371366608e-05, "loss": 0.0099, "step": 60875 }, { "epoch": 2.84, "learning_rate": 2.439439564383006e-05, "loss": 0.0559, "step": 60880 }, { "epoch": 2.84, "learning_rate": 2.439392757399404e-05, "loss": 0.1163, "step": 60885 }, { "epoch": 2.84, "learning_rate": 2.4393459504158022e-05, "loss": 0.0899, "step": 60890 }, { "epoch": 2.84, "learning_rate": 2.4392991434322002e-05, "loss": 0.0595, "step": 60895 }, { "epoch": 2.84, "learning_rate": 2.4392523364485982e-05, "loss": 0.0824, "step": 60900 }, { "epoch": 2.84, "learning_rate": 2.4392055294649962e-05, "loss": 0.0924, "step": 60905 }, { "epoch": 2.84, "learning_rate": 2.4391587224813945e-05, "loss": 0.2858, "step": 60910 }, { "epoch": 2.84, "learning_rate": 2.4391119154977925e-05, "loss": 0.1593, "step": 60915 }, { "epoch": 2.84, "learning_rate": 2.43906510851419e-05, "loss": 0.0711, "step": 60920 }, { "epoch": 2.84, "learning_rate": 2.4390183015305885e-05, "loss": 0.0241, "step": 60925 }, { "epoch": 2.84, "learning_rate": 2.4389714945469864e-05, "loss": 0.058, "step": 60930 }, { "epoch": 2.84, "learning_rate": 2.4389246875633844e-05, "loss": 0.0993, "step": 60935 }, { "epoch": 2.84, "learning_rate": 2.4388778805797824e-05, "loss": 0.0619, "step": 60940 }, { "epoch": 2.84, "learning_rate": 2.4388310735961807e-05, "loss": 0.0851, "step": 60945 }, { "epoch": 2.84, "learning_rate": 2.4387842666125787e-05, "loss": 0.0801, "step": 60950 }, { "epoch": 2.84, "learning_rate": 2.4387374596289767e-05, "loss": 0.1628, "step": 60955 }, { "epoch": 2.84, "learning_rate": 2.4386906526453747e-05, "loss": 0.2158, "step": 60960 }, { "epoch": 2.84, "learning_rate": 2.438643845661773e-05, "loss": 0.2365, "step": 60965 }, { "epoch": 2.84, "learning_rate": 2.438597038678171e-05, "loss": 0.074, "step": 60970 }, { "epoch": 2.85, "learning_rate": 2.438550231694569e-05, "loss": 0.0071, "step": 60975 }, { "epoch": 2.85, "learning_rate": 2.4385034247109673e-05, "loss": 0.0436, "step": 60980 }, { "epoch": 2.85, "learning_rate": 2.438456617727365e-05, "loss": 0.0791, "step": 60985 }, { "epoch": 2.85, "learning_rate": 2.438409810743763e-05, "loss": 0.0707, "step": 60990 }, { "epoch": 2.85, "learning_rate": 2.438363003760161e-05, "loss": 0.0873, "step": 60995 }, { "epoch": 2.85, "learning_rate": 2.4383161967765592e-05, "loss": 0.089, "step": 61000 }, { "epoch": 2.85, "learning_rate": 2.4382693897929572e-05, "loss": 0.1736, "step": 61005 }, { "epoch": 2.85, "learning_rate": 2.438222582809355e-05, "loss": 0.1735, "step": 61010 }, { "epoch": 2.85, "learning_rate": 2.438175775825753e-05, "loss": 0.2373, "step": 61015 }, { "epoch": 2.85, "learning_rate": 2.4381289688421515e-05, "loss": 0.0902, "step": 61020 }, { "epoch": 2.85, "learning_rate": 2.4380821618585494e-05, "loss": 0.0455, "step": 61025 }, { "epoch": 2.85, "learning_rate": 2.4380353548749474e-05, "loss": 0.0767, "step": 61030 }, { "epoch": 2.85, "learning_rate": 2.4379885478913454e-05, "loss": 0.0361, "step": 61035 }, { "epoch": 2.85, "learning_rate": 2.4379417409077437e-05, "loss": 0.0537, "step": 61040 }, { "epoch": 2.85, "learning_rate": 2.4378949339241414e-05, "loss": 0.0887, "step": 61045 }, { "epoch": 2.85, "learning_rate": 2.4378481269405394e-05, "loss": 0.0741, "step": 61050 }, { "epoch": 2.85, "learning_rate": 2.4378013199569377e-05, "loss": 0.1016, "step": 61055 }, { "epoch": 2.85, "learning_rate": 2.4377545129733357e-05, "loss": 0.2068, "step": 61060 }, { "epoch": 2.85, "learning_rate": 2.4377077059897336e-05, "loss": 0.1699, "step": 61065 }, { "epoch": 2.85, "learning_rate": 2.4376608990061316e-05, "loss": 0.0394, "step": 61070 }, { "epoch": 2.85, "learning_rate": 2.43761409202253e-05, "loss": 0.0519, "step": 61075 }, { "epoch": 2.85, "learning_rate": 2.437567285038928e-05, "loss": 0.0488, "step": 61080 }, { "epoch": 2.85, "learning_rate": 2.437520478055326e-05, "loss": 0.0327, "step": 61085 }, { "epoch": 2.85, "learning_rate": 2.437473671071724e-05, "loss": 0.1175, "step": 61090 }, { "epoch": 2.85, "learning_rate": 2.4374268640881222e-05, "loss": 0.0967, "step": 61095 }, { "epoch": 2.85, "learning_rate": 2.4373800571045202e-05, "loss": 0.1732, "step": 61100 }, { "epoch": 2.85, "learning_rate": 2.437333250120918e-05, "loss": 0.0984, "step": 61105 }, { "epoch": 2.85, "learning_rate": 2.437286443137316e-05, "loss": 0.2185, "step": 61110 }, { "epoch": 2.85, "learning_rate": 2.437239636153714e-05, "loss": 0.2051, "step": 61115 }, { "epoch": 2.85, "learning_rate": 2.437192829170112e-05, "loss": 0.0535, "step": 61120 }, { "epoch": 2.85, "learning_rate": 2.43714602218651e-05, "loss": 0.0825, "step": 61125 }, { "epoch": 2.85, "learning_rate": 2.4370992152029084e-05, "loss": 0.0412, "step": 61130 }, { "epoch": 2.85, "learning_rate": 2.4370524082193064e-05, "loss": 0.0505, "step": 61135 }, { "epoch": 2.85, "learning_rate": 2.4370056012357044e-05, "loss": 0.0527, "step": 61140 }, { "epoch": 2.85, "learning_rate": 2.4369587942521024e-05, "loss": 0.2246, "step": 61145 }, { "epoch": 2.85, "learning_rate": 2.4369119872685007e-05, "loss": 0.1156, "step": 61150 }, { "epoch": 2.85, "learning_rate": 2.4368651802848987e-05, "loss": 0.1205, "step": 61155 }, { "epoch": 2.85, "learning_rate": 2.4368183733012966e-05, "loss": 0.2444, "step": 61160 }, { "epoch": 2.85, "learning_rate": 2.436771566317695e-05, "loss": 0.2647, "step": 61165 }, { "epoch": 2.85, "learning_rate": 2.4367247593340926e-05, "loss": 0.078, "step": 61170 }, { "epoch": 2.85, "learning_rate": 2.4366779523504906e-05, "loss": 0.0541, "step": 61175 }, { "epoch": 2.85, "learning_rate": 2.4366311453668886e-05, "loss": 0.0712, "step": 61180 }, { "epoch": 2.85, "learning_rate": 2.436584338383287e-05, "loss": 0.0953, "step": 61185 }, { "epoch": 2.86, "learning_rate": 2.436537531399685e-05, "loss": 0.0654, "step": 61190 }, { "epoch": 2.86, "learning_rate": 2.436490724416083e-05, "loss": 0.0437, "step": 61195 }, { "epoch": 2.86, "learning_rate": 2.436443917432481e-05, "loss": 0.1437, "step": 61200 }, { "epoch": 2.86, "learning_rate": 2.436397110448879e-05, "loss": 0.1068, "step": 61205 }, { "epoch": 2.86, "learning_rate": 2.436350303465277e-05, "loss": 0.1847, "step": 61210 }, { "epoch": 2.86, "learning_rate": 2.436303496481675e-05, "loss": 0.3212, "step": 61215 }, { "epoch": 2.86, "learning_rate": 2.436256689498073e-05, "loss": 0.068, "step": 61220 }, { "epoch": 2.86, "learning_rate": 2.4362098825144714e-05, "loss": 0.0279, "step": 61225 }, { "epoch": 2.86, "learning_rate": 2.4361630755308694e-05, "loss": 0.0346, "step": 61230 }, { "epoch": 2.86, "learning_rate": 2.436116268547267e-05, "loss": 0.043, "step": 61235 }, { "epoch": 2.86, "learning_rate": 2.4360694615636654e-05, "loss": 0.0752, "step": 61240 }, { "epoch": 2.86, "learning_rate": 2.4360226545800634e-05, "loss": 0.075, "step": 61245 }, { "epoch": 2.86, "learning_rate": 2.4359758475964613e-05, "loss": 0.1264, "step": 61250 }, { "epoch": 2.86, "learning_rate": 2.4359290406128593e-05, "loss": 0.1511, "step": 61255 }, { "epoch": 2.86, "learning_rate": 2.4358822336292576e-05, "loss": 0.1892, "step": 61260 }, { "epoch": 2.86, "learning_rate": 2.4358354266456556e-05, "loss": 0.2657, "step": 61265 }, { "epoch": 2.86, "learning_rate": 2.4357886196620536e-05, "loss": 0.0295, "step": 61270 }, { "epoch": 2.86, "learning_rate": 2.4357418126784516e-05, "loss": 0.0323, "step": 61275 }, { "epoch": 2.86, "learning_rate": 2.43569500569485e-05, "loss": 0.0585, "step": 61280 }, { "epoch": 2.86, "learning_rate": 2.435648198711248e-05, "loss": 0.1003, "step": 61285 }, { "epoch": 2.86, "learning_rate": 2.435601391727646e-05, "loss": 0.124, "step": 61290 }, { "epoch": 2.86, "learning_rate": 2.4355545847440442e-05, "loss": 0.0432, "step": 61295 }, { "epoch": 2.86, "learning_rate": 2.4355077777604418e-05, "loss": 0.153, "step": 61300 }, { "epoch": 2.86, "learning_rate": 2.4354609707768398e-05, "loss": 0.2346, "step": 61305 }, { "epoch": 2.86, "learning_rate": 2.4354141637932378e-05, "loss": 0.3643, "step": 61310 }, { "epoch": 2.86, "learning_rate": 2.435367356809636e-05, "loss": 0.2115, "step": 61315 }, { "epoch": 2.86, "learning_rate": 2.435320549826034e-05, "loss": 0.0354, "step": 61320 }, { "epoch": 2.86, "learning_rate": 2.435273742842432e-05, "loss": 0.0395, "step": 61325 }, { "epoch": 2.86, "learning_rate": 2.43522693585883e-05, "loss": 0.0449, "step": 61330 }, { "epoch": 2.86, "learning_rate": 2.4351801288752284e-05, "loss": 0.0886, "step": 61335 }, { "epoch": 2.86, "learning_rate": 2.4351333218916264e-05, "loss": 0.0702, "step": 61340 }, { "epoch": 2.86, "learning_rate": 2.4350865149080243e-05, "loss": 0.0505, "step": 61345 }, { "epoch": 2.86, "learning_rate": 2.4350397079244227e-05, "loss": 0.1158, "step": 61350 }, { "epoch": 2.86, "learning_rate": 2.4349929009408206e-05, "loss": 0.0956, "step": 61355 }, { "epoch": 2.86, "learning_rate": 2.4349460939572183e-05, "loss": 0.1982, "step": 61360 }, { "epoch": 2.86, "learning_rate": 2.4348992869736163e-05, "loss": 0.3243, "step": 61365 }, { "epoch": 2.86, "learning_rate": 2.4348524799900146e-05, "loss": 0.0685, "step": 61370 }, { "epoch": 2.86, "learning_rate": 2.4348056730064126e-05, "loss": 0.0371, "step": 61375 }, { "epoch": 2.86, "learning_rate": 2.4347588660228106e-05, "loss": 0.0439, "step": 61380 }, { "epoch": 2.86, "learning_rate": 2.4347120590392085e-05, "loss": 0.0727, "step": 61385 }, { "epoch": 2.86, "learning_rate": 2.434665252055607e-05, "loss": 0.1047, "step": 61390 }, { "epoch": 2.86, "learning_rate": 2.434618445072005e-05, "loss": 0.0731, "step": 61395 }, { "epoch": 2.87, "learning_rate": 2.4345716380884028e-05, "loss": 0.1474, "step": 61400 }, { "epoch": 2.87, "learning_rate": 2.434524831104801e-05, "loss": 0.1503, "step": 61405 }, { "epoch": 2.87, "learning_rate": 2.434478024121199e-05, "loss": 0.1643, "step": 61410 }, { "epoch": 2.87, "learning_rate": 2.434431217137597e-05, "loss": 0.2506, "step": 61415 }, { "epoch": 2.87, "learning_rate": 2.434384410153995e-05, "loss": 0.0481, "step": 61420 }, { "epoch": 2.87, "learning_rate": 2.434337603170393e-05, "loss": 0.0373, "step": 61425 }, { "epoch": 2.87, "learning_rate": 2.434290796186791e-05, "loss": 0.0473, "step": 61430 }, { "epoch": 2.87, "learning_rate": 2.434243989203189e-05, "loss": 0.0797, "step": 61435 }, { "epoch": 2.87, "learning_rate": 2.434197182219587e-05, "loss": 0.0475, "step": 61440 }, { "epoch": 2.87, "learning_rate": 2.4341503752359853e-05, "loss": 0.058, "step": 61445 }, { "epoch": 2.87, "learning_rate": 2.4341035682523833e-05, "loss": 0.1259, "step": 61450 }, { "epoch": 2.87, "learning_rate": 2.4340567612687813e-05, "loss": 0.1367, "step": 61455 }, { "epoch": 2.87, "learning_rate": 2.4340099542851793e-05, "loss": 0.2598, "step": 61460 }, { "epoch": 2.87, "learning_rate": 2.4339631473015776e-05, "loss": 0.3622, "step": 61465 }, { "epoch": 2.87, "learning_rate": 2.4339163403179756e-05, "loss": 0.0815, "step": 61470 }, { "epoch": 2.87, "learning_rate": 2.4338695333343736e-05, "loss": 0.0363, "step": 61475 }, { "epoch": 2.87, "learning_rate": 2.433822726350772e-05, "loss": 0.0565, "step": 61480 }, { "epoch": 2.87, "learning_rate": 2.43377591936717e-05, "loss": 0.0606, "step": 61485 }, { "epoch": 2.87, "learning_rate": 2.4337291123835675e-05, "loss": 0.0921, "step": 61490 }, { "epoch": 2.87, "learning_rate": 2.4336823053999655e-05, "loss": 0.1805, "step": 61495 }, { "epoch": 2.87, "learning_rate": 2.4336354984163638e-05, "loss": 0.1077, "step": 61500 }, { "epoch": 2.87, "learning_rate": 2.4335886914327618e-05, "loss": 0.1032, "step": 61505 }, { "epoch": 2.87, "learning_rate": 2.4335418844491598e-05, "loss": 0.1976, "step": 61510 }, { "epoch": 2.87, "learning_rate": 2.4334950774655578e-05, "loss": 0.2539, "step": 61515 }, { "epoch": 2.87, "learning_rate": 2.433448270481956e-05, "loss": 0.0383, "step": 61520 }, { "epoch": 2.87, "learning_rate": 2.433401463498354e-05, "loss": 0.0519, "step": 61525 }, { "epoch": 2.87, "learning_rate": 2.433354656514752e-05, "loss": 0.0411, "step": 61530 }, { "epoch": 2.87, "learning_rate": 2.4333078495311504e-05, "loss": 0.045, "step": 61535 }, { "epoch": 2.87, "learning_rate": 2.4332610425475483e-05, "loss": 0.0384, "step": 61540 }, { "epoch": 2.87, "learning_rate": 2.4332142355639463e-05, "loss": 0.1072, "step": 61545 }, { "epoch": 2.87, "learning_rate": 2.433167428580344e-05, "loss": 0.1773, "step": 61550 }, { "epoch": 2.87, "learning_rate": 2.4331206215967423e-05, "loss": 0.1756, "step": 61555 }, { "epoch": 2.87, "learning_rate": 2.4330738146131403e-05, "loss": 0.3195, "step": 61560 }, { "epoch": 2.87, "learning_rate": 2.4330270076295383e-05, "loss": 0.3297, "step": 61565 }, { "epoch": 2.87, "learning_rate": 2.4329802006459362e-05, "loss": 0.0287, "step": 61570 }, { "epoch": 2.87, "learning_rate": 2.4329333936623346e-05, "loss": 0.0272, "step": 61575 }, { "epoch": 2.87, "learning_rate": 2.4328865866787325e-05, "loss": 0.0611, "step": 61580 }, { "epoch": 2.87, "learning_rate": 2.4328397796951305e-05, "loss": 0.0606, "step": 61585 }, { "epoch": 2.87, "learning_rate": 2.432792972711529e-05, "loss": 0.0713, "step": 61590 }, { "epoch": 2.87, "learning_rate": 2.4327461657279268e-05, "loss": 0.1068, "step": 61595 }, { "epoch": 2.87, "learning_rate": 2.4326993587443248e-05, "loss": 0.1738, "step": 61600 }, { "epoch": 2.87, "learning_rate": 2.4326525517607228e-05, "loss": 0.2613, "step": 61605 }, { "epoch": 2.87, "learning_rate": 2.432605744777121e-05, "loss": 0.1856, "step": 61610 }, { "epoch": 2.88, "learning_rate": 2.4325589377935187e-05, "loss": 0.2767, "step": 61615 }, { "epoch": 2.88, "learning_rate": 2.4325121308099167e-05, "loss": 0.0698, "step": 61620 }, { "epoch": 2.88, "learning_rate": 2.4324653238263147e-05, "loss": 0.0076, "step": 61625 }, { "epoch": 2.88, "learning_rate": 2.432418516842713e-05, "loss": 0.0321, "step": 61630 }, { "epoch": 2.88, "learning_rate": 2.432371709859111e-05, "loss": 0.0851, "step": 61635 }, { "epoch": 2.88, "learning_rate": 2.432324902875509e-05, "loss": 0.1152, "step": 61640 }, { "epoch": 2.88, "learning_rate": 2.432278095891907e-05, "loss": 0.0597, "step": 61645 }, { "epoch": 2.88, "learning_rate": 2.4322312889083053e-05, "loss": 0.1604, "step": 61650 }, { "epoch": 2.88, "learning_rate": 2.4321844819247033e-05, "loss": 0.1788, "step": 61655 }, { "epoch": 2.88, "learning_rate": 2.4321376749411013e-05, "loss": 0.1336, "step": 61660 }, { "epoch": 2.88, "learning_rate": 2.4320908679574996e-05, "loss": 0.1934, "step": 61665 }, { "epoch": 2.88, "learning_rate": 2.4320440609738976e-05, "loss": 0.0273, "step": 61670 }, { "epoch": 2.88, "learning_rate": 2.4319972539902955e-05, "loss": 0.0323, "step": 61675 }, { "epoch": 2.88, "learning_rate": 2.4319504470066932e-05, "loss": 0.0202, "step": 61680 }, { "epoch": 2.88, "learning_rate": 2.4319036400230915e-05, "loss": 0.0769, "step": 61685 }, { "epoch": 2.88, "learning_rate": 2.4318568330394895e-05, "loss": 0.083, "step": 61690 }, { "epoch": 2.88, "learning_rate": 2.4318100260558875e-05, "loss": 0.0766, "step": 61695 }, { "epoch": 2.88, "learning_rate": 2.4317632190722855e-05, "loss": 0.1065, "step": 61700 }, { "epoch": 2.88, "learning_rate": 2.4317164120886838e-05, "loss": 0.1781, "step": 61705 }, { "epoch": 2.88, "learning_rate": 2.4316696051050818e-05, "loss": 0.0665, "step": 61710 }, { "epoch": 2.88, "learning_rate": 2.4316227981214797e-05, "loss": 0.2226, "step": 61715 }, { "epoch": 2.88, "learning_rate": 2.431575991137878e-05, "loss": 0.081, "step": 61720 }, { "epoch": 2.88, "learning_rate": 2.431529184154276e-05, "loss": 0.0485, "step": 61725 }, { "epoch": 2.88, "learning_rate": 2.431482377170674e-05, "loss": 0.0235, "step": 61730 }, { "epoch": 2.88, "learning_rate": 2.431435570187072e-05, "loss": 0.069, "step": 61735 }, { "epoch": 2.88, "learning_rate": 2.43138876320347e-05, "loss": 0.0751, "step": 61740 }, { "epoch": 2.88, "learning_rate": 2.431341956219868e-05, "loss": 0.1192, "step": 61745 }, { "epoch": 2.88, "learning_rate": 2.431295149236266e-05, "loss": 0.1559, "step": 61750 }, { "epoch": 2.88, "learning_rate": 2.431248342252664e-05, "loss": 0.22, "step": 61755 }, { "epoch": 2.88, "learning_rate": 2.4312015352690623e-05, "loss": 0.2909, "step": 61760 }, { "epoch": 2.88, "learning_rate": 2.4311547282854602e-05, "loss": 0.2338, "step": 61765 }, { "epoch": 2.88, "learning_rate": 2.4311079213018582e-05, "loss": 0.0535, "step": 61770 }, { "epoch": 2.88, "learning_rate": 2.4310611143182565e-05, "loss": 0.0662, "step": 61775 }, { "epoch": 2.88, "learning_rate": 2.4310143073346545e-05, "loss": 0.0611, "step": 61780 }, { "epoch": 2.88, "learning_rate": 2.4309675003510525e-05, "loss": 0.0273, "step": 61785 }, { "epoch": 2.88, "learning_rate": 2.4309206933674505e-05, "loss": 0.0856, "step": 61790 }, { "epoch": 2.88, "learning_rate": 2.4308738863838488e-05, "loss": 0.1327, "step": 61795 }, { "epoch": 2.88, "learning_rate": 2.4308270794002468e-05, "loss": 0.1152, "step": 61800 }, { "epoch": 2.88, "learning_rate": 2.4307802724166444e-05, "loss": 0.1681, "step": 61805 }, { "epoch": 2.88, "learning_rate": 2.4307334654330424e-05, "loss": 0.1321, "step": 61810 }, { "epoch": 2.88, "learning_rate": 2.4306866584494407e-05, "loss": 0.1941, "step": 61815 }, { "epoch": 2.88, "learning_rate": 2.4306398514658387e-05, "loss": 0.0509, "step": 61820 }, { "epoch": 2.88, "learning_rate": 2.4305930444822367e-05, "loss": 0.0275, "step": 61825 }, { "epoch": 2.89, "learning_rate": 2.4305462374986347e-05, "loss": 0.0283, "step": 61830 }, { "epoch": 2.89, "learning_rate": 2.430499430515033e-05, "loss": 0.0961, "step": 61835 }, { "epoch": 2.89, "learning_rate": 2.430452623531431e-05, "loss": 0.0573, "step": 61840 }, { "epoch": 2.89, "learning_rate": 2.430405816547829e-05, "loss": 0.1085, "step": 61845 }, { "epoch": 2.89, "learning_rate": 2.4303590095642273e-05, "loss": 0.1402, "step": 61850 }, { "epoch": 2.89, "learning_rate": 2.4303122025806253e-05, "loss": 0.1818, "step": 61855 }, { "epoch": 2.89, "learning_rate": 2.4302653955970232e-05, "loss": 0.175, "step": 61860 }, { "epoch": 2.89, "learning_rate": 2.4302185886134212e-05, "loss": 0.3747, "step": 61865 }, { "epoch": 2.89, "learning_rate": 2.4301717816298192e-05, "loss": 0.0274, "step": 61870 }, { "epoch": 2.89, "learning_rate": 2.4301249746462172e-05, "loss": 0.0117, "step": 61875 }, { "epoch": 2.89, "learning_rate": 2.430078167662615e-05, "loss": 0.0528, "step": 61880 }, { "epoch": 2.89, "learning_rate": 2.430031360679013e-05, "loss": 0.0439, "step": 61885 }, { "epoch": 2.89, "learning_rate": 2.4299845536954115e-05, "loss": 0.0745, "step": 61890 }, { "epoch": 2.89, "learning_rate": 2.4299377467118095e-05, "loss": 0.1921, "step": 61895 }, { "epoch": 2.89, "learning_rate": 2.4298909397282074e-05, "loss": 0.0633, "step": 61900 }, { "epoch": 2.89, "learning_rate": 2.4298441327446058e-05, "loss": 0.1569, "step": 61905 }, { "epoch": 2.89, "learning_rate": 2.4297973257610037e-05, "loss": 0.1319, "step": 61910 }, { "epoch": 2.89, "learning_rate": 2.4297505187774017e-05, "loss": 0.1784, "step": 61915 }, { "epoch": 2.89, "learning_rate": 2.4297037117937997e-05, "loss": 0.0167, "step": 61920 }, { "epoch": 2.89, "learning_rate": 2.429656904810198e-05, "loss": 0.04, "step": 61925 }, { "epoch": 2.89, "learning_rate": 2.4296100978265957e-05, "loss": 0.0985, "step": 61930 }, { "epoch": 2.89, "learning_rate": 2.4295632908429936e-05, "loss": 0.0926, "step": 61935 }, { "epoch": 2.89, "learning_rate": 2.4295164838593916e-05, "loss": 0.0379, "step": 61940 }, { "epoch": 2.89, "learning_rate": 2.42946967687579e-05, "loss": 0.2004, "step": 61945 }, { "epoch": 2.89, "learning_rate": 2.429422869892188e-05, "loss": 0.0933, "step": 61950 }, { "epoch": 2.89, "learning_rate": 2.429376062908586e-05, "loss": 0.168, "step": 61955 }, { "epoch": 2.89, "learning_rate": 2.4293292559249842e-05, "loss": 0.1649, "step": 61960 }, { "epoch": 2.89, "learning_rate": 2.4292824489413822e-05, "loss": 0.2429, "step": 61965 }, { "epoch": 2.89, "learning_rate": 2.4292356419577802e-05, "loss": 0.094, "step": 61970 }, { "epoch": 2.89, "learning_rate": 2.4291888349741782e-05, "loss": 0.0499, "step": 61975 }, { "epoch": 2.89, "learning_rate": 2.4291420279905765e-05, "loss": 0.0279, "step": 61980 }, { "epoch": 2.89, "learning_rate": 2.4290952210069745e-05, "loss": 0.0924, "step": 61985 }, { "epoch": 2.89, "learning_rate": 2.4290484140233725e-05, "loss": 0.0661, "step": 61990 }, { "epoch": 2.89, "learning_rate": 2.42900160703977e-05, "loss": 0.1303, "step": 61995 }, { "epoch": 2.89, "learning_rate": 2.4289548000561684e-05, "loss": 0.1382, "step": 62000 }, { "epoch": 2.89, "learning_rate": 2.4289079930725664e-05, "loss": 0.0747, "step": 62005 }, { "epoch": 2.89, "learning_rate": 2.4288611860889644e-05, "loss": 0.1682, "step": 62010 }, { "epoch": 2.89, "learning_rate": 2.4288143791053627e-05, "loss": 0.2472, "step": 62015 }, { "epoch": 2.89, "learning_rate": 2.4287675721217607e-05, "loss": 0.0919, "step": 62020 }, { "epoch": 2.89, "learning_rate": 2.4287207651381587e-05, "loss": 0.0293, "step": 62025 }, { "epoch": 2.89, "learning_rate": 2.4286739581545567e-05, "loss": 0.0307, "step": 62030 }, { "epoch": 2.89, "learning_rate": 2.428627151170955e-05, "loss": 0.0959, "step": 62035 }, { "epoch": 2.89, "learning_rate": 2.428580344187353e-05, "loss": 0.0687, "step": 62040 }, { "epoch": 2.9, "learning_rate": 2.428533537203751e-05, "loss": 0.0993, "step": 62045 }, { "epoch": 2.9, "learning_rate": 2.428486730220149e-05, "loss": 0.2027, "step": 62050 }, { "epoch": 2.9, "learning_rate": 2.428439923236547e-05, "loss": 0.1111, "step": 62055 }, { "epoch": 2.9, "learning_rate": 2.428393116252945e-05, "loss": 0.2029, "step": 62060 }, { "epoch": 2.9, "learning_rate": 2.428346309269343e-05, "loss": 0.1714, "step": 62065 }, { "epoch": 2.9, "learning_rate": 2.428299502285741e-05, "loss": 0.0518, "step": 62070 }, { "epoch": 2.9, "learning_rate": 2.428252695302139e-05, "loss": 0.0558, "step": 62075 }, { "epoch": 2.9, "learning_rate": 2.428205888318537e-05, "loss": 0.0818, "step": 62080 }, { "epoch": 2.9, "learning_rate": 2.428159081334935e-05, "loss": 0.1121, "step": 62085 }, { "epoch": 2.9, "learning_rate": 2.4281122743513335e-05, "loss": 0.0769, "step": 62090 }, { "epoch": 2.9, "learning_rate": 2.4280654673677314e-05, "loss": 0.0717, "step": 62095 }, { "epoch": 2.9, "learning_rate": 2.4280186603841294e-05, "loss": 0.0822, "step": 62100 }, { "epoch": 2.9, "learning_rate": 2.4279718534005274e-05, "loss": 0.1347, "step": 62105 }, { "epoch": 2.9, "learning_rate": 2.4279250464169257e-05, "loss": 0.2519, "step": 62110 }, { "epoch": 2.9, "learning_rate": 2.4278782394333237e-05, "loss": 0.232, "step": 62115 }, { "epoch": 2.9, "learning_rate": 2.4278314324497213e-05, "loss": 0.0362, "step": 62120 }, { "epoch": 2.9, "learning_rate": 2.4277846254661193e-05, "loss": 0.0422, "step": 62125 }, { "epoch": 2.9, "learning_rate": 2.4277378184825176e-05, "loss": 0.0254, "step": 62130 }, { "epoch": 2.9, "learning_rate": 2.4276910114989156e-05, "loss": 0.0787, "step": 62135 }, { "epoch": 2.9, "learning_rate": 2.4276442045153136e-05, "loss": 0.11, "step": 62140 }, { "epoch": 2.9, "learning_rate": 2.427597397531712e-05, "loss": 0.101, "step": 62145 }, { "epoch": 2.9, "learning_rate": 2.42755059054811e-05, "loss": 0.1464, "step": 62150 }, { "epoch": 2.9, "learning_rate": 2.427503783564508e-05, "loss": 0.135, "step": 62155 }, { "epoch": 2.9, "learning_rate": 2.427456976580906e-05, "loss": 0.2666, "step": 62160 }, { "epoch": 2.9, "learning_rate": 2.4274101695973042e-05, "loss": 0.1826, "step": 62165 }, { "epoch": 2.9, "learning_rate": 2.4273633626137022e-05, "loss": 0.0506, "step": 62170 }, { "epoch": 2.9, "learning_rate": 2.4273165556301e-05, "loss": 0.0364, "step": 62175 }, { "epoch": 2.9, "learning_rate": 2.427269748646498e-05, "loss": 0.0439, "step": 62180 }, { "epoch": 2.9, "learning_rate": 2.427222941662896e-05, "loss": 0.0592, "step": 62185 }, { "epoch": 2.9, "learning_rate": 2.427176134679294e-05, "loss": 0.0831, "step": 62190 }, { "epoch": 2.9, "learning_rate": 2.427129327695692e-05, "loss": 0.1056, "step": 62195 }, { "epoch": 2.9, "learning_rate": 2.4270825207120904e-05, "loss": 0.125, "step": 62200 }, { "epoch": 2.9, "learning_rate": 2.4270357137284884e-05, "loss": 0.2117, "step": 62205 }, { "epoch": 2.9, "learning_rate": 2.4269889067448864e-05, "loss": 0.3004, "step": 62210 }, { "epoch": 2.9, "learning_rate": 2.4269420997612844e-05, "loss": 0.2234, "step": 62215 }, { "epoch": 2.9, "learning_rate": 2.4268952927776827e-05, "loss": 0.0448, "step": 62220 }, { "epoch": 2.9, "learning_rate": 2.4268484857940807e-05, "loss": 0.0236, "step": 62225 }, { "epoch": 2.9, "learning_rate": 2.4268016788104786e-05, "loss": 0.082, "step": 62230 }, { "epoch": 2.9, "learning_rate": 2.4267548718268766e-05, "loss": 0.0562, "step": 62235 }, { "epoch": 2.9, "learning_rate": 2.426708064843275e-05, "loss": 0.0415, "step": 62240 }, { "epoch": 2.9, "learning_rate": 2.4266612578596726e-05, "loss": 0.0682, "step": 62245 }, { "epoch": 2.9, "learning_rate": 2.4266144508760706e-05, "loss": 0.092, "step": 62250 }, { "epoch": 2.9, "learning_rate": 2.4265676438924685e-05, "loss": 0.137, "step": 62255 }, { "epoch": 2.91, "learning_rate": 2.426520836908867e-05, "loss": 0.2428, "step": 62260 }, { "epoch": 2.91, "learning_rate": 2.426474029925265e-05, "loss": 0.276, "step": 62265 }, { "epoch": 2.91, "learning_rate": 2.4264272229416628e-05, "loss": 0.0869, "step": 62270 }, { "epoch": 2.91, "learning_rate": 2.426380415958061e-05, "loss": 0.0545, "step": 62275 }, { "epoch": 2.91, "learning_rate": 2.426333608974459e-05, "loss": 0.0699, "step": 62280 }, { "epoch": 2.91, "learning_rate": 2.426286801990857e-05, "loss": 0.0621, "step": 62285 }, { "epoch": 2.91, "learning_rate": 2.426239995007255e-05, "loss": 0.0954, "step": 62290 }, { "epoch": 2.91, "learning_rate": 2.4261931880236534e-05, "loss": 0.1235, "step": 62295 }, { "epoch": 2.91, "learning_rate": 2.4261463810400514e-05, "loss": 0.1832, "step": 62300 }, { "epoch": 2.91, "learning_rate": 2.4260995740564494e-05, "loss": 0.1592, "step": 62305 }, { "epoch": 2.91, "learning_rate": 2.426052767072847e-05, "loss": 0.1928, "step": 62310 }, { "epoch": 2.91, "learning_rate": 2.4260059600892453e-05, "loss": 0.2388, "step": 62315 }, { "epoch": 2.91, "learning_rate": 2.4259591531056433e-05, "loss": 0.0617, "step": 62320 }, { "epoch": 2.91, "learning_rate": 2.4259123461220413e-05, "loss": 0.0122, "step": 62325 }, { "epoch": 2.91, "learning_rate": 2.4258655391384396e-05, "loss": 0.0475, "step": 62330 }, { "epoch": 2.91, "learning_rate": 2.4258187321548376e-05, "loss": 0.0282, "step": 62335 }, { "epoch": 2.91, "learning_rate": 2.4257719251712356e-05, "loss": 0.0403, "step": 62340 }, { "epoch": 2.91, "learning_rate": 2.4257251181876336e-05, "loss": 0.0951, "step": 62345 }, { "epoch": 2.91, "learning_rate": 2.425678311204032e-05, "loss": 0.1381, "step": 62350 }, { "epoch": 2.91, "learning_rate": 2.42563150422043e-05, "loss": 0.1644, "step": 62355 }, { "epoch": 2.91, "learning_rate": 2.425584697236828e-05, "loss": 0.1559, "step": 62360 }, { "epoch": 2.91, "learning_rate": 2.425537890253226e-05, "loss": 0.2036, "step": 62365 }, { "epoch": 2.91, "learning_rate": 2.425491083269624e-05, "loss": 0.0376, "step": 62370 }, { "epoch": 2.91, "learning_rate": 2.4254442762860218e-05, "loss": 0.0281, "step": 62375 }, { "epoch": 2.91, "learning_rate": 2.4253974693024198e-05, "loss": 0.0222, "step": 62380 }, { "epoch": 2.91, "learning_rate": 2.425350662318818e-05, "loss": 0.06, "step": 62385 }, { "epoch": 2.91, "learning_rate": 2.425303855335216e-05, "loss": 0.0387, "step": 62390 }, { "epoch": 2.91, "learning_rate": 2.425257048351614e-05, "loss": 0.0702, "step": 62395 }, { "epoch": 2.91, "learning_rate": 2.425210241368012e-05, "loss": 0.062, "step": 62400 }, { "epoch": 2.91, "learning_rate": 2.4251634343844104e-05, "loss": 0.1915, "step": 62405 }, { "epoch": 2.91, "learning_rate": 2.4251166274008084e-05, "loss": 0.1869, "step": 62410 }, { "epoch": 2.91, "learning_rate": 2.4250698204172063e-05, "loss": 0.354, "step": 62415 }, { "epoch": 2.91, "learning_rate": 2.4250230134336043e-05, "loss": 0.0541, "step": 62420 }, { "epoch": 2.91, "learning_rate": 2.4249762064500026e-05, "loss": 0.0214, "step": 62425 }, { "epoch": 2.91, "learning_rate": 2.4249293994664006e-05, "loss": 0.0189, "step": 62430 }, { "epoch": 2.91, "learning_rate": 2.4248825924827983e-05, "loss": 0.0817, "step": 62435 }, { "epoch": 2.91, "learning_rate": 2.4248357854991962e-05, "loss": 0.0589, "step": 62440 }, { "epoch": 2.91, "learning_rate": 2.4247889785155946e-05, "loss": 0.1001, "step": 62445 }, { "epoch": 2.91, "learning_rate": 2.4247421715319925e-05, "loss": 0.1036, "step": 62450 }, { "epoch": 2.91, "learning_rate": 2.4246953645483905e-05, "loss": 0.1051, "step": 62455 }, { "epoch": 2.91, "learning_rate": 2.424648557564789e-05, "loss": 0.2078, "step": 62460 }, { "epoch": 2.91, "learning_rate": 2.4246017505811868e-05, "loss": 0.2313, "step": 62465 }, { "epoch": 2.91, "learning_rate": 2.4245549435975848e-05, "loss": 0.0199, "step": 62470 }, { "epoch": 2.92, "learning_rate": 2.4245081366139828e-05, "loss": 0.0428, "step": 62475 }, { "epoch": 2.92, "learning_rate": 2.424461329630381e-05, "loss": 0.0405, "step": 62480 }, { "epoch": 2.92, "learning_rate": 2.424414522646779e-05, "loss": 0.0835, "step": 62485 }, { "epoch": 2.92, "learning_rate": 2.424367715663177e-05, "loss": 0.0966, "step": 62490 }, { "epoch": 2.92, "learning_rate": 2.424320908679575e-05, "loss": 0.1012, "step": 62495 }, { "epoch": 2.92, "learning_rate": 2.424274101695973e-05, "loss": 0.069, "step": 62500 }, { "epoch": 2.92, "learning_rate": 2.424227294712371e-05, "loss": 0.261, "step": 62505 }, { "epoch": 2.92, "learning_rate": 2.424180487728769e-05, "loss": 0.4236, "step": 62510 }, { "epoch": 2.92, "learning_rate": 2.4241336807451673e-05, "loss": 0.2733, "step": 62515 }, { "epoch": 2.92, "learning_rate": 2.4240868737615653e-05, "loss": 0.0231, "step": 62520 }, { "epoch": 2.92, "learning_rate": 2.4240400667779633e-05, "loss": 0.0412, "step": 62525 }, { "epoch": 2.92, "learning_rate": 2.4239932597943613e-05, "loss": 0.0379, "step": 62530 }, { "epoch": 2.92, "learning_rate": 2.4239464528107596e-05, "loss": 0.0517, "step": 62535 }, { "epoch": 2.92, "learning_rate": 2.4238996458271576e-05, "loss": 0.0839, "step": 62540 }, { "epoch": 2.92, "learning_rate": 2.4238528388435556e-05, "loss": 0.1312, "step": 62545 }, { "epoch": 2.92, "learning_rate": 2.4238060318599535e-05, "loss": 0.1746, "step": 62550 }, { "epoch": 2.92, "learning_rate": 2.423759224876352e-05, "loss": 0.1599, "step": 62555 }, { "epoch": 2.92, "learning_rate": 2.4237124178927495e-05, "loss": 0.1256, "step": 62560 }, { "epoch": 2.92, "learning_rate": 2.4236656109091475e-05, "loss": 0.2098, "step": 62565 }, { "epoch": 2.92, "learning_rate": 2.4236188039255458e-05, "loss": 0.018, "step": 62570 }, { "epoch": 2.92, "learning_rate": 2.4235719969419438e-05, "loss": 0.0437, "step": 62575 }, { "epoch": 2.92, "learning_rate": 2.4235251899583418e-05, "loss": 0.0274, "step": 62580 }, { "epoch": 2.92, "learning_rate": 2.4234783829747397e-05, "loss": 0.0552, "step": 62585 }, { "epoch": 2.92, "learning_rate": 2.423431575991138e-05, "loss": 0.0478, "step": 62590 }, { "epoch": 2.92, "learning_rate": 2.423384769007536e-05, "loss": 0.154, "step": 62595 }, { "epoch": 2.92, "learning_rate": 2.423337962023934e-05, "loss": 0.0941, "step": 62600 }, { "epoch": 2.92, "learning_rate": 2.423291155040332e-05, "loss": 0.1631, "step": 62605 }, { "epoch": 2.92, "learning_rate": 2.4232443480567303e-05, "loss": 0.2482, "step": 62610 }, { "epoch": 2.92, "learning_rate": 2.4231975410731283e-05, "loss": 0.2467, "step": 62615 }, { "epoch": 2.92, "learning_rate": 2.4231507340895263e-05, "loss": 0.0199, "step": 62620 }, { "epoch": 2.92, "learning_rate": 2.423103927105924e-05, "loss": 0.0346, "step": 62625 }, { "epoch": 2.92, "learning_rate": 2.4230571201223223e-05, "loss": 0.0191, "step": 62630 }, { "epoch": 2.92, "learning_rate": 2.4230103131387202e-05, "loss": 0.0524, "step": 62635 }, { "epoch": 2.92, "learning_rate": 2.4229635061551182e-05, "loss": 0.0683, "step": 62640 }, { "epoch": 2.92, "learning_rate": 2.4229166991715165e-05, "loss": 0.1245, "step": 62645 }, { "epoch": 2.92, "learning_rate": 2.4228698921879145e-05, "loss": 0.1553, "step": 62650 }, { "epoch": 2.92, "learning_rate": 2.4228230852043125e-05, "loss": 0.1299, "step": 62655 }, { "epoch": 2.92, "learning_rate": 2.4227762782207105e-05, "loss": 0.2061, "step": 62660 }, { "epoch": 2.92, "learning_rate": 2.4227294712371088e-05, "loss": 0.2275, "step": 62665 }, { "epoch": 2.92, "learning_rate": 2.4226826642535068e-05, "loss": 0.0413, "step": 62670 }, { "epoch": 2.92, "learning_rate": 2.4226358572699048e-05, "loss": 0.0218, "step": 62675 }, { "epoch": 2.92, "learning_rate": 2.4225890502863028e-05, "loss": 0.0585, "step": 62680 }, { "epoch": 2.92, "learning_rate": 2.422542243302701e-05, "loss": 0.0461, "step": 62685 }, { "epoch": 2.93, "learning_rate": 2.4224954363190987e-05, "loss": 0.0453, "step": 62690 }, { "epoch": 2.93, "learning_rate": 2.4224486293354967e-05, "loss": 0.0638, "step": 62695 }, { "epoch": 2.93, "learning_rate": 2.422401822351895e-05, "loss": 0.1183, "step": 62700 }, { "epoch": 2.93, "learning_rate": 2.422355015368293e-05, "loss": 0.1473, "step": 62705 }, { "epoch": 2.93, "learning_rate": 2.422308208384691e-05, "loss": 0.2651, "step": 62710 }, { "epoch": 2.93, "learning_rate": 2.422261401401089e-05, "loss": 0.3302, "step": 62715 }, { "epoch": 2.93, "learning_rate": 2.4222145944174873e-05, "loss": 0.0361, "step": 62720 }, { "epoch": 2.93, "learning_rate": 2.4221677874338853e-05, "loss": 0.0223, "step": 62725 }, { "epoch": 2.93, "learning_rate": 2.4221209804502832e-05, "loss": 0.0328, "step": 62730 }, { "epoch": 2.93, "learning_rate": 2.4220741734666812e-05, "loss": 0.0705, "step": 62735 }, { "epoch": 2.93, "learning_rate": 2.4220273664830796e-05, "loss": 0.0713, "step": 62740 }, { "epoch": 2.93, "learning_rate": 2.4219805594994775e-05, "loss": 0.1249, "step": 62745 }, { "epoch": 2.93, "learning_rate": 2.4219337525158752e-05, "loss": 0.1163, "step": 62750 }, { "epoch": 2.93, "learning_rate": 2.4218869455322735e-05, "loss": 0.1602, "step": 62755 }, { "epoch": 2.93, "learning_rate": 2.4218401385486715e-05, "loss": 0.3166, "step": 62760 }, { "epoch": 2.93, "learning_rate": 2.4217933315650695e-05, "loss": 0.2019, "step": 62765 }, { "epoch": 2.93, "learning_rate": 2.4217465245814674e-05, "loss": 0.0548, "step": 62770 }, { "epoch": 2.93, "learning_rate": 2.4216997175978658e-05, "loss": 0.0238, "step": 62775 }, { "epoch": 2.93, "learning_rate": 2.4216529106142637e-05, "loss": 0.0893, "step": 62780 }, { "epoch": 2.93, "learning_rate": 2.4216061036306617e-05, "loss": 0.0386, "step": 62785 }, { "epoch": 2.93, "learning_rate": 2.4215592966470597e-05, "loss": 0.1056, "step": 62790 }, { "epoch": 2.93, "learning_rate": 2.421512489663458e-05, "loss": 0.1014, "step": 62795 }, { "epoch": 2.93, "learning_rate": 2.421465682679856e-05, "loss": 0.0761, "step": 62800 }, { "epoch": 2.93, "learning_rate": 2.421418875696254e-05, "loss": 0.1406, "step": 62805 }, { "epoch": 2.93, "learning_rate": 2.4213720687126523e-05, "loss": 0.2066, "step": 62810 }, { "epoch": 2.93, "learning_rate": 2.42132526172905e-05, "loss": 0.1796, "step": 62815 }, { "epoch": 2.93, "learning_rate": 2.421278454745448e-05, "loss": 0.0369, "step": 62820 }, { "epoch": 2.93, "learning_rate": 2.421231647761846e-05, "loss": 0.0343, "step": 62825 }, { "epoch": 2.93, "learning_rate": 2.4211848407782442e-05, "loss": 0.0331, "step": 62830 }, { "epoch": 2.93, "learning_rate": 2.4211380337946422e-05, "loss": 0.0711, "step": 62835 }, { "epoch": 2.93, "learning_rate": 2.4210912268110402e-05, "loss": 0.0932, "step": 62840 }, { "epoch": 2.93, "learning_rate": 2.4210444198274382e-05, "loss": 0.0931, "step": 62845 }, { "epoch": 2.93, "learning_rate": 2.4209976128438365e-05, "loss": 0.0469, "step": 62850 }, { "epoch": 2.93, "learning_rate": 2.4209508058602345e-05, "loss": 0.1935, "step": 62855 }, { "epoch": 2.93, "learning_rate": 2.4209039988766325e-05, "loss": 0.1608, "step": 62860 }, { "epoch": 2.93, "learning_rate": 2.4208571918930305e-05, "loss": 0.2193, "step": 62865 }, { "epoch": 2.93, "learning_rate": 2.4208103849094288e-05, "loss": 0.0788, "step": 62870 }, { "epoch": 2.93, "learning_rate": 2.4207635779258268e-05, "loss": 0.0601, "step": 62875 }, { "epoch": 2.93, "learning_rate": 2.4207167709422244e-05, "loss": 0.034, "step": 62880 }, { "epoch": 2.93, "learning_rate": 2.4206699639586227e-05, "loss": 0.0355, "step": 62885 }, { "epoch": 2.93, "learning_rate": 2.4206231569750207e-05, "loss": 0.0392, "step": 62890 }, { "epoch": 2.93, "learning_rate": 2.4205763499914187e-05, "loss": 0.0807, "step": 62895 }, { "epoch": 2.94, "learning_rate": 2.4205295430078167e-05, "loss": 0.1078, "step": 62900 }, { "epoch": 2.94, "learning_rate": 2.420482736024215e-05, "loss": 0.1747, "step": 62905 }, { "epoch": 2.94, "learning_rate": 2.420435929040613e-05, "loss": 0.1978, "step": 62910 }, { "epoch": 2.94, "learning_rate": 2.420389122057011e-05, "loss": 0.2202, "step": 62915 }, { "epoch": 2.94, "learning_rate": 2.420342315073409e-05, "loss": 0.0395, "step": 62920 }, { "epoch": 2.94, "learning_rate": 2.4202955080898072e-05, "loss": 0.0087, "step": 62925 }, { "epoch": 2.94, "learning_rate": 2.4202487011062052e-05, "loss": 0.0778, "step": 62930 }, { "epoch": 2.94, "learning_rate": 2.4202018941226032e-05, "loss": 0.0473, "step": 62935 }, { "epoch": 2.94, "learning_rate": 2.4201550871390012e-05, "loss": 0.1018, "step": 62940 }, { "epoch": 2.94, "learning_rate": 2.4201082801553992e-05, "loss": 0.0669, "step": 62945 }, { "epoch": 2.94, "learning_rate": 2.420061473171797e-05, "loss": 0.0629, "step": 62950 }, { "epoch": 2.94, "learning_rate": 2.420014666188195e-05, "loss": 0.1694, "step": 62955 }, { "epoch": 2.94, "learning_rate": 2.4199678592045935e-05, "loss": 0.3145, "step": 62960 }, { "epoch": 2.94, "learning_rate": 2.4199210522209914e-05, "loss": 0.2555, "step": 62965 }, { "epoch": 2.94, "learning_rate": 2.4198742452373894e-05, "loss": 0.0611, "step": 62970 }, { "epoch": 2.94, "learning_rate": 2.4198274382537874e-05, "loss": 0.0464, "step": 62975 }, { "epoch": 2.94, "learning_rate": 2.4197806312701857e-05, "loss": 0.0323, "step": 62980 }, { "epoch": 2.94, "learning_rate": 2.4197338242865837e-05, "loss": 0.0762, "step": 62985 }, { "epoch": 2.94, "learning_rate": 2.4196870173029817e-05, "loss": 0.0858, "step": 62990 }, { "epoch": 2.94, "learning_rate": 2.41964021031938e-05, "loss": 0.1115, "step": 62995 }, { "epoch": 2.94, "learning_rate": 2.419593403335778e-05, "loss": 0.0466, "step": 63000 }, { "epoch": 2.94, "learning_rate": 2.4195465963521756e-05, "loss": 0.1256, "step": 63005 }, { "epoch": 2.94, "learning_rate": 2.4194997893685736e-05, "loss": 0.2292, "step": 63010 }, { "epoch": 2.94, "learning_rate": 2.419452982384972e-05, "loss": 0.2535, "step": 63015 }, { "epoch": 2.94, "learning_rate": 2.41940617540137e-05, "loss": 0.0445, "step": 63020 }, { "epoch": 2.94, "learning_rate": 2.419359368417768e-05, "loss": 0.0609, "step": 63025 }, { "epoch": 2.94, "learning_rate": 2.419312561434166e-05, "loss": 0.0444, "step": 63030 }, { "epoch": 2.94, "learning_rate": 2.4192657544505642e-05, "loss": 0.0541, "step": 63035 }, { "epoch": 2.94, "learning_rate": 2.4192189474669622e-05, "loss": 0.0605, "step": 63040 }, { "epoch": 2.94, "learning_rate": 2.41917214048336e-05, "loss": 0.1205, "step": 63045 }, { "epoch": 2.94, "learning_rate": 2.419125333499758e-05, "loss": 0.0461, "step": 63050 }, { "epoch": 2.94, "learning_rate": 2.4190785265161565e-05, "loss": 0.168, "step": 63055 }, { "epoch": 2.94, "learning_rate": 2.4190317195325545e-05, "loss": 0.1311, "step": 63060 }, { "epoch": 2.94, "learning_rate": 2.4189849125489524e-05, "loss": 0.1908, "step": 63065 }, { "epoch": 2.94, "learning_rate": 2.4189381055653504e-05, "loss": 0.0581, "step": 63070 }, { "epoch": 2.94, "learning_rate": 2.4188912985817484e-05, "loss": 0.0383, "step": 63075 }, { "epoch": 2.94, "learning_rate": 2.4188444915981464e-05, "loss": 0.0243, "step": 63080 }, { "epoch": 2.94, "learning_rate": 2.4187976846145444e-05, "loss": 0.0817, "step": 63085 }, { "epoch": 2.94, "learning_rate": 2.4187508776309427e-05, "loss": 0.1185, "step": 63090 }, { "epoch": 2.94, "learning_rate": 2.4187040706473407e-05, "loss": 0.1229, "step": 63095 }, { "epoch": 2.94, "learning_rate": 2.4186572636637386e-05, "loss": 0.1517, "step": 63100 }, { "epoch": 2.94, "learning_rate": 2.4186104566801366e-05, "loss": 0.2088, "step": 63105 }, { "epoch": 2.94, "learning_rate": 2.418563649696535e-05, "loss": 0.231, "step": 63110 }, { "epoch": 2.95, "learning_rate": 2.418516842712933e-05, "loss": 0.2014, "step": 63115 }, { "epoch": 2.95, "learning_rate": 2.418470035729331e-05, "loss": 0.0771, "step": 63120 }, { "epoch": 2.95, "learning_rate": 2.4184232287457292e-05, "loss": 0.0238, "step": 63125 }, { "epoch": 2.95, "learning_rate": 2.418376421762127e-05, "loss": 0.0214, "step": 63130 }, { "epoch": 2.95, "learning_rate": 2.418329614778525e-05, "loss": 0.0765, "step": 63135 }, { "epoch": 2.95, "learning_rate": 2.418282807794923e-05, "loss": 0.1006, "step": 63140 }, { "epoch": 2.95, "learning_rate": 2.418236000811321e-05, "loss": 0.0525, "step": 63145 }, { "epoch": 2.95, "learning_rate": 2.418189193827719e-05, "loss": 0.0991, "step": 63150 }, { "epoch": 2.95, "learning_rate": 2.418142386844117e-05, "loss": 0.1976, "step": 63155 }, { "epoch": 2.95, "learning_rate": 2.418095579860515e-05, "loss": 0.1167, "step": 63160 }, { "epoch": 2.95, "learning_rate": 2.4180487728769134e-05, "loss": 0.3467, "step": 63165 }, { "epoch": 2.95, "learning_rate": 2.4180019658933114e-05, "loss": 0.0368, "step": 63170 }, { "epoch": 2.95, "learning_rate": 2.4179551589097094e-05, "loss": 0.0271, "step": 63175 }, { "epoch": 2.95, "learning_rate": 2.4179083519261077e-05, "loss": 0.0217, "step": 63180 }, { "epoch": 2.95, "learning_rate": 2.4178615449425057e-05, "loss": 0.0944, "step": 63185 }, { "epoch": 2.95, "learning_rate": 2.4178147379589037e-05, "loss": 0.0368, "step": 63190 }, { "epoch": 2.95, "learning_rate": 2.4177679309753013e-05, "loss": 0.1101, "step": 63195 }, { "epoch": 2.95, "learning_rate": 2.4177211239916996e-05, "loss": 0.0614, "step": 63200 }, { "epoch": 2.95, "learning_rate": 2.4176743170080976e-05, "loss": 0.1767, "step": 63205 }, { "epoch": 2.95, "learning_rate": 2.4176275100244956e-05, "loss": 0.2781, "step": 63210 }, { "epoch": 2.95, "learning_rate": 2.4175807030408936e-05, "loss": 0.4507, "step": 63215 }, { "epoch": 2.95, "learning_rate": 2.417533896057292e-05, "loss": 0.0527, "step": 63220 }, { "epoch": 2.95, "learning_rate": 2.41748708907369e-05, "loss": 0.0222, "step": 63225 }, { "epoch": 2.95, "learning_rate": 2.417440282090088e-05, "loss": 0.0543, "step": 63230 }, { "epoch": 2.95, "learning_rate": 2.417393475106486e-05, "loss": 0.0463, "step": 63235 }, { "epoch": 2.95, "learning_rate": 2.417346668122884e-05, "loss": 0.0956, "step": 63240 }, { "epoch": 2.95, "learning_rate": 2.417299861139282e-05, "loss": 0.0393, "step": 63245 }, { "epoch": 2.95, "learning_rate": 2.41725305415568e-05, "loss": 0.0857, "step": 63250 }, { "epoch": 2.95, "learning_rate": 2.417206247172078e-05, "loss": 0.1616, "step": 63255 }, { "epoch": 2.95, "learning_rate": 2.417159440188476e-05, "loss": 0.2549, "step": 63260 }, { "epoch": 2.95, "learning_rate": 2.417112633204874e-05, "loss": 0.3067, "step": 63265 }, { "epoch": 2.95, "learning_rate": 2.417065826221272e-05, "loss": 0.0083, "step": 63270 }, { "epoch": 2.95, "learning_rate": 2.4170190192376704e-05, "loss": 0.0653, "step": 63275 }, { "epoch": 2.95, "learning_rate": 2.4169722122540684e-05, "loss": 0.0114, "step": 63280 }, { "epoch": 2.95, "learning_rate": 2.4169254052704663e-05, "loss": 0.0451, "step": 63285 }, { "epoch": 2.95, "learning_rate": 2.4168785982868643e-05, "loss": 0.0472, "step": 63290 }, { "epoch": 2.95, "learning_rate": 2.4168317913032626e-05, "loss": 0.0913, "step": 63295 }, { "epoch": 2.95, "learning_rate": 2.4167849843196606e-05, "loss": 0.1354, "step": 63300 }, { "epoch": 2.95, "learning_rate": 2.4167381773360586e-05, "loss": 0.0834, "step": 63305 }, { "epoch": 2.95, "learning_rate": 2.416691370352457e-05, "loss": 0.1842, "step": 63310 }, { "epoch": 2.95, "learning_rate": 2.416644563368855e-05, "loss": 0.2099, "step": 63315 }, { "epoch": 2.95, "learning_rate": 2.4165977563852526e-05, "loss": 0.1157, "step": 63320 }, { "epoch": 2.95, "learning_rate": 2.4165509494016505e-05, "loss": 0.0709, "step": 63325 }, { "epoch": 2.96, "learning_rate": 2.416504142418049e-05, "loss": 0.0374, "step": 63330 }, { "epoch": 2.96, "learning_rate": 2.416457335434447e-05, "loss": 0.1054, "step": 63335 }, { "epoch": 2.96, "learning_rate": 2.4164105284508448e-05, "loss": 0.0839, "step": 63340 }, { "epoch": 2.96, "learning_rate": 2.4163637214672428e-05, "loss": 0.1125, "step": 63345 }, { "epoch": 2.96, "learning_rate": 2.416316914483641e-05, "loss": 0.0766, "step": 63350 }, { "epoch": 2.96, "learning_rate": 2.416270107500039e-05, "loss": 0.1371, "step": 63355 }, { "epoch": 2.96, "learning_rate": 2.416223300516437e-05, "loss": 0.1878, "step": 63360 }, { "epoch": 2.96, "learning_rate": 2.4161764935328354e-05, "loss": 0.3318, "step": 63365 }, { "epoch": 2.96, "learning_rate": 2.4161296865492334e-05, "loss": 0.0314, "step": 63370 }, { "epoch": 2.96, "learning_rate": 2.4160828795656314e-05, "loss": 0.049, "step": 63375 }, { "epoch": 2.96, "learning_rate": 2.4160360725820293e-05, "loss": 0.0562, "step": 63380 }, { "epoch": 2.96, "learning_rate": 2.4159892655984273e-05, "loss": 0.0936, "step": 63385 }, { "epoch": 2.96, "learning_rate": 2.4159424586148253e-05, "loss": 0.1393, "step": 63390 }, { "epoch": 2.96, "learning_rate": 2.4158956516312233e-05, "loss": 0.1414, "step": 63395 }, { "epoch": 2.96, "learning_rate": 2.4158488446476213e-05, "loss": 0.1436, "step": 63400 }, { "epoch": 2.96, "learning_rate": 2.4158020376640196e-05, "loss": 0.1471, "step": 63405 }, { "epoch": 2.96, "learning_rate": 2.4157552306804176e-05, "loss": 0.2445, "step": 63410 }, { "epoch": 2.96, "learning_rate": 2.4157084236968156e-05, "loss": 0.2649, "step": 63415 }, { "epoch": 2.96, "learning_rate": 2.415661616713214e-05, "loss": 0.0298, "step": 63420 }, { "epoch": 2.96, "learning_rate": 2.415614809729612e-05, "loss": 0.0304, "step": 63425 }, { "epoch": 2.96, "learning_rate": 2.41556800274601e-05, "loss": 0.0624, "step": 63430 }, { "epoch": 2.96, "learning_rate": 2.4155211957624078e-05, "loss": 0.0395, "step": 63435 }, { "epoch": 2.96, "learning_rate": 2.415474388778806e-05, "loss": 0.1157, "step": 63440 }, { "epoch": 2.96, "learning_rate": 2.4154275817952038e-05, "loss": 0.1552, "step": 63445 }, { "epoch": 2.96, "learning_rate": 2.4153807748116018e-05, "loss": 0.1065, "step": 63450 }, { "epoch": 2.96, "learning_rate": 2.4153339678279998e-05, "loss": 0.106, "step": 63455 }, { "epoch": 2.96, "learning_rate": 2.415287160844398e-05, "loss": 0.2909, "step": 63460 }, { "epoch": 2.96, "learning_rate": 2.415240353860796e-05, "loss": 0.2588, "step": 63465 }, { "epoch": 2.96, "learning_rate": 2.415193546877194e-05, "loss": 0.067, "step": 63470 }, { "epoch": 2.96, "learning_rate": 2.415146739893592e-05, "loss": 0.0726, "step": 63475 }, { "epoch": 2.96, "learning_rate": 2.4150999329099903e-05, "loss": 0.0086, "step": 63480 }, { "epoch": 2.96, "learning_rate": 2.4150531259263883e-05, "loss": 0.122, "step": 63485 }, { "epoch": 2.96, "learning_rate": 2.4150063189427863e-05, "loss": 0.0687, "step": 63490 }, { "epoch": 2.96, "learning_rate": 2.4149595119591846e-05, "loss": 0.0654, "step": 63495 }, { "epoch": 2.96, "learning_rate": 2.4149127049755826e-05, "loss": 0.0585, "step": 63500 }, { "epoch": 2.96, "learning_rate": 2.4148658979919806e-05, "loss": 0.1256, "step": 63505 }, { "epoch": 2.96, "learning_rate": 2.4148190910083782e-05, "loss": 0.2653, "step": 63510 }, { "epoch": 2.96, "learning_rate": 2.4147722840247766e-05, "loss": 0.2189, "step": 63515 }, { "epoch": 2.96, "learning_rate": 2.4147254770411745e-05, "loss": 0.0423, "step": 63520 }, { "epoch": 2.96, "learning_rate": 2.4146786700575725e-05, "loss": 0.0367, "step": 63525 }, { "epoch": 2.96, "learning_rate": 2.4146318630739705e-05, "loss": 0.044, "step": 63530 }, { "epoch": 2.96, "learning_rate": 2.4145850560903688e-05, "loss": 0.0619, "step": 63535 }, { "epoch": 2.96, "learning_rate": 2.4145382491067668e-05, "loss": 0.0601, "step": 63540 }, { "epoch": 2.97, "learning_rate": 2.4144914421231648e-05, "loss": 0.0659, "step": 63545 }, { "epoch": 2.97, "learning_rate": 2.414444635139563e-05, "loss": 0.1026, "step": 63550 }, { "epoch": 2.97, "learning_rate": 2.414397828155961e-05, "loss": 0.1708, "step": 63555 }, { "epoch": 2.97, "learning_rate": 2.414351021172359e-05, "loss": 0.1861, "step": 63560 }, { "epoch": 2.97, "learning_rate": 2.414304214188757e-05, "loss": 0.2588, "step": 63565 }, { "epoch": 2.97, "learning_rate": 2.4142574072051554e-05, "loss": 0.0468, "step": 63570 }, { "epoch": 2.97, "learning_rate": 2.414210600221553e-05, "loss": 0.0252, "step": 63575 }, { "epoch": 2.97, "learning_rate": 2.414163793237951e-05, "loss": 0.0415, "step": 63580 }, { "epoch": 2.97, "learning_rate": 2.414116986254349e-05, "loss": 0.0742, "step": 63585 }, { "epoch": 2.97, "learning_rate": 2.4140701792707473e-05, "loss": 0.0556, "step": 63590 }, { "epoch": 2.97, "learning_rate": 2.4140233722871453e-05, "loss": 0.0523, "step": 63595 }, { "epoch": 2.97, "learning_rate": 2.4139765653035433e-05, "loss": 0.129, "step": 63600 }, { "epoch": 2.97, "learning_rate": 2.4139297583199416e-05, "loss": 0.1882, "step": 63605 }, { "epoch": 2.97, "learning_rate": 2.4138829513363396e-05, "loss": 0.2611, "step": 63610 }, { "epoch": 2.97, "learning_rate": 2.4138361443527375e-05, "loss": 0.1537, "step": 63615 }, { "epoch": 2.97, "learning_rate": 2.4137893373691355e-05, "loss": 0.0419, "step": 63620 }, { "epoch": 2.97, "learning_rate": 2.413742530385534e-05, "loss": 0.0177, "step": 63625 }, { "epoch": 2.97, "learning_rate": 2.4136957234019318e-05, "loss": 0.0815, "step": 63630 }, { "epoch": 2.97, "learning_rate": 2.4136489164183295e-05, "loss": 0.057, "step": 63635 }, { "epoch": 2.97, "learning_rate": 2.4136021094347275e-05, "loss": 0.0947, "step": 63640 }, { "epoch": 2.97, "learning_rate": 2.4135553024511258e-05, "loss": 0.1252, "step": 63645 }, { "epoch": 2.97, "learning_rate": 2.4135084954675238e-05, "loss": 0.0784, "step": 63650 }, { "epoch": 2.97, "learning_rate": 2.4134616884839217e-05, "loss": 0.0991, "step": 63655 }, { "epoch": 2.97, "learning_rate": 2.4134148815003197e-05, "loss": 0.1926, "step": 63660 }, { "epoch": 2.97, "learning_rate": 2.413368074516718e-05, "loss": 0.4223, "step": 63665 }, { "epoch": 2.97, "learning_rate": 2.413321267533116e-05, "loss": 0.0536, "step": 63670 }, { "epoch": 2.97, "learning_rate": 2.413274460549514e-05, "loss": 0.0268, "step": 63675 }, { "epoch": 2.97, "learning_rate": 2.4132276535659123e-05, "loss": 0.0167, "step": 63680 }, { "epoch": 2.97, "learning_rate": 2.4131808465823103e-05, "loss": 0.0705, "step": 63685 }, { "epoch": 2.97, "learning_rate": 2.4131340395987083e-05, "loss": 0.0285, "step": 63690 }, { "epoch": 2.97, "learning_rate": 2.4130872326151063e-05, "loss": 0.0866, "step": 63695 }, { "epoch": 2.97, "learning_rate": 2.4130404256315042e-05, "loss": 0.0753, "step": 63700 }, { "epoch": 2.97, "learning_rate": 2.4129936186479022e-05, "loss": 0.1502, "step": 63705 }, { "epoch": 2.97, "learning_rate": 2.4129468116643002e-05, "loss": 0.2183, "step": 63710 }, { "epoch": 2.97, "learning_rate": 2.4129000046806982e-05, "loss": 0.2109, "step": 63715 }, { "epoch": 2.97, "learning_rate": 2.4128531976970965e-05, "loss": 0.084, "step": 63720 }, { "epoch": 2.97, "learning_rate": 2.4128063907134945e-05, "loss": 0.0534, "step": 63725 }, { "epoch": 2.97, "learning_rate": 2.4127595837298925e-05, "loss": 0.0606, "step": 63730 }, { "epoch": 2.97, "learning_rate": 2.4127127767462908e-05, "loss": 0.0455, "step": 63735 }, { "epoch": 2.97, "learning_rate": 2.4126659697626888e-05, "loss": 0.0354, "step": 63740 }, { "epoch": 2.97, "learning_rate": 2.4126191627790868e-05, "loss": 0.0265, "step": 63745 }, { "epoch": 2.97, "learning_rate": 2.4125723557954847e-05, "loss": 0.156, "step": 63750 }, { "epoch": 2.97, "learning_rate": 2.412525548811883e-05, "loss": 0.1119, "step": 63755 }, { "epoch": 2.98, "learning_rate": 2.4124787418282807e-05, "loss": 0.1648, "step": 63760 }, { "epoch": 2.98, "learning_rate": 2.4124319348446787e-05, "loss": 0.2408, "step": 63765 }, { "epoch": 2.98, "learning_rate": 2.4123851278610767e-05, "loss": 0.0432, "step": 63770 }, { "epoch": 2.98, "learning_rate": 2.412338320877475e-05, "loss": 0.0339, "step": 63775 }, { "epoch": 2.98, "learning_rate": 2.412291513893873e-05, "loss": 0.0596, "step": 63780 }, { "epoch": 2.98, "learning_rate": 2.412244706910271e-05, "loss": 0.0279, "step": 63785 }, { "epoch": 2.98, "learning_rate": 2.4121978999266693e-05, "loss": 0.0724, "step": 63790 }, { "epoch": 2.98, "learning_rate": 2.4121510929430673e-05, "loss": 0.0428, "step": 63795 }, { "epoch": 2.98, "learning_rate": 2.4121042859594652e-05, "loss": 0.1633, "step": 63800 }, { "epoch": 2.98, "learning_rate": 2.4120574789758632e-05, "loss": 0.1748, "step": 63805 }, { "epoch": 2.98, "learning_rate": 2.4120106719922615e-05, "loss": 0.2044, "step": 63810 }, { "epoch": 2.98, "learning_rate": 2.4119638650086595e-05, "loss": 0.3987, "step": 63815 }, { "epoch": 2.98, "learning_rate": 2.4119170580250575e-05, "loss": 0.053, "step": 63820 }, { "epoch": 2.98, "learning_rate": 2.411870251041455e-05, "loss": 0.0649, "step": 63825 }, { "epoch": 2.98, "learning_rate": 2.4118234440578535e-05, "loss": 0.0428, "step": 63830 }, { "epoch": 2.98, "learning_rate": 2.4117766370742514e-05, "loss": 0.0502, "step": 63835 }, { "epoch": 2.98, "learning_rate": 2.4117298300906494e-05, "loss": 0.0718, "step": 63840 }, { "epoch": 2.98, "learning_rate": 2.4116830231070474e-05, "loss": 0.0835, "step": 63845 }, { "epoch": 2.98, "learning_rate": 2.4116362161234457e-05, "loss": 0.11, "step": 63850 }, { "epoch": 2.98, "learning_rate": 2.4115894091398437e-05, "loss": 0.1532, "step": 63855 }, { "epoch": 2.98, "learning_rate": 2.4115426021562417e-05, "loss": 0.2339, "step": 63860 }, { "epoch": 2.98, "learning_rate": 2.41149579517264e-05, "loss": 0.342, "step": 63865 }, { "epoch": 2.98, "learning_rate": 2.411448988189038e-05, "loss": 0.0502, "step": 63870 }, { "epoch": 2.98, "learning_rate": 2.411402181205436e-05, "loss": 0.0223, "step": 63875 }, { "epoch": 2.98, "learning_rate": 2.411355374221834e-05, "loss": 0.026, "step": 63880 }, { "epoch": 2.98, "learning_rate": 2.4113085672382323e-05, "loss": 0.0227, "step": 63885 }, { "epoch": 2.98, "learning_rate": 2.41126176025463e-05, "loss": 0.1067, "step": 63890 }, { "epoch": 2.98, "learning_rate": 2.411214953271028e-05, "loss": 0.0624, "step": 63895 }, { "epoch": 2.98, "learning_rate": 2.411168146287426e-05, "loss": 0.1022, "step": 63900 }, { "epoch": 2.98, "learning_rate": 2.4111213393038242e-05, "loss": 0.1, "step": 63905 }, { "epoch": 2.98, "learning_rate": 2.4110745323202222e-05, "loss": 0.3413, "step": 63910 }, { "epoch": 2.98, "learning_rate": 2.4110277253366202e-05, "loss": 0.1548, "step": 63915 }, { "epoch": 2.98, "learning_rate": 2.4109809183530185e-05, "loss": 0.0298, "step": 63920 }, { "epoch": 2.98, "learning_rate": 2.4109341113694165e-05, "loss": 0.0499, "step": 63925 }, { "epoch": 2.98, "learning_rate": 2.4108873043858145e-05, "loss": 0.0825, "step": 63930 }, { "epoch": 2.98, "learning_rate": 2.4108404974022124e-05, "loss": 0.0281, "step": 63935 }, { "epoch": 2.98, "learning_rate": 2.4107936904186108e-05, "loss": 0.0823, "step": 63940 }, { "epoch": 2.98, "learning_rate": 2.4107468834350087e-05, "loss": 0.0811, "step": 63945 }, { "epoch": 2.98, "learning_rate": 2.4107000764514064e-05, "loss": 0.1565, "step": 63950 }, { "epoch": 2.98, "learning_rate": 2.4106532694678044e-05, "loss": 0.1216, "step": 63955 }, { "epoch": 2.98, "learning_rate": 2.4106064624842027e-05, "loss": 0.3768, "step": 63960 }, { "epoch": 2.98, "learning_rate": 2.4105596555006007e-05, "loss": 0.2631, "step": 63965 }, { "epoch": 2.98, "learning_rate": 2.4105128485169987e-05, "loss": 0.0358, "step": 63970 }, { "epoch": 2.99, "learning_rate": 2.410466041533397e-05, "loss": 0.0213, "step": 63975 }, { "epoch": 2.99, "learning_rate": 2.410419234549795e-05, "loss": 0.0607, "step": 63980 }, { "epoch": 2.99, "learning_rate": 2.410372427566193e-05, "loss": 0.0393, "step": 63985 }, { "epoch": 2.99, "learning_rate": 2.410325620582591e-05, "loss": 0.076, "step": 63990 }, { "epoch": 2.99, "learning_rate": 2.4102788135989892e-05, "loss": 0.1323, "step": 63995 }, { "epoch": 2.99, "learning_rate": 2.4102320066153872e-05, "loss": 0.1455, "step": 64000 }, { "epoch": 2.99, "learning_rate": 2.4101851996317852e-05, "loss": 0.2264, "step": 64005 }, { "epoch": 2.99, "learning_rate": 2.4101383926481832e-05, "loss": 0.1147, "step": 64010 }, { "epoch": 2.99, "learning_rate": 2.410091585664581e-05, "loss": 0.21, "step": 64015 }, { "epoch": 2.99, "learning_rate": 2.410044778680979e-05, "loss": 0.0451, "step": 64020 }, { "epoch": 2.99, "learning_rate": 2.409997971697377e-05, "loss": 0.029, "step": 64025 }, { "epoch": 2.99, "learning_rate": 2.409951164713775e-05, "loss": 0.069, "step": 64030 }, { "epoch": 2.99, "learning_rate": 2.4099043577301734e-05, "loss": 0.0565, "step": 64035 }, { "epoch": 2.99, "learning_rate": 2.4098575507465714e-05, "loss": 0.0509, "step": 64040 }, { "epoch": 2.99, "learning_rate": 2.4098107437629694e-05, "loss": 0.1369, "step": 64045 }, { "epoch": 2.99, "learning_rate": 2.4097639367793677e-05, "loss": 0.2191, "step": 64050 }, { "epoch": 2.99, "learning_rate": 2.4097171297957657e-05, "loss": 0.1317, "step": 64055 }, { "epoch": 2.99, "learning_rate": 2.4096703228121637e-05, "loss": 0.1902, "step": 64060 }, { "epoch": 2.99, "learning_rate": 2.4096235158285617e-05, "loss": 0.2301, "step": 64065 }, { "epoch": 2.99, "learning_rate": 2.40957670884496e-05, "loss": 0.0941, "step": 64070 }, { "epoch": 2.99, "learning_rate": 2.409529901861358e-05, "loss": 0.0223, "step": 64075 }, { "epoch": 2.99, "learning_rate": 2.4094830948777556e-05, "loss": 0.0256, "step": 64080 }, { "epoch": 2.99, "learning_rate": 2.4094362878941536e-05, "loss": 0.0912, "step": 64085 }, { "epoch": 2.99, "learning_rate": 2.409389480910552e-05, "loss": 0.0955, "step": 64090 }, { "epoch": 2.99, "learning_rate": 2.40934267392695e-05, "loss": 0.0974, "step": 64095 }, { "epoch": 2.99, "learning_rate": 2.409295866943348e-05, "loss": 0.0965, "step": 64100 }, { "epoch": 2.99, "learning_rate": 2.4092490599597462e-05, "loss": 0.1043, "step": 64105 }, { "epoch": 2.99, "learning_rate": 2.4092022529761442e-05, "loss": 0.1681, "step": 64110 }, { "epoch": 2.99, "learning_rate": 2.409155445992542e-05, "loss": 0.3309, "step": 64115 }, { "epoch": 2.99, "learning_rate": 2.40910863900894e-05, "loss": 0.0548, "step": 64120 }, { "epoch": 2.99, "learning_rate": 2.4090618320253385e-05, "loss": 0.0593, "step": 64125 }, { "epoch": 2.99, "learning_rate": 2.4090150250417364e-05, "loss": 0.02, "step": 64130 }, { "epoch": 2.99, "learning_rate": 2.4089682180581344e-05, "loss": 0.0772, "step": 64135 }, { "epoch": 2.99, "learning_rate": 2.408921411074532e-05, "loss": 0.0398, "step": 64140 }, { "epoch": 2.99, "learning_rate": 2.4088746040909304e-05, "loss": 0.0954, "step": 64145 }, { "epoch": 2.99, "learning_rate": 2.4088277971073284e-05, "loss": 0.0995, "step": 64150 }, { "epoch": 2.99, "learning_rate": 2.4087809901237263e-05, "loss": 0.11, "step": 64155 }, { "epoch": 2.99, "learning_rate": 2.4087341831401247e-05, "loss": 0.1444, "step": 64160 }, { "epoch": 2.99, "learning_rate": 2.4086873761565227e-05, "loss": 0.2479, "step": 64165 }, { "epoch": 2.99, "learning_rate": 2.4086405691729206e-05, "loss": 0.0511, "step": 64170 }, { "epoch": 2.99, "learning_rate": 2.4085937621893186e-05, "loss": 0.0359, "step": 64175 }, { "epoch": 2.99, "learning_rate": 2.408546955205717e-05, "loss": 0.0519, "step": 64180 }, { "epoch": 2.99, "learning_rate": 2.408500148222115e-05, "loss": 0.0396, "step": 64185 }, { "epoch": 3.0, "learning_rate": 2.408453341238513e-05, "loss": 0.091, "step": 64190 }, { "epoch": 3.0, "learning_rate": 2.408406534254911e-05, "loss": 0.1266, "step": 64195 }, { "epoch": 3.0, "learning_rate": 2.4083597272713092e-05, "loss": 0.1414, "step": 64200 }, { "epoch": 3.0, "learning_rate": 2.408312920287707e-05, "loss": 0.1365, "step": 64205 }, { "epoch": 3.0, "learning_rate": 2.4082661133041048e-05, "loss": 0.2704, "step": 64210 }, { "epoch": 3.0, "learning_rate": 2.408219306320503e-05, "loss": 0.248, "step": 64215 }, { "epoch": 3.0, "learning_rate": 2.408172499336901e-05, "loss": 0.0238, "step": 64220 }, { "epoch": 3.0, "learning_rate": 2.408125692353299e-05, "loss": 0.041, "step": 64225 }, { "epoch": 3.0, "learning_rate": 2.408078885369697e-05, "loss": 0.0768, "step": 64230 }, { "epoch": 3.0, "learning_rate": 2.4080320783860954e-05, "loss": 0.1015, "step": 64235 }, { "epoch": 3.0, "learning_rate": 2.4079852714024934e-05, "loss": 0.0356, "step": 64240 }, { "epoch": 3.0, "learning_rate": 2.4079384644188914e-05, "loss": 0.0705, "step": 64245 }, { "epoch": 3.0, "learning_rate": 2.4078916574352894e-05, "loss": 0.1228, "step": 64250 }, { "epoch": 3.0, "learning_rate": 2.4078448504516877e-05, "loss": 0.1553, "step": 64255 }, { "epoch": 3.0, "learning_rate": 2.4077980434680857e-05, "loss": 0.1956, "step": 64260 }, { "epoch": 3.0, "learning_rate": 2.4077512364844836e-05, "loss": 0.3237, "step": 64265 }, { "epoch": 3.0, "learning_rate": 2.4077044295008813e-05, "loss": 0.0495, "step": 64270 }, { "epoch": 3.0, "learning_rate": 2.4076576225172796e-05, "loss": 0.0439, "step": 64275 }, { "epoch": 3.0, "learning_rate": 2.4076108155336776e-05, "loss": 0.0511, "step": 64280 }, { "epoch": 3.0, "learning_rate": 2.4075640085500756e-05, "loss": 0.0824, "step": 64285 }, { "epoch": 3.0, "learning_rate": 2.407517201566474e-05, "loss": 0.1603, "step": 64290 }, { "epoch": 3.0, "eval_cer": 0.011836236095430231, "eval_loss": 0.024219496175646782, "eval_runtime": 393.8928, "eval_samples_per_second": 48.363, "eval_steps_per_second": 12.092, "eval_wer": 0.10190570947497321, "step": 64293 } ], "max_steps": 321465, "num_train_epochs": 15, "total_flos": 7.733298332493346e+18, "trial_name": null, "trial_params": null }