diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,5008 @@ +{ + "best_metric": 5.987858719646799, + "best_model_checkpoint": "./checkpoint-17000", + "epoch": 99.01, + "eval_steps": 1000, + "global_step": 20000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 4.4e-07, + "loss": 1.216, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 9.400000000000001e-07, + "loss": 0.7368, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 1.44e-06, + "loss": 0.4, + "step": 75 + }, + { + "epoch": 0.01, + "learning_rate": 1.94e-06, + "loss": 0.3277, + "step": 100 + }, + { + "epoch": 0.01, + "learning_rate": 2.4400000000000004e-06, + "loss": 0.2631, + "step": 125 + }, + { + "epoch": 0.01, + "learning_rate": 2.9400000000000002e-06, + "loss": 0.2379, + "step": 150 + }, + { + "epoch": 0.01, + "learning_rate": 3.44e-06, + "loss": 0.2115, + "step": 175 + }, + { + "epoch": 0.01, + "learning_rate": 3.94e-06, + "loss": 0.1865, + "step": 200 + }, + { + "epoch": 1.0, + "learning_rate": 4.440000000000001e-06, + "loss": 0.1732, + "step": 225 + }, + { + "epoch": 1.0, + "learning_rate": 4.94e-06, + "loss": 0.1896, + "step": 250 + }, + { + "epoch": 1.0, + "learning_rate": 5.4400000000000004e-06, + "loss": 0.1618, + "step": 275 + }, + { + "epoch": 1.0, + "learning_rate": 5.94e-06, + "loss": 0.15, + "step": 300 + }, + { + "epoch": 1.01, + "learning_rate": 6.440000000000001e-06, + "loss": 0.1295, + "step": 325 + }, + { + "epoch": 1.01, + "learning_rate": 6.9400000000000005e-06, + "loss": 0.1241, + "step": 350 + }, + { + "epoch": 1.01, + "learning_rate": 7.440000000000001e-06, + "loss": 0.1089, + "step": 375 + }, + { + "epoch": 1.01, + "learning_rate": 7.94e-06, + "loss": 0.1029, + "step": 400 + }, + { + "epoch": 2.0, + "learning_rate": 8.44e-06, + "loss": 0.0887, + "step": 425 + }, + { + "epoch": 2.0, + "learning_rate": 8.94e-06, + "loss": 0.111, + "step": 450 + }, + { + "epoch": 2.0, + "learning_rate": 9.42e-06, + "loss": 0.09, + "step": 475 + }, + { + "epoch": 2.0, + "learning_rate": 9.920000000000002e-06, + "loss": 0.0851, + "step": 500 + }, + { + "epoch": 2.01, + "learning_rate": 9.98923076923077e-06, + "loss": 0.074, + "step": 525 + }, + { + "epoch": 2.01, + "learning_rate": 9.976410256410257e-06, + "loss": 0.0696, + "step": 550 + }, + { + "epoch": 2.01, + "learning_rate": 9.963589743589744e-06, + "loss": 0.0608, + "step": 575 + }, + { + "epoch": 2.01, + "learning_rate": 9.950769230769232e-06, + "loss": 0.0564, + "step": 600 + }, + { + "epoch": 3.0, + "learning_rate": 9.937948717948719e-06, + "loss": 0.051, + "step": 625 + }, + { + "epoch": 3.0, + "learning_rate": 9.925128205128206e-06, + "loss": 0.0646, + "step": 650 + }, + { + "epoch": 3.0, + "learning_rate": 9.912307692307693e-06, + "loss": 0.0537, + "step": 675 + }, + { + "epoch": 3.0, + "learning_rate": 9.899487179487181e-06, + "loss": 0.0533, + "step": 700 + }, + { + "epoch": 3.01, + "learning_rate": 9.886666666666668e-06, + "loss": 0.0463, + "step": 725 + }, + { + "epoch": 3.01, + "learning_rate": 9.873846153846155e-06, + "loss": 0.0387, + "step": 750 + }, + { + "epoch": 3.01, + "learning_rate": 9.861025641025642e-06, + "loss": 0.0357, + "step": 775 + }, + { + "epoch": 3.01, + "learning_rate": 9.84820512820513e-06, + "loss": 0.0318, + "step": 800 + }, + { + "epoch": 4.0, + "learning_rate": 9.835384615384617e-06, + "loss": 0.0312, + "step": 825 + }, + { + "epoch": 4.0, + "learning_rate": 9.822564102564104e-06, + "loss": 0.0371, + "step": 850 + }, + { + "epoch": 4.0, + "learning_rate": 9.80974358974359e-06, + "loss": 0.0317, + "step": 875 + }, + { + "epoch": 4.0, + "learning_rate": 9.796923076923077e-06, + "loss": 0.0322, + "step": 900 + }, + { + "epoch": 4.01, + "learning_rate": 9.784102564102564e-06, + "loss": 0.0258, + "step": 925 + }, + { + "epoch": 4.01, + "learning_rate": 9.771282051282051e-06, + "loss": 0.0241, + "step": 950 + }, + { + "epoch": 4.01, + "learning_rate": 9.75846153846154e-06, + "loss": 0.0201, + "step": 975 + }, + { + "epoch": 4.01, + "learning_rate": 9.745641025641026e-06, + "loss": 0.0185, + "step": 1000 + }, + { + "epoch": 4.01, + "eval_loss": 0.18961435556411743, + "eval_runtime": 2281.0062, + "eval_samples_per_second": 2.87, + "eval_steps_per_second": 0.18, + "eval_wer": 6.356926048565121, + "step": 1000 + }, + { + "epoch": 5.0, + "learning_rate": 9.732820512820513e-06, + "loss": 0.018, + "step": 1025 + }, + { + "epoch": 5.0, + "learning_rate": 9.72e-06, + "loss": 0.02, + "step": 1050 + }, + { + "epoch": 5.0, + "learning_rate": 9.707179487179487e-06, + "loss": 0.0205, + "step": 1075 + }, + { + "epoch": 5.0, + "learning_rate": 9.694358974358975e-06, + "loss": 0.018, + "step": 1100 + }, + { + "epoch": 5.01, + "learning_rate": 9.681538461538462e-06, + "loss": 0.0183, + "step": 1125 + }, + { + "epoch": 5.01, + "learning_rate": 9.668717948717949e-06, + "loss": 0.0145, + "step": 1150 + }, + { + "epoch": 5.01, + "learning_rate": 9.655897435897436e-06, + "loss": 0.0117, + "step": 1175 + }, + { + "epoch": 5.01, + "learning_rate": 9.643076923076924e-06, + "loss": 0.0116, + "step": 1200 + }, + { + "epoch": 6.0, + "learning_rate": 9.630256410256411e-06, + "loss": 0.0151, + "step": 1225 + }, + { + "epoch": 6.0, + "learning_rate": 9.617435897435898e-06, + "loss": 0.016, + "step": 1250 + }, + { + "epoch": 6.0, + "learning_rate": 9.604615384615385e-06, + "loss": 0.0143, + "step": 1275 + }, + { + "epoch": 6.0, + "learning_rate": 9.591794871794873e-06, + "loss": 0.0126, + "step": 1300 + }, + { + "epoch": 6.01, + "learning_rate": 9.57897435897436e-06, + "loss": 0.0101, + "step": 1325 + }, + { + "epoch": 6.01, + "learning_rate": 9.566153846153847e-06, + "loss": 0.0097, + "step": 1350 + }, + { + "epoch": 6.01, + "learning_rate": 9.553333333333334e-06, + "loss": 0.0088, + "step": 1375 + }, + { + "epoch": 6.01, + "learning_rate": 9.540512820512822e-06, + "loss": 0.0097, + "step": 1400 + }, + { + "epoch": 7.0, + "learning_rate": 9.52769230769231e-06, + "loss": 0.0099, + "step": 1425 + }, + { + "epoch": 7.0, + "learning_rate": 9.514871794871796e-06, + "loss": 0.0105, + "step": 1450 + }, + { + "epoch": 7.0, + "learning_rate": 9.502051282051283e-06, + "loss": 0.0086, + "step": 1475 + }, + { + "epoch": 7.0, + "learning_rate": 9.48923076923077e-06, + "loss": 0.0098, + "step": 1500 + }, + { + "epoch": 7.01, + "learning_rate": 9.476410256410257e-06, + "loss": 0.0069, + "step": 1525 + }, + { + "epoch": 7.01, + "learning_rate": 9.463589743589743e-06, + "loss": 0.0071, + "step": 1550 + }, + { + "epoch": 7.01, + "learning_rate": 9.450769230769232e-06, + "loss": 0.0082, + "step": 1575 + }, + { + "epoch": 7.01, + "learning_rate": 9.437948717948719e-06, + "loss": 0.0068, + "step": 1600 + }, + { + "epoch": 8.0, + "learning_rate": 9.425128205128206e-06, + "loss": 0.0085, + "step": 1625 + }, + { + "epoch": 8.0, + "learning_rate": 9.412307692307692e-06, + "loss": 0.0081, + "step": 1650 + }, + { + "epoch": 8.0, + "learning_rate": 9.39948717948718e-06, + "loss": 0.006, + "step": 1675 + }, + { + "epoch": 8.01, + "learning_rate": 9.386666666666668e-06, + "loss": 0.0055, + "step": 1700 + }, + { + "epoch": 8.01, + "learning_rate": 9.373846153846155e-06, + "loss": 0.0066, + "step": 1725 + }, + { + "epoch": 8.01, + "learning_rate": 9.361025641025641e-06, + "loss": 0.006, + "step": 1750 + }, + { + "epoch": 8.01, + "learning_rate": 9.348205128205128e-06, + "loss": 0.0078, + "step": 1775 + }, + { + "epoch": 8.01, + "learning_rate": 9.335384615384617e-06, + "loss": 0.0059, + "step": 1800 + }, + { + "epoch": 9.0, + "learning_rate": 9.322564102564104e-06, + "loss": 0.0077, + "step": 1825 + }, + { + "epoch": 9.0, + "learning_rate": 9.30974358974359e-06, + "loss": 0.007, + "step": 1850 + }, + { + "epoch": 9.0, + "learning_rate": 9.296923076923077e-06, + "loss": 0.0065, + "step": 1875 + }, + { + "epoch": 9.01, + "learning_rate": 9.284102564102566e-06, + "loss": 0.0064, + "step": 1900 + }, + { + "epoch": 9.01, + "learning_rate": 9.271282051282053e-06, + "loss": 0.0074, + "step": 1925 + }, + { + "epoch": 9.01, + "learning_rate": 9.25846153846154e-06, + "loss": 0.0054, + "step": 1950 + }, + { + "epoch": 9.01, + "learning_rate": 9.245641025641026e-06, + "loss": 0.0085, + "step": 1975 + }, + { + "epoch": 9.01, + "learning_rate": 9.232820512820515e-06, + "loss": 0.0067, + "step": 2000 + }, + { + "epoch": 9.01, + "eval_loss": 0.20833465456962585, + "eval_runtime": 2187.6524, + "eval_samples_per_second": 2.992, + "eval_steps_per_second": 0.187, + "eval_wer": 6.3862444812362025, + "step": 2000 + }, + { + "epoch": 10.0, + "learning_rate": 9.220000000000002e-06, + "loss": 0.0067, + "step": 2025 + }, + { + "epoch": 10.0, + "learning_rate": 9.207179487179488e-06, + "loss": 0.0058, + "step": 2050 + }, + { + "epoch": 10.0, + "learning_rate": 9.194358974358975e-06, + "loss": 0.0041, + "step": 2075 + }, + { + "epoch": 10.01, + "learning_rate": 9.181538461538464e-06, + "loss": 0.0056, + "step": 2100 + }, + { + "epoch": 10.01, + "learning_rate": 9.168717948717949e-06, + "loss": 0.0061, + "step": 2125 + }, + { + "epoch": 10.01, + "learning_rate": 9.155897435897436e-06, + "loss": 0.0046, + "step": 2150 + }, + { + "epoch": 10.01, + "learning_rate": 9.143076923076924e-06, + "loss": 0.0055, + "step": 2175 + }, + { + "epoch": 10.01, + "learning_rate": 9.130256410256411e-06, + "loss": 0.0044, + "step": 2200 + }, + { + "epoch": 11.0, + "learning_rate": 9.117435897435898e-06, + "loss": 0.0051, + "step": 2225 + }, + { + "epoch": 11.0, + "learning_rate": 9.104615384615385e-06, + "loss": 0.0058, + "step": 2250 + }, + { + "epoch": 11.0, + "learning_rate": 9.091794871794873e-06, + "loss": 0.0058, + "step": 2275 + }, + { + "epoch": 11.01, + "learning_rate": 9.07897435897436e-06, + "loss": 0.0041, + "step": 2300 + }, + { + "epoch": 11.01, + "learning_rate": 9.066153846153847e-06, + "loss": 0.0039, + "step": 2325 + }, + { + "epoch": 11.01, + "learning_rate": 9.053333333333334e-06, + "loss": 0.0038, + "step": 2350 + }, + { + "epoch": 11.01, + "learning_rate": 9.04051282051282e-06, + "loss": 0.0041, + "step": 2375 + }, + { + "epoch": 11.01, + "learning_rate": 9.027692307692309e-06, + "loss": 0.0043, + "step": 2400 + }, + { + "epoch": 12.0, + "learning_rate": 9.014871794871796e-06, + "loss": 0.005, + "step": 2425 + }, + { + "epoch": 12.0, + "learning_rate": 9.002051282051283e-06, + "loss": 0.0044, + "step": 2450 + }, + { + "epoch": 12.0, + "learning_rate": 8.98923076923077e-06, + "loss": 0.0042, + "step": 2475 + }, + { + "epoch": 12.01, + "learning_rate": 8.976410256410258e-06, + "loss": 0.0039, + "step": 2500 + }, + { + "epoch": 12.01, + "learning_rate": 8.963589743589745e-06, + "loss": 0.0028, + "step": 2525 + }, + { + "epoch": 12.01, + "learning_rate": 8.950769230769232e-06, + "loss": 0.0027, + "step": 2550 + }, + { + "epoch": 12.01, + "learning_rate": 8.937948717948718e-06, + "loss": 0.0035, + "step": 2575 + }, + { + "epoch": 12.01, + "learning_rate": 8.925128205128207e-06, + "loss": 0.0031, + "step": 2600 + }, + { + "epoch": 13.0, + "learning_rate": 8.912307692307694e-06, + "loss": 0.0043, + "step": 2625 + }, + { + "epoch": 13.0, + "learning_rate": 8.89948717948718e-06, + "loss": 0.0035, + "step": 2650 + }, + { + "epoch": 13.0, + "learning_rate": 8.886666666666667e-06, + "loss": 0.0025, + "step": 2675 + }, + { + "epoch": 13.01, + "learning_rate": 8.873846153846156e-06, + "loss": 0.0037, + "step": 2700 + }, + { + "epoch": 13.01, + "learning_rate": 8.861025641025641e-06, + "loss": 0.0021, + "step": 2725 + }, + { + "epoch": 13.01, + "learning_rate": 8.848205128205128e-06, + "loss": 0.0017, + "step": 2750 + }, + { + "epoch": 13.01, + "learning_rate": 8.835384615384616e-06, + "loss": 0.002, + "step": 2775 + }, + { + "epoch": 13.01, + "learning_rate": 8.822564102564103e-06, + "loss": 0.0019, + "step": 2800 + }, + { + "epoch": 14.0, + "learning_rate": 8.80974358974359e-06, + "loss": 0.0023, + "step": 2825 + }, + { + "epoch": 14.0, + "learning_rate": 8.796923076923077e-06, + "loss": 0.0043, + "step": 2850 + }, + { + "epoch": 14.0, + "learning_rate": 8.784102564102565e-06, + "loss": 0.0027, + "step": 2875 + }, + { + "epoch": 14.01, + "learning_rate": 8.771282051282052e-06, + "loss": 0.0037, + "step": 2900 + }, + { + "epoch": 14.01, + "learning_rate": 8.758461538461539e-06, + "loss": 0.0031, + "step": 2925 + }, + { + "epoch": 14.01, + "learning_rate": 8.745641025641026e-06, + "loss": 0.0034, + "step": 2950 + }, + { + "epoch": 14.01, + "learning_rate": 8.732820512820513e-06, + "loss": 0.0049, + "step": 2975 + }, + { + "epoch": 14.01, + "learning_rate": 8.720000000000001e-06, + "loss": 0.0038, + "step": 3000 + }, + { + "epoch": 14.01, + "eval_loss": 0.2291453778743744, + "eval_runtime": 2242.0438, + "eval_samples_per_second": 2.92, + "eval_steps_per_second": 0.183, + "eval_wer": 6.462127483443708, + "step": 3000 + }, + { + "epoch": 15.0, + "learning_rate": 8.707179487179488e-06, + "loss": 0.0034, + "step": 3025 + }, + { + "epoch": 15.0, + "learning_rate": 8.694358974358975e-06, + "loss": 0.0057, + "step": 3050 + }, + { + "epoch": 15.0, + "learning_rate": 8.681538461538462e-06, + "loss": 0.0058, + "step": 3075 + }, + { + "epoch": 15.01, + "learning_rate": 8.66871794871795e-06, + "loss": 0.0062, + "step": 3100 + }, + { + "epoch": 15.01, + "learning_rate": 8.655897435897437e-06, + "loss": 0.0059, + "step": 3125 + }, + { + "epoch": 15.01, + "learning_rate": 8.643076923076924e-06, + "loss": 0.0058, + "step": 3150 + }, + { + "epoch": 15.01, + "learning_rate": 8.63025641025641e-06, + "loss": 0.0067, + "step": 3175 + }, + { + "epoch": 15.01, + "learning_rate": 8.6174358974359e-06, + "loss": 0.0071, + "step": 3200 + }, + { + "epoch": 16.0, + "learning_rate": 8.604615384615386e-06, + "loss": 0.0091, + "step": 3225 + }, + { + "epoch": 16.0, + "learning_rate": 8.591794871794873e-06, + "loss": 0.0064, + "step": 3250 + }, + { + "epoch": 16.0, + "learning_rate": 8.57897435897436e-06, + "loss": 0.0044, + "step": 3275 + }, + { + "epoch": 16.0, + "learning_rate": 8.566153846153848e-06, + "loss": 0.0055, + "step": 3300 + }, + { + "epoch": 16.01, + "learning_rate": 8.553333333333333e-06, + "loss": 0.0063, + "step": 3325 + }, + { + "epoch": 16.01, + "learning_rate": 8.54051282051282e-06, + "loss": 0.0057, + "step": 3350 + }, + { + "epoch": 16.01, + "learning_rate": 8.527692307692309e-06, + "loss": 0.0049, + "step": 3375 + }, + { + "epoch": 16.01, + "learning_rate": 8.514871794871795e-06, + "loss": 0.0041, + "step": 3400 + }, + { + "epoch": 17.0, + "learning_rate": 8.502051282051282e-06, + "loss": 0.0037, + "step": 3425 + }, + { + "epoch": 17.0, + "learning_rate": 8.489230769230769e-06, + "loss": 0.0055, + "step": 3450 + }, + { + "epoch": 17.0, + "learning_rate": 8.476410256410258e-06, + "loss": 0.0048, + "step": 3475 + }, + { + "epoch": 17.0, + "learning_rate": 8.463589743589744e-06, + "loss": 0.0068, + "step": 3500 + }, + { + "epoch": 17.01, + "learning_rate": 8.450769230769231e-06, + "loss": 0.0054, + "step": 3525 + }, + { + "epoch": 17.01, + "learning_rate": 8.437948717948718e-06, + "loss": 0.0033, + "step": 3550 + }, + { + "epoch": 17.01, + "learning_rate": 8.425128205128205e-06, + "loss": 0.0048, + "step": 3575 + }, + { + "epoch": 17.01, + "learning_rate": 8.412307692307693e-06, + "loss": 0.0029, + "step": 3600 + }, + { + "epoch": 18.0, + "learning_rate": 8.39948717948718e-06, + "loss": 0.0031, + "step": 3625 + }, + { + "epoch": 18.0, + "learning_rate": 8.386666666666667e-06, + "loss": 0.0029, + "step": 3650 + }, + { + "epoch": 18.0, + "learning_rate": 8.373846153846154e-06, + "loss": 0.0054, + "step": 3675 + }, + { + "epoch": 18.0, + "learning_rate": 8.361025641025642e-06, + "loss": 0.0042, + "step": 3700 + }, + { + "epoch": 18.01, + "learning_rate": 8.34820512820513e-06, + "loss": 0.0022, + "step": 3725 + }, + { + "epoch": 18.01, + "learning_rate": 8.335384615384616e-06, + "loss": 0.0029, + "step": 3750 + }, + { + "epoch": 18.01, + "learning_rate": 8.322564102564103e-06, + "loss": 0.0038, + "step": 3775 + }, + { + "epoch": 18.01, + "learning_rate": 8.309743589743591e-06, + "loss": 0.0025, + "step": 3800 + }, + { + "epoch": 19.0, + "learning_rate": 8.296923076923078e-06, + "loss": 0.0017, + "step": 3825 + }, + { + "epoch": 19.0, + "learning_rate": 8.284102564102565e-06, + "loss": 0.0018, + "step": 3850 + }, + { + "epoch": 19.0, + "learning_rate": 8.271282051282052e-06, + "loss": 0.0027, + "step": 3875 + }, + { + "epoch": 19.0, + "learning_rate": 8.25846153846154e-06, + "loss": 0.0024, + "step": 3900 + }, + { + "epoch": 19.01, + "learning_rate": 8.245641025641027e-06, + "loss": 0.0018, + "step": 3925 + }, + { + "epoch": 19.01, + "learning_rate": 8.232820512820512e-06, + "loss": 0.0023, + "step": 3950 + }, + { + "epoch": 19.01, + "learning_rate": 8.220000000000001e-06, + "loss": 0.0028, + "step": 3975 + }, + { + "epoch": 19.01, + "learning_rate": 8.207179487179488e-06, + "loss": 0.0022, + "step": 4000 + }, + { + "epoch": 19.01, + "eval_loss": 0.24121767282485962, + "eval_runtime": 2209.526, + "eval_samples_per_second": 2.963, + "eval_steps_per_second": 0.186, + "eval_wer": 6.479373620309051, + "step": 4000 + }, + { + "epoch": 20.0, + "learning_rate": 8.194358974358975e-06, + "loss": 0.0018, + "step": 4025 + }, + { + "epoch": 20.0, + "learning_rate": 8.181538461538461e-06, + "loss": 0.0019, + "step": 4050 + }, + { + "epoch": 20.0, + "learning_rate": 8.16871794871795e-06, + "loss": 0.005, + "step": 4075 + }, + { + "epoch": 20.0, + "learning_rate": 8.155897435897437e-06, + "loss": 0.0026, + "step": 4100 + }, + { + "epoch": 20.01, + "learning_rate": 8.143076923076924e-06, + "loss": 0.0031, + "step": 4125 + }, + { + "epoch": 20.01, + "learning_rate": 8.13025641025641e-06, + "loss": 0.0026, + "step": 4150 + }, + { + "epoch": 20.01, + "learning_rate": 8.117435897435897e-06, + "loss": 0.0027, + "step": 4175 + }, + { + "epoch": 20.01, + "learning_rate": 8.104615384615386e-06, + "loss": 0.0019, + "step": 4200 + }, + { + "epoch": 21.0, + "learning_rate": 8.091794871794873e-06, + "loss": 0.0023, + "step": 4225 + }, + { + "epoch": 21.0, + "learning_rate": 8.07897435897436e-06, + "loss": 0.0017, + "step": 4250 + }, + { + "epoch": 21.0, + "learning_rate": 8.066153846153846e-06, + "loss": 0.0026, + "step": 4275 + }, + { + "epoch": 21.0, + "learning_rate": 8.053333333333335e-06, + "loss": 0.0022, + "step": 4300 + }, + { + "epoch": 21.01, + "learning_rate": 8.040512820512822e-06, + "loss": 0.0027, + "step": 4325 + }, + { + "epoch": 21.01, + "learning_rate": 8.027692307692308e-06, + "loss": 0.0034, + "step": 4350 + }, + { + "epoch": 21.01, + "learning_rate": 8.014871794871795e-06, + "loss": 0.0037, + "step": 4375 + }, + { + "epoch": 21.01, + "learning_rate": 8.002051282051284e-06, + "loss": 0.0037, + "step": 4400 + }, + { + "epoch": 22.0, + "learning_rate": 7.98923076923077e-06, + "loss": 0.003, + "step": 4425 + }, + { + "epoch": 22.0, + "learning_rate": 7.976410256410257e-06, + "loss": 0.0031, + "step": 4450 + }, + { + "epoch": 22.0, + "learning_rate": 7.963589743589744e-06, + "loss": 0.0046, + "step": 4475 + }, + { + "epoch": 22.0, + "learning_rate": 7.950769230769233e-06, + "loss": 0.0047, + "step": 4500 + }, + { + "epoch": 22.01, + "learning_rate": 7.93794871794872e-06, + "loss": 0.0041, + "step": 4525 + }, + { + "epoch": 22.01, + "learning_rate": 7.925128205128205e-06, + "loss": 0.0034, + "step": 4550 + }, + { + "epoch": 22.01, + "learning_rate": 7.912307692307693e-06, + "loss": 0.0034, + "step": 4575 + }, + { + "epoch": 22.01, + "learning_rate": 7.89948717948718e-06, + "loss": 0.0037, + "step": 4600 + }, + { + "epoch": 23.0, + "learning_rate": 7.886666666666667e-06, + "loss": 0.0036, + "step": 4625 + }, + { + "epoch": 23.0, + "learning_rate": 7.873846153846154e-06, + "loss": 0.0032, + "step": 4650 + }, + { + "epoch": 23.0, + "learning_rate": 7.861025641025642e-06, + "loss": 0.003, + "step": 4675 + }, + { + "epoch": 23.0, + "learning_rate": 7.848205128205129e-06, + "loss": 0.0035, + "step": 4700 + }, + { + "epoch": 23.01, + "learning_rate": 7.835384615384616e-06, + "loss": 0.0041, + "step": 4725 + }, + { + "epoch": 23.01, + "learning_rate": 7.822564102564103e-06, + "loss": 0.0028, + "step": 4750 + }, + { + "epoch": 23.01, + "learning_rate": 7.80974358974359e-06, + "loss": 0.0043, + "step": 4775 + }, + { + "epoch": 23.01, + "learning_rate": 7.796923076923078e-06, + "loss": 0.0026, + "step": 4800 + }, + { + "epoch": 24.0, + "learning_rate": 7.784102564102565e-06, + "loss": 0.0027, + "step": 4825 + }, + { + "epoch": 24.0, + "learning_rate": 7.771282051282052e-06, + "loss": 0.0033, + "step": 4850 + }, + { + "epoch": 24.0, + "learning_rate": 7.758461538461538e-06, + "loss": 0.0026, + "step": 4875 + }, + { + "epoch": 24.0, + "learning_rate": 7.745641025641027e-06, + "loss": 0.002, + "step": 4900 + }, + { + "epoch": 24.01, + "learning_rate": 7.732820512820514e-06, + "loss": 0.0023, + "step": 4925 + }, + { + "epoch": 24.01, + "learning_rate": 7.72e-06, + "loss": 0.0015, + "step": 4950 + }, + { + "epoch": 24.01, + "learning_rate": 7.707179487179487e-06, + "loss": 0.0014, + "step": 4975 + }, + { + "epoch": 24.01, + "learning_rate": 7.694358974358976e-06, + "loss": 0.0013, + "step": 5000 + }, + { + "epoch": 24.01, + "eval_loss": 0.2515004873275757, + "eval_runtime": 2203.2241, + "eval_samples_per_second": 2.971, + "eval_steps_per_second": 0.186, + "eval_wer": 6.46730132450331, + "step": 5000 + }, + { + "epoch": 25.0, + "learning_rate": 7.681538461538463e-06, + "loss": 0.0018, + "step": 5025 + }, + { + "epoch": 25.0, + "learning_rate": 7.66871794871795e-06, + "loss": 0.0021, + "step": 5050 + }, + { + "epoch": 25.0, + "learning_rate": 7.655897435897436e-06, + "loss": 0.0025, + "step": 5075 + }, + { + "epoch": 25.0, + "learning_rate": 7.643076923076925e-06, + "loss": 0.0023, + "step": 5100 + }, + { + "epoch": 25.01, + "learning_rate": 7.630256410256412e-06, + "loss": 0.0019, + "step": 5125 + }, + { + "epoch": 25.01, + "learning_rate": 7.617435897435898e-06, + "loss": 0.002, + "step": 5150 + }, + { + "epoch": 25.01, + "learning_rate": 7.604615384615385e-06, + "loss": 0.0023, + "step": 5175 + }, + { + "epoch": 25.01, + "learning_rate": 7.591794871794872e-06, + "loss": 0.0032, + "step": 5200 + }, + { + "epoch": 26.0, + "learning_rate": 7.578974358974359e-06, + "loss": 0.002, + "step": 5225 + }, + { + "epoch": 26.0, + "learning_rate": 7.566153846153847e-06, + "loss": 0.0019, + "step": 5250 + }, + { + "epoch": 26.0, + "learning_rate": 7.553333333333334e-06, + "loss": 0.002, + "step": 5275 + }, + { + "epoch": 26.0, + "learning_rate": 7.540512820512821e-06, + "loss": 0.0014, + "step": 5300 + }, + { + "epoch": 26.01, + "learning_rate": 7.527692307692308e-06, + "loss": 0.0016, + "step": 5325 + }, + { + "epoch": 26.01, + "learning_rate": 7.514871794871795e-06, + "loss": 0.0016, + "step": 5350 + }, + { + "epoch": 26.01, + "learning_rate": 7.5020512820512826e-06, + "loss": 0.0017, + "step": 5375 + }, + { + "epoch": 26.01, + "learning_rate": 7.489230769230769e-06, + "loss": 0.002, + "step": 5400 + }, + { + "epoch": 27.0, + "learning_rate": 7.476410256410257e-06, + "loss": 0.0028, + "step": 5425 + }, + { + "epoch": 27.0, + "learning_rate": 7.463589743589744e-06, + "loss": 0.0015, + "step": 5450 + }, + { + "epoch": 27.0, + "learning_rate": 7.4507692307692316e-06, + "loss": 0.0021, + "step": 5475 + }, + { + "epoch": 27.0, + "learning_rate": 7.437948717948718e-06, + "loss": 0.003, + "step": 5500 + }, + { + "epoch": 27.01, + "learning_rate": 7.425128205128206e-06, + "loss": 0.003, + "step": 5525 + }, + { + "epoch": 27.01, + "learning_rate": 7.412307692307693e-06, + "loss": 0.0018, + "step": 5550 + }, + { + "epoch": 27.01, + "learning_rate": 7.3994871794871806e-06, + "loss": 0.0024, + "step": 5575 + }, + { + "epoch": 27.01, + "learning_rate": 7.386666666666667e-06, + "loss": 0.0015, + "step": 5600 + }, + { + "epoch": 28.0, + "learning_rate": 7.373846153846155e-06, + "loss": 0.0018, + "step": 5625 + }, + { + "epoch": 28.0, + "learning_rate": 7.361025641025642e-06, + "loss": 0.0021, + "step": 5650 + }, + { + "epoch": 28.0, + "learning_rate": 7.3482051282051295e-06, + "loss": 0.0019, + "step": 5675 + }, + { + "epoch": 28.0, + "learning_rate": 7.335384615384616e-06, + "loss": 0.0027, + "step": 5700 + }, + { + "epoch": 28.01, + "learning_rate": 7.322564102564104e-06, + "loss": 0.0026, + "step": 5725 + }, + { + "epoch": 28.01, + "learning_rate": 7.309743589743591e-06, + "loss": 0.0028, + "step": 5750 + }, + { + "epoch": 28.01, + "learning_rate": 7.296923076923077e-06, + "loss": 0.002, + "step": 5775 + }, + { + "epoch": 28.01, + "learning_rate": 7.2841025641025645e-06, + "loss": 0.0025, + "step": 5800 + }, + { + "epoch": 29.0, + "learning_rate": 7.271282051282051e-06, + "loss": 0.002, + "step": 5825 + }, + { + "epoch": 29.0, + "learning_rate": 7.258461538461539e-06, + "loss": 0.0033, + "step": 5850 + }, + { + "epoch": 29.0, + "learning_rate": 7.245641025641026e-06, + "loss": 0.0021, + "step": 5875 + }, + { + "epoch": 29.0, + "learning_rate": 7.2328205128205135e-06, + "loss": 0.0043, + "step": 5900 + }, + { + "epoch": 29.01, + "learning_rate": 7.22e-06, + "loss": 0.0028, + "step": 5925 + }, + { + "epoch": 29.01, + "learning_rate": 7.207179487179487e-06, + "loss": 0.0028, + "step": 5950 + }, + { + "epoch": 29.01, + "learning_rate": 7.194358974358975e-06, + "loss": 0.0015, + "step": 5975 + }, + { + "epoch": 29.01, + "learning_rate": 7.181538461538462e-06, + "loss": 0.0023, + "step": 6000 + }, + { + "epoch": 29.01, + "eval_loss": 0.25697147846221924, + "eval_runtime": 2192.4122, + "eval_samples_per_second": 2.986, + "eval_steps_per_second": 0.187, + "eval_wer": 6.643211920529801, + "step": 6000 + }, + { + "epoch": 30.0, + "learning_rate": 7.168717948717949e-06, + "loss": 0.004, + "step": 6025 + }, + { + "epoch": 30.0, + "learning_rate": 7.155897435897436e-06, + "loss": 0.0029, + "step": 6050 + }, + { + "epoch": 30.0, + "learning_rate": 7.143076923076924e-06, + "loss": 0.0023, + "step": 6075 + }, + { + "epoch": 30.0, + "learning_rate": 7.130256410256411e-06, + "loss": 0.0029, + "step": 6100 + }, + { + "epoch": 30.01, + "learning_rate": 7.117435897435898e-06, + "loss": 0.0032, + "step": 6125 + }, + { + "epoch": 30.01, + "learning_rate": 7.104615384615385e-06, + "loss": 0.0018, + "step": 6150 + }, + { + "epoch": 30.01, + "learning_rate": 7.091794871794873e-06, + "loss": 0.0022, + "step": 6175 + }, + { + "epoch": 30.01, + "learning_rate": 7.07897435897436e-06, + "loss": 0.0018, + "step": 6200 + }, + { + "epoch": 31.0, + "learning_rate": 7.066153846153847e-06, + "loss": 0.0015, + "step": 6225 + }, + { + "epoch": 31.0, + "learning_rate": 7.053333333333334e-06, + "loss": 0.0019, + "step": 6250 + }, + { + "epoch": 31.0, + "learning_rate": 7.040512820512822e-06, + "loss": 0.0011, + "step": 6275 + }, + { + "epoch": 31.0, + "learning_rate": 7.027692307692309e-06, + "loss": 0.0013, + "step": 6300 + }, + { + "epoch": 31.01, + "learning_rate": 7.014871794871796e-06, + "loss": 0.0015, + "step": 6325 + }, + { + "epoch": 31.01, + "learning_rate": 7.002051282051283e-06, + "loss": 0.0016, + "step": 6350 + }, + { + "epoch": 31.01, + "learning_rate": 6.989230769230769e-06, + "loss": 0.0011, + "step": 6375 + }, + { + "epoch": 31.01, + "learning_rate": 6.976410256410257e-06, + "loss": 0.0009, + "step": 6400 + }, + { + "epoch": 32.0, + "learning_rate": 6.963589743589744e-06, + "loss": 0.0025, + "step": 6425 + }, + { + "epoch": 32.0, + "learning_rate": 6.950769230769231e-06, + "loss": 0.0027, + "step": 6450 + }, + { + "epoch": 32.0, + "learning_rate": 6.937948717948718e-06, + "loss": 0.0014, + "step": 6475 + }, + { + "epoch": 32.01, + "learning_rate": 6.925128205128206e-06, + "loss": 0.0037, + "step": 6500 + }, + { + "epoch": 32.01, + "learning_rate": 6.912307692307693e-06, + "loss": 0.0037, + "step": 6525 + }, + { + "epoch": 32.01, + "learning_rate": 6.899487179487179e-06, + "loss": 0.0025, + "step": 6550 + }, + { + "epoch": 32.01, + "learning_rate": 6.886666666666667e-06, + "loss": 0.0016, + "step": 6575 + }, + { + "epoch": 32.01, + "learning_rate": 6.873846153846154e-06, + "loss": 0.0015, + "step": 6600 + }, + { + "epoch": 33.0, + "learning_rate": 6.861025641025642e-06, + "loss": 0.0012, + "step": 6625 + }, + { + "epoch": 33.0, + "learning_rate": 6.848205128205128e-06, + "loss": 0.0013, + "step": 6650 + }, + { + "epoch": 33.0, + "learning_rate": 6.835384615384616e-06, + "loss": 0.0014, + "step": 6675 + }, + { + "epoch": 33.01, + "learning_rate": 6.822564102564103e-06, + "loss": 0.0042, + "step": 6700 + }, + { + "epoch": 33.01, + "learning_rate": 6.8097435897435906e-06, + "loss": 0.0019, + "step": 6725 + }, + { + "epoch": 33.01, + "learning_rate": 6.796923076923077e-06, + "loss": 0.0011, + "step": 6750 + }, + { + "epoch": 33.01, + "learning_rate": 6.784102564102565e-06, + "loss": 0.0008, + "step": 6775 + }, + { + "epoch": 33.01, + "learning_rate": 6.771282051282052e-06, + "loss": 0.0008, + "step": 6800 + }, + { + "epoch": 34.0, + "learning_rate": 6.7584615384615396e-06, + "loss": 0.0018, + "step": 6825 + }, + { + "epoch": 34.0, + "learning_rate": 6.745641025641026e-06, + "loss": 0.0041, + "step": 6850 + }, + { + "epoch": 34.0, + "learning_rate": 6.732820512820514e-06, + "loss": 0.0023, + "step": 6875 + }, + { + "epoch": 34.01, + "learning_rate": 6.720000000000001e-06, + "loss": 0.0021, + "step": 6900 + }, + { + "epoch": 34.01, + "learning_rate": 6.7071794871794886e-06, + "loss": 0.0023, + "step": 6925 + }, + { + "epoch": 34.01, + "learning_rate": 6.694358974358975e-06, + "loss": 0.0024, + "step": 6950 + }, + { + "epoch": 34.01, + "learning_rate": 6.681538461538463e-06, + "loss": 0.002, + "step": 6975 + }, + { + "epoch": 34.01, + "learning_rate": 6.668717948717949e-06, + "loss": 0.0018, + "step": 7000 + }, + { + "epoch": 34.01, + "eval_loss": 0.24735534191131592, + "eval_runtime": 2185.0211, + "eval_samples_per_second": 2.996, + "eval_steps_per_second": 0.188, + "eval_wer": 6.638038079470199, + "step": 7000 + }, + { + "epoch": 35.0, + "learning_rate": 6.655897435897436e-06, + "loss": 0.0017, + "step": 7025 + }, + { + "epoch": 35.0, + "learning_rate": 6.6430769230769235e-06, + "loss": 0.0021, + "step": 7050 + }, + { + "epoch": 35.0, + "learning_rate": 6.63025641025641e-06, + "loss": 0.0011, + "step": 7075 + }, + { + "epoch": 35.01, + "learning_rate": 6.617435897435898e-06, + "loss": 0.0013, + "step": 7100 + }, + { + "epoch": 35.01, + "learning_rate": 6.604615384615385e-06, + "loss": 0.0012, + "step": 7125 + }, + { + "epoch": 35.01, + "learning_rate": 6.5917948717948725e-06, + "loss": 0.0008, + "step": 7150 + }, + { + "epoch": 35.01, + "learning_rate": 6.578974358974359e-06, + "loss": 0.0007, + "step": 7175 + }, + { + "epoch": 35.01, + "learning_rate": 6.566153846153846e-06, + "loss": 0.0007, + "step": 7200 + }, + { + "epoch": 36.0, + "learning_rate": 6.553333333333334e-06, + "loss": 0.0009, + "step": 7225 + }, + { + "epoch": 36.0, + "learning_rate": 6.540512820512821e-06, + "loss": 0.001, + "step": 7250 + }, + { + "epoch": 36.0, + "learning_rate": 6.527692307692308e-06, + "loss": 0.001, + "step": 7275 + }, + { + "epoch": 36.01, + "learning_rate": 6.514871794871795e-06, + "loss": 0.0017, + "step": 7300 + }, + { + "epoch": 36.01, + "learning_rate": 6.502051282051283e-06, + "loss": 0.0021, + "step": 7325 + }, + { + "epoch": 36.01, + "learning_rate": 6.48923076923077e-06, + "loss": 0.0023, + "step": 7350 + }, + { + "epoch": 36.01, + "learning_rate": 6.476410256410257e-06, + "loss": 0.0016, + "step": 7375 + }, + { + "epoch": 36.01, + "learning_rate": 6.463589743589744e-06, + "loss": 0.0013, + "step": 7400 + }, + { + "epoch": 37.0, + "learning_rate": 6.450769230769232e-06, + "loss": 0.0013, + "step": 7425 + }, + { + "epoch": 37.0, + "learning_rate": 6.437948717948719e-06, + "loss": 0.0011, + "step": 7450 + }, + { + "epoch": 37.0, + "learning_rate": 6.425128205128206e-06, + "loss": 0.0006, + "step": 7475 + }, + { + "epoch": 37.01, + "learning_rate": 6.412307692307693e-06, + "loss": 0.0015, + "step": 7500 + }, + { + "epoch": 37.01, + "learning_rate": 6.399487179487181e-06, + "loss": 0.0019, + "step": 7525 + }, + { + "epoch": 37.01, + "learning_rate": 6.386666666666668e-06, + "loss": 0.0011, + "step": 7550 + }, + { + "epoch": 37.01, + "learning_rate": 6.373846153846155e-06, + "loss": 0.0014, + "step": 7575 + }, + { + "epoch": 37.01, + "learning_rate": 6.361025641025641e-06, + "loss": 0.0015, + "step": 7600 + }, + { + "epoch": 38.0, + "learning_rate": 6.348205128205128e-06, + "loss": 0.0018, + "step": 7625 + }, + { + "epoch": 38.0, + "learning_rate": 6.335384615384616e-06, + "loss": 0.0025, + "step": 7650 + }, + { + "epoch": 38.0, + "learning_rate": 6.322564102564103e-06, + "loss": 0.0022, + "step": 7675 + }, + { + "epoch": 38.01, + "learning_rate": 6.30974358974359e-06, + "loss": 0.0029, + "step": 7700 + }, + { + "epoch": 38.01, + "learning_rate": 6.296923076923077e-06, + "loss": 0.0022, + "step": 7725 + }, + { + "epoch": 38.01, + "learning_rate": 6.284102564102565e-06, + "loss": 0.0015, + "step": 7750 + }, + { + "epoch": 38.01, + "learning_rate": 6.271282051282052e-06, + "loss": 0.0022, + "step": 7775 + }, + { + "epoch": 38.01, + "learning_rate": 6.258461538461538e-06, + "loss": 0.0017, + "step": 7800 + }, + { + "epoch": 39.0, + "learning_rate": 6.245641025641026e-06, + "loss": 0.0018, + "step": 7825 + }, + { + "epoch": 39.0, + "learning_rate": 6.232820512820513e-06, + "loss": 0.0031, + "step": 7850 + }, + { + "epoch": 39.0, + "learning_rate": 6.220000000000001e-06, + "loss": 0.0013, + "step": 7875 + }, + { + "epoch": 39.01, + "learning_rate": 6.207179487179487e-06, + "loss": 0.0015, + "step": 7900 + }, + { + "epoch": 39.01, + "learning_rate": 6.194358974358975e-06, + "loss": 0.0017, + "step": 7925 + }, + { + "epoch": 39.01, + "learning_rate": 6.181538461538462e-06, + "loss": 0.0015, + "step": 7950 + }, + { + "epoch": 39.01, + "learning_rate": 6.16871794871795e-06, + "loss": 0.0018, + "step": 7975 + }, + { + "epoch": 39.01, + "learning_rate": 6.155897435897436e-06, + "loss": 0.0017, + "step": 8000 + }, + { + "epoch": 39.01, + "eval_loss": 0.2529506981372833, + "eval_runtime": 2279.7931, + "eval_samples_per_second": 2.871, + "eval_steps_per_second": 0.18, + "eval_wer": 6.931222406181016, + "step": 8000 + }, + { + "epoch": 40.0, + "learning_rate": 6.143076923076924e-06, + "loss": 0.0025, + "step": 8025 + }, + { + "epoch": 40.0, + "learning_rate": 6.130256410256411e-06, + "loss": 0.0024, + "step": 8050 + }, + { + "epoch": 40.0, + "learning_rate": 6.1174358974358986e-06, + "loss": 0.0022, + "step": 8075 + }, + { + "epoch": 40.01, + "learning_rate": 6.104615384615385e-06, + "loss": 0.0011, + "step": 8100 + }, + { + "epoch": 40.01, + "learning_rate": 6.091794871794873e-06, + "loss": 0.0009, + "step": 8125 + }, + { + "epoch": 40.01, + "learning_rate": 6.07897435897436e-06, + "loss": 0.0011, + "step": 8150 + }, + { + "epoch": 40.01, + "learning_rate": 6.0661538461538476e-06, + "loss": 0.0011, + "step": 8175 + }, + { + "epoch": 40.01, + "learning_rate": 6.0533333333333335e-06, + "loss": 0.0015, + "step": 8200 + }, + { + "epoch": 41.0, + "learning_rate": 6.04051282051282e-06, + "loss": 0.0021, + "step": 8225 + }, + { + "epoch": 41.0, + "learning_rate": 6.027692307692308e-06, + "loss": 0.0025, + "step": 8250 + }, + { + "epoch": 41.0, + "learning_rate": 6.014871794871795e-06, + "loss": 0.0013, + "step": 8275 + }, + { + "epoch": 41.01, + "learning_rate": 6.0020512820512825e-06, + "loss": 0.0014, + "step": 8300 + }, + { + "epoch": 41.01, + "learning_rate": 5.989230769230769e-06, + "loss": 0.0007, + "step": 8325 + }, + { + "epoch": 41.01, + "learning_rate": 5.976410256410257e-06, + "loss": 0.0007, + "step": 8350 + }, + { + "epoch": 41.01, + "learning_rate": 5.963589743589744e-06, + "loss": 0.0011, + "step": 8375 + }, + { + "epoch": 41.01, + "learning_rate": 5.950769230769231e-06, + "loss": 0.0005, + "step": 8400 + }, + { + "epoch": 42.0, + "learning_rate": 5.937948717948718e-06, + "loss": 0.0008, + "step": 8425 + }, + { + "epoch": 42.0, + "learning_rate": 5.925128205128205e-06, + "loss": 0.0008, + "step": 8450 + }, + { + "epoch": 42.0, + "learning_rate": 5.912307692307693e-06, + "loss": 0.0005, + "step": 8475 + }, + { + "epoch": 42.01, + "learning_rate": 5.89948717948718e-06, + "loss": 0.0006, + "step": 8500 + }, + { + "epoch": 42.01, + "learning_rate": 5.886666666666667e-06, + "loss": 0.0005, + "step": 8525 + }, + { + "epoch": 42.01, + "learning_rate": 5.873846153846154e-06, + "loss": 0.0012, + "step": 8550 + }, + { + "epoch": 42.01, + "learning_rate": 5.861025641025642e-06, + "loss": 0.0006, + "step": 8575 + }, + { + "epoch": 42.01, + "learning_rate": 5.848205128205129e-06, + "loss": 0.0004, + "step": 8600 + }, + { + "epoch": 43.0, + "learning_rate": 5.835384615384616e-06, + "loss": 0.0008, + "step": 8625 + }, + { + "epoch": 43.0, + "learning_rate": 5.822564102564103e-06, + "loss": 0.0006, + "step": 8650 + }, + { + "epoch": 43.0, + "learning_rate": 5.809743589743591e-06, + "loss": 0.001, + "step": 8675 + }, + { + "epoch": 43.01, + "learning_rate": 5.796923076923078e-06, + "loss": 0.0004, + "step": 8700 + }, + { + "epoch": 43.01, + "learning_rate": 5.784102564102565e-06, + "loss": 0.0003, + "step": 8725 + }, + { + "epoch": 43.01, + "learning_rate": 5.771282051282052e-06, + "loss": 0.0003, + "step": 8750 + }, + { + "epoch": 43.01, + "learning_rate": 5.75846153846154e-06, + "loss": 0.0002, + "step": 8775 + }, + { + "epoch": 43.01, + "learning_rate": 5.745641025641027e-06, + "loss": 0.0002, + "step": 8800 + }, + { + "epoch": 44.0, + "learning_rate": 5.732820512820513e-06, + "loss": 0.0003, + "step": 8825 + }, + { + "epoch": 44.0, + "learning_rate": 5.72e-06, + "loss": 0.0003, + "step": 8850 + }, + { + "epoch": 44.0, + "learning_rate": 5.707179487179487e-06, + "loss": 0.0002, + "step": 8875 + }, + { + "epoch": 44.01, + "learning_rate": 5.694358974358975e-06, + "loss": 0.0004, + "step": 8900 + }, + { + "epoch": 44.01, + "learning_rate": 5.681538461538462e-06, + "loss": 0.0002, + "step": 8925 + }, + { + "epoch": 44.01, + "learning_rate": 5.668717948717949e-06, + "loss": 0.0002, + "step": 8950 + }, + { + "epoch": 44.01, + "learning_rate": 5.655897435897436e-06, + "loss": 0.0002, + "step": 8975 + }, + { + "epoch": 44.01, + "learning_rate": 5.643076923076923e-06, + "loss": 0.0001, + "step": 9000 + }, + { + "epoch": 44.01, + "eval_loss": 0.27584660053253174, + "eval_runtime": 2185.9433, + "eval_samples_per_second": 2.995, + "eval_steps_per_second": 0.188, + "eval_wer": 6.23792770419426, + "step": 9000 + }, + { + "epoch": 45.0, + "learning_rate": 5.630256410256411e-06, + "loss": 0.0002, + "step": 9025 + }, + { + "epoch": 45.0, + "learning_rate": 5.6174358974358974e-06, + "loss": 0.0001, + "step": 9050 + }, + { + "epoch": 45.0, + "learning_rate": 5.604615384615385e-06, + "loss": 0.0001, + "step": 9075 + }, + { + "epoch": 45.01, + "learning_rate": 5.591794871794872e-06, + "loss": 0.0001, + "step": 9100 + }, + { + "epoch": 45.01, + "learning_rate": 5.57897435897436e-06, + "loss": 0.0001, + "step": 9125 + }, + { + "epoch": 45.01, + "learning_rate": 5.566153846153846e-06, + "loss": 0.0001, + "step": 9150 + }, + { + "epoch": 45.01, + "learning_rate": 5.553333333333334e-06, + "loss": 0.0001, + "step": 9175 + }, + { + "epoch": 45.01, + "learning_rate": 5.540512820512821e-06, + "loss": 0.0001, + "step": 9200 + }, + { + "epoch": 46.0, + "learning_rate": 5.527692307692309e-06, + "loss": 0.0001, + "step": 9225 + }, + { + "epoch": 46.0, + "learning_rate": 5.514871794871795e-06, + "loss": 0.0001, + "step": 9250 + }, + { + "epoch": 46.0, + "learning_rate": 5.502051282051283e-06, + "loss": 0.0001, + "step": 9275 + }, + { + "epoch": 46.01, + "learning_rate": 5.48923076923077e-06, + "loss": 0.0001, + "step": 9300 + }, + { + "epoch": 46.01, + "learning_rate": 5.476410256410258e-06, + "loss": 0.0001, + "step": 9325 + }, + { + "epoch": 46.01, + "learning_rate": 5.463589743589744e-06, + "loss": 0.0001, + "step": 9350 + }, + { + "epoch": 46.01, + "learning_rate": 5.450769230769232e-06, + "loss": 0.0001, + "step": 9375 + }, + { + "epoch": 46.01, + "learning_rate": 5.437948717948719e-06, + "loss": 0.0001, + "step": 9400 + }, + { + "epoch": 47.0, + "learning_rate": 5.425128205128205e-06, + "loss": 0.0001, + "step": 9425 + }, + { + "epoch": 47.0, + "learning_rate": 5.4123076923076925e-06, + "loss": 0.0001, + "step": 9450 + }, + { + "epoch": 47.0, + "learning_rate": 5.399487179487179e-06, + "loss": 0.0001, + "step": 9475 + }, + { + "epoch": 47.01, + "learning_rate": 5.386666666666667e-06, + "loss": 0.0001, + "step": 9500 + }, + { + "epoch": 47.01, + "learning_rate": 5.373846153846154e-06, + "loss": 0.0001, + "step": 9525 + }, + { + "epoch": 47.01, + "learning_rate": 5.3610256410256415e-06, + "loss": 0.0001, + "step": 9550 + }, + { + "epoch": 47.01, + "learning_rate": 5.348205128205128e-06, + "loss": 0.0001, + "step": 9575 + }, + { + "epoch": 47.01, + "learning_rate": 5.335384615384615e-06, + "loss": 0.0001, + "step": 9600 + }, + { + "epoch": 48.0, + "learning_rate": 5.322564102564103e-06, + "loss": 0.0001, + "step": 9625 + }, + { + "epoch": 48.0, + "learning_rate": 5.30974358974359e-06, + "loss": 0.0001, + "step": 9650 + }, + { + "epoch": 48.0, + "learning_rate": 5.296923076923077e-06, + "loss": 0.0001, + "step": 9675 + }, + { + "epoch": 48.01, + "learning_rate": 5.284102564102564e-06, + "loss": 0.0001, + "step": 9700 + }, + { + "epoch": 48.01, + "learning_rate": 5.271282051282052e-06, + "loss": 0.0001, + "step": 9725 + }, + { + "epoch": 48.01, + "learning_rate": 5.258461538461539e-06, + "loss": 0.0001, + "step": 9750 + }, + { + "epoch": 48.01, + "learning_rate": 5.245641025641026e-06, + "loss": 0.0001, + "step": 9775 + }, + { + "epoch": 48.01, + "learning_rate": 5.232820512820513e-06, + "loss": 0.0001, + "step": 9800 + }, + { + "epoch": 49.0, + "learning_rate": 5.220000000000001e-06, + "loss": 0.0001, + "step": 9825 + }, + { + "epoch": 49.0, + "learning_rate": 5.207179487179488e-06, + "loss": 0.0001, + "step": 9850 + }, + { + "epoch": 49.0, + "learning_rate": 5.194358974358975e-06, + "loss": 0.0001, + "step": 9875 + }, + { + "epoch": 49.01, + "learning_rate": 5.181538461538462e-06, + "loss": 0.0001, + "step": 9900 + }, + { + "epoch": 49.01, + "learning_rate": 5.16871794871795e-06, + "loss": 0.0001, + "step": 9925 + }, + { + "epoch": 49.01, + "learning_rate": 5.155897435897437e-06, + "loss": 0.0001, + "step": 9950 + }, + { + "epoch": 49.01, + "learning_rate": 5.143076923076924e-06, + "loss": 0.0001, + "step": 9975 + }, + { + "epoch": 49.01, + "learning_rate": 5.130256410256411e-06, + "loss": 0.0001, + "step": 10000 + }, + { + "epoch": 49.01, + "eval_loss": 0.2951536476612091, + "eval_runtime": 2247.8627, + "eval_samples_per_second": 2.912, + "eval_steps_per_second": 0.182, + "eval_wer": 6.1241032008830025, + "step": 10000 + }, + { + "epoch": 50.0, + "learning_rate": 5.117435897435897e-06, + "loss": 0.0001, + "step": 10025 + }, + { + "epoch": 50.0, + "learning_rate": 5.104615384615385e-06, + "loss": 0.0001, + "step": 10050 + }, + { + "epoch": 50.0, + "learning_rate": 5.091794871794872e-06, + "loss": 0.0001, + "step": 10075 + }, + { + "epoch": 50.01, + "learning_rate": 5.078974358974359e-06, + "loss": 0.0001, + "step": 10100 + }, + { + "epoch": 50.01, + "learning_rate": 5.066153846153846e-06, + "loss": 0.0001, + "step": 10125 + }, + { + "epoch": 50.01, + "learning_rate": 5.053333333333334e-06, + "loss": 0.0001, + "step": 10150 + }, + { + "epoch": 50.01, + "learning_rate": 5.040512820512821e-06, + "loss": 0.0001, + "step": 10175 + }, + { + "epoch": 50.01, + "learning_rate": 5.027692307692308e-06, + "loss": 0.0001, + "step": 10200 + }, + { + "epoch": 51.0, + "learning_rate": 5.014871794871795e-06, + "loss": 0.0001, + "step": 10225 + }, + { + "epoch": 51.0, + "learning_rate": 5.002051282051282e-06, + "loss": 0.0001, + "step": 10250 + }, + { + "epoch": 51.0, + "learning_rate": 4.98923076923077e-06, + "loss": 0.0001, + "step": 10275 + }, + { + "epoch": 51.01, + "learning_rate": 4.9764102564102564e-06, + "loss": 0.0001, + "step": 10300 + }, + { + "epoch": 51.01, + "learning_rate": 4.963589743589744e-06, + "loss": 0.0001, + "step": 10325 + }, + { + "epoch": 51.01, + "learning_rate": 4.950769230769231e-06, + "loss": 0.0001, + "step": 10350 + }, + { + "epoch": 51.01, + "learning_rate": 4.937948717948719e-06, + "loss": 0.0001, + "step": 10375 + }, + { + "epoch": 51.01, + "learning_rate": 4.9251282051282054e-06, + "loss": 0.0001, + "step": 10400 + }, + { + "epoch": 52.0, + "learning_rate": 4.912307692307693e-06, + "loss": 0.0001, + "step": 10425 + }, + { + "epoch": 52.0, + "learning_rate": 4.89948717948718e-06, + "loss": 0.0001, + "step": 10450 + }, + { + "epoch": 52.0, + "learning_rate": 4.886666666666668e-06, + "loss": 0.0001, + "step": 10475 + }, + { + "epoch": 52.01, + "learning_rate": 4.873846153846154e-06, + "loss": 0.0001, + "step": 10500 + }, + { + "epoch": 52.01, + "learning_rate": 4.861025641025641e-06, + "loss": 0.0001, + "step": 10525 + }, + { + "epoch": 52.01, + "learning_rate": 4.848205128205128e-06, + "loss": 0.0001, + "step": 10550 + }, + { + "epoch": 52.01, + "learning_rate": 4.835384615384616e-06, + "loss": 0.0001, + "step": 10575 + }, + { + "epoch": 52.01, + "learning_rate": 4.8225641025641026e-06, + "loss": 0.0001, + "step": 10600 + }, + { + "epoch": 53.0, + "learning_rate": 4.80974358974359e-06, + "loss": 0.0001, + "step": 10625 + }, + { + "epoch": 53.0, + "learning_rate": 4.796923076923077e-06, + "loss": 0.0001, + "step": 10650 + }, + { + "epoch": 53.0, + "learning_rate": 4.784102564102565e-06, + "loss": 0.0001, + "step": 10675 + }, + { + "epoch": 53.01, + "learning_rate": 4.7712820512820516e-06, + "loss": 0.0001, + "step": 10700 + }, + { + "epoch": 53.01, + "learning_rate": 4.758461538461539e-06, + "loss": 0.0001, + "step": 10725 + }, + { + "epoch": 53.01, + "learning_rate": 4.745641025641026e-06, + "loss": 0.0001, + "step": 10750 + }, + { + "epoch": 53.01, + "learning_rate": 4.732820512820514e-06, + "loss": 0.0001, + "step": 10775 + }, + { + "epoch": 53.01, + "learning_rate": 4.7200000000000005e-06, + "loss": 0.0001, + "step": 10800 + }, + { + "epoch": 54.0, + "learning_rate": 4.707179487179487e-06, + "loss": 0.0001, + "step": 10825 + }, + { + "epoch": 54.0, + "learning_rate": 4.694358974358974e-06, + "loss": 0.0001, + "step": 10850 + }, + { + "epoch": 54.0, + "learning_rate": 4.681538461538462e-06, + "loss": 0.0001, + "step": 10875 + }, + { + "epoch": 54.01, + "learning_rate": 4.668717948717949e-06, + "loss": 0.0001, + "step": 10900 + }, + { + "epoch": 54.01, + "learning_rate": 4.655897435897436e-06, + "loss": 0.0001, + "step": 10925 + }, + { + "epoch": 54.01, + "learning_rate": 4.643076923076923e-06, + "loss": 0.0001, + "step": 10950 + }, + { + "epoch": 54.01, + "learning_rate": 4.630256410256411e-06, + "loss": 0.0001, + "step": 10975 + }, + { + "epoch": 54.01, + "learning_rate": 4.617435897435898e-06, + "loss": 0.0001, + "step": 11000 + }, + { + "epoch": 54.01, + "eval_loss": 0.3056494891643524, + "eval_runtime": 2202.9041, + "eval_samples_per_second": 2.972, + "eval_steps_per_second": 0.186, + "eval_wer": 6.049944812362031, + "step": 11000 + }, + { + "epoch": 55.0, + "learning_rate": 4.604615384615385e-06, + "loss": 0.0001, + "step": 11025 + }, + { + "epoch": 55.0, + "learning_rate": 4.591794871794872e-06, + "loss": 0.0001, + "step": 11050 + }, + { + "epoch": 55.0, + "learning_rate": 4.57897435897436e-06, + "loss": 0.0001, + "step": 11075 + }, + { + "epoch": 55.01, + "learning_rate": 4.566153846153847e-06, + "loss": 0.0001, + "step": 11100 + }, + { + "epoch": 55.01, + "learning_rate": 4.5533333333333335e-06, + "loss": 0.0001, + "step": 11125 + }, + { + "epoch": 55.01, + "learning_rate": 4.54051282051282e-06, + "loss": 0.0001, + "step": 11150 + }, + { + "epoch": 55.01, + "learning_rate": 4.527692307692308e-06, + "loss": 0.0001, + "step": 11175 + }, + { + "epoch": 55.01, + "learning_rate": 4.514871794871795e-06, + "loss": 0.0001, + "step": 11200 + }, + { + "epoch": 56.0, + "learning_rate": 4.5020512820512825e-06, + "loss": 0.0001, + "step": 11225 + }, + { + "epoch": 56.0, + "learning_rate": 4.489230769230769e-06, + "loss": 0.0001, + "step": 11250 + }, + { + "epoch": 56.0, + "learning_rate": 4.476410256410257e-06, + "loss": 0.0001, + "step": 11275 + }, + { + "epoch": 56.01, + "learning_rate": 4.463589743589744e-06, + "loss": 0.0001, + "step": 11300 + }, + { + "epoch": 56.01, + "learning_rate": 4.4507692307692315e-06, + "loss": 0.0001, + "step": 11325 + }, + { + "epoch": 56.01, + "learning_rate": 4.437948717948718e-06, + "loss": 0.0001, + "step": 11350 + }, + { + "epoch": 56.01, + "learning_rate": 4.425128205128206e-06, + "loss": 0.0001, + "step": 11375 + }, + { + "epoch": 56.01, + "learning_rate": 4.412307692307693e-06, + "loss": 0.0001, + "step": 11400 + }, + { + "epoch": 57.0, + "learning_rate": 4.39948717948718e-06, + "loss": 0.0001, + "step": 11425 + }, + { + "epoch": 57.0, + "learning_rate": 4.3866666666666665e-06, + "loss": 0.0001, + "step": 11450 + }, + { + "epoch": 57.0, + "learning_rate": 4.373846153846154e-06, + "loss": 0.0001, + "step": 11475 + }, + { + "epoch": 57.01, + "learning_rate": 4.361025641025641e-06, + "loss": 0.0001, + "step": 11500 + }, + { + "epoch": 57.01, + "learning_rate": 4.348205128205129e-06, + "loss": 0.0001, + "step": 11525 + }, + { + "epoch": 57.01, + "learning_rate": 4.3353846153846154e-06, + "loss": 0.0001, + "step": 11550 + }, + { + "epoch": 57.01, + "learning_rate": 4.322564102564103e-06, + "loss": 0.0001, + "step": 11575 + }, + { + "epoch": 57.01, + "learning_rate": 4.30974358974359e-06, + "loss": 0.0001, + "step": 11600 + }, + { + "epoch": 58.0, + "learning_rate": 4.296923076923078e-06, + "loss": 0.0001, + "step": 11625 + }, + { + "epoch": 58.0, + "learning_rate": 4.2841025641025644e-06, + "loss": 0.0001, + "step": 11650 + }, + { + "epoch": 58.0, + "learning_rate": 4.271282051282052e-06, + "loss": 0.0001, + "step": 11675 + }, + { + "epoch": 58.01, + "learning_rate": 4.258461538461539e-06, + "loss": 0.0001, + "step": 11700 + }, + { + "epoch": 58.01, + "learning_rate": 4.245641025641026e-06, + "loss": 0.0001, + "step": 11725 + }, + { + "epoch": 58.01, + "learning_rate": 4.232820512820513e-06, + "loss": 0.0001, + "step": 11750 + }, + { + "epoch": 58.01, + "learning_rate": 4.22e-06, + "loss": 0.0001, + "step": 11775 + }, + { + "epoch": 58.01, + "learning_rate": 4.207179487179487e-06, + "loss": 0.0, + "step": 11800 + }, + { + "epoch": 59.0, + "learning_rate": 4.194358974358975e-06, + "loss": 0.0, + "step": 11825 + }, + { + "epoch": 59.0, + "learning_rate": 4.1815384615384616e-06, + "loss": 0.0001, + "step": 11850 + }, + { + "epoch": 59.0, + "learning_rate": 4.168717948717949e-06, + "loss": 0.0, + "step": 11875 + }, + { + "epoch": 59.01, + "learning_rate": 4.155897435897436e-06, + "loss": 0.0001, + "step": 11900 + }, + { + "epoch": 59.01, + "learning_rate": 4.143076923076924e-06, + "loss": 0.0, + "step": 11925 + }, + { + "epoch": 59.01, + "learning_rate": 4.1302564102564106e-06, + "loss": 0.0, + "step": 11950 + }, + { + "epoch": 59.01, + "learning_rate": 4.117435897435898e-06, + "loss": 0.0, + "step": 11975 + }, + { + "epoch": 59.01, + "learning_rate": 4.104615384615385e-06, + "loss": 0.0, + "step": 12000 + }, + { + "epoch": 59.01, + "eval_loss": 0.31518158316612244, + "eval_runtime": 2241.8161, + "eval_samples_per_second": 2.92, + "eval_steps_per_second": 0.183, + "eval_wer": 5.994757174392936, + "step": 12000 + }, + { + "epoch": 60.0, + "learning_rate": 4.091794871794872e-06, + "loss": 0.0, + "step": 12025 + }, + { + "epoch": 60.0, + "learning_rate": 4.078974358974359e-06, + "loss": 0.0, + "step": 12050 + }, + { + "epoch": 60.0, + "learning_rate": 4.066153846153846e-06, + "loss": 0.0, + "step": 12075 + }, + { + "epoch": 60.01, + "learning_rate": 4.053333333333333e-06, + "loss": 0.0, + "step": 12100 + }, + { + "epoch": 60.01, + "learning_rate": 4.040512820512821e-06, + "loss": 0.0, + "step": 12125 + }, + { + "epoch": 60.01, + "learning_rate": 4.027692307692308e-06, + "loss": 0.0, + "step": 12150 + }, + { + "epoch": 60.01, + "learning_rate": 4.014871794871795e-06, + "loss": 0.0, + "step": 12175 + }, + { + "epoch": 60.01, + "learning_rate": 4.002051282051282e-06, + "loss": 0.0, + "step": 12200 + }, + { + "epoch": 61.0, + "learning_rate": 3.98923076923077e-06, + "loss": 0.0, + "step": 12225 + }, + { + "epoch": 61.0, + "learning_rate": 3.976410256410257e-06, + "loss": 0.0, + "step": 12250 + }, + { + "epoch": 61.0, + "learning_rate": 3.963589743589744e-06, + "loss": 0.0, + "step": 12275 + }, + { + "epoch": 61.01, + "learning_rate": 3.950769230769231e-06, + "loss": 0.0, + "step": 12300 + }, + { + "epoch": 61.01, + "learning_rate": 3.937948717948718e-06, + "loss": 0.0, + "step": 12325 + }, + { + "epoch": 61.01, + "learning_rate": 3.925128205128205e-06, + "loss": 0.0, + "step": 12350 + }, + { + "epoch": 61.01, + "learning_rate": 3.9123076923076925e-06, + "loss": 0.0, + "step": 12375 + }, + { + "epoch": 61.01, + "learning_rate": 3.899487179487179e-06, + "loss": 0.0, + "step": 12400 + }, + { + "epoch": 62.0, + "learning_rate": 3.886666666666667e-06, + "loss": 0.0, + "step": 12425 + }, + { + "epoch": 62.0, + "learning_rate": 3.873846153846154e-06, + "loss": 0.0, + "step": 12450 + }, + { + "epoch": 62.0, + "learning_rate": 3.8610256410256415e-06, + "loss": 0.0, + "step": 12475 + }, + { + "epoch": 62.01, + "learning_rate": 3.848205128205128e-06, + "loss": 0.0, + "step": 12500 + }, + { + "epoch": 62.01, + "learning_rate": 3.835384615384616e-06, + "loss": 0.0, + "step": 12525 + }, + { + "epoch": 62.01, + "learning_rate": 3.822564102564103e-06, + "loss": 0.0, + "step": 12550 + }, + { + "epoch": 62.01, + "learning_rate": 3.80974358974359e-06, + "loss": 0.0, + "step": 12575 + }, + { + "epoch": 62.01, + "learning_rate": 3.7969230769230773e-06, + "loss": 0.0, + "step": 12600 + }, + { + "epoch": 63.0, + "learning_rate": 3.784102564102564e-06, + "loss": 0.0, + "step": 12625 + }, + { + "epoch": 63.0, + "learning_rate": 3.7712820512820514e-06, + "loss": 0.0, + "step": 12650 + }, + { + "epoch": 63.0, + "learning_rate": 3.7584615384615386e-06, + "loss": 0.0, + "step": 12675 + }, + { + "epoch": 63.01, + "learning_rate": 3.745641025641026e-06, + "loss": 0.0, + "step": 12700 + }, + { + "epoch": 63.01, + "learning_rate": 3.732820512820513e-06, + "loss": 0.0, + "step": 12725 + }, + { + "epoch": 63.01, + "learning_rate": 3.7200000000000004e-06, + "loss": 0.0, + "step": 12750 + }, + { + "epoch": 63.01, + "learning_rate": 3.7071794871794876e-06, + "loss": 0.0, + "step": 12775 + }, + { + "epoch": 63.01, + "learning_rate": 3.694358974358975e-06, + "loss": 0.0, + "step": 12800 + }, + { + "epoch": 64.0, + "learning_rate": 3.681538461538462e-06, + "loss": 0.0, + "step": 12825 + }, + { + "epoch": 64.0, + "learning_rate": 3.6687179487179494e-06, + "loss": 0.0, + "step": 12850 + }, + { + "epoch": 64.0, + "learning_rate": 3.655897435897436e-06, + "loss": 0.0, + "step": 12875 + }, + { + "epoch": 64.0, + "learning_rate": 3.6430769230769234e-06, + "loss": 0.0, + "step": 12900 + }, + { + "epoch": 64.01, + "learning_rate": 3.6302564102564103e-06, + "loss": 0.0, + "step": 12925 + }, + { + "epoch": 64.01, + "learning_rate": 3.6174358974358975e-06, + "loss": 0.0, + "step": 12950 + }, + { + "epoch": 64.01, + "learning_rate": 3.6046153846153848e-06, + "loss": 0.0, + "step": 12975 + }, + { + "epoch": 64.01, + "learning_rate": 3.591794871794872e-06, + "loss": 0.0, + "step": 13000 + }, + { + "epoch": 64.01, + "eval_loss": 0.3244025409221649, + "eval_runtime": 2199.1403, + "eval_samples_per_second": 2.977, + "eval_steps_per_second": 0.186, + "eval_wer": 6.030974061810154, + "step": 13000 + }, + { + "epoch": 65.0, + "learning_rate": 3.5789743589743593e-06, + "loss": 0.0, + "step": 13025 + }, + { + "epoch": 65.0, + "learning_rate": 3.5661538461538465e-06, + "loss": 0.0, + "step": 13050 + }, + { + "epoch": 65.0, + "learning_rate": 3.5533333333333338e-06, + "loss": 0.0, + "step": 13075 + }, + { + "epoch": 65.0, + "learning_rate": 3.540512820512821e-06, + "loss": 0.0, + "step": 13100 + }, + { + "epoch": 65.01, + "learning_rate": 3.5276923076923083e-06, + "loss": 0.0, + "step": 13125 + }, + { + "epoch": 65.01, + "learning_rate": 3.5148717948717955e-06, + "loss": 0.0, + "step": 13150 + }, + { + "epoch": 65.01, + "learning_rate": 3.5020512820512823e-06, + "loss": 0.0, + "step": 13175 + }, + { + "epoch": 65.01, + "learning_rate": 3.4892307692307696e-06, + "loss": 0.0, + "step": 13200 + }, + { + "epoch": 66.0, + "learning_rate": 3.4764102564102564e-06, + "loss": 0.0, + "step": 13225 + }, + { + "epoch": 66.0, + "learning_rate": 3.4635897435897436e-06, + "loss": 0.0, + "step": 13250 + }, + { + "epoch": 66.0, + "learning_rate": 3.450769230769231e-06, + "loss": 0.0, + "step": 13275 + }, + { + "epoch": 66.0, + "learning_rate": 3.437948717948718e-06, + "loss": 0.0, + "step": 13300 + }, + { + "epoch": 66.01, + "learning_rate": 3.4251282051282054e-06, + "loss": 0.0, + "step": 13325 + }, + { + "epoch": 66.01, + "learning_rate": 3.4123076923076926e-06, + "loss": 0.0, + "step": 13350 + }, + { + "epoch": 66.01, + "learning_rate": 3.39948717948718e-06, + "loss": 0.0, + "step": 13375 + }, + { + "epoch": 66.01, + "learning_rate": 3.386666666666667e-06, + "loss": 0.0, + "step": 13400 + }, + { + "epoch": 67.0, + "learning_rate": 3.3738461538461544e-06, + "loss": 0.0, + "step": 13425 + }, + { + "epoch": 67.0, + "learning_rate": 3.3610256410256416e-06, + "loss": 0.0, + "step": 13450 + }, + { + "epoch": 67.0, + "learning_rate": 3.3482051282051285e-06, + "loss": 0.0, + "step": 13475 + }, + { + "epoch": 67.0, + "learning_rate": 3.3353846153846157e-06, + "loss": 0.0, + "step": 13500 + }, + { + "epoch": 67.01, + "learning_rate": 3.322564102564103e-06, + "loss": 0.0, + "step": 13525 + }, + { + "epoch": 67.01, + "learning_rate": 3.3097435897435898e-06, + "loss": 0.0, + "step": 13550 + }, + { + "epoch": 67.01, + "learning_rate": 3.296923076923077e-06, + "loss": 0.0, + "step": 13575 + }, + { + "epoch": 67.01, + "learning_rate": 3.2841025641025643e-06, + "loss": 0.0, + "step": 13600 + }, + { + "epoch": 68.0, + "learning_rate": 3.2712820512820515e-06, + "loss": 0.0, + "step": 13625 + }, + { + "epoch": 68.0, + "learning_rate": 3.2584615384615388e-06, + "loss": 0.0, + "step": 13650 + }, + { + "epoch": 68.0, + "learning_rate": 3.245641025641026e-06, + "loss": 0.0, + "step": 13675 + }, + { + "epoch": 68.0, + "learning_rate": 3.2328205128205133e-06, + "loss": 0.0, + "step": 13700 + }, + { + "epoch": 68.01, + "learning_rate": 3.2200000000000005e-06, + "loss": 0.0, + "step": 13725 + }, + { + "epoch": 68.01, + "learning_rate": 3.2071794871794878e-06, + "loss": 0.0, + "step": 13750 + }, + { + "epoch": 68.01, + "learning_rate": 3.194358974358975e-06, + "loss": 0.0, + "step": 13775 + }, + { + "epoch": 68.01, + "learning_rate": 3.181538461538462e-06, + "loss": 0.0, + "step": 13800 + }, + { + "epoch": 69.0, + "learning_rate": 3.168717948717949e-06, + "loss": 0.0, + "step": 13825 + }, + { + "epoch": 69.0, + "learning_rate": 3.155897435897436e-06, + "loss": 0.0, + "step": 13850 + }, + { + "epoch": 69.0, + "learning_rate": 3.143076923076923e-06, + "loss": 0.0, + "step": 13875 + }, + { + "epoch": 69.0, + "learning_rate": 3.1302564102564104e-06, + "loss": 0.0, + "step": 13900 + }, + { + "epoch": 69.01, + "learning_rate": 3.1174358974358976e-06, + "loss": 0.0, + "step": 13925 + }, + { + "epoch": 69.01, + "learning_rate": 3.104615384615385e-06, + "loss": 0.0, + "step": 13950 + }, + { + "epoch": 69.01, + "learning_rate": 3.091794871794872e-06, + "loss": 0.0, + "step": 13975 + }, + { + "epoch": 69.01, + "learning_rate": 3.0789743589743594e-06, + "loss": 0.0, + "step": 14000 + }, + { + "epoch": 69.01, + "eval_loss": 0.33362671732902527, + "eval_runtime": 2201.1397, + "eval_samples_per_second": 2.974, + "eval_steps_per_second": 0.186, + "eval_wer": 6.058567880794702, + "step": 14000 + }, + { + "epoch": 70.0, + "learning_rate": 3.0661538461538466e-06, + "loss": 0.0, + "step": 14025 + }, + { + "epoch": 70.0, + "learning_rate": 3.053333333333334e-06, + "loss": 0.0, + "step": 14050 + }, + { + "epoch": 70.0, + "learning_rate": 3.040512820512821e-06, + "loss": 0.0, + "step": 14075 + }, + { + "epoch": 70.0, + "learning_rate": 3.027692307692308e-06, + "loss": 0.0, + "step": 14100 + }, + { + "epoch": 70.01, + "learning_rate": 3.014871794871795e-06, + "loss": 0.0, + "step": 14125 + }, + { + "epoch": 70.01, + "learning_rate": 3.002051282051282e-06, + "loss": 0.0, + "step": 14150 + }, + { + "epoch": 70.01, + "learning_rate": 2.9892307692307693e-06, + "loss": 0.0, + "step": 14175 + }, + { + "epoch": 70.01, + "learning_rate": 2.9764102564102565e-06, + "loss": 0.0, + "step": 14200 + }, + { + "epoch": 71.0, + "learning_rate": 2.9635897435897438e-06, + "loss": 0.0, + "step": 14225 + }, + { + "epoch": 71.0, + "learning_rate": 2.950769230769231e-06, + "loss": 0.0, + "step": 14250 + }, + { + "epoch": 71.0, + "learning_rate": 2.9379487179487183e-06, + "loss": 0.0, + "step": 14275 + }, + { + "epoch": 71.0, + "learning_rate": 2.9251282051282055e-06, + "loss": 0.0, + "step": 14300 + }, + { + "epoch": 71.01, + "learning_rate": 2.9123076923076928e-06, + "loss": 0.0, + "step": 14325 + }, + { + "epoch": 71.01, + "learning_rate": 2.89948717948718e-06, + "loss": 0.0, + "step": 14350 + }, + { + "epoch": 71.01, + "learning_rate": 2.8866666666666673e-06, + "loss": 0.0, + "step": 14375 + }, + { + "epoch": 71.01, + "learning_rate": 2.873846153846154e-06, + "loss": 0.0, + "step": 14400 + }, + { + "epoch": 72.0, + "learning_rate": 2.8610256410256413e-06, + "loss": 0.0, + "step": 14425 + }, + { + "epoch": 72.0, + "learning_rate": 2.848205128205128e-06, + "loss": 0.0, + "step": 14450 + }, + { + "epoch": 72.0, + "learning_rate": 2.8353846153846154e-06, + "loss": 0.0, + "step": 14475 + }, + { + "epoch": 72.0, + "learning_rate": 2.8225641025641027e-06, + "loss": 0.0, + "step": 14500 + }, + { + "epoch": 72.01, + "learning_rate": 2.80974358974359e-06, + "loss": 0.0, + "step": 14525 + }, + { + "epoch": 72.01, + "learning_rate": 2.796923076923077e-06, + "loss": 0.0, + "step": 14550 + }, + { + "epoch": 72.01, + "learning_rate": 2.7841025641025644e-06, + "loss": 0.0, + "step": 14575 + }, + { + "epoch": 72.01, + "learning_rate": 2.7712820512820516e-06, + "loss": 0.0, + "step": 14600 + }, + { + "epoch": 73.0, + "learning_rate": 2.758461538461539e-06, + "loss": 0.0, + "step": 14625 + }, + { + "epoch": 73.0, + "learning_rate": 2.745641025641026e-06, + "loss": 0.0, + "step": 14650 + }, + { + "epoch": 73.0, + "learning_rate": 2.7328205128205134e-06, + "loss": 0.0, + "step": 14675 + }, + { + "epoch": 73.0, + "learning_rate": 2.7200000000000002e-06, + "loss": 0.0, + "step": 14700 + }, + { + "epoch": 73.01, + "learning_rate": 2.7071794871794875e-06, + "loss": 0.0, + "step": 14725 + }, + { + "epoch": 73.01, + "learning_rate": 2.6943589743589743e-06, + "loss": 0.0, + "step": 14750 + }, + { + "epoch": 73.01, + "learning_rate": 2.6815384615384615e-06, + "loss": 0.0, + "step": 14775 + }, + { + "epoch": 73.01, + "learning_rate": 2.6687179487179488e-06, + "loss": 0.0, + "step": 14800 + }, + { + "epoch": 74.0, + "learning_rate": 2.655897435897436e-06, + "loss": 0.0, + "step": 14825 + }, + { + "epoch": 74.0, + "learning_rate": 2.6430769230769233e-06, + "loss": 0.0, + "step": 14850 + }, + { + "epoch": 74.0, + "learning_rate": 2.6302564102564105e-06, + "loss": 0.0, + "step": 14875 + }, + { + "epoch": 74.0, + "learning_rate": 2.6174358974358978e-06, + "loss": 0.0, + "step": 14900 + }, + { + "epoch": 74.01, + "learning_rate": 2.604615384615385e-06, + "loss": 0.0, + "step": 14925 + }, + { + "epoch": 74.01, + "learning_rate": 2.5917948717948723e-06, + "loss": 0.0, + "step": 14950 + }, + { + "epoch": 74.01, + "learning_rate": 2.5789743589743595e-06, + "loss": 0.0, + "step": 14975 + }, + { + "epoch": 74.01, + "learning_rate": 2.5661538461538463e-06, + "loss": 0.0, + "step": 15000 + }, + { + "epoch": 74.01, + "eval_loss": 0.3427940309047699, + "eval_runtime": 2275.0151, + "eval_samples_per_second": 2.877, + "eval_steps_per_second": 0.18, + "eval_wer": 6.0344232891832235, + "step": 15000 + }, + { + "epoch": 75.0, + "learning_rate": 2.5533333333333336e-06, + "loss": 0.0, + "step": 15025 + }, + { + "epoch": 75.0, + "learning_rate": 2.5405128205128204e-06, + "loss": 0.0, + "step": 15050 + }, + { + "epoch": 75.0, + "learning_rate": 2.5276923076923077e-06, + "loss": 0.0, + "step": 15075 + }, + { + "epoch": 75.0, + "learning_rate": 2.514871794871795e-06, + "loss": 0.0, + "step": 15100 + }, + { + "epoch": 75.01, + "learning_rate": 2.502051282051282e-06, + "loss": 0.0, + "step": 15125 + }, + { + "epoch": 75.01, + "learning_rate": 2.4892307692307694e-06, + "loss": 0.0, + "step": 15150 + }, + { + "epoch": 75.01, + "learning_rate": 2.4764102564102567e-06, + "loss": 0.0, + "step": 15175 + }, + { + "epoch": 75.01, + "learning_rate": 2.463589743589744e-06, + "loss": 0.0, + "step": 15200 + }, + { + "epoch": 76.0, + "learning_rate": 2.450769230769231e-06, + "loss": 0.0, + "step": 15225 + }, + { + "epoch": 76.0, + "learning_rate": 2.4379487179487184e-06, + "loss": 0.0, + "step": 15250 + }, + { + "epoch": 76.0, + "learning_rate": 2.4251282051282056e-06, + "loss": 0.0, + "step": 15275 + }, + { + "epoch": 76.0, + "learning_rate": 2.4123076923076925e-06, + "loss": 0.0, + "step": 15300 + }, + { + "epoch": 76.01, + "learning_rate": 2.3994871794871797e-06, + "loss": 0.0, + "step": 15325 + }, + { + "epoch": 76.01, + "learning_rate": 2.386666666666667e-06, + "loss": 0.0, + "step": 15350 + }, + { + "epoch": 76.01, + "learning_rate": 2.3738461538461542e-06, + "loss": 0.0, + "step": 15375 + }, + { + "epoch": 76.01, + "learning_rate": 2.3610256410256415e-06, + "loss": 0.0, + "step": 15400 + }, + { + "epoch": 77.0, + "learning_rate": 2.3482051282051287e-06, + "loss": 0.0, + "step": 15425 + }, + { + "epoch": 77.0, + "learning_rate": 2.3353846153846155e-06, + "loss": 0.0, + "step": 15450 + }, + { + "epoch": 77.0, + "learning_rate": 2.3225641025641028e-06, + "loss": 0.0, + "step": 15475 + }, + { + "epoch": 77.0, + "learning_rate": 2.30974358974359e-06, + "loss": 0.0, + "step": 15500 + }, + { + "epoch": 77.01, + "learning_rate": 2.2969230769230773e-06, + "loss": 0.0, + "step": 15525 + }, + { + "epoch": 77.01, + "learning_rate": 2.2841025641025645e-06, + "loss": 0.0, + "step": 15550 + }, + { + "epoch": 77.01, + "learning_rate": 2.2712820512820518e-06, + "loss": 0.0, + "step": 15575 + }, + { + "epoch": 77.01, + "learning_rate": 2.2584615384615386e-06, + "loss": 0.0, + "step": 15600 + }, + { + "epoch": 78.0, + "learning_rate": 2.245641025641026e-06, + "loss": 0.0, + "step": 15625 + }, + { + "epoch": 78.0, + "learning_rate": 2.232820512820513e-06, + "loss": 0.0, + "step": 15650 + }, + { + "epoch": 78.0, + "learning_rate": 2.2200000000000003e-06, + "loss": 0.0, + "step": 15675 + }, + { + "epoch": 78.0, + "learning_rate": 2.2071794871794876e-06, + "loss": 0.0, + "step": 15700 + }, + { + "epoch": 78.01, + "learning_rate": 2.194358974358975e-06, + "loss": 0.0, + "step": 15725 + }, + { + "epoch": 78.01, + "learning_rate": 2.1815384615384617e-06, + "loss": 0.0, + "step": 15750 + }, + { + "epoch": 78.01, + "learning_rate": 2.168717948717949e-06, + "loss": 0.0, + "step": 15775 + }, + { + "epoch": 78.01, + "learning_rate": 2.155897435897436e-06, + "loss": 0.0, + "step": 15800 + }, + { + "epoch": 79.0, + "learning_rate": 2.1430769230769234e-06, + "loss": 0.0, + "step": 15825 + }, + { + "epoch": 79.0, + "learning_rate": 2.1302564102564107e-06, + "loss": 0.0, + "step": 15850 + }, + { + "epoch": 79.0, + "learning_rate": 2.117435897435898e-06, + "loss": 0.0, + "step": 15875 + }, + { + "epoch": 79.0, + "learning_rate": 2.1046153846153847e-06, + "loss": 0.0, + "step": 15900 + }, + { + "epoch": 79.01, + "learning_rate": 2.091794871794872e-06, + "loss": 0.0, + "step": 15925 + }, + { + "epoch": 79.01, + "learning_rate": 2.0789743589743592e-06, + "loss": 0.0, + "step": 15950 + }, + { + "epoch": 79.01, + "learning_rate": 2.0661538461538465e-06, + "loss": 0.0, + "step": 15975 + }, + { + "epoch": 79.01, + "learning_rate": 2.0533333333333337e-06, + "loss": 0.0, + "step": 16000 + }, + { + "epoch": 79.01, + "eval_loss": 0.3518044054508209, + "eval_runtime": 2226.9704, + "eval_samples_per_second": 2.939, + "eval_steps_per_second": 0.184, + "eval_wer": 6.001655629139073, + "step": 16000 + }, + { + "epoch": 80.0, + "learning_rate": 2.040512820512821e-06, + "loss": 0.0, + "step": 16025 + }, + { + "epoch": 80.0, + "learning_rate": 2.027692307692308e-06, + "loss": 0.0, + "step": 16050 + }, + { + "epoch": 80.0, + "learning_rate": 2.014871794871795e-06, + "loss": 0.0, + "step": 16075 + }, + { + "epoch": 80.0, + "learning_rate": 2.0020512820512823e-06, + "loss": 0.0, + "step": 16100 + }, + { + "epoch": 80.01, + "learning_rate": 1.9892307692307695e-06, + "loss": 0.0, + "step": 16125 + }, + { + "epoch": 80.01, + "learning_rate": 1.9764102564102568e-06, + "loss": 0.0, + "step": 16150 + }, + { + "epoch": 80.01, + "learning_rate": 1.963589743589744e-06, + "loss": 0.0, + "step": 16175 + }, + { + "epoch": 80.01, + "learning_rate": 1.950769230769231e-06, + "loss": 0.0, + "step": 16200 + }, + { + "epoch": 81.0, + "learning_rate": 1.937948717948718e-06, + "loss": 0.0, + "step": 16225 + }, + { + "epoch": 81.0, + "learning_rate": 1.9251282051282054e-06, + "loss": 0.0, + "step": 16250 + }, + { + "epoch": 81.0, + "learning_rate": 1.9123076923076926e-06, + "loss": 0.0, + "step": 16275 + }, + { + "epoch": 81.0, + "learning_rate": 1.8994871794871796e-06, + "loss": 0.0, + "step": 16300 + }, + { + "epoch": 81.01, + "learning_rate": 1.8866666666666669e-06, + "loss": 0.0, + "step": 16325 + }, + { + "epoch": 81.01, + "learning_rate": 1.873846153846154e-06, + "loss": 0.0, + "step": 16350 + }, + { + "epoch": 81.01, + "learning_rate": 1.8610256410256412e-06, + "loss": 0.0, + "step": 16375 + }, + { + "epoch": 81.01, + "learning_rate": 1.8482051282051284e-06, + "loss": 0.0, + "step": 16400 + }, + { + "epoch": 82.0, + "learning_rate": 1.8353846153846155e-06, + "loss": 0.0, + "step": 16425 + }, + { + "epoch": 82.0, + "learning_rate": 1.8225641025641027e-06, + "loss": 0.0, + "step": 16450 + }, + { + "epoch": 82.0, + "learning_rate": 1.80974358974359e-06, + "loss": 0.0, + "step": 16475 + }, + { + "epoch": 82.0, + "learning_rate": 1.796923076923077e-06, + "loss": 0.0, + "step": 16500 + }, + { + "epoch": 82.01, + "learning_rate": 1.7841025641025642e-06, + "loss": 0.0, + "step": 16525 + }, + { + "epoch": 82.01, + "learning_rate": 1.7712820512820515e-06, + "loss": 0.0, + "step": 16550 + }, + { + "epoch": 82.01, + "learning_rate": 1.7584615384615385e-06, + "loss": 0.0, + "step": 16575 + }, + { + "epoch": 82.01, + "learning_rate": 1.7456410256410258e-06, + "loss": 0.0, + "step": 16600 + }, + { + "epoch": 83.0, + "learning_rate": 1.732820512820513e-06, + "loss": 0.0, + "step": 16625 + }, + { + "epoch": 83.0, + "learning_rate": 1.72e-06, + "loss": 0.0, + "step": 16650 + }, + { + "epoch": 83.0, + "learning_rate": 1.7071794871794873e-06, + "loss": 0.0, + "step": 16675 + }, + { + "epoch": 83.0, + "learning_rate": 1.6943589743589745e-06, + "loss": 0.0, + "step": 16700 + }, + { + "epoch": 83.01, + "learning_rate": 1.6815384615384616e-06, + "loss": 0.0, + "step": 16725 + }, + { + "epoch": 83.01, + "learning_rate": 1.6687179487179488e-06, + "loss": 0.0, + "step": 16750 + }, + { + "epoch": 83.01, + "learning_rate": 1.655897435897436e-06, + "loss": 0.0, + "step": 16775 + }, + { + "epoch": 83.01, + "learning_rate": 1.6430769230769233e-06, + "loss": 0.0, + "step": 16800 + }, + { + "epoch": 84.0, + "learning_rate": 1.6302564102564104e-06, + "loss": 0.0, + "step": 16825 + }, + { + "epoch": 84.0, + "learning_rate": 1.6174358974358976e-06, + "loss": 0.0, + "step": 16850 + }, + { + "epoch": 84.0, + "learning_rate": 1.6046153846153846e-06, + "loss": 0.0, + "step": 16875 + }, + { + "epoch": 84.0, + "learning_rate": 1.5917948717948719e-06, + "loss": 0.0, + "step": 16900 + }, + { + "epoch": 84.01, + "learning_rate": 1.5789743589743591e-06, + "loss": 0.0, + "step": 16925 + }, + { + "epoch": 84.01, + "learning_rate": 1.5661538461538464e-06, + "loss": 0.0, + "step": 16950 + }, + { + "epoch": 84.01, + "learning_rate": 1.5533333333333334e-06, + "loss": 0.0, + "step": 16975 + }, + { + "epoch": 84.01, + "learning_rate": 1.5405128205128207e-06, + "loss": 0.0, + "step": 17000 + }, + { + "epoch": 84.01, + "eval_loss": 0.3600834906101227, + "eval_runtime": 2243.6524, + "eval_samples_per_second": 2.918, + "eval_steps_per_second": 0.183, + "eval_wer": 5.987858719646799, + "step": 17000 + }, + { + "epoch": 85.0, + "learning_rate": 1.5276923076923077e-06, + "loss": 0.0, + "step": 17025 + }, + { + "epoch": 85.0, + "learning_rate": 1.514871794871795e-06, + "loss": 0.0, + "step": 17050 + }, + { + "epoch": 85.0, + "learning_rate": 1.5020512820512822e-06, + "loss": 0.0, + "step": 17075 + }, + { + "epoch": 85.0, + "learning_rate": 1.4892307692307695e-06, + "loss": 0.0, + "step": 17100 + }, + { + "epoch": 85.01, + "learning_rate": 1.4764102564102565e-06, + "loss": 0.0, + "step": 17125 + }, + { + "epoch": 85.01, + "learning_rate": 1.4635897435897437e-06, + "loss": 0.0, + "step": 17150 + }, + { + "epoch": 85.01, + "learning_rate": 1.4507692307692308e-06, + "loss": 0.0, + "step": 17175 + }, + { + "epoch": 85.01, + "learning_rate": 1.437948717948718e-06, + "loss": 0.0, + "step": 17200 + }, + { + "epoch": 86.0, + "learning_rate": 1.4251282051282053e-06, + "loss": 0.0, + "step": 17225 + }, + { + "epoch": 86.0, + "learning_rate": 1.4123076923076925e-06, + "loss": 0.0, + "step": 17250 + }, + { + "epoch": 86.0, + "learning_rate": 1.3994871794871796e-06, + "loss": 0.0, + "step": 17275 + }, + { + "epoch": 86.0, + "learning_rate": 1.3866666666666668e-06, + "loss": 0.0, + "step": 17300 + }, + { + "epoch": 86.01, + "learning_rate": 1.3738461538461538e-06, + "loss": 0.0, + "step": 17325 + }, + { + "epoch": 86.01, + "learning_rate": 1.361025641025641e-06, + "loss": 0.0, + "step": 17350 + }, + { + "epoch": 86.01, + "learning_rate": 1.3482051282051283e-06, + "loss": 0.0, + "step": 17375 + }, + { + "epoch": 86.01, + "learning_rate": 1.3353846153846156e-06, + "loss": 0.0, + "step": 17400 + }, + { + "epoch": 87.0, + "learning_rate": 1.3225641025641026e-06, + "loss": 0.0, + "step": 17425 + }, + { + "epoch": 87.0, + "learning_rate": 1.3097435897435899e-06, + "loss": 0.0, + "step": 17450 + }, + { + "epoch": 87.0, + "learning_rate": 1.296923076923077e-06, + "loss": 0.0, + "step": 17475 + }, + { + "epoch": 87.0, + "learning_rate": 1.2841025641025641e-06, + "loss": 0.0, + "step": 17500 + }, + { + "epoch": 87.01, + "learning_rate": 1.2712820512820514e-06, + "loss": 0.0, + "step": 17525 + }, + { + "epoch": 87.01, + "learning_rate": 1.2584615384615386e-06, + "loss": 0.0, + "step": 17550 + }, + { + "epoch": 87.01, + "learning_rate": 1.2456410256410257e-06, + "loss": 0.0, + "step": 17575 + }, + { + "epoch": 87.01, + "learning_rate": 1.232820512820513e-06, + "loss": 0.0, + "step": 17600 + }, + { + "epoch": 88.0, + "learning_rate": 1.2200000000000002e-06, + "loss": 0.0, + "step": 17625 + }, + { + "epoch": 88.0, + "learning_rate": 1.2071794871794872e-06, + "loss": 0.0, + "step": 17650 + }, + { + "epoch": 88.0, + "learning_rate": 1.1943589743589745e-06, + "loss": 0.0, + "step": 17675 + }, + { + "epoch": 88.0, + "learning_rate": 1.1815384615384617e-06, + "loss": 0.0, + "step": 17700 + }, + { + "epoch": 88.01, + "learning_rate": 1.1687179487179487e-06, + "loss": 0.0, + "step": 17725 + }, + { + "epoch": 88.01, + "learning_rate": 1.155897435897436e-06, + "loss": 0.0, + "step": 17750 + }, + { + "epoch": 88.01, + "learning_rate": 1.1430769230769232e-06, + "loss": 0.0, + "step": 17775 + }, + { + "epoch": 88.01, + "learning_rate": 1.1302564102564103e-06, + "loss": 0.0, + "step": 17800 + }, + { + "epoch": 89.0, + "learning_rate": 1.1174358974358975e-06, + "loss": 0.0, + "step": 17825 + }, + { + "epoch": 89.0, + "learning_rate": 1.1046153846153848e-06, + "loss": 0.0, + "step": 17850 + }, + { + "epoch": 89.0, + "learning_rate": 1.0917948717948718e-06, + "loss": 0.0, + "step": 17875 + }, + { + "epoch": 89.0, + "learning_rate": 1.078974358974359e-06, + "loss": 0.0, + "step": 17900 + }, + { + "epoch": 89.01, + "learning_rate": 1.0661538461538463e-06, + "loss": 0.0, + "step": 17925 + }, + { + "epoch": 89.01, + "learning_rate": 1.0533333333333333e-06, + "loss": 0.0, + "step": 17950 + }, + { + "epoch": 89.01, + "learning_rate": 1.0405128205128206e-06, + "loss": 0.0, + "step": 17975 + }, + { + "epoch": 89.01, + "learning_rate": 1.0276923076923078e-06, + "loss": 0.0, + "step": 18000 + }, + { + "epoch": 89.01, + "eval_loss": 0.36747556924819946, + "eval_runtime": 2228.9623, + "eval_samples_per_second": 2.937, + "eval_steps_per_second": 0.184, + "eval_wer": 6.010278697571744, + "step": 18000 + }, + { + "epoch": 90.0, + "learning_rate": 1.0148717948717949e-06, + "loss": 0.0, + "step": 18025 + }, + { + "epoch": 90.0, + "learning_rate": 1.0020512820512821e-06, + "loss": 0.0, + "step": 18050 + }, + { + "epoch": 90.0, + "learning_rate": 9.892307692307694e-07, + "loss": 0.0, + "step": 18075 + }, + { + "epoch": 90.0, + "learning_rate": 9.764102564102564e-07, + "loss": 0.0, + "step": 18100 + }, + { + "epoch": 90.01, + "learning_rate": 9.635897435897437e-07, + "loss": 0.0, + "step": 18125 + }, + { + "epoch": 90.01, + "learning_rate": 9.507692307692308e-07, + "loss": 0.0, + "step": 18150 + }, + { + "epoch": 90.01, + "learning_rate": 9.37948717948718e-07, + "loss": 0.0, + "step": 18175 + }, + { + "epoch": 90.01, + "learning_rate": 9.251282051282053e-07, + "loss": 0.0, + "step": 18200 + }, + { + "epoch": 91.0, + "learning_rate": 9.123076923076923e-07, + "loss": 0.0, + "step": 18225 + }, + { + "epoch": 91.0, + "learning_rate": 8.994871794871796e-07, + "loss": 0.0, + "step": 18250 + }, + { + "epoch": 91.0, + "learning_rate": 8.866666666666668e-07, + "loss": 0.0, + "step": 18275 + }, + { + "epoch": 91.0, + "learning_rate": 8.738461538461539e-07, + "loss": 0.0, + "step": 18300 + }, + { + "epoch": 91.01, + "learning_rate": 8.610256410256411e-07, + "loss": 0.0, + "step": 18325 + }, + { + "epoch": 91.01, + "learning_rate": 8.482051282051284e-07, + "loss": 0.0, + "step": 18350 + }, + { + "epoch": 91.01, + "learning_rate": 8.353846153846154e-07, + "loss": 0.0, + "step": 18375 + }, + { + "epoch": 91.01, + "learning_rate": 8.225641025641026e-07, + "loss": 0.0, + "step": 18400 + }, + { + "epoch": 92.0, + "learning_rate": 8.097435897435899e-07, + "loss": 0.0, + "step": 18425 + }, + { + "epoch": 92.0, + "learning_rate": 7.969230769230769e-07, + "loss": 0.0, + "step": 18450 + }, + { + "epoch": 92.0, + "learning_rate": 7.841025641025642e-07, + "loss": 0.0, + "step": 18475 + }, + { + "epoch": 92.0, + "learning_rate": 7.712820512820514e-07, + "loss": 0.0, + "step": 18500 + }, + { + "epoch": 92.01, + "learning_rate": 7.584615384615385e-07, + "loss": 0.0, + "step": 18525 + }, + { + "epoch": 92.01, + "learning_rate": 7.456410256410257e-07, + "loss": 0.0, + "step": 18550 + }, + { + "epoch": 92.01, + "learning_rate": 7.32820512820513e-07, + "loss": 0.0, + "step": 18575 + }, + { + "epoch": 92.01, + "learning_rate": 7.2e-07, + "loss": 0.0, + "step": 18600 + }, + { + "epoch": 93.0, + "learning_rate": 7.071794871794872e-07, + "loss": 0.0, + "step": 18625 + }, + { + "epoch": 93.0, + "learning_rate": 6.943589743589745e-07, + "loss": 0.0, + "step": 18650 + }, + { + "epoch": 93.0, + "learning_rate": 6.815384615384615e-07, + "loss": 0.0, + "step": 18675 + }, + { + "epoch": 93.0, + "learning_rate": 6.687179487179488e-07, + "loss": 0.0, + "step": 18700 + }, + { + "epoch": 93.01, + "learning_rate": 6.55897435897436e-07, + "loss": 0.0, + "step": 18725 + }, + { + "epoch": 93.01, + "learning_rate": 6.43076923076923e-07, + "loss": 0.0, + "step": 18750 + }, + { + "epoch": 93.01, + "learning_rate": 6.302564102564103e-07, + "loss": 0.0, + "step": 18775 + }, + { + "epoch": 93.01, + "learning_rate": 6.174358974358975e-07, + "loss": 0.0, + "step": 18800 + }, + { + "epoch": 94.0, + "learning_rate": 6.046153846153847e-07, + "loss": 0.0, + "step": 18825 + }, + { + "epoch": 94.0, + "learning_rate": 5.917948717948718e-07, + "loss": 0.0, + "step": 18850 + }, + { + "epoch": 94.0, + "learning_rate": 5.789743589743591e-07, + "loss": 0.0, + "step": 18875 + }, + { + "epoch": 94.0, + "learning_rate": 5.661538461538462e-07, + "loss": 0.0, + "step": 18900 + }, + { + "epoch": 94.01, + "learning_rate": 5.533333333333334e-07, + "loss": 0.0, + "step": 18925 + }, + { + "epoch": 94.01, + "learning_rate": 5.405128205128206e-07, + "loss": 0.0, + "step": 18950 + }, + { + "epoch": 94.01, + "learning_rate": 5.276923076923078e-07, + "loss": 0.0, + "step": 18975 + }, + { + "epoch": 94.01, + "learning_rate": 5.148717948717949e-07, + "loss": 0.0, + "step": 19000 + }, + { + "epoch": 94.01, + "eval_loss": 0.3728844225406647, + "eval_runtime": 2227.807, + "eval_samples_per_second": 2.938, + "eval_steps_per_second": 0.184, + "eval_wer": 6.006829470198675, + "step": 19000 + }, + { + "epoch": 95.0, + "learning_rate": 5.020512820512821e-07, + "loss": 0.0, + "step": 19025 + }, + { + "epoch": 95.0, + "learning_rate": 4.892307692307693e-07, + "loss": 0.0, + "step": 19050 + }, + { + "epoch": 95.0, + "learning_rate": 4.7641025641025643e-07, + "loss": 0.0, + "step": 19075 + }, + { + "epoch": 95.0, + "learning_rate": 4.635897435897437e-07, + "loss": 0.0, + "step": 19100 + }, + { + "epoch": 95.01, + "learning_rate": 4.507692307692308e-07, + "loss": 0.0, + "step": 19125 + }, + { + "epoch": 95.01, + "learning_rate": 4.3794871794871796e-07, + "loss": 0.0, + "step": 19150 + }, + { + "epoch": 95.01, + "learning_rate": 4.251282051282052e-07, + "loss": 0.0, + "step": 19175 + }, + { + "epoch": 95.01, + "learning_rate": 4.1230769230769235e-07, + "loss": 0.0, + "step": 19200 + }, + { + "epoch": 96.0, + "learning_rate": 3.994871794871795e-07, + "loss": 0.0, + "step": 19225 + }, + { + "epoch": 96.0, + "learning_rate": 3.8666666666666674e-07, + "loss": 0.0, + "step": 19250 + }, + { + "epoch": 96.0, + "learning_rate": 3.738461538461539e-07, + "loss": 0.0, + "step": 19275 + }, + { + "epoch": 96.0, + "learning_rate": 3.61025641025641e-07, + "loss": 0.0, + "step": 19300 + }, + { + "epoch": 96.01, + "learning_rate": 3.4820512820512827e-07, + "loss": 0.0, + "step": 19325 + }, + { + "epoch": 96.01, + "learning_rate": 3.353846153846154e-07, + "loss": 0.0, + "step": 19350 + }, + { + "epoch": 96.01, + "learning_rate": 3.2256410256410256e-07, + "loss": 0.0, + "step": 19375 + }, + { + "epoch": 96.01, + "learning_rate": 3.097435897435898e-07, + "loss": 0.0, + "step": 19400 + }, + { + "epoch": 97.0, + "learning_rate": 2.9692307692307695e-07, + "loss": 0.0, + "step": 19425 + }, + { + "epoch": 97.0, + "learning_rate": 2.8410256410256414e-07, + "loss": 0.0, + "step": 19450 + }, + { + "epoch": 97.0, + "learning_rate": 2.7128205128205134e-07, + "loss": 0.0, + "step": 19475 + }, + { + "epoch": 97.0, + "learning_rate": 2.584615384615385e-07, + "loss": 0.0, + "step": 19500 + }, + { + "epoch": 97.01, + "learning_rate": 2.456410256410257e-07, + "loss": 0.0, + "step": 19525 + }, + { + "epoch": 97.01, + "learning_rate": 2.3282051282051284e-07, + "loss": 0.0, + "step": 19550 + }, + { + "epoch": 97.01, + "learning_rate": 2.2e-07, + "loss": 0.0, + "step": 19575 + }, + { + "epoch": 97.01, + "learning_rate": 2.071794871794872e-07, + "loss": 0.0, + "step": 19600 + }, + { + "epoch": 98.0, + "learning_rate": 1.9435897435897437e-07, + "loss": 0.0, + "step": 19625 + }, + { + "epoch": 98.0, + "learning_rate": 1.8153846153846154e-07, + "loss": 0.0, + "step": 19650 + }, + { + "epoch": 98.0, + "learning_rate": 1.6871794871794874e-07, + "loss": 0.0, + "step": 19675 + }, + { + "epoch": 98.0, + "learning_rate": 1.558974358974359e-07, + "loss": 0.0, + "step": 19700 + }, + { + "epoch": 98.01, + "learning_rate": 1.430769230769231e-07, + "loss": 0.0, + "step": 19725 + }, + { + "epoch": 98.01, + "learning_rate": 1.3025641025641027e-07, + "loss": 0.0, + "step": 19750 + }, + { + "epoch": 98.01, + "learning_rate": 1.1743589743589745e-07, + "loss": 0.0, + "step": 19775 + }, + { + "epoch": 98.01, + "learning_rate": 1.0461538461538462e-07, + "loss": 0.0, + "step": 19800 + }, + { + "epoch": 99.0, + "learning_rate": 9.179487179487179e-08, + "loss": 0.0, + "step": 19825 + }, + { + "epoch": 99.0, + "learning_rate": 7.897435897435898e-08, + "loss": 0.0, + "step": 19850 + }, + { + "epoch": 99.0, + "learning_rate": 6.615384615384615e-08, + "loss": 0.0, + "step": 19875 + }, + { + "epoch": 99.0, + "learning_rate": 5.3333333333333334e-08, + "loss": 0.0, + "step": 19900 + }, + { + "epoch": 99.01, + "learning_rate": 4.0512820512820516e-08, + "loss": 0.0, + "step": 19925 + }, + { + "epoch": 99.01, + "learning_rate": 2.769230769230769e-08, + "loss": 0.0, + "step": 19950 + }, + { + "epoch": 99.01, + "learning_rate": 1.4871794871794873e-08, + "loss": 0.0, + "step": 19975 + }, + { + "epoch": 99.01, + "learning_rate": 2.051282051282051e-09, + "loss": 0.0, + "step": 20000 + }, + { + "epoch": 99.01, + "eval_loss": 0.375347375869751, + "eval_runtime": 2231.3267, + "eval_samples_per_second": 2.934, + "eval_steps_per_second": 0.184, + "eval_wer": 6.01717715231788, + "step": 20000 + }, + { + "epoch": 99.01, + "step": 20000, + "total_flos": 2.716385004011523e+21, + "train_loss": 0.008869889611534018, + "train_runtime": 372775.1926, + "train_samples_per_second": 3.434, + "train_steps_per_second": 0.054 + } + ], + "logging_steps": 25, + "max_steps": 20000, + "num_train_epochs": 9223372036854775807, + "save_steps": 1000, + "total_flos": 2.716385004011523e+21, + "trial_name": null, + "trial_params": null +}