{ "best_metric": 24.655888136333843, "best_model_checkpoint": "whisper-medium-mn-5/checkpoint-11000", "epoch": 6.695069993913573, "global_step": 11000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.02, "learning_rate": 4.2000000000000006e-07, "loss": 4.3886, "step": 25 }, { "epoch": 0.03, "learning_rate": 9.200000000000001e-07, "loss": 3.5928, "step": 50 }, { "epoch": 0.05, "learning_rate": 1.42e-06, "loss": 2.7058, "step": 75 }, { "epoch": 0.06, "learning_rate": 1.9200000000000003e-06, "loss": 2.0017, "step": 100 }, { "epoch": 0.08, "learning_rate": 2.42e-06, "loss": 1.5276, "step": 125 }, { "epoch": 0.09, "learning_rate": 2.92e-06, "loss": 1.2181, "step": 150 }, { "epoch": 0.11, "learning_rate": 3.4200000000000007e-06, "loss": 1.0112, "step": 175 }, { "epoch": 0.12, "learning_rate": 3.920000000000001e-06, "loss": 0.7822, "step": 200 }, { "epoch": 0.14, "learning_rate": 4.42e-06, "loss": 0.718, "step": 225 }, { "epoch": 0.15, "learning_rate": 4.92e-06, "loss": 0.6322, "step": 250 }, { "epoch": 0.17, "learning_rate": 5.420000000000001e-06, "loss": 0.5869, "step": 275 }, { "epoch": 0.18, "learning_rate": 5.92e-06, "loss": 0.5581, "step": 300 }, { "epoch": 0.2, "learning_rate": 6.42e-06, "loss": 0.5241, "step": 325 }, { "epoch": 0.21, "learning_rate": 6.92e-06, "loss": 0.4871, "step": 350 }, { "epoch": 0.23, "learning_rate": 7.420000000000001e-06, "loss": 0.4674, "step": 375 }, { "epoch": 0.24, "learning_rate": 7.92e-06, "loss": 0.4481, "step": 400 }, { "epoch": 0.26, "learning_rate": 8.42e-06, "loss": 0.4244, "step": 425 }, { "epoch": 0.27, "learning_rate": 8.920000000000001e-06, "loss": 0.4134, "step": 450 }, { "epoch": 0.29, "learning_rate": 9.42e-06, "loss": 0.3976, "step": 475 }, { "epoch": 0.3, "learning_rate": 9.920000000000002e-06, "loss": 0.3942, "step": 500 }, { "epoch": 0.32, "learning_rate": 9.98551724137931e-06, "loss": 0.3958, "step": 525 }, { "epoch": 0.33, "learning_rate": 9.968275862068967e-06, "loss": 0.3737, "step": 550 }, { "epoch": 0.35, "learning_rate": 9.951034482758622e-06, "loss": 0.3571, "step": 575 }, { "epoch": 0.37, "learning_rate": 9.933793103448277e-06, "loss": 0.3455, "step": 600 }, { "epoch": 0.38, "learning_rate": 9.916551724137932e-06, "loss": 0.329, "step": 625 }, { "epoch": 0.4, "learning_rate": 9.899310344827588e-06, "loss": 0.3312, "step": 650 }, { "epoch": 0.41, "learning_rate": 9.882068965517241e-06, "loss": 0.3447, "step": 675 }, { "epoch": 0.43, "learning_rate": 9.864827586206898e-06, "loss": 0.3286, "step": 700 }, { "epoch": 0.44, "learning_rate": 9.847586206896553e-06, "loss": 0.3113, "step": 725 }, { "epoch": 0.46, "learning_rate": 9.830344827586208e-06, "loss": 0.3142, "step": 750 }, { "epoch": 0.47, "learning_rate": 9.813103448275862e-06, "loss": 0.3118, "step": 775 }, { "epoch": 0.49, "learning_rate": 9.795862068965517e-06, "loss": 0.3092, "step": 800 }, { "epoch": 0.5, "learning_rate": 9.778620689655172e-06, "loss": 0.2868, "step": 825 }, { "epoch": 0.52, "learning_rate": 9.761379310344829e-06, "loss": 0.2943, "step": 850 }, { "epoch": 0.53, "learning_rate": 9.744137931034484e-06, "loss": 0.2962, "step": 875 }, { "epoch": 0.55, "learning_rate": 9.726896551724139e-06, "loss": 0.2875, "step": 900 }, { "epoch": 0.56, "learning_rate": 9.709655172413795e-06, "loss": 0.2834, "step": 925 }, { "epoch": 0.58, "learning_rate": 9.692413793103448e-06, "loss": 0.2824, "step": 950 }, { "epoch": 0.59, "learning_rate": 9.675172413793105e-06, "loss": 0.2725, "step": 975 }, { "epoch": 0.61, "learning_rate": 9.65793103448276e-06, "loss": 0.2657, "step": 1000 }, { "epoch": 0.61, "eval_cer": 13.107321345767428, "eval_loss": 0.31748202443122864, "eval_runtime": 2311.6568, "eval_samples_per_second": 0.816, "eval_steps_per_second": 0.026, "eval_wer": 37.393489185055714, "step": 1000 }, { "epoch": 0.62, "learning_rate": 9.640689655172415e-06, "loss": 0.2719, "step": 1025 }, { "epoch": 0.64, "learning_rate": 9.62344827586207e-06, "loss": 0.2673, "step": 1050 }, { "epoch": 0.65, "learning_rate": 9.606206896551726e-06, "loss": 0.2549, "step": 1075 }, { "epoch": 0.67, "learning_rate": 9.58896551724138e-06, "loss": 0.2644, "step": 1100 }, { "epoch": 0.68, "learning_rate": 9.571724137931036e-06, "loss": 0.2652, "step": 1125 }, { "epoch": 0.7, "learning_rate": 9.55448275862069e-06, "loss": 0.2544, "step": 1150 }, { "epoch": 0.72, "learning_rate": 9.537241379310345e-06, "loss": 0.2696, "step": 1175 }, { "epoch": 0.73, "learning_rate": 9.52e-06, "loss": 0.2627, "step": 1200 }, { "epoch": 0.75, "learning_rate": 9.502758620689655e-06, "loss": 0.2429, "step": 1225 }, { "epoch": 0.76, "learning_rate": 9.48551724137931e-06, "loss": 0.2573, "step": 1250 }, { "epoch": 0.78, "learning_rate": 9.468275862068967e-06, "loss": 0.2547, "step": 1275 }, { "epoch": 0.79, "learning_rate": 9.451034482758622e-06, "loss": 0.2445, "step": 1300 }, { "epoch": 0.81, "learning_rate": 9.433793103448276e-06, "loss": 0.2407, "step": 1325 }, { "epoch": 0.82, "learning_rate": 9.416551724137933e-06, "loss": 0.2439, "step": 1350 }, { "epoch": 0.84, "learning_rate": 9.399310344827586e-06, "loss": 0.2428, "step": 1375 }, { "epoch": 0.85, "learning_rate": 9.382068965517243e-06, "loss": 0.2426, "step": 1400 }, { "epoch": 0.87, "learning_rate": 9.364827586206898e-06, "loss": 0.2404, "step": 1425 }, { "epoch": 0.88, "learning_rate": 9.347586206896552e-06, "loss": 0.2441, "step": 1450 }, { "epoch": 0.9, "learning_rate": 9.330344827586207e-06, "loss": 0.2293, "step": 1475 }, { "epoch": 0.91, "learning_rate": 9.313103448275864e-06, "loss": 0.2318, "step": 1500 }, { "epoch": 0.93, "learning_rate": 9.295862068965517e-06, "loss": 0.2227, "step": 1525 }, { "epoch": 0.94, "learning_rate": 9.278620689655174e-06, "loss": 0.2338, "step": 1550 }, { "epoch": 0.96, "learning_rate": 9.261379310344828e-06, "loss": 0.2326, "step": 1575 }, { "epoch": 0.97, "learning_rate": 9.244137931034483e-06, "loss": 0.2268, "step": 1600 }, { "epoch": 0.99, "learning_rate": 9.226896551724138e-06, "loss": 0.2335, "step": 1625 }, { "epoch": 1.0, "learning_rate": 9.209655172413793e-06, "loss": 0.2184, "step": 1650 }, { "epoch": 1.02, "learning_rate": 9.192413793103448e-06, "loss": 0.1802, "step": 1675 }, { "epoch": 1.03, "learning_rate": 9.175172413793105e-06, "loss": 0.1804, "step": 1700 }, { "epoch": 1.05, "learning_rate": 9.15793103448276e-06, "loss": 0.1907, "step": 1725 }, { "epoch": 1.07, "learning_rate": 9.140689655172414e-06, "loss": 0.1762, "step": 1750 }, { "epoch": 1.08, "learning_rate": 9.12344827586207e-06, "loss": 0.1755, "step": 1775 }, { "epoch": 1.1, "learning_rate": 9.106206896551724e-06, "loss": 0.1831, "step": 1800 }, { "epoch": 1.11, "learning_rate": 9.08896551724138e-06, "loss": 0.1786, "step": 1825 }, { "epoch": 1.13, "learning_rate": 9.071724137931035e-06, "loss": 0.171, "step": 1850 }, { "epoch": 1.14, "learning_rate": 9.05448275862069e-06, "loss": 0.1767, "step": 1875 }, { "epoch": 1.16, "learning_rate": 9.037241379310345e-06, "loss": 0.1807, "step": 1900 }, { "epoch": 1.17, "learning_rate": 9.020000000000002e-06, "loss": 0.1766, "step": 1925 }, { "epoch": 1.19, "learning_rate": 9.002758620689655e-06, "loss": 0.1825, "step": 1950 }, { "epoch": 1.2, "learning_rate": 8.985517241379311e-06, "loss": 0.1792, "step": 1975 }, { "epoch": 1.22, "learning_rate": 8.968275862068966e-06, "loss": 0.1708, "step": 2000 }, { "epoch": 1.22, "eval_cer": 10.228931595308019, "eval_loss": 0.2588365077972412, "eval_runtime": 2189.388, "eval_samples_per_second": 0.861, "eval_steps_per_second": 0.027, "eval_wer": 29.730172602141142, "step": 2000 }, { "epoch": 1.23, "learning_rate": 8.951034482758621e-06, "loss": 0.1721, "step": 2025 }, { "epoch": 1.25, "learning_rate": 8.933793103448276e-06, "loss": 0.1777, "step": 2050 }, { "epoch": 1.26, "learning_rate": 8.916551724137931e-06, "loss": 0.1727, "step": 2075 }, { "epoch": 1.28, "learning_rate": 8.899310344827588e-06, "loss": 0.186, "step": 2100 }, { "epoch": 1.29, "learning_rate": 8.882068965517242e-06, "loss": 0.1815, "step": 2125 }, { "epoch": 1.31, "learning_rate": 8.864827586206897e-06, "loss": 0.1849, "step": 2150 }, { "epoch": 1.32, "learning_rate": 8.847586206896552e-06, "loss": 0.1737, "step": 2175 }, { "epoch": 1.34, "learning_rate": 8.830344827586209e-06, "loss": 0.1746, "step": 2200 }, { "epoch": 1.35, "learning_rate": 8.813103448275862e-06, "loss": 0.1741, "step": 2225 }, { "epoch": 1.37, "learning_rate": 8.795862068965518e-06, "loss": 0.1769, "step": 2250 }, { "epoch": 1.38, "learning_rate": 8.778620689655173e-06, "loss": 0.1702, "step": 2275 }, { "epoch": 1.4, "learning_rate": 8.761379310344828e-06, "loss": 0.1806, "step": 2300 }, { "epoch": 1.42, "learning_rate": 8.744137931034483e-06, "loss": 0.1803, "step": 2325 }, { "epoch": 1.43, "learning_rate": 8.72689655172414e-06, "loss": 0.1786, "step": 2350 }, { "epoch": 1.45, "learning_rate": 8.709655172413793e-06, "loss": 0.1764, "step": 2375 }, { "epoch": 1.46, "learning_rate": 8.69241379310345e-06, "loss": 0.1736, "step": 2400 }, { "epoch": 1.48, "learning_rate": 8.675172413793104e-06, "loss": 0.1695, "step": 2425 }, { "epoch": 1.49, "learning_rate": 8.657931034482759e-06, "loss": 0.1671, "step": 2450 }, { "epoch": 1.51, "learning_rate": 8.640689655172414e-06, "loss": 0.1738, "step": 2475 }, { "epoch": 1.52, "learning_rate": 8.623448275862069e-06, "loss": 0.1792, "step": 2500 }, { "epoch": 1.54, "learning_rate": 8.606206896551725e-06, "loss": 0.1753, "step": 2525 }, { "epoch": 1.55, "learning_rate": 8.58896551724138e-06, "loss": 0.1655, "step": 2550 }, { "epoch": 1.57, "learning_rate": 8.571724137931035e-06, "loss": 0.1759, "step": 2575 }, { "epoch": 1.58, "learning_rate": 8.55448275862069e-06, "loss": 0.1748, "step": 2600 }, { "epoch": 1.6, "learning_rate": 8.537241379310347e-06, "loss": 0.1788, "step": 2625 }, { "epoch": 1.61, "learning_rate": 8.52e-06, "loss": 0.1727, "step": 2650 }, { "epoch": 1.63, "learning_rate": 8.502758620689656e-06, "loss": 0.1745, "step": 2675 }, { "epoch": 1.64, "learning_rate": 8.485517241379311e-06, "loss": 0.166, "step": 2700 }, { "epoch": 1.66, "learning_rate": 8.468275862068966e-06, "loss": 0.1583, "step": 2725 }, { "epoch": 1.67, "learning_rate": 8.451034482758621e-06, "loss": 0.1703, "step": 2750 }, { "epoch": 1.69, "learning_rate": 8.433793103448277e-06, "loss": 0.1679, "step": 2775 }, { "epoch": 1.7, "learning_rate": 8.41655172413793e-06, "loss": 0.1683, "step": 2800 }, { "epoch": 1.72, "learning_rate": 8.399310344827587e-06, "loss": 0.1656, "step": 2825 }, { "epoch": 1.73, "learning_rate": 8.382068965517242e-06, "loss": 0.172, "step": 2850 }, { "epoch": 1.75, "learning_rate": 8.364827586206897e-06, "loss": 0.1724, "step": 2875 }, { "epoch": 1.77, "learning_rate": 8.347586206896552e-06, "loss": 0.1659, "step": 2900 }, { "epoch": 1.78, "learning_rate": 8.330344827586208e-06, "loss": 0.1696, "step": 2925 }, { "epoch": 1.8, "learning_rate": 8.313103448275863e-06, "loss": 0.1697, "step": 2950 }, { "epoch": 1.81, "learning_rate": 8.295862068965518e-06, "loss": 0.1655, "step": 2975 }, { "epoch": 1.83, "learning_rate": 8.278620689655173e-06, "loss": 0.167, "step": 3000 }, { "epoch": 1.83, "eval_cer": 9.5462586085493, "eval_loss": 0.22871485352516174, "eval_runtime": 2206.2854, "eval_samples_per_second": 0.855, "eval_steps_per_second": 0.027, "eval_wer": 27.386934673366838, "step": 3000 }, { "epoch": 1.84, "learning_rate": 8.261379310344828e-06, "loss": 0.1725, "step": 3025 }, { "epoch": 1.86, "learning_rate": 8.244137931034484e-06, "loss": 0.1704, "step": 3050 }, { "epoch": 1.87, "learning_rate": 8.226896551724138e-06, "loss": 0.1674, "step": 3075 }, { "epoch": 1.89, "learning_rate": 8.209655172413794e-06, "loss": 0.1677, "step": 3100 }, { "epoch": 1.9, "learning_rate": 8.192413793103449e-06, "loss": 0.1639, "step": 3125 }, { "epoch": 1.92, "learning_rate": 8.175172413793104e-06, "loss": 0.1711, "step": 3150 }, { "epoch": 1.93, "learning_rate": 8.157931034482759e-06, "loss": 0.1649, "step": 3175 }, { "epoch": 1.95, "learning_rate": 8.140689655172415e-06, "loss": 0.1706, "step": 3200 }, { "epoch": 1.96, "learning_rate": 8.123448275862069e-06, "loss": 0.1719, "step": 3225 }, { "epoch": 1.98, "learning_rate": 8.106206896551725e-06, "loss": 0.1668, "step": 3250 }, { "epoch": 1.99, "learning_rate": 8.08896551724138e-06, "loss": 0.1711, "step": 3275 }, { "epoch": 2.01, "learning_rate": 8.071724137931035e-06, "loss": 0.1291, "step": 3300 }, { "epoch": 2.02, "learning_rate": 8.054482758620691e-06, "loss": 0.1108, "step": 3325 }, { "epoch": 2.04, "learning_rate": 8.037241379310346e-06, "loss": 0.1101, "step": 3350 }, { "epoch": 2.05, "learning_rate": 8.020000000000001e-06, "loss": 0.1166, "step": 3375 }, { "epoch": 2.07, "learning_rate": 8.002758620689656e-06, "loss": 0.1076, "step": 3400 }, { "epoch": 2.08, "learning_rate": 7.985517241379311e-06, "loss": 0.1103, "step": 3425 }, { "epoch": 2.1, "learning_rate": 7.968275862068966e-06, "loss": 0.1174, "step": 3450 }, { "epoch": 2.12, "learning_rate": 7.951034482758622e-06, "loss": 0.111, "step": 3475 }, { "epoch": 2.13, "learning_rate": 7.933793103448275e-06, "loss": 0.1156, "step": 3500 }, { "epoch": 2.15, "learning_rate": 7.916551724137932e-06, "loss": 0.1206, "step": 3525 }, { "epoch": 2.16, "learning_rate": 7.899310344827587e-06, "loss": 0.1138, "step": 3550 }, { "epoch": 2.18, "learning_rate": 7.882068965517242e-06, "loss": 0.115, "step": 3575 }, { "epoch": 2.19, "learning_rate": 7.864827586206897e-06, "loss": 0.1121, "step": 3600 }, { "epoch": 2.21, "learning_rate": 7.847586206896553e-06, "loss": 0.1104, "step": 3625 }, { "epoch": 2.22, "learning_rate": 7.830344827586206e-06, "loss": 0.1135, "step": 3650 }, { "epoch": 2.24, "learning_rate": 7.813103448275863e-06, "loss": 0.1145, "step": 3675 }, { "epoch": 2.25, "learning_rate": 7.795862068965518e-06, "loss": 0.1155, "step": 3700 }, { "epoch": 2.27, "learning_rate": 7.778620689655173e-06, "loss": 0.1121, "step": 3725 }, { "epoch": 2.28, "learning_rate": 7.76137931034483e-06, "loss": 0.1139, "step": 3750 }, { "epoch": 2.3, "learning_rate": 7.744137931034484e-06, "loss": 0.1135, "step": 3775 }, { "epoch": 2.31, "learning_rate": 7.726896551724139e-06, "loss": 0.1176, "step": 3800 }, { "epoch": 2.33, "learning_rate": 7.709655172413794e-06, "loss": 0.1179, "step": 3825 }, { "epoch": 2.34, "learning_rate": 7.692413793103449e-06, "loss": 0.1105, "step": 3850 }, { "epoch": 2.36, "learning_rate": 7.675172413793104e-06, "loss": 0.1157, "step": 3875 }, { "epoch": 2.37, "learning_rate": 7.65793103448276e-06, "loss": 0.1179, "step": 3900 }, { "epoch": 2.39, "learning_rate": 7.640689655172413e-06, "loss": 0.1138, "step": 3925 }, { "epoch": 2.4, "learning_rate": 7.62344827586207e-06, "loss": 0.1124, "step": 3950 }, { "epoch": 2.42, "learning_rate": 7.606206896551725e-06, "loss": 0.1057, "step": 3975 }, { "epoch": 2.43, "learning_rate": 7.58896551724138e-06, "loss": 0.112, "step": 4000 }, { "epoch": 2.43, "eval_cer": 8.879042368660588, "eval_loss": 0.23367805778980255, "eval_runtime": 2188.4072, "eval_samples_per_second": 0.862, "eval_steps_per_second": 0.027, "eval_wer": 25.617216517369457, "step": 4000 }, { "epoch": 2.45, "learning_rate": 7.5717241379310345e-06, "loss": 0.1116, "step": 4025 }, { "epoch": 2.47, "learning_rate": 7.55448275862069e-06, "loss": 0.1098, "step": 4050 }, { "epoch": 2.48, "learning_rate": 7.537241379310345e-06, "loss": 0.113, "step": 4075 }, { "epoch": 2.5, "learning_rate": 7.520000000000001e-06, "loss": 0.1102, "step": 4100 }, { "epoch": 2.51, "learning_rate": 7.5027586206896566e-06, "loss": 0.1131, "step": 4125 }, { "epoch": 2.53, "learning_rate": 7.485517241379311e-06, "loss": 0.1129, "step": 4150 }, { "epoch": 2.54, "learning_rate": 7.468275862068966e-06, "loss": 0.1069, "step": 4175 }, { "epoch": 2.56, "learning_rate": 7.451034482758621e-06, "loss": 0.1117, "step": 4200 }, { "epoch": 2.57, "learning_rate": 7.433793103448277e-06, "loss": 0.1193, "step": 4225 }, { "epoch": 2.59, "learning_rate": 7.416551724137932e-06, "loss": 0.1148, "step": 4250 }, { "epoch": 2.6, "learning_rate": 7.3993103448275875e-06, "loss": 0.1069, "step": 4275 }, { "epoch": 2.62, "learning_rate": 7.3820689655172415e-06, "loss": 0.1165, "step": 4300 }, { "epoch": 2.63, "learning_rate": 7.364827586206897e-06, "loss": 0.1186, "step": 4325 }, { "epoch": 2.65, "learning_rate": 7.347586206896552e-06, "loss": 0.1116, "step": 4350 }, { "epoch": 2.66, "learning_rate": 7.330344827586208e-06, "loss": 0.1114, "step": 4375 }, { "epoch": 2.68, "learning_rate": 7.313103448275863e-06, "loss": 0.1111, "step": 4400 }, { "epoch": 2.69, "learning_rate": 7.2958620689655175e-06, "loss": 0.1125, "step": 4425 }, { "epoch": 2.71, "learning_rate": 7.278620689655172e-06, "loss": 0.1136, "step": 4450 }, { "epoch": 2.72, "learning_rate": 7.261379310344828e-06, "loss": 0.1066, "step": 4475 }, { "epoch": 2.74, "learning_rate": 7.244137931034483e-06, "loss": 0.1117, "step": 4500 }, { "epoch": 2.75, "learning_rate": 7.226896551724139e-06, "loss": 0.1101, "step": 4525 }, { "epoch": 2.77, "learning_rate": 7.2096551724137944e-06, "loss": 0.1136, "step": 4550 }, { "epoch": 2.78, "learning_rate": 7.1924137931034485e-06, "loss": 0.1124, "step": 4575 }, { "epoch": 2.8, "learning_rate": 7.175172413793104e-06, "loss": 0.1175, "step": 4600 }, { "epoch": 2.81, "learning_rate": 7.157931034482759e-06, "loss": 0.1124, "step": 4625 }, { "epoch": 2.83, "learning_rate": 7.140689655172415e-06, "loss": 0.1107, "step": 4650 }, { "epoch": 2.85, "learning_rate": 7.12344827586207e-06, "loss": 0.1143, "step": 4675 }, { "epoch": 2.86, "learning_rate": 7.106206896551725e-06, "loss": 0.1114, "step": 4700 }, { "epoch": 2.88, "learning_rate": 7.088965517241379e-06, "loss": 0.1108, "step": 4725 }, { "epoch": 2.89, "learning_rate": 7.071724137931035e-06, "loss": 0.1119, "step": 4750 }, { "epoch": 2.91, "learning_rate": 7.05448275862069e-06, "loss": 0.1142, "step": 4775 }, { "epoch": 2.92, "learning_rate": 7.037241379310346e-06, "loss": 0.1147, "step": 4800 }, { "epoch": 2.94, "learning_rate": 7.0200000000000006e-06, "loss": 0.1161, "step": 4825 }, { "epoch": 2.95, "learning_rate": 7.002758620689655e-06, "loss": 0.1083, "step": 4850 }, { "epoch": 2.97, "learning_rate": 6.98551724137931e-06, "loss": 0.1111, "step": 4875 }, { "epoch": 2.98, "learning_rate": 6.968275862068966e-06, "loss": 0.1159, "step": 4900 }, { "epoch": 3.0, "learning_rate": 6.951034482758622e-06, "loss": 0.1153, "step": 4925 }, { "epoch": 3.01, "learning_rate": 6.933793103448277e-06, "loss": 0.0801, "step": 4950 }, { "epoch": 3.03, "learning_rate": 6.916551724137932e-06, "loss": 0.0654, "step": 4975 }, { "epoch": 3.04, "learning_rate": 6.899310344827586e-06, "loss": 0.066, "step": 5000 }, { "epoch": 3.04, "eval_cer": 8.721898775482165, "eval_loss": 0.2409059852361679, "eval_runtime": 2216.746, "eval_samples_per_second": 0.851, "eval_steps_per_second": 0.027, "eval_wer": 24.96176534848154, "step": 5000 }, { "epoch": 3.06, "learning_rate": 6.882068965517242e-06, "loss": 0.0621, "step": 5025 }, { "epoch": 3.07, "learning_rate": 6.864827586206897e-06, "loss": 0.0684, "step": 5050 }, { "epoch": 3.09, "learning_rate": 6.847586206896553e-06, "loss": 0.0664, "step": 5075 }, { "epoch": 3.1, "learning_rate": 6.8303448275862075e-06, "loss": 0.0662, "step": 5100 }, { "epoch": 3.12, "learning_rate": 6.813103448275863e-06, "loss": 0.07, "step": 5125 }, { "epoch": 3.13, "learning_rate": 6.795862068965517e-06, "loss": 0.0652, "step": 5150 }, { "epoch": 3.15, "learning_rate": 6.778620689655173e-06, "loss": 0.0679, "step": 5175 }, { "epoch": 3.16, "learning_rate": 6.761379310344828e-06, "loss": 0.0693, "step": 5200 }, { "epoch": 3.18, "learning_rate": 6.7441379310344836e-06, "loss": 0.0632, "step": 5225 }, { "epoch": 3.2, "learning_rate": 6.7268965517241384e-06, "loss": 0.0683, "step": 5250 }, { "epoch": 3.21, "learning_rate": 6.709655172413793e-06, "loss": 0.0678, "step": 5275 }, { "epoch": 3.23, "learning_rate": 6.692413793103448e-06, "loss": 0.0683, "step": 5300 }, { "epoch": 3.24, "learning_rate": 6.675172413793104e-06, "loss": 0.0673, "step": 5325 }, { "epoch": 3.26, "learning_rate": 6.65793103448276e-06, "loss": 0.0702, "step": 5350 }, { "epoch": 3.27, "learning_rate": 6.6406896551724145e-06, "loss": 0.0699, "step": 5375 }, { "epoch": 3.29, "learning_rate": 6.62344827586207e-06, "loss": 0.0684, "step": 5400 }, { "epoch": 3.3, "learning_rate": 6.606206896551724e-06, "loss": 0.0694, "step": 5425 }, { "epoch": 3.32, "learning_rate": 6.58896551724138e-06, "loss": 0.0707, "step": 5450 }, { "epoch": 3.33, "learning_rate": 6.571724137931035e-06, "loss": 0.064, "step": 5475 }, { "epoch": 3.35, "learning_rate": 6.5544827586206905e-06, "loss": 0.066, "step": 5500 }, { "epoch": 3.36, "learning_rate": 6.537241379310345e-06, "loss": 0.0655, "step": 5525 }, { "epoch": 3.38, "learning_rate": 6.520000000000001e-06, "loss": 0.064, "step": 5550 }, { "epoch": 3.39, "learning_rate": 6.502758620689655e-06, "loss": 0.0713, "step": 5575 }, { "epoch": 3.41, "learning_rate": 6.485517241379311e-06, "loss": 0.0684, "step": 5600 }, { "epoch": 3.42, "learning_rate": 6.468275862068966e-06, "loss": 0.0724, "step": 5625 }, { "epoch": 3.44, "learning_rate": 6.4510344827586214e-06, "loss": 0.0708, "step": 5650 }, { "epoch": 3.45, "learning_rate": 6.433793103448276e-06, "loss": 0.0669, "step": 5675 }, { "epoch": 3.47, "learning_rate": 6.416551724137931e-06, "loss": 0.0665, "step": 5700 }, { "epoch": 3.48, "learning_rate": 6.399310344827587e-06, "loss": 0.071, "step": 5725 }, { "epoch": 3.5, "learning_rate": 6.382068965517242e-06, "loss": 0.0734, "step": 5750 }, { "epoch": 3.51, "learning_rate": 6.3648275862068975e-06, "loss": 0.0649, "step": 5775 }, { "epoch": 3.53, "learning_rate": 6.347586206896552e-06, "loss": 0.067, "step": 5800 }, { "epoch": 3.55, "learning_rate": 6.330344827586208e-06, "loss": 0.0671, "step": 5825 }, { "epoch": 3.56, "learning_rate": 6.313103448275862e-06, "loss": 0.0742, "step": 5850 }, { "epoch": 3.58, "learning_rate": 6.295862068965518e-06, "loss": 0.0701, "step": 5875 }, { "epoch": 3.59, "learning_rate": 6.278620689655173e-06, "loss": 0.0692, "step": 5900 }, { "epoch": 3.61, "learning_rate": 6.261379310344828e-06, "loss": 0.0725, "step": 5925 }, { "epoch": 3.62, "learning_rate": 6.244137931034483e-06, "loss": 0.0682, "step": 5950 }, { "epoch": 3.64, "learning_rate": 6.226896551724139e-06, "loss": 0.0698, "step": 5975 }, { "epoch": 3.65, "learning_rate": 6.209655172413793e-06, "loss": 0.0702, "step": 6000 }, { "epoch": 3.65, "eval_cer": 8.83954179332612, "eval_loss": 0.24995653331279755, "eval_runtime": 2191.0157, "eval_samples_per_second": 0.861, "eval_steps_per_second": 0.027, "eval_wer": 25.305877212147692, "step": 6000 }, { "epoch": 3.67, "learning_rate": 6.192413793103449e-06, "loss": 0.0661, "step": 6025 }, { "epoch": 3.68, "learning_rate": 6.175172413793104e-06, "loss": 0.0681, "step": 6050 }, { "epoch": 3.7, "learning_rate": 6.157931034482759e-06, "loss": 0.0695, "step": 6075 }, { "epoch": 3.71, "learning_rate": 6.140689655172414e-06, "loss": 0.0739, "step": 6100 }, { "epoch": 3.73, "learning_rate": 6.123448275862069e-06, "loss": 0.0647, "step": 6125 }, { "epoch": 3.74, "learning_rate": 6.106206896551725e-06, "loss": 0.0711, "step": 6150 }, { "epoch": 3.76, "learning_rate": 6.08896551724138e-06, "loss": 0.0652, "step": 6175 }, { "epoch": 3.77, "learning_rate": 6.071724137931035e-06, "loss": 0.0716, "step": 6200 }, { "epoch": 3.79, "learning_rate": 6.05448275862069e-06, "loss": 0.0699, "step": 6225 }, { "epoch": 3.8, "learning_rate": 6.037241379310346e-06, "loss": 0.0704, "step": 6250 }, { "epoch": 3.82, "learning_rate": 6.02e-06, "loss": 0.0656, "step": 6275 }, { "epoch": 3.83, "learning_rate": 6.002758620689656e-06, "loss": 0.0682, "step": 6300 }, { "epoch": 3.85, "learning_rate": 5.9855172413793105e-06, "loss": 0.0689, "step": 6325 }, { "epoch": 3.86, "learning_rate": 5.968275862068966e-06, "loss": 0.0679, "step": 6350 }, { "epoch": 3.88, "learning_rate": 5.951034482758621e-06, "loss": 0.0689, "step": 6375 }, { "epoch": 3.9, "learning_rate": 5.933793103448277e-06, "loss": 0.0698, "step": 6400 }, { "epoch": 3.91, "learning_rate": 5.916551724137931e-06, "loss": 0.0674, "step": 6425 }, { "epoch": 3.93, "learning_rate": 5.899310344827587e-06, "loss": 0.0699, "step": 6450 }, { "epoch": 3.94, "learning_rate": 5.8820689655172415e-06, "loss": 0.0673, "step": 6475 }, { "epoch": 3.96, "learning_rate": 5.864827586206897e-06, "loss": 0.0671, "step": 6500 }, { "epoch": 3.97, "learning_rate": 5.847586206896552e-06, "loss": 0.0717, "step": 6525 }, { "epoch": 3.99, "learning_rate": 5.830344827586208e-06, "loss": 0.0706, "step": 6550 }, { "epoch": 4.0, "learning_rate": 5.813103448275863e-06, "loss": 0.0639, "step": 6575 }, { "epoch": 4.02, "learning_rate": 5.7958620689655175e-06, "loss": 0.0362, "step": 6600 }, { "epoch": 4.03, "learning_rate": 5.778620689655173e-06, "loss": 0.0328, "step": 6625 }, { "epoch": 4.05, "learning_rate": 5.761379310344828e-06, "loss": 0.0355, "step": 6650 }, { "epoch": 4.06, "learning_rate": 5.744137931034484e-06, "loss": 0.0334, "step": 6675 }, { "epoch": 4.08, "learning_rate": 5.726896551724138e-06, "loss": 0.0358, "step": 6700 }, { "epoch": 4.09, "learning_rate": 5.7096551724137936e-06, "loss": 0.0337, "step": 6725 }, { "epoch": 4.11, "learning_rate": 5.692413793103448e-06, "loss": 0.0355, "step": 6750 }, { "epoch": 4.12, "learning_rate": 5.675172413793104e-06, "loss": 0.0338, "step": 6775 }, { "epoch": 4.14, "learning_rate": 5.657931034482759e-06, "loss": 0.0347, "step": 6800 }, { "epoch": 4.15, "learning_rate": 5.640689655172415e-06, "loss": 0.033, "step": 6825 }, { "epoch": 4.17, "learning_rate": 5.623448275862069e-06, "loss": 0.0376, "step": 6850 }, { "epoch": 4.18, "learning_rate": 5.6062068965517245e-06, "loss": 0.0353, "step": 6875 }, { "epoch": 4.2, "learning_rate": 5.588965517241379e-06, "loss": 0.0369, "step": 6900 }, { "epoch": 4.21, "learning_rate": 5.571724137931035e-06, "loss": 0.0363, "step": 6925 }, { "epoch": 4.23, "learning_rate": 5.554482758620691e-06, "loss": 0.0353, "step": 6950 }, { "epoch": 4.25, "learning_rate": 5.537241379310346e-06, "loss": 0.0348, "step": 6975 }, { "epoch": 4.26, "learning_rate": 5.5200000000000005e-06, "loss": 0.0338, "step": 7000 }, { "epoch": 4.26, "eval_cer": 8.612842839232657, "eval_loss": 0.26946696639060974, "eval_runtime": 2199.8568, "eval_samples_per_second": 0.857, "eval_steps_per_second": 0.027, "eval_wer": 24.781516277037362, "step": 7000 }, { "epoch": 4.28, "learning_rate": 5.502758620689655e-06, "loss": 0.0344, "step": 7025 }, { "epoch": 4.29, "learning_rate": 5.485517241379311e-06, "loss": 0.0349, "step": 7050 }, { "epoch": 4.31, "learning_rate": 5.468275862068966e-06, "loss": 0.0362, "step": 7075 }, { "epoch": 4.32, "learning_rate": 5.451034482758622e-06, "loss": 0.0325, "step": 7100 }, { "epoch": 4.34, "learning_rate": 5.433793103448276e-06, "loss": 0.0352, "step": 7125 }, { "epoch": 4.35, "learning_rate": 5.4165517241379314e-06, "loss": 0.0357, "step": 7150 }, { "epoch": 4.37, "learning_rate": 5.399310344827586e-06, "loss": 0.0357, "step": 7175 }, { "epoch": 4.38, "learning_rate": 5.382068965517242e-06, "loss": 0.0347, "step": 7200 }, { "epoch": 4.4, "learning_rate": 5.364827586206897e-06, "loss": 0.0379, "step": 7225 }, { "epoch": 4.41, "learning_rate": 5.347586206896553e-06, "loss": 0.0334, "step": 7250 }, { "epoch": 4.43, "learning_rate": 5.330344827586207e-06, "loss": 0.036, "step": 7275 }, { "epoch": 4.44, "learning_rate": 5.313103448275862e-06, "loss": 0.0356, "step": 7300 }, { "epoch": 4.46, "learning_rate": 5.295862068965517e-06, "loss": 0.0362, "step": 7325 }, { "epoch": 4.47, "learning_rate": 5.278620689655173e-06, "loss": 0.0364, "step": 7350 }, { "epoch": 4.49, "learning_rate": 5.261379310344829e-06, "loss": 0.0344, "step": 7375 }, { "epoch": 4.5, "learning_rate": 5.2441379310344835e-06, "loss": 0.0363, "step": 7400 }, { "epoch": 4.52, "learning_rate": 5.226896551724138e-06, "loss": 0.0359, "step": 7425 }, { "epoch": 4.53, "learning_rate": 5.209655172413793e-06, "loss": 0.0367, "step": 7450 }, { "epoch": 4.55, "learning_rate": 5.192413793103449e-06, "loss": 0.0335, "step": 7475 }, { "epoch": 4.56, "learning_rate": 5.175172413793104e-06, "loss": 0.0363, "step": 7500 }, { "epoch": 4.58, "learning_rate": 5.1579310344827596e-06, "loss": 0.0315, "step": 7525 }, { "epoch": 4.6, "learning_rate": 5.140689655172414e-06, "loss": 0.0359, "step": 7550 }, { "epoch": 4.61, "learning_rate": 5.123448275862069e-06, "loss": 0.037, "step": 7575 }, { "epoch": 4.63, "learning_rate": 5.106206896551724e-06, "loss": 0.0347, "step": 7600 }, { "epoch": 4.64, "learning_rate": 5.08896551724138e-06, "loss": 0.0364, "step": 7625 }, { "epoch": 4.66, "learning_rate": 5.071724137931035e-06, "loss": 0.0373, "step": 7650 }, { "epoch": 4.67, "learning_rate": 5.0544827586206905e-06, "loss": 0.0325, "step": 7675 }, { "epoch": 4.69, "learning_rate": 5.0372413793103445e-06, "loss": 0.0348, "step": 7700 }, { "epoch": 4.7, "learning_rate": 5.02e-06, "loss": 0.0375, "step": 7725 }, { "epoch": 4.72, "learning_rate": 5.002758620689656e-06, "loss": 0.0347, "step": 7750 }, { "epoch": 4.73, "learning_rate": 4.985517241379311e-06, "loss": 0.0378, "step": 7775 }, { "epoch": 4.75, "learning_rate": 4.968275862068966e-06, "loss": 0.0376, "step": 7800 }, { "epoch": 4.76, "learning_rate": 4.951034482758621e-06, "loss": 0.0366, "step": 7825 }, { "epoch": 4.78, "learning_rate": 4.933793103448276e-06, "loss": 0.0328, "step": 7850 }, { "epoch": 4.79, "learning_rate": 4.916551724137931e-06, "loss": 0.0384, "step": 7875 }, { "epoch": 4.81, "learning_rate": 4.899310344827586e-06, "loss": 0.0374, "step": 7900 }, { "epoch": 4.82, "learning_rate": 4.882068965517242e-06, "loss": 0.035, "step": 7925 }, { "epoch": 4.84, "learning_rate": 4.8648275862068974e-06, "loss": 0.0344, "step": 7950 }, { "epoch": 4.85, "learning_rate": 4.847586206896552e-06, "loss": 0.0357, "step": 7975 }, { "epoch": 4.87, "learning_rate": 4.830344827586207e-06, "loss": 0.0351, "step": 8000 }, { "epoch": 4.87, "eval_cer": 8.811204424064439, "eval_loss": 0.2785949110984802, "eval_runtime": 2203.1021, "eval_samples_per_second": 0.856, "eval_steps_per_second": 0.027, "eval_wer": 25.032772558444393, "step": 8000 }, { "epoch": 4.88, "learning_rate": 4.813103448275863e-06, "loss": 0.0373, "step": 8025 }, { "epoch": 4.9, "learning_rate": 4.795862068965518e-06, "loss": 0.035, "step": 8050 }, { "epoch": 4.91, "learning_rate": 4.778620689655173e-06, "loss": 0.0333, "step": 8075 }, { "epoch": 4.93, "learning_rate": 4.761379310344828e-06, "loss": 0.036, "step": 8100 }, { "epoch": 4.95, "learning_rate": 4.744137931034483e-06, "loss": 0.0363, "step": 8125 }, { "epoch": 4.96, "learning_rate": 4.726896551724138e-06, "loss": 0.0357, "step": 8150 }, { "epoch": 4.98, "learning_rate": 4.709655172413794e-06, "loss": 0.0348, "step": 8175 }, { "epoch": 4.99, "learning_rate": 4.692413793103449e-06, "loss": 0.0364, "step": 8200 }, { "epoch": 5.01, "learning_rate": 4.6751724137931035e-06, "loss": 0.0277, "step": 8225 }, { "epoch": 5.02, "learning_rate": 4.657931034482759e-06, "loss": 0.0147, "step": 8250 }, { "epoch": 5.04, "learning_rate": 4.640689655172414e-06, "loss": 0.0152, "step": 8275 }, { "epoch": 5.05, "learning_rate": 4.623448275862069e-06, "loss": 0.0148, "step": 8300 }, { "epoch": 5.07, "learning_rate": 4.606206896551724e-06, "loss": 0.0156, "step": 8325 }, { "epoch": 5.08, "learning_rate": 4.58896551724138e-06, "loss": 0.014, "step": 8350 }, { "epoch": 5.1, "learning_rate": 4.571724137931035e-06, "loss": 0.0143, "step": 8375 }, { "epoch": 5.11, "learning_rate": 4.55448275862069e-06, "loss": 0.0143, "step": 8400 }, { "epoch": 5.13, "learning_rate": 4.537241379310345e-06, "loss": 0.0146, "step": 8425 }, { "epoch": 5.14, "learning_rate": 4.520000000000001e-06, "loss": 0.0141, "step": 8450 }, { "epoch": 5.16, "learning_rate": 4.502758620689656e-06, "loss": 0.0151, "step": 8475 }, { "epoch": 5.17, "learning_rate": 4.4855172413793105e-06, "loss": 0.0141, "step": 8500 }, { "epoch": 5.19, "learning_rate": 4.468275862068966e-06, "loss": 0.0148, "step": 8525 }, { "epoch": 5.2, "learning_rate": 4.451034482758621e-06, "loss": 0.0147, "step": 8550 }, { "epoch": 5.22, "learning_rate": 4.433793103448276e-06, "loss": 0.0145, "step": 8575 }, { "epoch": 5.23, "learning_rate": 4.416551724137932e-06, "loss": 0.0141, "step": 8600 }, { "epoch": 5.25, "learning_rate": 4.3993103448275866e-06, "loss": 0.0153, "step": 8625 }, { "epoch": 5.26, "learning_rate": 4.3820689655172414e-06, "loss": 0.0153, "step": 8650 }, { "epoch": 5.28, "learning_rate": 4.364827586206897e-06, "loss": 0.0148, "step": 8675 }, { "epoch": 5.3, "learning_rate": 4.347586206896552e-06, "loss": 0.0146, "step": 8700 }, { "epoch": 5.31, "learning_rate": 4.330344827586207e-06, "loss": 0.0151, "step": 8725 }, { "epoch": 5.33, "learning_rate": 4.313103448275863e-06, "loss": 0.0145, "step": 8750 }, { "epoch": 5.34, "learning_rate": 4.2958620689655175e-06, "loss": 0.0148, "step": 8775 }, { "epoch": 5.36, "learning_rate": 4.278620689655173e-06, "loss": 0.015, "step": 8800 }, { "epoch": 5.37, "learning_rate": 4.261379310344828e-06, "loss": 0.0169, "step": 8825 }, { "epoch": 5.39, "learning_rate": 4.244137931034483e-06, "loss": 0.0143, "step": 8850 }, { "epoch": 5.4, "learning_rate": 4.226896551724139e-06, "loss": 0.0143, "step": 8875 }, { "epoch": 5.42, "learning_rate": 4.2096551724137935e-06, "loss": 0.0166, "step": 8900 }, { "epoch": 5.43, "learning_rate": 4.192413793103448e-06, "loss": 0.0145, "step": 8925 }, { "epoch": 5.45, "learning_rate": 4.175172413793104e-06, "loss": 0.015, "step": 8950 }, { "epoch": 5.46, "learning_rate": 4.157931034482759e-06, "loss": 0.0156, "step": 8975 }, { "epoch": 5.48, "learning_rate": 4.140689655172414e-06, "loss": 0.0152, "step": 9000 }, { "epoch": 5.48, "eval_cer": 8.65835437168324, "eval_loss": 0.29967886209487915, "eval_runtime": 2187.4802, "eval_samples_per_second": 0.862, "eval_steps_per_second": 0.027, "eval_wer": 24.677736508630108, "step": 9000 }, { "epoch": 5.49, "learning_rate": 4.1234482758620696e-06, "loss": 0.0152, "step": 9025 }, { "epoch": 5.51, "learning_rate": 4.1062068965517244e-06, "loss": 0.0155, "step": 9050 }, { "epoch": 5.52, "learning_rate": 4.088965517241379e-06, "loss": 0.0145, "step": 9075 }, { "epoch": 5.54, "learning_rate": 4.071724137931035e-06, "loss": 0.0146, "step": 9100 }, { "epoch": 5.55, "learning_rate": 4.05448275862069e-06, "loss": 0.0152, "step": 9125 }, { "epoch": 5.57, "learning_rate": 4.037241379310346e-06, "loss": 0.0149, "step": 9150 }, { "epoch": 5.58, "learning_rate": 4.0200000000000005e-06, "loss": 0.0146, "step": 9175 }, { "epoch": 5.6, "learning_rate": 4.002758620689655e-06, "loss": 0.0151, "step": 9200 }, { "epoch": 5.61, "learning_rate": 3.985517241379311e-06, "loss": 0.0155, "step": 9225 }, { "epoch": 5.63, "learning_rate": 3.968275862068966e-06, "loss": 0.0148, "step": 9250 }, { "epoch": 5.65, "learning_rate": 3.951034482758621e-06, "loss": 0.0152, "step": 9275 }, { "epoch": 5.66, "learning_rate": 3.9337931034482765e-06, "loss": 0.014, "step": 9300 }, { "epoch": 5.68, "learning_rate": 3.916551724137931e-06, "loss": 0.0142, "step": 9325 }, { "epoch": 5.69, "learning_rate": 3.899310344827586e-06, "loss": 0.0162, "step": 9350 }, { "epoch": 5.71, "learning_rate": 3.882068965517242e-06, "loss": 0.0147, "step": 9375 }, { "epoch": 5.72, "learning_rate": 3.864827586206897e-06, "loss": 0.0146, "step": 9400 }, { "epoch": 5.74, "learning_rate": 3.847586206896552e-06, "loss": 0.0153, "step": 9425 }, { "epoch": 5.75, "learning_rate": 3.8303448275862074e-06, "loss": 0.0156, "step": 9450 }, { "epoch": 5.77, "learning_rate": 3.8131034482758623e-06, "loss": 0.0143, "step": 9475 }, { "epoch": 5.78, "learning_rate": 3.795862068965517e-06, "loss": 0.0143, "step": 9500 }, { "epoch": 5.8, "learning_rate": 3.7786206896551725e-06, "loss": 0.0155, "step": 9525 }, { "epoch": 5.81, "learning_rate": 3.7613793103448278e-06, "loss": 0.0149, "step": 9550 }, { "epoch": 5.83, "learning_rate": 3.7441379310344835e-06, "loss": 0.0139, "step": 9575 }, { "epoch": 5.84, "learning_rate": 3.7268965517241383e-06, "loss": 0.0151, "step": 9600 }, { "epoch": 5.86, "learning_rate": 3.7096551724137936e-06, "loss": 0.0158, "step": 9625 }, { "epoch": 5.87, "learning_rate": 3.6924137931034485e-06, "loss": 0.0153, "step": 9650 }, { "epoch": 5.89, "learning_rate": 3.675172413793104e-06, "loss": 0.0146, "step": 9675 }, { "epoch": 5.9, "learning_rate": 3.657931034482759e-06, "loss": 0.0147, "step": 9700 }, { "epoch": 5.92, "learning_rate": 3.640689655172414e-06, "loss": 0.0151, "step": 9725 }, { "epoch": 5.93, "learning_rate": 3.6234482758620693e-06, "loss": 0.0152, "step": 9750 }, { "epoch": 5.95, "learning_rate": 3.6062068965517246e-06, "loss": 0.0147, "step": 9775 }, { "epoch": 5.96, "learning_rate": 3.5889655172413794e-06, "loss": 0.0151, "step": 9800 }, { "epoch": 5.98, "learning_rate": 3.5717241379310347e-06, "loss": 0.0155, "step": 9825 }, { "epoch": 6.0, "learning_rate": 3.55448275862069e-06, "loss": 0.0139, "step": 9850 }, { "epoch": 6.01, "learning_rate": 3.537241379310345e-06, "loss": 0.0088, "step": 9875 }, { "epoch": 6.03, "learning_rate": 3.52e-06, "loss": 0.006, "step": 9900 }, { "epoch": 6.04, "learning_rate": 3.502758620689655e-06, "loss": 0.0059, "step": 9925 }, { "epoch": 6.06, "learning_rate": 3.4855172413793103e-06, "loss": 0.0049, "step": 9950 }, { "epoch": 6.07, "learning_rate": 3.468275862068966e-06, "loss": 0.005, "step": 9975 }, { "epoch": 6.09, "learning_rate": 3.4510344827586214e-06, "loss": 0.0052, "step": 10000 }, { "epoch": 6.09, "eval_cer": 8.7073007367716, "eval_loss": 0.31903430819511414, "eval_runtime": 2191.4289, "eval_samples_per_second": 0.861, "eval_steps_per_second": 0.027, "eval_wer": 24.939916976185277, "step": 10000 }, { "epoch": 6.1, "learning_rate": 3.4337931034482762e-06, "loss": 0.0052, "step": 10025 }, { "epoch": 6.12, "learning_rate": 3.4165517241379315e-06, "loss": 0.0056, "step": 10050 }, { "epoch": 6.13, "learning_rate": 3.3993103448275864e-06, "loss": 0.0059, "step": 10075 }, { "epoch": 6.15, "learning_rate": 3.3820689655172417e-06, "loss": 0.005, "step": 10100 }, { "epoch": 6.16, "learning_rate": 3.364827586206897e-06, "loss": 0.0051, "step": 10125 }, { "epoch": 6.18, "learning_rate": 3.347586206896552e-06, "loss": 0.006, "step": 10150 }, { "epoch": 6.19, "learning_rate": 3.330344827586207e-06, "loss": 0.0052, "step": 10175 }, { "epoch": 6.21, "learning_rate": 3.3131034482758624e-06, "loss": 0.0051, "step": 10200 }, { "epoch": 6.22, "learning_rate": 3.2958620689655173e-06, "loss": 0.0053, "step": 10225 }, { "epoch": 6.24, "learning_rate": 3.2786206896551726e-06, "loss": 0.0051, "step": 10250 }, { "epoch": 6.25, "learning_rate": 3.261379310344828e-06, "loss": 0.0052, "step": 10275 }, { "epoch": 6.27, "learning_rate": 3.2441379310344828e-06, "loss": 0.0053, "step": 10300 }, { "epoch": 6.28, "learning_rate": 3.226896551724138e-06, "loss": 0.0047, "step": 10325 }, { "epoch": 6.3, "learning_rate": 3.209655172413793e-06, "loss": 0.0053, "step": 10350 }, { "epoch": 6.31, "learning_rate": 3.1924137931034486e-06, "loss": 0.0055, "step": 10375 }, { "epoch": 6.33, "learning_rate": 3.175172413793104e-06, "loss": 0.0057, "step": 10400 }, { "epoch": 6.35, "learning_rate": 3.1579310344827592e-06, "loss": 0.0053, "step": 10425 }, { "epoch": 6.36, "learning_rate": 3.140689655172414e-06, "loss": 0.0051, "step": 10450 }, { "epoch": 6.38, "learning_rate": 3.1234482758620694e-06, "loss": 0.0052, "step": 10475 }, { "epoch": 6.39, "learning_rate": 3.1062068965517243e-06, "loss": 0.0057, "step": 10500 }, { "epoch": 6.41, "learning_rate": 3.0889655172413796e-06, "loss": 0.0051, "step": 10525 }, { "epoch": 6.42, "learning_rate": 3.071724137931035e-06, "loss": 0.0048, "step": 10550 }, { "epoch": 6.44, "learning_rate": 3.0544827586206897e-06, "loss": 0.0055, "step": 10575 }, { "epoch": 6.45, "learning_rate": 3.037241379310345e-06, "loss": 0.0052, "step": 10600 }, { "epoch": 6.47, "learning_rate": 3.0200000000000003e-06, "loss": 0.0055, "step": 10625 }, { "epoch": 6.48, "learning_rate": 3.002758620689655e-06, "loss": 0.0059, "step": 10650 }, { "epoch": 6.5, "learning_rate": 2.9855172413793105e-06, "loss": 0.0051, "step": 10675 }, { "epoch": 6.51, "learning_rate": 2.9682758620689658e-06, "loss": 0.0054, "step": 10700 }, { "epoch": 6.53, "learning_rate": 2.9510344827586206e-06, "loss": 0.0059, "step": 10725 }, { "epoch": 6.54, "learning_rate": 2.933793103448276e-06, "loss": 0.0058, "step": 10750 }, { "epoch": 6.56, "learning_rate": 2.9165517241379316e-06, "loss": 0.006, "step": 10775 }, { "epoch": 6.57, "learning_rate": 2.8993103448275865e-06, "loss": 0.0047, "step": 10800 }, { "epoch": 6.59, "learning_rate": 2.882068965517242e-06, "loss": 0.0052, "step": 10825 }, { "epoch": 6.6, "learning_rate": 2.864827586206897e-06, "loss": 0.0049, "step": 10850 }, { "epoch": 6.62, "learning_rate": 2.847586206896552e-06, "loss": 0.0049, "step": 10875 }, { "epoch": 6.63, "learning_rate": 2.8303448275862073e-06, "loss": 0.0049, "step": 10900 }, { "epoch": 6.65, "learning_rate": 2.813103448275862e-06, "loss": 0.0053, "step": 10925 }, { "epoch": 6.66, "learning_rate": 2.7958620689655174e-06, "loss": 0.0053, "step": 10950 }, { "epoch": 6.68, "learning_rate": 2.7793103448275864e-06, "loss": 0.0052, "step": 10975 }, { "epoch": 6.7, "learning_rate": 2.7620689655172417e-06, "loss": 0.0056, "step": 11000 }, { "epoch": 6.7, "eval_cer": 8.672093702234358, "eval_loss": 0.3261697292327881, "eval_runtime": 2191.5507, "eval_samples_per_second": 0.861, "eval_steps_per_second": 0.027, "eval_wer": 24.655888136333843, "step": 11000 } ], "max_steps": 15000, "num_train_epochs": 10, "total_flos": 3.5914882574647296e+20, "trial_name": null, "trial_params": null }