{ "best_metric": 5.309030539895549, "best_model_checkpoint": "./checkpoint-20000", "epoch": 116.61807580174927, "eval_steps": 1000, "global_step": 20000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.15, "learning_rate": 4.4e-07, "loss": 6.5991, "step": 25 }, { "epoch": 0.29, "learning_rate": 9.400000000000001e-07, "loss": 5.0457, "step": 50 }, { "epoch": 0.44, "learning_rate": 1.44e-06, "loss": 3.7648, "step": 75 }, { "epoch": 0.58, "learning_rate": 1.94e-06, "loss": 2.6623, "step": 100 }, { "epoch": 0.73, "learning_rate": 2.4400000000000004e-06, "loss": 1.9696, "step": 125 }, { "epoch": 0.87, "learning_rate": 2.9400000000000002e-06, "loss": 1.4657, "step": 150 }, { "epoch": 1.02, "learning_rate": 3.44e-06, "loss": 1.194, "step": 175 }, { "epoch": 1.17, "learning_rate": 3.94e-06, "loss": 0.8863, "step": 200 }, { "epoch": 1.31, "learning_rate": 4.440000000000001e-06, "loss": 0.751, "step": 225 }, { "epoch": 1.46, "learning_rate": 4.94e-06, "loss": 0.6216, "step": 250 }, { "epoch": 1.6, "learning_rate": 5.4400000000000004e-06, "loss": 0.5416, "step": 275 }, { "epoch": 1.75, "learning_rate": 5.94e-06, "loss": 0.4472, "step": 300 }, { "epoch": 1.9, "learning_rate": 6.440000000000001e-06, "loss": 0.4225, "step": 325 }, { "epoch": 2.04, "learning_rate": 6.9400000000000005e-06, "loss": 0.3439, "step": 350 }, { "epoch": 2.19, "learning_rate": 7.440000000000001e-06, "loss": 0.3089, "step": 375 }, { "epoch": 2.33, "learning_rate": 7.94e-06, "loss": 0.2844, "step": 400 }, { "epoch": 2.48, "learning_rate": 8.44e-06, "loss": 0.2741, "step": 425 }, { "epoch": 2.62, "learning_rate": 8.94e-06, "loss": 0.2675, "step": 450 }, { "epoch": 2.77, "learning_rate": 9.440000000000001e-06, "loss": 0.2609, "step": 475 }, { "epoch": 2.92, "learning_rate": 9.940000000000001e-06, "loss": 0.2495, "step": 500 }, { "epoch": 3.06, "learning_rate": 9.98871794871795e-06, "loss": 0.214, "step": 525 }, { "epoch": 3.21, "learning_rate": 9.975897435897436e-06, "loss": 0.1678, "step": 550 }, { "epoch": 3.35, "learning_rate": 9.963076923076925e-06, "loss": 0.1623, "step": 575 }, { "epoch": 3.5, "learning_rate": 9.950256410256412e-06, "loss": 0.1752, "step": 600 }, { "epoch": 3.64, "learning_rate": 9.937435897435898e-06, "loss": 0.1626, "step": 625 }, { "epoch": 3.79, "learning_rate": 9.924615384615385e-06, "loss": 0.1743, "step": 650 }, { "epoch": 3.94, "learning_rate": 9.911794871794874e-06, "loss": 0.1594, "step": 675 }, { "epoch": 4.08, "learning_rate": 9.89897435897436e-06, "loss": 0.135, "step": 700 }, { "epoch": 4.23, "learning_rate": 9.886153846153846e-06, "loss": 0.1072, "step": 725 }, { "epoch": 4.37, "learning_rate": 9.873333333333334e-06, "loss": 0.1138, "step": 750 }, { "epoch": 4.52, "learning_rate": 9.860512820512821e-06, "loss": 0.1255, "step": 775 }, { "epoch": 4.66, "learning_rate": 9.847692307692308e-06, "loss": 0.1224, "step": 800 }, { "epoch": 4.81, "learning_rate": 9.834871794871795e-06, "loss": 0.1131, "step": 825 }, { "epoch": 4.96, "learning_rate": 9.822051282051283e-06, "loss": 0.1078, "step": 850 }, { "epoch": 5.1, "learning_rate": 9.80923076923077e-06, "loss": 0.092, "step": 875 }, { "epoch": 5.25, "learning_rate": 9.796410256410257e-06, "loss": 0.0891, "step": 900 }, { "epoch": 5.39, "learning_rate": 9.783589743589744e-06, "loss": 0.0808, "step": 925 }, { "epoch": 5.54, "learning_rate": 9.770769230769232e-06, "loss": 0.0897, "step": 950 }, { "epoch": 5.69, "learning_rate": 9.757948717948719e-06, "loss": 0.0853, "step": 975 }, { "epoch": 5.83, "learning_rate": 9.745128205128206e-06, "loss": 0.0761, "step": 1000 }, { "epoch": 5.83, "eval_loss": 0.1530763804912567, "eval_runtime": 1312.7715, "eval_samples_per_second": 4.873, "eval_steps_per_second": 0.305, "eval_wer": 6.095873828381697, "step": 1000 }, { "epoch": 5.98, "learning_rate": 9.732307692307693e-06, "loss": 0.0847, "step": 1025 }, { "epoch": 6.12, "learning_rate": 9.71948717948718e-06, "loss": 0.065, "step": 1050 }, { "epoch": 6.27, "learning_rate": 9.706666666666668e-06, "loss": 0.0654, "step": 1075 }, { "epoch": 6.41, "learning_rate": 9.693846153846155e-06, "loss": 0.0579, "step": 1100 }, { "epoch": 6.56, "learning_rate": 9.681025641025642e-06, "loss": 0.0604, "step": 1125 }, { "epoch": 6.71, "learning_rate": 9.668205128205129e-06, "loss": 0.0543, "step": 1150 }, { "epoch": 6.85, "learning_rate": 9.655384615384617e-06, "loss": 0.0567, "step": 1175 }, { "epoch": 7.0, "learning_rate": 9.642564102564104e-06, "loss": 0.0576, "step": 1200 }, { "epoch": 7.14, "learning_rate": 9.62974358974359e-06, "loss": 0.053, "step": 1225 }, { "epoch": 7.29, "learning_rate": 9.616923076923077e-06, "loss": 0.0485, "step": 1250 }, { "epoch": 7.43, "learning_rate": 9.604102564102566e-06, "loss": 0.0414, "step": 1275 }, { "epoch": 7.58, "learning_rate": 9.591282051282053e-06, "loss": 0.0401, "step": 1300 }, { "epoch": 7.73, "learning_rate": 9.578461538461538e-06, "loss": 0.04, "step": 1325 }, { "epoch": 7.87, "learning_rate": 9.565641025641026e-06, "loss": 0.0452, "step": 1350 }, { "epoch": 8.02, "learning_rate": 9.552820512820513e-06, "loss": 0.0471, "step": 1375 }, { "epoch": 8.16, "learning_rate": 9.54e-06, "loss": 0.0297, "step": 1400 }, { "epoch": 8.31, "learning_rate": 9.527179487179487e-06, "loss": 0.032, "step": 1425 }, { "epoch": 8.45, "learning_rate": 9.514358974358975e-06, "loss": 0.0288, "step": 1450 }, { "epoch": 8.6, "learning_rate": 9.501538461538462e-06, "loss": 0.0361, "step": 1475 }, { "epoch": 8.75, "learning_rate": 9.488717948717949e-06, "loss": 0.0284, "step": 1500 }, { "epoch": 8.89, "learning_rate": 9.475897435897436e-06, "loss": 0.0309, "step": 1525 }, { "epoch": 9.04, "learning_rate": 9.463076923076924e-06, "loss": 0.0287, "step": 1550 }, { "epoch": 9.18, "learning_rate": 9.450256410256411e-06, "loss": 0.024, "step": 1575 }, { "epoch": 9.33, "learning_rate": 9.437435897435898e-06, "loss": 0.0206, "step": 1600 }, { "epoch": 9.48, "learning_rate": 9.424615384615385e-06, "loss": 0.0244, "step": 1625 }, { "epoch": 9.62, "learning_rate": 9.411794871794872e-06, "loss": 0.0209, "step": 1650 }, { "epoch": 9.77, "learning_rate": 9.39897435897436e-06, "loss": 0.0221, "step": 1675 }, { "epoch": 9.91, "learning_rate": 9.386153846153847e-06, "loss": 0.0243, "step": 1700 }, { "epoch": 10.06, "learning_rate": 9.373333333333334e-06, "loss": 0.0213, "step": 1725 }, { "epoch": 10.2, "learning_rate": 9.36051282051282e-06, "loss": 0.0172, "step": 1750 }, { "epoch": 10.35, "learning_rate": 9.34769230769231e-06, "loss": 0.0155, "step": 1775 }, { "epoch": 10.5, "learning_rate": 9.334871794871796e-06, "loss": 0.016, "step": 1800 }, { "epoch": 10.64, "learning_rate": 9.322051282051283e-06, "loss": 0.0181, "step": 1825 }, { "epoch": 10.79, "learning_rate": 9.30923076923077e-06, "loss": 0.018, "step": 1850 }, { "epoch": 10.93, "learning_rate": 9.296410256410258e-06, "loss": 0.0224, "step": 1875 }, { "epoch": 11.08, "learning_rate": 9.283589743589745e-06, "loss": 0.0142, "step": 1900 }, { "epoch": 11.22, "learning_rate": 9.270769230769232e-06, "loss": 0.0148, "step": 1925 }, { "epoch": 11.37, "learning_rate": 9.257948717948719e-06, "loss": 0.013, "step": 1950 }, { "epoch": 11.52, "learning_rate": 9.245128205128206e-06, "loss": 0.0141, "step": 1975 }, { "epoch": 11.66, "learning_rate": 9.232307692307692e-06, "loss": 0.0148, "step": 2000 }, { "epoch": 11.66, "eval_loss": 0.1874324083328247, "eval_runtime": 1309.2409, "eval_samples_per_second": 4.886, "eval_steps_per_second": 0.306, "eval_wer": 5.732715387541936, "step": 2000 }, { "epoch": 11.81, "learning_rate": 9.21948717948718e-06, "loss": 0.0135, "step": 2025 }, { "epoch": 11.95, "learning_rate": 9.206666666666668e-06, "loss": 0.0148, "step": 2050 }, { "epoch": 12.1, "learning_rate": 9.193846153846155e-06, "loss": 0.0127, "step": 2075 }, { "epoch": 12.24, "learning_rate": 9.181025641025641e-06, "loss": 0.0146, "step": 2100 }, { "epoch": 12.39, "learning_rate": 9.168205128205128e-06, "loss": 0.0115, "step": 2125 }, { "epoch": 12.54, "learning_rate": 9.155384615384617e-06, "loss": 0.011, "step": 2150 }, { "epoch": 12.68, "learning_rate": 9.142564102564104e-06, "loss": 0.0098, "step": 2175 }, { "epoch": 12.83, "learning_rate": 9.12974358974359e-06, "loss": 0.0146, "step": 2200 }, { "epoch": 12.97, "learning_rate": 9.116923076923077e-06, "loss": 0.0133, "step": 2225 }, { "epoch": 13.12, "learning_rate": 9.104102564102566e-06, "loss": 0.0094, "step": 2250 }, { "epoch": 13.27, "learning_rate": 9.091282051282053e-06, "loss": 0.0103, "step": 2275 }, { "epoch": 13.41, "learning_rate": 9.07846153846154e-06, "loss": 0.0108, "step": 2300 }, { "epoch": 13.56, "learning_rate": 9.065641025641026e-06, "loss": 0.0078, "step": 2325 }, { "epoch": 13.7, "learning_rate": 9.052820512820513e-06, "loss": 0.0114, "step": 2350 }, { "epoch": 13.85, "learning_rate": 9.040000000000002e-06, "loss": 0.0096, "step": 2375 }, { "epoch": 13.99, "learning_rate": 9.027179487179488e-06, "loss": 0.01, "step": 2400 }, { "epoch": 14.14, "learning_rate": 9.014358974358975e-06, "loss": 0.0074, "step": 2425 }, { "epoch": 14.29, "learning_rate": 9.001538461538462e-06, "loss": 0.0087, "step": 2450 }, { "epoch": 14.43, "learning_rate": 8.98871794871795e-06, "loss": 0.0082, "step": 2475 }, { "epoch": 14.58, "learning_rate": 8.975897435897437e-06, "loss": 0.0075, "step": 2500 }, { "epoch": 14.72, "learning_rate": 8.963076923076924e-06, "loss": 0.0081, "step": 2525 }, { "epoch": 14.87, "learning_rate": 8.950256410256411e-06, "loss": 0.0102, "step": 2550 }, { "epoch": 15.01, "learning_rate": 8.937435897435898e-06, "loss": 0.0104, "step": 2575 }, { "epoch": 15.16, "learning_rate": 8.924615384615385e-06, "loss": 0.0086, "step": 2600 }, { "epoch": 15.31, "learning_rate": 8.911794871794871e-06, "loss": 0.0086, "step": 2625 }, { "epoch": 15.45, "learning_rate": 8.89897435897436e-06, "loss": 0.008, "step": 2650 }, { "epoch": 15.6, "learning_rate": 8.886153846153847e-06, "loss": 0.0083, "step": 2675 }, { "epoch": 15.74, "learning_rate": 8.873333333333334e-06, "loss": 0.0076, "step": 2700 }, { "epoch": 15.89, "learning_rate": 8.86051282051282e-06, "loss": 0.0069, "step": 2725 }, { "epoch": 16.03, "learning_rate": 8.847692307692309e-06, "loss": 0.0072, "step": 2750 }, { "epoch": 16.18, "learning_rate": 8.834871794871796e-06, "loss": 0.0064, "step": 2775 }, { "epoch": 16.33, "learning_rate": 8.822051282051283e-06, "loss": 0.0065, "step": 2800 }, { "epoch": 16.47, "learning_rate": 8.80923076923077e-06, "loss": 0.0067, "step": 2825 }, { "epoch": 16.62, "learning_rate": 8.796410256410258e-06, "loss": 0.0068, "step": 2850 }, { "epoch": 16.76, "learning_rate": 8.783589743589745e-06, "loss": 0.0071, "step": 2875 }, { "epoch": 16.91, "learning_rate": 8.770769230769232e-06, "loss": 0.0072, "step": 2900 }, { "epoch": 17.06, "learning_rate": 8.757948717948718e-06, "loss": 0.0067, "step": 2925 }, { "epoch": 17.2, "learning_rate": 8.745128205128205e-06, "loss": 0.0065, "step": 2950 }, { "epoch": 17.35, "learning_rate": 8.732307692307694e-06, "loss": 0.007, "step": 2975 }, { "epoch": 17.49, "learning_rate": 8.71948717948718e-06, "loss": 0.0076, "step": 3000 }, { "epoch": 17.49, "eval_loss": 0.2061571180820465, "eval_runtime": 1309.7163, "eval_samples_per_second": 4.884, "eval_steps_per_second": 0.305, "eval_wer": 5.758655276173347, "step": 3000 }, { "epoch": 17.64, "learning_rate": 8.706666666666667e-06, "loss": 0.0063, "step": 3025 }, { "epoch": 17.78, "learning_rate": 8.693846153846154e-06, "loss": 0.0059, "step": 3050 }, { "epoch": 17.93, "learning_rate": 8.681025641025643e-06, "loss": 0.0066, "step": 3075 }, { "epoch": 18.08, "learning_rate": 8.66820512820513e-06, "loss": 0.0059, "step": 3100 }, { "epoch": 18.22, "learning_rate": 8.655384615384616e-06, "loss": 0.0057, "step": 3125 }, { "epoch": 18.37, "learning_rate": 8.642564102564103e-06, "loss": 0.0055, "step": 3150 }, { "epoch": 18.51, "learning_rate": 8.62974358974359e-06, "loss": 0.0064, "step": 3175 }, { "epoch": 18.66, "learning_rate": 8.616923076923077e-06, "loss": 0.0044, "step": 3200 }, { "epoch": 18.8, "learning_rate": 8.604102564102564e-06, "loss": 0.0047, "step": 3225 }, { "epoch": 18.95, "learning_rate": 8.591282051282052e-06, "loss": 0.0063, "step": 3250 }, { "epoch": 19.1, "learning_rate": 8.578461538461539e-06, "loss": 0.005, "step": 3275 }, { "epoch": 19.24, "learning_rate": 8.565641025641026e-06, "loss": 0.0052, "step": 3300 }, { "epoch": 19.39, "learning_rate": 8.552820512820513e-06, "loss": 0.0052, "step": 3325 }, { "epoch": 19.53, "learning_rate": 8.540000000000001e-06, "loss": 0.0051, "step": 3350 }, { "epoch": 19.68, "learning_rate": 8.527179487179488e-06, "loss": 0.0048, "step": 3375 }, { "epoch": 19.83, "learning_rate": 8.514358974358975e-06, "loss": 0.0045, "step": 3400 }, { "epoch": 19.97, "learning_rate": 8.501538461538462e-06, "loss": 0.0055, "step": 3425 }, { "epoch": 20.12, "learning_rate": 8.48871794871795e-06, "loss": 0.004, "step": 3450 }, { "epoch": 20.26, "learning_rate": 8.475897435897437e-06, "loss": 0.0047, "step": 3475 }, { "epoch": 20.41, "learning_rate": 8.463076923076924e-06, "loss": 0.0051, "step": 3500 }, { "epoch": 20.55, "learning_rate": 8.45025641025641e-06, "loss": 0.0059, "step": 3525 }, { "epoch": 20.7, "learning_rate": 8.437435897435898e-06, "loss": 0.005, "step": 3550 }, { "epoch": 20.85, "learning_rate": 8.424615384615386e-06, "loss": 0.0054, "step": 3575 }, { "epoch": 20.99, "learning_rate": 8.411794871794873e-06, "loss": 0.0058, "step": 3600 }, { "epoch": 21.14, "learning_rate": 8.39897435897436e-06, "loss": 0.0069, "step": 3625 }, { "epoch": 21.28, "learning_rate": 8.386153846153847e-06, "loss": 0.0054, "step": 3650 }, { "epoch": 21.43, "learning_rate": 8.373333333333335e-06, "loss": 0.0049, "step": 3675 }, { "epoch": 21.57, "learning_rate": 8.360512820512822e-06, "loss": 0.0047, "step": 3700 }, { "epoch": 21.72, "learning_rate": 8.347692307692309e-06, "loss": 0.005, "step": 3725 }, { "epoch": 21.87, "learning_rate": 8.334871794871796e-06, "loss": 0.006, "step": 3750 }, { "epoch": 22.01, "learning_rate": 8.322051282051282e-06, "loss": 0.0054, "step": 3775 }, { "epoch": 22.16, "learning_rate": 8.30923076923077e-06, "loss": 0.0044, "step": 3800 }, { "epoch": 22.3, "learning_rate": 8.296410256410256e-06, "loss": 0.0035, "step": 3825 }, { "epoch": 22.45, "learning_rate": 8.283589743589745e-06, "loss": 0.0037, "step": 3850 }, { "epoch": 22.59, "learning_rate": 8.270769230769231e-06, "loss": 0.0038, "step": 3875 }, { "epoch": 22.74, "learning_rate": 8.257948717948718e-06, "loss": 0.0052, "step": 3900 }, { "epoch": 22.89, "learning_rate": 8.245128205128205e-06, "loss": 0.0048, "step": 3925 }, { "epoch": 23.03, "learning_rate": 8.232307692307694e-06, "loss": 0.0052, "step": 3950 }, { "epoch": 23.18, "learning_rate": 8.21948717948718e-06, "loss": 0.0042, "step": 3975 }, { "epoch": 23.32, "learning_rate": 8.206666666666667e-06, "loss": 0.0035, "step": 4000 }, { "epoch": 23.32, "eval_loss": 0.21957869827747345, "eval_runtime": 1309.1664, "eval_samples_per_second": 4.886, "eval_steps_per_second": 0.306, "eval_wer": 5.449105938505171, "step": 4000 }, { "epoch": 23.47, "learning_rate": 8.193846153846154e-06, "loss": 0.004, "step": 4025 }, { "epoch": 23.62, "learning_rate": 8.181025641025642e-06, "loss": 0.0044, "step": 4050 }, { "epoch": 23.76, "learning_rate": 8.16820512820513e-06, "loss": 0.0044, "step": 4075 }, { "epoch": 23.91, "learning_rate": 8.155384615384616e-06, "loss": 0.005, "step": 4100 }, { "epoch": 24.05, "learning_rate": 8.142564102564103e-06, "loss": 0.0053, "step": 4125 }, { "epoch": 24.2, "learning_rate": 8.12974358974359e-06, "loss": 0.0061, "step": 4150 }, { "epoch": 24.34, "learning_rate": 8.116923076923078e-06, "loss": 0.005, "step": 4175 }, { "epoch": 24.49, "learning_rate": 8.104102564102565e-06, "loss": 0.0047, "step": 4200 }, { "epoch": 24.64, "learning_rate": 8.091282051282052e-06, "loss": 0.0043, "step": 4225 }, { "epoch": 24.78, "learning_rate": 8.078461538461539e-06, "loss": 0.0071, "step": 4250 }, { "epoch": 24.93, "learning_rate": 8.065641025641027e-06, "loss": 0.0057, "step": 4275 }, { "epoch": 25.07, "learning_rate": 8.052820512820514e-06, "loss": 0.0049, "step": 4300 }, { "epoch": 25.22, "learning_rate": 8.040000000000001e-06, "loss": 0.0052, "step": 4325 }, { "epoch": 25.36, "learning_rate": 8.027179487179488e-06, "loss": 0.004, "step": 4350 }, { "epoch": 25.51, "learning_rate": 8.014358974358975e-06, "loss": 0.0047, "step": 4375 }, { "epoch": 25.66, "learning_rate": 8.001538461538461e-06, "loss": 0.0045, "step": 4400 }, { "epoch": 25.8, "learning_rate": 7.988717948717948e-06, "loss": 0.006, "step": 4425 }, { "epoch": 25.95, "learning_rate": 7.975897435897437e-06, "loss": 0.0037, "step": 4450 }, { "epoch": 26.09, "learning_rate": 7.963076923076924e-06, "loss": 0.0052, "step": 4475 }, { "epoch": 26.24, "learning_rate": 7.95025641025641e-06, "loss": 0.0035, "step": 4500 }, { "epoch": 26.38, "learning_rate": 7.937435897435897e-06, "loss": 0.0039, "step": 4525 }, { "epoch": 26.53, "learning_rate": 7.924615384615386e-06, "loss": 0.0031, "step": 4550 }, { "epoch": 26.68, "learning_rate": 7.911794871794873e-06, "loss": 0.0027, "step": 4575 }, { "epoch": 26.82, "learning_rate": 7.89897435897436e-06, "loss": 0.0028, "step": 4600 }, { "epoch": 26.97, "learning_rate": 7.886153846153846e-06, "loss": 0.0029, "step": 4625 }, { "epoch": 27.11, "learning_rate": 7.873333333333335e-06, "loss": 0.003, "step": 4650 }, { "epoch": 27.26, "learning_rate": 7.860512820512822e-06, "loss": 0.0028, "step": 4675 }, { "epoch": 27.41, "learning_rate": 7.847692307692308e-06, "loss": 0.0029, "step": 4700 }, { "epoch": 27.55, "learning_rate": 7.834871794871795e-06, "loss": 0.0031, "step": 4725 }, { "epoch": 27.7, "learning_rate": 7.822051282051282e-06, "loss": 0.003, "step": 4750 }, { "epoch": 27.84, "learning_rate": 7.80923076923077e-06, "loss": 0.0032, "step": 4775 }, { "epoch": 27.99, "learning_rate": 7.796410256410257e-06, "loss": 0.0026, "step": 4800 }, { "epoch": 28.13, "learning_rate": 7.783589743589744e-06, "loss": 0.0035, "step": 4825 }, { "epoch": 28.28, "learning_rate": 7.770769230769231e-06, "loss": 0.0028, "step": 4850 }, { "epoch": 28.43, "learning_rate": 7.75794871794872e-06, "loss": 0.0031, "step": 4875 }, { "epoch": 28.57, "learning_rate": 7.745128205128206e-06, "loss": 0.0043, "step": 4900 }, { "epoch": 28.72, "learning_rate": 7.732307692307693e-06, "loss": 0.0026, "step": 4925 }, { "epoch": 28.86, "learning_rate": 7.71948717948718e-06, "loss": 0.0028, "step": 4950 }, { "epoch": 29.01, "learning_rate": 7.706666666666669e-06, "loss": 0.0034, "step": 4975 }, { "epoch": 29.15, "learning_rate": 7.693846153846154e-06, "loss": 0.0029, "step": 5000 }, { "epoch": 29.15, "eval_loss": 0.2264678031206131, "eval_runtime": 1308.8714, "eval_samples_per_second": 4.887, "eval_steps_per_second": 0.306, "eval_wer": 5.589181337114792, "step": 5000 }, { "epoch": 29.3, "learning_rate": 7.68102564102564e-06, "loss": 0.0051, "step": 5025 }, { "epoch": 29.45, "learning_rate": 7.668205128205129e-06, "loss": 0.0031, "step": 5050 }, { "epoch": 29.59, "learning_rate": 7.655384615384616e-06, "loss": 0.0041, "step": 5075 }, { "epoch": 29.74, "learning_rate": 7.642564102564103e-06, "loss": 0.0032, "step": 5100 }, { "epoch": 29.88, "learning_rate": 7.62974358974359e-06, "loss": 0.0038, "step": 5125 }, { "epoch": 30.03, "learning_rate": 7.616923076923077e-06, "loss": 0.0037, "step": 5150 }, { "epoch": 30.17, "learning_rate": 7.604102564102565e-06, "loss": 0.0034, "step": 5175 }, { "epoch": 30.32, "learning_rate": 7.591282051282052e-06, "loss": 0.0047, "step": 5200 }, { "epoch": 30.47, "learning_rate": 7.578461538461539e-06, "loss": 0.0032, "step": 5225 }, { "epoch": 30.61, "learning_rate": 7.565641025641026e-06, "loss": 0.0032, "step": 5250 }, { "epoch": 30.76, "learning_rate": 7.552820512820514e-06, "loss": 0.0033, "step": 5275 }, { "epoch": 30.9, "learning_rate": 7.540000000000001e-06, "loss": 0.0032, "step": 5300 }, { "epoch": 31.05, "learning_rate": 7.5271794871794875e-06, "loss": 0.0041, "step": 5325 }, { "epoch": 31.2, "learning_rate": 7.514358974358975e-06, "loss": 0.0031, "step": 5350 }, { "epoch": 31.34, "learning_rate": 7.501538461538462e-06, "loss": 0.003, "step": 5375 }, { "epoch": 31.49, "learning_rate": 7.48871794871795e-06, "loss": 0.0039, "step": 5400 }, { "epoch": 31.63, "learning_rate": 7.4758974358974365e-06, "loss": 0.0033, "step": 5425 }, { "epoch": 31.78, "learning_rate": 7.463076923076924e-06, "loss": 0.0032, "step": 5450 }, { "epoch": 31.92, "learning_rate": 7.450256410256411e-06, "loss": 0.0025, "step": 5475 }, { "epoch": 32.07, "learning_rate": 7.437435897435899e-06, "loss": 0.0028, "step": 5500 }, { "epoch": 32.22, "learning_rate": 7.4246153846153855e-06, "loss": 0.0032, "step": 5525 }, { "epoch": 32.36, "learning_rate": 7.411794871794873e-06, "loss": 0.0038, "step": 5550 }, { "epoch": 32.51, "learning_rate": 7.39897435897436e-06, "loss": 0.0026, "step": 5575 }, { "epoch": 32.65, "learning_rate": 7.386153846153846e-06, "loss": 0.003, "step": 5600 }, { "epoch": 32.8, "learning_rate": 7.373333333333334e-06, "loss": 0.0028, "step": 5625 }, { "epoch": 32.94, "learning_rate": 7.3605128205128204e-06, "loss": 0.0029, "step": 5650 }, { "epoch": 33.09, "learning_rate": 7.347692307692308e-06, "loss": 0.0022, "step": 5675 }, { "epoch": 33.24, "learning_rate": 7.334871794871795e-06, "loss": 0.0026, "step": 5700 }, { "epoch": 33.38, "learning_rate": 7.322051282051283e-06, "loss": 0.0017, "step": 5725 }, { "epoch": 33.53, "learning_rate": 7.309230769230769e-06, "loss": 0.0024, "step": 5750 }, { "epoch": 33.67, "learning_rate": 7.296410256410257e-06, "loss": 0.0026, "step": 5775 }, { "epoch": 33.82, "learning_rate": 7.283589743589744e-06, "loss": 0.003, "step": 5800 }, { "epoch": 33.97, "learning_rate": 7.270769230769232e-06, "loss": 0.0027, "step": 5825 }, { "epoch": 34.11, "learning_rate": 7.257948717948718e-06, "loss": 0.0025, "step": 5850 }, { "epoch": 34.26, "learning_rate": 7.245128205128206e-06, "loss": 0.0035, "step": 5875 }, { "epoch": 34.4, "learning_rate": 7.232307692307693e-06, "loss": 0.0034, "step": 5900 }, { "epoch": 34.55, "learning_rate": 7.21948717948718e-06, "loss": 0.0031, "step": 5925 }, { "epoch": 34.69, "learning_rate": 7.206666666666667e-06, "loss": 0.0028, "step": 5950 }, { "epoch": 34.84, "learning_rate": 7.193846153846154e-06, "loss": 0.0024, "step": 5975 }, { "epoch": 34.99, "learning_rate": 7.181025641025642e-06, "loss": 0.0027, "step": 6000 }, { "epoch": 34.99, "eval_loss": 0.23758915066719055, "eval_runtime": 1303.8706, "eval_samples_per_second": 4.906, "eval_steps_per_second": 0.307, "eval_wer": 5.836474942067582, "step": 6000 }, { "epoch": 35.13, "learning_rate": 7.168205128205129e-06, "loss": 0.003, "step": 6025 }, { "epoch": 35.28, "learning_rate": 7.155384615384616e-06, "loss": 0.0026, "step": 6050 }, { "epoch": 35.42, "learning_rate": 7.142564102564103e-06, "loss": 0.0022, "step": 6075 }, { "epoch": 35.57, "learning_rate": 7.129743589743591e-06, "loss": 0.0023, "step": 6100 }, { "epoch": 35.71, "learning_rate": 7.116923076923078e-06, "loss": 0.0031, "step": 6125 }, { "epoch": 35.86, "learning_rate": 7.104102564102565e-06, "loss": 0.0019, "step": 6150 }, { "epoch": 36.01, "learning_rate": 7.091282051282052e-06, "loss": 0.002, "step": 6175 }, { "epoch": 36.15, "learning_rate": 7.078461538461538e-06, "loss": 0.0026, "step": 6200 }, { "epoch": 36.3, "learning_rate": 7.065641025641026e-06, "loss": 0.0026, "step": 6225 }, { "epoch": 36.44, "learning_rate": 7.052820512820513e-06, "loss": 0.0023, "step": 6250 }, { "epoch": 36.59, "learning_rate": 7.04e-06, "loss": 0.0019, "step": 6275 }, { "epoch": 36.73, "learning_rate": 7.027179487179487e-06, "loss": 0.003, "step": 6300 }, { "epoch": 36.88, "learning_rate": 7.014358974358975e-06, "loss": 0.0024, "step": 6325 }, { "epoch": 37.03, "learning_rate": 7.001538461538462e-06, "loss": 0.0026, "step": 6350 }, { "epoch": 37.17, "learning_rate": 6.988717948717949e-06, "loss": 0.002, "step": 6375 }, { "epoch": 37.32, "learning_rate": 6.975897435897436e-06, "loss": 0.0024, "step": 6400 }, { "epoch": 37.46, "learning_rate": 6.963076923076924e-06, "loss": 0.002, "step": 6425 }, { "epoch": 37.61, "learning_rate": 6.950256410256411e-06, "loss": 0.0014, "step": 6450 }, { "epoch": 37.76, "learning_rate": 6.937435897435898e-06, "loss": 0.0017, "step": 6475 }, { "epoch": 37.9, "learning_rate": 6.924615384615385e-06, "loss": 0.0017, "step": 6500 }, { "epoch": 38.05, "learning_rate": 6.911794871794872e-06, "loss": 0.0017, "step": 6525 }, { "epoch": 38.19, "learning_rate": 6.89897435897436e-06, "loss": 0.0017, "step": 6550 }, { "epoch": 38.34, "learning_rate": 6.8861538461538465e-06, "loss": 0.0013, "step": 6575 }, { "epoch": 38.48, "learning_rate": 6.873333333333334e-06, "loss": 0.0017, "step": 6600 }, { "epoch": 38.63, "learning_rate": 6.860512820512821e-06, "loss": 0.0017, "step": 6625 }, { "epoch": 38.78, "learning_rate": 6.847692307692309e-06, "loss": 0.0016, "step": 6650 }, { "epoch": 38.92, "learning_rate": 6.8348717948717955e-06, "loss": 0.0014, "step": 6675 }, { "epoch": 39.07, "learning_rate": 6.822051282051283e-06, "loss": 0.0017, "step": 6700 }, { "epoch": 39.21, "learning_rate": 6.80923076923077e-06, "loss": 0.002, "step": 6725 }, { "epoch": 39.36, "learning_rate": 6.796410256410258e-06, "loss": 0.0021, "step": 6750 }, { "epoch": 39.5, "learning_rate": 6.7835897435897445e-06, "loss": 0.0025, "step": 6775 }, { "epoch": 39.65, "learning_rate": 6.770769230769232e-06, "loss": 0.002, "step": 6800 }, { "epoch": 39.8, "learning_rate": 6.757948717948718e-06, "loss": 0.0024, "step": 6825 }, { "epoch": 39.94, "learning_rate": 6.745128205128205e-06, "loss": 0.0033, "step": 6850 }, { "epoch": 40.09, "learning_rate": 6.732307692307693e-06, "loss": 0.0028, "step": 6875 }, { "epoch": 40.23, "learning_rate": 6.7194871794871794e-06, "loss": 0.0021, "step": 6900 }, { "epoch": 40.38, "learning_rate": 6.706666666666667e-06, "loss": 0.0024, "step": 6925 }, { "epoch": 40.52, "learning_rate": 6.693846153846154e-06, "loss": 0.0027, "step": 6950 }, { "epoch": 40.67, "learning_rate": 6.681025641025642e-06, "loss": 0.0032, "step": 6975 }, { "epoch": 40.82, "learning_rate": 6.6682051282051284e-06, "loss": 0.0028, "step": 7000 }, { "epoch": 40.82, "eval_loss": 0.23964348435401917, "eval_runtime": 1302.1874, "eval_samples_per_second": 4.913, "eval_steps_per_second": 0.307, "eval_wer": 5.69639954345796, "step": 7000 }, { "epoch": 40.96, "learning_rate": 6.655384615384616e-06, "loss": 0.0022, "step": 7025 }, { "epoch": 41.11, "learning_rate": 6.642564102564103e-06, "loss": 0.0029, "step": 7050 }, { "epoch": 41.25, "learning_rate": 6.629743589743591e-06, "loss": 0.0024, "step": 7075 }, { "epoch": 41.4, "learning_rate": 6.616923076923077e-06, "loss": 0.0023, "step": 7100 }, { "epoch": 41.55, "learning_rate": 6.604102564102565e-06, "loss": 0.0021, "step": 7125 }, { "epoch": 41.69, "learning_rate": 6.591282051282052e-06, "loss": 0.0017, "step": 7150 }, { "epoch": 41.84, "learning_rate": 6.578461538461539e-06, "loss": 0.0018, "step": 7175 }, { "epoch": 41.98, "learning_rate": 6.565641025641026e-06, "loss": 0.002, "step": 7200 }, { "epoch": 42.13, "learning_rate": 6.552820512820513e-06, "loss": 0.0017, "step": 7225 }, { "epoch": 42.27, "learning_rate": 6.540000000000001e-06, "loss": 0.0019, "step": 7250 }, { "epoch": 42.42, "learning_rate": 6.527179487179488e-06, "loss": 0.0016, "step": 7275 }, { "epoch": 42.57, "learning_rate": 6.514358974358975e-06, "loss": 0.0023, "step": 7300 }, { "epoch": 42.71, "learning_rate": 6.501538461538462e-06, "loss": 0.0019, "step": 7325 }, { "epoch": 42.86, "learning_rate": 6.48871794871795e-06, "loss": 0.0024, "step": 7350 }, { "epoch": 43.0, "learning_rate": 6.475897435897437e-06, "loss": 0.0019, "step": 7375 }, { "epoch": 43.15, "learning_rate": 6.463076923076924e-06, "loss": 0.0016, "step": 7400 }, { "epoch": 43.29, "learning_rate": 6.45025641025641e-06, "loss": 0.0022, "step": 7425 }, { "epoch": 43.44, "learning_rate": 6.437435897435897e-06, "loss": 0.0022, "step": 7450 }, { "epoch": 43.59, "learning_rate": 6.424615384615385e-06, "loss": 0.0015, "step": 7475 }, { "epoch": 43.73, "learning_rate": 6.411794871794872e-06, "loss": 0.0013, "step": 7500 }, { "epoch": 43.88, "learning_rate": 6.398974358974359e-06, "loss": 0.0012, "step": 7525 }, { "epoch": 44.02, "learning_rate": 6.386153846153846e-06, "loss": 0.0016, "step": 7550 }, { "epoch": 44.17, "learning_rate": 6.373333333333334e-06, "loss": 0.0021, "step": 7575 }, { "epoch": 44.31, "learning_rate": 6.360512820512821e-06, "loss": 0.0024, "step": 7600 }, { "epoch": 44.46, "learning_rate": 6.347692307692308e-06, "loss": 0.0022, "step": 7625 }, { "epoch": 44.61, "learning_rate": 6.334871794871795e-06, "loss": 0.0015, "step": 7650 }, { "epoch": 44.75, "learning_rate": 6.322051282051283e-06, "loss": 0.0014, "step": 7675 }, { "epoch": 44.9, "learning_rate": 6.30923076923077e-06, "loss": 0.0017, "step": 7700 }, { "epoch": 45.04, "learning_rate": 6.296410256410257e-06, "loss": 0.0014, "step": 7725 }, { "epoch": 45.19, "learning_rate": 6.283589743589744e-06, "loss": 0.0011, "step": 7750 }, { "epoch": 45.34, "learning_rate": 6.270769230769231e-06, "loss": 0.0011, "step": 7775 }, { "epoch": 45.48, "learning_rate": 6.257948717948719e-06, "loss": 0.0018, "step": 7800 }, { "epoch": 45.63, "learning_rate": 6.2451282051282055e-06, "loss": 0.0011, "step": 7825 }, { "epoch": 45.77, "learning_rate": 6.232307692307693e-06, "loss": 0.0014, "step": 7850 }, { "epoch": 45.92, "learning_rate": 6.21948717948718e-06, "loss": 0.0015, "step": 7875 }, { "epoch": 46.06, "learning_rate": 6.206666666666668e-06, "loss": 0.002, "step": 7900 }, { "epoch": 46.21, "learning_rate": 6.1938461538461545e-06, "loss": 0.0012, "step": 7925 }, { "epoch": 46.36, "learning_rate": 6.181025641025642e-06, "loss": 0.0011, "step": 7950 }, { "epoch": 46.5, "learning_rate": 6.168205128205129e-06, "loss": 0.0019, "step": 7975 }, { "epoch": 46.65, "learning_rate": 6.155384615384617e-06, "loss": 0.0021, "step": 8000 }, { "epoch": 46.65, "eval_loss": 0.24486812949180603, "eval_runtime": 1302.806, "eval_samples_per_second": 4.91, "eval_steps_per_second": 0.307, "eval_wer": 5.481963130771625, "step": 8000 }, { "epoch": 46.79, "learning_rate": 6.142564102564103e-06, "loss": 0.0018, "step": 8025 }, { "epoch": 46.94, "learning_rate": 6.1297435897435895e-06, "loss": 0.0012, "step": 8050 }, { "epoch": 47.08, "learning_rate": 6.116923076923077e-06, "loss": 0.002, "step": 8075 }, { "epoch": 47.23, "learning_rate": 6.104102564102564e-06, "loss": 0.0015, "step": 8100 }, { "epoch": 47.38, "learning_rate": 6.091282051282052e-06, "loss": 0.0013, "step": 8125 }, { "epoch": 47.52, "learning_rate": 6.0784615384615384e-06, "loss": 0.0024, "step": 8150 }, { "epoch": 47.67, "learning_rate": 6.065641025641026e-06, "loss": 0.0019, "step": 8175 }, { "epoch": 47.81, "learning_rate": 6.052820512820513e-06, "loss": 0.0023, "step": 8200 }, { "epoch": 47.96, "learning_rate": 6.040000000000001e-06, "loss": 0.0018, "step": 8225 }, { "epoch": 48.1, "learning_rate": 6.0271794871794874e-06, "loss": 0.0019, "step": 8250 }, { "epoch": 48.25, "learning_rate": 6.014358974358975e-06, "loss": 0.0018, "step": 8275 }, { "epoch": 48.4, "learning_rate": 6.001538461538462e-06, "loss": 0.0014, "step": 8300 }, { "epoch": 48.54, "learning_rate": 5.98871794871795e-06, "loss": 0.0021, "step": 8325 }, { "epoch": 48.69, "learning_rate": 5.9758974358974364e-06, "loss": 0.0021, "step": 8350 }, { "epoch": 48.83, "learning_rate": 5.963076923076923e-06, "loss": 0.0018, "step": 8375 }, { "epoch": 48.98, "learning_rate": 5.950256410256411e-06, "loss": 0.0016, "step": 8400 }, { "epoch": 49.13, "learning_rate": 5.937435897435898e-06, "loss": 0.0014, "step": 8425 }, { "epoch": 49.27, "learning_rate": 5.9246153846153854e-06, "loss": 0.0013, "step": 8450 }, { "epoch": 49.42, "learning_rate": 5.911794871794872e-06, "loss": 0.0018, "step": 8475 }, { "epoch": 49.56, "learning_rate": 5.89897435897436e-06, "loss": 0.0021, "step": 8500 }, { "epoch": 49.71, "learning_rate": 5.886153846153847e-06, "loss": 0.0018, "step": 8525 }, { "epoch": 49.85, "learning_rate": 5.873333333333334e-06, "loss": 0.0011, "step": 8550 }, { "epoch": 50.0, "learning_rate": 5.860512820512821e-06, "loss": 0.0016, "step": 8575 }, { "epoch": 50.15, "learning_rate": 5.847692307692309e-06, "loss": 0.0011, "step": 8600 }, { "epoch": 50.29, "learning_rate": 5.834871794871796e-06, "loss": 0.0009, "step": 8625 }, { "epoch": 50.44, "learning_rate": 5.822051282051282e-06, "loss": 0.0014, "step": 8650 }, { "epoch": 50.58, "learning_rate": 5.809230769230769e-06, "loss": 0.0012, "step": 8675 }, { "epoch": 50.73, "learning_rate": 5.796410256410256e-06, "loss": 0.001, "step": 8700 }, { "epoch": 50.87, "learning_rate": 5.783589743589744e-06, "loss": 0.0011, "step": 8725 }, { "epoch": 51.02, "learning_rate": 5.770769230769231e-06, "loss": 0.0009, "step": 8750 }, { "epoch": 51.17, "learning_rate": 5.757948717948718e-06, "loss": 0.0008, "step": 8775 }, { "epoch": 51.31, "learning_rate": 5.745128205128205e-06, "loss": 0.0009, "step": 8800 }, { "epoch": 51.46, "learning_rate": 5.732307692307693e-06, "loss": 0.0012, "step": 8825 }, { "epoch": 51.6, "learning_rate": 5.71948717948718e-06, "loss": 0.0007, "step": 8850 }, { "epoch": 51.75, "learning_rate": 5.706666666666667e-06, "loss": 0.0006, "step": 8875 }, { "epoch": 51.9, "learning_rate": 5.693846153846154e-06, "loss": 0.0007, "step": 8900 }, { "epoch": 52.04, "learning_rate": 5.681025641025642e-06, "loss": 0.0008, "step": 8925 }, { "epoch": 52.19, "learning_rate": 5.668205128205129e-06, "loss": 0.0008, "step": 8950 }, { "epoch": 52.33, "learning_rate": 5.6553846153846155e-06, "loss": 0.0007, "step": 8975 }, { "epoch": 52.48, "learning_rate": 5.642564102564103e-06, "loss": 0.0012, "step": 9000 }, { "epoch": 52.48, "eval_loss": 0.2438248246908188, "eval_runtime": 1302.7379, "eval_samples_per_second": 4.91, "eval_steps_per_second": 0.307, "eval_wer": 5.449105938505171, "step": 9000 }, { "epoch": 52.62, "learning_rate": 5.62974358974359e-06, "loss": 0.0015, "step": 9025 }, { "epoch": 52.77, "learning_rate": 5.616923076923078e-06, "loss": 0.001, "step": 9050 }, { "epoch": 52.92, "learning_rate": 5.6041025641025645e-06, "loss": 0.0009, "step": 9075 }, { "epoch": 53.06, "learning_rate": 5.591282051282052e-06, "loss": 0.0012, "step": 9100 }, { "epoch": 53.21, "learning_rate": 5.578461538461539e-06, "loss": 0.0013, "step": 9125 }, { "epoch": 53.35, "learning_rate": 5.565641025641027e-06, "loss": 0.0015, "step": 9150 }, { "epoch": 53.5, "learning_rate": 5.5528205128205135e-06, "loss": 0.0014, "step": 9175 }, { "epoch": 53.64, "learning_rate": 5.540000000000001e-06, "loss": 0.0017, "step": 9200 }, { "epoch": 53.79, "learning_rate": 5.527179487179488e-06, "loss": 0.0012, "step": 9225 }, { "epoch": 53.94, "learning_rate": 5.514358974358974e-06, "loss": 0.0011, "step": 9250 }, { "epoch": 54.08, "learning_rate": 5.501538461538462e-06, "loss": 0.0007, "step": 9275 }, { "epoch": 54.23, "learning_rate": 5.4887179487179485e-06, "loss": 0.0008, "step": 9300 }, { "epoch": 54.37, "learning_rate": 5.475897435897436e-06, "loss": 0.0014, "step": 9325 }, { "epoch": 54.52, "learning_rate": 5.463076923076923e-06, "loss": 0.0014, "step": 9350 }, { "epoch": 54.66, "learning_rate": 5.450256410256411e-06, "loss": 0.0018, "step": 9375 }, { "epoch": 54.81, "learning_rate": 5.4374358974358975e-06, "loss": 0.001, "step": 9400 }, { "epoch": 54.96, "learning_rate": 5.424615384615385e-06, "loss": 0.0016, "step": 9425 }, { "epoch": 55.1, "learning_rate": 5.411794871794872e-06, "loss": 0.0016, "step": 9450 }, { "epoch": 55.25, "learning_rate": 5.39897435897436e-06, "loss": 0.0009, "step": 9475 }, { "epoch": 55.39, "learning_rate": 5.3861538461538464e-06, "loss": 0.0007, "step": 9500 }, { "epoch": 55.54, "learning_rate": 5.373333333333334e-06, "loss": 0.0009, "step": 9525 }, { "epoch": 55.69, "learning_rate": 5.360512820512821e-06, "loss": 0.0013, "step": 9550 }, { "epoch": 55.83, "learning_rate": 5.347692307692308e-06, "loss": 0.0021, "step": 9575 }, { "epoch": 55.98, "learning_rate": 5.3348717948717954e-06, "loss": 0.0014, "step": 9600 }, { "epoch": 56.12, "learning_rate": 5.322051282051282e-06, "loss": 0.0015, "step": 9625 }, { "epoch": 56.27, "learning_rate": 5.30923076923077e-06, "loss": 0.0017, "step": 9650 }, { "epoch": 56.41, "learning_rate": 5.296410256410257e-06, "loss": 0.0024, "step": 9675 }, { "epoch": 56.56, "learning_rate": 5.2835897435897444e-06, "loss": 0.0022, "step": 9700 }, { "epoch": 56.71, "learning_rate": 5.270769230769231e-06, "loss": 0.0021, "step": 9725 }, { "epoch": 56.85, "learning_rate": 5.257948717948719e-06, "loss": 0.0027, "step": 9750 }, { "epoch": 57.0, "learning_rate": 5.245128205128206e-06, "loss": 0.0019, "step": 9775 }, { "epoch": 57.14, "learning_rate": 5.2323076923076934e-06, "loss": 0.0024, "step": 9800 }, { "epoch": 57.29, "learning_rate": 5.21948717948718e-06, "loss": 0.0012, "step": 9825 }, { "epoch": 57.43, "learning_rate": 5.206666666666668e-06, "loss": 0.0014, "step": 9850 }, { "epoch": 57.58, "learning_rate": 5.193846153846154e-06, "loss": 0.0016, "step": 9875 }, { "epoch": 57.73, "learning_rate": 5.181025641025641e-06, "loss": 0.0011, "step": 9900 }, { "epoch": 57.87, "learning_rate": 5.168205128205128e-06, "loss": 0.0013, "step": 9925 }, { "epoch": 58.02, "learning_rate": 5.155384615384615e-06, "loss": 0.0017, "step": 9950 }, { "epoch": 58.16, "learning_rate": 5.142564102564103e-06, "loss": 0.0011, "step": 9975 }, { "epoch": 58.31, "learning_rate": 5.12974358974359e-06, "loss": 0.0014, "step": 10000 }, { "epoch": 58.31, "eval_loss": 0.24899841845035553, "eval_runtime": 1306.8079, "eval_samples_per_second": 4.895, "eval_steps_per_second": 0.306, "eval_wer": 5.558053470757099, "step": 10000 }, { "epoch": 58.45, "learning_rate": 5.116923076923077e-06, "loss": 0.0016, "step": 10025 }, { "epoch": 58.6, "learning_rate": 5.104102564102564e-06, "loss": 0.0016, "step": 10050 }, { "epoch": 58.75, "learning_rate": 5.091282051282052e-06, "loss": 0.0023, "step": 10075 }, { "epoch": 58.89, "learning_rate": 5.078461538461539e-06, "loss": 0.0015, "step": 10100 }, { "epoch": 59.04, "learning_rate": 5.065641025641026e-06, "loss": 0.0013, "step": 10125 }, { "epoch": 59.18, "learning_rate": 5.052820512820513e-06, "loss": 0.0008, "step": 10150 }, { "epoch": 59.33, "learning_rate": 5.04e-06, "loss": 0.0008, "step": 10175 }, { "epoch": 59.48, "learning_rate": 5.027179487179488e-06, "loss": 0.001, "step": 10200 }, { "epoch": 59.62, "learning_rate": 5.0143589743589745e-06, "loss": 0.0015, "step": 10225 }, { "epoch": 59.77, "learning_rate": 5.001538461538462e-06, "loss": 0.0019, "step": 10250 }, { "epoch": 59.91, "learning_rate": 4.988717948717949e-06, "loss": 0.0012, "step": 10275 }, { "epoch": 60.06, "learning_rate": 4.975897435897437e-06, "loss": 0.0012, "step": 10300 }, { "epoch": 60.2, "learning_rate": 4.9630769230769235e-06, "loss": 0.0008, "step": 10325 }, { "epoch": 60.35, "learning_rate": 4.95025641025641e-06, "loss": 0.001, "step": 10350 }, { "epoch": 60.5, "learning_rate": 4.937435897435898e-06, "loss": 0.0015, "step": 10375 }, { "epoch": 60.64, "learning_rate": 4.924615384615385e-06, "loss": 0.0013, "step": 10400 }, { "epoch": 60.79, "learning_rate": 4.9117948717948725e-06, "loss": 0.0009, "step": 10425 }, { "epoch": 60.93, "learning_rate": 4.898974358974359e-06, "loss": 0.0008, "step": 10450 }, { "epoch": 61.08, "learning_rate": 4.886153846153847e-06, "loss": 0.0009, "step": 10475 }, { "epoch": 61.22, "learning_rate": 4.873333333333334e-06, "loss": 0.0006, "step": 10500 }, { "epoch": 61.37, "learning_rate": 4.860512820512821e-06, "loss": 0.0009, "step": 10525 }, { "epoch": 61.52, "learning_rate": 4.847692307692308e-06, "loss": 0.0008, "step": 10550 }, { "epoch": 61.66, "learning_rate": 4.834871794871795e-06, "loss": 0.0012, "step": 10575 }, { "epoch": 61.81, "learning_rate": 4.822051282051283e-06, "loss": 0.001, "step": 10600 }, { "epoch": 61.95, "learning_rate": 4.80923076923077e-06, "loss": 0.0008, "step": 10625 }, { "epoch": 62.1, "learning_rate": 4.7964102564102565e-06, "loss": 0.0007, "step": 10650 }, { "epoch": 62.24, "learning_rate": 4.783589743589744e-06, "loss": 0.0013, "step": 10675 }, { "epoch": 62.39, "learning_rate": 4.770769230769231e-06, "loss": 0.0008, "step": 10700 }, { "epoch": 62.54, "learning_rate": 4.757948717948719e-06, "loss": 0.0009, "step": 10725 }, { "epoch": 62.68, "learning_rate": 4.7451282051282055e-06, "loss": 0.0009, "step": 10750 }, { "epoch": 62.83, "learning_rate": 4.732307692307693e-06, "loss": 0.0009, "step": 10775 }, { "epoch": 62.97, "learning_rate": 4.71948717948718e-06, "loss": 0.0007, "step": 10800 }, { "epoch": 63.12, "learning_rate": 4.706666666666667e-06, "loss": 0.0005, "step": 10825 }, { "epoch": 63.27, "learning_rate": 4.6938461538461544e-06, "loss": 0.0005, "step": 10850 }, { "epoch": 63.41, "learning_rate": 4.681025641025641e-06, "loss": 0.0004, "step": 10875 }, { "epoch": 63.56, "learning_rate": 4.668205128205129e-06, "loss": 0.0007, "step": 10900 }, { "epoch": 63.7, "learning_rate": 4.655384615384616e-06, "loss": 0.0004, "step": 10925 }, { "epoch": 63.85, "learning_rate": 4.642564102564103e-06, "loss": 0.0005, "step": 10950 }, { "epoch": 63.99, "learning_rate": 4.62974358974359e-06, "loss": 0.0012, "step": 10975 }, { "epoch": 64.14, "learning_rate": 4.616923076923077e-06, "loss": 0.0009, "step": 11000 }, { "epoch": 64.14, "eval_loss": 0.2461892068386078, "eval_runtime": 1302.5891, "eval_samples_per_second": 4.911, "eval_steps_per_second": 0.307, "eval_wer": 5.3695569467021755, "step": 11000 }, { "epoch": 64.29, "learning_rate": 4.604102564102565e-06, "loss": 0.0009, "step": 11025 }, { "epoch": 64.43, "learning_rate": 4.591282051282052e-06, "loss": 0.0004, "step": 11050 }, { "epoch": 64.58, "learning_rate": 4.578461538461539e-06, "loss": 0.0004, "step": 11075 }, { "epoch": 64.72, "learning_rate": 4.565641025641026e-06, "loss": 0.0005, "step": 11100 }, { "epoch": 64.87, "learning_rate": 4.552820512820513e-06, "loss": 0.0005, "step": 11125 }, { "epoch": 65.01, "learning_rate": 4.540000000000001e-06, "loss": 0.0004, "step": 11150 }, { "epoch": 65.16, "learning_rate": 4.527179487179487e-06, "loss": 0.0003, "step": 11175 }, { "epoch": 65.31, "learning_rate": 4.514358974358975e-06, "loss": 0.0004, "step": 11200 }, { "epoch": 65.45, "learning_rate": 4.501538461538462e-06, "loss": 0.0006, "step": 11225 }, { "epoch": 65.6, "learning_rate": 4.488717948717949e-06, "loss": 0.0006, "step": 11250 }, { "epoch": 65.74, "learning_rate": 4.476410256410257e-06, "loss": 0.0013, "step": 11275 }, { "epoch": 65.89, "learning_rate": 4.463589743589744e-06, "loss": 0.0011, "step": 11300 }, { "epoch": 66.03, "learning_rate": 4.4507692307692315e-06, "loss": 0.0009, "step": 11325 }, { "epoch": 66.18, "learning_rate": 4.437948717948718e-06, "loss": 0.0008, "step": 11350 }, { "epoch": 66.33, "learning_rate": 4.425128205128206e-06, "loss": 0.0008, "step": 11375 }, { "epoch": 66.47, "learning_rate": 4.412307692307693e-06, "loss": 0.0007, "step": 11400 }, { "epoch": 66.62, "learning_rate": 4.39948717948718e-06, "loss": 0.0009, "step": 11425 }, { "epoch": 66.76, "learning_rate": 4.3866666666666665e-06, "loss": 0.0018, "step": 11450 }, { "epoch": 66.91, "learning_rate": 4.373846153846154e-06, "loss": 0.0019, "step": 11475 }, { "epoch": 67.06, "learning_rate": 4.361025641025641e-06, "loss": 0.0009, "step": 11500 }, { "epoch": 67.2, "learning_rate": 4.348205128205129e-06, "loss": 0.0012, "step": 11525 }, { "epoch": 67.35, "learning_rate": 4.3353846153846154e-06, "loss": 0.0007, "step": 11550 }, { "epoch": 67.49, "learning_rate": 4.322564102564103e-06, "loss": 0.0005, "step": 11575 }, { "epoch": 67.64, "learning_rate": 4.30974358974359e-06, "loss": 0.0009, "step": 11600 }, { "epoch": 67.78, "learning_rate": 4.296923076923078e-06, "loss": 0.0007, "step": 11625 }, { "epoch": 67.93, "learning_rate": 4.2841025641025644e-06, "loss": 0.0012, "step": 11650 }, { "epoch": 68.08, "learning_rate": 4.271282051282052e-06, "loss": 0.0015, "step": 11675 }, { "epoch": 68.22, "learning_rate": 4.258461538461539e-06, "loss": 0.0006, "step": 11700 }, { "epoch": 68.37, "learning_rate": 4.245641025641026e-06, "loss": 0.0005, "step": 11725 }, { "epoch": 68.51, "learning_rate": 4.232820512820513e-06, "loss": 0.0007, "step": 11750 }, { "epoch": 68.66, "learning_rate": 4.22e-06, "loss": 0.0004, "step": 11775 }, { "epoch": 68.8, "learning_rate": 4.207179487179487e-06, "loss": 0.0008, "step": 11800 }, { "epoch": 68.95, "learning_rate": 4.194358974358975e-06, "loss": 0.0007, "step": 11825 }, { "epoch": 69.1, "learning_rate": 4.1815384615384616e-06, "loss": 0.0008, "step": 11850 }, { "epoch": 69.24, "learning_rate": 4.168717948717949e-06, "loss": 0.0013, "step": 11875 }, { "epoch": 69.39, "learning_rate": 4.155897435897436e-06, "loss": 0.0006, "step": 11900 }, { "epoch": 69.53, "learning_rate": 4.143076923076924e-06, "loss": 0.0005, "step": 11925 }, { "epoch": 69.68, "learning_rate": 4.1302564102564106e-06, "loss": 0.0005, "step": 11950 }, { "epoch": 69.83, "learning_rate": 4.117435897435898e-06, "loss": 0.0014, "step": 11975 }, { "epoch": 69.97, "learning_rate": 4.104615384615385e-06, "loss": 0.0006, "step": 12000 }, { "epoch": 69.97, "eval_loss": 0.2597556710243225, "eval_runtime": 1303.3583, "eval_samples_per_second": 4.908, "eval_steps_per_second": 0.307, "eval_wer": 5.630685158925051, "step": 12000 }, { "epoch": 70.12, "learning_rate": 4.091794871794872e-06, "loss": 0.0006, "step": 12025 }, { "epoch": 70.26, "learning_rate": 4.078974358974359e-06, "loss": 0.0006, "step": 12050 }, { "epoch": 70.41, "learning_rate": 4.066153846153846e-06, "loss": 0.0007, "step": 12075 }, { "epoch": 70.55, "learning_rate": 4.053333333333333e-06, "loss": 0.0007, "step": 12100 }, { "epoch": 70.7, "learning_rate": 4.0410256410256415e-06, "loss": 0.0009, "step": 12125 }, { "epoch": 70.85, "learning_rate": 4.028205128205129e-06, "loss": 0.0005, "step": 12150 }, { "epoch": 70.99, "learning_rate": 4.015384615384615e-06, "loss": 0.0005, "step": 12175 }, { "epoch": 71.14, "learning_rate": 4.002564102564103e-06, "loss": 0.0005, "step": 12200 }, { "epoch": 71.28, "learning_rate": 3.98974358974359e-06, "loss": 0.0007, "step": 12225 }, { "epoch": 71.43, "learning_rate": 3.976923076923077e-06, "loss": 0.0004, "step": 12250 }, { "epoch": 71.57, "learning_rate": 3.964102564102564e-06, "loss": 0.0007, "step": 12275 }, { "epoch": 71.72, "learning_rate": 3.951282051282052e-06, "loss": 0.0005, "step": 12300 }, { "epoch": 71.87, "learning_rate": 3.938461538461539e-06, "loss": 0.0005, "step": 12325 }, { "epoch": 72.01, "learning_rate": 3.925641025641026e-06, "loss": 0.0005, "step": 12350 }, { "epoch": 72.16, "learning_rate": 3.912820512820513e-06, "loss": 0.0007, "step": 12375 }, { "epoch": 72.3, "learning_rate": 3.900000000000001e-06, "loss": 0.0007, "step": 12400 }, { "epoch": 72.45, "learning_rate": 3.887179487179488e-06, "loss": 0.0004, "step": 12425 }, { "epoch": 72.59, "learning_rate": 3.874358974358975e-06, "loss": 0.001, "step": 12450 }, { "epoch": 72.74, "learning_rate": 3.861538461538462e-06, "loss": 0.0005, "step": 12475 }, { "epoch": 72.89, "learning_rate": 3.848717948717949e-06, "loss": 0.001, "step": 12500 }, { "epoch": 73.03, "learning_rate": 3.835897435897436e-06, "loss": 0.0007, "step": 12525 }, { "epoch": 73.18, "learning_rate": 3.823076923076923e-06, "loss": 0.0008, "step": 12550 }, { "epoch": 73.32, "learning_rate": 3.8102564102564107e-06, "loss": 0.0004, "step": 12575 }, { "epoch": 73.47, "learning_rate": 3.7974358974358975e-06, "loss": 0.0006, "step": 12600 }, { "epoch": 73.62, "learning_rate": 3.7846153846153847e-06, "loss": 0.001, "step": 12625 }, { "epoch": 73.76, "learning_rate": 3.771794871794872e-06, "loss": 0.0012, "step": 12650 }, { "epoch": 73.91, "learning_rate": 3.7589743589743592e-06, "loss": 0.0006, "step": 12675 }, { "epoch": 74.05, "learning_rate": 3.7461538461538465e-06, "loss": 0.0007, "step": 12700 }, { "epoch": 74.2, "learning_rate": 3.7333333333333337e-06, "loss": 0.0007, "step": 12725 }, { "epoch": 74.34, "learning_rate": 3.720512820512821e-06, "loss": 0.0005, "step": 12750 }, { "epoch": 74.49, "learning_rate": 3.7076923076923082e-06, "loss": 0.0005, "step": 12775 }, { "epoch": 74.64, "learning_rate": 3.694871794871795e-06, "loss": 0.0006, "step": 12800 }, { "epoch": 74.78, "learning_rate": 3.6820512820512823e-06, "loss": 0.001, "step": 12825 }, { "epoch": 74.93, "learning_rate": 3.6692307692307695e-06, "loss": 0.0015, "step": 12850 }, { "epoch": 75.07, "learning_rate": 3.656410256410257e-06, "loss": 0.0012, "step": 12875 }, { "epoch": 75.22, "learning_rate": 3.6435897435897436e-06, "loss": 0.0011, "step": 12900 }, { "epoch": 75.36, "learning_rate": 3.630769230769231e-06, "loss": 0.0008, "step": 12925 }, { "epoch": 75.51, "learning_rate": 3.617948717948718e-06, "loss": 0.0006, "step": 12950 }, { "epoch": 75.66, "learning_rate": 3.6051282051282054e-06, "loss": 0.0011, "step": 12975 }, { "epoch": 75.8, "learning_rate": 3.5923076923076926e-06, "loss": 0.0008, "step": 13000 }, { "epoch": 75.8, "eval_loss": 0.25434035062789917, "eval_runtime": 1303.9744, "eval_samples_per_second": 4.906, "eval_steps_per_second": 0.307, "eval_wer": 5.601286618476118, "step": 13000 }, { "epoch": 75.95, "learning_rate": 3.57948717948718e-06, "loss": 0.0007, "step": 13025 }, { "epoch": 76.09, "learning_rate": 3.566666666666667e-06, "loss": 0.0006, "step": 13050 }, { "epoch": 76.24, "learning_rate": 3.5538461538461544e-06, "loss": 0.0007, "step": 13075 }, { "epoch": 76.38, "learning_rate": 3.541025641025641e-06, "loss": 0.0006, "step": 13100 }, { "epoch": 76.53, "learning_rate": 3.5282051282051284e-06, "loss": 0.0007, "step": 13125 }, { "epoch": 76.68, "learning_rate": 3.5153846153846157e-06, "loss": 0.0007, "step": 13150 }, { "epoch": 76.82, "learning_rate": 3.502564102564103e-06, "loss": 0.001, "step": 13175 }, { "epoch": 76.97, "learning_rate": 3.4897435897435897e-06, "loss": 0.0005, "step": 13200 }, { "epoch": 77.11, "learning_rate": 3.476923076923077e-06, "loss": 0.0005, "step": 13225 }, { "epoch": 77.26, "learning_rate": 3.4641025641025642e-06, "loss": 0.0007, "step": 13250 }, { "epoch": 77.41, "learning_rate": 3.4512820512820515e-06, "loss": 0.0005, "step": 13275 }, { "epoch": 77.55, "learning_rate": 3.4384615384615387e-06, "loss": 0.0005, "step": 13300 }, { "epoch": 77.7, "learning_rate": 3.425641025641026e-06, "loss": 0.0005, "step": 13325 }, { "epoch": 77.84, "learning_rate": 3.4128205128205132e-06, "loss": 0.0006, "step": 13350 }, { "epoch": 77.99, "learning_rate": 3.4000000000000005e-06, "loss": 0.0006, "step": 13375 }, { "epoch": 78.13, "learning_rate": 3.3871794871794873e-06, "loss": 0.0009, "step": 13400 }, { "epoch": 78.28, "learning_rate": 3.3743589743589746e-06, "loss": 0.0004, "step": 13425 }, { "epoch": 78.43, "learning_rate": 3.361538461538462e-06, "loss": 0.0007, "step": 13450 }, { "epoch": 78.57, "learning_rate": 3.348717948717949e-06, "loss": 0.0004, "step": 13475 }, { "epoch": 78.72, "learning_rate": 3.3358974358974363e-06, "loss": 0.0005, "step": 13500 }, { "epoch": 78.86, "learning_rate": 3.323076923076923e-06, "loss": 0.0003, "step": 13525 }, { "epoch": 79.01, "learning_rate": 3.3102564102564104e-06, "loss": 0.0003, "step": 13550 }, { "epoch": 79.15, "learning_rate": 3.2974358974358976e-06, "loss": 0.0004, "step": 13575 }, { "epoch": 79.3, "learning_rate": 3.284615384615385e-06, "loss": 0.0003, "step": 13600 }, { "epoch": 79.45, "learning_rate": 3.271794871794872e-06, "loss": 0.0003, "step": 13625 }, { "epoch": 79.59, "learning_rate": 3.2589743589743594e-06, "loss": 0.0003, "step": 13650 }, { "epoch": 79.74, "learning_rate": 3.2461538461538466e-06, "loss": 0.0006, "step": 13675 }, { "epoch": 79.88, "learning_rate": 3.2333333333333334e-06, "loss": 0.0009, "step": 13700 }, { "epoch": 80.03, "learning_rate": 3.2205128205128207e-06, "loss": 0.0004, "step": 13725 }, { "epoch": 80.17, "learning_rate": 3.207692307692308e-06, "loss": 0.0002, "step": 13750 }, { "epoch": 80.32, "learning_rate": 3.194871794871795e-06, "loss": 0.0004, "step": 13775 }, { "epoch": 80.47, "learning_rate": 3.1820512820512824e-06, "loss": 0.0003, "step": 13800 }, { "epoch": 80.61, "learning_rate": 3.1692307692307693e-06, "loss": 0.0003, "step": 13825 }, { "epoch": 80.76, "learning_rate": 3.1564102564102565e-06, "loss": 0.0003, "step": 13850 }, { "epoch": 80.9, "learning_rate": 3.1435897435897437e-06, "loss": 0.0004, "step": 13875 }, { "epoch": 81.05, "learning_rate": 3.130769230769231e-06, "loss": 0.0004, "step": 13900 }, { "epoch": 81.2, "learning_rate": 3.1179487179487182e-06, "loss": 0.0012, "step": 13925 }, { "epoch": 81.34, "learning_rate": 3.1051282051282055e-06, "loss": 0.0006, "step": 13950 }, { "epoch": 81.49, "learning_rate": 3.0923076923076927e-06, "loss": 0.0006, "step": 13975 }, { "epoch": 81.63, "learning_rate": 3.07948717948718e-06, "loss": 0.0003, "step": 14000 }, { "epoch": 81.63, "eval_loss": 0.2581544518470764, "eval_runtime": 1306.1881, "eval_samples_per_second": 4.897, "eval_steps_per_second": 0.306, "eval_wer": 5.360910317158372, "step": 14000 }, { "epoch": 81.78, "learning_rate": 3.066666666666667e-06, "loss": 0.0003, "step": 14025 }, { "epoch": 81.92, "learning_rate": 3.053846153846154e-06, "loss": 0.0004, "step": 14050 }, { "epoch": 82.07, "learning_rate": 3.0410256410256413e-06, "loss": 0.0006, "step": 14075 }, { "epoch": 82.22, "learning_rate": 3.0282051282051286e-06, "loss": 0.0003, "step": 14100 }, { "epoch": 82.36, "learning_rate": 3.0153846153846154e-06, "loss": 0.0005, "step": 14125 }, { "epoch": 82.51, "learning_rate": 3.0025641025641026e-06, "loss": 0.0005, "step": 14150 }, { "epoch": 82.65, "learning_rate": 2.98974358974359e-06, "loss": 0.0004, "step": 14175 }, { "epoch": 82.8, "learning_rate": 2.976923076923077e-06, "loss": 0.0005, "step": 14200 }, { "epoch": 82.94, "learning_rate": 2.9641025641025644e-06, "loss": 0.0003, "step": 14225 }, { "epoch": 83.09, "learning_rate": 2.9512820512820516e-06, "loss": 0.0004, "step": 14250 }, { "epoch": 83.24, "learning_rate": 2.938461538461539e-06, "loss": 0.0007, "step": 14275 }, { "epoch": 83.38, "learning_rate": 2.925641025641026e-06, "loss": 0.0005, "step": 14300 }, { "epoch": 83.53, "learning_rate": 2.912820512820513e-06, "loss": 0.0005, "step": 14325 }, { "epoch": 83.67, "learning_rate": 2.9e-06, "loss": 0.0005, "step": 14350 }, { "epoch": 83.82, "learning_rate": 2.8871794871794874e-06, "loss": 0.0005, "step": 14375 }, { "epoch": 83.97, "learning_rate": 2.8743589743589747e-06, "loss": 0.0006, "step": 14400 }, { "epoch": 84.11, "learning_rate": 2.8615384615384615e-06, "loss": 0.0007, "step": 14425 }, { "epoch": 84.26, "learning_rate": 2.8487179487179488e-06, "loss": 0.0006, "step": 14450 }, { "epoch": 84.4, "learning_rate": 2.835897435897436e-06, "loss": 0.001, "step": 14475 }, { "epoch": 84.55, "learning_rate": 2.8230769230769233e-06, "loss": 0.0004, "step": 14500 }, { "epoch": 84.69, "learning_rate": 2.8102564102564105e-06, "loss": 0.0003, "step": 14525 }, { "epoch": 84.84, "learning_rate": 2.7974358974358977e-06, "loss": 0.0003, "step": 14550 }, { "epoch": 84.99, "learning_rate": 2.784615384615385e-06, "loss": 0.0003, "step": 14575 }, { "epoch": 85.13, "learning_rate": 2.7717948717948722e-06, "loss": 0.0008, "step": 14600 }, { "epoch": 85.28, "learning_rate": 2.758974358974359e-06, "loss": 0.0006, "step": 14625 }, { "epoch": 85.42, "learning_rate": 2.7461538461538463e-06, "loss": 0.0004, "step": 14650 }, { "epoch": 85.57, "learning_rate": 2.7333333333333336e-06, "loss": 0.0003, "step": 14675 }, { "epoch": 85.71, "learning_rate": 2.720512820512821e-06, "loss": 0.0004, "step": 14700 }, { "epoch": 85.86, "learning_rate": 2.7076923076923076e-06, "loss": 0.0003, "step": 14725 }, { "epoch": 86.01, "learning_rate": 2.694871794871795e-06, "loss": 0.0002, "step": 14750 }, { "epoch": 86.15, "learning_rate": 2.682051282051282e-06, "loss": 0.0006, "step": 14775 }, { "epoch": 86.3, "learning_rate": 2.6692307692307694e-06, "loss": 0.0004, "step": 14800 }, { "epoch": 86.44, "learning_rate": 2.6564102564102566e-06, "loss": 0.0002, "step": 14825 }, { "epoch": 86.59, "learning_rate": 2.643589743589744e-06, "loss": 0.0003, "step": 14850 }, { "epoch": 86.73, "learning_rate": 2.630769230769231e-06, "loss": 0.0003, "step": 14875 }, { "epoch": 86.88, "learning_rate": 2.6179487179487184e-06, "loss": 0.0007, "step": 14900 }, { "epoch": 87.03, "learning_rate": 2.605128205128205e-06, "loss": 0.0003, "step": 14925 }, { "epoch": 87.17, "learning_rate": 2.5923076923076924e-06, "loss": 0.0003, "step": 14950 }, { "epoch": 87.32, "learning_rate": 2.5794871794871797e-06, "loss": 0.0005, "step": 14975 }, { "epoch": 87.46, "learning_rate": 2.566666666666667e-06, "loss": 0.0003, "step": 15000 }, { "epoch": 87.46, "eval_loss": 0.2590600848197937, "eval_runtime": 1305.7515, "eval_samples_per_second": 4.899, "eval_steps_per_second": 0.306, "eval_wer": 5.340158406253242, "step": 15000 }, { "epoch": 87.61, "learning_rate": 2.5538461538461538e-06, "loss": 0.0003, "step": 15025 }, { "epoch": 87.76, "learning_rate": 2.541025641025641e-06, "loss": 0.0002, "step": 15050 }, { "epoch": 87.9, "learning_rate": 2.5282051282051283e-06, "loss": 0.0003, "step": 15075 }, { "epoch": 88.05, "learning_rate": 2.5153846153846155e-06, "loss": 0.0002, "step": 15100 }, { "epoch": 88.19, "learning_rate": 2.5025641025641028e-06, "loss": 0.0002, "step": 15125 }, { "epoch": 88.34, "learning_rate": 2.48974358974359e-06, "loss": 0.0003, "step": 15150 }, { "epoch": 88.48, "learning_rate": 2.4769230769230773e-06, "loss": 0.0003, "step": 15175 }, { "epoch": 88.63, "learning_rate": 2.464102564102564e-06, "loss": 0.0003, "step": 15200 }, { "epoch": 88.78, "learning_rate": 2.4512820512820513e-06, "loss": 0.0006, "step": 15225 }, { "epoch": 88.92, "learning_rate": 2.4384615384615386e-06, "loss": 0.0004, "step": 15250 }, { "epoch": 89.07, "learning_rate": 2.425641025641026e-06, "loss": 0.0003, "step": 15275 }, { "epoch": 89.21, "learning_rate": 2.412820512820513e-06, "loss": 0.0003, "step": 15300 }, { "epoch": 89.36, "learning_rate": 2.4000000000000003e-06, "loss": 0.0005, "step": 15325 }, { "epoch": 89.5, "learning_rate": 2.387179487179487e-06, "loss": 0.0004, "step": 15350 }, { "epoch": 89.65, "learning_rate": 2.3743589743589744e-06, "loss": 0.0003, "step": 15375 }, { "epoch": 89.8, "learning_rate": 2.3615384615384616e-06, "loss": 0.0002, "step": 15400 }, { "epoch": 89.94, "learning_rate": 2.348717948717949e-06, "loss": 0.0003, "step": 15425 }, { "epoch": 90.09, "learning_rate": 2.335897435897436e-06, "loss": 0.0002, "step": 15450 }, { "epoch": 90.23, "learning_rate": 2.3230769230769234e-06, "loss": 0.0002, "step": 15475 }, { "epoch": 90.38, "learning_rate": 2.31025641025641e-06, "loss": 0.0003, "step": 15500 }, { "epoch": 90.52, "learning_rate": 2.2974358974358975e-06, "loss": 0.0003, "step": 15525 }, { "epoch": 90.67, "learning_rate": 2.2846153846153847e-06, "loss": 0.0002, "step": 15550 }, { "epoch": 90.82, "learning_rate": 2.271794871794872e-06, "loss": 0.0002, "step": 15575 }, { "epoch": 90.96, "learning_rate": 2.258974358974359e-06, "loss": 0.0002, "step": 15600 }, { "epoch": 91.11, "learning_rate": 2.2461538461538464e-06, "loss": 0.0003, "step": 15625 }, { "epoch": 91.25, "learning_rate": 2.2333333333333333e-06, "loss": 0.0003, "step": 15650 }, { "epoch": 91.4, "learning_rate": 2.2205128205128205e-06, "loss": 0.0002, "step": 15675 }, { "epoch": 91.55, "learning_rate": 2.2076923076923078e-06, "loss": 0.0002, "step": 15700 }, { "epoch": 91.69, "learning_rate": 2.194871794871795e-06, "loss": 0.0002, "step": 15725 }, { "epoch": 91.84, "learning_rate": 2.1820512820512823e-06, "loss": 0.0003, "step": 15750 }, { "epoch": 91.98, "learning_rate": 2.1692307692307695e-06, "loss": 0.0002, "step": 15775 }, { "epoch": 92.13, "learning_rate": 2.1564102564102563e-06, "loss": 0.0002, "step": 15800 }, { "epoch": 92.27, "learning_rate": 2.1435897435897436e-06, "loss": 0.0002, "step": 15825 }, { "epoch": 92.42, "learning_rate": 2.130769230769231e-06, "loss": 0.0002, "step": 15850 }, { "epoch": 92.57, "learning_rate": 2.117948717948718e-06, "loss": 0.0002, "step": 15875 }, { "epoch": 92.71, "learning_rate": 2.1051282051282053e-06, "loss": 0.0002, "step": 15900 }, { "epoch": 92.86, "learning_rate": 2.0923076923076926e-06, "loss": 0.0002, "step": 15925 }, { "epoch": 93.0, "learning_rate": 2.0794871794871794e-06, "loss": 0.0001, "step": 15950 }, { "epoch": 93.15, "learning_rate": 2.0666666666666666e-06, "loss": 0.0004, "step": 15975 }, { "epoch": 93.29, "learning_rate": 2.053846153846154e-06, "loss": 0.0003, "step": 16000 }, { "epoch": 93.29, "eval_loss": 0.26566943526268005, "eval_runtime": 1305.7711, "eval_samples_per_second": 4.899, "eval_steps_per_second": 0.306, "eval_wer": 5.360910317158372, "step": 16000 }, { "epoch": 93.44, "learning_rate": 2.041025641025641e-06, "loss": 0.0002, "step": 16025 }, { "epoch": 93.59, "learning_rate": 2.0282051282051284e-06, "loss": 0.0002, "step": 16050 }, { "epoch": 93.73, "learning_rate": 2.0153846153846156e-06, "loss": 0.0002, "step": 16075 }, { "epoch": 93.88, "learning_rate": 2.0025641025641025e-06, "loss": 0.0003, "step": 16100 }, { "epoch": 94.02, "learning_rate": 1.9897435897435897e-06, "loss": 0.0003, "step": 16125 }, { "epoch": 94.17, "learning_rate": 1.976923076923077e-06, "loss": 0.0003, "step": 16150 }, { "epoch": 94.31, "learning_rate": 1.964102564102564e-06, "loss": 0.0002, "step": 16175 }, { "epoch": 94.46, "learning_rate": 1.9512820512820515e-06, "loss": 0.0003, "step": 16200 }, { "epoch": 94.61, "learning_rate": 1.9384615384615387e-06, "loss": 0.0004, "step": 16225 }, { "epoch": 94.75, "learning_rate": 1.9256410256410255e-06, "loss": 0.0002, "step": 16250 }, { "epoch": 94.9, "learning_rate": 1.9128205128205128e-06, "loss": 0.0002, "step": 16275 }, { "epoch": 95.04, "learning_rate": 1.9000000000000002e-06, "loss": 0.0002, "step": 16300 }, { "epoch": 95.19, "learning_rate": 1.8871794871794875e-06, "loss": 0.0002, "step": 16325 }, { "epoch": 95.34, "learning_rate": 1.8743589743589747e-06, "loss": 0.0001, "step": 16350 }, { "epoch": 95.48, "learning_rate": 1.8615384615384616e-06, "loss": 0.0002, "step": 16375 }, { "epoch": 95.63, "learning_rate": 1.8487179487179488e-06, "loss": 0.0003, "step": 16400 }, { "epoch": 95.77, "learning_rate": 1.835897435897436e-06, "loss": 0.0004, "step": 16425 }, { "epoch": 95.92, "learning_rate": 1.8230769230769233e-06, "loss": 0.0003, "step": 16450 }, { "epoch": 96.06, "learning_rate": 1.8102564102564105e-06, "loss": 0.0003, "step": 16475 }, { "epoch": 96.21, "learning_rate": 1.7974358974358978e-06, "loss": 0.0002, "step": 16500 }, { "epoch": 96.36, "learning_rate": 1.7846153846153846e-06, "loss": 0.0009, "step": 16525 }, { "epoch": 96.5, "learning_rate": 1.7717948717948719e-06, "loss": 0.0003, "step": 16550 }, { "epoch": 96.65, "learning_rate": 1.7589743589743591e-06, "loss": 0.0002, "step": 16575 }, { "epoch": 96.79, "learning_rate": 1.7461538461538464e-06, "loss": 0.0007, "step": 16600 }, { "epoch": 96.94, "learning_rate": 1.7333333333333336e-06, "loss": 0.0002, "step": 16625 }, { "epoch": 97.08, "learning_rate": 1.7205128205128209e-06, "loss": 0.0002, "step": 16650 }, { "epoch": 97.23, "learning_rate": 1.7076923076923077e-06, "loss": 0.0001, "step": 16675 }, { "epoch": 97.38, "learning_rate": 1.694871794871795e-06, "loss": 0.0003, "step": 16700 }, { "epoch": 97.52, "learning_rate": 1.6820512820512822e-06, "loss": 0.0007, "step": 16725 }, { "epoch": 97.67, "learning_rate": 1.6692307692307694e-06, "loss": 0.0003, "step": 16750 }, { "epoch": 97.81, "learning_rate": 1.6564102564102567e-06, "loss": 0.0001, "step": 16775 }, { "epoch": 97.96, "learning_rate": 1.643589743589744e-06, "loss": 0.0004, "step": 16800 }, { "epoch": 98.1, "learning_rate": 1.6307692307692307e-06, "loss": 0.0002, "step": 16825 }, { "epoch": 98.25, "learning_rate": 1.617948717948718e-06, "loss": 0.0002, "step": 16850 }, { "epoch": 98.4, "learning_rate": 1.6051282051282052e-06, "loss": 0.0002, "step": 16875 }, { "epoch": 98.54, "learning_rate": 1.5923076923076925e-06, "loss": 0.0002, "step": 16900 }, { "epoch": 98.69, "learning_rate": 1.5794871794871797e-06, "loss": 0.0002, "step": 16925 }, { "epoch": 98.83, "learning_rate": 1.566666666666667e-06, "loss": 0.0002, "step": 16950 }, { "epoch": 98.98, "learning_rate": 1.5538461538461538e-06, "loss": 0.0003, "step": 16975 }, { "epoch": 99.13, "learning_rate": 1.541025641025641e-06, "loss": 0.0002, "step": 17000 }, { "epoch": 99.13, "eval_loss": 0.26610660552978516, "eval_runtime": 1305.0531, "eval_samples_per_second": 4.902, "eval_steps_per_second": 0.307, "eval_wer": 5.386850205789783, "step": 17000 }, { "epoch": 99.27, "learning_rate": 1.5282051282051283e-06, "loss": 0.0002, "step": 17025 }, { "epoch": 99.42, "learning_rate": 1.5153846153846156e-06, "loss": 0.0002, "step": 17050 }, { "epoch": 99.56, "learning_rate": 1.5025641025641028e-06, "loss": 0.0003, "step": 17075 }, { "epoch": 99.71, "learning_rate": 1.48974358974359e-06, "loss": 0.0002, "step": 17100 }, { "epoch": 99.85, "learning_rate": 1.476923076923077e-06, "loss": 0.0003, "step": 17125 }, { "epoch": 100.0, "learning_rate": 1.4641025641025641e-06, "loss": 0.0001, "step": 17150 }, { "epoch": 100.15, "learning_rate": 1.4512820512820514e-06, "loss": 0.0001, "step": 17175 }, { "epoch": 100.29, "learning_rate": 1.4384615384615386e-06, "loss": 0.0001, "step": 17200 }, { "epoch": 100.44, "learning_rate": 1.4256410256410259e-06, "loss": 0.0001, "step": 17225 }, { "epoch": 100.58, "learning_rate": 1.4128205128205131e-06, "loss": 0.0002, "step": 17250 }, { "epoch": 100.73, "learning_rate": 1.4000000000000001e-06, "loss": 0.0002, "step": 17275 }, { "epoch": 100.87, "learning_rate": 1.3871794871794872e-06, "loss": 0.0002, "step": 17300 }, { "epoch": 101.02, "learning_rate": 1.3743589743589744e-06, "loss": 0.0002, "step": 17325 }, { "epoch": 101.17, "learning_rate": 1.3615384615384617e-06, "loss": 0.0001, "step": 17350 }, { "epoch": 101.31, "learning_rate": 1.348717948717949e-06, "loss": 0.0004, "step": 17375 }, { "epoch": 101.46, "learning_rate": 1.3358974358974362e-06, "loss": 0.0003, "step": 17400 }, { "epoch": 101.6, "learning_rate": 1.3230769230769232e-06, "loss": 0.0002, "step": 17425 }, { "epoch": 101.75, "learning_rate": 1.3102564102564102e-06, "loss": 0.0002, "step": 17450 }, { "epoch": 101.9, "learning_rate": 1.2974358974358975e-06, "loss": 0.0001, "step": 17475 }, { "epoch": 102.04, "learning_rate": 1.2846153846153847e-06, "loss": 0.0003, "step": 17500 }, { "epoch": 102.19, "learning_rate": 1.271794871794872e-06, "loss": 0.0002, "step": 17525 }, { "epoch": 102.33, "learning_rate": 1.2589743589743592e-06, "loss": 0.0002, "step": 17550 }, { "epoch": 102.48, "learning_rate": 1.2461538461538463e-06, "loss": 0.0002, "step": 17575 }, { "epoch": 102.62, "learning_rate": 1.2333333333333335e-06, "loss": 0.0001, "step": 17600 }, { "epoch": 102.77, "learning_rate": 1.2205128205128206e-06, "loss": 0.0002, "step": 17625 }, { "epoch": 102.92, "learning_rate": 1.2076923076923078e-06, "loss": 0.0002, "step": 17650 }, { "epoch": 103.06, "learning_rate": 1.194871794871795e-06, "loss": 0.0002, "step": 17675 }, { "epoch": 103.21, "learning_rate": 1.182051282051282e-06, "loss": 0.0004, "step": 17700 }, { "epoch": 103.35, "learning_rate": 1.1692307692307693e-06, "loss": 0.0002, "step": 17725 }, { "epoch": 103.5, "learning_rate": 1.1564102564102566e-06, "loss": 0.0001, "step": 17750 }, { "epoch": 103.64, "learning_rate": 1.1435897435897436e-06, "loss": 0.0002, "step": 17775 }, { "epoch": 103.79, "learning_rate": 1.1307692307692309e-06, "loss": 0.0001, "step": 17800 }, { "epoch": 103.94, "learning_rate": 1.1179487179487181e-06, "loss": 0.0001, "step": 17825 }, { "epoch": 104.08, "learning_rate": 1.1051282051282052e-06, "loss": 0.0002, "step": 17850 }, { "epoch": 104.23, "learning_rate": 1.0923076923076924e-06, "loss": 0.0002, "step": 17875 }, { "epoch": 104.37, "learning_rate": 1.0794871794871797e-06, "loss": 0.0001, "step": 17900 }, { "epoch": 104.52, "learning_rate": 1.066666666666667e-06, "loss": 0.0001, "step": 17925 }, { "epoch": 104.66, "learning_rate": 1.053846153846154e-06, "loss": 0.0001, "step": 17950 }, { "epoch": 104.81, "learning_rate": 1.0410256410256412e-06, "loss": 0.0002, "step": 17975 }, { "epoch": 104.96, "learning_rate": 1.0282051282051284e-06, "loss": 0.0001, "step": 18000 }, { "epoch": 104.96, "eval_loss": 0.27037352323532104, "eval_runtime": 1304.1303, "eval_samples_per_second": 4.905, "eval_steps_per_second": 0.307, "eval_wer": 5.317677169439353, "step": 18000 }, { "epoch": 105.1, "learning_rate": 1.0153846153846155e-06, "loss": 0.0001, "step": 18025 }, { "epoch": 105.25, "learning_rate": 1.0025641025641027e-06, "loss": 0.0002, "step": 18050 }, { "epoch": 105.39, "learning_rate": 9.8974358974359e-07, "loss": 0.0002, "step": 18075 }, { "epoch": 105.54, "learning_rate": 9.76923076923077e-07, "loss": 0.0001, "step": 18100 }, { "epoch": 105.69, "learning_rate": 9.641025641025642e-07, "loss": 0.0001, "step": 18125 }, { "epoch": 105.83, "learning_rate": 9.512820512820514e-07, "loss": 0.0001, "step": 18150 }, { "epoch": 105.98, "learning_rate": 9.384615384615385e-07, "loss": 0.0002, "step": 18175 }, { "epoch": 106.12, "learning_rate": 9.256410256410258e-07, "loss": 0.0001, "step": 18200 }, { "epoch": 106.27, "learning_rate": 9.128205128205129e-07, "loss": 0.0002, "step": 18225 }, { "epoch": 106.41, "learning_rate": 9.000000000000001e-07, "loss": 0.0002, "step": 18250 }, { "epoch": 106.56, "learning_rate": 8.871794871794873e-07, "loss": 0.0001, "step": 18275 }, { "epoch": 106.71, "learning_rate": 8.743589743589745e-07, "loss": 0.0001, "step": 18300 }, { "epoch": 106.85, "learning_rate": 8.615384615384616e-07, "loss": 0.0001, "step": 18325 }, { "epoch": 107.0, "learning_rate": 8.487179487179488e-07, "loss": 0.0001, "step": 18350 }, { "epoch": 107.14, "learning_rate": 8.35897435897436e-07, "loss": 0.0001, "step": 18375 }, { "epoch": 107.29, "learning_rate": 8.230769230769231e-07, "loss": 0.0001, "step": 18400 }, { "epoch": 107.43, "learning_rate": 8.102564102564104e-07, "loss": 0.0001, "step": 18425 }, { "epoch": 107.58, "learning_rate": 7.974358974358975e-07, "loss": 0.0001, "step": 18450 }, { "epoch": 107.73, "learning_rate": 7.846153846153847e-07, "loss": 0.0002, "step": 18475 }, { "epoch": 107.87, "learning_rate": 7.717948717948719e-07, "loss": 0.0001, "step": 18500 }, { "epoch": 108.02, "learning_rate": 7.58974358974359e-07, "loss": 0.0002, "step": 18525 }, { "epoch": 108.16, "learning_rate": 7.461538461538462e-07, "loss": 0.0001, "step": 18550 }, { "epoch": 108.31, "learning_rate": 7.333333333333334e-07, "loss": 0.0001, "step": 18575 }, { "epoch": 108.45, "learning_rate": 7.205128205128206e-07, "loss": 0.0001, "step": 18600 }, { "epoch": 108.6, "learning_rate": 7.076923076923077e-07, "loss": 0.0001, "step": 18625 }, { "epoch": 108.75, "learning_rate": 6.94871794871795e-07, "loss": 0.0001, "step": 18650 }, { "epoch": 108.89, "learning_rate": 6.820512820512821e-07, "loss": 0.0001, "step": 18675 }, { "epoch": 109.04, "learning_rate": 6.692307692307693e-07, "loss": 0.0001, "step": 18700 }, { "epoch": 109.18, "learning_rate": 6.564102564102565e-07, "loss": 0.0001, "step": 18725 }, { "epoch": 109.33, "learning_rate": 6.435897435897436e-07, "loss": 0.0001, "step": 18750 }, { "epoch": 109.48, "learning_rate": 6.307692307692308e-07, "loss": 0.0001, "step": 18775 }, { "epoch": 109.62, "learning_rate": 6.17948717948718e-07, "loss": 0.0001, "step": 18800 }, { "epoch": 109.77, "learning_rate": 6.051282051282052e-07, "loss": 0.0001, "step": 18825 }, { "epoch": 109.91, "learning_rate": 5.923076923076924e-07, "loss": 0.0001, "step": 18850 }, { "epoch": 110.06, "learning_rate": 5.794871794871796e-07, "loss": 0.0001, "step": 18875 }, { "epoch": 110.2, "learning_rate": 5.666666666666667e-07, "loss": 0.0001, "step": 18900 }, { "epoch": 110.35, "learning_rate": 5.53846153846154e-07, "loss": 0.0001, "step": 18925 }, { "epoch": 110.5, "learning_rate": 5.410256410256411e-07, "loss": 0.0001, "step": 18950 }, { "epoch": 110.64, "learning_rate": 5.282051282051282e-07, "loss": 0.0001, "step": 18975 }, { "epoch": 110.79, "learning_rate": 5.153846153846155e-07, "loss": 0.0001, "step": 19000 }, { "epoch": 110.79, "eval_loss": 0.2749752700328827, "eval_runtime": 1305.7419, "eval_samples_per_second": 4.899, "eval_steps_per_second": 0.306, "eval_wer": 5.315947843530592, "step": 19000 }, { "epoch": 110.93, "learning_rate": 5.025641025641026e-07, "loss": 0.0001, "step": 19025 }, { "epoch": 111.08, "learning_rate": 4.897435897435898e-07, "loss": 0.0001, "step": 19050 }, { "epoch": 111.22, "learning_rate": 4.76923076923077e-07, "loss": 0.0001, "step": 19075 }, { "epoch": 111.37, "learning_rate": 4.641025641025641e-07, "loss": 0.0001, "step": 19100 }, { "epoch": 111.52, "learning_rate": 4.5128205128205136e-07, "loss": 0.0001, "step": 19125 }, { "epoch": 111.66, "learning_rate": 4.384615384615385e-07, "loss": 0.0001, "step": 19150 }, { "epoch": 111.81, "learning_rate": 4.2564102564102564e-07, "loss": 0.0001, "step": 19175 }, { "epoch": 111.95, "learning_rate": 4.128205128205129e-07, "loss": 0.0001, "step": 19200 }, { "epoch": 112.1, "learning_rate": 4.0000000000000003e-07, "loss": 0.0001, "step": 19225 }, { "epoch": 112.24, "learning_rate": 3.871794871794872e-07, "loss": 0.0002, "step": 19250 }, { "epoch": 112.39, "learning_rate": 3.743589743589744e-07, "loss": 0.0002, "step": 19275 }, { "epoch": 112.54, "learning_rate": 3.6153846153846157e-07, "loss": 0.0001, "step": 19300 }, { "epoch": 112.68, "learning_rate": 3.487179487179487e-07, "loss": 0.0001, "step": 19325 }, { "epoch": 112.83, "learning_rate": 3.3589743589743596e-07, "loss": 0.0001, "step": 19350 }, { "epoch": 112.97, "learning_rate": 3.230769230769231e-07, "loss": 0.0001, "step": 19375 }, { "epoch": 113.12, "learning_rate": 3.102564102564103e-07, "loss": 0.0001, "step": 19400 }, { "epoch": 113.27, "learning_rate": 2.9743589743589744e-07, "loss": 0.0001, "step": 19425 }, { "epoch": 113.41, "learning_rate": 2.8461538461538463e-07, "loss": 0.0001, "step": 19450 }, { "epoch": 113.56, "learning_rate": 2.717948717948718e-07, "loss": 0.0001, "step": 19475 }, { "epoch": 113.7, "learning_rate": 2.5897435897435897e-07, "loss": 0.0002, "step": 19500 }, { "epoch": 113.85, "learning_rate": 2.4615384615384616e-07, "loss": 0.0001, "step": 19525 }, { "epoch": 113.99, "learning_rate": 2.3333333333333336e-07, "loss": 0.0001, "step": 19550 }, { "epoch": 114.14, "learning_rate": 2.2051282051282053e-07, "loss": 0.0001, "step": 19575 }, { "epoch": 114.29, "learning_rate": 2.076923076923077e-07, "loss": 0.0002, "step": 19600 }, { "epoch": 114.43, "learning_rate": 1.948717948717949e-07, "loss": 0.0001, "step": 19625 }, { "epoch": 114.58, "learning_rate": 1.8205128205128206e-07, "loss": 0.0003, "step": 19650 }, { "epoch": 114.72, "learning_rate": 1.6923076923076923e-07, "loss": 0.0001, "step": 19675 }, { "epoch": 114.87, "learning_rate": 1.5641025641025642e-07, "loss": 0.0001, "step": 19700 }, { "epoch": 115.01, "learning_rate": 1.435897435897436e-07, "loss": 0.0001, "step": 19725 }, { "epoch": 115.16, "learning_rate": 1.3076923076923079e-07, "loss": 0.0001, "step": 19750 }, { "epoch": 115.31, "learning_rate": 1.1794871794871797e-07, "loss": 0.0001, "step": 19775 }, { "epoch": 115.45, "learning_rate": 1.0512820512820514e-07, "loss": 0.0001, "step": 19800 }, { "epoch": 115.6, "learning_rate": 9.230769230769232e-08, "loss": 0.0001, "step": 19825 }, { "epoch": 115.74, "learning_rate": 7.94871794871795e-08, "loss": 0.0001, "step": 19850 }, { "epoch": 115.89, "learning_rate": 6.666666666666668e-08, "loss": 0.0001, "step": 19875 }, { "epoch": 116.03, "learning_rate": 5.384615384615385e-08, "loss": 0.0001, "step": 19900 }, { "epoch": 116.18, "learning_rate": 4.102564102564103e-08, "loss": 0.0001, "step": 19925 }, { "epoch": 116.33, "learning_rate": 2.820512820512821e-08, "loss": 0.0001, "step": 19950 }, { "epoch": 116.47, "learning_rate": 1.5384615384615385e-08, "loss": 0.0001, "step": 19975 }, { "epoch": 116.62, "learning_rate": 2.5641025641025645e-09, "loss": 0.0001, "step": 20000 }, { "epoch": 116.62, "eval_loss": 0.2734989523887634, "eval_runtime": 1306.0978, "eval_samples_per_second": 4.898, "eval_steps_per_second": 0.306, "eval_wer": 5.309030539895549, "step": 20000 }, { "epoch": 116.62, "step": 20000, "total_flos": 4.338945087307776e+21, "train_loss": 0.041992762515088544, "train_runtime": 138702.4363, "train_samples_per_second": 9.228, "train_steps_per_second": 0.144 } ], "logging_steps": 25, "max_steps": 20000, "num_input_tokens_seen": 0, "num_train_epochs": 117, "save_steps": 1000, "total_flos": 4.338945087307776e+21, "train_batch_size": 32, "trial_name": null, "trial_params": null }