{ "best_metric": 12.237466436164343, "best_model_checkpoint": "./checkpoint-8000", "epoch": 16.0368, "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": 4.4e-07, "loss": 0.7171, "step": 25 }, { "epoch": 0.01, "learning_rate": 9.400000000000001e-07, "loss": 0.6414, "step": 50 }, { "epoch": 0.01, "learning_rate": 1.44e-06, "loss": 0.5087, "step": 75 }, { "epoch": 0.01, "learning_rate": 1.94e-06, "loss": 0.2921, "step": 100 }, { "epoch": 0.01, "learning_rate": 2.4400000000000004e-06, "loss": 0.2483, "step": 125 }, { "epoch": 0.01, "learning_rate": 2.9400000000000002e-06, "loss": 0.2364, "step": 150 }, { "epoch": 0.02, "learning_rate": 3.44e-06, "loss": 0.2065, "step": 175 }, { "epoch": 0.02, "learning_rate": 3.94e-06, "loss": 0.1832, "step": 200 }, { "epoch": 0.02, "learning_rate": 4.440000000000001e-06, "loss": 0.188, "step": 225 }, { "epoch": 0.03, "learning_rate": 4.94e-06, "loss": 0.2169, "step": 250 }, { "epoch": 0.03, "learning_rate": 5.4400000000000004e-06, "loss": 0.2031, "step": 275 }, { "epoch": 0.03, "learning_rate": 5.94e-06, "loss": 0.1825, "step": 300 }, { "epoch": 0.03, "learning_rate": 6.440000000000001e-06, "loss": 0.1908, "step": 325 }, { "epoch": 0.04, "learning_rate": 6.9400000000000005e-06, "loss": 0.1763, "step": 350 }, { "epoch": 0.04, "learning_rate": 7.440000000000001e-06, "loss": 0.1476, "step": 375 }, { "epoch": 0.04, "learning_rate": 7.94e-06, "loss": 0.1632, "step": 400 }, { "epoch": 0.04, "learning_rate": 8.44e-06, "loss": 0.1562, "step": 425 }, { "epoch": 0.04, "learning_rate": 8.94e-06, "loss": 0.1768, "step": 450 }, { "epoch": 0.05, "learning_rate": 9.440000000000001e-06, "loss": 0.1971, "step": 475 }, { "epoch": 0.05, "learning_rate": 9.940000000000001e-06, "loss": 0.2098, "step": 500 }, { "epoch": 0.05, "learning_rate": 9.976842105263158e-06, "loss": 0.2049, "step": 525 }, { "epoch": 0.06, "learning_rate": 9.950526315789475e-06, "loss": 0.1977, "step": 550 }, { "epoch": 0.06, "learning_rate": 9.92421052631579e-06, "loss": 0.1858, "step": 575 }, { "epoch": 0.06, "learning_rate": 9.897894736842107e-06, "loss": 0.1871, "step": 600 }, { "epoch": 1.0, "learning_rate": 9.871578947368422e-06, "loss": 0.1797, "step": 625 }, { "epoch": 1.0, "learning_rate": 9.845263157894738e-06, "loss": 0.1823, "step": 650 }, { "epoch": 1.01, "learning_rate": 9.818947368421053e-06, "loss": 0.1492, "step": 675 }, { "epoch": 1.01, "learning_rate": 9.79263157894737e-06, "loss": 0.1422, "step": 700 }, { "epoch": 1.01, "learning_rate": 9.766315789473685e-06, "loss": 0.1452, "step": 725 }, { "epoch": 1.01, "learning_rate": 9.74e-06, "loss": 0.14, "step": 750 }, { "epoch": 1.02, "learning_rate": 9.713684210526317e-06, "loss": 0.1121, "step": 775 }, { "epoch": 1.02, "learning_rate": 9.687368421052632e-06, "loss": 0.1062, "step": 800 }, { "epoch": 1.02, "learning_rate": 9.661052631578948e-06, "loss": 0.0976, "step": 825 }, { "epoch": 1.02, "learning_rate": 9.634736842105265e-06, "loss": 0.1127, "step": 850 }, { "epoch": 1.03, "learning_rate": 9.60842105263158e-06, "loss": 0.1042, "step": 875 }, { "epoch": 1.03, "learning_rate": 9.582105263157897e-06, "loss": 0.0952, "step": 900 }, { "epoch": 1.03, "learning_rate": 9.555789473684211e-06, "loss": 0.0981, "step": 925 }, { "epoch": 1.03, "learning_rate": 9.529473684210528e-06, "loss": 0.0792, "step": 950 }, { "epoch": 1.04, "learning_rate": 9.503157894736843e-06, "loss": 0.0719, "step": 975 }, { "epoch": 1.04, "learning_rate": 9.476842105263158e-06, "loss": 0.0731, "step": 1000 }, { "epoch": 1.04, "eval_loss": 0.21826446056365967, "eval_runtime": 1088.3225, "eval_samples_per_second": 8.848, "eval_steps_per_second": 0.553, "eval_wer": 13.058937482375155, "step": 1000 }, { "epoch": 1.04, "learning_rate": 9.450526315789475e-06, "loss": 0.072, "step": 1025 }, { "epoch": 1.04, "learning_rate": 9.42421052631579e-06, "loss": 0.0736, "step": 1050 }, { "epoch": 1.05, "learning_rate": 9.397894736842106e-06, "loss": 0.0894, "step": 1075 }, { "epoch": 1.05, "learning_rate": 9.371578947368421e-06, "loss": 0.0928, "step": 1100 }, { "epoch": 1.05, "learning_rate": 9.345263157894738e-06, "loss": 0.0842, "step": 1125 }, { "epoch": 1.05, "learning_rate": 9.318947368421053e-06, "loss": 0.0791, "step": 1150 }, { "epoch": 1.06, "learning_rate": 9.292631578947368e-06, "loss": 0.0694, "step": 1175 }, { "epoch": 1.06, "learning_rate": 9.266315789473685e-06, "loss": 0.0765, "step": 1200 }, { "epoch": 2.0, "learning_rate": 9.240000000000001e-06, "loss": 0.0801, "step": 1225 }, { "epoch": 2.0, "learning_rate": 9.213684210526316e-06, "loss": 0.0751, "step": 1250 }, { "epoch": 2.01, "learning_rate": 9.187368421052633e-06, "loss": 0.0664, "step": 1275 }, { "epoch": 2.01, "learning_rate": 9.161052631578948e-06, "loss": 0.0609, "step": 1300 }, { "epoch": 2.01, "learning_rate": 9.134736842105265e-06, "loss": 0.0603, "step": 1325 }, { "epoch": 2.01, "learning_rate": 9.10842105263158e-06, "loss": 0.058, "step": 1350 }, { "epoch": 2.02, "learning_rate": 9.082105263157896e-06, "loss": 0.045, "step": 1375 }, { "epoch": 2.02, "learning_rate": 9.055789473684211e-06, "loss": 0.0436, "step": 1400 }, { "epoch": 2.02, "learning_rate": 9.029473684210526e-06, "loss": 0.0377, "step": 1425 }, { "epoch": 2.02, "learning_rate": 9.003157894736843e-06, "loss": 0.0471, "step": 1450 }, { "epoch": 2.03, "learning_rate": 8.976842105263158e-06, "loss": 0.0462, "step": 1475 }, { "epoch": 2.03, "learning_rate": 8.950526315789474e-06, "loss": 0.041, "step": 1500 }, { "epoch": 2.03, "learning_rate": 8.92421052631579e-06, "loss": 0.0428, "step": 1525 }, { "epoch": 2.03, "learning_rate": 8.897894736842106e-06, "loss": 0.0326, "step": 1550 }, { "epoch": 2.04, "learning_rate": 8.871578947368423e-06, "loss": 0.0309, "step": 1575 }, { "epoch": 2.04, "learning_rate": 8.845263157894738e-06, "loss": 0.0334, "step": 1600 }, { "epoch": 2.04, "learning_rate": 8.818947368421054e-06, "loss": 0.0329, "step": 1625 }, { "epoch": 2.04, "learning_rate": 8.79263157894737e-06, "loss": 0.0332, "step": 1650 }, { "epoch": 2.05, "learning_rate": 8.766315789473684e-06, "loss": 0.0364, "step": 1675 }, { "epoch": 2.05, "learning_rate": 8.740000000000001e-06, "loss": 0.0437, "step": 1700 }, { "epoch": 2.05, "learning_rate": 8.713684210526316e-06, "loss": 0.0437, "step": 1725 }, { "epoch": 2.05, "learning_rate": 8.687368421052633e-06, "loss": 0.0321, "step": 1750 }, { "epoch": 2.06, "learning_rate": 8.661052631578948e-06, "loss": 0.0326, "step": 1775 }, { "epoch": 2.06, "learning_rate": 8.634736842105264e-06, "loss": 0.0335, "step": 1800 }, { "epoch": 3.0, "learning_rate": 8.60842105263158e-06, "loss": 0.0355, "step": 1825 }, { "epoch": 3.0, "learning_rate": 8.582105263157894e-06, "loss": 0.0335, "step": 1850 }, { "epoch": 3.01, "learning_rate": 8.555789473684213e-06, "loss": 0.0301, "step": 1875 }, { "epoch": 3.01, "learning_rate": 8.529473684210528e-06, "loss": 0.0251, "step": 1900 }, { "epoch": 3.01, "learning_rate": 8.503157894736843e-06, "loss": 0.0266, "step": 1925 }, { "epoch": 3.01, "learning_rate": 8.47684210526316e-06, "loss": 0.0255, "step": 1950 }, { "epoch": 3.02, "learning_rate": 8.450526315789474e-06, "loss": 0.0207, "step": 1975 }, { "epoch": 3.02, "learning_rate": 8.42421052631579e-06, "loss": 0.0194, "step": 2000 }, { "epoch": 3.02, "eval_loss": 0.23903788626194, "eval_runtime": 934.1349, "eval_samples_per_second": 10.309, "eval_steps_per_second": 0.644, "eval_wer": 12.802687559004916, "step": 2000 }, { "epoch": 3.02, "learning_rate": 8.397894736842106e-06, "loss": 0.0182, "step": 2025 }, { "epoch": 3.02, "learning_rate": 8.371578947368422e-06, "loss": 0.0206, "step": 2050 }, { "epoch": 3.03, "learning_rate": 8.345263157894737e-06, "loss": 0.0196, "step": 2075 }, { "epoch": 3.03, "learning_rate": 8.318947368421052e-06, "loss": 0.0185, "step": 2100 }, { "epoch": 3.03, "learning_rate": 8.292631578947369e-06, "loss": 0.0177, "step": 2125 }, { "epoch": 3.03, "learning_rate": 8.266315789473684e-06, "loss": 0.0153, "step": 2150 }, { "epoch": 3.04, "learning_rate": 8.24e-06, "loss": 0.0142, "step": 2175 }, { "epoch": 3.04, "learning_rate": 8.213684210526316e-06, "loss": 0.0146, "step": 2200 }, { "epoch": 3.04, "learning_rate": 8.187368421052632e-06, "loss": 0.0144, "step": 2225 }, { "epoch": 3.04, "learning_rate": 8.161052631578949e-06, "loss": 0.0133, "step": 2250 }, { "epoch": 3.05, "learning_rate": 8.134736842105264e-06, "loss": 0.0172, "step": 2275 }, { "epoch": 3.05, "learning_rate": 8.10842105263158e-06, "loss": 0.0206, "step": 2300 }, { "epoch": 3.05, "learning_rate": 8.082105263157896e-06, "loss": 0.0196, "step": 2325 }, { "epoch": 3.05, "learning_rate": 8.05578947368421e-06, "loss": 0.0178, "step": 2350 }, { "epoch": 3.06, "learning_rate": 8.029473684210527e-06, "loss": 0.0124, "step": 2375 }, { "epoch": 3.06, "learning_rate": 8.003157894736842e-06, "loss": 0.0164, "step": 2400 }, { "epoch": 4.0, "learning_rate": 7.976842105263159e-06, "loss": 0.0164, "step": 2425 }, { "epoch": 4.0, "learning_rate": 7.950526315789474e-06, "loss": 0.0163, "step": 2450 }, { "epoch": 4.01, "learning_rate": 7.92421052631579e-06, "loss": 0.0132, "step": 2475 }, { "epoch": 4.01, "learning_rate": 7.897894736842106e-06, "loss": 0.0106, "step": 2500 }, { "epoch": 4.01, "learning_rate": 7.87157894736842e-06, "loss": 0.012, "step": 2525 }, { "epoch": 4.01, "learning_rate": 7.845263157894739e-06, "loss": 0.0127, "step": 2550 }, { "epoch": 4.02, "learning_rate": 7.818947368421054e-06, "loss": 0.0092, "step": 2575 }, { "epoch": 4.02, "learning_rate": 7.792631578947369e-06, "loss": 0.0095, "step": 2600 }, { "epoch": 4.02, "learning_rate": 7.766315789473685e-06, "loss": 0.0093, "step": 2625 }, { "epoch": 4.02, "learning_rate": 7.74e-06, "loss": 0.0113, "step": 2650 }, { "epoch": 4.03, "learning_rate": 7.713684210526317e-06, "loss": 0.0108, "step": 2675 }, { "epoch": 4.03, "learning_rate": 7.687368421052632e-06, "loss": 0.0095, "step": 2700 }, { "epoch": 4.03, "learning_rate": 7.661052631578949e-06, "loss": 0.0094, "step": 2725 }, { "epoch": 4.03, "learning_rate": 7.634736842105264e-06, "loss": 0.0073, "step": 2750 }, { "epoch": 4.04, "learning_rate": 7.6084210526315795e-06, "loss": 0.0071, "step": 2775 }, { "epoch": 4.04, "learning_rate": 7.582105263157895e-06, "loss": 0.0069, "step": 2800 }, { "epoch": 4.04, "learning_rate": 7.555789473684211e-06, "loss": 0.0069, "step": 2825 }, { "epoch": 4.04, "learning_rate": 7.529473684210526e-06, "loss": 0.0083, "step": 2850 }, { "epoch": 4.05, "learning_rate": 7.503157894736842e-06, "loss": 0.0107, "step": 2875 }, { "epoch": 4.05, "learning_rate": 7.476842105263159e-06, "loss": 0.0098, "step": 2900 }, { "epoch": 4.05, "learning_rate": 7.4505263157894744e-06, "loss": 0.0118, "step": 2925 }, { "epoch": 4.05, "learning_rate": 7.42421052631579e-06, "loss": 0.0101, "step": 2950 }, { "epoch": 4.06, "learning_rate": 7.397894736842106e-06, "loss": 0.0074, "step": 2975 }, { "epoch": 4.06, "learning_rate": 7.371578947368422e-06, "loss": 0.0067, "step": 3000 }, { "epoch": 4.06, "eval_loss": 0.25243330001831055, "eval_runtime": 931.7389, "eval_samples_per_second": 10.336, "eval_steps_per_second": 0.646, "eval_wer": 12.58321992128591, "step": 3000 }, { "epoch": 5.0, "learning_rate": 7.345263157894738e-06, "loss": 0.0071, "step": 3025 }, { "epoch": 5.0, "learning_rate": 7.3189473684210535e-06, "loss": 0.0081, "step": 3050 }, { "epoch": 5.01, "learning_rate": 7.292631578947369e-06, "loss": 0.0073, "step": 3075 }, { "epoch": 5.01, "learning_rate": 7.266315789473684e-06, "loss": 0.0063, "step": 3100 }, { "epoch": 5.01, "learning_rate": 7.24e-06, "loss": 0.006, "step": 3125 }, { "epoch": 5.01, "learning_rate": 7.213684210526316e-06, "loss": 0.0075, "step": 3150 }, { "epoch": 5.02, "learning_rate": 7.187368421052632e-06, "loss": 0.0058, "step": 3175 }, { "epoch": 5.02, "learning_rate": 7.161052631578948e-06, "loss": 0.0051, "step": 3200 }, { "epoch": 5.02, "learning_rate": 7.134736842105264e-06, "loss": 0.0059, "step": 3225 }, { "epoch": 5.02, "learning_rate": 7.10842105263158e-06, "loss": 0.0061, "step": 3250 }, { "epoch": 5.03, "learning_rate": 7.082105263157896e-06, "loss": 0.007, "step": 3275 }, { "epoch": 5.03, "learning_rate": 7.055789473684212e-06, "loss": 0.006, "step": 3300 }, { "epoch": 5.03, "learning_rate": 7.0294736842105275e-06, "loss": 0.0071, "step": 3325 }, { "epoch": 5.03, "learning_rate": 7.0031578947368425e-06, "loss": 0.0062, "step": 3350 }, { "epoch": 5.04, "learning_rate": 6.976842105263158e-06, "loss": 0.0038, "step": 3375 }, { "epoch": 5.04, "learning_rate": 6.950526315789474e-06, "loss": 0.0036, "step": 3400 }, { "epoch": 5.04, "learning_rate": 6.92421052631579e-06, "loss": 0.0035, "step": 3425 }, { "epoch": 5.04, "learning_rate": 6.897894736842106e-06, "loss": 0.004, "step": 3450 }, { "epoch": 5.05, "learning_rate": 6.871578947368422e-06, "loss": 0.0047, "step": 3475 }, { "epoch": 5.05, "learning_rate": 6.845263157894737e-06, "loss": 0.005, "step": 3500 }, { "epoch": 5.05, "learning_rate": 6.818947368421052e-06, "loss": 0.0057, "step": 3525 }, { "epoch": 5.05, "learning_rate": 6.792631578947368e-06, "loss": 0.0053, "step": 3550 }, { "epoch": 5.06, "learning_rate": 6.766315789473685e-06, "loss": 0.0034, "step": 3575 }, { "epoch": 5.06, "learning_rate": 6.740000000000001e-06, "loss": 0.0042, "step": 3600 }, { "epoch": 6.0, "learning_rate": 6.7136842105263165e-06, "loss": 0.0044, "step": 3625 }, { "epoch": 6.0, "learning_rate": 6.687368421052632e-06, "loss": 0.0039, "step": 3650 }, { "epoch": 6.01, "learning_rate": 6.661052631578948e-06, "loss": 0.0037, "step": 3675 }, { "epoch": 6.01, "learning_rate": 6.634736842105264e-06, "loss": 0.0037, "step": 3700 }, { "epoch": 6.01, "learning_rate": 6.60842105263158e-06, "loss": 0.004, "step": 3725 }, { "epoch": 6.01, "learning_rate": 6.582105263157896e-06, "loss": 0.0042, "step": 3750 }, { "epoch": 6.02, "learning_rate": 6.5557894736842106e-06, "loss": 0.0033, "step": 3775 }, { "epoch": 6.02, "learning_rate": 6.529473684210526e-06, "loss": 0.0023, "step": 3800 }, { "epoch": 6.02, "learning_rate": 6.503157894736842e-06, "loss": 0.0039, "step": 3825 }, { "epoch": 6.02, "learning_rate": 6.476842105263158e-06, "loss": 0.0029, "step": 3850 }, { "epoch": 6.03, "learning_rate": 6.450526315789474e-06, "loss": 0.0036, "step": 3875 }, { "epoch": 6.03, "learning_rate": 6.42421052631579e-06, "loss": 0.0027, "step": 3900 }, { "epoch": 6.03, "learning_rate": 6.397894736842106e-06, "loss": 0.0031, "step": 3925 }, { "epoch": 6.03, "learning_rate": 6.371578947368422e-06, "loss": 0.0025, "step": 3950 }, { "epoch": 6.04, "learning_rate": 6.345263157894738e-06, "loss": 0.0028, "step": 3975 }, { "epoch": 6.04, "learning_rate": 6.318947368421054e-06, "loss": 0.0025, "step": 4000 }, { "epoch": 6.04, "eval_loss": 0.2725428342819214, "eval_runtime": 933.2375, "eval_samples_per_second": 10.319, "eval_steps_per_second": 0.645, "eval_wer": 12.324517845538923, "step": 4000 }, { "epoch": 6.04, "learning_rate": 6.292631578947369e-06, "loss": 0.002, "step": 4025 }, { "epoch": 6.04, "learning_rate": 6.266315789473685e-06, "loss": 0.002, "step": 4050 }, { "epoch": 6.05, "learning_rate": 6.24e-06, "loss": 0.0026, "step": 4075 }, { "epoch": 6.05, "learning_rate": 6.213684210526316e-06, "loss": 0.0028, "step": 4100 }, { "epoch": 6.05, "learning_rate": 6.187368421052632e-06, "loss": 0.003, "step": 4125 }, { "epoch": 6.05, "learning_rate": 6.161052631578948e-06, "loss": 0.0026, "step": 4150 }, { "epoch": 6.06, "learning_rate": 6.134736842105264e-06, "loss": 0.0024, "step": 4175 }, { "epoch": 6.06, "learning_rate": 6.108421052631579e-06, "loss": 0.0027, "step": 4200 }, { "epoch": 7.0, "learning_rate": 6.0821052631578945e-06, "loss": 0.0036, "step": 4225 }, { "epoch": 7.0, "learning_rate": 6.055789473684212e-06, "loss": 0.0028, "step": 4250 }, { "epoch": 7.01, "learning_rate": 6.029473684210527e-06, "loss": 0.0023, "step": 4275 }, { "epoch": 7.01, "learning_rate": 6.003157894736843e-06, "loss": 0.0025, "step": 4300 }, { "epoch": 7.01, "learning_rate": 5.976842105263159e-06, "loss": 0.0025, "step": 4325 }, { "epoch": 7.01, "learning_rate": 5.950526315789474e-06, "loss": 0.002, "step": 4350 }, { "epoch": 7.02, "learning_rate": 5.92421052631579e-06, "loss": 0.002, "step": 4375 }, { "epoch": 7.02, "learning_rate": 5.897894736842106e-06, "loss": 0.0018, "step": 4400 }, { "epoch": 7.02, "learning_rate": 5.871578947368422e-06, "loss": 0.0021, "step": 4425 }, { "epoch": 7.02, "learning_rate": 5.845263157894737e-06, "loss": 0.0016, "step": 4450 }, { "epoch": 7.03, "learning_rate": 5.818947368421053e-06, "loss": 0.002, "step": 4475 }, { "epoch": 7.03, "learning_rate": 5.7926315789473685e-06, "loss": 0.0027, "step": 4500 }, { "epoch": 7.03, "learning_rate": 5.766315789473684e-06, "loss": 0.0017, "step": 4525 }, { "epoch": 7.03, "learning_rate": 5.74e-06, "loss": 0.002, "step": 4550 }, { "epoch": 7.04, "learning_rate": 5.713684210526316e-06, "loss": 0.0017, "step": 4575 }, { "epoch": 7.04, "learning_rate": 5.687368421052633e-06, "loss": 0.0013, "step": 4600 }, { "epoch": 7.04, "learning_rate": 5.661052631578948e-06, "loss": 0.0014, "step": 4625 }, { "epoch": 7.04, "learning_rate": 5.634736842105264e-06, "loss": 0.0016, "step": 4650 }, { "epoch": 7.05, "learning_rate": 5.60842105263158e-06, "loss": 0.003, "step": 4675 }, { "epoch": 7.05, "learning_rate": 5.582105263157895e-06, "loss": 0.002, "step": 4700 }, { "epoch": 7.05, "learning_rate": 5.555789473684211e-06, "loss": 0.0019, "step": 4725 }, { "epoch": 7.05, "learning_rate": 5.529473684210527e-06, "loss": 0.0022, "step": 4750 }, { "epoch": 7.06, "learning_rate": 5.5031578947368425e-06, "loss": 0.0021, "step": 4775 }, { "epoch": 7.06, "learning_rate": 5.476842105263158e-06, "loss": 0.0015, "step": 4800 }, { "epoch": 8.0, "learning_rate": 5.450526315789474e-06, "loss": 0.0019, "step": 4825 }, { "epoch": 8.0, "learning_rate": 5.42421052631579e-06, "loss": 0.0028, "step": 4850 }, { "epoch": 8.01, "learning_rate": 5.397894736842105e-06, "loss": 0.002, "step": 4875 }, { "epoch": 8.01, "learning_rate": 5.371578947368421e-06, "loss": 0.002, "step": 4900 }, { "epoch": 8.01, "learning_rate": 5.345263157894738e-06, "loss": 0.002, "step": 4925 }, { "epoch": 8.01, "learning_rate": 5.318947368421053e-06, "loss": 0.002, "step": 4950 }, { "epoch": 8.02, "learning_rate": 5.292631578947369e-06, "loss": 0.0016, "step": 4975 }, { "epoch": 8.02, "learning_rate": 5.266315789473685e-06, "loss": 0.0017, "step": 5000 }, { "epoch": 8.02, "eval_loss": 0.28544357419013977, "eval_runtime": 938.3685, "eval_samples_per_second": 10.262, "eval_steps_per_second": 0.642, "eval_wer": 12.704601463934969, "step": 5000 }, { "epoch": 8.02, "learning_rate": 5.240000000000001e-06, "loss": 0.0016, "step": 5025 }, { "epoch": 8.02, "learning_rate": 5.2136842105263165e-06, "loss": 0.002, "step": 5050 }, { "epoch": 8.03, "learning_rate": 5.187368421052632e-06, "loss": 0.0016, "step": 5075 }, { "epoch": 8.03, "learning_rate": 5.161052631578948e-06, "loss": 0.0015, "step": 5100 }, { "epoch": 8.03, "learning_rate": 5.134736842105263e-06, "loss": 0.0015, "step": 5125 }, { "epoch": 8.03, "learning_rate": 5.108421052631579e-06, "loss": 0.001, "step": 5150 }, { "epoch": 8.04, "learning_rate": 5.082105263157895e-06, "loss": 0.001, "step": 5175 }, { "epoch": 8.04, "learning_rate": 5.0557894736842105e-06, "loss": 0.0013, "step": 5200 }, { "epoch": 8.04, "learning_rate": 5.029473684210526e-06, "loss": 0.0014, "step": 5225 }, { "epoch": 8.04, "learning_rate": 5.003157894736842e-06, "loss": 0.0017, "step": 5250 }, { "epoch": 8.05, "learning_rate": 4.976842105263158e-06, "loss": 0.002, "step": 5275 }, { "epoch": 8.05, "learning_rate": 4.950526315789474e-06, "loss": 0.0013, "step": 5300 }, { "epoch": 8.05, "learning_rate": 4.92421052631579e-06, "loss": 0.0014, "step": 5325 }, { "epoch": 8.05, "learning_rate": 4.897894736842106e-06, "loss": 0.0014, "step": 5350 }, { "epoch": 8.06, "learning_rate": 4.871578947368421e-06, "loss": 0.0016, "step": 5375 }, { "epoch": 8.06, "learning_rate": 4.845263157894737e-06, "loss": 0.0019, "step": 5400 }, { "epoch": 9.0, "learning_rate": 4.818947368421053e-06, "loss": 0.0014, "step": 5425 }, { "epoch": 9.0, "learning_rate": 4.792631578947369e-06, "loss": 0.0016, "step": 5450 }, { "epoch": 9.01, "learning_rate": 4.7663157894736845e-06, "loss": 0.0019, "step": 5475 }, { "epoch": 9.01, "learning_rate": 4.74e-06, "loss": 0.0015, "step": 5500 }, { "epoch": 9.01, "learning_rate": 4.713684210526316e-06, "loss": 0.0012, "step": 5525 }, { "epoch": 9.01, "learning_rate": 4.687368421052632e-06, "loss": 0.0011, "step": 5550 }, { "epoch": 9.02, "learning_rate": 4.661052631578948e-06, "loss": 0.001, "step": 5575 }, { "epoch": 9.02, "learning_rate": 4.634736842105264e-06, "loss": 0.0009, "step": 5600 }, { "epoch": 9.02, "learning_rate": 4.6084210526315794e-06, "loss": 0.001, "step": 5625 }, { "epoch": 9.02, "learning_rate": 4.582105263157895e-06, "loss": 0.0008, "step": 5650 }, { "epoch": 9.03, "learning_rate": 4.55578947368421e-06, "loss": 0.0012, "step": 5675 }, { "epoch": 9.03, "learning_rate": 4.529473684210527e-06, "loss": 0.0011, "step": 5700 }, { "epoch": 9.03, "learning_rate": 4.503157894736843e-06, "loss": 0.0014, "step": 5725 }, { "epoch": 9.03, "learning_rate": 4.4768421052631585e-06, "loss": 0.001, "step": 5750 }, { "epoch": 9.04, "learning_rate": 4.450526315789474e-06, "loss": 0.0009, "step": 5775 }, { "epoch": 9.04, "learning_rate": 4.424210526315789e-06, "loss": 0.001, "step": 5800 }, { "epoch": 9.04, "learning_rate": 4.397894736842105e-06, "loss": 0.0009, "step": 5825 }, { "epoch": 9.04, "learning_rate": 4.371578947368421e-06, "loss": 0.0017, "step": 5850 }, { "epoch": 9.05, "learning_rate": 4.345263157894738e-06, "loss": 0.0012, "step": 5875 }, { "epoch": 9.05, "learning_rate": 4.3189473684210535e-06, "loss": 0.001, "step": 5900 }, { "epoch": 9.05, "learning_rate": 4.2926315789473684e-06, "loss": 0.0011, "step": 5925 }, { "epoch": 9.05, "learning_rate": 4.266315789473684e-06, "loss": 0.0009, "step": 5950 }, { "epoch": 9.06, "learning_rate": 4.24e-06, "loss": 0.0012, "step": 5975 }, { "epoch": 9.06, "learning_rate": 4.213684210526316e-06, "loss": 0.0009, "step": 6000 }, { "epoch": 9.06, "eval_loss": 0.2915147542953491, "eval_runtime": 902.4505, "eval_samples_per_second": 10.671, "eval_steps_per_second": 0.667, "eval_wer": 12.507203197606701, "step": 6000 }, { "epoch": 10.0, "learning_rate": 4.1873684210526325e-06, "loss": 0.002, "step": 6025 }, { "epoch": 10.0, "learning_rate": 4.1610526315789475e-06, "loss": 0.0008, "step": 6050 }, { "epoch": 10.01, "learning_rate": 4.134736842105263e-06, "loss": 0.001, "step": 6075 }, { "epoch": 10.01, "learning_rate": 4.108421052631579e-06, "loss": 0.0008, "step": 6100 }, { "epoch": 10.01, "learning_rate": 4.082105263157895e-06, "loss": 0.0007, "step": 6125 }, { "epoch": 10.01, "learning_rate": 4.055789473684211e-06, "loss": 0.0009, "step": 6150 }, { "epoch": 10.02, "learning_rate": 4.029473684210527e-06, "loss": 0.0008, "step": 6175 }, { "epoch": 10.02, "learning_rate": 4.0031578947368424e-06, "loss": 0.0007, "step": 6200 }, { "epoch": 10.02, "learning_rate": 3.976842105263158e-06, "loss": 0.0006, "step": 6225 }, { "epoch": 10.02, "learning_rate": 3.950526315789474e-06, "loss": 0.0007, "step": 6250 }, { "epoch": 10.03, "learning_rate": 3.92421052631579e-06, "loss": 0.0011, "step": 6275 }, { "epoch": 10.03, "learning_rate": 3.897894736842106e-06, "loss": 0.0016, "step": 6300 }, { "epoch": 10.03, "learning_rate": 3.8715789473684215e-06, "loss": 0.0007, "step": 6325 }, { "epoch": 10.03, "learning_rate": 3.845263157894737e-06, "loss": 0.0007, "step": 6350 }, { "epoch": 10.04, "learning_rate": 3.818947368421053e-06, "loss": 0.0006, "step": 6375 }, { "epoch": 10.04, "learning_rate": 3.792631578947369e-06, "loss": 0.0006, "step": 6400 }, { "epoch": 10.04, "learning_rate": 3.766315789473685e-06, "loss": 0.0006, "step": 6425 }, { "epoch": 10.04, "learning_rate": 3.74e-06, "loss": 0.0006, "step": 6450 }, { "epoch": 10.05, "learning_rate": 3.713684210526316e-06, "loss": 0.0008, "step": 6475 }, { "epoch": 10.05, "learning_rate": 3.687368421052632e-06, "loss": 0.0009, "step": 6500 }, { "epoch": 10.05, "learning_rate": 3.6610526315789472e-06, "loss": 0.0009, "step": 6525 }, { "epoch": 10.05, "learning_rate": 3.6347368421052635e-06, "loss": 0.001, "step": 6550 }, { "epoch": 10.06, "learning_rate": 3.6084210526315793e-06, "loss": 0.0007, "step": 6575 }, { "epoch": 10.06, "learning_rate": 3.582105263157895e-06, "loss": 0.0007, "step": 6600 }, { "epoch": 11.0, "learning_rate": 3.555789473684211e-06, "loss": 0.0007, "step": 6625 }, { "epoch": 11.0, "learning_rate": 3.5294736842105263e-06, "loss": 0.0007, "step": 6650 }, { "epoch": 11.01, "learning_rate": 3.503157894736842e-06, "loss": 0.0006, "step": 6675 }, { "epoch": 11.01, "learning_rate": 3.476842105263158e-06, "loss": 0.0009, "step": 6700 }, { "epoch": 11.01, "learning_rate": 3.450526315789474e-06, "loss": 0.0006, "step": 6725 }, { "epoch": 11.01, "learning_rate": 3.42421052631579e-06, "loss": 0.0006, "step": 6750 }, { "epoch": 11.02, "learning_rate": 3.3978947368421054e-06, "loss": 0.0008, "step": 6775 }, { "epoch": 11.02, "learning_rate": 3.3715789473684212e-06, "loss": 0.0006, "step": 6800 }, { "epoch": 11.02, "learning_rate": 3.345263157894737e-06, "loss": 0.0006, "step": 6825 }, { "epoch": 11.02, "learning_rate": 3.318947368421053e-06, "loss": 0.0005, "step": 6850 }, { "epoch": 11.03, "learning_rate": 3.292631578947369e-06, "loss": 0.0006, "step": 6875 }, { "epoch": 11.03, "learning_rate": 3.2663157894736845e-06, "loss": 0.0011, "step": 6900 }, { "epoch": 11.03, "learning_rate": 3.2400000000000003e-06, "loss": 0.0006, "step": 6925 }, { "epoch": 11.03, "learning_rate": 3.213684210526316e-06, "loss": 0.0006, "step": 6950 }, { "epoch": 11.04, "learning_rate": 3.187368421052632e-06, "loss": 0.0005, "step": 6975 }, { "epoch": 11.04, "learning_rate": 3.1610526315789474e-06, "loss": 0.0005, "step": 7000 }, { "epoch": 11.04, "eval_loss": 0.3005782663822174, "eval_runtime": 933.6239, "eval_samples_per_second": 10.315, "eval_steps_per_second": 0.645, "eval_wer": 12.247275045671339, "step": 7000 }, { "epoch": 11.04, "learning_rate": 3.134736842105263e-06, "loss": 0.0005, "step": 7025 }, { "epoch": 11.04, "learning_rate": 3.1084210526315794e-06, "loss": 0.0005, "step": 7050 }, { "epoch": 11.05, "learning_rate": 3.0821052631578952e-06, "loss": 0.0005, "step": 7075 }, { "epoch": 11.05, "learning_rate": 3.055789473684211e-06, "loss": 0.0006, "step": 7100 }, { "epoch": 11.05, "learning_rate": 3.0294736842105264e-06, "loss": 0.0006, "step": 7125 }, { "epoch": 11.05, "learning_rate": 3.0031578947368423e-06, "loss": 0.0006, "step": 7150 }, { "epoch": 11.06, "learning_rate": 2.976842105263158e-06, "loss": 0.0005, "step": 7175 }, { "epoch": 11.06, "learning_rate": 2.9505263157894735e-06, "loss": 0.0005, "step": 7200 }, { "epoch": 12.0, "learning_rate": 2.92421052631579e-06, "loss": 0.0005, "step": 7225 }, { "epoch": 12.0, "learning_rate": 2.8978947368421055e-06, "loss": 0.0005, "step": 7250 }, { "epoch": 12.01, "learning_rate": 2.8715789473684214e-06, "loss": 0.0008, "step": 7275 }, { "epoch": 12.01, "learning_rate": 2.845263157894737e-06, "loss": 0.0005, "step": 7300 }, { "epoch": 12.01, "learning_rate": 2.8189473684210526e-06, "loss": 0.0008, "step": 7325 }, { "epoch": 12.01, "learning_rate": 2.7926315789473684e-06, "loss": 0.0005, "step": 7350 }, { "epoch": 12.02, "learning_rate": 2.766315789473684e-06, "loss": 0.0005, "step": 7375 }, { "epoch": 12.02, "learning_rate": 2.7400000000000004e-06, "loss": 0.0005, "step": 7400 }, { "epoch": 12.02, "learning_rate": 2.7136842105263163e-06, "loss": 0.0004, "step": 7425 }, { "epoch": 12.02, "learning_rate": 2.6873684210526317e-06, "loss": 0.0005, "step": 7450 }, { "epoch": 12.03, "learning_rate": 2.6610526315789475e-06, "loss": 0.0005, "step": 7475 }, { "epoch": 12.03, "learning_rate": 2.6347368421052633e-06, "loss": 0.0006, "step": 7500 }, { "epoch": 12.03, "learning_rate": 2.608421052631579e-06, "loss": 0.0005, "step": 7525 }, { "epoch": 12.03, "learning_rate": 2.5821052631578954e-06, "loss": 0.0004, "step": 7550 }, { "epoch": 12.04, "learning_rate": 2.5557894736842108e-06, "loss": 0.0004, "step": 7575 }, { "epoch": 12.04, "learning_rate": 2.5294736842105266e-06, "loss": 0.0005, "step": 7600 }, { "epoch": 12.04, "learning_rate": 2.5031578947368424e-06, "loss": 0.0004, "step": 7625 }, { "epoch": 12.04, "learning_rate": 2.476842105263158e-06, "loss": 0.0004, "step": 7650 }, { "epoch": 12.05, "learning_rate": 2.4505263157894736e-06, "loss": 0.0005, "step": 7675 }, { "epoch": 12.05, "learning_rate": 2.42421052631579e-06, "loss": 0.0005, "step": 7700 }, { "epoch": 12.05, "learning_rate": 2.3978947368421052e-06, "loss": 0.0005, "step": 7725 }, { "epoch": 12.05, "learning_rate": 2.371578947368421e-06, "loss": 0.0005, "step": 7750 }, { "epoch": 12.06, "learning_rate": 2.3452631578947373e-06, "loss": 0.0004, "step": 7775 }, { "epoch": 12.06, "learning_rate": 2.3189473684210527e-06, "loss": 0.0004, "step": 7800 }, { "epoch": 12.06, "learning_rate": 2.2926315789473685e-06, "loss": 0.0004, "step": 7825 }, { "epoch": 13.0, "learning_rate": 2.2663157894736843e-06, "loss": 0.0009, "step": 7850 }, { "epoch": 13.0, "learning_rate": 2.24e-06, "loss": 0.0004, "step": 7875 }, { "epoch": 13.01, "learning_rate": 2.213684210526316e-06, "loss": 0.0005, "step": 7900 }, { "epoch": 13.01, "learning_rate": 2.187368421052632e-06, "loss": 0.0004, "step": 7925 }, { "epoch": 13.01, "learning_rate": 2.1610526315789476e-06, "loss": 0.0004, "step": 7950 }, { "epoch": 13.01, "learning_rate": 2.1347368421052634e-06, "loss": 0.0004, "step": 7975 }, { "epoch": 13.02, "learning_rate": 2.1084210526315792e-06, "loss": 0.0004, "step": 8000 }, { "epoch": 13.02, "eval_loss": 0.30600619316101074, "eval_runtime": 901.1215, "eval_samples_per_second": 10.687, "eval_steps_per_second": 0.668, "eval_wer": 12.237466436164343, "step": 8000 }, { "epoch": 13.02, "learning_rate": 2.082105263157895e-06, "loss": 0.0004, "step": 8025 }, { "epoch": 13.02, "learning_rate": 2.055789473684211e-06, "loss": 0.0004, "step": 8050 }, { "epoch": 13.02, "learning_rate": 2.0294736842105263e-06, "loss": 0.0004, "step": 8075 }, { "epoch": 13.03, "learning_rate": 2.003157894736842e-06, "loss": 0.0004, "step": 8100 }, { "epoch": 13.03, "learning_rate": 1.976842105263158e-06, "loss": 0.0004, "step": 8125 }, { "epoch": 13.03, "learning_rate": 1.9505263157894737e-06, "loss": 0.0004, "step": 8150 }, { "epoch": 13.03, "learning_rate": 1.9242105263157896e-06, "loss": 0.0004, "step": 8175 }, { "epoch": 13.04, "learning_rate": 1.8978947368421056e-06, "loss": 0.0003, "step": 8200 }, { "epoch": 13.04, "learning_rate": 1.8715789473684212e-06, "loss": 0.0004, "step": 8225 }, { "epoch": 13.04, "learning_rate": 1.845263157894737e-06, "loss": 0.0004, "step": 8250 }, { "epoch": 13.04, "learning_rate": 1.8189473684210528e-06, "loss": 0.0004, "step": 8275 }, { "epoch": 13.05, "learning_rate": 1.7926315789473686e-06, "loss": 0.0004, "step": 8300 }, { "epoch": 13.05, "learning_rate": 1.7663157894736843e-06, "loss": 0.0004, "step": 8325 }, { "epoch": 13.05, "learning_rate": 1.74e-06, "loss": 0.0004, "step": 8350 }, { "epoch": 13.05, "learning_rate": 1.713684210526316e-06, "loss": 0.0004, "step": 8375 }, { "epoch": 13.06, "learning_rate": 1.6873684210526317e-06, "loss": 0.0004, "step": 8400 }, { "epoch": 13.06, "learning_rate": 1.6610526315789473e-06, "loss": 0.0004, "step": 8425 }, { "epoch": 14.0, "learning_rate": 1.6347368421052633e-06, "loss": 0.0004, "step": 8450 }, { "epoch": 14.0, "learning_rate": 1.6084210526315792e-06, "loss": 0.0004, "step": 8475 }, { "epoch": 14.01, "learning_rate": 1.5821052631578948e-06, "loss": 0.0005, "step": 8500 }, { "epoch": 14.01, "learning_rate": 1.5557894736842106e-06, "loss": 0.0003, "step": 8525 }, { "epoch": 14.01, "learning_rate": 1.5294736842105264e-06, "loss": 0.0004, "step": 8550 }, { "epoch": 14.01, "learning_rate": 1.5031578947368422e-06, "loss": 0.0004, "step": 8575 }, { "epoch": 14.02, "learning_rate": 1.4768421052631578e-06, "loss": 0.0004, "step": 8600 }, { "epoch": 14.02, "learning_rate": 1.4505263157894739e-06, "loss": 0.0003, "step": 8625 }, { "epoch": 14.02, "learning_rate": 1.4242105263157897e-06, "loss": 0.0003, "step": 8650 }, { "epoch": 14.02, "learning_rate": 1.3978947368421053e-06, "loss": 0.0004, "step": 8675 }, { "epoch": 14.03, "learning_rate": 1.3715789473684213e-06, "loss": 0.0003, "step": 8700 }, { "epoch": 14.03, "learning_rate": 1.345263157894737e-06, "loss": 0.0003, "step": 8725 }, { "epoch": 14.03, "learning_rate": 1.3189473684210527e-06, "loss": 0.0003, "step": 8750 }, { "epoch": 14.03, "learning_rate": 1.2926315789473683e-06, "loss": 0.0003, "step": 8775 }, { "epoch": 14.04, "learning_rate": 1.2663157894736844e-06, "loss": 0.0003, "step": 8800 }, { "epoch": 14.04, "learning_rate": 1.2400000000000002e-06, "loss": 0.0003, "step": 8825 }, { "epoch": 14.04, "learning_rate": 1.213684210526316e-06, "loss": 0.0003, "step": 8850 }, { "epoch": 14.04, "learning_rate": 1.1873684210526316e-06, "loss": 0.0003, "step": 8875 }, { "epoch": 14.05, "learning_rate": 1.1610526315789474e-06, "loss": 0.0004, "step": 8900 }, { "epoch": 14.05, "learning_rate": 1.1347368421052633e-06, "loss": 0.0004, "step": 8925 }, { "epoch": 14.05, "learning_rate": 1.108421052631579e-06, "loss": 0.0004, "step": 8950 }, { "epoch": 14.05, "learning_rate": 1.082105263157895e-06, "loss": 0.0003, "step": 8975 }, { "epoch": 14.06, "learning_rate": 1.0557894736842105e-06, "loss": 0.0003, "step": 9000 }, { "epoch": 14.06, "eval_loss": 0.3128972053527832, "eval_runtime": 930.4653, "eval_samples_per_second": 10.35, "eval_steps_per_second": 0.647, "eval_wer": 12.296318093206311, "step": 9000 }, { "epoch": 14.06, "learning_rate": 1.0294736842105265e-06, "loss": 0.0003, "step": 9025 }, { "epoch": 15.0, "learning_rate": 1.0031578947368421e-06, "loss": 0.0004, "step": 9050 }, { "epoch": 15.0, "learning_rate": 9.76842105263158e-07, "loss": 0.0003, "step": 9075 }, { "epoch": 15.01, "learning_rate": 9.505263157894738e-07, "loss": 0.0004, "step": 9100 }, { "epoch": 15.01, "learning_rate": 9.242105263157895e-07, "loss": 0.0003, "step": 9125 }, { "epoch": 15.01, "learning_rate": 8.978947368421054e-07, "loss": 0.0003, "step": 9150 }, { "epoch": 15.01, "learning_rate": 8.715789473684211e-07, "loss": 0.0003, "step": 9175 }, { "epoch": 15.02, "learning_rate": 8.452631578947369e-07, "loss": 0.0003, "step": 9200 }, { "epoch": 15.02, "learning_rate": 8.189473684210527e-07, "loss": 0.0003, "step": 9225 }, { "epoch": 15.02, "learning_rate": 7.926315789473685e-07, "loss": 0.0003, "step": 9250 }, { "epoch": 15.02, "learning_rate": 7.663157894736843e-07, "loss": 0.0003, "step": 9275 }, { "epoch": 15.03, "learning_rate": 7.4e-07, "loss": 0.0004, "step": 9300 }, { "epoch": 15.03, "learning_rate": 7.136842105263158e-07, "loss": 0.0003, "step": 9325 }, { "epoch": 15.03, "learning_rate": 6.873684210526315e-07, "loss": 0.0003, "step": 9350 }, { "epoch": 15.03, "learning_rate": 6.610526315789475e-07, "loss": 0.0003, "step": 9375 }, { "epoch": 15.04, "learning_rate": 6.347368421052633e-07, "loss": 0.0003, "step": 9400 }, { "epoch": 15.04, "learning_rate": 6.08421052631579e-07, "loss": 0.0003, "step": 9425 }, { "epoch": 15.04, "learning_rate": 5.821052631578948e-07, "loss": 0.0003, "step": 9450 }, { "epoch": 15.04, "learning_rate": 5.557894736842106e-07, "loss": 0.0003, "step": 9475 }, { "epoch": 15.05, "learning_rate": 5.294736842105263e-07, "loss": 0.0003, "step": 9500 }, { "epoch": 15.05, "learning_rate": 5.031578947368422e-07, "loss": 0.0003, "step": 9525 }, { "epoch": 15.05, "learning_rate": 4.768421052631579e-07, "loss": 0.0003, "step": 9550 }, { "epoch": 15.05, "learning_rate": 4.505263157894737e-07, "loss": 0.0003, "step": 9575 }, { "epoch": 15.06, "learning_rate": 4.242105263157895e-07, "loss": 0.0003, "step": 9600 }, { "epoch": 15.06, "learning_rate": 3.9789473684210533e-07, "loss": 0.0003, "step": 9625 }, { "epoch": 16.0, "learning_rate": 3.715789473684211e-07, "loss": 0.0003, "step": 9650 }, { "epoch": 16.0, "learning_rate": 3.4526315789473686e-07, "loss": 0.0003, "step": 9675 }, { "epoch": 16.01, "learning_rate": 3.1894736842105263e-07, "loss": 0.0004, "step": 9700 }, { "epoch": 16.01, "learning_rate": 2.9263157894736845e-07, "loss": 0.0003, "step": 9725 }, { "epoch": 16.01, "learning_rate": 2.663157894736842e-07, "loss": 0.0003, "step": 9750 }, { "epoch": 16.01, "learning_rate": 2.4000000000000003e-07, "loss": 0.0003, "step": 9775 }, { "epoch": 16.02, "learning_rate": 2.136842105263158e-07, "loss": 0.0003, "step": 9800 }, { "epoch": 16.02, "learning_rate": 1.8736842105263162e-07, "loss": 0.0003, "step": 9825 }, { "epoch": 16.02, "learning_rate": 1.6105263157894738e-07, "loss": 0.0003, "step": 9850 }, { "epoch": 16.02, "learning_rate": 1.3473684210526317e-07, "loss": 0.0003, "step": 9875 }, { "epoch": 16.03, "learning_rate": 1.0842105263157895e-07, "loss": 0.0003, "step": 9900 }, { "epoch": 16.03, "learning_rate": 8.210526315789474e-08, "loss": 0.0003, "step": 9925 }, { "epoch": 16.03, "learning_rate": 5.578947368421053e-08, "loss": 0.0003, "step": 9950 }, { "epoch": 16.03, "learning_rate": 2.9473684210526318e-08, "loss": 0.0003, "step": 9975 }, { "epoch": 16.04, "learning_rate": 3.1578947368421054e-09, "loss": 0.0003, "step": 10000 }, { "epoch": 16.04, "eval_loss": 0.3156881630420685, "eval_runtime": 932.5213, "eval_samples_per_second": 10.327, "eval_steps_per_second": 0.646, "eval_wer": 12.29877024558306, "step": 10000 }, { "epoch": 16.04, "step": 10000, "total_flos": 9.231962390396928e+19, "train_loss": 0.026919007622404022, "train_runtime": 42713.0008, "train_samples_per_second": 7.492, "train_steps_per_second": 0.234 } ], "max_steps": 10000, "num_train_epochs": 9223372036854775807, "total_flos": 9.231962390396928e+19, "trial_name": null, "trial_params": null }