{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.8625442354255912, "global_step": 10000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 2e-09, "loss": 23.1708, "step": 5 }, { "epoch": 0.0, "learning_rate": 1.2e-08, "loss": 17.4647, "step": 10 }, { "epoch": 0.0, "learning_rate": 2.2e-08, "loss": 16.726, "step": 15 }, { "epoch": 0.0, "learning_rate": 3.2e-08, "loss": 16.6276, "step": 20 }, { "epoch": 0.0, "learning_rate": 4.2e-08, "loss": 14.8649, "step": 25 }, { "epoch": 0.01, "learning_rate": 5.1999999999999996e-08, "loss": 15.1187, "step": 30 }, { "epoch": 0.01, "learning_rate": 6.2e-08, "loss": 14.6901, "step": 35 }, { "epoch": 0.01, "learning_rate": 7.2e-08, "loss": 14.3665, "step": 40 }, { "epoch": 0.01, "learning_rate": 8.2e-08, "loss": 14.6373, "step": 45 }, { "epoch": 0.01, "learning_rate": 9.199999999999999e-08, "loss": 14.0672, "step": 50 }, { "epoch": 0.01, "learning_rate": 1.0199999999999999e-07, "loss": 18.1758, "step": 55 }, { "epoch": 0.01, "learning_rate": 1.12e-07, "loss": 18.3919, "step": 60 }, { "epoch": 0.01, "learning_rate": 1.2199999999999998e-07, "loss": 17.8389, "step": 65 }, { "epoch": 0.01, "learning_rate": 1.32e-07, "loss": 15.098, "step": 70 }, { "epoch": 0.01, "learning_rate": 1.4199999999999997e-07, "loss": 15.2273, "step": 75 }, { "epoch": 0.01, "learning_rate": 1.5199999999999998e-07, "loss": 14.9402, "step": 80 }, { "epoch": 0.02, "learning_rate": 1.62e-07, "loss": 14.2399, "step": 85 }, { "epoch": 0.02, "learning_rate": 1.7199999999999998e-07, "loss": 14.0017, "step": 90 }, { "epoch": 0.02, "learning_rate": 1.82e-07, "loss": 14.2729, "step": 95 }, { "epoch": 0.02, "learning_rate": 1.92e-07, "loss": 14.4501, "step": 100 }, { "epoch": 0.02, "learning_rate": 2.02e-07, "loss": 17.7063, "step": 105 }, { "epoch": 0.02, "learning_rate": 2.12e-07, "loss": 17.2415, "step": 110 }, { "epoch": 0.02, "learning_rate": 2.22e-07, "loss": 17.3164, "step": 115 }, { "epoch": 0.02, "learning_rate": 2.32e-07, "loss": 15.1226, "step": 120 }, { "epoch": 0.02, "learning_rate": 2.4199999999999997e-07, "loss": 14.7632, "step": 125 }, { "epoch": 0.02, "learning_rate": 2.52e-07, "loss": 14.269, "step": 130 }, { "epoch": 0.03, "learning_rate": 2.62e-07, "loss": 14.3178, "step": 135 }, { "epoch": 0.03, "learning_rate": 2.72e-07, "loss": 13.9972, "step": 140 }, { "epoch": 0.03, "learning_rate": 2.8199999999999996e-07, "loss": 14.4335, "step": 145 }, { "epoch": 0.03, "learning_rate": 2.9199999999999997e-07, "loss": 14.1463, "step": 150 }, { "epoch": 0.03, "learning_rate": 3.02e-07, "loss": 16.1839, "step": 155 }, { "epoch": 0.03, "learning_rate": 3.12e-07, "loss": 17.0019, "step": 160 }, { "epoch": 0.03, "learning_rate": 3.22e-07, "loss": 16.061, "step": 165 }, { "epoch": 0.03, "learning_rate": 3.32e-07, "loss": 13.9992, "step": 170 }, { "epoch": 0.03, "learning_rate": 3.42e-07, "loss": 14.0072, "step": 175 }, { "epoch": 0.03, "learning_rate": 3.52e-07, "loss": 13.4026, "step": 180 }, { "epoch": 0.03, "learning_rate": 3.62e-07, "loss": 13.1469, "step": 185 }, { "epoch": 0.04, "learning_rate": 3.72e-07, "loss": 13.2259, "step": 190 }, { "epoch": 0.04, "learning_rate": 3.82e-07, "loss": 13.0371, "step": 195 }, { "epoch": 0.04, "learning_rate": 3.8999999999999997e-07, "loss": 14.4267, "step": 200 }, { "epoch": 0.04, "learning_rate": 4e-07, "loss": 17.5423, "step": 205 }, { "epoch": 0.04, "learning_rate": 4.0999999999999994e-07, "loss": 15.3916, "step": 210 }, { "epoch": 0.04, "learning_rate": 4.1999999999999995e-07, "loss": 14.2848, "step": 215 }, { "epoch": 0.04, "learning_rate": 4.2999999999999996e-07, "loss": 15.4372, "step": 220 }, { "epoch": 0.04, "learning_rate": 4.3999999999999997e-07, "loss": 12.914, "step": 225 }, { "epoch": 0.04, "learning_rate": 4.5e-07, "loss": 12.8913, "step": 230 }, { "epoch": 0.04, "learning_rate": 4.6e-07, "loss": 12.2806, "step": 235 }, { "epoch": 0.04, "learning_rate": 4.6999999999999995e-07, "loss": 11.734, "step": 240 }, { "epoch": 0.05, "learning_rate": 4.8e-07, "loss": 10.8097, "step": 245 }, { "epoch": 0.05, "learning_rate": 4.9e-07, "loss": 11.4272, "step": 250 }, { "epoch": 0.05, "learning_rate": 5e-07, "loss": 14.283, "step": 255 }, { "epoch": 0.05, "learning_rate": 5.1e-07, "loss": 12.6282, "step": 260 }, { "epoch": 0.05, "learning_rate": 5.2e-07, "loss": 13.4567, "step": 265 }, { "epoch": 0.05, "learning_rate": 5.3e-07, "loss": 11.2645, "step": 270 }, { "epoch": 0.05, "learning_rate": 5.4e-07, "loss": 10.2447, "step": 275 }, { "epoch": 0.05, "learning_rate": 5.5e-07, "loss": 10.9771, "step": 280 }, { "epoch": 0.05, "learning_rate": 5.6e-07, "loss": 10.9162, "step": 285 }, { "epoch": 0.05, "learning_rate": 5.699999999999999e-07, "loss": 10.2676, "step": 290 }, { "epoch": 0.05, "learning_rate": 5.8e-07, "loss": 8.9776, "step": 295 }, { "epoch": 0.06, "learning_rate": 5.9e-07, "loss": 9.0533, "step": 300 }, { "epoch": 0.06, "learning_rate": 5.979999999999999e-07, "loss": 14.5471, "step": 305 }, { "epoch": 0.06, "learning_rate": 6.079999999999999e-07, "loss": 12.7666, "step": 310 }, { "epoch": 0.06, "learning_rate": 6.18e-07, "loss": 9.7622, "step": 315 }, { "epoch": 0.06, "learning_rate": 6.28e-07, "loss": 9.0693, "step": 320 }, { "epoch": 0.06, "learning_rate": 6.38e-07, "loss": 9.1909, "step": 325 }, { "epoch": 0.06, "learning_rate": 6.48e-07, "loss": 9.2916, "step": 330 }, { "epoch": 0.06, "learning_rate": 6.58e-07, "loss": 7.4726, "step": 335 }, { "epoch": 0.06, "learning_rate": 6.68e-07, "loss": 9.438, "step": 340 }, { "epoch": 0.06, "learning_rate": 6.78e-07, "loss": 6.7858, "step": 345 }, { "epoch": 0.07, "learning_rate": 6.879999999999999e-07, "loss": 6.6149, "step": 350 }, { "epoch": 0.07, "learning_rate": 6.979999999999999e-07, "loss": 10.9064, "step": 355 }, { "epoch": 0.07, "learning_rate": 7.079999999999999e-07, "loss": 6.9364, "step": 360 }, { "epoch": 0.07, "learning_rate": 7.179999999999999e-07, "loss": 8.5146, "step": 365 }, { "epoch": 0.07, "learning_rate": 7.28e-07, "loss": 5.9193, "step": 370 }, { "epoch": 0.07, "learning_rate": 7.38e-07, "loss": 6.343, "step": 375 }, { "epoch": 0.07, "learning_rate": 7.48e-07, "loss": 4.9407, "step": 380 }, { "epoch": 0.07, "learning_rate": 7.58e-07, "loss": 6.2009, "step": 385 }, { "epoch": 0.07, "learning_rate": 7.68e-07, "loss": 6.2122, "step": 390 }, { "epoch": 0.07, "learning_rate": 7.78e-07, "loss": 6.305, "step": 395 }, { "epoch": 0.07, "learning_rate": 7.88e-07, "loss": 6.0134, "step": 400 }, { "epoch": 0.08, "learning_rate": 7.98e-07, "loss": 6.3882, "step": 405 }, { "epoch": 0.08, "learning_rate": 8.08e-07, "loss": 7.8588, "step": 410 }, { "epoch": 0.08, "learning_rate": 8.179999999999999e-07, "loss": 4.2314, "step": 415 }, { "epoch": 0.08, "learning_rate": 8.28e-07, "loss": 5.1779, "step": 420 }, { "epoch": 0.08, "learning_rate": 8.38e-07, "loss": 3.6178, "step": 425 }, { "epoch": 0.08, "learning_rate": 8.48e-07, "loss": 3.8777, "step": 430 }, { "epoch": 0.08, "learning_rate": 8.58e-07, "loss": 6.0639, "step": 435 }, { "epoch": 0.08, "learning_rate": 8.68e-07, "loss": 4.7344, "step": 440 }, { "epoch": 0.08, "learning_rate": 8.78e-07, "loss": 4.4478, "step": 445 }, { "epoch": 0.08, "learning_rate": 8.88e-07, "loss": 3.6034, "step": 450 }, { "epoch": 0.08, "learning_rate": 8.98e-07, "loss": 3.4378, "step": 455 }, { "epoch": 0.09, "learning_rate": 9.08e-07, "loss": 3.8416, "step": 460 }, { "epoch": 0.09, "learning_rate": 9.18e-07, "loss": 3.3023, "step": 465 }, { "epoch": 0.09, "learning_rate": 9.28e-07, "loss": 3.3863, "step": 470 }, { "epoch": 0.09, "learning_rate": 9.379999999999998e-07, "loss": 3.2869, "step": 475 }, { "epoch": 0.09, "learning_rate": 9.479999999999999e-07, "loss": 3.3075, "step": 480 }, { "epoch": 0.09, "learning_rate": 9.58e-07, "loss": 4.247, "step": 485 }, { "epoch": 0.09, "learning_rate": 9.679999999999999e-07, "loss": 3.5028, "step": 490 }, { "epoch": 0.09, "learning_rate": 9.78e-07, "loss": 5.2398, "step": 495 }, { "epoch": 0.09, "learning_rate": 9.88e-07, "loss": 4.051, "step": 500 }, { "epoch": 0.09, "eval_loss": 3.1419003009796143, "eval_runtime": 101.742, "eval_samples_per_second": 11.736, "eval_steps_per_second": 1.474, "eval_wer": 1.0, "step": 500 }, { "epoch": 0.09, "learning_rate": 9.98e-07, "loss": 4.4604, "step": 505 }, { "epoch": 0.09, "learning_rate": 1.008e-06, "loss": 3.2391, "step": 510 }, { "epoch": 0.1, "learning_rate": 1.018e-06, "loss": 3.1886, "step": 515 }, { "epoch": 0.1, "learning_rate": 1.028e-06, "loss": 3.132, "step": 520 }, { "epoch": 0.1, "learning_rate": 1.038e-06, "loss": 3.2774, "step": 525 }, { "epoch": 0.1, "learning_rate": 1.048e-06, "loss": 3.6977, "step": 530 }, { "epoch": 0.1, "learning_rate": 1.058e-06, "loss": 3.3265, "step": 535 }, { "epoch": 0.1, "learning_rate": 1.068e-06, "loss": 3.095, "step": 540 }, { "epoch": 0.1, "learning_rate": 1.078e-06, "loss": 3.3518, "step": 545 }, { "epoch": 0.1, "learning_rate": 1.088e-06, "loss": 3.1893, "step": 550 }, { "epoch": 0.1, "learning_rate": 1.0980000000000001e-06, "loss": 3.1898, "step": 555 }, { "epoch": 0.1, "learning_rate": 1.108e-06, "loss": 4.1224, "step": 560 }, { "epoch": 0.11, "learning_rate": 1.1180000000000001e-06, "loss": 3.0569, "step": 565 }, { "epoch": 0.11, "learning_rate": 1.1279999999999998e-06, "loss": 5.0761, "step": 570 }, { "epoch": 0.11, "learning_rate": 1.138e-06, "loss": 4.0581, "step": 575 }, { "epoch": 0.11, "learning_rate": 1.1479999999999999e-06, "loss": 3.6377, "step": 580 }, { "epoch": 0.11, "learning_rate": 1.158e-06, "loss": 3.4788, "step": 585 }, { "epoch": 0.11, "learning_rate": 1.1679999999999999e-06, "loss": 3.0537, "step": 590 }, { "epoch": 0.11, "learning_rate": 1.178e-06, "loss": 3.0412, "step": 595 }, { "epoch": 0.11, "learning_rate": 1.1879999999999999e-06, "loss": 3.1245, "step": 600 }, { "epoch": 0.11, "learning_rate": 1.1979999999999998e-06, "loss": 3.2805, "step": 605 }, { "epoch": 0.11, "learning_rate": 1.208e-06, "loss": 2.9896, "step": 610 }, { "epoch": 0.11, "learning_rate": 1.2179999999999998e-06, "loss": 3.5087, "step": 615 }, { "epoch": 0.12, "learning_rate": 1.228e-06, "loss": 3.4038, "step": 620 }, { "epoch": 0.12, "learning_rate": 1.2379999999999998e-06, "loss": 3.5193, "step": 625 }, { "epoch": 0.12, "learning_rate": 1.248e-06, "loss": 3.0089, "step": 630 }, { "epoch": 0.12, "learning_rate": 1.2579999999999999e-06, "loss": 2.9954, "step": 635 }, { "epoch": 0.12, "learning_rate": 1.268e-06, "loss": 3.0579, "step": 640 }, { "epoch": 0.12, "learning_rate": 1.2779999999999999e-06, "loss": 3.8326, "step": 645 }, { "epoch": 0.12, "learning_rate": 1.288e-06, "loss": 3.2384, "step": 650 }, { "epoch": 0.12, "learning_rate": 1.298e-06, "loss": 3.452, "step": 655 }, { "epoch": 0.12, "learning_rate": 1.308e-06, "loss": 3.7447, "step": 660 }, { "epoch": 0.12, "learning_rate": 1.318e-06, "loss": 3.5435, "step": 665 }, { "epoch": 0.12, "learning_rate": 1.328e-06, "loss": 3.0208, "step": 670 }, { "epoch": 0.13, "learning_rate": 1.338e-06, "loss": 3.3091, "step": 675 }, { "epoch": 0.13, "learning_rate": 1.348e-06, "loss": 3.3663, "step": 680 }, { "epoch": 0.13, "learning_rate": 1.358e-06, "loss": 2.9651, "step": 685 }, { "epoch": 0.13, "learning_rate": 1.368e-06, "loss": 2.9384, "step": 690 }, { "epoch": 0.13, "learning_rate": 1.3779999999999998e-06, "loss": 2.9415, "step": 695 }, { "epoch": 0.13, "learning_rate": 1.3879999999999999e-06, "loss": 2.9731, "step": 700 }, { "epoch": 0.13, "learning_rate": 1.3979999999999998e-06, "loss": 3.4581, "step": 705 }, { "epoch": 0.13, "learning_rate": 1.408e-06, "loss": 2.9852, "step": 710 }, { "epoch": 0.13, "learning_rate": 1.4179999999999998e-06, "loss": 2.9559, "step": 715 }, { "epoch": 0.13, "learning_rate": 1.428e-06, "loss": 3.2803, "step": 720 }, { "epoch": 0.14, "learning_rate": 1.4379999999999998e-06, "loss": 2.9441, "step": 725 }, { "epoch": 0.14, "learning_rate": 1.448e-06, "loss": 3.9254, "step": 730 }, { "epoch": 0.14, "learning_rate": 1.4579999999999998e-06, "loss": 3.1551, "step": 735 }, { "epoch": 0.14, "learning_rate": 1.468e-06, "loss": 5.0411, "step": 740 }, { "epoch": 0.14, "learning_rate": 1.4779999999999999e-06, "loss": 2.9128, "step": 745 }, { "epoch": 0.14, "learning_rate": 1.488e-06, "loss": 3.168, "step": 750 }, { "epoch": 0.14, "learning_rate": 1.4979999999999999e-06, "loss": 3.1407, "step": 755 }, { "epoch": 0.14, "learning_rate": 1.508e-06, "loss": 3.2372, "step": 760 }, { "epoch": 0.14, "learning_rate": 1.518e-06, "loss": 3.2442, "step": 765 }, { "epoch": 0.14, "learning_rate": 1.528e-06, "loss": 2.9669, "step": 770 }, { "epoch": 0.14, "learning_rate": 1.538e-06, "loss": 3.0412, "step": 775 }, { "epoch": 0.15, "learning_rate": 1.548e-06, "loss": 2.918, "step": 780 }, { "epoch": 0.15, "learning_rate": 1.558e-06, "loss": 3.2461, "step": 785 }, { "epoch": 0.15, "learning_rate": 1.568e-06, "loss": 2.9181, "step": 790 }, { "epoch": 0.15, "learning_rate": 1.578e-06, "loss": 3.2525, "step": 795 }, { "epoch": 0.15, "learning_rate": 1.588e-06, "loss": 3.3201, "step": 800 }, { "epoch": 0.15, "learning_rate": 1.598e-06, "loss": 3.0265, "step": 805 }, { "epoch": 0.15, "learning_rate": 1.608e-06, "loss": 2.886, "step": 810 }, { "epoch": 0.15, "learning_rate": 1.618e-06, "loss": 3.0627, "step": 815 }, { "epoch": 0.15, "learning_rate": 1.628e-06, "loss": 2.9099, "step": 820 }, { "epoch": 0.15, "learning_rate": 1.6379999999999998e-06, "loss": 2.9201, "step": 825 }, { "epoch": 0.15, "learning_rate": 1.648e-06, "loss": 3.0767, "step": 830 }, { "epoch": 0.16, "learning_rate": 1.6579999999999998e-06, "loss": 2.8778, "step": 835 }, { "epoch": 0.16, "learning_rate": 1.668e-06, "loss": 2.995, "step": 840 }, { "epoch": 0.16, "learning_rate": 1.6779999999999999e-06, "loss": 2.8796, "step": 845 }, { "epoch": 0.16, "learning_rate": 1.6879999999999998e-06, "loss": 2.9054, "step": 850 }, { "epoch": 0.16, "learning_rate": 1.6979999999999999e-06, "loss": 3.3809, "step": 855 }, { "epoch": 0.16, "learning_rate": 1.7079999999999998e-06, "loss": 2.9066, "step": 860 }, { "epoch": 0.16, "learning_rate": 1.718e-06, "loss": 2.9675, "step": 865 }, { "epoch": 0.16, "learning_rate": 1.7279999999999998e-06, "loss": 3.2022, "step": 870 }, { "epoch": 0.16, "learning_rate": 1.738e-06, "loss": 2.9072, "step": 875 }, { "epoch": 0.16, "learning_rate": 1.7479999999999998e-06, "loss": 3.569, "step": 880 }, { "epoch": 0.16, "learning_rate": 1.758e-06, "loss": 3.073, "step": 885 }, { "epoch": 0.17, "learning_rate": 1.7679999999999998e-06, "loss": 3.1634, "step": 890 }, { "epoch": 0.17, "learning_rate": 1.778e-06, "loss": 2.9443, "step": 895 }, { "epoch": 0.17, "learning_rate": 1.7879999999999999e-06, "loss": 2.9246, "step": 900 }, { "epoch": 0.17, "learning_rate": 1.798e-06, "loss": 2.9731, "step": 905 }, { "epoch": 0.17, "learning_rate": 1.8079999999999999e-06, "loss": 2.8715, "step": 910 }, { "epoch": 0.17, "learning_rate": 1.818e-06, "loss": 2.8709, "step": 915 }, { "epoch": 0.17, "learning_rate": 1.828e-06, "loss": 2.8895, "step": 920 }, { "epoch": 0.17, "learning_rate": 1.838e-06, "loss": 2.9853, "step": 925 }, { "epoch": 0.17, "learning_rate": 1.848e-06, "loss": 2.9241, "step": 930 }, { "epoch": 0.17, "learning_rate": 1.858e-06, "loss": 2.8844, "step": 935 }, { "epoch": 0.18, "learning_rate": 1.868e-06, "loss": 2.871, "step": 940 }, { "epoch": 0.18, "learning_rate": 1.8779999999999998e-06, "loss": 2.8814, "step": 945 }, { "epoch": 0.18, "learning_rate": 1.8879999999999998e-06, "loss": 2.8748, "step": 950 }, { "epoch": 0.18, "learning_rate": 1.8979999999999999e-06, "loss": 3.0096, "step": 955 }, { "epoch": 0.18, "learning_rate": 1.9079999999999998e-06, "loss": 2.8701, "step": 960 }, { "epoch": 0.18, "learning_rate": 1.9179999999999997e-06, "loss": 2.8968, "step": 965 }, { "epoch": 0.18, "learning_rate": 1.928e-06, "loss": 3.0461, "step": 970 }, { "epoch": 0.18, "learning_rate": 1.938e-06, "loss": 2.9494, "step": 975 }, { "epoch": 0.18, "learning_rate": 1.948e-06, "loss": 4.5448, "step": 980 }, { "epoch": 0.18, "learning_rate": 1.9579999999999997e-06, "loss": 2.9924, "step": 985 }, { "epoch": 0.18, "learning_rate": 1.968e-06, "loss": 2.9612, "step": 990 }, { "epoch": 0.19, "learning_rate": 1.978e-06, "loss": 3.2622, "step": 995 }, { "epoch": 0.19, "learning_rate": 1.988e-06, "loss": 2.8626, "step": 1000 }, { "epoch": 0.19, "eval_loss": 2.8933510780334473, "eval_runtime": 138.873, "eval_samples_per_second": 8.598, "eval_steps_per_second": 1.08, "eval_wer": 1.0, "step": 1000 }, { "epoch": 0.19, "learning_rate": 1.9979999999999998e-06, "loss": 2.8847, "step": 1005 }, { "epoch": 0.19, "learning_rate": 1.9991784760731156e-06, "loss": 2.8753, "step": 1010 }, { "epoch": 0.19, "learning_rate": 1.99815157116451e-06, "loss": 2.9105, "step": 1015 }, { "epoch": 0.19, "learning_rate": 1.9971246662559046e-06, "loss": 2.9129, "step": 1020 }, { "epoch": 0.19, "learning_rate": 1.9960977613472992e-06, "loss": 3.202, "step": 1025 }, { "epoch": 0.19, "learning_rate": 1.995070856438694e-06, "loss": 3.0123, "step": 1030 }, { "epoch": 0.19, "learning_rate": 1.9940439515300882e-06, "loss": 3.4388, "step": 1035 }, { "epoch": 0.19, "learning_rate": 1.993017046621483e-06, "loss": 2.9159, "step": 1040 }, { "epoch": 0.19, "learning_rate": 1.991990141712877e-06, "loss": 2.9784, "step": 1045 }, { "epoch": 0.2, "learning_rate": 1.990963236804272e-06, "loss": 2.8644, "step": 1050 }, { "epoch": 0.2, "learning_rate": 1.989936331895666e-06, "loss": 2.8842, "step": 1055 }, { "epoch": 0.2, "learning_rate": 1.988909426987061e-06, "loss": 2.9332, "step": 1060 }, { "epoch": 0.2, "learning_rate": 1.9878825220784556e-06, "loss": 2.8966, "step": 1065 }, { "epoch": 0.2, "learning_rate": 1.9868556171698503e-06, "loss": 2.9699, "step": 1070 }, { "epoch": 0.2, "learning_rate": 1.9858287122612445e-06, "loss": 2.8659, "step": 1075 }, { "epoch": 0.2, "learning_rate": 1.9848018073526392e-06, "loss": 2.9186, "step": 1080 }, { "epoch": 0.2, "learning_rate": 1.9837749024440335e-06, "loss": 2.871, "step": 1085 }, { "epoch": 0.2, "learning_rate": 1.982747997535428e-06, "loss": 2.8712, "step": 1090 }, { "epoch": 0.2, "learning_rate": 1.9817210926268225e-06, "loss": 2.839, "step": 1095 }, { "epoch": 0.2, "learning_rate": 1.980694187718217e-06, "loss": 2.9186, "step": 1100 }, { "epoch": 0.21, "learning_rate": 1.979667282809612e-06, "loss": 3.1007, "step": 1105 }, { "epoch": 0.21, "learning_rate": 1.9786403779010066e-06, "loss": 2.8781, "step": 1110 }, { "epoch": 0.21, "learning_rate": 1.977613472992401e-06, "loss": 3.292, "step": 1115 }, { "epoch": 0.21, "learning_rate": 1.9765865680837955e-06, "loss": 3.0817, "step": 1120 }, { "epoch": 0.21, "learning_rate": 1.97555966317519e-06, "loss": 2.9705, "step": 1125 }, { "epoch": 0.21, "learning_rate": 1.9745327582665845e-06, "loss": 2.9304, "step": 1130 }, { "epoch": 0.21, "learning_rate": 1.9735058533579788e-06, "loss": 2.8766, "step": 1135 }, { "epoch": 0.21, "learning_rate": 1.9724789484493735e-06, "loss": 2.8477, "step": 1140 }, { "epoch": 0.21, "learning_rate": 1.971452043540768e-06, "loss": 2.9345, "step": 1145 }, { "epoch": 0.21, "learning_rate": 1.9704251386321624e-06, "loss": 2.9276, "step": 1150 }, { "epoch": 0.22, "learning_rate": 1.969398233723557e-06, "loss": 2.9444, "step": 1155 }, { "epoch": 0.22, "learning_rate": 1.968371328814952e-06, "loss": 2.9099, "step": 1160 }, { "epoch": 0.22, "learning_rate": 1.967344423906346e-06, "loss": 2.8898, "step": 1165 }, { "epoch": 0.22, "learning_rate": 1.966317518997741e-06, "loss": 2.9451, "step": 1170 }, { "epoch": 0.22, "learning_rate": 1.965290614089135e-06, "loss": 2.87, "step": 1175 }, { "epoch": 0.22, "learning_rate": 1.9642637091805298e-06, "loss": 2.9508, "step": 1180 }, { "epoch": 0.22, "learning_rate": 1.9632368042719245e-06, "loss": 2.9362, "step": 1185 }, { "epoch": 0.22, "learning_rate": 1.9622098993633187e-06, "loss": 2.9366, "step": 1190 }, { "epoch": 0.22, "learning_rate": 1.9611829944547134e-06, "loss": 3.0603, "step": 1195 }, { "epoch": 0.22, "learning_rate": 1.960156089546108e-06, "loss": 2.8994, "step": 1200 }, { "epoch": 0.22, "learning_rate": 1.9591291846375024e-06, "loss": 3.0001, "step": 1205 }, { "epoch": 0.23, "learning_rate": 1.958102279728897e-06, "loss": 2.8472, "step": 1210 }, { "epoch": 0.23, "learning_rate": 1.9570753748202914e-06, "loss": 2.9021, "step": 1215 }, { "epoch": 0.23, "learning_rate": 1.956048469911686e-06, "loss": 2.8638, "step": 1220 }, { "epoch": 0.23, "learning_rate": 1.9550215650030808e-06, "loss": 2.8506, "step": 1225 }, { "epoch": 0.23, "learning_rate": 1.953994660094475e-06, "loss": 2.8397, "step": 1230 }, { "epoch": 0.23, "learning_rate": 1.9529677551858697e-06, "loss": 3.468, "step": 1235 }, { "epoch": 0.23, "learning_rate": 1.9519408502772644e-06, "loss": 2.9221, "step": 1240 }, { "epoch": 0.23, "learning_rate": 1.9509139453686587e-06, "loss": 3.2824, "step": 1245 }, { "epoch": 0.23, "learning_rate": 1.9498870404600534e-06, "loss": 2.9593, "step": 1250 }, { "epoch": 0.23, "learning_rate": 1.9488601355514477e-06, "loss": 2.9068, "step": 1255 }, { "epoch": 0.23, "learning_rate": 1.9478332306428424e-06, "loss": 2.8651, "step": 1260 }, { "epoch": 0.24, "learning_rate": 1.946806325734237e-06, "loss": 2.8539, "step": 1265 }, { "epoch": 0.24, "learning_rate": 1.9457794208256313e-06, "loss": 2.955, "step": 1270 }, { "epoch": 0.24, "learning_rate": 1.944752515917026e-06, "loss": 3.0867, "step": 1275 }, { "epoch": 0.24, "learning_rate": 1.9437256110084207e-06, "loss": 2.894, "step": 1280 }, { "epoch": 0.24, "learning_rate": 1.942698706099815e-06, "loss": 2.8836, "step": 1285 }, { "epoch": 0.24, "learning_rate": 1.9416718011912097e-06, "loss": 2.8627, "step": 1290 }, { "epoch": 0.24, "learning_rate": 1.940644896282604e-06, "loss": 2.8886, "step": 1295 }, { "epoch": 0.24, "learning_rate": 1.9396179913739987e-06, "loss": 2.91, "step": 1300 }, { "epoch": 0.24, "learning_rate": 1.938591086465393e-06, "loss": 2.8789, "step": 1305 }, { "epoch": 0.24, "learning_rate": 1.9375641815567877e-06, "loss": 2.9238, "step": 1310 }, { "epoch": 0.24, "learning_rate": 1.9365372766481823e-06, "loss": 2.9038, "step": 1315 }, { "epoch": 0.25, "learning_rate": 1.935510371739577e-06, "loss": 2.8612, "step": 1320 }, { "epoch": 0.25, "learning_rate": 1.9344834668309713e-06, "loss": 2.8983, "step": 1325 }, { "epoch": 0.25, "learning_rate": 1.933456561922366e-06, "loss": 2.878, "step": 1330 }, { "epoch": 0.25, "learning_rate": 1.9324296570137603e-06, "loss": 2.8546, "step": 1335 }, { "epoch": 0.25, "learning_rate": 1.931402752105155e-06, "loss": 3.1733, "step": 1340 }, { "epoch": 0.25, "learning_rate": 1.9303758471965493e-06, "loss": 3.0895, "step": 1345 }, { "epoch": 0.25, "learning_rate": 1.929348942287944e-06, "loss": 2.8654, "step": 1350 }, { "epoch": 0.25, "learning_rate": 1.9283220373793387e-06, "loss": 2.9303, "step": 1355 }, { "epoch": 0.25, "learning_rate": 1.9272951324707333e-06, "loss": 2.9429, "step": 1360 }, { "epoch": 0.25, "learning_rate": 1.9262682275621276e-06, "loss": 2.8373, "step": 1365 }, { "epoch": 0.26, "learning_rate": 1.9252413226535223e-06, "loss": 2.9104, "step": 1370 }, { "epoch": 0.26, "learning_rate": 1.9242144177449166e-06, "loss": 2.8777, "step": 1375 }, { "epoch": 0.26, "learning_rate": 1.9231875128363113e-06, "loss": 3.0137, "step": 1380 }, { "epoch": 0.26, "learning_rate": 1.9221606079277056e-06, "loss": 2.9153, "step": 1385 }, { "epoch": 0.26, "learning_rate": 1.9211337030191003e-06, "loss": 2.8599, "step": 1390 }, { "epoch": 0.26, "learning_rate": 1.920106798110495e-06, "loss": 3.5802, "step": 1395 }, { "epoch": 0.26, "learning_rate": 1.9190798932018897e-06, "loss": 2.8061, "step": 1400 }, { "epoch": 0.26, "learning_rate": 1.918052988293284e-06, "loss": 2.9472, "step": 1405 }, { "epoch": 0.26, "learning_rate": 1.9170260833846786e-06, "loss": 2.8901, "step": 1410 }, { "epoch": 0.26, "learning_rate": 1.915999178476073e-06, "loss": 3.3134, "step": 1415 }, { "epoch": 0.26, "learning_rate": 1.9149722735674676e-06, "loss": 2.8888, "step": 1420 }, { "epoch": 0.27, "learning_rate": 1.913945368658862e-06, "loss": 2.8508, "step": 1425 }, { "epoch": 0.27, "learning_rate": 1.9129184637502566e-06, "loss": 2.8544, "step": 1430 }, { "epoch": 0.27, "learning_rate": 1.9118915588416513e-06, "loss": 2.8625, "step": 1435 }, { "epoch": 0.27, "learning_rate": 1.9108646539330455e-06, "loss": 2.8544, "step": 1440 }, { "epoch": 0.27, "learning_rate": 1.9098377490244402e-06, "loss": 2.8768, "step": 1445 }, { "epoch": 0.27, "learning_rate": 1.908810844115835e-06, "loss": 2.917, "step": 1450 }, { "epoch": 0.27, "learning_rate": 1.907783939207229e-06, "loss": 2.8553, "step": 1455 }, { "epoch": 0.27, "learning_rate": 1.906757034298624e-06, "loss": 2.8264, "step": 1460 }, { "epoch": 0.27, "learning_rate": 1.9057301293900184e-06, "loss": 2.9097, "step": 1465 }, { "epoch": 0.27, "learning_rate": 1.9047032244814129e-06, "loss": 3.0831, "step": 1470 }, { "epoch": 0.27, "learning_rate": 1.9036763195728074e-06, "loss": 2.8505, "step": 1475 }, { "epoch": 0.28, "learning_rate": 1.902649414664202e-06, "loss": 2.9628, "step": 1480 }, { "epoch": 0.28, "learning_rate": 1.9016225097555965e-06, "loss": 3.1449, "step": 1485 }, { "epoch": 0.28, "learning_rate": 1.900595604846991e-06, "loss": 2.8451, "step": 1490 }, { "epoch": 0.28, "learning_rate": 1.8995686999383855e-06, "loss": 2.8117, "step": 1495 }, { "epoch": 0.28, "learning_rate": 1.8985417950297802e-06, "loss": 2.8544, "step": 1500 }, { "epoch": 0.28, "eval_loss": 2.8667774200439453, "eval_runtime": 138.7148, "eval_samples_per_second": 8.608, "eval_steps_per_second": 1.081, "eval_wer": 1.0, "step": 1500 }, { "epoch": 0.28, "learning_rate": 1.8975148901211747e-06, "loss": 2.867, "step": 1505 }, { "epoch": 0.28, "learning_rate": 1.8964879852125692e-06, "loss": 2.8454, "step": 1510 }, { "epoch": 0.28, "learning_rate": 1.8954610803039637e-06, "loss": 2.8422, "step": 1515 }, { "epoch": 0.28, "learning_rate": 1.8944341753953584e-06, "loss": 2.9662, "step": 1520 }, { "epoch": 0.28, "learning_rate": 1.8934072704867528e-06, "loss": 2.9235, "step": 1525 }, { "epoch": 0.28, "learning_rate": 1.8923803655781473e-06, "loss": 2.8577, "step": 1530 }, { "epoch": 0.29, "learning_rate": 1.8913534606695418e-06, "loss": 3.5179, "step": 1535 }, { "epoch": 0.29, "learning_rate": 1.8903265557609365e-06, "loss": 2.8604, "step": 1540 }, { "epoch": 0.29, "learning_rate": 1.889299650852331e-06, "loss": 2.8331, "step": 1545 }, { "epoch": 0.29, "learning_rate": 1.8882727459437255e-06, "loss": 2.8456, "step": 1550 }, { "epoch": 0.29, "learning_rate": 1.88724584103512e-06, "loss": 2.9088, "step": 1555 }, { "epoch": 0.29, "learning_rate": 1.8862189361265147e-06, "loss": 2.9141, "step": 1560 }, { "epoch": 0.29, "learning_rate": 1.8851920312179091e-06, "loss": 2.8593, "step": 1565 }, { "epoch": 0.29, "learning_rate": 1.8841651263093036e-06, "loss": 2.8582, "step": 1570 }, { "epoch": 0.29, "learning_rate": 1.8831382214006981e-06, "loss": 2.8991, "step": 1575 }, { "epoch": 0.29, "learning_rate": 1.8821113164920928e-06, "loss": 2.9633, "step": 1580 }, { "epoch": 0.3, "learning_rate": 1.881084411583487e-06, "loss": 2.8986, "step": 1585 }, { "epoch": 0.3, "learning_rate": 1.8800575066748818e-06, "loss": 2.8375, "step": 1590 }, { "epoch": 0.3, "learning_rate": 1.8790306017662763e-06, "loss": 2.9018, "step": 1595 }, { "epoch": 0.3, "learning_rate": 1.878003696857671e-06, "loss": 2.9049, "step": 1600 }, { "epoch": 0.3, "learning_rate": 1.8769767919490652e-06, "loss": 2.8384, "step": 1605 }, { "epoch": 0.3, "learning_rate": 1.87594988704046e-06, "loss": 2.8415, "step": 1610 }, { "epoch": 0.3, "learning_rate": 1.8749229821318544e-06, "loss": 3.1296, "step": 1615 }, { "epoch": 0.3, "learning_rate": 1.8738960772232491e-06, "loss": 2.9117, "step": 1620 }, { "epoch": 0.3, "learning_rate": 1.8728691723146434e-06, "loss": 2.8328, "step": 1625 }, { "epoch": 0.3, "learning_rate": 1.871842267406038e-06, "loss": 2.8658, "step": 1630 }, { "epoch": 0.3, "learning_rate": 1.8708153624974326e-06, "loss": 2.8803, "step": 1635 }, { "epoch": 0.31, "learning_rate": 1.8697884575888273e-06, "loss": 2.8543, "step": 1640 }, { "epoch": 0.31, "learning_rate": 1.8687615526802215e-06, "loss": 2.8685, "step": 1645 }, { "epoch": 0.31, "learning_rate": 1.8677346477716162e-06, "loss": 2.9056, "step": 1650 }, { "epoch": 0.31, "learning_rate": 1.8667077428630107e-06, "loss": 2.8946, "step": 1655 }, { "epoch": 0.31, "learning_rate": 1.8656808379544054e-06, "loss": 2.9043, "step": 1660 }, { "epoch": 0.31, "learning_rate": 1.8646539330457997e-06, "loss": 3.2464, "step": 1665 }, { "epoch": 0.31, "learning_rate": 1.8636270281371944e-06, "loss": 2.8617, "step": 1670 }, { "epoch": 0.31, "learning_rate": 1.8626001232285889e-06, "loss": 2.8053, "step": 1675 }, { "epoch": 0.31, "learning_rate": 1.8615732183199836e-06, "loss": 2.846, "step": 1680 }, { "epoch": 0.31, "learning_rate": 1.860546313411378e-06, "loss": 2.9291, "step": 1685 }, { "epoch": 0.31, "learning_rate": 1.8595194085027725e-06, "loss": 2.8725, "step": 1690 }, { "epoch": 0.32, "learning_rate": 1.8584925035941672e-06, "loss": 2.86, "step": 1695 }, { "epoch": 0.32, "learning_rate": 1.8574655986855617e-06, "loss": 2.8348, "step": 1700 }, { "epoch": 0.32, "learning_rate": 1.8564386937769562e-06, "loss": 2.899, "step": 1705 }, { "epoch": 0.32, "learning_rate": 1.8554117888683507e-06, "loss": 2.9245, "step": 1710 }, { "epoch": 0.32, "learning_rate": 1.8543848839597454e-06, "loss": 2.8713, "step": 1715 }, { "epoch": 0.32, "learning_rate": 1.8533579790511399e-06, "loss": 2.8944, "step": 1720 }, { "epoch": 0.32, "learning_rate": 1.8523310741425344e-06, "loss": 2.9117, "step": 1725 }, { "epoch": 0.32, "learning_rate": 1.8513041692339288e-06, "loss": 3.192, "step": 1730 }, { "epoch": 0.32, "learning_rate": 1.8502772643253235e-06, "loss": 2.9314, "step": 1735 }, { "epoch": 0.32, "learning_rate": 1.8492503594167178e-06, "loss": 2.8147, "step": 1740 }, { "epoch": 0.33, "learning_rate": 1.8482234545081125e-06, "loss": 2.8892, "step": 1745 }, { "epoch": 0.33, "learning_rate": 1.847196549599507e-06, "loss": 2.8516, "step": 1750 }, { "epoch": 0.33, "learning_rate": 1.8461696446909017e-06, "loss": 2.8361, "step": 1755 }, { "epoch": 0.33, "learning_rate": 1.845142739782296e-06, "loss": 2.8513, "step": 1760 }, { "epoch": 0.33, "learning_rate": 1.8441158348736907e-06, "loss": 2.8812, "step": 1765 }, { "epoch": 0.33, "learning_rate": 1.8430889299650851e-06, "loss": 2.9485, "step": 1770 }, { "epoch": 0.33, "learning_rate": 1.8420620250564798e-06, "loss": 2.8524, "step": 1775 }, { "epoch": 0.33, "learning_rate": 1.8410351201478741e-06, "loss": 3.5713, "step": 1780 }, { "epoch": 0.33, "learning_rate": 1.8400082152392688e-06, "loss": 2.8501, "step": 1785 }, { "epoch": 0.33, "learning_rate": 1.8389813103306633e-06, "loss": 2.8184, "step": 1790 }, { "epoch": 0.33, "learning_rate": 1.837954405422058e-06, "loss": 2.8074, "step": 1795 }, { "epoch": 0.34, "learning_rate": 1.8369275005134523e-06, "loss": 2.8764, "step": 1800 }, { "epoch": 0.34, "learning_rate": 1.835900595604847e-06, "loss": 2.9097, "step": 1805 }, { "epoch": 0.34, "learning_rate": 1.8348736906962415e-06, "loss": 2.8483, "step": 1810 }, { "epoch": 0.34, "learning_rate": 1.8338467857876361e-06, "loss": 2.8487, "step": 1815 }, { "epoch": 0.34, "learning_rate": 1.8328198808790304e-06, "loss": 2.8502, "step": 1820 }, { "epoch": 0.34, "learning_rate": 1.8317929759704251e-06, "loss": 2.838, "step": 1825 }, { "epoch": 0.34, "learning_rate": 1.8307660710618196e-06, "loss": 2.8157, "step": 1830 }, { "epoch": 0.34, "learning_rate": 1.8297391661532143e-06, "loss": 2.8304, "step": 1835 }, { "epoch": 0.34, "learning_rate": 1.8287122612446086e-06, "loss": 2.8154, "step": 1840 }, { "epoch": 0.34, "learning_rate": 1.8276853563360033e-06, "loss": 2.795, "step": 1845 }, { "epoch": 0.34, "learning_rate": 1.8266584514273978e-06, "loss": 2.8491, "step": 1850 }, { "epoch": 0.35, "learning_rate": 1.8256315465187925e-06, "loss": 2.8307, "step": 1855 }, { "epoch": 0.35, "learning_rate": 1.8246046416101867e-06, "loss": 2.8661, "step": 1860 }, { "epoch": 0.35, "learning_rate": 1.8235777367015814e-06, "loss": 2.847, "step": 1865 }, { "epoch": 0.35, "learning_rate": 1.822550831792976e-06, "loss": 2.9903, "step": 1870 }, { "epoch": 0.35, "learning_rate": 1.8215239268843706e-06, "loss": 2.8249, "step": 1875 }, { "epoch": 0.35, "learning_rate": 1.8204970219757649e-06, "loss": 2.8958, "step": 1880 }, { "epoch": 0.35, "learning_rate": 1.8194701170671596e-06, "loss": 2.8271, "step": 1885 }, { "epoch": 0.35, "learning_rate": 1.818443212158554e-06, "loss": 2.7845, "step": 1890 }, { "epoch": 0.35, "learning_rate": 1.8174163072499488e-06, "loss": 2.809, "step": 1895 }, { "epoch": 0.35, "learning_rate": 1.816389402341343e-06, "loss": 2.7484, "step": 1900 }, { "epoch": 0.35, "learning_rate": 1.8153624974327377e-06, "loss": 2.8483, "step": 1905 }, { "epoch": 0.36, "learning_rate": 1.8143355925241322e-06, "loss": 2.793, "step": 1910 }, { "epoch": 0.36, "learning_rate": 1.8133086876155267e-06, "loss": 2.9283, "step": 1915 }, { "epoch": 0.36, "learning_rate": 1.8122817827069212e-06, "loss": 2.8682, "step": 1920 }, { "epoch": 0.36, "learning_rate": 1.8112548777983159e-06, "loss": 3.1691, "step": 1925 }, { "epoch": 0.36, "learning_rate": 1.8102279728897104e-06, "loss": 2.7775, "step": 1930 }, { "epoch": 0.36, "learning_rate": 1.8092010679811048e-06, "loss": 2.7845, "step": 1935 }, { "epoch": 0.36, "learning_rate": 1.8081741630724993e-06, "loss": 2.8187, "step": 1940 }, { "epoch": 0.36, "learning_rate": 1.807147258163894e-06, "loss": 3.2324, "step": 1945 }, { "epoch": 0.36, "learning_rate": 1.8061203532552885e-06, "loss": 2.762, "step": 1950 }, { "epoch": 0.36, "learning_rate": 1.805093448346683e-06, "loss": 2.8946, "step": 1955 }, { "epoch": 0.37, "learning_rate": 1.8040665434380775e-06, "loss": 2.8937, "step": 1960 }, { "epoch": 0.37, "learning_rate": 1.8030396385294722e-06, "loss": 2.8476, "step": 1965 }, { "epoch": 0.37, "learning_rate": 1.8020127336208667e-06, "loss": 2.7975, "step": 1970 }, { "epoch": 0.37, "learning_rate": 1.8009858287122612e-06, "loss": 2.7058, "step": 1975 }, { "epoch": 0.37, "learning_rate": 1.7999589238036556e-06, "loss": 2.7794, "step": 1980 }, { "epoch": 0.37, "learning_rate": 1.7989320188950503e-06, "loss": 2.7986, "step": 1985 }, { "epoch": 0.37, "learning_rate": 1.7979051139864448e-06, "loss": 2.7735, "step": 1990 }, { "epoch": 0.37, "learning_rate": 1.7968782090778393e-06, "loss": 3.2428, "step": 1995 }, { "epoch": 0.37, "learning_rate": 1.7958513041692338e-06, "loss": 2.8012, "step": 2000 }, { "epoch": 0.37, "eval_loss": 2.72654390335083, "eval_runtime": 158.5883, "eval_samples_per_second": 7.529, "eval_steps_per_second": 0.946, "eval_wer": 1.0, "step": 2000 }, { "epoch": 0.37, "learning_rate": 1.7948243992606285e-06, "loss": 2.788, "step": 2005 }, { "epoch": 0.37, "learning_rate": 1.793797494352023e-06, "loss": 2.8242, "step": 2010 }, { "epoch": 0.38, "learning_rate": 1.7927705894434175e-06, "loss": 2.9338, "step": 2015 }, { "epoch": 0.38, "learning_rate": 1.791743684534812e-06, "loss": 2.7148, "step": 2020 }, { "epoch": 0.38, "learning_rate": 1.7907167796262066e-06, "loss": 4.2617, "step": 2025 }, { "epoch": 0.38, "learning_rate": 1.7896898747176011e-06, "loss": 2.6996, "step": 2030 }, { "epoch": 0.38, "learning_rate": 1.7886629698089956e-06, "loss": 2.803, "step": 2035 }, { "epoch": 0.38, "learning_rate": 1.78763606490039e-06, "loss": 2.8994, "step": 2040 }, { "epoch": 0.38, "learning_rate": 1.7866091599917848e-06, "loss": 2.7159, "step": 2045 }, { "epoch": 0.38, "learning_rate": 1.785582255083179e-06, "loss": 2.727, "step": 2050 }, { "epoch": 0.38, "learning_rate": 1.7845553501745738e-06, "loss": 2.748, "step": 2055 }, { "epoch": 0.38, "learning_rate": 1.7835284452659682e-06, "loss": 3.1073, "step": 2060 }, { "epoch": 0.38, "learning_rate": 1.782501540357363e-06, "loss": 2.7317, "step": 2065 }, { "epoch": 0.39, "learning_rate": 1.7814746354487572e-06, "loss": 2.7702, "step": 2070 }, { "epoch": 0.39, "learning_rate": 1.780447730540152e-06, "loss": 2.6403, "step": 2075 }, { "epoch": 0.39, "learning_rate": 1.7794208256315464e-06, "loss": 2.6535, "step": 2080 }, { "epoch": 0.39, "learning_rate": 1.778393920722941e-06, "loss": 2.6274, "step": 2085 }, { "epoch": 0.39, "learning_rate": 1.7773670158143354e-06, "loss": 2.6284, "step": 2090 }, { "epoch": 0.39, "learning_rate": 1.77634011090573e-06, "loss": 2.7832, "step": 2095 }, { "epoch": 0.39, "learning_rate": 1.7753132059971246e-06, "loss": 2.6126, "step": 2100 }, { "epoch": 0.39, "learning_rate": 1.7742863010885192e-06, "loss": 2.6608, "step": 2105 }, { "epoch": 0.39, "learning_rate": 1.7732593961799135e-06, "loss": 2.7854, "step": 2110 }, { "epoch": 0.39, "learning_rate": 1.7722324912713082e-06, "loss": 2.6561, "step": 2115 }, { "epoch": 0.39, "learning_rate": 1.7712055863627027e-06, "loss": 2.5986, "step": 2120 }, { "epoch": 0.4, "learning_rate": 1.7701786814540974e-06, "loss": 2.6772, "step": 2125 }, { "epoch": 0.4, "learning_rate": 1.7691517765454917e-06, "loss": 2.5899, "step": 2130 }, { "epoch": 0.4, "learning_rate": 1.7681248716368864e-06, "loss": 2.6522, "step": 2135 }, { "epoch": 0.4, "learning_rate": 1.7670979667282809e-06, "loss": 2.7009, "step": 2140 }, { "epoch": 0.4, "learning_rate": 1.7660710618196756e-06, "loss": 2.5363, "step": 2145 }, { "epoch": 0.4, "learning_rate": 1.7650441569110698e-06, "loss": 2.5242, "step": 2150 }, { "epoch": 0.4, "learning_rate": 1.7640172520024645e-06, "loss": 2.535, "step": 2155 }, { "epoch": 0.4, "learning_rate": 1.762990347093859e-06, "loss": 2.6498, "step": 2160 }, { "epoch": 0.4, "learning_rate": 1.7619634421852537e-06, "loss": 2.7191, "step": 2165 }, { "epoch": 0.4, "learning_rate": 1.760936537276648e-06, "loss": 2.5465, "step": 2170 }, { "epoch": 0.41, "learning_rate": 1.7599096323680427e-06, "loss": 2.5029, "step": 2175 }, { "epoch": 0.41, "learning_rate": 1.7588827274594372e-06, "loss": 3.0624, "step": 2180 }, { "epoch": 0.41, "learning_rate": 1.7578558225508319e-06, "loss": 2.5327, "step": 2185 }, { "epoch": 0.41, "learning_rate": 1.7568289176422261e-06, "loss": 2.5604, "step": 2190 }, { "epoch": 0.41, "learning_rate": 1.7558020127336208e-06, "loss": 2.6551, "step": 2195 }, { "epoch": 0.41, "learning_rate": 1.7547751078250153e-06, "loss": 2.7544, "step": 2200 }, { "epoch": 0.41, "learning_rate": 1.75374820291641e-06, "loss": 2.8209, "step": 2205 }, { "epoch": 0.41, "learning_rate": 1.7527212980078043e-06, "loss": 2.7117, "step": 2210 }, { "epoch": 0.41, "learning_rate": 1.751694393099199e-06, "loss": 2.563, "step": 2215 }, { "epoch": 0.41, "learning_rate": 1.7506674881905935e-06, "loss": 2.585, "step": 2220 }, { "epoch": 0.41, "learning_rate": 1.749640583281988e-06, "loss": 2.6235, "step": 2225 }, { "epoch": 0.42, "learning_rate": 1.7486136783733824e-06, "loss": 2.5745, "step": 2230 }, { "epoch": 0.42, "learning_rate": 1.7475867734647771e-06, "loss": 2.5579, "step": 2235 }, { "epoch": 0.42, "learning_rate": 1.7465598685561716e-06, "loss": 2.6564, "step": 2240 }, { "epoch": 0.42, "learning_rate": 1.745532963647566e-06, "loss": 2.6688, "step": 2245 }, { "epoch": 0.42, "learning_rate": 1.7445060587389606e-06, "loss": 2.8039, "step": 2250 }, { "epoch": 0.42, "learning_rate": 1.7434791538303553e-06, "loss": 2.6522, "step": 2255 }, { "epoch": 0.42, "learning_rate": 1.7424522489217498e-06, "loss": 3.1135, "step": 2260 }, { "epoch": 0.42, "learning_rate": 1.7414253440131443e-06, "loss": 2.4431, "step": 2265 }, { "epoch": 0.42, "learning_rate": 1.7403984391045387e-06, "loss": 2.4361, "step": 2270 }, { "epoch": 0.42, "learning_rate": 1.7393715341959334e-06, "loss": 2.4744, "step": 2275 }, { "epoch": 0.42, "learning_rate": 1.738344629287328e-06, "loss": 2.4935, "step": 2280 }, { "epoch": 0.43, "learning_rate": 1.7373177243787224e-06, "loss": 2.5459, "step": 2285 }, { "epoch": 0.43, "learning_rate": 1.7362908194701169e-06, "loss": 3.1449, "step": 2290 }, { "epoch": 0.43, "learning_rate": 1.7352639145615116e-06, "loss": 2.9531, "step": 2295 }, { "epoch": 0.43, "learning_rate": 1.734237009652906e-06, "loss": 2.7413, "step": 2300 }, { "epoch": 0.43, "learning_rate": 1.7332101047443006e-06, "loss": 2.5559, "step": 2305 }, { "epoch": 0.43, "learning_rate": 1.732183199835695e-06, "loss": 2.5224, "step": 2310 }, { "epoch": 0.43, "learning_rate": 1.7311562949270897e-06, "loss": 2.355, "step": 2315 }, { "epoch": 0.43, "learning_rate": 1.7301293900184842e-06, "loss": 2.3918, "step": 2320 }, { "epoch": 0.43, "learning_rate": 1.7291024851098787e-06, "loss": 2.5834, "step": 2325 }, { "epoch": 0.43, "learning_rate": 1.7280755802012732e-06, "loss": 2.3767, "step": 2330 }, { "epoch": 0.43, "learning_rate": 1.7270486752926679e-06, "loss": 2.5311, "step": 2335 }, { "epoch": 0.44, "learning_rate": 1.7260217703840624e-06, "loss": 2.3856, "step": 2340 }, { "epoch": 0.44, "learning_rate": 1.7249948654754569e-06, "loss": 3.6706, "step": 2345 }, { "epoch": 0.44, "learning_rate": 1.7239679605668513e-06, "loss": 2.2921, "step": 2350 }, { "epoch": 0.44, "learning_rate": 1.722941055658246e-06, "loss": 2.6198, "step": 2355 }, { "epoch": 0.44, "learning_rate": 1.7219141507496403e-06, "loss": 2.5341, "step": 2360 }, { "epoch": 0.44, "learning_rate": 1.720887245841035e-06, "loss": 2.5211, "step": 2365 }, { "epoch": 0.44, "learning_rate": 1.7198603409324295e-06, "loss": 2.3758, "step": 2370 }, { "epoch": 0.44, "learning_rate": 1.7188334360238242e-06, "loss": 2.3367, "step": 2375 }, { "epoch": 0.44, "learning_rate": 1.7178065311152185e-06, "loss": 2.4326, "step": 2380 }, { "epoch": 0.44, "learning_rate": 1.7167796262066132e-06, "loss": 2.2972, "step": 2385 }, { "epoch": 0.45, "learning_rate": 1.7157527212980076e-06, "loss": 2.4532, "step": 2390 }, { "epoch": 0.45, "learning_rate": 1.7147258163894023e-06, "loss": 2.3251, "step": 2395 }, { "epoch": 0.45, "learning_rate": 1.7136989114807966e-06, "loss": 2.2711, "step": 2400 }, { "epoch": 0.45, "learning_rate": 1.7126720065721913e-06, "loss": 2.4129, "step": 2405 }, { "epoch": 0.45, "learning_rate": 1.7116451016635858e-06, "loss": 2.4588, "step": 2410 }, { "epoch": 0.45, "learning_rate": 1.7106181967549805e-06, "loss": 2.5394, "step": 2415 }, { "epoch": 0.45, "learning_rate": 1.7095912918463748e-06, "loss": 2.3018, "step": 2420 }, { "epoch": 0.45, "learning_rate": 1.7085643869377695e-06, "loss": 2.5115, "step": 2425 }, { "epoch": 0.45, "learning_rate": 1.707537482029164e-06, "loss": 2.259, "step": 2430 }, { "epoch": 0.45, "learning_rate": 1.7065105771205587e-06, "loss": 2.3676, "step": 2435 }, { "epoch": 0.45, "learning_rate": 1.705483672211953e-06, "loss": 2.6998, "step": 2440 }, { "epoch": 0.46, "learning_rate": 1.7044567673033476e-06, "loss": 2.4527, "step": 2445 }, { "epoch": 0.46, "learning_rate": 1.703429862394742e-06, "loss": 2.3832, "step": 2450 }, { "epoch": 0.46, "learning_rate": 1.7024029574861368e-06, "loss": 2.3461, "step": 2455 }, { "epoch": 0.46, "learning_rate": 1.701376052577531e-06, "loss": 2.2367, "step": 2460 }, { "epoch": 0.46, "learning_rate": 1.7003491476689258e-06, "loss": 2.2527, "step": 2465 }, { "epoch": 0.46, "learning_rate": 1.6993222427603203e-06, "loss": 2.2838, "step": 2470 }, { "epoch": 0.46, "learning_rate": 1.698295337851715e-06, "loss": 2.5507, "step": 2475 }, { "epoch": 0.46, "learning_rate": 1.6972684329431092e-06, "loss": 2.3646, "step": 2480 }, { "epoch": 0.46, "learning_rate": 1.696241528034504e-06, "loss": 2.248, "step": 2485 }, { "epoch": 0.46, "learning_rate": 1.6952146231258984e-06, "loss": 2.1861, "step": 2490 }, { "epoch": 0.46, "learning_rate": 1.6941877182172931e-06, "loss": 2.2259, "step": 2495 }, { "epoch": 0.47, "learning_rate": 1.6931608133086874e-06, "loss": 2.3309, "step": 2500 }, { "epoch": 0.47, "eval_loss": 2.0526249408721924, "eval_runtime": 162.7261, "eval_samples_per_second": 7.337, "eval_steps_per_second": 0.922, "eval_wer": 1.0, "step": 2500 }, { "epoch": 0.47, "learning_rate": 1.692133908400082e-06, "loss": 2.2051, "step": 2505 }, { "epoch": 0.47, "learning_rate": 1.6911070034914766e-06, "loss": 2.1754, "step": 2510 }, { "epoch": 0.47, "learning_rate": 1.690080098582871e-06, "loss": 2.2147, "step": 2515 }, { "epoch": 0.47, "learning_rate": 1.6890531936742657e-06, "loss": 2.0856, "step": 2520 }, { "epoch": 0.47, "learning_rate": 1.6880262887656602e-06, "loss": 2.1742, "step": 2525 }, { "epoch": 0.47, "learning_rate": 1.686999383857055e-06, "loss": 2.2649, "step": 2530 }, { "epoch": 0.47, "learning_rate": 1.6859724789484492e-06, "loss": 2.5063, "step": 2535 }, { "epoch": 0.47, "learning_rate": 1.684945574039844e-06, "loss": 2.2344, "step": 2540 }, { "epoch": 0.47, "learning_rate": 1.6839186691312384e-06, "loss": 2.4601, "step": 2545 }, { "epoch": 0.47, "learning_rate": 1.682891764222633e-06, "loss": 2.2304, "step": 2550 }, { "epoch": 0.48, "learning_rate": 1.6818648593140274e-06, "loss": 2.0962, "step": 2555 }, { "epoch": 0.48, "learning_rate": 1.680837954405422e-06, "loss": 2.1322, "step": 2560 }, { "epoch": 0.48, "learning_rate": 1.6798110494968165e-06, "loss": 2.1372, "step": 2565 }, { "epoch": 0.48, "learning_rate": 1.6787841445882112e-06, "loss": 2.1914, "step": 2570 }, { "epoch": 0.48, "learning_rate": 1.6777572396796055e-06, "loss": 2.321, "step": 2575 }, { "epoch": 0.48, "learning_rate": 1.6767303347710002e-06, "loss": 2.1254, "step": 2580 }, { "epoch": 0.48, "learning_rate": 1.6757034298623947e-06, "loss": 2.2112, "step": 2585 }, { "epoch": 0.48, "learning_rate": 1.6746765249537894e-06, "loss": 2.3469, "step": 2590 }, { "epoch": 0.48, "learning_rate": 1.6736496200451837e-06, "loss": 2.8565, "step": 2595 }, { "epoch": 0.48, "learning_rate": 1.6726227151365784e-06, "loss": 2.4071, "step": 2600 }, { "epoch": 0.49, "learning_rate": 1.6715958102279728e-06, "loss": 2.4843, "step": 2605 }, { "epoch": 0.49, "learning_rate": 1.6705689053193675e-06, "loss": 2.2168, "step": 2610 }, { "epoch": 0.49, "learning_rate": 1.6695420004107618e-06, "loss": 2.5841, "step": 2615 }, { "epoch": 0.49, "learning_rate": 1.6685150955021565e-06, "loss": 2.2403, "step": 2620 }, { "epoch": 0.49, "learning_rate": 1.667488190593551e-06, "loss": 2.0698, "step": 2625 }, { "epoch": 0.49, "learning_rate": 1.6664612856849457e-06, "loss": 2.3543, "step": 2630 }, { "epoch": 0.49, "learning_rate": 1.66543438077634e-06, "loss": 2.3407, "step": 2635 }, { "epoch": 0.49, "learning_rate": 1.6644074758677347e-06, "loss": 2.4341, "step": 2640 }, { "epoch": 0.49, "learning_rate": 1.6633805709591291e-06, "loss": 2.2227, "step": 2645 }, { "epoch": 0.49, "learning_rate": 1.6623536660505238e-06, "loss": 2.1824, "step": 2650 }, { "epoch": 0.49, "learning_rate": 1.6613267611419181e-06, "loss": 2.0214, "step": 2655 }, { "epoch": 0.5, "learning_rate": 1.6602998562333128e-06, "loss": 2.0735, "step": 2660 }, { "epoch": 0.5, "learning_rate": 1.6592729513247073e-06, "loss": 2.5132, "step": 2665 }, { "epoch": 0.5, "learning_rate": 1.658246046416102e-06, "loss": 2.3907, "step": 2670 }, { "epoch": 0.5, "learning_rate": 1.6572191415074963e-06, "loss": 2.051, "step": 2675 }, { "epoch": 0.5, "learning_rate": 1.656192236598891e-06, "loss": 2.2869, "step": 2680 }, { "epoch": 0.5, "learning_rate": 1.6551653316902854e-06, "loss": 2.0994, "step": 2685 }, { "epoch": 0.5, "learning_rate": 1.65413842678168e-06, "loss": 2.0096, "step": 2690 }, { "epoch": 0.5, "learning_rate": 1.6531115218730744e-06, "loss": 2.0538, "step": 2695 }, { "epoch": 0.5, "learning_rate": 1.6520846169644691e-06, "loss": 2.2783, "step": 2700 }, { "epoch": 0.5, "learning_rate": 1.6510577120558636e-06, "loss": 2.2621, "step": 2705 }, { "epoch": 0.5, "learning_rate": 1.650030807147258e-06, "loss": 2.152, "step": 2710 }, { "epoch": 0.51, "learning_rate": 1.6490039022386526e-06, "loss": 2.1277, "step": 2715 }, { "epoch": 0.51, "learning_rate": 1.6479769973300473e-06, "loss": 1.9992, "step": 2720 }, { "epoch": 0.51, "learning_rate": 1.6469500924214417e-06, "loss": 2.0336, "step": 2725 }, { "epoch": 0.51, "learning_rate": 1.6459231875128362e-06, "loss": 2.1393, "step": 2730 }, { "epoch": 0.51, "learning_rate": 1.6448962826042307e-06, "loss": 1.9702, "step": 2735 }, { "epoch": 0.51, "learning_rate": 1.6438693776956254e-06, "loss": 2.0614, "step": 2740 }, { "epoch": 0.51, "learning_rate": 1.64284247278702e-06, "loss": 2.0463, "step": 2745 }, { "epoch": 0.51, "learning_rate": 1.6418155678784144e-06, "loss": 2.027, "step": 2750 }, { "epoch": 0.51, "learning_rate": 1.6407886629698089e-06, "loss": 1.9492, "step": 2755 }, { "epoch": 0.51, "learning_rate": 1.6397617580612036e-06, "loss": 2.1043, "step": 2760 }, { "epoch": 0.51, "learning_rate": 1.638734853152598e-06, "loss": 2.0343, "step": 2765 }, { "epoch": 0.52, "learning_rate": 1.6377079482439925e-06, "loss": 2.1734, "step": 2770 }, { "epoch": 0.52, "learning_rate": 1.636681043335387e-06, "loss": 2.2739, "step": 2775 }, { "epoch": 0.52, "learning_rate": 1.6356541384267817e-06, "loss": 1.9545, "step": 2780 }, { "epoch": 0.52, "learning_rate": 1.6346272335181762e-06, "loss": 2.1237, "step": 2785 }, { "epoch": 0.52, "learning_rate": 1.6336003286095707e-06, "loss": 1.9395, "step": 2790 }, { "epoch": 0.52, "learning_rate": 1.6325734237009652e-06, "loss": 2.1862, "step": 2795 }, { "epoch": 0.52, "learning_rate": 1.6315465187923599e-06, "loss": 2.2151, "step": 2800 }, { "epoch": 0.52, "learning_rate": 1.6305196138837544e-06, "loss": 1.9663, "step": 2805 }, { "epoch": 0.52, "learning_rate": 1.6294927089751488e-06, "loss": 1.9221, "step": 2810 }, { "epoch": 0.52, "learning_rate": 1.6284658040665433e-06, "loss": 2.4582, "step": 2815 }, { "epoch": 0.53, "learning_rate": 1.627438899157938e-06, "loss": 2.0649, "step": 2820 }, { "epoch": 0.53, "learning_rate": 1.6264119942493323e-06, "loss": 2.2205, "step": 2825 }, { "epoch": 0.53, "learning_rate": 1.625385089340727e-06, "loss": 2.0404, "step": 2830 }, { "epoch": 0.53, "learning_rate": 1.6243581844321215e-06, "loss": 2.0706, "step": 2835 }, { "epoch": 0.53, "learning_rate": 1.6233312795235162e-06, "loss": 1.9393, "step": 2840 }, { "epoch": 0.53, "learning_rate": 1.6223043746149104e-06, "loss": 1.8826, "step": 2845 }, { "epoch": 0.53, "learning_rate": 1.6212774697063051e-06, "loss": 2.0004, "step": 2850 }, { "epoch": 0.53, "learning_rate": 1.6202505647976996e-06, "loss": 1.7917, "step": 2855 }, { "epoch": 0.53, "learning_rate": 1.6192236598890943e-06, "loss": 2.2027, "step": 2860 }, { "epoch": 0.53, "learning_rate": 1.6181967549804886e-06, "loss": 1.814, "step": 2865 }, { "epoch": 0.53, "learning_rate": 1.6171698500718833e-06, "loss": 2.1004, "step": 2870 }, { "epoch": 0.54, "learning_rate": 1.6161429451632778e-06, "loss": 2.0044, "step": 2875 }, { "epoch": 0.54, "learning_rate": 1.6151160402546725e-06, "loss": 1.9016, "step": 2880 }, { "epoch": 0.54, "learning_rate": 1.6140891353460668e-06, "loss": 1.8864, "step": 2885 }, { "epoch": 0.54, "learning_rate": 1.6130622304374615e-06, "loss": 1.9027, "step": 2890 }, { "epoch": 0.54, "learning_rate": 1.612035325528856e-06, "loss": 2.0984, "step": 2895 }, { "epoch": 0.54, "learning_rate": 1.6110084206202506e-06, "loss": 1.8449, "step": 2900 }, { "epoch": 0.54, "learning_rate": 1.609981515711645e-06, "loss": 1.9141, "step": 2905 }, { "epoch": 0.54, "learning_rate": 1.6089546108030396e-06, "loss": 1.8873, "step": 2910 }, { "epoch": 0.54, "learning_rate": 1.607927705894434e-06, "loss": 2.0334, "step": 2915 }, { "epoch": 0.54, "learning_rate": 1.6069008009858288e-06, "loss": 2.0702, "step": 2920 }, { "epoch": 0.54, "learning_rate": 1.605873896077223e-06, "loss": 2.1452, "step": 2925 }, { "epoch": 0.55, "learning_rate": 1.6048469911686178e-06, "loss": 1.7758, "step": 2930 }, { "epoch": 0.55, "learning_rate": 1.6038200862600122e-06, "loss": 2.2563, "step": 2935 }, { "epoch": 0.55, "learning_rate": 1.602793181351407e-06, "loss": 1.8165, "step": 2940 }, { "epoch": 0.55, "learning_rate": 1.6017662764428012e-06, "loss": 1.8777, "step": 2945 }, { "epoch": 0.55, "learning_rate": 1.600739371534196e-06, "loss": 1.869, "step": 2950 }, { "epoch": 0.55, "learning_rate": 1.5997124666255904e-06, "loss": 2.1495, "step": 2955 }, { "epoch": 0.55, "learning_rate": 1.598685561716985e-06, "loss": 1.7904, "step": 2960 }, { "epoch": 0.55, "learning_rate": 1.5976586568083794e-06, "loss": 2.1359, "step": 2965 }, { "epoch": 0.55, "learning_rate": 1.596631751899774e-06, "loss": 1.73, "step": 2970 }, { "epoch": 0.55, "learning_rate": 1.5956048469911685e-06, "loss": 1.7582, "step": 2975 }, { "epoch": 0.56, "learning_rate": 1.594577942082563e-06, "loss": 1.8494, "step": 2980 }, { "epoch": 0.56, "learning_rate": 1.5935510371739575e-06, "loss": 1.8766, "step": 2985 }, { "epoch": 0.56, "learning_rate": 1.5925241322653522e-06, "loss": 1.8829, "step": 2990 }, { "epoch": 0.56, "learning_rate": 1.5914972273567467e-06, "loss": 1.6666, "step": 2995 }, { "epoch": 0.56, "learning_rate": 1.5904703224481412e-06, "loss": 1.993, "step": 3000 }, { "epoch": 0.56, "eval_loss": 1.505236029624939, "eval_runtime": 167.3686, "eval_samples_per_second": 7.134, "eval_steps_per_second": 0.896, "eval_wer": 1.0, "step": 3000 }, { "epoch": 0.56, "learning_rate": 1.5894434175395357e-06, "loss": 1.6019, "step": 3005 }, { "epoch": 0.56, "learning_rate": 1.5884165126309304e-06, "loss": 2.1837, "step": 3010 }, { "epoch": 0.56, "learning_rate": 1.5873896077223248e-06, "loss": 1.8788, "step": 3015 }, { "epoch": 0.56, "learning_rate": 1.5863627028137193e-06, "loss": 2.2183, "step": 3020 }, { "epoch": 0.56, "learning_rate": 1.5853357979051138e-06, "loss": 1.7065, "step": 3025 }, { "epoch": 0.56, "learning_rate": 1.5843088929965085e-06, "loss": 1.6309, "step": 3030 }, { "epoch": 0.57, "learning_rate": 1.583281988087903e-06, "loss": 1.9461, "step": 3035 }, { "epoch": 0.57, "learning_rate": 1.5822550831792975e-06, "loss": 1.8182, "step": 3040 }, { "epoch": 0.57, "learning_rate": 1.581228178270692e-06, "loss": 1.8182, "step": 3045 }, { "epoch": 0.57, "learning_rate": 1.5802012733620867e-06, "loss": 1.95, "step": 3050 }, { "epoch": 0.57, "learning_rate": 1.5791743684534812e-06, "loss": 1.6864, "step": 3055 }, { "epoch": 0.57, "learning_rate": 1.5781474635448756e-06, "loss": 1.7599, "step": 3060 }, { "epoch": 0.57, "learning_rate": 1.5771205586362701e-06, "loss": 1.9663, "step": 3065 }, { "epoch": 0.57, "learning_rate": 1.5760936537276648e-06, "loss": 1.8025, "step": 3070 }, { "epoch": 0.57, "learning_rate": 1.5750667488190593e-06, "loss": 2.1133, "step": 3075 }, { "epoch": 0.57, "learning_rate": 1.5740398439104538e-06, "loss": 1.928, "step": 3080 }, { "epoch": 0.57, "learning_rate": 1.5730129390018483e-06, "loss": 1.6224, "step": 3085 }, { "epoch": 0.58, "learning_rate": 1.571986034093243e-06, "loss": 1.6727, "step": 3090 }, { "epoch": 0.58, "learning_rate": 1.5709591291846375e-06, "loss": 1.7127, "step": 3095 }, { "epoch": 0.58, "learning_rate": 1.569932224276032e-06, "loss": 1.6791, "step": 3100 }, { "epoch": 0.58, "learning_rate": 1.5689053193674264e-06, "loss": 1.9341, "step": 3105 }, { "epoch": 0.58, "learning_rate": 1.5678784144588211e-06, "loss": 1.9801, "step": 3110 }, { "epoch": 0.58, "learning_rate": 1.5668515095502154e-06, "loss": 2.084, "step": 3115 }, { "epoch": 0.58, "learning_rate": 1.56582460464161e-06, "loss": 1.7563, "step": 3120 }, { "epoch": 0.58, "learning_rate": 1.5647976997330046e-06, "loss": 1.6595, "step": 3125 }, { "epoch": 0.58, "learning_rate": 1.5637707948243993e-06, "loss": 1.5875, "step": 3130 }, { "epoch": 0.58, "learning_rate": 1.5627438899157935e-06, "loss": 2.1124, "step": 3135 }, { "epoch": 0.58, "learning_rate": 1.5617169850071882e-06, "loss": 1.8638, "step": 3140 }, { "epoch": 0.59, "learning_rate": 1.5606900800985827e-06, "loss": 1.6619, "step": 3145 }, { "epoch": 0.59, "learning_rate": 1.5596631751899774e-06, "loss": 1.7096, "step": 3150 }, { "epoch": 0.59, "learning_rate": 1.5586362702813717e-06, "loss": 1.796, "step": 3155 }, { "epoch": 0.59, "learning_rate": 1.5576093653727664e-06, "loss": 1.6431, "step": 3160 }, { "epoch": 0.59, "learning_rate": 1.5565824604641609e-06, "loss": 1.6813, "step": 3165 }, { "epoch": 0.59, "learning_rate": 1.5555555555555556e-06, "loss": 1.6838, "step": 3170 }, { "epoch": 0.59, "learning_rate": 1.5545286506469499e-06, "loss": 1.7729, "step": 3175 }, { "epoch": 0.59, "learning_rate": 1.5535017457383445e-06, "loss": 1.8684, "step": 3180 }, { "epoch": 0.59, "learning_rate": 1.552474840829739e-06, "loss": 1.6152, "step": 3185 }, { "epoch": 0.59, "learning_rate": 1.5514479359211337e-06, "loss": 1.8852, "step": 3190 }, { "epoch": 0.6, "learning_rate": 1.550421031012528e-06, "loss": 1.638, "step": 3195 }, { "epoch": 0.6, "learning_rate": 1.5493941261039227e-06, "loss": 1.7184, "step": 3200 }, { "epoch": 0.6, "learning_rate": 1.5483672211953172e-06, "loss": 1.8423, "step": 3205 }, { "epoch": 0.6, "learning_rate": 1.5473403162867119e-06, "loss": 1.996, "step": 3210 }, { "epoch": 0.6, "learning_rate": 1.5463134113781062e-06, "loss": 1.6101, "step": 3215 }, { "epoch": 0.6, "learning_rate": 1.5452865064695009e-06, "loss": 1.6341, "step": 3220 }, { "epoch": 0.6, "learning_rate": 1.5442596015608953e-06, "loss": 1.7689, "step": 3225 }, { "epoch": 0.6, "learning_rate": 1.54323269665229e-06, "loss": 1.5537, "step": 3230 }, { "epoch": 0.6, "learning_rate": 1.5422057917436843e-06, "loss": 1.6625, "step": 3235 }, { "epoch": 0.6, "learning_rate": 1.541178886835079e-06, "loss": 2.0027, "step": 3240 }, { "epoch": 0.6, "learning_rate": 1.5401519819264735e-06, "loss": 1.5174, "step": 3245 }, { "epoch": 0.61, "learning_rate": 1.5391250770178682e-06, "loss": 1.7209, "step": 3250 }, { "epoch": 0.61, "learning_rate": 1.5380981721092625e-06, "loss": 1.6119, "step": 3255 }, { "epoch": 0.61, "learning_rate": 1.5370712672006572e-06, "loss": 1.5694, "step": 3260 }, { "epoch": 0.61, "learning_rate": 1.5360443622920516e-06, "loss": 1.8602, "step": 3265 }, { "epoch": 0.61, "learning_rate": 1.5350174573834463e-06, "loss": 1.5486, "step": 3270 }, { "epoch": 0.61, "learning_rate": 1.5339905524748406e-06, "loss": 2.5045, "step": 3275 }, { "epoch": 0.61, "learning_rate": 1.5329636475662353e-06, "loss": 1.6657, "step": 3280 }, { "epoch": 0.61, "learning_rate": 1.5319367426576298e-06, "loss": 1.8253, "step": 3285 }, { "epoch": 0.61, "learning_rate": 1.5309098377490243e-06, "loss": 1.7489, "step": 3290 }, { "epoch": 0.61, "learning_rate": 1.5298829328404188e-06, "loss": 2.0765, "step": 3295 }, { "epoch": 0.61, "learning_rate": 1.5288560279318135e-06, "loss": 1.7706, "step": 3300 }, { "epoch": 0.62, "learning_rate": 1.527829123023208e-06, "loss": 1.6185, "step": 3305 }, { "epoch": 0.62, "learning_rate": 1.5268022181146024e-06, "loss": 1.4862, "step": 3310 }, { "epoch": 0.62, "learning_rate": 1.525775313205997e-06, "loss": 1.4534, "step": 3315 }, { "epoch": 0.62, "learning_rate": 1.5247484082973916e-06, "loss": 1.7926, "step": 3320 }, { "epoch": 0.62, "learning_rate": 1.523721503388786e-06, "loss": 1.9686, "step": 3325 }, { "epoch": 0.62, "learning_rate": 1.5226945984801806e-06, "loss": 1.7371, "step": 3330 }, { "epoch": 0.62, "learning_rate": 1.521667693571575e-06, "loss": 1.8109, "step": 3335 }, { "epoch": 0.62, "learning_rate": 1.5206407886629698e-06, "loss": 1.835, "step": 3340 }, { "epoch": 0.62, "learning_rate": 1.5196138837543643e-06, "loss": 1.4894, "step": 3345 }, { "epoch": 0.62, "learning_rate": 1.5185869788457587e-06, "loss": 1.8841, "step": 3350 }, { "epoch": 0.62, "learning_rate": 1.5175600739371532e-06, "loss": 1.3907, "step": 3355 }, { "epoch": 0.63, "learning_rate": 1.516533169028548e-06, "loss": 1.7033, "step": 3360 }, { "epoch": 0.63, "learning_rate": 1.5155062641199426e-06, "loss": 1.5871, "step": 3365 }, { "epoch": 0.63, "learning_rate": 1.5144793592113369e-06, "loss": 1.7982, "step": 3370 }, { "epoch": 0.63, "learning_rate": 1.5134524543027316e-06, "loss": 1.4018, "step": 3375 }, { "epoch": 0.63, "learning_rate": 1.512425549394126e-06, "loss": 1.8385, "step": 3380 }, { "epoch": 0.63, "learning_rate": 1.5113986444855208e-06, "loss": 1.7863, "step": 3385 }, { "epoch": 0.63, "learning_rate": 1.510371739576915e-06, "loss": 1.7658, "step": 3390 }, { "epoch": 0.63, "learning_rate": 1.5093448346683097e-06, "loss": 1.9278, "step": 3395 }, { "epoch": 0.63, "learning_rate": 1.5083179297597042e-06, "loss": 1.523, "step": 3400 }, { "epoch": 0.63, "learning_rate": 1.507291024851099e-06, "loss": 3.4477, "step": 3405 }, { "epoch": 0.64, "learning_rate": 1.5062641199424932e-06, "loss": 2.0776, "step": 3410 }, { "epoch": 0.64, "learning_rate": 1.5052372150338879e-06, "loss": 1.6868, "step": 3415 }, { "epoch": 0.64, "learning_rate": 1.5042103101252824e-06, "loss": 1.3986, "step": 3420 }, { "epoch": 0.64, "learning_rate": 1.503183405216677e-06, "loss": 1.5797, "step": 3425 }, { "epoch": 0.64, "learning_rate": 1.5021565003080713e-06, "loss": 1.7646, "step": 3430 }, { "epoch": 0.64, "learning_rate": 1.501129595399466e-06, "loss": 1.761, "step": 3435 }, { "epoch": 0.64, "learning_rate": 1.5001026904908605e-06, "loss": 1.4129, "step": 3440 }, { "epoch": 0.64, "learning_rate": 1.499075785582255e-06, "loss": 1.9737, "step": 3445 }, { "epoch": 0.64, "learning_rate": 1.4980488806736495e-06, "loss": 1.8094, "step": 3450 }, { "epoch": 0.64, "learning_rate": 1.4970219757650442e-06, "loss": 1.4816, "step": 3455 }, { "epoch": 0.64, "learning_rate": 1.4959950708564387e-06, "loss": 1.492, "step": 3460 }, { "epoch": 0.65, "learning_rate": 1.4949681659478332e-06, "loss": 1.5756, "step": 3465 }, { "epoch": 0.65, "learning_rate": 1.4939412610392276e-06, "loss": 1.6382, "step": 3470 }, { "epoch": 0.65, "learning_rate": 1.4929143561306223e-06, "loss": 1.5102, "step": 3475 }, { "epoch": 0.65, "learning_rate": 1.4918874512220168e-06, "loss": 1.4742, "step": 3480 }, { "epoch": 0.65, "learning_rate": 1.4908605463134113e-06, "loss": 1.8775, "step": 3485 }, { "epoch": 0.65, "learning_rate": 1.4898336414048058e-06, "loss": 1.4767, "step": 3490 }, { "epoch": 0.65, "learning_rate": 1.4888067364962005e-06, "loss": 1.3756, "step": 3495 }, { "epoch": 0.65, "learning_rate": 1.487779831587595e-06, "loss": 1.374, "step": 3500 }, { "epoch": 0.65, "eval_loss": 1.0981903076171875, "eval_runtime": 166.6258, "eval_samples_per_second": 7.166, "eval_steps_per_second": 0.9, "eval_wer": 0.9991021036428938, "step": 3500 }, { "epoch": 0.65, "learning_rate": 1.4867529266789895e-06, "loss": 1.3723, "step": 3505 }, { "epoch": 0.65, "learning_rate": 1.485726021770384e-06, "loss": 1.4567, "step": 3510 }, { "epoch": 0.65, "learning_rate": 1.4846991168617786e-06, "loss": 1.7109, "step": 3515 }, { "epoch": 0.66, "learning_rate": 1.4836722119531731e-06, "loss": 1.6903, "step": 3520 }, { "epoch": 0.66, "learning_rate": 1.4826453070445676e-06, "loss": 1.6524, "step": 3525 }, { "epoch": 0.66, "learning_rate": 1.481618402135962e-06, "loss": 1.6053, "step": 3530 }, { "epoch": 0.66, "learning_rate": 1.4805914972273568e-06, "loss": 1.842, "step": 3535 }, { "epoch": 0.66, "learning_rate": 1.4795645923187513e-06, "loss": 1.5981, "step": 3540 }, { "epoch": 0.66, "learning_rate": 1.4785376874101458e-06, "loss": 1.6901, "step": 3545 }, { "epoch": 0.66, "learning_rate": 1.4775107825015403e-06, "loss": 1.3535, "step": 3550 }, { "epoch": 0.66, "learning_rate": 1.476483877592935e-06, "loss": 1.3337, "step": 3555 }, { "epoch": 0.66, "learning_rate": 1.4754569726843294e-06, "loss": 1.9106, "step": 3560 }, { "epoch": 0.66, "learning_rate": 1.474430067775724e-06, "loss": 1.3896, "step": 3565 }, { "epoch": 0.66, "learning_rate": 1.4734031628671184e-06, "loss": 1.5954, "step": 3570 }, { "epoch": 0.67, "learning_rate": 1.472376257958513e-06, "loss": 1.5729, "step": 3575 }, { "epoch": 0.67, "learning_rate": 1.4713493530499074e-06, "loss": 1.4061, "step": 3580 }, { "epoch": 0.67, "learning_rate": 1.470322448141302e-06, "loss": 1.3232, "step": 3585 }, { "epoch": 0.67, "learning_rate": 1.4692955432326966e-06, "loss": 1.4051, "step": 3590 }, { "epoch": 0.67, "learning_rate": 1.4682686383240913e-06, "loss": 1.3958, "step": 3595 }, { "epoch": 0.67, "learning_rate": 1.4672417334154855e-06, "loss": 2.2725, "step": 3600 }, { "epoch": 0.67, "learning_rate": 1.4662148285068802e-06, "loss": 1.627, "step": 3605 }, { "epoch": 0.67, "learning_rate": 1.4651879235982747e-06, "loss": 1.7266, "step": 3610 }, { "epoch": 0.67, "learning_rate": 1.4641610186896694e-06, "loss": 1.4118, "step": 3615 }, { "epoch": 0.67, "learning_rate": 1.4631341137810637e-06, "loss": 1.6395, "step": 3620 }, { "epoch": 0.68, "learning_rate": 1.4621072088724584e-06, "loss": 1.5691, "step": 3625 }, { "epoch": 0.68, "learning_rate": 1.4610803039638529e-06, "loss": 1.3964, "step": 3630 }, { "epoch": 0.68, "learning_rate": 1.4600533990552476e-06, "loss": 1.3778, "step": 3635 }, { "epoch": 0.68, "learning_rate": 1.4590264941466418e-06, "loss": 1.6348, "step": 3640 }, { "epoch": 0.68, "learning_rate": 1.4579995892380365e-06, "loss": 1.2501, "step": 3645 }, { "epoch": 0.68, "learning_rate": 1.456972684329431e-06, "loss": 1.2765, "step": 3650 }, { "epoch": 0.68, "learning_rate": 1.4559457794208257e-06, "loss": 1.8279, "step": 3655 }, { "epoch": 0.68, "learning_rate": 1.45491887451222e-06, "loss": 1.3734, "step": 3660 }, { "epoch": 0.68, "learning_rate": 1.4538919696036147e-06, "loss": 1.8975, "step": 3665 }, { "epoch": 0.68, "learning_rate": 1.4528650646950092e-06, "loss": 1.3493, "step": 3670 }, { "epoch": 0.68, "learning_rate": 1.4518381597864039e-06, "loss": 1.6834, "step": 3675 }, { "epoch": 0.69, "learning_rate": 1.4508112548777981e-06, "loss": 1.2842, "step": 3680 }, { "epoch": 0.69, "learning_rate": 1.4497843499691928e-06, "loss": 1.5672, "step": 3685 }, { "epoch": 0.69, "learning_rate": 1.4487574450605873e-06, "loss": 1.415, "step": 3690 }, { "epoch": 0.69, "learning_rate": 1.447730540151982e-06, "loss": 1.7638, "step": 3695 }, { "epoch": 0.69, "learning_rate": 1.4467036352433763e-06, "loss": 2.0034, "step": 3700 }, { "epoch": 0.69, "learning_rate": 1.445676730334771e-06, "loss": 1.2874, "step": 3705 }, { "epoch": 0.69, "learning_rate": 1.4446498254261655e-06, "loss": 1.5264, "step": 3710 }, { "epoch": 0.69, "learning_rate": 1.4436229205175602e-06, "loss": 1.5826, "step": 3715 }, { "epoch": 0.69, "learning_rate": 1.4425960156089544e-06, "loss": 1.4389, "step": 3720 }, { "epoch": 0.69, "learning_rate": 1.4415691107003491e-06, "loss": 1.3155, "step": 3725 }, { "epoch": 0.69, "learning_rate": 1.4405422057917436e-06, "loss": 1.229, "step": 3730 }, { "epoch": 0.7, "learning_rate": 1.4395153008831383e-06, "loss": 1.5056, "step": 3735 }, { "epoch": 0.7, "learning_rate": 1.4384883959745326e-06, "loss": 1.7682, "step": 3740 }, { "epoch": 0.7, "learning_rate": 1.4374614910659273e-06, "loss": 1.5526, "step": 3745 }, { "epoch": 0.7, "learning_rate": 1.4364345861573218e-06, "loss": 1.6273, "step": 3750 }, { "epoch": 0.7, "learning_rate": 1.4354076812487163e-06, "loss": 1.1467, "step": 3755 }, { "epoch": 0.7, "learning_rate": 1.4343807763401107e-06, "loss": 1.3847, "step": 3760 }, { "epoch": 0.7, "learning_rate": 1.4333538714315054e-06, "loss": 1.3596, "step": 3765 }, { "epoch": 0.7, "learning_rate": 1.4323269665229e-06, "loss": 1.5988, "step": 3770 }, { "epoch": 0.7, "learning_rate": 1.4313000616142944e-06, "loss": 1.5569, "step": 3775 }, { "epoch": 0.7, "learning_rate": 1.430273156705689e-06, "loss": 1.4909, "step": 3780 }, { "epoch": 0.7, "learning_rate": 1.4292462517970836e-06, "loss": 1.1364, "step": 3785 }, { "epoch": 0.71, "learning_rate": 1.428219346888478e-06, "loss": 1.4173, "step": 3790 }, { "epoch": 0.71, "learning_rate": 1.4271924419798726e-06, "loss": 1.4494, "step": 3795 }, { "epoch": 0.71, "learning_rate": 1.426165537071267e-06, "loss": 1.8525, "step": 3800 }, { "epoch": 0.71, "learning_rate": 1.4251386321626617e-06, "loss": 1.6446, "step": 3805 }, { "epoch": 0.71, "learning_rate": 1.4241117272540562e-06, "loss": 1.5038, "step": 3810 }, { "epoch": 0.71, "learning_rate": 1.4230848223454507e-06, "loss": 1.4557, "step": 3815 }, { "epoch": 0.71, "learning_rate": 1.4220579174368452e-06, "loss": 1.1329, "step": 3820 }, { "epoch": 0.71, "learning_rate": 1.42103101252824e-06, "loss": 1.4263, "step": 3825 }, { "epoch": 0.71, "learning_rate": 1.4200041076196344e-06, "loss": 1.1607, "step": 3830 }, { "epoch": 0.71, "learning_rate": 1.4189772027110289e-06, "loss": 1.5083, "step": 3835 }, { "epoch": 0.72, "learning_rate": 1.4179502978024234e-06, "loss": 1.6398, "step": 3840 }, { "epoch": 0.72, "learning_rate": 1.416923392893818e-06, "loss": 1.1638, "step": 3845 }, { "epoch": 0.72, "learning_rate": 1.4158964879852125e-06, "loss": 1.5206, "step": 3850 }, { "epoch": 0.72, "learning_rate": 1.414869583076607e-06, "loss": 1.4404, "step": 3855 }, { "epoch": 0.72, "learning_rate": 1.4138426781680015e-06, "loss": 1.4838, "step": 3860 }, { "epoch": 0.72, "learning_rate": 1.4128157732593962e-06, "loss": 1.3232, "step": 3865 }, { "epoch": 0.72, "learning_rate": 1.4117888683507907e-06, "loss": 1.4703, "step": 3870 }, { "epoch": 0.72, "learning_rate": 1.4107619634421852e-06, "loss": 1.2905, "step": 3875 }, { "epoch": 0.72, "learning_rate": 1.4097350585335797e-06, "loss": 1.4413, "step": 3880 }, { "epoch": 0.72, "learning_rate": 1.4087081536249744e-06, "loss": 1.1968, "step": 3885 }, { "epoch": 0.72, "learning_rate": 1.4076812487163686e-06, "loss": 1.3222, "step": 3890 }, { "epoch": 0.73, "learning_rate": 1.4066543438077633e-06, "loss": 1.1413, "step": 3895 }, { "epoch": 0.73, "learning_rate": 1.4056274388991578e-06, "loss": 1.5517, "step": 3900 }, { "epoch": 0.73, "learning_rate": 1.4046005339905525e-06, "loss": 1.1597, "step": 3905 }, { "epoch": 0.73, "learning_rate": 1.4035736290819468e-06, "loss": 1.2781, "step": 3910 }, { "epoch": 0.73, "learning_rate": 1.4025467241733415e-06, "loss": 1.1736, "step": 3915 }, { "epoch": 0.73, "learning_rate": 1.401519819264736e-06, "loss": 1.3079, "step": 3920 }, { "epoch": 0.73, "learning_rate": 1.4004929143561307e-06, "loss": 1.2027, "step": 3925 }, { "epoch": 0.73, "learning_rate": 1.399466009447525e-06, "loss": 1.2086, "step": 3930 }, { "epoch": 0.73, "learning_rate": 1.3984391045389196e-06, "loss": 1.243, "step": 3935 }, { "epoch": 0.73, "learning_rate": 1.3974121996303141e-06, "loss": 1.6102, "step": 3940 }, { "epoch": 0.73, "learning_rate": 1.3963852947217088e-06, "loss": 1.2809, "step": 3945 }, { "epoch": 0.74, "learning_rate": 1.395358389813103e-06, "loss": 1.1862, "step": 3950 }, { "epoch": 0.74, "learning_rate": 1.3943314849044978e-06, "loss": 1.274, "step": 3955 }, { "epoch": 0.74, "learning_rate": 1.3933045799958923e-06, "loss": 1.6845, "step": 3960 }, { "epoch": 0.74, "learning_rate": 1.392277675087287e-06, "loss": 1.8259, "step": 3965 }, { "epoch": 0.74, "learning_rate": 1.3912507701786812e-06, "loss": 1.5038, "step": 3970 }, { "epoch": 0.74, "learning_rate": 1.390223865270076e-06, "loss": 1.3773, "step": 3975 }, { "epoch": 0.74, "learning_rate": 1.3891969603614704e-06, "loss": 1.2645, "step": 3980 }, { "epoch": 0.74, "learning_rate": 1.3881700554528651e-06, "loss": 1.2258, "step": 3985 }, { "epoch": 0.74, "learning_rate": 1.3871431505442594e-06, "loss": 1.2284, "step": 3990 }, { "epoch": 0.74, "learning_rate": 1.386116245635654e-06, "loss": 1.4279, "step": 3995 }, { "epoch": 0.75, "learning_rate": 1.3850893407270486e-06, "loss": 1.4324, "step": 4000 }, { "epoch": 0.75, "eval_loss": 0.871010959148407, "eval_runtime": 165.2205, "eval_samples_per_second": 7.227, "eval_steps_per_second": 0.908, "eval_wer": 0.9969214982042073, "step": 4000 }, { "epoch": 0.75, "learning_rate": 1.3840624358184433e-06, "loss": 1.4432, "step": 4005 }, { "epoch": 0.75, "learning_rate": 1.3830355309098375e-06, "loss": 1.1459, "step": 4010 }, { "epoch": 0.75, "learning_rate": 1.3820086260012322e-06, "loss": 1.8922, "step": 4015 }, { "epoch": 0.75, "learning_rate": 1.3809817210926267e-06, "loss": 1.5279, "step": 4020 }, { "epoch": 0.75, "learning_rate": 1.3799548161840214e-06, "loss": 1.4555, "step": 4025 }, { "epoch": 0.75, "learning_rate": 1.3789279112754157e-06, "loss": 1.255, "step": 4030 }, { "epoch": 0.75, "learning_rate": 1.3779010063668104e-06, "loss": 1.0926, "step": 4035 }, { "epoch": 0.75, "learning_rate": 1.3768741014582049e-06, "loss": 1.423, "step": 4040 }, { "epoch": 0.75, "learning_rate": 1.3758471965495994e-06, "loss": 1.0813, "step": 4045 }, { "epoch": 0.75, "learning_rate": 1.3748202916409938e-06, "loss": 1.3814, "step": 4050 }, { "epoch": 0.76, "learning_rate": 1.3737933867323885e-06, "loss": 1.2505, "step": 4055 }, { "epoch": 0.76, "learning_rate": 1.372766481823783e-06, "loss": 2.0445, "step": 4060 }, { "epoch": 0.76, "learning_rate": 1.3717395769151775e-06, "loss": 1.3996, "step": 4065 }, { "epoch": 0.76, "learning_rate": 1.370712672006572e-06, "loss": 1.3825, "step": 4070 }, { "epoch": 0.76, "learning_rate": 1.3696857670979667e-06, "loss": 1.5115, "step": 4075 }, { "epoch": 0.76, "learning_rate": 1.3686588621893612e-06, "loss": 1.3877, "step": 4080 }, { "epoch": 0.76, "learning_rate": 1.3676319572807557e-06, "loss": 1.0905, "step": 4085 }, { "epoch": 0.76, "learning_rate": 1.3666050523721501e-06, "loss": 1.3608, "step": 4090 }, { "epoch": 0.76, "learning_rate": 1.3655781474635448e-06, "loss": 1.3166, "step": 4095 }, { "epoch": 0.76, "learning_rate": 1.3645512425549393e-06, "loss": 1.0832, "step": 4100 }, { "epoch": 0.76, "learning_rate": 1.3635243376463338e-06, "loss": 1.248, "step": 4105 }, { "epoch": 0.77, "learning_rate": 1.3624974327377283e-06, "loss": 1.1094, "step": 4110 }, { "epoch": 0.77, "learning_rate": 1.361470527829123e-06, "loss": 1.1145, "step": 4115 }, { "epoch": 0.77, "learning_rate": 1.3604436229205175e-06, "loss": 1.5423, "step": 4120 }, { "epoch": 0.77, "learning_rate": 1.359416718011912e-06, "loss": 1.1091, "step": 4125 }, { "epoch": 0.77, "learning_rate": 1.3583898131033065e-06, "loss": 1.3596, "step": 4130 }, { "epoch": 0.77, "learning_rate": 1.3573629081947012e-06, "loss": 1.1707, "step": 4135 }, { "epoch": 0.77, "learning_rate": 1.3563360032860956e-06, "loss": 1.3573, "step": 4140 }, { "epoch": 0.77, "learning_rate": 1.3553090983774901e-06, "loss": 1.4772, "step": 4145 }, { "epoch": 0.77, "learning_rate": 1.3542821934688846e-06, "loss": 1.1288, "step": 4150 }, { "epoch": 0.77, "learning_rate": 1.3532552885602793e-06, "loss": 1.1474, "step": 4155 }, { "epoch": 0.77, "learning_rate": 1.3522283836516738e-06, "loss": 1.2226, "step": 4160 }, { "epoch": 0.78, "learning_rate": 1.3512014787430683e-06, "loss": 1.6267, "step": 4165 }, { "epoch": 0.78, "learning_rate": 1.3501745738344628e-06, "loss": 1.4794, "step": 4170 }, { "epoch": 0.78, "learning_rate": 1.3491476689258575e-06, "loss": 1.2631, "step": 4175 }, { "epoch": 0.78, "learning_rate": 1.3481207640172517e-06, "loss": 1.4549, "step": 4180 }, { "epoch": 0.78, "learning_rate": 1.3470938591086464e-06, "loss": 1.434, "step": 4185 }, { "epoch": 0.78, "learning_rate": 1.346066954200041e-06, "loss": 1.2872, "step": 4190 }, { "epoch": 0.78, "learning_rate": 1.3450400492914356e-06, "loss": 1.3279, "step": 4195 }, { "epoch": 0.78, "learning_rate": 1.3440131443828303e-06, "loss": 1.3805, "step": 4200 }, { "epoch": 0.78, "learning_rate": 1.3429862394742246e-06, "loss": 1.1369, "step": 4205 }, { "epoch": 0.78, "learning_rate": 1.3419593345656193e-06, "loss": 1.1013, "step": 4210 }, { "epoch": 0.79, "learning_rate": 1.3409324296570138e-06, "loss": 0.9662, "step": 4215 }, { "epoch": 0.79, "learning_rate": 1.3399055247484082e-06, "loss": 1.093, "step": 4220 }, { "epoch": 0.79, "learning_rate": 1.3388786198398027e-06, "loss": 1.2849, "step": 4225 }, { "epoch": 0.79, "learning_rate": 1.3378517149311974e-06, "loss": 1.0428, "step": 4230 }, { "epoch": 0.79, "learning_rate": 1.336824810022592e-06, "loss": 1.083, "step": 4235 }, { "epoch": 0.79, "learning_rate": 1.3357979051139864e-06, "loss": 1.3622, "step": 4240 }, { "epoch": 0.79, "learning_rate": 1.3347710002053809e-06, "loss": 1.1327, "step": 4245 }, { "epoch": 0.79, "learning_rate": 1.3337440952967756e-06, "loss": 1.0384, "step": 4250 }, { "epoch": 0.79, "learning_rate": 1.33271719038817e-06, "loss": 0.9773, "step": 4255 }, { "epoch": 0.79, "learning_rate": 1.3316902854795645e-06, "loss": 1.2673, "step": 4260 }, { "epoch": 0.79, "learning_rate": 1.330663380570959e-06, "loss": 1.3151, "step": 4265 }, { "epoch": 0.8, "learning_rate": 1.3296364756623537e-06, "loss": 1.4965, "step": 4270 }, { "epoch": 0.8, "learning_rate": 1.3286095707537482e-06, "loss": 1.3185, "step": 4275 }, { "epoch": 0.8, "learning_rate": 1.3275826658451427e-06, "loss": 1.6872, "step": 4280 }, { "epoch": 0.8, "learning_rate": 1.3265557609365372e-06, "loss": 1.6085, "step": 4285 }, { "epoch": 0.8, "learning_rate": 1.3255288560279319e-06, "loss": 1.3537, "step": 4290 }, { "epoch": 0.8, "learning_rate": 1.3245019511193264e-06, "loss": 1.1366, "step": 4295 }, { "epoch": 0.8, "learning_rate": 1.3234750462107209e-06, "loss": 1.1825, "step": 4300 }, { "epoch": 0.8, "learning_rate": 1.3224481413021153e-06, "loss": 1.3218, "step": 4305 }, { "epoch": 0.8, "learning_rate": 1.32142123639351e-06, "loss": 1.1001, "step": 4310 }, { "epoch": 0.8, "learning_rate": 1.3203943314849045e-06, "loss": 1.0751, "step": 4315 }, { "epoch": 0.8, "learning_rate": 1.319367426576299e-06, "loss": 1.3475, "step": 4320 }, { "epoch": 0.81, "learning_rate": 1.3183405216676935e-06, "loss": 1.0859, "step": 4325 }, { "epoch": 0.81, "learning_rate": 1.3173136167590882e-06, "loss": 1.4619, "step": 4330 }, { "epoch": 0.81, "learning_rate": 1.3162867118504827e-06, "loss": 1.1734, "step": 4335 }, { "epoch": 0.81, "learning_rate": 1.3152598069418772e-06, "loss": 1.0587, "step": 4340 }, { "epoch": 0.81, "learning_rate": 1.3142329020332716e-06, "loss": 1.0159, "step": 4345 }, { "epoch": 0.81, "learning_rate": 1.3132059971246663e-06, "loss": 0.9952, "step": 4350 }, { "epoch": 0.81, "learning_rate": 1.3121790922160606e-06, "loss": 0.9823, "step": 4355 }, { "epoch": 0.81, "learning_rate": 1.3111521873074553e-06, "loss": 1.209, "step": 4360 }, { "epoch": 0.81, "learning_rate": 1.3101252823988498e-06, "loss": 1.5698, "step": 4365 }, { "epoch": 0.81, "learning_rate": 1.3090983774902445e-06, "loss": 1.3529, "step": 4370 }, { "epoch": 0.81, "learning_rate": 1.3080714725816388e-06, "loss": 1.2588, "step": 4375 }, { "epoch": 0.82, "learning_rate": 1.3070445676730335e-06, "loss": 1.6183, "step": 4380 }, { "epoch": 0.82, "learning_rate": 1.306017662764428e-06, "loss": 1.0698, "step": 4385 }, { "epoch": 0.82, "learning_rate": 1.3049907578558226e-06, "loss": 1.0928, "step": 4390 }, { "epoch": 0.82, "learning_rate": 1.303963852947217e-06, "loss": 1.027, "step": 4395 }, { "epoch": 0.82, "learning_rate": 1.3029369480386116e-06, "loss": 1.2732, "step": 4400 }, { "epoch": 0.82, "learning_rate": 1.301910043130006e-06, "loss": 1.0801, "step": 4405 }, { "epoch": 0.82, "learning_rate": 1.3008831382214008e-06, "loss": 1.6345, "step": 4410 }, { "epoch": 0.82, "learning_rate": 1.299856233312795e-06, "loss": 1.228, "step": 4415 }, { "epoch": 0.82, "learning_rate": 1.2988293284041898e-06, "loss": 1.1848, "step": 4420 }, { "epoch": 0.82, "learning_rate": 1.2978024234955842e-06, "loss": 1.0078, "step": 4425 }, { "epoch": 0.83, "learning_rate": 1.296775518586979e-06, "loss": 1.0556, "step": 4430 }, { "epoch": 0.83, "learning_rate": 1.2957486136783732e-06, "loss": 1.0646, "step": 4435 }, { "epoch": 0.83, "learning_rate": 1.294721708769768e-06, "loss": 1.1601, "step": 4440 }, { "epoch": 0.83, "learning_rate": 1.2936948038611624e-06, "loss": 1.0673, "step": 4445 }, { "epoch": 0.83, "learning_rate": 1.292667898952557e-06, "loss": 1.4266, "step": 4450 }, { "epoch": 0.83, "learning_rate": 1.2916409940439514e-06, "loss": 1.0005, "step": 4455 }, { "epoch": 0.83, "learning_rate": 1.290614089135346e-06, "loss": 1.0665, "step": 4460 }, { "epoch": 0.83, "learning_rate": 1.2895871842267406e-06, "loss": 1.7098, "step": 4465 }, { "epoch": 0.83, "learning_rate": 1.2885602793181353e-06, "loss": 1.0159, "step": 4470 }, { "epoch": 0.83, "learning_rate": 1.2875333744095295e-06, "loss": 1.0682, "step": 4475 }, { "epoch": 0.83, "learning_rate": 1.2865064695009242e-06, "loss": 1.029, "step": 4480 }, { "epoch": 0.84, "learning_rate": 1.2854795645923187e-06, "loss": 1.2077, "step": 4485 }, { "epoch": 0.84, "learning_rate": 1.2844526596837134e-06, "loss": 1.1243, "step": 4490 }, { "epoch": 0.84, "learning_rate": 1.2834257547751077e-06, "loss": 0.8508, "step": 4495 }, { "epoch": 0.84, "learning_rate": 1.2823988498665024e-06, "loss": 1.5273, "step": 4500 }, { "epoch": 0.84, "eval_loss": 0.7541612386703491, "eval_runtime": 160.776, "eval_samples_per_second": 7.426, "eval_steps_per_second": 0.933, "eval_wer": 0.9961518727552591, "step": 4500 }, { "epoch": 0.84, "learning_rate": 1.2813719449578969e-06, "loss": 1.4633, "step": 4505 }, { "epoch": 0.84, "learning_rate": 1.2803450400492913e-06, "loss": 1.0911, "step": 4510 }, { "epoch": 0.84, "learning_rate": 1.2793181351406858e-06, "loss": 1.039, "step": 4515 }, { "epoch": 0.84, "learning_rate": 1.2782912302320805e-06, "loss": 1.1547, "step": 4520 }, { "epoch": 0.84, "learning_rate": 1.277264325323475e-06, "loss": 0.9209, "step": 4525 }, { "epoch": 0.84, "learning_rate": 1.2762374204148695e-06, "loss": 0.9886, "step": 4530 }, { "epoch": 0.84, "learning_rate": 1.275210515506264e-06, "loss": 1.0997, "step": 4535 }, { "epoch": 0.85, "learning_rate": 1.2741836105976587e-06, "loss": 1.9769, "step": 4540 }, { "epoch": 0.85, "learning_rate": 1.2731567056890532e-06, "loss": 0.9819, "step": 4545 }, { "epoch": 0.85, "learning_rate": 1.2721298007804476e-06, "loss": 1.8481, "step": 4550 }, { "epoch": 0.85, "learning_rate": 1.2711028958718421e-06, "loss": 1.2041, "step": 4555 }, { "epoch": 0.85, "learning_rate": 1.2700759909632368e-06, "loss": 1.3081, "step": 4560 }, { "epoch": 0.85, "learning_rate": 1.2690490860546313e-06, "loss": 1.3219, "step": 4565 }, { "epoch": 0.85, "learning_rate": 1.2680221811460258e-06, "loss": 1.3714, "step": 4570 }, { "epoch": 0.85, "learning_rate": 1.2669952762374203e-06, "loss": 0.9208, "step": 4575 }, { "epoch": 0.85, "learning_rate": 1.265968371328815e-06, "loss": 1.0268, "step": 4580 }, { "epoch": 0.85, "learning_rate": 1.2649414664202095e-06, "loss": 0.9589, "step": 4585 }, { "epoch": 0.85, "learning_rate": 1.263914561511604e-06, "loss": 1.1938, "step": 4590 }, { "epoch": 0.86, "learning_rate": 1.2628876566029984e-06, "loss": 1.0504, "step": 4595 }, { "epoch": 0.86, "learning_rate": 1.2618607516943931e-06, "loss": 1.4992, "step": 4600 }, { "epoch": 0.86, "learning_rate": 1.2608338467857876e-06, "loss": 1.0982, "step": 4605 }, { "epoch": 0.86, "learning_rate": 1.259806941877182e-06, "loss": 1.1808, "step": 4610 }, { "epoch": 0.86, "learning_rate": 1.2587800369685766e-06, "loss": 0.9891, "step": 4615 }, { "epoch": 0.86, "learning_rate": 1.2577531320599713e-06, "loss": 0.9634, "step": 4620 }, { "epoch": 0.86, "learning_rate": 1.2567262271513658e-06, "loss": 1.253, "step": 4625 }, { "epoch": 0.86, "learning_rate": 1.2556993222427603e-06, "loss": 0.9934, "step": 4630 }, { "epoch": 0.86, "learning_rate": 1.2546724173341547e-06, "loss": 0.99, "step": 4635 }, { "epoch": 0.86, "learning_rate": 1.2536455124255494e-06, "loss": 1.1702, "step": 4640 }, { "epoch": 0.87, "learning_rate": 1.2526186075169437e-06, "loss": 1.2641, "step": 4645 }, { "epoch": 0.87, "learning_rate": 1.2515917026083384e-06, "loss": 0.9113, "step": 4650 }, { "epoch": 0.87, "learning_rate": 1.2505647976997329e-06, "loss": 0.8693, "step": 4655 }, { "epoch": 0.87, "learning_rate": 1.2495378927911276e-06, "loss": 1.0559, "step": 4660 }, { "epoch": 0.87, "learning_rate": 1.2485109878825219e-06, "loss": 1.177, "step": 4665 }, { "epoch": 0.87, "learning_rate": 1.2474840829739166e-06, "loss": 1.0942, "step": 4670 }, { "epoch": 0.87, "learning_rate": 1.246457178065311e-06, "loss": 1.0034, "step": 4675 }, { "epoch": 0.87, "learning_rate": 1.2454302731567057e-06, "loss": 0.8607, "step": 4680 }, { "epoch": 0.87, "learning_rate": 1.2444033682481e-06, "loss": 1.5524, "step": 4685 }, { "epoch": 0.87, "learning_rate": 1.2433764633394947e-06, "loss": 1.2187, "step": 4690 }, { "epoch": 0.87, "learning_rate": 1.2423495584308892e-06, "loss": 1.05, "step": 4695 }, { "epoch": 0.88, "learning_rate": 1.2413226535222839e-06, "loss": 1.1299, "step": 4700 }, { "epoch": 0.88, "learning_rate": 1.2402957486136782e-06, "loss": 1.0083, "step": 4705 }, { "epoch": 0.88, "learning_rate": 1.2392688437050729e-06, "loss": 1.0147, "step": 4710 }, { "epoch": 0.88, "learning_rate": 1.2382419387964673e-06, "loss": 1.1144, "step": 4715 }, { "epoch": 0.88, "learning_rate": 1.237215033887862e-06, "loss": 0.9494, "step": 4720 }, { "epoch": 0.88, "learning_rate": 1.2361881289792563e-06, "loss": 0.8921, "step": 4725 }, { "epoch": 0.88, "learning_rate": 1.235161224070651e-06, "loss": 0.9891, "step": 4730 }, { "epoch": 0.88, "learning_rate": 1.2341343191620455e-06, "loss": 1.0606, "step": 4735 }, { "epoch": 0.88, "learning_rate": 1.2331074142534402e-06, "loss": 1.1937, "step": 4740 }, { "epoch": 0.88, "learning_rate": 1.2320805093448345e-06, "loss": 0.9862, "step": 4745 }, { "epoch": 0.88, "learning_rate": 1.2310536044362292e-06, "loss": 1.3468, "step": 4750 }, { "epoch": 0.89, "learning_rate": 1.2300266995276237e-06, "loss": 0.8859, "step": 4755 }, { "epoch": 0.89, "learning_rate": 1.2289997946190183e-06, "loss": 0.9742, "step": 4760 }, { "epoch": 0.89, "learning_rate": 1.2279728897104126e-06, "loss": 0.898, "step": 4765 }, { "epoch": 0.89, "learning_rate": 1.2269459848018073e-06, "loss": 1.799, "step": 4770 }, { "epoch": 0.89, "learning_rate": 1.2259190798932018e-06, "loss": 0.8833, "step": 4775 }, { "epoch": 0.89, "learning_rate": 1.2248921749845965e-06, "loss": 1.2739, "step": 4780 }, { "epoch": 0.89, "learning_rate": 1.2238652700759908e-06, "loss": 0.9587, "step": 4785 }, { "epoch": 0.89, "learning_rate": 1.2228383651673855e-06, "loss": 1.0981, "step": 4790 }, { "epoch": 0.89, "learning_rate": 1.22181146025878e-06, "loss": 1.0444, "step": 4795 }, { "epoch": 0.89, "learning_rate": 1.2207845553501747e-06, "loss": 1.3144, "step": 4800 }, { "epoch": 0.89, "learning_rate": 1.219757650441569e-06, "loss": 0.8613, "step": 4805 }, { "epoch": 0.9, "learning_rate": 1.2187307455329636e-06, "loss": 0.9722, "step": 4810 }, { "epoch": 0.9, "learning_rate": 1.2177038406243581e-06, "loss": 1.3232, "step": 4815 }, { "epoch": 0.9, "learning_rate": 1.2166769357157526e-06, "loss": 1.3468, "step": 4820 }, { "epoch": 0.9, "learning_rate": 1.215650030807147e-06, "loss": 0.9843, "step": 4825 }, { "epoch": 0.9, "learning_rate": 1.2146231258985418e-06, "loss": 1.0025, "step": 4830 }, { "epoch": 0.9, "learning_rate": 1.2135962209899363e-06, "loss": 1.192, "step": 4835 }, { "epoch": 0.9, "learning_rate": 1.2125693160813307e-06, "loss": 1.1494, "step": 4840 }, { "epoch": 0.9, "learning_rate": 1.2115424111727252e-06, "loss": 1.2217, "step": 4845 }, { "epoch": 0.9, "learning_rate": 1.21051550626412e-06, "loss": 1.058, "step": 4850 }, { "epoch": 0.9, "learning_rate": 1.2094886013555144e-06, "loss": 0.8975, "step": 4855 }, { "epoch": 0.91, "learning_rate": 1.208461696446909e-06, "loss": 1.0316, "step": 4860 }, { "epoch": 0.91, "learning_rate": 1.2074347915383034e-06, "loss": 1.3582, "step": 4865 }, { "epoch": 0.91, "learning_rate": 1.206407886629698e-06, "loss": 1.1331, "step": 4870 }, { "epoch": 0.91, "learning_rate": 1.2053809817210926e-06, "loss": 1.0301, "step": 4875 }, { "epoch": 0.91, "learning_rate": 1.204354076812487e-06, "loss": 0.9586, "step": 4880 }, { "epoch": 0.91, "learning_rate": 1.2033271719038815e-06, "loss": 0.8966, "step": 4885 }, { "epoch": 0.91, "learning_rate": 1.2023002669952762e-06, "loss": 1.3665, "step": 4890 }, { "epoch": 0.91, "learning_rate": 1.2012733620866707e-06, "loss": 1.1276, "step": 4895 }, { "epoch": 0.91, "learning_rate": 1.2002464571780652e-06, "loss": 1.4535, "step": 4900 }, { "epoch": 0.91, "learning_rate": 1.1992195522694597e-06, "loss": 0.8732, "step": 4905 }, { "epoch": 0.91, "learning_rate": 1.1981926473608544e-06, "loss": 1.1848, "step": 4910 }, { "epoch": 0.92, "learning_rate": 1.1971657424522489e-06, "loss": 0.976, "step": 4915 }, { "epoch": 0.92, "learning_rate": 1.1961388375436434e-06, "loss": 1.2866, "step": 4920 }, { "epoch": 0.92, "learning_rate": 1.1951119326350378e-06, "loss": 0.8825, "step": 4925 }, { "epoch": 0.92, "learning_rate": 1.1940850277264325e-06, "loss": 0.8137, "step": 4930 }, { "epoch": 0.92, "learning_rate": 1.193058122817827e-06, "loss": 1.1454, "step": 4935 }, { "epoch": 0.92, "learning_rate": 1.1920312179092215e-06, "loss": 0.9308, "step": 4940 }, { "epoch": 0.92, "learning_rate": 1.191004313000616e-06, "loss": 0.9336, "step": 4945 }, { "epoch": 0.92, "learning_rate": 1.1899774080920107e-06, "loss": 0.9893, "step": 4950 }, { "epoch": 0.92, "learning_rate": 1.188950503183405e-06, "loss": 1.312, "step": 4955 }, { "epoch": 0.92, "learning_rate": 1.1879235982747997e-06, "loss": 1.1768, "step": 4960 }, { "epoch": 0.92, "learning_rate": 1.1868966933661941e-06, "loss": 1.1879, "step": 4965 }, { "epoch": 0.93, "learning_rate": 1.1858697884575888e-06, "loss": 0.8554, "step": 4970 }, { "epoch": 0.93, "learning_rate": 1.1848428835489831e-06, "loss": 1.1773, "step": 4975 }, { "epoch": 0.93, "learning_rate": 1.1838159786403778e-06, "loss": 1.2384, "step": 4980 }, { "epoch": 0.93, "learning_rate": 1.1827890737317723e-06, "loss": 1.0867, "step": 4985 }, { "epoch": 0.93, "learning_rate": 1.181762168823167e-06, "loss": 0.9638, "step": 4990 }, { "epoch": 0.93, "learning_rate": 1.1807352639145613e-06, "loss": 0.9975, "step": 4995 }, { "epoch": 0.93, "learning_rate": 1.179708359005956e-06, "loss": 1.6983, "step": 5000 }, { "epoch": 0.93, "eval_loss": 0.6864951848983765, "eval_runtime": 158.2129, "eval_samples_per_second": 7.547, "eval_steps_per_second": 0.948, "eval_wer": 0.9917906618778861, "step": 5000 }, { "epoch": 0.93, "learning_rate": 1.1786814540973504e-06, "loss": 1.3032, "step": 5005 }, { "epoch": 0.93, "learning_rate": 1.1776545491887451e-06, "loss": 0.8693, "step": 5010 }, { "epoch": 0.93, "learning_rate": 1.1766276442801394e-06, "loss": 1.4751, "step": 5015 }, { "epoch": 0.93, "learning_rate": 1.1756007393715341e-06, "loss": 1.2441, "step": 5020 }, { "epoch": 0.94, "learning_rate": 1.1745738344629286e-06, "loss": 0.9813, "step": 5025 }, { "epoch": 0.94, "learning_rate": 1.1735469295543233e-06, "loss": 0.9869, "step": 5030 }, { "epoch": 0.94, "learning_rate": 1.1725200246457176e-06, "loss": 1.0416, "step": 5035 }, { "epoch": 0.94, "learning_rate": 1.1714931197371123e-06, "loss": 1.0176, "step": 5040 }, { "epoch": 0.94, "learning_rate": 1.170466214828507e-06, "loss": 0.9778, "step": 5045 }, { "epoch": 0.94, "learning_rate": 1.1694393099199014e-06, "loss": 0.9427, "step": 5050 }, { "epoch": 0.94, "learning_rate": 1.168412405011296e-06, "loss": 1.2824, "step": 5055 }, { "epoch": 0.94, "learning_rate": 1.1673855001026904e-06, "loss": 1.0722, "step": 5060 }, { "epoch": 0.94, "learning_rate": 1.1663585951940851e-06, "loss": 0.8688, "step": 5065 }, { "epoch": 0.94, "learning_rate": 1.1653316902854796e-06, "loss": 0.8749, "step": 5070 }, { "epoch": 0.95, "learning_rate": 1.164304785376874e-06, "loss": 1.1906, "step": 5075 }, { "epoch": 0.95, "learning_rate": 1.1632778804682686e-06, "loss": 1.2363, "step": 5080 }, { "epoch": 0.95, "learning_rate": 1.1622509755596633e-06, "loss": 1.2716, "step": 5085 }, { "epoch": 0.95, "learning_rate": 1.1612240706510578e-06, "loss": 1.2981, "step": 5090 }, { "epoch": 0.95, "learning_rate": 1.1601971657424522e-06, "loss": 1.1415, "step": 5095 }, { "epoch": 0.95, "learning_rate": 1.1591702608338467e-06, "loss": 0.9733, "step": 5100 }, { "epoch": 0.95, "learning_rate": 1.1581433559252414e-06, "loss": 1.0701, "step": 5105 }, { "epoch": 0.95, "learning_rate": 1.1571164510166357e-06, "loss": 1.2484, "step": 5110 }, { "epoch": 0.95, "learning_rate": 1.1560895461080304e-06, "loss": 1.1807, "step": 5115 }, { "epoch": 0.95, "learning_rate": 1.1550626411994249e-06, "loss": 1.4136, "step": 5120 }, { "epoch": 0.95, "learning_rate": 1.1540357362908196e-06, "loss": 0.888, "step": 5125 }, { "epoch": 0.96, "learning_rate": 1.1530088313822138e-06, "loss": 1.2419, "step": 5130 }, { "epoch": 0.96, "learning_rate": 1.1519819264736085e-06, "loss": 1.2136, "step": 5135 }, { "epoch": 0.96, "learning_rate": 1.150955021565003e-06, "loss": 1.1895, "step": 5140 }, { "epoch": 0.96, "learning_rate": 1.1499281166563977e-06, "loss": 0.895, "step": 5145 }, { "epoch": 0.96, "learning_rate": 1.148901211747792e-06, "loss": 1.2032, "step": 5150 }, { "epoch": 0.96, "learning_rate": 1.1478743068391867e-06, "loss": 0.9632, "step": 5155 }, { "epoch": 0.96, "learning_rate": 1.1468474019305812e-06, "loss": 1.2399, "step": 5160 }, { "epoch": 0.96, "learning_rate": 1.1458204970219759e-06, "loss": 1.0117, "step": 5165 }, { "epoch": 0.96, "learning_rate": 1.1447935921133701e-06, "loss": 1.2522, "step": 5170 }, { "epoch": 0.96, "learning_rate": 1.1437666872047648e-06, "loss": 0.915, "step": 5175 }, { "epoch": 0.96, "learning_rate": 1.1427397822961593e-06, "loss": 1.1109, "step": 5180 }, { "epoch": 0.97, "learning_rate": 1.141712877387554e-06, "loss": 1.1202, "step": 5185 }, { "epoch": 0.97, "learning_rate": 1.1406859724789483e-06, "loss": 1.1674, "step": 5190 }, { "epoch": 0.97, "learning_rate": 1.139659067570343e-06, "loss": 1.0151, "step": 5195 }, { "epoch": 0.97, "learning_rate": 1.1386321626617375e-06, "loss": 1.1584, "step": 5200 }, { "epoch": 0.97, "learning_rate": 1.1376052577531322e-06, "loss": 1.1947, "step": 5205 }, { "epoch": 0.97, "learning_rate": 1.1365783528445265e-06, "loss": 1.4659, "step": 5210 }, { "epoch": 0.97, "learning_rate": 1.1355514479359211e-06, "loss": 1.3269, "step": 5215 }, { "epoch": 0.97, "learning_rate": 1.1345245430273156e-06, "loss": 0.8676, "step": 5220 }, { "epoch": 0.97, "learning_rate": 1.1334976381187103e-06, "loss": 1.4621, "step": 5225 }, { "epoch": 0.97, "learning_rate": 1.1324707332101046e-06, "loss": 1.1258, "step": 5230 }, { "epoch": 0.98, "learning_rate": 1.1314438283014993e-06, "loss": 0.9158, "step": 5235 }, { "epoch": 0.98, "learning_rate": 1.1304169233928938e-06, "loss": 1.1497, "step": 5240 }, { "epoch": 0.98, "learning_rate": 1.1293900184842885e-06, "loss": 1.1792, "step": 5245 }, { "epoch": 0.98, "learning_rate": 1.1283631135756828e-06, "loss": 0.9564, "step": 5250 }, { "epoch": 0.98, "learning_rate": 1.1273362086670775e-06, "loss": 1.108, "step": 5255 }, { "epoch": 0.98, "learning_rate": 1.126309303758472e-06, "loss": 0.9212, "step": 5260 }, { "epoch": 0.98, "learning_rate": 1.1252823988498666e-06, "loss": 0.9636, "step": 5265 }, { "epoch": 0.98, "learning_rate": 1.124255493941261e-06, "loss": 0.8813, "step": 5270 }, { "epoch": 0.98, "learning_rate": 1.1232285890326556e-06, "loss": 0.9706, "step": 5275 }, { "epoch": 0.98, "learning_rate": 1.12220168412405e-06, "loss": 1.3708, "step": 5280 }, { "epoch": 0.98, "learning_rate": 1.1211747792154446e-06, "loss": 1.3527, "step": 5285 }, { "epoch": 0.99, "learning_rate": 1.120147874306839e-06, "loss": 0.897, "step": 5290 }, { "epoch": 0.99, "learning_rate": 1.1191209693982338e-06, "loss": 1.1322, "step": 5295 }, { "epoch": 0.99, "learning_rate": 1.1180940644896282e-06, "loss": 1.3505, "step": 5300 }, { "epoch": 0.99, "learning_rate": 1.1170671595810227e-06, "loss": 0.815, "step": 5305 }, { "epoch": 0.99, "learning_rate": 1.1160402546724172e-06, "loss": 0.9654, "step": 5310 }, { "epoch": 0.99, "learning_rate": 1.115013349763812e-06, "loss": 0.9672, "step": 5315 }, { "epoch": 0.99, "learning_rate": 1.1139864448552064e-06, "loss": 1.0071, "step": 5320 }, { "epoch": 0.99, "learning_rate": 1.1129595399466009e-06, "loss": 0.8848, "step": 5325 }, { "epoch": 0.99, "learning_rate": 1.1119326350379954e-06, "loss": 0.8209, "step": 5330 }, { "epoch": 0.99, "learning_rate": 1.11090573012939e-06, "loss": 0.8231, "step": 5335 }, { "epoch": 0.99, "learning_rate": 1.1098788252207845e-06, "loss": 0.9784, "step": 5340 }, { "epoch": 1.0, "learning_rate": 1.108851920312179e-06, "loss": 1.1011, "step": 5345 }, { "epoch": 1.0, "learning_rate": 1.1078250154035735e-06, "loss": 0.8685, "step": 5350 }, { "epoch": 1.0, "learning_rate": 1.1067981104949682e-06, "loss": 1.1179, "step": 5355 }, { "epoch": 1.0, "learning_rate": 1.1057712055863627e-06, "loss": 0.9418, "step": 5360 }, { "epoch": 1.0, "learning_rate": 1.1047443006777572e-06, "loss": 0.8272, "step": 5365 }, { "epoch": 1.0, "learning_rate": 1.1037173957691517e-06, "loss": 1.1035, "step": 5370 }, { "epoch": 1.0, "learning_rate": 1.1026904908605464e-06, "loss": 1.1308, "step": 5375 }, { "epoch": 1.0, "learning_rate": 1.1016635859519409e-06, "loss": 1.3084, "step": 5380 }, { "epoch": 1.0, "learning_rate": 1.1006366810433353e-06, "loss": 1.1212, "step": 5385 }, { "epoch": 1.0, "learning_rate": 1.0996097761347298e-06, "loss": 1.1658, "step": 5390 }, { "epoch": 1.0, "learning_rate": 1.0985828712261245e-06, "loss": 1.0412, "step": 5395 }, { "epoch": 1.01, "learning_rate": 1.097555966317519e-06, "loss": 0.9328, "step": 5400 }, { "epoch": 1.01, "learning_rate": 1.0965290614089135e-06, "loss": 1.2895, "step": 5405 }, { "epoch": 1.01, "learning_rate": 1.095502156500308e-06, "loss": 0.865, "step": 5410 }, { "epoch": 1.01, "learning_rate": 1.0944752515917027e-06, "loss": 0.8867, "step": 5415 }, { "epoch": 1.01, "learning_rate": 1.093448346683097e-06, "loss": 1.02, "step": 5420 }, { "epoch": 1.01, "learning_rate": 1.0924214417744916e-06, "loss": 1.03, "step": 5425 }, { "epoch": 1.01, "learning_rate": 1.0913945368658861e-06, "loss": 1.2617, "step": 5430 }, { "epoch": 1.01, "learning_rate": 1.0903676319572808e-06, "loss": 1.2242, "step": 5435 }, { "epoch": 1.01, "learning_rate": 1.089340727048675e-06, "loss": 0.9322, "step": 5440 }, { "epoch": 1.01, "learning_rate": 1.0883138221400698e-06, "loss": 1.1522, "step": 5445 }, { "epoch": 1.02, "learning_rate": 1.0872869172314643e-06, "loss": 1.1224, "step": 5450 }, { "epoch": 1.02, "learning_rate": 1.086260012322859e-06, "loss": 1.2487, "step": 5455 }, { "epoch": 1.02, "learning_rate": 1.0852331074142532e-06, "loss": 0.9581, "step": 5460 }, { "epoch": 1.02, "learning_rate": 1.084206202505648e-06, "loss": 1.442, "step": 5465 }, { "epoch": 1.02, "learning_rate": 1.0831792975970424e-06, "loss": 1.3162, "step": 5470 }, { "epoch": 1.02, "learning_rate": 1.0821523926884371e-06, "loss": 0.9452, "step": 5475 }, { "epoch": 1.02, "learning_rate": 1.0811254877798314e-06, "loss": 0.8652, "step": 5480 }, { "epoch": 1.02, "learning_rate": 1.080098582871226e-06, "loss": 1.0966, "step": 5485 }, { "epoch": 1.02, "learning_rate": 1.0790716779626206e-06, "loss": 0.8736, "step": 5490 }, { "epoch": 1.02, "learning_rate": 1.0780447730540153e-06, "loss": 0.8189, "step": 5495 }, { "epoch": 1.02, "learning_rate": 1.0770178681454096e-06, "loss": 1.3183, "step": 5500 }, { "epoch": 1.02, "eval_loss": 0.657252311706543, "eval_runtime": 155.1549, "eval_samples_per_second": 7.696, "eval_steps_per_second": 0.967, "eval_wer": 0.9889687018984095, "step": 5500 }, { "epoch": 1.03, "learning_rate": 1.0759909632368042e-06, "loss": 1.0584, "step": 5505 }, { "epoch": 1.03, "learning_rate": 1.0749640583281987e-06, "loss": 1.2603, "step": 5510 }, { "epoch": 1.03, "learning_rate": 1.0739371534195934e-06, "loss": 1.2782, "step": 5515 }, { "epoch": 1.03, "learning_rate": 1.0729102485109877e-06, "loss": 0.9035, "step": 5520 }, { "epoch": 1.03, "learning_rate": 1.0718833436023824e-06, "loss": 0.8609, "step": 5525 }, { "epoch": 1.03, "learning_rate": 1.0708564386937769e-06, "loss": 1.2478, "step": 5530 }, { "epoch": 1.03, "learning_rate": 1.0698295337851716e-06, "loss": 0.9603, "step": 5535 }, { "epoch": 1.03, "learning_rate": 1.0688026288765659e-06, "loss": 0.8891, "step": 5540 }, { "epoch": 1.03, "learning_rate": 1.0677757239679606e-06, "loss": 1.302, "step": 5545 }, { "epoch": 1.03, "learning_rate": 1.066748819059355e-06, "loss": 1.0154, "step": 5550 }, { "epoch": 1.03, "learning_rate": 1.0657219141507497e-06, "loss": 1.2031, "step": 5555 }, { "epoch": 1.04, "learning_rate": 1.064695009242144e-06, "loss": 1.3591, "step": 5560 }, { "epoch": 1.04, "learning_rate": 1.0636681043335387e-06, "loss": 1.8304, "step": 5565 }, { "epoch": 1.04, "learning_rate": 1.0626411994249332e-06, "loss": 0.8764, "step": 5570 }, { "epoch": 1.04, "learning_rate": 1.0616142945163277e-06, "loss": 0.8704, "step": 5575 }, { "epoch": 1.04, "learning_rate": 1.0605873896077222e-06, "loss": 0.9308, "step": 5580 }, { "epoch": 1.04, "learning_rate": 1.0595604846991169e-06, "loss": 0.8924, "step": 5585 }, { "epoch": 1.04, "learning_rate": 1.0585335797905113e-06, "loss": 0.8275, "step": 5590 }, { "epoch": 1.04, "learning_rate": 1.0575066748819058e-06, "loss": 0.9068, "step": 5595 }, { "epoch": 1.04, "learning_rate": 1.0564797699733003e-06, "loss": 1.1158, "step": 5600 }, { "epoch": 1.04, "learning_rate": 1.055452865064695e-06, "loss": 0.8869, "step": 5605 }, { "epoch": 1.04, "learning_rate": 1.0544259601560895e-06, "loss": 0.9003, "step": 5610 }, { "epoch": 1.05, "learning_rate": 1.053399055247484e-06, "loss": 1.0639, "step": 5615 }, { "epoch": 1.05, "learning_rate": 1.0523721503388785e-06, "loss": 1.3936, "step": 5620 }, { "epoch": 1.05, "learning_rate": 1.0513452454302732e-06, "loss": 1.2128, "step": 5625 }, { "epoch": 1.05, "learning_rate": 1.0503183405216676e-06, "loss": 1.1496, "step": 5630 }, { "epoch": 1.05, "learning_rate": 1.0492914356130621e-06, "loss": 1.1112, "step": 5635 }, { "epoch": 1.05, "learning_rate": 1.0482645307044566e-06, "loss": 1.406, "step": 5640 }, { "epoch": 1.05, "learning_rate": 1.0472376257958513e-06, "loss": 0.9241, "step": 5645 }, { "epoch": 1.05, "learning_rate": 1.0462107208872458e-06, "loss": 0.9112, "step": 5650 }, { "epoch": 1.05, "learning_rate": 1.0451838159786403e-06, "loss": 1.1099, "step": 5655 }, { "epoch": 1.05, "learning_rate": 1.0441569110700348e-06, "loss": 1.0678, "step": 5660 }, { "epoch": 1.06, "learning_rate": 1.0431300061614295e-06, "loss": 1.6929, "step": 5665 }, { "epoch": 1.06, "learning_rate": 1.042103101252824e-06, "loss": 0.8691, "step": 5670 }, { "epoch": 1.06, "learning_rate": 1.0410761963442184e-06, "loss": 0.7781, "step": 5675 }, { "epoch": 1.06, "learning_rate": 1.040049291435613e-06, "loss": 0.7718, "step": 5680 }, { "epoch": 1.06, "learning_rate": 1.0390223865270076e-06, "loss": 0.8345, "step": 5685 }, { "epoch": 1.06, "learning_rate": 1.037995481618402e-06, "loss": 0.8584, "step": 5690 }, { "epoch": 1.06, "learning_rate": 1.0369685767097966e-06, "loss": 0.8594, "step": 5695 }, { "epoch": 1.06, "learning_rate": 1.035941671801191e-06, "loss": 1.1204, "step": 5700 }, { "epoch": 1.06, "learning_rate": 1.0349147668925858e-06, "loss": 1.0265, "step": 5705 }, { "epoch": 1.06, "learning_rate": 1.03388786198398e-06, "loss": 0.7591, "step": 5710 }, { "epoch": 1.06, "learning_rate": 1.0328609570753747e-06, "loss": 0.8985, "step": 5715 }, { "epoch": 1.07, "learning_rate": 1.0318340521667692e-06, "loss": 1.3364, "step": 5720 }, { "epoch": 1.07, "learning_rate": 1.030807147258164e-06, "loss": 0.89, "step": 5725 }, { "epoch": 1.07, "learning_rate": 1.0297802423495582e-06, "loss": 1.0093, "step": 5730 }, { "epoch": 1.07, "learning_rate": 1.0287533374409529e-06, "loss": 0.7526, "step": 5735 }, { "epoch": 1.07, "learning_rate": 1.0277264325323474e-06, "loss": 0.8598, "step": 5740 }, { "epoch": 1.07, "learning_rate": 1.026699527623742e-06, "loss": 0.8158, "step": 5745 }, { "epoch": 1.07, "learning_rate": 1.0256726227151363e-06, "loss": 0.9379, "step": 5750 }, { "epoch": 1.07, "learning_rate": 1.024645717806531e-06, "loss": 1.0578, "step": 5755 }, { "epoch": 1.07, "learning_rate": 1.0236188128979255e-06, "loss": 0.9005, "step": 5760 }, { "epoch": 1.07, "learning_rate": 1.0225919079893202e-06, "loss": 0.7995, "step": 5765 }, { "epoch": 1.07, "learning_rate": 1.0215650030807145e-06, "loss": 1.4633, "step": 5770 }, { "epoch": 1.08, "learning_rate": 1.0205380981721092e-06, "loss": 0.9068, "step": 5775 }, { "epoch": 1.08, "learning_rate": 1.0195111932635037e-06, "loss": 1.0527, "step": 5780 }, { "epoch": 1.08, "learning_rate": 1.0184842883548984e-06, "loss": 1.186, "step": 5785 }, { "epoch": 1.08, "learning_rate": 1.0174573834462926e-06, "loss": 0.8254, "step": 5790 }, { "epoch": 1.08, "learning_rate": 1.0164304785376873e-06, "loss": 1.1384, "step": 5795 }, { "epoch": 1.08, "learning_rate": 1.0154035736290818e-06, "loss": 0.9161, "step": 5800 }, { "epoch": 1.08, "learning_rate": 1.0143766687204765e-06, "loss": 1.0261, "step": 5805 }, { "epoch": 1.08, "learning_rate": 1.0133497638118708e-06, "loss": 0.8808, "step": 5810 }, { "epoch": 1.08, "learning_rate": 1.0123228589032655e-06, "loss": 0.9922, "step": 5815 }, { "epoch": 1.08, "learning_rate": 1.01129595399466e-06, "loss": 1.5194, "step": 5820 }, { "epoch": 1.08, "learning_rate": 1.0102690490860547e-06, "loss": 0.7695, "step": 5825 }, { "epoch": 1.09, "learning_rate": 1.009242144177449e-06, "loss": 1.1249, "step": 5830 }, { "epoch": 1.09, "learning_rate": 1.0082152392688437e-06, "loss": 0.9865, "step": 5835 }, { "epoch": 1.09, "learning_rate": 1.0071883343602381e-06, "loss": 1.1333, "step": 5840 }, { "epoch": 1.09, "learning_rate": 1.0061614294516328e-06, "loss": 0.7846, "step": 5845 }, { "epoch": 1.09, "learning_rate": 1.005134524543027e-06, "loss": 1.7098, "step": 5850 }, { "epoch": 1.09, "learning_rate": 1.0041076196344218e-06, "loss": 1.1977, "step": 5855 }, { "epoch": 1.09, "learning_rate": 1.0030807147258163e-06, "loss": 1.0122, "step": 5860 }, { "epoch": 1.09, "learning_rate": 1.002053809817211e-06, "loss": 0.926, "step": 5865 }, { "epoch": 1.09, "learning_rate": 1.0010269049086053e-06, "loss": 0.9618, "step": 5870 }, { "epoch": 1.09, "learning_rate": 1e-06, "loss": 1.3674, "step": 5875 }, { "epoch": 1.1, "learning_rate": 9.989730950913944e-07, "loss": 1.2644, "step": 5880 }, { "epoch": 1.1, "learning_rate": 9.97946190182789e-07, "loss": 0.9674, "step": 5885 }, { "epoch": 1.1, "learning_rate": 9.969192852741836e-07, "loss": 1.0945, "step": 5890 }, { "epoch": 1.1, "learning_rate": 9.958923803655781e-07, "loss": 0.7358, "step": 5895 }, { "epoch": 1.1, "learning_rate": 9.948654754569726e-07, "loss": 1.0663, "step": 5900 }, { "epoch": 1.1, "learning_rate": 9.93838570548367e-07, "loss": 1.1117, "step": 5905 }, { "epoch": 1.1, "learning_rate": 9.928116656397618e-07, "loss": 0.8097, "step": 5910 }, { "epoch": 1.1, "learning_rate": 9.917847607311563e-07, "loss": 0.9344, "step": 5915 }, { "epoch": 1.1, "learning_rate": 9.907578558225507e-07, "loss": 0.9088, "step": 5920 }, { "epoch": 1.1, "learning_rate": 9.897309509139452e-07, "loss": 0.7678, "step": 5925 }, { "epoch": 1.1, "learning_rate": 9.8870404600534e-07, "loss": 0.8153, "step": 5930 }, { "epoch": 1.11, "learning_rate": 9.876771410967344e-07, "loss": 1.0027, "step": 5935 }, { "epoch": 1.11, "learning_rate": 9.86650236188129e-07, "loss": 0.8837, "step": 5940 }, { "epoch": 1.11, "learning_rate": 9.856233312795234e-07, "loss": 0.837, "step": 5945 }, { "epoch": 1.11, "learning_rate": 9.84596426370918e-07, "loss": 1.0137, "step": 5950 }, { "epoch": 1.11, "learning_rate": 9.835695214623126e-07, "loss": 1.3639, "step": 5955 }, { "epoch": 1.11, "learning_rate": 9.82542616553707e-07, "loss": 1.2383, "step": 5960 }, { "epoch": 1.11, "learning_rate": 9.815157116451015e-07, "loss": 0.8519, "step": 5965 }, { "epoch": 1.11, "learning_rate": 9.804888067364962e-07, "loss": 0.8987, "step": 5970 }, { "epoch": 1.11, "learning_rate": 9.794619018278907e-07, "loss": 0.7677, "step": 5975 }, { "epoch": 1.11, "learning_rate": 9.784349969192852e-07, "loss": 0.7758, "step": 5980 }, { "epoch": 1.11, "learning_rate": 9.774080920106797e-07, "loss": 0.7488, "step": 5985 }, { "epoch": 1.12, "learning_rate": 9.763811871020744e-07, "loss": 1.3031, "step": 5990 }, { "epoch": 1.12, "learning_rate": 9.753542821934689e-07, "loss": 0.9354, "step": 5995 }, { "epoch": 1.12, "learning_rate": 9.743273772848634e-07, "loss": 0.8463, "step": 6000 }, { "epoch": 1.12, "eval_loss": 0.6284204125404358, "eval_runtime": 159.2114, "eval_samples_per_second": 7.499, "eval_steps_per_second": 0.942, "eval_wer": 0.9790918419702411, "step": 6000 }, { "epoch": 1.12, "learning_rate": 9.733004723762578e-07, "loss": 1.0615, "step": 6005 }, { "epoch": 1.12, "learning_rate": 9.722735674676525e-07, "loss": 0.7658, "step": 6010 }, { "epoch": 1.12, "learning_rate": 9.71246662559047e-07, "loss": 0.7916, "step": 6015 }, { "epoch": 1.12, "learning_rate": 9.702197576504415e-07, "loss": 0.8727, "step": 6020 }, { "epoch": 1.12, "learning_rate": 9.69192852741836e-07, "loss": 1.3318, "step": 6025 }, { "epoch": 1.12, "learning_rate": 9.681659478332305e-07, "loss": 0.9762, "step": 6030 }, { "epoch": 1.12, "learning_rate": 9.671390429246252e-07, "loss": 1.13, "step": 6035 }, { "epoch": 1.12, "learning_rate": 9.661121380160197e-07, "loss": 0.9907, "step": 6040 }, { "epoch": 1.13, "learning_rate": 9.650852331074141e-07, "loss": 0.8157, "step": 6045 }, { "epoch": 1.13, "learning_rate": 9.640583281988086e-07, "loss": 1.3409, "step": 6050 }, { "epoch": 1.13, "learning_rate": 9.630314232902033e-07, "loss": 0.7981, "step": 6055 }, { "epoch": 1.13, "learning_rate": 9.620045183815978e-07, "loss": 1.1387, "step": 6060 }, { "epoch": 1.13, "learning_rate": 9.609776134729923e-07, "loss": 1.3179, "step": 6065 }, { "epoch": 1.13, "learning_rate": 9.599507085643868e-07, "loss": 1.3169, "step": 6070 }, { "epoch": 1.13, "learning_rate": 9.589238036557815e-07, "loss": 0.859, "step": 6075 }, { "epoch": 1.13, "learning_rate": 9.57896898747176e-07, "loss": 0.8268, "step": 6080 }, { "epoch": 1.13, "learning_rate": 9.568699938385704e-07, "loss": 0.8662, "step": 6085 }, { "epoch": 1.13, "learning_rate": 9.55843088929965e-07, "loss": 1.1457, "step": 6090 }, { "epoch": 1.14, "learning_rate": 9.548161840213596e-07, "loss": 0.884, "step": 6095 }, { "epoch": 1.14, "learning_rate": 9.537892791127541e-07, "loss": 1.0815, "step": 6100 }, { "epoch": 1.14, "learning_rate": 9.527623742041486e-07, "loss": 0.8797, "step": 6105 }, { "epoch": 1.14, "learning_rate": 9.517354692955432e-07, "loss": 0.9832, "step": 6110 }, { "epoch": 1.14, "learning_rate": 9.507085643869377e-07, "loss": 1.0398, "step": 6115 }, { "epoch": 1.14, "learning_rate": 9.496816594783323e-07, "loss": 0.9358, "step": 6120 }, { "epoch": 1.14, "learning_rate": 9.486547545697267e-07, "loss": 0.7413, "step": 6125 }, { "epoch": 1.14, "learning_rate": 9.476278496611213e-07, "loss": 0.8723, "step": 6130 }, { "epoch": 1.14, "learning_rate": 9.466009447525158e-07, "loss": 0.8005, "step": 6135 }, { "epoch": 1.14, "learning_rate": 9.455740398439104e-07, "loss": 1.2658, "step": 6140 }, { "epoch": 1.14, "learning_rate": 9.445471349353049e-07, "loss": 0.7385, "step": 6145 }, { "epoch": 1.15, "learning_rate": 9.435202300266995e-07, "loss": 1.0049, "step": 6150 }, { "epoch": 1.15, "learning_rate": 9.42493325118094e-07, "loss": 0.9273, "step": 6155 }, { "epoch": 1.15, "learning_rate": 9.414664202094886e-07, "loss": 0.8961, "step": 6160 }, { "epoch": 1.15, "learning_rate": 9.404395153008831e-07, "loss": 0.9023, "step": 6165 }, { "epoch": 1.15, "learning_rate": 9.394126103922776e-07, "loss": 0.9846, "step": 6170 }, { "epoch": 1.15, "learning_rate": 9.383857054836721e-07, "loss": 0.8104, "step": 6175 }, { "epoch": 1.15, "learning_rate": 9.373588005750666e-07, "loss": 1.1124, "step": 6180 }, { "epoch": 1.15, "learning_rate": 9.363318956664612e-07, "loss": 0.9646, "step": 6185 }, { "epoch": 1.15, "learning_rate": 9.353049907578557e-07, "loss": 1.0231, "step": 6190 }, { "epoch": 1.15, "learning_rate": 9.342780858492503e-07, "loss": 1.2962, "step": 6195 }, { "epoch": 1.15, "learning_rate": 9.332511809406448e-07, "loss": 0.8414, "step": 6200 }, { "epoch": 1.16, "learning_rate": 9.322242760320394e-07, "loss": 0.7878, "step": 6205 }, { "epoch": 1.16, "learning_rate": 9.311973711234338e-07, "loss": 0.944, "step": 6210 }, { "epoch": 1.16, "learning_rate": 9.301704662148284e-07, "loss": 0.942, "step": 6215 }, { "epoch": 1.16, "learning_rate": 9.291435613062229e-07, "loss": 1.3707, "step": 6220 }, { "epoch": 1.16, "learning_rate": 9.281166563976175e-07, "loss": 0.8415, "step": 6225 }, { "epoch": 1.16, "learning_rate": 9.27089751489012e-07, "loss": 0.8805, "step": 6230 }, { "epoch": 1.16, "learning_rate": 9.260628465804066e-07, "loss": 0.891, "step": 6235 }, { "epoch": 1.16, "learning_rate": 9.250359416718011e-07, "loss": 0.7415, "step": 6240 }, { "epoch": 1.16, "learning_rate": 9.240090367631957e-07, "loss": 0.8968, "step": 6245 }, { "epoch": 1.16, "learning_rate": 9.229821318545901e-07, "loss": 0.7583, "step": 6250 }, { "epoch": 1.17, "learning_rate": 9.219552269459847e-07, "loss": 0.7122, "step": 6255 }, { "epoch": 1.17, "learning_rate": 9.209283220373792e-07, "loss": 1.2121, "step": 6260 }, { "epoch": 1.17, "learning_rate": 9.199014171287738e-07, "loss": 1.0367, "step": 6265 }, { "epoch": 1.17, "learning_rate": 9.188745122201683e-07, "loss": 0.7003, "step": 6270 }, { "epoch": 1.17, "learning_rate": 9.178476073115629e-07, "loss": 1.0541, "step": 6275 }, { "epoch": 1.17, "learning_rate": 9.168207024029574e-07, "loss": 0.9233, "step": 6280 }, { "epoch": 1.17, "learning_rate": 9.15793797494352e-07, "loss": 1.0809, "step": 6285 }, { "epoch": 1.17, "learning_rate": 9.147668925857465e-07, "loss": 0.7956, "step": 6290 }, { "epoch": 1.17, "learning_rate": 9.137399876771411e-07, "loss": 0.9461, "step": 6295 }, { "epoch": 1.17, "learning_rate": 9.127130827685356e-07, "loss": 1.3418, "step": 6300 }, { "epoch": 1.17, "learning_rate": 9.116861778599302e-07, "loss": 0.785, "step": 6305 }, { "epoch": 1.18, "learning_rate": 9.106592729513247e-07, "loss": 0.8314, "step": 6310 }, { "epoch": 1.18, "learning_rate": 9.096323680427193e-07, "loss": 0.8523, "step": 6315 }, { "epoch": 1.18, "learning_rate": 9.086054631341138e-07, "loss": 1.278, "step": 6320 }, { "epoch": 1.18, "learning_rate": 9.075785582255084e-07, "loss": 1.2236, "step": 6325 }, { "epoch": 1.18, "learning_rate": 9.065516533169029e-07, "loss": 0.9903, "step": 6330 }, { "epoch": 1.18, "learning_rate": 9.055247484082975e-07, "loss": 1.452, "step": 6335 }, { "epoch": 1.18, "learning_rate": 9.044978434996919e-07, "loss": 1.2451, "step": 6340 }, { "epoch": 1.18, "learning_rate": 9.034709385910864e-07, "loss": 1.02, "step": 6345 }, { "epoch": 1.18, "learning_rate": 9.02444033682481e-07, "loss": 1.0055, "step": 6350 }, { "epoch": 1.18, "learning_rate": 9.014171287738755e-07, "loss": 0.8727, "step": 6355 }, { "epoch": 1.18, "learning_rate": 9.003902238652701e-07, "loss": 0.8138, "step": 6360 }, { "epoch": 1.19, "learning_rate": 8.993633189566646e-07, "loss": 0.8879, "step": 6365 }, { "epoch": 1.19, "learning_rate": 8.983364140480592e-07, "loss": 1.0248, "step": 6370 }, { "epoch": 1.19, "learning_rate": 8.973095091394536e-07, "loss": 0.8572, "step": 6375 }, { "epoch": 1.19, "learning_rate": 8.962826042308482e-07, "loss": 0.7173, "step": 6380 }, { "epoch": 1.19, "learning_rate": 8.952556993222427e-07, "loss": 0.8795, "step": 6385 }, { "epoch": 1.19, "learning_rate": 8.942287944136373e-07, "loss": 1.6804, "step": 6390 }, { "epoch": 1.19, "learning_rate": 8.932018895050318e-07, "loss": 1.1026, "step": 6395 }, { "epoch": 1.19, "learning_rate": 8.921749845964264e-07, "loss": 0.8339, "step": 6400 }, { "epoch": 1.19, "learning_rate": 8.911480796878209e-07, "loss": 1.3447, "step": 6405 }, { "epoch": 1.19, "learning_rate": 8.901211747792155e-07, "loss": 0.9072, "step": 6410 }, { "epoch": 1.19, "learning_rate": 8.8909426987061e-07, "loss": 1.6232, "step": 6415 }, { "epoch": 1.2, "learning_rate": 8.880673649620045e-07, "loss": 0.7329, "step": 6420 }, { "epoch": 1.2, "learning_rate": 8.87040460053399e-07, "loss": 0.8971, "step": 6425 }, { "epoch": 1.2, "learning_rate": 8.860135551447936e-07, "loss": 0.8014, "step": 6430 }, { "epoch": 1.2, "learning_rate": 8.849866502361881e-07, "loss": 0.902, "step": 6435 }, { "epoch": 1.2, "learning_rate": 8.839597453275827e-07, "loss": 0.9787, "step": 6440 }, { "epoch": 1.2, "learning_rate": 8.829328404189772e-07, "loss": 1.3367, "step": 6445 }, { "epoch": 1.2, "learning_rate": 8.819059355103718e-07, "loss": 1.0189, "step": 6450 }, { "epoch": 1.2, "learning_rate": 8.808790306017663e-07, "loss": 0.8373, "step": 6455 }, { "epoch": 1.2, "learning_rate": 8.798521256931608e-07, "loss": 0.8355, "step": 6460 }, { "epoch": 1.2, "learning_rate": 8.788252207845553e-07, "loss": 0.8937, "step": 6465 }, { "epoch": 1.21, "learning_rate": 8.777983158759499e-07, "loss": 1.0145, "step": 6470 }, { "epoch": 1.21, "learning_rate": 8.767714109673444e-07, "loss": 0.8532, "step": 6475 }, { "epoch": 1.21, "learning_rate": 8.75744506058739e-07, "loss": 1.2985, "step": 6480 }, { "epoch": 1.21, "learning_rate": 8.747176011501335e-07, "loss": 0.9596, "step": 6485 }, { "epoch": 1.21, "learning_rate": 8.736906962415281e-07, "loss": 1.1915, "step": 6490 }, { "epoch": 1.21, "learning_rate": 8.726637913329226e-07, "loss": 0.7417, "step": 6495 }, { "epoch": 1.21, "learning_rate": 8.71636886424317e-07, "loss": 0.8899, "step": 6500 }, { "epoch": 1.21, "eval_loss": 0.6119966506958008, "eval_runtime": 161.8006, "eval_samples_per_second": 7.379, "eval_steps_per_second": 0.927, "eval_wer": 0.9730631092868137, "step": 6500 }, { "epoch": 1.21, "learning_rate": 8.706099815157116e-07, "loss": 1.0632, "step": 6505 }, { "epoch": 1.21, "learning_rate": 8.695830766071061e-07, "loss": 0.9989, "step": 6510 }, { "epoch": 1.21, "learning_rate": 8.685561716985007e-07, "loss": 1.0238, "step": 6515 }, { "epoch": 1.21, "learning_rate": 8.675292667898952e-07, "loss": 0.9555, "step": 6520 }, { "epoch": 1.22, "learning_rate": 8.665023618812898e-07, "loss": 0.9623, "step": 6525 }, { "epoch": 1.22, "learning_rate": 8.654754569726843e-07, "loss": 0.6852, "step": 6530 }, { "epoch": 1.22, "learning_rate": 8.644485520640789e-07, "loss": 0.9969, "step": 6535 }, { "epoch": 1.22, "learning_rate": 8.634216471554734e-07, "loss": 0.8759, "step": 6540 }, { "epoch": 1.22, "learning_rate": 8.623947422468679e-07, "loss": 0.972, "step": 6545 }, { "epoch": 1.22, "learning_rate": 8.613678373382624e-07, "loss": 0.7837, "step": 6550 }, { "epoch": 1.22, "learning_rate": 8.60340932429657e-07, "loss": 1.3972, "step": 6555 }, { "epoch": 1.22, "learning_rate": 8.593140275210515e-07, "loss": 0.9475, "step": 6560 }, { "epoch": 1.22, "learning_rate": 8.582871226124461e-07, "loss": 0.9438, "step": 6565 }, { "epoch": 1.22, "learning_rate": 8.572602177038406e-07, "loss": 0.848, "step": 6570 }, { "epoch": 1.22, "learning_rate": 8.562333127952352e-07, "loss": 0.9746, "step": 6575 }, { "epoch": 1.23, "learning_rate": 8.552064078866297e-07, "loss": 0.859, "step": 6580 }, { "epoch": 1.23, "learning_rate": 8.541795029780242e-07, "loss": 1.0748, "step": 6585 }, { "epoch": 1.23, "learning_rate": 8.531525980694187e-07, "loss": 0.8909, "step": 6590 }, { "epoch": 1.23, "learning_rate": 8.521256931608133e-07, "loss": 0.7974, "step": 6595 }, { "epoch": 1.23, "learning_rate": 8.510987882522078e-07, "loss": 0.9178, "step": 6600 }, { "epoch": 1.23, "learning_rate": 8.500718833436024e-07, "loss": 0.8336, "step": 6605 }, { "epoch": 1.23, "learning_rate": 8.490449784349969e-07, "loss": 0.9486, "step": 6610 }, { "epoch": 1.23, "learning_rate": 8.480180735263915e-07, "loss": 0.8703, "step": 6615 }, { "epoch": 1.23, "learning_rate": 8.46991168617786e-07, "loss": 1.1615, "step": 6620 }, { "epoch": 1.23, "learning_rate": 8.459642637091806e-07, "loss": 0.8364, "step": 6625 }, { "epoch": 1.23, "learning_rate": 8.44937358800575e-07, "loss": 0.9219, "step": 6630 }, { "epoch": 1.24, "learning_rate": 8.439104538919696e-07, "loss": 0.8983, "step": 6635 }, { "epoch": 1.24, "learning_rate": 8.428835489833641e-07, "loss": 1.0514, "step": 6640 }, { "epoch": 1.24, "learning_rate": 8.418566440747586e-07, "loss": 0.7211, "step": 6645 }, { "epoch": 1.24, "learning_rate": 8.408297391661532e-07, "loss": 0.9361, "step": 6650 }, { "epoch": 1.24, "learning_rate": 8.398028342575477e-07, "loss": 0.818, "step": 6655 }, { "epoch": 1.24, "learning_rate": 8.387759293489423e-07, "loss": 0.8167, "step": 6660 }, { "epoch": 1.24, "learning_rate": 8.377490244403367e-07, "loss": 0.9332, "step": 6665 }, { "epoch": 1.24, "learning_rate": 8.367221195317313e-07, "loss": 1.3069, "step": 6670 }, { "epoch": 1.24, "learning_rate": 8.356952146231258e-07, "loss": 0.8066, "step": 6675 }, { "epoch": 1.24, "learning_rate": 8.346683097145204e-07, "loss": 0.7699, "step": 6680 }, { "epoch": 1.25, "learning_rate": 8.336414048059149e-07, "loss": 1.1784, "step": 6685 }, { "epoch": 1.25, "learning_rate": 8.326144998973095e-07, "loss": 0.9819, "step": 6690 }, { "epoch": 1.25, "learning_rate": 8.31587594988704e-07, "loss": 0.7979, "step": 6695 }, { "epoch": 1.25, "learning_rate": 8.305606900800986e-07, "loss": 0.8063, "step": 6700 }, { "epoch": 1.25, "learning_rate": 8.29533785171493e-07, "loss": 0.7892, "step": 6705 }, { "epoch": 1.25, "learning_rate": 8.285068802628876e-07, "loss": 0.8912, "step": 6710 }, { "epoch": 1.25, "learning_rate": 8.274799753542821e-07, "loss": 0.9465, "step": 6715 }, { "epoch": 1.25, "learning_rate": 8.264530704456767e-07, "loss": 0.7973, "step": 6720 }, { "epoch": 1.25, "learning_rate": 8.254261655370712e-07, "loss": 0.7016, "step": 6725 }, { "epoch": 1.25, "learning_rate": 8.243992606284658e-07, "loss": 1.0314, "step": 6730 }, { "epoch": 1.25, "learning_rate": 8.233723557198603e-07, "loss": 0.837, "step": 6735 }, { "epoch": 1.26, "learning_rate": 8.223454508112549e-07, "loss": 1.0261, "step": 6740 }, { "epoch": 1.26, "learning_rate": 8.213185459026494e-07, "loss": 0.907, "step": 6745 }, { "epoch": 1.26, "learning_rate": 8.20291640994044e-07, "loss": 1.3548, "step": 6750 }, { "epoch": 1.26, "learning_rate": 8.192647360854384e-07, "loss": 0.6939, "step": 6755 }, { "epoch": 1.26, "learning_rate": 8.18237831176833e-07, "loss": 0.642, "step": 6760 }, { "epoch": 1.26, "learning_rate": 8.172109262682275e-07, "loss": 0.7524, "step": 6765 }, { "epoch": 1.26, "learning_rate": 8.161840213596221e-07, "loss": 1.3072, "step": 6770 }, { "epoch": 1.26, "learning_rate": 8.151571164510166e-07, "loss": 0.9227, "step": 6775 }, { "epoch": 1.26, "learning_rate": 8.141302115424112e-07, "loss": 1.0372, "step": 6780 }, { "epoch": 1.26, "learning_rate": 8.131033066338057e-07, "loss": 1.0262, "step": 6785 }, { "epoch": 1.26, "learning_rate": 8.120764017252003e-07, "loss": 0.7881, "step": 6790 }, { "epoch": 1.27, "learning_rate": 8.110494968165947e-07, "loss": 1.0526, "step": 6795 }, { "epoch": 1.27, "learning_rate": 8.100225919079892e-07, "loss": 1.1853, "step": 6800 }, { "epoch": 1.27, "learning_rate": 8.089956869993838e-07, "loss": 0.7666, "step": 6805 }, { "epoch": 1.27, "learning_rate": 8.079687820907783e-07, "loss": 0.8543, "step": 6810 }, { "epoch": 1.27, "learning_rate": 8.069418771821729e-07, "loss": 0.8792, "step": 6815 }, { "epoch": 1.27, "learning_rate": 8.059149722735674e-07, "loss": 0.8891, "step": 6820 }, { "epoch": 1.27, "learning_rate": 8.04888067364962e-07, "loss": 0.6812, "step": 6825 }, { "epoch": 1.27, "learning_rate": 8.038611624563565e-07, "loss": 0.9066, "step": 6830 }, { "epoch": 1.27, "learning_rate": 8.02834257547751e-07, "loss": 0.951, "step": 6835 }, { "epoch": 1.27, "learning_rate": 8.018073526391455e-07, "loss": 1.108, "step": 6840 }, { "epoch": 1.27, "learning_rate": 8.007804477305401e-07, "loss": 0.7586, "step": 6845 }, { "epoch": 1.28, "learning_rate": 7.997535428219346e-07, "loss": 0.8866, "step": 6850 }, { "epoch": 1.28, "learning_rate": 7.987266379133292e-07, "loss": 0.7629, "step": 6855 }, { "epoch": 1.28, "learning_rate": 7.976997330047237e-07, "loss": 0.8082, "step": 6860 }, { "epoch": 1.28, "learning_rate": 7.966728280961183e-07, "loss": 0.8103, "step": 6865 }, { "epoch": 1.28, "learning_rate": 7.956459231875128e-07, "loss": 0.8515, "step": 6870 }, { "epoch": 1.28, "learning_rate": 7.946190182789073e-07, "loss": 0.971, "step": 6875 }, { "epoch": 1.28, "learning_rate": 7.935921133703018e-07, "loss": 0.8346, "step": 6880 }, { "epoch": 1.28, "learning_rate": 7.925652084616964e-07, "loss": 0.9302, "step": 6885 }, { "epoch": 1.28, "learning_rate": 7.915383035530909e-07, "loss": 0.9843, "step": 6890 }, { "epoch": 1.28, "learning_rate": 7.905113986444855e-07, "loss": 1.0542, "step": 6895 }, { "epoch": 1.29, "learning_rate": 7.8948449373588e-07, "loss": 1.1789, "step": 6900 }, { "epoch": 1.29, "learning_rate": 7.884575888272746e-07, "loss": 0.9403, "step": 6905 }, { "epoch": 1.29, "learning_rate": 7.874306839186691e-07, "loss": 0.9338, "step": 6910 }, { "epoch": 1.29, "learning_rate": 7.864037790100636e-07, "loss": 0.8633, "step": 6915 }, { "epoch": 1.29, "learning_rate": 7.853768741014581e-07, "loss": 0.8621, "step": 6920 }, { "epoch": 1.29, "learning_rate": 7.843499691928527e-07, "loss": 0.8301, "step": 6925 }, { "epoch": 1.29, "learning_rate": 7.833230642842472e-07, "loss": 0.8555, "step": 6930 }, { "epoch": 1.29, "learning_rate": 7.822961593756418e-07, "loss": 0.8075, "step": 6935 }, { "epoch": 1.29, "learning_rate": 7.812692544670363e-07, "loss": 1.0482, "step": 6940 }, { "epoch": 1.29, "learning_rate": 7.802423495584308e-07, "loss": 1.2927, "step": 6945 }, { "epoch": 1.29, "learning_rate": 7.792154446498254e-07, "loss": 0.7703, "step": 6950 }, { "epoch": 1.3, "learning_rate": 7.781885397412198e-07, "loss": 1.0229, "step": 6955 }, { "epoch": 1.3, "learning_rate": 7.771616348326144e-07, "loss": 0.6933, "step": 6960 }, { "epoch": 1.3, "learning_rate": 7.761347299240089e-07, "loss": 0.8685, "step": 6965 }, { "epoch": 1.3, "learning_rate": 7.751078250154035e-07, "loss": 0.9672, "step": 6970 }, { "epoch": 1.3, "learning_rate": 7.74080920106798e-07, "loss": 1.1863, "step": 6975 }, { "epoch": 1.3, "learning_rate": 7.730540151981926e-07, "loss": 0.7385, "step": 6980 }, { "epoch": 1.3, "learning_rate": 7.720271102895871e-07, "loss": 0.9416, "step": 6985 }, { "epoch": 1.3, "learning_rate": 7.710002053809817e-07, "loss": 0.9813, "step": 6990 }, { "epoch": 1.3, "learning_rate": 7.699733004723762e-07, "loss": 1.1032, "step": 6995 }, { "epoch": 1.3, "learning_rate": 7.689463955637707e-07, "loss": 0.9412, "step": 7000 }, { "epoch": 1.3, "eval_loss": 0.5923011898994446, "eval_runtime": 153.2623, "eval_samples_per_second": 7.791, "eval_steps_per_second": 0.979, "eval_wer": 0.952539763981529, "step": 7000 }, { "epoch": 1.3, "learning_rate": 7.679194906551652e-07, "loss": 0.7374, "step": 7005 }, { "epoch": 1.31, "learning_rate": 7.668925857465598e-07, "loss": 0.9419, "step": 7010 }, { "epoch": 1.31, "learning_rate": 7.658656808379543e-07, "loss": 0.8655, "step": 7015 }, { "epoch": 1.31, "learning_rate": 7.648387759293489e-07, "loss": 1.0748, "step": 7020 }, { "epoch": 1.31, "learning_rate": 7.638118710207434e-07, "loss": 0.9836, "step": 7025 }, { "epoch": 1.31, "learning_rate": 7.62784966112138e-07, "loss": 0.7297, "step": 7030 }, { "epoch": 1.31, "learning_rate": 7.617580612035325e-07, "loss": 1.3443, "step": 7035 }, { "epoch": 1.31, "learning_rate": 7.60731156294927e-07, "loss": 0.894, "step": 7040 }, { "epoch": 1.31, "learning_rate": 7.597042513863215e-07, "loss": 1.1128, "step": 7045 }, { "epoch": 1.31, "learning_rate": 7.586773464777161e-07, "loss": 0.7503, "step": 7050 }, { "epoch": 1.31, "learning_rate": 7.576504415691106e-07, "loss": 0.8865, "step": 7055 }, { "epoch": 1.31, "learning_rate": 7.566235366605052e-07, "loss": 0.8906, "step": 7060 }, { "epoch": 1.32, "learning_rate": 7.555966317518997e-07, "loss": 0.8262, "step": 7065 }, { "epoch": 1.32, "learning_rate": 7.545697268432943e-07, "loss": 0.9334, "step": 7070 }, { "epoch": 1.32, "learning_rate": 7.535428219346888e-07, "loss": 0.7228, "step": 7075 }, { "epoch": 1.32, "learning_rate": 7.525159170260834e-07, "loss": 0.6956, "step": 7080 }, { "epoch": 1.32, "learning_rate": 7.514890121174778e-07, "loss": 0.8124, "step": 7085 }, { "epoch": 1.32, "learning_rate": 7.504621072088724e-07, "loss": 1.0577, "step": 7090 }, { "epoch": 1.32, "learning_rate": 7.494352023002669e-07, "loss": 1.0066, "step": 7095 }, { "epoch": 1.32, "learning_rate": 7.484082973916614e-07, "loss": 1.1995, "step": 7100 }, { "epoch": 1.32, "learning_rate": 7.47381392483056e-07, "loss": 0.7794, "step": 7105 }, { "epoch": 1.32, "learning_rate": 7.463544875744505e-07, "loss": 0.9095, "step": 7110 }, { "epoch": 1.33, "learning_rate": 7.453275826658451e-07, "loss": 0.8585, "step": 7115 }, { "epoch": 1.33, "learning_rate": 7.443006777572395e-07, "loss": 1.0227, "step": 7120 }, { "epoch": 1.33, "learning_rate": 7.432737728486341e-07, "loss": 0.7072, "step": 7125 }, { "epoch": 1.33, "learning_rate": 7.422468679400286e-07, "loss": 0.765, "step": 7130 }, { "epoch": 1.33, "learning_rate": 7.412199630314233e-07, "loss": 0.7863, "step": 7135 }, { "epoch": 1.33, "learning_rate": 7.401930581228178e-07, "loss": 0.7918, "step": 7140 }, { "epoch": 1.33, "learning_rate": 7.391661532142124e-07, "loss": 1.0758, "step": 7145 }, { "epoch": 1.33, "learning_rate": 7.381392483056069e-07, "loss": 0.7628, "step": 7150 }, { "epoch": 1.33, "learning_rate": 7.371123433970015e-07, "loss": 0.9962, "step": 7155 }, { "epoch": 1.33, "learning_rate": 7.36085438488396e-07, "loss": 1.0403, "step": 7160 }, { "epoch": 1.33, "learning_rate": 7.350585335797905e-07, "loss": 0.775, "step": 7165 }, { "epoch": 1.34, "learning_rate": 7.34031628671185e-07, "loss": 0.8383, "step": 7170 }, { "epoch": 1.34, "learning_rate": 7.330047237625796e-07, "loss": 0.8059, "step": 7175 }, { "epoch": 1.34, "learning_rate": 7.319778188539741e-07, "loss": 0.715, "step": 7180 }, { "epoch": 1.34, "learning_rate": 7.309509139453687e-07, "loss": 0.7442, "step": 7185 }, { "epoch": 1.34, "learning_rate": 7.299240090367632e-07, "loss": 0.7807, "step": 7190 }, { "epoch": 1.34, "learning_rate": 7.288971041281578e-07, "loss": 0.8463, "step": 7195 }, { "epoch": 1.34, "learning_rate": 7.278701992195523e-07, "loss": 0.8495, "step": 7200 }, { "epoch": 1.34, "learning_rate": 7.268432943109469e-07, "loss": 0.8013, "step": 7205 }, { "epoch": 1.34, "learning_rate": 7.258163894023413e-07, "loss": 1.0864, "step": 7210 }, { "epoch": 1.34, "learning_rate": 7.247894844937359e-07, "loss": 1.1016, "step": 7215 }, { "epoch": 1.34, "learning_rate": 7.237625795851304e-07, "loss": 0.917, "step": 7220 }, { "epoch": 1.35, "learning_rate": 7.22735674676525e-07, "loss": 0.7113, "step": 7225 }, { "epoch": 1.35, "learning_rate": 7.217087697679195e-07, "loss": 1.215, "step": 7230 }, { "epoch": 1.35, "learning_rate": 7.206818648593141e-07, "loss": 0.9089, "step": 7235 }, { "epoch": 1.35, "learning_rate": 7.196549599507086e-07, "loss": 1.0031, "step": 7240 }, { "epoch": 1.35, "learning_rate": 7.186280550421032e-07, "loss": 0.9265, "step": 7245 }, { "epoch": 1.35, "learning_rate": 7.176011501334976e-07, "loss": 0.7035, "step": 7250 }, { "epoch": 1.35, "learning_rate": 7.165742452248922e-07, "loss": 0.8049, "step": 7255 }, { "epoch": 1.35, "learning_rate": 7.155473403162867e-07, "loss": 0.87, "step": 7260 }, { "epoch": 1.35, "learning_rate": 7.145204354076812e-07, "loss": 0.9135, "step": 7265 }, { "epoch": 1.35, "learning_rate": 7.134935304990758e-07, "loss": 0.9293, "step": 7270 }, { "epoch": 1.36, "learning_rate": 7.124666255904703e-07, "loss": 0.8794, "step": 7275 }, { "epoch": 1.36, "learning_rate": 7.114397206818649e-07, "loss": 0.7732, "step": 7280 }, { "epoch": 1.36, "learning_rate": 7.104128157732594e-07, "loss": 0.784, "step": 7285 }, { "epoch": 1.36, "learning_rate": 7.093859108646539e-07, "loss": 1.573, "step": 7290 }, { "epoch": 1.36, "learning_rate": 7.083590059560484e-07, "loss": 0.956, "step": 7295 }, { "epoch": 1.36, "learning_rate": 7.07332101047443e-07, "loss": 1.0869, "step": 7300 }, { "epoch": 1.36, "learning_rate": 7.063051961388375e-07, "loss": 0.8412, "step": 7305 }, { "epoch": 1.36, "learning_rate": 7.052782912302321e-07, "loss": 1.2126, "step": 7310 }, { "epoch": 1.36, "learning_rate": 7.042513863216266e-07, "loss": 0.8477, "step": 7315 }, { "epoch": 1.36, "learning_rate": 7.032244814130212e-07, "loss": 1.3195, "step": 7320 }, { "epoch": 1.36, "learning_rate": 7.021975765044157e-07, "loss": 0.7801, "step": 7325 }, { "epoch": 1.37, "learning_rate": 7.011706715958103e-07, "loss": 0.8952, "step": 7330 }, { "epoch": 1.37, "learning_rate": 7.001437666872047e-07, "loss": 0.9682, "step": 7335 }, { "epoch": 1.37, "learning_rate": 6.991168617785993e-07, "loss": 0.7288, "step": 7340 }, { "epoch": 1.37, "learning_rate": 6.980899568699938e-07, "loss": 0.855, "step": 7345 }, { "epoch": 1.37, "learning_rate": 6.970630519613884e-07, "loss": 0.9398, "step": 7350 }, { "epoch": 1.37, "learning_rate": 6.960361470527829e-07, "loss": 0.9362, "step": 7355 }, { "epoch": 1.37, "learning_rate": 6.950092421441775e-07, "loss": 0.9882, "step": 7360 }, { "epoch": 1.37, "learning_rate": 6.93982337235572e-07, "loss": 1.0757, "step": 7365 }, { "epoch": 1.37, "learning_rate": 6.929554323269666e-07, "loss": 0.8157, "step": 7370 }, { "epoch": 1.37, "learning_rate": 6.91928527418361e-07, "loss": 0.8172, "step": 7375 }, { "epoch": 1.37, "learning_rate": 6.909016225097556e-07, "loss": 0.737, "step": 7380 }, { "epoch": 1.38, "learning_rate": 6.898747176011501e-07, "loss": 0.7614, "step": 7385 }, { "epoch": 1.38, "learning_rate": 6.888478126925447e-07, "loss": 1.1501, "step": 7390 }, { "epoch": 1.38, "learning_rate": 6.878209077839392e-07, "loss": 0.6488, "step": 7395 }, { "epoch": 1.38, "learning_rate": 6.867940028753338e-07, "loss": 1.0218, "step": 7400 }, { "epoch": 1.38, "learning_rate": 6.857670979667283e-07, "loss": 0.7974, "step": 7405 }, { "epoch": 1.38, "learning_rate": 6.847401930581228e-07, "loss": 0.9662, "step": 7410 }, { "epoch": 1.38, "learning_rate": 6.837132881495173e-07, "loss": 0.7973, "step": 7415 }, { "epoch": 1.38, "learning_rate": 6.826863832409118e-07, "loss": 1.1172, "step": 7420 }, { "epoch": 1.38, "learning_rate": 6.816594783323064e-07, "loss": 1.118, "step": 7425 }, { "epoch": 1.38, "learning_rate": 6.806325734237009e-07, "loss": 1.2301, "step": 7430 }, { "epoch": 1.38, "learning_rate": 6.796056685150955e-07, "loss": 1.0008, "step": 7435 }, { "epoch": 1.39, "learning_rate": 6.7857876360649e-07, "loss": 0.9601, "step": 7440 }, { "epoch": 1.39, "learning_rate": 6.775518586978846e-07, "loss": 0.7135, "step": 7445 }, { "epoch": 1.39, "learning_rate": 6.765249537892791e-07, "loss": 0.8211, "step": 7450 }, { "epoch": 1.39, "learning_rate": 6.754980488806736e-07, "loss": 1.1097, "step": 7455 }, { "epoch": 1.39, "learning_rate": 6.744711439720681e-07, "loss": 1.0741, "step": 7460 }, { "epoch": 1.39, "learning_rate": 6.734442390634627e-07, "loss": 1.2393, "step": 7465 }, { "epoch": 1.39, "learning_rate": 6.724173341548572e-07, "loss": 0.8408, "step": 7470 }, { "epoch": 1.39, "learning_rate": 6.713904292462518e-07, "loss": 0.7914, "step": 7475 }, { "epoch": 1.39, "learning_rate": 6.703635243376463e-07, "loss": 0.7918, "step": 7480 }, { "epoch": 1.39, "learning_rate": 6.693366194290409e-07, "loss": 0.9031, "step": 7485 }, { "epoch": 1.4, "learning_rate": 6.683097145204354e-07, "loss": 0.7893, "step": 7490 }, { "epoch": 1.4, "learning_rate": 6.6728280961183e-07, "loss": 0.7198, "step": 7495 }, { "epoch": 1.4, "learning_rate": 6.662559047032244e-07, "loss": 0.8052, "step": 7500 }, { "epoch": 1.4, "eval_loss": 0.5810225009918213, "eval_runtime": 145.8096, "eval_samples_per_second": 8.189, "eval_steps_per_second": 1.029, "eval_wer": 0.9427911749615188, "step": 7500 }, { "epoch": 1.4, "learning_rate": 6.65228999794619e-07, "loss": 1.2298, "step": 7505 }, { "epoch": 1.4, "learning_rate": 6.642020948860135e-07, "loss": 0.7371, "step": 7510 }, { "epoch": 1.4, "learning_rate": 6.631751899774081e-07, "loss": 1.2392, "step": 7515 }, { "epoch": 1.4, "learning_rate": 6.621482850688026e-07, "loss": 0.7397, "step": 7520 }, { "epoch": 1.4, "learning_rate": 6.611213801601972e-07, "loss": 0.8519, "step": 7525 }, { "epoch": 1.4, "learning_rate": 6.600944752515917e-07, "loss": 1.2469, "step": 7530 }, { "epoch": 1.4, "learning_rate": 6.590675703429863e-07, "loss": 0.7008, "step": 7535 }, { "epoch": 1.4, "learning_rate": 6.580406654343807e-07, "loss": 0.8012, "step": 7540 }, { "epoch": 1.41, "learning_rate": 6.570137605257753e-07, "loss": 1.1256, "step": 7545 }, { "epoch": 1.41, "learning_rate": 6.559868556171698e-07, "loss": 0.7622, "step": 7550 }, { "epoch": 1.41, "learning_rate": 6.549599507085644e-07, "loss": 0.8032, "step": 7555 }, { "epoch": 1.41, "learning_rate": 6.539330457999589e-07, "loss": 0.8004, "step": 7560 }, { "epoch": 1.41, "learning_rate": 6.529061408913534e-07, "loss": 0.7468, "step": 7565 }, { "epoch": 1.41, "learning_rate": 6.51879235982748e-07, "loss": 0.8619, "step": 7570 }, { "epoch": 1.41, "learning_rate": 6.508523310741425e-07, "loss": 0.7856, "step": 7575 }, { "epoch": 1.41, "learning_rate": 6.49825426165537e-07, "loss": 0.7295, "step": 7580 }, { "epoch": 1.41, "learning_rate": 6.487985212569315e-07, "loss": 0.9886, "step": 7585 }, { "epoch": 1.41, "learning_rate": 6.477716163483261e-07, "loss": 0.9888, "step": 7590 }, { "epoch": 1.41, "learning_rate": 6.467447114397206e-07, "loss": 0.9019, "step": 7595 }, { "epoch": 1.42, "learning_rate": 6.457178065311152e-07, "loss": 0.7629, "step": 7600 }, { "epoch": 1.42, "learning_rate": 6.446909016225097e-07, "loss": 0.8729, "step": 7605 }, { "epoch": 1.42, "learning_rate": 6.436639967139043e-07, "loss": 0.8466, "step": 7610 }, { "epoch": 1.42, "learning_rate": 6.426370918052988e-07, "loss": 0.7365, "step": 7615 }, { "epoch": 1.42, "learning_rate": 6.416101868966933e-07, "loss": 1.0247, "step": 7620 }, { "epoch": 1.42, "learning_rate": 6.405832819880878e-07, "loss": 1.0193, "step": 7625 }, { "epoch": 1.42, "learning_rate": 6.395563770794824e-07, "loss": 1.3307, "step": 7630 }, { "epoch": 1.42, "learning_rate": 6.385294721708769e-07, "loss": 0.9158, "step": 7635 }, { "epoch": 1.42, "learning_rate": 6.375025672622715e-07, "loss": 1.2267, "step": 7640 }, { "epoch": 1.42, "learning_rate": 6.36475662353666e-07, "loss": 1.0363, "step": 7645 }, { "epoch": 1.42, "learning_rate": 6.354487574450606e-07, "loss": 0.9707, "step": 7650 }, { "epoch": 1.43, "learning_rate": 6.344218525364551e-07, "loss": 0.8471, "step": 7655 }, { "epoch": 1.43, "learning_rate": 6.336003286095707e-07, "loss": 2.2894, "step": 7660 }, { "epoch": 1.43, "learning_rate": 6.325734237009652e-07, "loss": 1.0268, "step": 7665 }, { "epoch": 1.43, "learning_rate": 6.315465187923598e-07, "loss": 0.7682, "step": 7670 }, { "epoch": 1.43, "learning_rate": 6.305196138837543e-07, "loss": 0.6118, "step": 7675 }, { "epoch": 1.43, "learning_rate": 6.294927089751488e-07, "loss": 0.877, "step": 7680 }, { "epoch": 1.43, "learning_rate": 6.284658040665434e-07, "loss": 0.6714, "step": 7685 }, { "epoch": 1.43, "learning_rate": 6.274388991579379e-07, "loss": 1.2271, "step": 7690 }, { "epoch": 1.43, "learning_rate": 6.264119942493324e-07, "loss": 0.7157, "step": 7695 }, { "epoch": 1.43, "learning_rate": 6.253850893407269e-07, "loss": 0.649, "step": 7700 }, { "epoch": 1.44, "learning_rate": 6.243581844321215e-07, "loss": 0.9921, "step": 7705 }, { "epoch": 1.44, "learning_rate": 6.23331279523516e-07, "loss": 0.8691, "step": 7710 }, { "epoch": 1.44, "learning_rate": 6.223043746149106e-07, "loss": 1.0678, "step": 7715 }, { "epoch": 1.44, "learning_rate": 6.212774697063051e-07, "loss": 0.7666, "step": 7720 }, { "epoch": 1.44, "learning_rate": 6.202505647976997e-07, "loss": 0.8856, "step": 7725 }, { "epoch": 1.44, "learning_rate": 6.192236598890942e-07, "loss": 1.169, "step": 7730 }, { "epoch": 1.44, "learning_rate": 6.181967549804888e-07, "loss": 0.8315, "step": 7735 }, { "epoch": 1.44, "learning_rate": 6.171698500718832e-07, "loss": 0.9041, "step": 7740 }, { "epoch": 1.44, "learning_rate": 6.161429451632778e-07, "loss": 0.7837, "step": 7745 }, { "epoch": 1.44, "learning_rate": 6.151160402546723e-07, "loss": 0.8858, "step": 7750 }, { "epoch": 1.44, "learning_rate": 6.140891353460669e-07, "loss": 0.9944, "step": 7755 }, { "epoch": 1.45, "learning_rate": 6.130622304374614e-07, "loss": 1.438, "step": 7760 }, { "epoch": 1.45, "learning_rate": 6.12035325528856e-07, "loss": 0.7522, "step": 7765 }, { "epoch": 1.45, "learning_rate": 6.110084206202505e-07, "loss": 1.161, "step": 7770 }, { "epoch": 1.45, "learning_rate": 6.099815157116451e-07, "loss": 0.8066, "step": 7775 }, { "epoch": 1.45, "learning_rate": 6.089546108030395e-07, "loss": 0.8732, "step": 7780 }, { "epoch": 1.45, "learning_rate": 6.079277058944341e-07, "loss": 0.8681, "step": 7785 }, { "epoch": 1.45, "learning_rate": 6.069008009858286e-07, "loss": 0.6792, "step": 7790 }, { "epoch": 1.45, "learning_rate": 6.058738960772232e-07, "loss": 0.7122, "step": 7795 }, { "epoch": 1.45, "learning_rate": 6.048469911686177e-07, "loss": 1.1585, "step": 7800 }, { "epoch": 1.45, "learning_rate": 6.038200862600123e-07, "loss": 0.8072, "step": 7805 }, { "epoch": 1.45, "learning_rate": 6.027931813514069e-07, "loss": 0.9174, "step": 7810 }, { "epoch": 1.46, "learning_rate": 6.017662764428015e-07, "loss": 0.9088, "step": 7815 }, { "epoch": 1.46, "learning_rate": 6.00739371534196e-07, "loss": 0.817, "step": 7820 }, { "epoch": 1.46, "learning_rate": 5.997124666255905e-07, "loss": 0.849, "step": 7825 }, { "epoch": 1.46, "learning_rate": 5.98685561716985e-07, "loss": 0.9686, "step": 7830 }, { "epoch": 1.46, "learning_rate": 5.976586568083796e-07, "loss": 0.9308, "step": 7835 }, { "epoch": 1.46, "learning_rate": 5.966317518997741e-07, "loss": 0.8584, "step": 7840 }, { "epoch": 1.46, "learning_rate": 5.956048469911686e-07, "loss": 0.8181, "step": 7845 }, { "epoch": 1.46, "learning_rate": 5.945779420825632e-07, "loss": 0.7327, "step": 7850 }, { "epoch": 1.46, "learning_rate": 5.935510371739577e-07, "loss": 0.6989, "step": 7855 }, { "epoch": 1.46, "learning_rate": 5.925241322653523e-07, "loss": 0.7766, "step": 7860 }, { "epoch": 1.46, "learning_rate": 5.914972273567467e-07, "loss": 1.4259, "step": 7865 }, { "epoch": 1.47, "learning_rate": 5.904703224481413e-07, "loss": 0.9504, "step": 7870 }, { "epoch": 1.47, "learning_rate": 5.894434175395358e-07, "loss": 0.8883, "step": 7875 }, { "epoch": 1.47, "learning_rate": 5.884165126309304e-07, "loss": 0.8198, "step": 7880 }, { "epoch": 1.47, "learning_rate": 5.873896077223249e-07, "loss": 1.1042, "step": 7885 }, { "epoch": 1.47, "learning_rate": 5.863627028137195e-07, "loss": 0.761, "step": 7890 }, { "epoch": 1.47, "learning_rate": 5.85335797905114e-07, "loss": 0.7181, "step": 7895 }, { "epoch": 1.47, "learning_rate": 5.843088929965086e-07, "loss": 0.7321, "step": 7900 }, { "epoch": 1.47, "learning_rate": 5.83281988087903e-07, "loss": 0.8182, "step": 7905 }, { "epoch": 1.47, "learning_rate": 5.822550831792976e-07, "loss": 0.8529, "step": 7910 }, { "epoch": 1.47, "learning_rate": 5.812281782706921e-07, "loss": 0.9153, "step": 7915 }, { "epoch": 1.48, "learning_rate": 5.802012733620867e-07, "loss": 1.3715, "step": 7920 }, { "epoch": 1.48, "learning_rate": 5.791743684534812e-07, "loss": 0.7765, "step": 7925 }, { "epoch": 1.48, "learning_rate": 5.781474635448758e-07, "loss": 0.7273, "step": 7930 }, { "epoch": 1.48, "learning_rate": 5.771205586362703e-07, "loss": 1.0647, "step": 7935 }, { "epoch": 1.48, "learning_rate": 5.760936537276649e-07, "loss": 0.7211, "step": 7940 }, { "epoch": 1.48, "learning_rate": 5.750667488190593e-07, "loss": 1.1039, "step": 7945 }, { "epoch": 1.48, "learning_rate": 5.740398439104539e-07, "loss": 0.9831, "step": 7950 }, { "epoch": 1.48, "learning_rate": 5.730129390018484e-07, "loss": 0.7514, "step": 7955 }, { "epoch": 1.48, "learning_rate": 5.71986034093243e-07, "loss": 0.9246, "step": 7960 }, { "epoch": 1.48, "learning_rate": 5.709591291846375e-07, "loss": 1.0028, "step": 7965 }, { "epoch": 1.48, "learning_rate": 5.699322242760321e-07, "loss": 0.9095, "step": 7970 }, { "epoch": 1.49, "learning_rate": 5.689053193674266e-07, "loss": 0.7569, "step": 7975 }, { "epoch": 1.49, "learning_rate": 5.678784144588212e-07, "loss": 0.6962, "step": 7980 }, { "epoch": 1.49, "learning_rate": 5.668515095502157e-07, "loss": 0.8134, "step": 7985 }, { "epoch": 1.49, "learning_rate": 5.658246046416102e-07, "loss": 1.0712, "step": 7990 }, { "epoch": 1.49, "learning_rate": 5.647976997330047e-07, "loss": 0.6945, "step": 7995 }, { "epoch": 1.49, "learning_rate": 5.637707948243992e-07, "loss": 0.9881, "step": 8000 }, { "epoch": 1.49, "eval_loss": 0.5700684189796448, "eval_runtime": 138.9442, "eval_samples_per_second": 8.593, "eval_steps_per_second": 1.08, "eval_wer": 0.9295792714212416, "step": 8000 }, { "epoch": 1.49, "learning_rate": 5.627438899157938e-07, "loss": 1.0634, "step": 8005 }, { "epoch": 1.49, "learning_rate": 5.617169850071883e-07, "loss": 0.7482, "step": 8010 }, { "epoch": 1.49, "learning_rate": 5.606900800985829e-07, "loss": 0.8581, "step": 8015 }, { "epoch": 1.49, "learning_rate": 5.596631751899774e-07, "loss": 1.0777, "step": 8020 }, { "epoch": 1.49, "learning_rate": 5.58636270281372e-07, "loss": 0.75, "step": 8025 }, { "epoch": 1.5, "learning_rate": 5.576093653727664e-07, "loss": 0.8893, "step": 8030 }, { "epoch": 1.5, "learning_rate": 5.56582460464161e-07, "loss": 0.7469, "step": 8035 }, { "epoch": 1.5, "learning_rate": 5.555555555555555e-07, "loss": 0.7758, "step": 8040 }, { "epoch": 1.5, "learning_rate": 5.545286506469501e-07, "loss": 1.1137, "step": 8045 }, { "epoch": 1.5, "learning_rate": 5.535017457383446e-07, "loss": 0.8816, "step": 8050 }, { "epoch": 1.5, "learning_rate": 5.524748408297392e-07, "loss": 0.8427, "step": 8055 }, { "epoch": 1.5, "learning_rate": 5.514479359211337e-07, "loss": 0.8506, "step": 8060 }, { "epoch": 1.5, "learning_rate": 5.504210310125283e-07, "loss": 0.9094, "step": 8065 }, { "epoch": 1.5, "learning_rate": 5.493941261039227e-07, "loss": 0.9327, "step": 8070 }, { "epoch": 1.5, "learning_rate": 5.483672211953173e-07, "loss": 0.7259, "step": 8075 }, { "epoch": 1.5, "learning_rate": 5.473403162867118e-07, "loss": 0.8508, "step": 8080 }, { "epoch": 1.51, "learning_rate": 5.463134113781064e-07, "loss": 0.9925, "step": 8085 }, { "epoch": 1.51, "learning_rate": 5.452865064695009e-07, "loss": 0.855, "step": 8090 }, { "epoch": 1.51, "learning_rate": 5.442596015608955e-07, "loss": 0.7268, "step": 8095 }, { "epoch": 1.51, "learning_rate": 5.4323269665229e-07, "loss": 0.7489, "step": 8100 }, { "epoch": 1.51, "learning_rate": 5.422057917436846e-07, "loss": 0.8293, "step": 8105 }, { "epoch": 1.51, "learning_rate": 5.41178886835079e-07, "loss": 0.8503, "step": 8110 }, { "epoch": 1.51, "learning_rate": 5.401519819264736e-07, "loss": 1.0528, "step": 8115 }, { "epoch": 1.51, "learning_rate": 5.391250770178681e-07, "loss": 1.0937, "step": 8120 }, { "epoch": 1.51, "learning_rate": 5.380981721092627e-07, "loss": 0.733, "step": 8125 }, { "epoch": 1.51, "learning_rate": 5.370712672006572e-07, "loss": 1.0224, "step": 8130 }, { "epoch": 1.52, "learning_rate": 5.360443622920518e-07, "loss": 0.8198, "step": 8135 }, { "epoch": 1.52, "learning_rate": 5.350174573834463e-07, "loss": 0.7389, "step": 8140 }, { "epoch": 1.52, "learning_rate": 5.339905524748408e-07, "loss": 0.9084, "step": 8145 }, { "epoch": 1.52, "learning_rate": 5.329636475662354e-07, "loss": 0.661, "step": 8150 }, { "epoch": 1.52, "learning_rate": 5.319367426576298e-07, "loss": 0.7891, "step": 8155 }, { "epoch": 1.52, "learning_rate": 5.309098377490244e-07, "loss": 0.8624, "step": 8160 }, { "epoch": 1.52, "learning_rate": 5.298829328404189e-07, "loss": 0.8217, "step": 8165 }, { "epoch": 1.52, "learning_rate": 5.288560279318135e-07, "loss": 0.7959, "step": 8170 }, { "epoch": 1.52, "learning_rate": 5.27829123023208e-07, "loss": 1.0279, "step": 8175 }, { "epoch": 1.52, "learning_rate": 5.268022181146026e-07, "loss": 0.7266, "step": 8180 }, { "epoch": 1.52, "learning_rate": 5.257753132059971e-07, "loss": 0.9282, "step": 8185 }, { "epoch": 1.53, "learning_rate": 5.247484082973917e-07, "loss": 0.8392, "step": 8190 }, { "epoch": 1.53, "learning_rate": 5.237215033887861e-07, "loss": 0.8131, "step": 8195 }, { "epoch": 1.53, "learning_rate": 5.226945984801807e-07, "loss": 0.8858, "step": 8200 }, { "epoch": 1.53, "learning_rate": 5.216676935715752e-07, "loss": 0.786, "step": 8205 }, { "epoch": 1.53, "learning_rate": 5.206407886629698e-07, "loss": 0.918, "step": 8210 }, { "epoch": 1.53, "learning_rate": 5.196138837543643e-07, "loss": 0.7505, "step": 8215 }, { "epoch": 1.53, "learning_rate": 5.185869788457589e-07, "loss": 0.7523, "step": 8220 }, { "epoch": 1.53, "learning_rate": 5.175600739371534e-07, "loss": 0.9177, "step": 8225 }, { "epoch": 1.53, "learning_rate": 5.16533169028548e-07, "loss": 0.9611, "step": 8230 }, { "epoch": 1.53, "learning_rate": 5.155062641199424e-07, "loss": 0.8993, "step": 8235 }, { "epoch": 1.53, "learning_rate": 5.14479359211337e-07, "loss": 0.9785, "step": 8240 }, { "epoch": 1.54, "learning_rate": 5.134524543027315e-07, "loss": 0.9, "step": 8245 }, { "epoch": 1.54, "learning_rate": 5.124255493941261e-07, "loss": 0.7835, "step": 8250 }, { "epoch": 1.54, "learning_rate": 5.113986444855206e-07, "loss": 1.0824, "step": 8255 }, { "epoch": 1.54, "learning_rate": 5.103717395769152e-07, "loss": 1.0807, "step": 8260 }, { "epoch": 1.54, "learning_rate": 5.093448346683097e-07, "loss": 0.9891, "step": 8265 }, { "epoch": 1.54, "learning_rate": 5.083179297597043e-07, "loss": 0.8318, "step": 8270 }, { "epoch": 1.54, "learning_rate": 5.072910248510988e-07, "loss": 0.7792, "step": 8275 }, { "epoch": 1.54, "learning_rate": 5.062641199424933e-07, "loss": 0.8312, "step": 8280 }, { "epoch": 1.54, "learning_rate": 5.052372150338878e-07, "loss": 0.6618, "step": 8285 }, { "epoch": 1.54, "learning_rate": 5.042103101252824e-07, "loss": 0.7102, "step": 8290 }, { "epoch": 1.54, "learning_rate": 5.031834052166769e-07, "loss": 0.9409, "step": 8295 }, { "epoch": 1.55, "learning_rate": 5.021565003080714e-07, "loss": 1.1354, "step": 8300 }, { "epoch": 1.55, "learning_rate": 5.01129595399466e-07, "loss": 0.7533, "step": 8305 }, { "epoch": 1.55, "learning_rate": 5.001026904908605e-07, "loss": 1.0336, "step": 8310 }, { "epoch": 1.55, "learning_rate": 4.990757855822551e-07, "loss": 0.8046, "step": 8315 }, { "epoch": 1.55, "learning_rate": 4.980488806736495e-07, "loss": 1.3911, "step": 8320 }, { "epoch": 1.55, "learning_rate": 4.970219757650441e-07, "loss": 0.7045, "step": 8325 }, { "epoch": 1.55, "learning_rate": 4.959950708564386e-07, "loss": 0.8076, "step": 8330 }, { "epoch": 1.55, "learning_rate": 4.949681659478332e-07, "loss": 0.9713, "step": 8335 }, { "epoch": 1.55, "learning_rate": 4.939412610392277e-07, "loss": 0.8131, "step": 8340 }, { "epoch": 1.55, "learning_rate": 4.929143561306223e-07, "loss": 0.683, "step": 8345 }, { "epoch": 1.56, "learning_rate": 4.918874512220168e-07, "loss": 1.0491, "step": 8350 }, { "epoch": 1.56, "learning_rate": 4.908605463134114e-07, "loss": 0.9126, "step": 8355 }, { "epoch": 1.56, "learning_rate": 4.898336414048058e-07, "loss": 0.7582, "step": 8360 }, { "epoch": 1.56, "learning_rate": 4.888067364962004e-07, "loss": 1.0113, "step": 8365 }, { "epoch": 1.56, "learning_rate": 4.877798315875949e-07, "loss": 0.7116, "step": 8370 }, { "epoch": 1.56, "learning_rate": 4.867529266789895e-07, "loss": 0.9096, "step": 8375 }, { "epoch": 1.56, "learning_rate": 4.85726021770384e-07, "loss": 0.795, "step": 8380 }, { "epoch": 1.56, "learning_rate": 4.846991168617786e-07, "loss": 0.6889, "step": 8385 }, { "epoch": 1.56, "learning_rate": 4.836722119531731e-07, "loss": 1.0997, "step": 8390 }, { "epoch": 1.56, "learning_rate": 4.826453070445677e-07, "loss": 0.7957, "step": 8395 }, { "epoch": 1.56, "learning_rate": 4.816184021359621e-07, "loss": 0.8872, "step": 8400 }, { "epoch": 1.57, "learning_rate": 4.805914972273567e-07, "loss": 0.8239, "step": 8405 }, { "epoch": 1.57, "learning_rate": 4.795645923187512e-07, "loss": 0.7891, "step": 8410 }, { "epoch": 1.57, "learning_rate": 4.785376874101458e-07, "loss": 1.098, "step": 8415 }, { "epoch": 1.57, "learning_rate": 4.775107825015403e-07, "loss": 0.9928, "step": 8420 }, { "epoch": 1.57, "learning_rate": 4.7648387759293484e-07, "loss": 0.7261, "step": 8425 }, { "epoch": 1.57, "learning_rate": 4.754569726843294e-07, "loss": 0.9117, "step": 8430 }, { "epoch": 1.57, "learning_rate": 4.744300677757239e-07, "loss": 0.7395, "step": 8435 }, { "epoch": 1.57, "learning_rate": 4.734031628671185e-07, "loss": 0.8122, "step": 8440 }, { "epoch": 1.57, "learning_rate": 4.7237625795851304e-07, "loss": 0.6871, "step": 8445 }, { "epoch": 1.57, "learning_rate": 4.713493530499076e-07, "loss": 0.8141, "step": 8450 }, { "epoch": 1.57, "learning_rate": 4.703224481413021e-07, "loss": 0.7518, "step": 8455 }, { "epoch": 1.58, "learning_rate": 4.6929554323269666e-07, "loss": 0.9715, "step": 8460 }, { "epoch": 1.58, "learning_rate": 4.682686383240912e-07, "loss": 0.9768, "step": 8465 }, { "epoch": 1.58, "learning_rate": 4.6724173341548573e-07, "loss": 0.7381, "step": 8470 }, { "epoch": 1.58, "learning_rate": 4.6621482850688027e-07, "loss": 1.0727, "step": 8475 }, { "epoch": 1.58, "learning_rate": 4.651879235982748e-07, "loss": 0.9019, "step": 8480 }, { "epoch": 1.58, "learning_rate": 4.6416101868966935e-07, "loss": 1.1234, "step": 8485 }, { "epoch": 1.58, "learning_rate": 4.631341137810639e-07, "loss": 0.9862, "step": 8490 }, { "epoch": 1.58, "learning_rate": 4.621072088724584e-07, "loss": 0.9288, "step": 8495 }, { "epoch": 1.58, "learning_rate": 4.6108030396385296e-07, "loss": 0.7656, "step": 8500 }, { "epoch": 1.58, "eval_loss": 0.5572460889816284, "eval_runtime": 135.2315, "eval_samples_per_second": 8.829, "eval_steps_per_second": 1.109, "eval_wer": 0.9211133914828117, "step": 8500 }, { "epoch": 1.58, "learning_rate": 4.600533990552475e-07, "loss": 1.2769, "step": 8505 }, { "epoch": 1.59, "learning_rate": 4.5902649414664204e-07, "loss": 0.7621, "step": 8510 }, { "epoch": 1.59, "learning_rate": 4.579995892380366e-07, "loss": 0.8506, "step": 8515 }, { "epoch": 1.59, "learning_rate": 4.569726843294311e-07, "loss": 0.8792, "step": 8520 }, { "epoch": 1.59, "learning_rate": 4.5594577942082565e-07, "loss": 0.8107, "step": 8525 }, { "epoch": 1.59, "learning_rate": 4.5491887451222014e-07, "loss": 0.7621, "step": 8530 }, { "epoch": 1.59, "learning_rate": 4.538919696036147e-07, "loss": 0.7552, "step": 8535 }, { "epoch": 1.59, "learning_rate": 4.528650646950092e-07, "loss": 0.6921, "step": 8540 }, { "epoch": 1.59, "learning_rate": 4.5183815978640375e-07, "loss": 0.614, "step": 8545 }, { "epoch": 1.59, "learning_rate": 4.508112548777983e-07, "loss": 0.7925, "step": 8550 }, { "epoch": 1.59, "learning_rate": 4.4978434996919283e-07, "loss": 0.8397, "step": 8555 }, { "epoch": 1.59, "learning_rate": 4.4875744506058737e-07, "loss": 0.9498, "step": 8560 }, { "epoch": 1.6, "learning_rate": 4.477305401519819e-07, "loss": 0.7446, "step": 8565 }, { "epoch": 1.6, "learning_rate": 4.4670363524337644e-07, "loss": 0.9555, "step": 8570 }, { "epoch": 1.6, "learning_rate": 4.45676730334771e-07, "loss": 0.8588, "step": 8575 }, { "epoch": 1.6, "learning_rate": 4.446498254261655e-07, "loss": 0.9522, "step": 8580 }, { "epoch": 1.6, "learning_rate": 4.4362292051756006e-07, "loss": 0.716, "step": 8585 }, { "epoch": 1.6, "learning_rate": 4.425960156089546e-07, "loss": 0.7492, "step": 8590 }, { "epoch": 1.6, "learning_rate": 4.4156911070034913e-07, "loss": 0.9591, "step": 8595 }, { "epoch": 1.6, "learning_rate": 4.4054220579174367e-07, "loss": 0.6838, "step": 8600 }, { "epoch": 1.6, "learning_rate": 4.395153008831382e-07, "loss": 0.6971, "step": 8605 }, { "epoch": 1.6, "learning_rate": 4.3848839597453275e-07, "loss": 0.7055, "step": 8610 }, { "epoch": 1.6, "learning_rate": 4.374614910659273e-07, "loss": 0.7494, "step": 8615 }, { "epoch": 1.61, "learning_rate": 4.364345861573218e-07, "loss": 0.7243, "step": 8620 }, { "epoch": 1.61, "learning_rate": 4.3540768124871636e-07, "loss": 0.8597, "step": 8625 }, { "epoch": 1.61, "learning_rate": 4.343807763401109e-07, "loss": 0.739, "step": 8630 }, { "epoch": 1.61, "learning_rate": 4.3335387143150544e-07, "loss": 0.838, "step": 8635 }, { "epoch": 1.61, "learning_rate": 4.323269665229e-07, "loss": 0.8557, "step": 8640 }, { "epoch": 1.61, "learning_rate": 4.313000616142945e-07, "loss": 0.6209, "step": 8645 }, { "epoch": 1.61, "learning_rate": 4.3027315670568905e-07, "loss": 0.7521, "step": 8650 }, { "epoch": 1.61, "learning_rate": 4.292462517970836e-07, "loss": 0.6777, "step": 8655 }, { "epoch": 1.61, "learning_rate": 4.282193468884781e-07, "loss": 0.7345, "step": 8660 }, { "epoch": 1.61, "learning_rate": 4.2719244197987266e-07, "loss": 0.7864, "step": 8665 }, { "epoch": 1.61, "learning_rate": 4.261655370712672e-07, "loss": 0.6808, "step": 8670 }, { "epoch": 1.62, "learning_rate": 4.2513863216266174e-07, "loss": 0.7143, "step": 8675 }, { "epoch": 1.62, "learning_rate": 4.241117272540562e-07, "loss": 1.0389, "step": 8680 }, { "epoch": 1.62, "learning_rate": 4.2308482234545076e-07, "loss": 0.7213, "step": 8685 }, { "epoch": 1.62, "learning_rate": 4.220579174368453e-07, "loss": 0.789, "step": 8690 }, { "epoch": 1.62, "learning_rate": 4.2103101252823984e-07, "loss": 0.6796, "step": 8695 }, { "epoch": 1.62, "learning_rate": 4.200041076196344e-07, "loss": 0.7828, "step": 8700 }, { "epoch": 1.62, "learning_rate": 4.189772027110289e-07, "loss": 0.7098, "step": 8705 }, { "epoch": 1.62, "learning_rate": 4.1795029780242345e-07, "loss": 0.7801, "step": 8710 }, { "epoch": 1.62, "learning_rate": 4.16923392893818e-07, "loss": 0.8762, "step": 8715 }, { "epoch": 1.62, "learning_rate": 4.1589648798521253e-07, "loss": 1.2351, "step": 8720 }, { "epoch": 1.63, "learning_rate": 4.1486958307660707e-07, "loss": 0.8635, "step": 8725 }, { "epoch": 1.63, "learning_rate": 4.138426781680016e-07, "loss": 0.7581, "step": 8730 }, { "epoch": 1.63, "learning_rate": 4.1281577325939614e-07, "loss": 0.7198, "step": 8735 }, { "epoch": 1.63, "learning_rate": 4.117888683507907e-07, "loss": 0.7464, "step": 8740 }, { "epoch": 1.63, "learning_rate": 4.107619634421852e-07, "loss": 0.7753, "step": 8745 }, { "epoch": 1.63, "learning_rate": 4.0973505853357976e-07, "loss": 0.7831, "step": 8750 }, { "epoch": 1.63, "learning_rate": 4.087081536249743e-07, "loss": 0.8637, "step": 8755 }, { "epoch": 1.63, "learning_rate": 4.0768124871636883e-07, "loss": 0.7104, "step": 8760 }, { "epoch": 1.63, "learning_rate": 4.0665434380776337e-07, "loss": 1.1407, "step": 8765 }, { "epoch": 1.63, "learning_rate": 4.056274388991579e-07, "loss": 1.8111, "step": 8770 }, { "epoch": 1.63, "learning_rate": 4.0460053399055245e-07, "loss": 0.9607, "step": 8775 }, { "epoch": 1.64, "learning_rate": 4.03573629081947e-07, "loss": 0.7323, "step": 8780 }, { "epoch": 1.64, "learning_rate": 4.025467241733415e-07, "loss": 0.7853, "step": 8785 }, { "epoch": 1.64, "learning_rate": 4.0151981926473606e-07, "loss": 1.0725, "step": 8790 }, { "epoch": 1.64, "learning_rate": 4.004929143561306e-07, "loss": 1.0457, "step": 8795 }, { "epoch": 1.64, "learning_rate": 3.9946600944752514e-07, "loss": 0.8052, "step": 8800 }, { "epoch": 1.64, "learning_rate": 3.984391045389197e-07, "loss": 0.9338, "step": 8805 }, { "epoch": 1.64, "learning_rate": 3.974121996303142e-07, "loss": 1.0658, "step": 8810 }, { "epoch": 1.64, "learning_rate": 3.9638529472170875e-07, "loss": 1.1294, "step": 8815 }, { "epoch": 1.64, "learning_rate": 3.953583898131033e-07, "loss": 0.8306, "step": 8820 }, { "epoch": 1.64, "learning_rate": 3.9433148490449783e-07, "loss": 0.9354, "step": 8825 }, { "epoch": 1.64, "learning_rate": 3.933045799958923e-07, "loss": 0.901, "step": 8830 }, { "epoch": 1.65, "learning_rate": 3.9227767508728685e-07, "loss": 0.888, "step": 8835 }, { "epoch": 1.65, "learning_rate": 3.912507701786814e-07, "loss": 0.8643, "step": 8840 }, { "epoch": 1.65, "learning_rate": 3.9022386527007593e-07, "loss": 0.8769, "step": 8845 }, { "epoch": 1.65, "learning_rate": 3.8919696036147046e-07, "loss": 0.7751, "step": 8850 }, { "epoch": 1.65, "learning_rate": 3.88170055452865e-07, "loss": 0.9784, "step": 8855 }, { "epoch": 1.65, "learning_rate": 3.871431505442596e-07, "loss": 0.6873, "step": 8860 }, { "epoch": 1.65, "learning_rate": 3.8611624563565413e-07, "loss": 0.8287, "step": 8865 }, { "epoch": 1.65, "learning_rate": 3.8508934072704867e-07, "loss": 1.0599, "step": 8870 }, { "epoch": 1.65, "learning_rate": 3.840624358184432e-07, "loss": 0.7779, "step": 8875 }, { "epoch": 1.65, "learning_rate": 3.8303553090983775e-07, "loss": 0.8061, "step": 8880 }, { "epoch": 1.65, "learning_rate": 3.820086260012323e-07, "loss": 0.7412, "step": 8885 }, { "epoch": 1.66, "learning_rate": 3.809817210926268e-07, "loss": 0.9013, "step": 8890 }, { "epoch": 1.66, "learning_rate": 3.7995481618402136e-07, "loss": 1.071, "step": 8895 }, { "epoch": 1.66, "learning_rate": 3.789279112754159e-07, "loss": 1.1361, "step": 8900 }, { "epoch": 1.66, "learning_rate": 3.7790100636681044e-07, "loss": 0.7694, "step": 8905 }, { "epoch": 1.66, "learning_rate": 3.76874101458205e-07, "loss": 0.9759, "step": 8910 }, { "epoch": 1.66, "learning_rate": 3.758471965495995e-07, "loss": 0.9244, "step": 8915 }, { "epoch": 1.66, "learning_rate": 3.7482029164099405e-07, "loss": 0.6793, "step": 8920 }, { "epoch": 1.66, "learning_rate": 3.737933867323886e-07, "loss": 0.7158, "step": 8925 }, { "epoch": 1.66, "learning_rate": 3.727664818237831e-07, "loss": 0.8618, "step": 8930 }, { "epoch": 1.66, "learning_rate": 3.7173957691517766e-07, "loss": 0.7628, "step": 8935 }, { "epoch": 1.67, "learning_rate": 3.707126720065722e-07, "loss": 0.7349, "step": 8940 }, { "epoch": 1.67, "learning_rate": 3.6968576709796674e-07, "loss": 0.7003, "step": 8945 }, { "epoch": 1.67, "learning_rate": 3.686588621893613e-07, "loss": 0.6532, "step": 8950 }, { "epoch": 1.67, "learning_rate": 3.676319572807558e-07, "loss": 0.9775, "step": 8955 }, { "epoch": 1.67, "learning_rate": 3.6660505237215035e-07, "loss": 0.8026, "step": 8960 }, { "epoch": 1.67, "learning_rate": 3.655781474635449e-07, "loss": 0.8922, "step": 8965 }, { "epoch": 1.67, "learning_rate": 3.6455124255493943e-07, "loss": 0.9833, "step": 8970 }, { "epoch": 1.67, "learning_rate": 3.6352433764633397e-07, "loss": 0.9331, "step": 8975 }, { "epoch": 1.67, "learning_rate": 3.624974327377285e-07, "loss": 0.6908, "step": 8980 }, { "epoch": 1.67, "learning_rate": 3.6147052782912304e-07, "loss": 0.7091, "step": 8985 }, { "epoch": 1.67, "learning_rate": 3.6044362292051753e-07, "loss": 0.9287, "step": 8990 }, { "epoch": 1.68, "learning_rate": 3.5941671801191207e-07, "loss": 0.9938, "step": 8995 }, { "epoch": 1.68, "learning_rate": 3.583898131033066e-07, "loss": 1.0478, "step": 9000 }, { "epoch": 1.68, "eval_loss": 0.548711895942688, "eval_runtime": 135.1827, "eval_samples_per_second": 8.832, "eval_steps_per_second": 1.11, "eval_wer": 0.9168804515135968, "step": 9000 }, { "epoch": 1.68, "learning_rate": 3.5736290819470114e-07, "loss": 0.8037, "step": 9005 }, { "epoch": 1.68, "learning_rate": 3.563360032860957e-07, "loss": 0.7756, "step": 9010 }, { "epoch": 1.68, "learning_rate": 3.553090983774902e-07, "loss": 0.9559, "step": 9015 }, { "epoch": 1.68, "learning_rate": 3.5428219346888476e-07, "loss": 0.979, "step": 9020 }, { "epoch": 1.68, "learning_rate": 3.532552885602793e-07, "loss": 0.5972, "step": 9025 }, { "epoch": 1.68, "learning_rate": 3.5222838365167383e-07, "loss": 0.7141, "step": 9030 }, { "epoch": 1.68, "learning_rate": 3.5120147874306837e-07, "loss": 1.1511, "step": 9035 }, { "epoch": 1.68, "learning_rate": 3.501745738344629e-07, "loss": 1.0922, "step": 9040 }, { "epoch": 1.68, "learning_rate": 3.4914766892585745e-07, "loss": 0.9412, "step": 9045 }, { "epoch": 1.69, "learning_rate": 3.48120764017252e-07, "loss": 0.9679, "step": 9050 }, { "epoch": 1.69, "learning_rate": 3.470938591086465e-07, "loss": 0.7127, "step": 9055 }, { "epoch": 1.69, "learning_rate": 3.4606695420004106e-07, "loss": 0.7559, "step": 9060 }, { "epoch": 1.69, "learning_rate": 3.450400492914356e-07, "loss": 0.894, "step": 9065 }, { "epoch": 1.69, "learning_rate": 3.4401314438283014e-07, "loss": 0.8835, "step": 9070 }, { "epoch": 1.69, "learning_rate": 3.429862394742247e-07, "loss": 0.9195, "step": 9075 }, { "epoch": 1.69, "learning_rate": 3.419593345656192e-07, "loss": 0.7117, "step": 9080 }, { "epoch": 1.69, "learning_rate": 3.4093242965701375e-07, "loss": 1.1162, "step": 9085 }, { "epoch": 1.69, "learning_rate": 3.399055247484083e-07, "loss": 0.7655, "step": 9090 }, { "epoch": 1.69, "learning_rate": 3.3887861983980283e-07, "loss": 0.9353, "step": 9095 }, { "epoch": 1.69, "learning_rate": 3.3785171493119737e-07, "loss": 0.8058, "step": 9100 }, { "epoch": 1.7, "learning_rate": 3.368248100225919e-07, "loss": 0.7053, "step": 9105 }, { "epoch": 1.7, "learning_rate": 3.3579790511398644e-07, "loss": 1.4235, "step": 9110 }, { "epoch": 1.7, "learning_rate": 3.34771000205381e-07, "loss": 0.7978, "step": 9115 }, { "epoch": 1.7, "learning_rate": 3.337440952967755e-07, "loss": 0.9128, "step": 9120 }, { "epoch": 1.7, "learning_rate": 3.3271719038817006e-07, "loss": 0.6848, "step": 9125 }, { "epoch": 1.7, "learning_rate": 3.316902854795646e-07, "loss": 0.6636, "step": 9130 }, { "epoch": 1.7, "learning_rate": 3.3066338057095913e-07, "loss": 0.9091, "step": 9135 }, { "epoch": 1.7, "learning_rate": 3.296364756623536e-07, "loss": 0.6772, "step": 9140 }, { "epoch": 1.7, "learning_rate": 3.2860957075374815e-07, "loss": 0.732, "step": 9145 }, { "epoch": 1.7, "learning_rate": 3.275826658451427e-07, "loss": 0.9508, "step": 9150 }, { "epoch": 1.71, "learning_rate": 3.2655576093653723e-07, "loss": 0.9088, "step": 9155 }, { "epoch": 1.71, "learning_rate": 3.2552885602793177e-07, "loss": 0.7601, "step": 9160 }, { "epoch": 1.71, "learning_rate": 3.245019511193263e-07, "loss": 1.1497, "step": 9165 }, { "epoch": 1.71, "learning_rate": 3.2347504621072084e-07, "loss": 1.4884, "step": 9170 }, { "epoch": 1.71, "learning_rate": 3.224481413021154e-07, "loss": 1.0249, "step": 9175 }, { "epoch": 1.71, "learning_rate": 3.214212363935099e-07, "loss": 0.6794, "step": 9180 }, { "epoch": 1.71, "learning_rate": 3.2039433148490446e-07, "loss": 0.792, "step": 9185 }, { "epoch": 1.71, "learning_rate": 3.19367426576299e-07, "loss": 0.8582, "step": 9190 }, { "epoch": 1.71, "learning_rate": 3.1834052166769353e-07, "loss": 0.7575, "step": 9195 }, { "epoch": 1.71, "learning_rate": 3.1731361675908807e-07, "loss": 0.8361, "step": 9200 }, { "epoch": 1.71, "learning_rate": 3.162867118504826e-07, "loss": 1.4164, "step": 9205 }, { "epoch": 1.72, "learning_rate": 3.1525980694187715e-07, "loss": 0.8392, "step": 9210 }, { "epoch": 1.72, "learning_rate": 3.142329020332717e-07, "loss": 0.6588, "step": 9215 }, { "epoch": 1.72, "learning_rate": 3.132059971246662e-07, "loss": 0.7318, "step": 9220 }, { "epoch": 1.72, "learning_rate": 3.1217909221606076e-07, "loss": 1.029, "step": 9225 }, { "epoch": 1.72, "learning_rate": 3.111521873074553e-07, "loss": 0.9353, "step": 9230 }, { "epoch": 1.72, "learning_rate": 3.1012528239884984e-07, "loss": 1.1036, "step": 9235 }, { "epoch": 1.72, "learning_rate": 3.090983774902444e-07, "loss": 0.8916, "step": 9240 }, { "epoch": 1.72, "learning_rate": 3.080714725816389e-07, "loss": 0.7161, "step": 9245 }, { "epoch": 1.72, "learning_rate": 3.0704456767303345e-07, "loss": 0.7532, "step": 9250 }, { "epoch": 1.72, "learning_rate": 3.06017662764428e-07, "loss": 0.9709, "step": 9255 }, { "epoch": 1.72, "learning_rate": 3.0499075785582253e-07, "loss": 1.0166, "step": 9260 }, { "epoch": 1.73, "learning_rate": 3.0396385294721707e-07, "loss": 1.0056, "step": 9265 }, { "epoch": 1.73, "learning_rate": 3.029369480386116e-07, "loss": 0.7948, "step": 9270 }, { "epoch": 1.73, "learning_rate": 3.0191004313000614e-07, "loss": 0.7811, "step": 9275 }, { "epoch": 1.73, "learning_rate": 3.0088313822140073e-07, "loss": 0.6995, "step": 9280 }, { "epoch": 1.73, "learning_rate": 2.9985623331279527e-07, "loss": 0.8626, "step": 9285 }, { "epoch": 1.73, "learning_rate": 2.988293284041898e-07, "loss": 1.0764, "step": 9290 }, { "epoch": 1.73, "learning_rate": 2.978024234955843e-07, "loss": 1.33, "step": 9295 }, { "epoch": 1.73, "learning_rate": 2.9677551858697883e-07, "loss": 0.8439, "step": 9300 }, { "epoch": 1.73, "learning_rate": 2.9574861367837337e-07, "loss": 0.7295, "step": 9305 }, { "epoch": 1.73, "learning_rate": 2.947217087697679e-07, "loss": 0.7173, "step": 9310 }, { "epoch": 1.73, "learning_rate": 2.9369480386116245e-07, "loss": 0.9574, "step": 9315 }, { "epoch": 1.74, "learning_rate": 2.92667898952557e-07, "loss": 1.0192, "step": 9320 }, { "epoch": 1.74, "learning_rate": 2.916409940439515e-07, "loss": 0.8545, "step": 9325 }, { "epoch": 1.74, "learning_rate": 2.9061408913534606e-07, "loss": 1.3505, "step": 9330 }, { "epoch": 1.74, "learning_rate": 2.895871842267406e-07, "loss": 0.7252, "step": 9335 }, { "epoch": 1.74, "learning_rate": 2.8856027931813514e-07, "loss": 0.7729, "step": 9340 }, { "epoch": 1.74, "learning_rate": 2.875333744095297e-07, "loss": 0.7349, "step": 9345 }, { "epoch": 1.74, "learning_rate": 2.865064695009242e-07, "loss": 0.7728, "step": 9350 }, { "epoch": 1.74, "learning_rate": 2.8547956459231875e-07, "loss": 0.8057, "step": 9355 }, { "epoch": 1.74, "learning_rate": 2.844526596837133e-07, "loss": 0.7852, "step": 9360 }, { "epoch": 1.74, "learning_rate": 2.8342575477510783e-07, "loss": 0.9283, "step": 9365 }, { "epoch": 1.75, "learning_rate": 2.8239884986650236e-07, "loss": 0.7183, "step": 9370 }, { "epoch": 1.75, "learning_rate": 2.813719449578969e-07, "loss": 1.1469, "step": 9375 }, { "epoch": 1.75, "learning_rate": 2.8034504004929144e-07, "loss": 0.7638, "step": 9380 }, { "epoch": 1.75, "learning_rate": 2.79318135140686e-07, "loss": 0.9751, "step": 9385 }, { "epoch": 1.75, "learning_rate": 2.782912302320805e-07, "loss": 0.8638, "step": 9390 }, { "epoch": 1.75, "learning_rate": 2.7726432532347505e-07, "loss": 0.6055, "step": 9395 }, { "epoch": 1.75, "learning_rate": 2.762374204148696e-07, "loss": 0.7411, "step": 9400 }, { "epoch": 1.75, "learning_rate": 2.7521051550626413e-07, "loss": 0.9885, "step": 9405 }, { "epoch": 1.75, "learning_rate": 2.7418361059765867e-07, "loss": 0.8003, "step": 9410 }, { "epoch": 1.75, "learning_rate": 2.731567056890532e-07, "loss": 1.071, "step": 9415 }, { "epoch": 1.75, "learning_rate": 2.7212980078044774e-07, "loss": 0.8145, "step": 9420 }, { "epoch": 1.76, "learning_rate": 2.711028958718423e-07, "loss": 0.6303, "step": 9425 }, { "epoch": 1.76, "learning_rate": 2.700759909632368e-07, "loss": 0.9826, "step": 9430 }, { "epoch": 1.76, "learning_rate": 2.6904908605463136e-07, "loss": 0.7654, "step": 9435 }, { "epoch": 1.76, "learning_rate": 2.680221811460259e-07, "loss": 1.1978, "step": 9440 }, { "epoch": 1.76, "learning_rate": 2.669952762374204e-07, "loss": 1.0603, "step": 9445 }, { "epoch": 1.76, "learning_rate": 2.659683713288149e-07, "loss": 0.8834, "step": 9450 }, { "epoch": 1.76, "learning_rate": 2.6494146642020946e-07, "loss": 0.7221, "step": 9455 }, { "epoch": 1.76, "learning_rate": 2.63914561511604e-07, "loss": 0.8836, "step": 9460 }, { "epoch": 1.76, "learning_rate": 2.6288765660299853e-07, "loss": 0.7714, "step": 9465 }, { "epoch": 1.76, "learning_rate": 2.6186075169439307e-07, "loss": 0.8994, "step": 9470 }, { "epoch": 1.76, "learning_rate": 2.608338467857876e-07, "loss": 0.6029, "step": 9475 }, { "epoch": 1.77, "learning_rate": 2.5980694187718215e-07, "loss": 1.0004, "step": 9480 }, { "epoch": 1.77, "learning_rate": 2.587800369685767e-07, "loss": 0.8633, "step": 9485 }, { "epoch": 1.77, "learning_rate": 2.577531320599712e-07, "loss": 0.8298, "step": 9490 }, { "epoch": 1.77, "learning_rate": 2.5672622715136576e-07, "loss": 0.7246, "step": 9495 }, { "epoch": 1.77, "learning_rate": 2.556993222427603e-07, "loss": 0.8339, "step": 9500 }, { "epoch": 1.77, "eval_loss": 0.5449791550636292, "eval_runtime": 132.473, "eval_samples_per_second": 9.013, "eval_steps_per_second": 1.132, "eval_wer": 0.9112365315546435, "step": 9500 }, { "epoch": 1.77, "learning_rate": 2.5467241733415484e-07, "loss": 1.1769, "step": 9505 }, { "epoch": 1.77, "learning_rate": 2.536455124255494e-07, "loss": 0.795, "step": 9510 }, { "epoch": 1.77, "learning_rate": 2.526186075169439e-07, "loss": 0.9244, "step": 9515 }, { "epoch": 1.77, "learning_rate": 2.5159170260833845e-07, "loss": 0.7642, "step": 9520 }, { "epoch": 1.77, "learning_rate": 2.50564797699733e-07, "loss": 1.0131, "step": 9525 }, { "epoch": 1.78, "learning_rate": 2.4953789279112753e-07, "loss": 0.7834, "step": 9530 }, { "epoch": 1.78, "learning_rate": 2.4851098788252207e-07, "loss": 0.9238, "step": 9535 }, { "epoch": 1.78, "learning_rate": 2.474840829739166e-07, "loss": 0.7334, "step": 9540 }, { "epoch": 1.78, "learning_rate": 2.4645717806531114e-07, "loss": 0.7504, "step": 9545 }, { "epoch": 1.78, "learning_rate": 2.454302731567057e-07, "loss": 0.6311, "step": 9550 }, { "epoch": 1.78, "learning_rate": 2.444033682481002e-07, "loss": 1.1449, "step": 9555 }, { "epoch": 1.78, "learning_rate": 2.4337646333949476e-07, "loss": 1.0693, "step": 9560 }, { "epoch": 1.78, "learning_rate": 2.423495584308893e-07, "loss": 0.8986, "step": 9565 }, { "epoch": 1.78, "learning_rate": 2.4132265352228383e-07, "loss": 1.0381, "step": 9570 }, { "epoch": 1.78, "learning_rate": 2.4029574861367837e-07, "loss": 0.9385, "step": 9575 }, { "epoch": 1.78, "learning_rate": 2.392688437050729e-07, "loss": 0.6327, "step": 9580 }, { "epoch": 1.79, "learning_rate": 2.3824193879646742e-07, "loss": 0.6624, "step": 9585 }, { "epoch": 1.79, "learning_rate": 2.3721503388786196e-07, "loss": 0.7345, "step": 9590 }, { "epoch": 1.79, "learning_rate": 2.3618812897925652e-07, "loss": 0.7158, "step": 9595 }, { "epoch": 1.79, "learning_rate": 2.3516122407065106e-07, "loss": 0.7815, "step": 9600 }, { "epoch": 1.79, "learning_rate": 2.341343191620456e-07, "loss": 1.0816, "step": 9605 }, { "epoch": 1.79, "learning_rate": 2.3310741425344014e-07, "loss": 0.8828, "step": 9610 }, { "epoch": 1.79, "learning_rate": 2.3208050934483467e-07, "loss": 0.9075, "step": 9615 }, { "epoch": 1.79, "learning_rate": 2.310536044362292e-07, "loss": 1.6703, "step": 9620 }, { "epoch": 1.79, "learning_rate": 2.3002669952762375e-07, "loss": 0.7218, "step": 9625 }, { "epoch": 1.79, "learning_rate": 2.289997946190183e-07, "loss": 0.7332, "step": 9630 }, { "epoch": 1.79, "learning_rate": 2.2797288971041283e-07, "loss": 0.6355, "step": 9635 }, { "epoch": 1.8, "learning_rate": 2.2694598480180734e-07, "loss": 0.8792, "step": 9640 }, { "epoch": 1.8, "learning_rate": 2.2591907989320188e-07, "loss": 0.7534, "step": 9645 }, { "epoch": 1.8, "learning_rate": 2.2489217498459641e-07, "loss": 0.738, "step": 9650 }, { "epoch": 1.8, "learning_rate": 2.2386527007599095e-07, "loss": 0.7342, "step": 9655 }, { "epoch": 1.8, "learning_rate": 2.228383651673855e-07, "loss": 0.7395, "step": 9660 }, { "epoch": 1.8, "learning_rate": 2.2181146025878003e-07, "loss": 0.6329, "step": 9665 }, { "epoch": 1.8, "learning_rate": 2.2078455535017457e-07, "loss": 0.8756, "step": 9670 }, { "epoch": 1.8, "learning_rate": 2.197576504415691e-07, "loss": 0.6017, "step": 9675 }, { "epoch": 1.8, "learning_rate": 2.1873074553296364e-07, "loss": 0.6754, "step": 9680 }, { "epoch": 1.8, "learning_rate": 2.1770384062435818e-07, "loss": 0.6871, "step": 9685 }, { "epoch": 1.8, "learning_rate": 2.1667693571575272e-07, "loss": 0.7573, "step": 9690 }, { "epoch": 1.81, "learning_rate": 2.1565003080714726e-07, "loss": 1.0602, "step": 9695 }, { "epoch": 1.81, "learning_rate": 2.146231258985418e-07, "loss": 0.8002, "step": 9700 }, { "epoch": 1.81, "learning_rate": 2.1359622098993633e-07, "loss": 0.9188, "step": 9705 }, { "epoch": 1.81, "learning_rate": 2.1256931608133087e-07, "loss": 0.8239, "step": 9710 }, { "epoch": 1.81, "learning_rate": 2.1154241117272538e-07, "loss": 0.7205, "step": 9715 }, { "epoch": 1.81, "learning_rate": 2.1051550626411992e-07, "loss": 0.9943, "step": 9720 }, { "epoch": 1.81, "learning_rate": 2.0948860135551446e-07, "loss": 0.7139, "step": 9725 }, { "epoch": 1.81, "learning_rate": 2.08461696446909e-07, "loss": 1.0076, "step": 9730 }, { "epoch": 1.81, "learning_rate": 2.0743479153830353e-07, "loss": 0.8668, "step": 9735 }, { "epoch": 1.81, "learning_rate": 2.0640788662969807e-07, "loss": 0.7373, "step": 9740 }, { "epoch": 1.82, "learning_rate": 2.053809817210926e-07, "loss": 0.92, "step": 9745 }, { "epoch": 1.82, "learning_rate": 2.0435407681248715e-07, "loss": 0.7847, "step": 9750 }, { "epoch": 1.82, "learning_rate": 2.0332717190388169e-07, "loss": 0.7019, "step": 9755 }, { "epoch": 1.82, "learning_rate": 2.0230026699527622e-07, "loss": 1.3355, "step": 9760 }, { "epoch": 1.82, "learning_rate": 2.0127336208667076e-07, "loss": 0.8364, "step": 9765 }, { "epoch": 1.82, "learning_rate": 2.002464571780653e-07, "loss": 1.0353, "step": 9770 }, { "epoch": 1.82, "learning_rate": 1.9921955226945984e-07, "loss": 0.799, "step": 9775 }, { "epoch": 1.82, "learning_rate": 1.9819264736085438e-07, "loss": 1.0127, "step": 9780 }, { "epoch": 1.82, "learning_rate": 1.9716574245224891e-07, "loss": 0.5976, "step": 9785 }, { "epoch": 1.82, "learning_rate": 1.9613883754364343e-07, "loss": 0.816, "step": 9790 }, { "epoch": 1.82, "learning_rate": 1.9511193263503796e-07, "loss": 0.8324, "step": 9795 }, { "epoch": 1.83, "learning_rate": 1.940850277264325e-07, "loss": 0.7665, "step": 9800 }, { "epoch": 1.83, "learning_rate": 1.9305812281782707e-07, "loss": 0.7966, "step": 9805 }, { "epoch": 1.83, "learning_rate": 1.920312179092216e-07, "loss": 0.8726, "step": 9810 }, { "epoch": 1.83, "learning_rate": 1.9100431300061614e-07, "loss": 0.7403, "step": 9815 }, { "epoch": 1.83, "learning_rate": 1.8997740809201068e-07, "loss": 0.9103, "step": 9820 }, { "epoch": 1.83, "learning_rate": 1.8895050318340522e-07, "loss": 0.6655, "step": 9825 }, { "epoch": 1.83, "learning_rate": 1.8792359827479976e-07, "loss": 0.6775, "step": 9830 }, { "epoch": 1.83, "learning_rate": 1.868966933661943e-07, "loss": 0.9323, "step": 9835 }, { "epoch": 1.83, "learning_rate": 1.8586978845758883e-07, "loss": 0.6396, "step": 9840 }, { "epoch": 1.83, "learning_rate": 1.8484288354898337e-07, "loss": 0.5921, "step": 9845 }, { "epoch": 1.83, "learning_rate": 1.838159786403779e-07, "loss": 0.8523, "step": 9850 }, { "epoch": 1.84, "learning_rate": 1.8278907373177245e-07, "loss": 0.7539, "step": 9855 }, { "epoch": 1.84, "learning_rate": 1.8176216882316698e-07, "loss": 0.7188, "step": 9860 }, { "epoch": 1.84, "learning_rate": 1.8073526391456152e-07, "loss": 0.9111, "step": 9865 }, { "epoch": 1.84, "learning_rate": 1.7970835900595603e-07, "loss": 0.9562, "step": 9870 }, { "epoch": 1.84, "learning_rate": 1.7868145409735057e-07, "loss": 0.6707, "step": 9875 }, { "epoch": 1.84, "learning_rate": 1.776545491887451e-07, "loss": 0.674, "step": 9880 }, { "epoch": 1.84, "learning_rate": 1.7662764428013965e-07, "loss": 0.7727, "step": 9885 }, { "epoch": 1.84, "learning_rate": 1.7560073937153419e-07, "loss": 0.9508, "step": 9890 }, { "epoch": 1.84, "learning_rate": 1.7457383446292872e-07, "loss": 1.0723, "step": 9895 }, { "epoch": 1.84, "learning_rate": 1.7354692955432326e-07, "loss": 0.9362, "step": 9900 }, { "epoch": 1.84, "learning_rate": 1.725200246457178e-07, "loss": 0.8443, "step": 9905 }, { "epoch": 1.85, "learning_rate": 1.7149311973711234e-07, "loss": 1.0858, "step": 9910 }, { "epoch": 1.85, "learning_rate": 1.7046621482850688e-07, "loss": 0.8252, "step": 9915 }, { "epoch": 1.85, "learning_rate": 1.6943930991990141e-07, "loss": 1.1394, "step": 9920 }, { "epoch": 1.85, "learning_rate": 1.6841240501129595e-07, "loss": 0.9659, "step": 9925 }, { "epoch": 1.85, "learning_rate": 1.673855001026905e-07, "loss": 0.8237, "step": 9930 }, { "epoch": 1.85, "learning_rate": 1.6635859519408503e-07, "loss": 1.0135, "step": 9935 }, { "epoch": 1.85, "learning_rate": 1.6533169028547957e-07, "loss": 0.5366, "step": 9940 }, { "epoch": 1.85, "learning_rate": 1.6430478537687408e-07, "loss": 0.7087, "step": 9945 }, { "epoch": 1.85, "learning_rate": 1.6327788046826862e-07, "loss": 0.7841, "step": 9950 }, { "epoch": 1.85, "learning_rate": 1.6225097555966315e-07, "loss": 0.7555, "step": 9955 }, { "epoch": 1.86, "learning_rate": 1.612240706510577e-07, "loss": 1.0564, "step": 9960 }, { "epoch": 1.86, "learning_rate": 1.6019716574245223e-07, "loss": 1.1673, "step": 9965 }, { "epoch": 1.86, "learning_rate": 1.5917026083384677e-07, "loss": 1.2251, "step": 9970 }, { "epoch": 1.86, "learning_rate": 1.581433559252413e-07, "loss": 0.6291, "step": 9975 }, { "epoch": 1.86, "learning_rate": 1.5711645101663584e-07, "loss": 0.9839, "step": 9980 }, { "epoch": 1.86, "learning_rate": 1.5608954610803038e-07, "loss": 0.6658, "step": 9985 }, { "epoch": 1.86, "learning_rate": 1.5506264119942492e-07, "loss": 0.6988, "step": 9990 }, { "epoch": 1.86, "learning_rate": 1.5403573629081946e-07, "loss": 1.0948, "step": 9995 }, { "epoch": 1.86, "learning_rate": 1.53008831382214e-07, "loss": 0.8519, "step": 10000 }, { "epoch": 1.86, "eval_loss": 0.5432996153831482, "eval_runtime": 131.7393, "eval_samples_per_second": 9.063, "eval_steps_per_second": 1.139, "eval_wer": 0.9117496151872755, "step": 10000 } ], "max_steps": 10738, "num_train_epochs": 2, "total_flos": 9.223744348159401e+17, "trial_name": null, "trial_params": null }