{ "best_metric": 30.36923749180686, "best_model_checkpoint": "whisper-small-mn-3/checkpoint-11000", "epoch": 9.129640900791236, "global_step": 15000, "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": 5.6947, "step": 25 }, { "epoch": 0.03, "learning_rate": 9.200000000000001e-07, "loss": 4.2083, "step": 50 }, { "epoch": 0.05, "learning_rate": 1.42e-06, "loss": 3.2539, "step": 75 }, { "epoch": 0.06, "learning_rate": 1.9200000000000003e-06, "loss": 2.5471, "step": 100 }, { "epoch": 0.08, "learning_rate": 2.42e-06, "loss": 2.0523, "step": 125 }, { "epoch": 0.09, "learning_rate": 2.92e-06, "loss": 1.654, "step": 150 }, { "epoch": 0.11, "learning_rate": 3.4200000000000007e-06, "loss": 1.4119, "step": 175 }, { "epoch": 0.12, "learning_rate": 3.920000000000001e-06, "loss": 1.1999, "step": 200 }, { "epoch": 0.14, "learning_rate": 4.42e-06, "loss": 1.1094, "step": 225 }, { "epoch": 0.15, "learning_rate": 4.92e-06, "loss": 0.9837, "step": 250 }, { "epoch": 0.17, "learning_rate": 5.420000000000001e-06, "loss": 0.9029, "step": 275 }, { "epoch": 0.18, "learning_rate": 5.92e-06, "loss": 0.8462, "step": 300 }, { "epoch": 0.2, "learning_rate": 6.42e-06, "loss": 0.778, "step": 325 }, { "epoch": 0.21, "learning_rate": 6.92e-06, "loss": 0.7245, "step": 350 }, { "epoch": 0.23, "learning_rate": 7.420000000000001e-06, "loss": 0.6804, "step": 375 }, { "epoch": 0.24, "learning_rate": 7.92e-06, "loss": 0.6289, "step": 400 }, { "epoch": 0.26, "learning_rate": 8.42e-06, "loss": 0.5689, "step": 425 }, { "epoch": 0.27, "learning_rate": 8.920000000000001e-06, "loss": 0.5493, "step": 450 }, { "epoch": 0.29, "learning_rate": 9.42e-06, "loss": 0.5268, "step": 475 }, { "epoch": 0.3, "learning_rate": 9.920000000000002e-06, "loss": 0.5195, "step": 500 }, { "epoch": 0.32, "learning_rate": 9.98551724137931e-06, "loss": 0.5135, "step": 525 }, { "epoch": 0.33, "learning_rate": 9.968275862068967e-06, "loss": 0.4805, "step": 550 }, { "epoch": 0.35, "learning_rate": 9.951034482758622e-06, "loss": 0.4608, "step": 575 }, { "epoch": 0.37, "learning_rate": 9.933793103448277e-06, "loss": 0.4424, "step": 600 }, { "epoch": 0.38, "learning_rate": 9.916551724137932e-06, "loss": 0.4288, "step": 625 }, { "epoch": 0.4, "learning_rate": 9.899310344827588e-06, "loss": 0.4252, "step": 650 }, { "epoch": 0.41, "learning_rate": 9.882068965517241e-06, "loss": 0.4356, "step": 675 }, { "epoch": 0.43, "learning_rate": 9.864827586206898e-06, "loss": 0.4174, "step": 700 }, { "epoch": 0.44, "learning_rate": 9.847586206896553e-06, "loss": 0.3962, "step": 725 }, { "epoch": 0.46, "learning_rate": 9.830344827586208e-06, "loss": 0.3975, "step": 750 }, { "epoch": 0.47, "learning_rate": 9.813103448275862e-06, "loss": 0.3973, "step": 775 }, { "epoch": 0.49, "learning_rate": 9.795862068965517e-06, "loss": 0.3886, "step": 800 }, { "epoch": 0.5, "learning_rate": 9.778620689655172e-06, "loss": 0.365, "step": 825 }, { "epoch": 0.52, "learning_rate": 9.761379310344829e-06, "loss": 0.3759, "step": 850 }, { "epoch": 0.53, "learning_rate": 9.744137931034484e-06, "loss": 0.3723, "step": 875 }, { "epoch": 0.55, "learning_rate": 9.726896551724139e-06, "loss": 0.3636, "step": 900 }, { "epoch": 0.56, "learning_rate": 9.709655172413795e-06, "loss": 0.361, "step": 925 }, { "epoch": 0.58, "learning_rate": 9.692413793103448e-06, "loss": 0.3557, "step": 950 }, { "epoch": 0.59, "learning_rate": 9.675172413793105e-06, "loss": 0.3399, "step": 975 }, { "epoch": 0.61, "learning_rate": 9.65793103448276e-06, "loss": 0.3408, "step": 1000 }, { "epoch": 0.61, "eval_cer": 17.38111185532485, "eval_loss": 0.4061901271343231, "eval_runtime": 799.3563, "eval_samples_per_second": 2.359, "eval_steps_per_second": 0.074, "eval_wer": 47.68407253659602, "step": 1000 }, { "epoch": 0.62, "learning_rate": 9.640689655172415e-06, "loss": 0.3427, "step": 1025 }, { "epoch": 0.64, "learning_rate": 9.62344827586207e-06, "loss": 0.3368, "step": 1050 }, { "epoch": 0.65, "learning_rate": 9.606206896551726e-06, "loss": 0.3219, "step": 1075 }, { "epoch": 0.67, "learning_rate": 9.58896551724138e-06, "loss": 0.3325, "step": 1100 }, { "epoch": 0.68, "learning_rate": 9.571724137931036e-06, "loss": 0.3302, "step": 1125 }, { "epoch": 0.7, "learning_rate": 9.55448275862069e-06, "loss": 0.3232, "step": 1150 }, { "epoch": 0.72, "learning_rate": 9.537241379310345e-06, "loss": 0.3406, "step": 1175 }, { "epoch": 0.73, "learning_rate": 9.52e-06, "loss": 0.3289, "step": 1200 }, { "epoch": 0.75, "learning_rate": 9.502758620689655e-06, "loss": 0.3095, "step": 1225 }, { "epoch": 0.76, "learning_rate": 9.48551724137931e-06, "loss": 0.3205, "step": 1250 }, { "epoch": 0.78, "learning_rate": 9.468275862068967e-06, "loss": 0.3187, "step": 1275 }, { "epoch": 0.79, "learning_rate": 9.451034482758622e-06, "loss": 0.3094, "step": 1300 }, { "epoch": 0.81, "learning_rate": 9.433793103448276e-06, "loss": 0.3037, "step": 1325 }, { "epoch": 0.82, "learning_rate": 9.416551724137933e-06, "loss": 0.3022, "step": 1350 }, { "epoch": 0.84, "learning_rate": 9.399310344827586e-06, "loss": 0.3045, "step": 1375 }, { "epoch": 0.85, "learning_rate": 9.382068965517243e-06, "loss": 0.3022, "step": 1400 }, { "epoch": 0.87, "learning_rate": 9.364827586206898e-06, "loss": 0.2994, "step": 1425 }, { "epoch": 0.88, "learning_rate": 9.347586206896552e-06, "loss": 0.3012, "step": 1450 }, { "epoch": 0.9, "learning_rate": 9.330344827586207e-06, "loss": 0.2853, "step": 1475 }, { "epoch": 0.91, "learning_rate": 9.313103448275864e-06, "loss": 0.2884, "step": 1500 }, { "epoch": 0.93, "learning_rate": 9.295862068965517e-06, "loss": 0.276, "step": 1525 }, { "epoch": 0.94, "learning_rate": 9.278620689655174e-06, "loss": 0.2871, "step": 1550 }, { "epoch": 0.96, "learning_rate": 9.261379310344828e-06, "loss": 0.2879, "step": 1575 }, { "epoch": 0.97, "learning_rate": 9.244137931034483e-06, "loss": 0.2825, "step": 1600 }, { "epoch": 0.99, "learning_rate": 9.226896551724138e-06, "loss": 0.2903, "step": 1625 }, { "epoch": 1.0, "learning_rate": 9.209655172413793e-06, "loss": 0.2756, "step": 1650 }, { "epoch": 1.02, "learning_rate": 9.192413793103448e-06, "loss": 0.241, "step": 1675 }, { "epoch": 1.03, "learning_rate": 9.175172413793105e-06, "loss": 0.2435, "step": 1700 }, { "epoch": 1.05, "learning_rate": 9.15793103448276e-06, "loss": 0.2525, "step": 1725 }, { "epoch": 1.07, "learning_rate": 9.140689655172414e-06, "loss": 0.237, "step": 1750 }, { "epoch": 1.08, "learning_rate": 9.12344827586207e-06, "loss": 0.2354, "step": 1775 }, { "epoch": 1.1, "learning_rate": 9.106206896551724e-06, "loss": 0.2416, "step": 1800 }, { "epoch": 1.11, "learning_rate": 9.08896551724138e-06, "loss": 0.2377, "step": 1825 }, { "epoch": 1.13, "learning_rate": 9.071724137931035e-06, "loss": 0.226, "step": 1850 }, { "epoch": 1.14, "learning_rate": 9.05448275862069e-06, "loss": 0.2374, "step": 1875 }, { "epoch": 1.16, "learning_rate": 9.037241379310345e-06, "loss": 0.2421, "step": 1900 }, { "epoch": 1.17, "learning_rate": 9.020000000000002e-06, "loss": 0.2357, "step": 1925 }, { "epoch": 1.19, "learning_rate": 9.002758620689655e-06, "loss": 0.2379, "step": 1950 }, { "epoch": 1.2, "learning_rate": 8.985517241379311e-06, "loss": 0.2384, "step": 1975 }, { "epoch": 1.22, "learning_rate": 8.968275862068966e-06, "loss": 0.2261, "step": 2000 }, { "epoch": 1.22, "eval_cer": 13.646590069898846, "eval_loss": 0.3261679410934448, "eval_runtime": 756.3181, "eval_samples_per_second": 2.494, "eval_steps_per_second": 0.078, "eval_wer": 37.80860825868473, "step": 2000 }, { "epoch": 1.23, "learning_rate": 8.951034482758621e-06, "loss": 0.228, "step": 2025 }, { "epoch": 1.25, "learning_rate": 8.933793103448276e-06, "loss": 0.2321, "step": 2050 }, { "epoch": 1.26, "learning_rate": 8.916551724137931e-06, "loss": 0.2271, "step": 2075 }, { "epoch": 1.28, "learning_rate": 8.899310344827588e-06, "loss": 0.2413, "step": 2100 }, { "epoch": 1.29, "learning_rate": 8.882068965517242e-06, "loss": 0.2382, "step": 2125 }, { "epoch": 1.31, "learning_rate": 8.864827586206897e-06, "loss": 0.2423, "step": 2150 }, { "epoch": 1.32, "learning_rate": 8.847586206896552e-06, "loss": 0.2299, "step": 2175 }, { "epoch": 1.34, "learning_rate": 8.830344827586209e-06, "loss": 0.226, "step": 2200 }, { "epoch": 1.35, "learning_rate": 8.813103448275862e-06, "loss": 0.2267, "step": 2225 }, { "epoch": 1.37, "learning_rate": 8.795862068965518e-06, "loss": 0.2268, "step": 2250 }, { "epoch": 1.38, "learning_rate": 8.778620689655173e-06, "loss": 0.224, "step": 2275 }, { "epoch": 1.4, "learning_rate": 8.761379310344828e-06, "loss": 0.2355, "step": 2300 }, { "epoch": 1.42, "learning_rate": 8.744137931034483e-06, "loss": 0.2333, "step": 2325 }, { "epoch": 1.43, "learning_rate": 8.72689655172414e-06, "loss": 0.2345, "step": 2350 }, { "epoch": 1.45, "learning_rate": 8.709655172413793e-06, "loss": 0.2283, "step": 2375 }, { "epoch": 1.46, "learning_rate": 8.69241379310345e-06, "loss": 0.2235, "step": 2400 }, { "epoch": 1.48, "learning_rate": 8.675172413793104e-06, "loss": 0.2213, "step": 2425 }, { "epoch": 1.49, "learning_rate": 8.657931034482759e-06, "loss": 0.2207, "step": 2450 }, { "epoch": 1.51, "learning_rate": 8.640689655172414e-06, "loss": 0.2262, "step": 2475 }, { "epoch": 1.52, "learning_rate": 8.623448275862069e-06, "loss": 0.2307, "step": 2500 }, { "epoch": 1.54, "learning_rate": 8.606206896551725e-06, "loss": 0.2263, "step": 2525 }, { "epoch": 1.55, "learning_rate": 8.58896551724138e-06, "loss": 0.2151, "step": 2550 }, { "epoch": 1.57, "learning_rate": 8.571724137931035e-06, "loss": 0.2296, "step": 2575 }, { "epoch": 1.58, "learning_rate": 8.55448275862069e-06, "loss": 0.2257, "step": 2600 }, { "epoch": 1.6, "learning_rate": 8.537241379310347e-06, "loss": 0.2284, "step": 2625 }, { "epoch": 1.61, "learning_rate": 8.52e-06, "loss": 0.2208, "step": 2650 }, { "epoch": 1.63, "learning_rate": 8.502758620689656e-06, "loss": 0.223, "step": 2675 }, { "epoch": 1.64, "learning_rate": 8.485517241379311e-06, "loss": 0.212, "step": 2700 }, { "epoch": 1.66, "learning_rate": 8.468275862068966e-06, "loss": 0.2053, "step": 2725 }, { "epoch": 1.67, "learning_rate": 8.451034482758621e-06, "loss": 0.2179, "step": 2750 }, { "epoch": 1.69, "learning_rate": 8.433793103448277e-06, "loss": 0.2143, "step": 2775 }, { "epoch": 1.7, "learning_rate": 8.41655172413793e-06, "loss": 0.2152, "step": 2800 }, { "epoch": 1.72, "learning_rate": 8.399310344827587e-06, "loss": 0.214, "step": 2825 }, { "epoch": 1.73, "learning_rate": 8.382068965517242e-06, "loss": 0.2215, "step": 2850 }, { "epoch": 1.75, "learning_rate": 8.364827586206897e-06, "loss": 0.2204, "step": 2875 }, { "epoch": 1.77, "learning_rate": 8.347586206896552e-06, "loss": 0.2147, "step": 2900 }, { "epoch": 1.78, "learning_rate": 8.330344827586208e-06, "loss": 0.2144, "step": 2925 }, { "epoch": 1.8, "learning_rate": 8.313103448275863e-06, "loss": 0.218, "step": 2950 }, { "epoch": 1.81, "learning_rate": 8.295862068965518e-06, "loss": 0.2112, "step": 2975 }, { "epoch": 1.83, "learning_rate": 8.278620689655173e-06, "loss": 0.2135, "step": 3000 }, { "epoch": 1.83, "eval_cer": 12.224569357858037, "eval_loss": 0.2862820625305176, "eval_runtime": 755.8427, "eval_samples_per_second": 2.495, "eval_steps_per_second": 0.078, "eval_wer": 33.71750054620931, "step": 3000 }, { "epoch": 1.84, "learning_rate": 8.261379310344828e-06, "loss": 0.2221, "step": 3025 }, { "epoch": 1.86, "learning_rate": 8.244137931034484e-06, "loss": 0.2182, "step": 3050 }, { "epoch": 1.87, "learning_rate": 8.226896551724138e-06, "loss": 0.2137, "step": 3075 }, { "epoch": 1.89, "learning_rate": 8.209655172413794e-06, "loss": 0.2138, "step": 3100 }, { "epoch": 1.9, "learning_rate": 8.192413793103449e-06, "loss": 0.2065, "step": 3125 }, { "epoch": 1.92, "learning_rate": 8.175172413793104e-06, "loss": 0.2171, "step": 3150 }, { "epoch": 1.93, "learning_rate": 8.157931034482759e-06, "loss": 0.2091, "step": 3175 }, { "epoch": 1.95, "learning_rate": 8.140689655172415e-06, "loss": 0.2155, "step": 3200 }, { "epoch": 1.96, "learning_rate": 8.123448275862069e-06, "loss": 0.2192, "step": 3225 }, { "epoch": 1.98, "learning_rate": 8.106206896551725e-06, "loss": 0.2122, "step": 3250 }, { "epoch": 1.99, "learning_rate": 8.08896551724138e-06, "loss": 0.2176, "step": 3275 }, { "epoch": 2.01, "learning_rate": 8.071724137931035e-06, "loss": 0.1802, "step": 3300 }, { "epoch": 2.02, "learning_rate": 8.054482758620691e-06, "loss": 0.1647, "step": 3325 }, { "epoch": 2.04, "learning_rate": 8.037241379310346e-06, "loss": 0.1652, "step": 3350 }, { "epoch": 2.05, "learning_rate": 8.020000000000001e-06, "loss": 0.1713, "step": 3375 }, { "epoch": 2.07, "learning_rate": 8.002758620689656e-06, "loss": 0.1581, "step": 3400 }, { "epoch": 2.08, "learning_rate": 7.985517241379311e-06, "loss": 0.1663, "step": 3425 }, { "epoch": 2.1, "learning_rate": 7.968275862068966e-06, "loss": 0.1737, "step": 3450 }, { "epoch": 2.12, "learning_rate": 7.951034482758622e-06, "loss": 0.1629, "step": 3475 }, { "epoch": 2.13, "learning_rate": 7.933793103448275e-06, "loss": 0.1709, "step": 3500 }, { "epoch": 2.15, "learning_rate": 7.916551724137932e-06, "loss": 0.1766, "step": 3525 }, { "epoch": 2.16, "learning_rate": 7.899310344827587e-06, "loss": 0.1686, "step": 3550 }, { "epoch": 2.18, "learning_rate": 7.882068965517242e-06, "loss": 0.1682, "step": 3575 }, { "epoch": 2.19, "learning_rate": 7.864827586206897e-06, "loss": 0.1668, "step": 3600 }, { "epoch": 2.21, "learning_rate": 7.847586206896553e-06, "loss": 0.1626, "step": 3625 }, { "epoch": 2.22, "learning_rate": 7.830344827586206e-06, "loss": 0.1666, "step": 3650 }, { "epoch": 2.24, "learning_rate": 7.813103448275863e-06, "loss": 0.1698, "step": 3675 }, { "epoch": 2.25, "learning_rate": 7.795862068965518e-06, "loss": 0.1662, "step": 3700 }, { "epoch": 2.27, "learning_rate": 7.778620689655173e-06, "loss": 0.1651, "step": 3725 }, { "epoch": 2.28, "learning_rate": 7.76137931034483e-06, "loss": 0.1684, "step": 3750 }, { "epoch": 2.3, "learning_rate": 7.744137931034484e-06, "loss": 0.1678, "step": 3775 }, { "epoch": 2.31, "learning_rate": 7.726896551724139e-06, "loss": 0.1702, "step": 3800 }, { "epoch": 2.33, "learning_rate": 7.709655172413794e-06, "loss": 0.1719, "step": 3825 }, { "epoch": 2.34, "learning_rate": 7.692413793103449e-06, "loss": 0.1634, "step": 3850 }, { "epoch": 2.36, "learning_rate": 7.675172413793104e-06, "loss": 0.1684, "step": 3875 }, { "epoch": 2.37, "learning_rate": 7.65793103448276e-06, "loss": 0.1715, "step": 3900 }, { "epoch": 2.39, "learning_rate": 7.640689655172413e-06, "loss": 0.1656, "step": 3925 }, { "epoch": 2.4, "learning_rate": 7.62344827586207e-06, "loss": 0.1649, "step": 3950 }, { "epoch": 2.42, "learning_rate": 7.606206896551725e-06, "loss": 0.1563, "step": 3975 }, { "epoch": 2.43, "learning_rate": 7.58896551724138e-06, "loss": 0.1643, "step": 4000 }, { "epoch": 2.43, "eval_cer": 11.452590722517044, "eval_loss": 0.28033071756362915, "eval_runtime": 751.4141, "eval_samples_per_second": 2.51, "eval_steps_per_second": 0.079, "eval_wer": 32.59777146602578, "step": 4000 }, { "epoch": 2.45, "learning_rate": 7.5717241379310345e-06, "loss": 0.1613, "step": 4025 }, { "epoch": 2.47, "learning_rate": 7.55448275862069e-06, "loss": 0.1593, "step": 4050 }, { "epoch": 2.48, "learning_rate": 7.537241379310345e-06, "loss": 0.1659, "step": 4075 }, { "epoch": 2.5, "learning_rate": 7.520000000000001e-06, "loss": 0.1603, "step": 4100 }, { "epoch": 2.51, "learning_rate": 7.5027586206896566e-06, "loss": 0.1656, "step": 4125 }, { "epoch": 2.53, "learning_rate": 7.485517241379311e-06, "loss": 0.1608, "step": 4150 }, { "epoch": 2.54, "learning_rate": 7.468275862068966e-06, "loss": 0.1556, "step": 4175 }, { "epoch": 2.56, "learning_rate": 7.451034482758621e-06, "loss": 0.1642, "step": 4200 }, { "epoch": 2.57, "learning_rate": 7.433793103448277e-06, "loss": 0.1719, "step": 4225 }, { "epoch": 2.59, "learning_rate": 7.416551724137932e-06, "loss": 0.1651, "step": 4250 }, { "epoch": 2.6, "learning_rate": 7.3993103448275875e-06, "loss": 0.1557, "step": 4275 }, { "epoch": 2.62, "learning_rate": 7.3820689655172415e-06, "loss": 0.1682, "step": 4300 }, { "epoch": 2.63, "learning_rate": 7.364827586206897e-06, "loss": 0.1696, "step": 4325 }, { "epoch": 2.65, "learning_rate": 7.347586206896552e-06, "loss": 0.1621, "step": 4350 }, { "epoch": 2.66, "learning_rate": 7.330344827586208e-06, "loss": 0.1619, "step": 4375 }, { "epoch": 2.68, "learning_rate": 7.313103448275863e-06, "loss": 0.1599, "step": 4400 }, { "epoch": 2.69, "learning_rate": 7.2958620689655175e-06, "loss": 0.1624, "step": 4425 }, { "epoch": 2.71, "learning_rate": 7.278620689655172e-06, "loss": 0.1653, "step": 4450 }, { "epoch": 2.72, "learning_rate": 7.261379310344828e-06, "loss": 0.156, "step": 4475 }, { "epoch": 2.74, "learning_rate": 7.244137931034483e-06, "loss": 0.1643, "step": 4500 }, { "epoch": 2.75, "learning_rate": 7.226896551724139e-06, "loss": 0.1594, "step": 4525 }, { "epoch": 2.77, "learning_rate": 7.2096551724137944e-06, "loss": 0.1651, "step": 4550 }, { "epoch": 2.78, "learning_rate": 7.1924137931034485e-06, "loss": 0.1625, "step": 4575 }, { "epoch": 2.8, "learning_rate": 7.175172413793104e-06, "loss": 0.1678, "step": 4600 }, { "epoch": 2.81, "learning_rate": 7.157931034482759e-06, "loss": 0.1638, "step": 4625 }, { "epoch": 2.83, "learning_rate": 7.140689655172415e-06, "loss": 0.1607, "step": 4650 }, { "epoch": 2.85, "learning_rate": 7.12344827586207e-06, "loss": 0.1636, "step": 4675 }, { "epoch": 2.86, "learning_rate": 7.106206896551725e-06, "loss": 0.1604, "step": 4700 }, { "epoch": 2.88, "learning_rate": 7.088965517241379e-06, "loss": 0.1589, "step": 4725 }, { "epoch": 2.89, "learning_rate": 7.071724137931035e-06, "loss": 0.1642, "step": 4750 }, { "epoch": 2.91, "learning_rate": 7.05448275862069e-06, "loss": 0.1669, "step": 4775 }, { "epoch": 2.92, "learning_rate": 7.037241379310346e-06, "loss": 0.1662, "step": 4800 }, { "epoch": 2.94, "learning_rate": 7.0200000000000006e-06, "loss": 0.1642, "step": 4825 }, { "epoch": 2.95, "learning_rate": 7.002758620689655e-06, "loss": 0.1573, "step": 4850 }, { "epoch": 2.97, "learning_rate": 6.98551724137931e-06, "loss": 0.1599, "step": 4875 }, { "epoch": 2.98, "learning_rate": 6.968275862068966e-06, "loss": 0.1628, "step": 4900 }, { "epoch": 3.0, "learning_rate": 6.951034482758622e-06, "loss": 0.1651, "step": 4925 }, { "epoch": 3.01, "learning_rate": 6.933793103448277e-06, "loss": 0.1346, "step": 4950 }, { "epoch": 3.03, "learning_rate": 6.916551724137932e-06, "loss": 0.1202, "step": 4975 }, { "epoch": 3.04, "learning_rate": 6.899310344827586e-06, "loss": 0.1198, "step": 5000 }, { "epoch": 3.04, "eval_cer": 11.053291428375152, "eval_loss": 0.2747246026992798, "eval_runtime": 726.45, "eval_samples_per_second": 2.596, "eval_steps_per_second": 0.081, "eval_wer": 31.112082149879832, "step": 5000 }, { "epoch": 3.06, "learning_rate": 6.882068965517242e-06, "loss": 0.1149, "step": 5025 }, { "epoch": 3.07, "learning_rate": 6.864827586206897e-06, "loss": 0.125, "step": 5050 }, { "epoch": 3.09, "learning_rate": 6.847586206896553e-06, "loss": 0.1199, "step": 5075 }, { "epoch": 3.1, "learning_rate": 6.8303448275862075e-06, "loss": 0.1225, "step": 5100 }, { "epoch": 3.12, "learning_rate": 6.813103448275863e-06, "loss": 0.1244, "step": 5125 }, { "epoch": 3.13, "learning_rate": 6.795862068965517e-06, "loss": 0.1212, "step": 5150 }, { "epoch": 3.15, "learning_rate": 6.778620689655173e-06, "loss": 0.1247, "step": 5175 }, { "epoch": 3.16, "learning_rate": 6.761379310344828e-06, "loss": 0.1249, "step": 5200 }, { "epoch": 3.18, "learning_rate": 6.7441379310344836e-06, "loss": 0.1161, "step": 5225 }, { "epoch": 3.2, "learning_rate": 6.7268965517241384e-06, "loss": 0.1231, "step": 5250 }, { "epoch": 3.21, "learning_rate": 6.709655172413793e-06, "loss": 0.1228, "step": 5275 }, { "epoch": 3.23, "learning_rate": 6.692413793103448e-06, "loss": 0.1225, "step": 5300 }, { "epoch": 3.24, "learning_rate": 6.675172413793104e-06, "loss": 0.1236, "step": 5325 }, { "epoch": 3.26, "learning_rate": 6.65793103448276e-06, "loss": 0.1283, "step": 5350 }, { "epoch": 3.27, "learning_rate": 6.6406896551724145e-06, "loss": 0.1276, "step": 5375 }, { "epoch": 3.29, "learning_rate": 6.62344827586207e-06, "loss": 0.1248, "step": 5400 }, { "epoch": 3.3, "learning_rate": 6.606206896551724e-06, "loss": 0.1264, "step": 5425 }, { "epoch": 3.32, "learning_rate": 6.58896551724138e-06, "loss": 0.127, "step": 5450 }, { "epoch": 3.33, "learning_rate": 6.571724137931035e-06, "loss": 0.1176, "step": 5475 }, { "epoch": 3.35, "learning_rate": 6.5544827586206905e-06, "loss": 0.1202, "step": 5500 }, { "epoch": 3.36, "learning_rate": 6.537241379310345e-06, "loss": 0.12, "step": 5525 }, { "epoch": 3.38, "learning_rate": 6.520000000000001e-06, "loss": 0.118, "step": 5550 }, { "epoch": 3.39, "learning_rate": 6.502758620689655e-06, "loss": 0.128, "step": 5575 }, { "epoch": 3.41, "learning_rate": 6.485517241379311e-06, "loss": 0.1253, "step": 5600 }, { "epoch": 3.42, "learning_rate": 6.468275862068966e-06, "loss": 0.1309, "step": 5625 }, { "epoch": 3.44, "learning_rate": 6.4510344827586214e-06, "loss": 0.1275, "step": 5650 }, { "epoch": 3.45, "learning_rate": 6.433793103448276e-06, "loss": 0.1244, "step": 5675 }, { "epoch": 3.47, "learning_rate": 6.416551724137931e-06, "loss": 0.12, "step": 5700 }, { "epoch": 3.48, "learning_rate": 6.399310344827587e-06, "loss": 0.1309, "step": 5725 }, { "epoch": 3.5, "learning_rate": 6.382068965517242e-06, "loss": 0.1268, "step": 5750 }, { "epoch": 3.51, "learning_rate": 6.3648275862068975e-06, "loss": 0.1198, "step": 5775 }, { "epoch": 3.53, "learning_rate": 6.347586206896552e-06, "loss": 0.123, "step": 5800 }, { "epoch": 3.55, "learning_rate": 6.330344827586208e-06, "loss": 0.1246, "step": 5825 }, { "epoch": 3.56, "learning_rate": 6.313103448275862e-06, "loss": 0.1322, "step": 5850 }, { "epoch": 3.58, "learning_rate": 6.295862068965518e-06, "loss": 0.1255, "step": 5875 }, { "epoch": 3.59, "learning_rate": 6.278620689655173e-06, "loss": 0.1258, "step": 5900 }, { "epoch": 3.61, "learning_rate": 6.261379310344828e-06, "loss": 0.13, "step": 5925 }, { "epoch": 3.62, "learning_rate": 6.244137931034483e-06, "loss": 0.1237, "step": 5950 }, { "epoch": 3.64, "learning_rate": 6.226896551724139e-06, "loss": 0.1243, "step": 5975 }, { "epoch": 3.65, "learning_rate": 6.209655172413793e-06, "loss": 0.1279, "step": 6000 }, { "epoch": 3.65, "eval_cer": 10.89271300255895, "eval_loss": 0.2756609320640564, "eval_runtime": 729.9117, "eval_samples_per_second": 2.584, "eval_steps_per_second": 0.081, "eval_wer": 30.72427354162115, "step": 6000 }, { "epoch": 3.67, "learning_rate": 6.192413793103449e-06, "loss": 0.1188, "step": 6025 }, { "epoch": 3.68, "learning_rate": 6.175172413793104e-06, "loss": 0.1233, "step": 6050 }, { "epoch": 3.7, "learning_rate": 6.157931034482759e-06, "loss": 0.1255, "step": 6075 }, { "epoch": 3.71, "learning_rate": 6.140689655172414e-06, "loss": 0.1304, "step": 6100 }, { "epoch": 3.73, "learning_rate": 6.123448275862069e-06, "loss": 0.1208, "step": 6125 }, { "epoch": 3.74, "learning_rate": 6.106206896551725e-06, "loss": 0.1297, "step": 6150 }, { "epoch": 3.76, "learning_rate": 6.08896551724138e-06, "loss": 0.1193, "step": 6175 }, { "epoch": 3.77, "learning_rate": 6.071724137931035e-06, "loss": 0.1271, "step": 6200 }, { "epoch": 3.79, "learning_rate": 6.05448275862069e-06, "loss": 0.1226, "step": 6225 }, { "epoch": 3.8, "learning_rate": 6.037241379310346e-06, "loss": 0.1242, "step": 6250 }, { "epoch": 3.82, "learning_rate": 6.02e-06, "loss": 0.118, "step": 6275 }, { "epoch": 3.83, "learning_rate": 6.002758620689656e-06, "loss": 0.1221, "step": 6300 }, { "epoch": 3.85, "learning_rate": 5.9855172413793105e-06, "loss": 0.1252, "step": 6325 }, { "epoch": 3.86, "learning_rate": 5.968275862068966e-06, "loss": 0.1231, "step": 6350 }, { "epoch": 3.88, "learning_rate": 5.951034482758621e-06, "loss": 0.1201, "step": 6375 }, { "epoch": 3.9, "learning_rate": 5.933793103448277e-06, "loss": 0.1226, "step": 6400 }, { "epoch": 3.91, "learning_rate": 5.916551724137931e-06, "loss": 0.1251, "step": 6425 }, { "epoch": 3.93, "learning_rate": 5.899310344827587e-06, "loss": 0.1236, "step": 6450 }, { "epoch": 3.94, "learning_rate": 5.8820689655172415e-06, "loss": 0.1231, "step": 6475 }, { "epoch": 3.96, "learning_rate": 5.864827586206897e-06, "loss": 0.1232, "step": 6500 }, { "epoch": 3.97, "learning_rate": 5.847586206896552e-06, "loss": 0.1289, "step": 6525 }, { "epoch": 3.99, "learning_rate": 5.830344827586208e-06, "loss": 0.1274, "step": 6550 }, { "epoch": 4.0, "learning_rate": 5.813103448275863e-06, "loss": 0.1222, "step": 6575 }, { "epoch": 4.02, "learning_rate": 5.7958620689655175e-06, "loss": 0.09, "step": 6600 }, { "epoch": 4.03, "learning_rate": 5.778620689655173e-06, "loss": 0.0883, "step": 6625 }, { "epoch": 4.05, "learning_rate": 5.761379310344828e-06, "loss": 0.0921, "step": 6650 }, { "epoch": 4.06, "learning_rate": 5.744137931034484e-06, "loss": 0.0861, "step": 6675 }, { "epoch": 4.08, "learning_rate": 5.726896551724138e-06, "loss": 0.0945, "step": 6700 }, { "epoch": 4.09, "learning_rate": 5.7096551724137936e-06, "loss": 0.086, "step": 6725 }, { "epoch": 4.11, "learning_rate": 5.692413793103448e-06, "loss": 0.0904, "step": 6750 }, { "epoch": 4.12, "learning_rate": 5.675172413793104e-06, "loss": 0.0871, "step": 6775 }, { "epoch": 4.14, "learning_rate": 5.657931034482759e-06, "loss": 0.0914, "step": 6800 }, { "epoch": 4.15, "learning_rate": 5.640689655172415e-06, "loss": 0.0879, "step": 6825 }, { "epoch": 4.17, "learning_rate": 5.623448275862069e-06, "loss": 0.0928, "step": 6850 }, { "epoch": 4.18, "learning_rate": 5.6062068965517245e-06, "loss": 0.0927, "step": 6875 }, { "epoch": 4.2, "learning_rate": 5.588965517241379e-06, "loss": 0.0936, "step": 6900 }, { "epoch": 4.21, "learning_rate": 5.571724137931035e-06, "loss": 0.0925, "step": 6925 }, { "epoch": 4.23, "learning_rate": 5.554482758620691e-06, "loss": 0.0937, "step": 6950 }, { "epoch": 4.25, "learning_rate": 5.537241379310346e-06, "loss": 0.0934, "step": 6975 }, { "epoch": 4.26, "learning_rate": 5.5200000000000005e-06, "loss": 0.0891, "step": 7000 }, { "epoch": 4.26, "eval_cer": 11.061019801810158, "eval_loss": 0.28782549500465393, "eval_runtime": 728.5291, "eval_samples_per_second": 2.589, "eval_steps_per_second": 0.081, "eval_wer": 30.920908892287525, "step": 7000 }, { "epoch": 4.28, "learning_rate": 5.502758620689655e-06, "loss": 0.0914, "step": 7025 }, { "epoch": 4.29, "learning_rate": 5.485517241379311e-06, "loss": 0.0936, "step": 7050 }, { "epoch": 4.31, "learning_rate": 5.468275862068966e-06, "loss": 0.0927, "step": 7075 }, { "epoch": 4.32, "learning_rate": 5.451034482758622e-06, "loss": 0.0852, "step": 7100 }, { "epoch": 4.34, "learning_rate": 5.433793103448276e-06, "loss": 0.0879, "step": 7125 }, { "epoch": 4.35, "learning_rate": 5.4165517241379314e-06, "loss": 0.0915, "step": 7150 }, { "epoch": 4.37, "learning_rate": 5.399310344827586e-06, "loss": 0.0921, "step": 7175 }, { "epoch": 4.38, "learning_rate": 5.382068965517242e-06, "loss": 0.0908, "step": 7200 }, { "epoch": 4.4, "learning_rate": 5.364827586206897e-06, "loss": 0.0955, "step": 7225 }, { "epoch": 4.41, "learning_rate": 5.347586206896553e-06, "loss": 0.0893, "step": 7250 }, { "epoch": 4.43, "learning_rate": 5.330344827586207e-06, "loss": 0.0927, "step": 7275 }, { "epoch": 4.44, "learning_rate": 5.313103448275862e-06, "loss": 0.0909, "step": 7300 }, { "epoch": 4.46, "learning_rate": 5.295862068965517e-06, "loss": 0.0923, "step": 7325 }, { "epoch": 4.47, "learning_rate": 5.278620689655173e-06, "loss": 0.0919, "step": 7350 }, { "epoch": 4.49, "learning_rate": 5.261379310344829e-06, "loss": 0.0884, "step": 7375 }, { "epoch": 4.5, "learning_rate": 5.2441379310344835e-06, "loss": 0.0913, "step": 7400 }, { "epoch": 4.52, "learning_rate": 5.226896551724138e-06, "loss": 0.0927, "step": 7425 }, { "epoch": 4.53, "learning_rate": 5.209655172413793e-06, "loss": 0.0917, "step": 7450 }, { "epoch": 4.55, "learning_rate": 5.192413793103449e-06, "loss": 0.087, "step": 7475 }, { "epoch": 4.56, "learning_rate": 5.175172413793104e-06, "loss": 0.0931, "step": 7500 }, { "epoch": 4.58, "learning_rate": 5.1579310344827596e-06, "loss": 0.0834, "step": 7525 }, { "epoch": 4.6, "learning_rate": 5.140689655172414e-06, "loss": 0.0929, "step": 7550 }, { "epoch": 4.61, "learning_rate": 5.123448275862069e-06, "loss": 0.0922, "step": 7575 }, { "epoch": 4.63, "learning_rate": 5.106206896551724e-06, "loss": 0.0932, "step": 7600 }, { "epoch": 4.64, "learning_rate": 5.08896551724138e-06, "loss": 0.0955, "step": 7625 }, { "epoch": 4.66, "learning_rate": 5.071724137931035e-06, "loss": 0.0961, "step": 7650 }, { "epoch": 4.67, "learning_rate": 5.0544827586206905e-06, "loss": 0.0854, "step": 7675 }, { "epoch": 4.69, "learning_rate": 5.0372413793103445e-06, "loss": 0.0905, "step": 7700 }, { "epoch": 4.7, "learning_rate": 5.02e-06, "loss": 0.0936, "step": 7725 }, { "epoch": 4.72, "learning_rate": 5.002758620689656e-06, "loss": 0.0879, "step": 7750 }, { "epoch": 4.73, "learning_rate": 4.985517241379311e-06, "loss": 0.0951, "step": 7775 }, { "epoch": 4.75, "learning_rate": 4.968275862068966e-06, "loss": 0.0951, "step": 7800 }, { "epoch": 4.76, "learning_rate": 4.951034482758621e-06, "loss": 0.0948, "step": 7825 }, { "epoch": 4.78, "learning_rate": 4.933793103448276e-06, "loss": 0.0862, "step": 7850 }, { "epoch": 4.79, "learning_rate": 4.916551724137931e-06, "loss": 0.0986, "step": 7875 }, { "epoch": 4.81, "learning_rate": 4.899310344827586e-06, "loss": 0.0937, "step": 7900 }, { "epoch": 4.82, "learning_rate": 4.882068965517242e-06, "loss": 0.0935, "step": 7925 }, { "epoch": 4.84, "learning_rate": 4.8648275862068974e-06, "loss": 0.0904, "step": 7950 }, { "epoch": 4.85, "learning_rate": 4.847586206896552e-06, "loss": 0.0952, "step": 7975 }, { "epoch": 4.87, "learning_rate": 4.830344827586207e-06, "loss": 0.0899, "step": 8000 }, { "epoch": 4.87, "eval_cer": 11.079911381317945, "eval_loss": 0.29063600301742554, "eval_runtime": 724.5205, "eval_samples_per_second": 2.603, "eval_steps_per_second": 0.081, "eval_wer": 30.66419051780642, "step": 8000 }, { "epoch": 4.88, "learning_rate": 4.813103448275863e-06, "loss": 0.0983, "step": 8025 }, { "epoch": 4.9, "learning_rate": 4.795862068965518e-06, "loss": 0.0897, "step": 8050 }, { "epoch": 4.91, "learning_rate": 4.778620689655173e-06, "loss": 0.0875, "step": 8075 }, { "epoch": 4.93, "learning_rate": 4.761379310344828e-06, "loss": 0.0944, "step": 8100 }, { "epoch": 4.95, "learning_rate": 4.744137931034483e-06, "loss": 0.0922, "step": 8125 }, { "epoch": 4.96, "learning_rate": 4.726896551724138e-06, "loss": 0.0923, "step": 8150 }, { "epoch": 4.98, "learning_rate": 4.709655172413794e-06, "loss": 0.089, "step": 8175 }, { "epoch": 4.99, "learning_rate": 4.692413793103449e-06, "loss": 0.0933, "step": 8200 }, { "epoch": 5.01, "learning_rate": 4.6751724137931035e-06, "loss": 0.0804, "step": 8225 }, { "epoch": 5.02, "learning_rate": 4.657931034482759e-06, "loss": 0.0646, "step": 8250 }, { "epoch": 5.04, "learning_rate": 4.640689655172414e-06, "loss": 0.0602, "step": 8275 }, { "epoch": 5.05, "learning_rate": 4.623448275862069e-06, "loss": 0.0639, "step": 8300 }, { "epoch": 5.07, "learning_rate": 4.606206896551724e-06, "loss": 0.0656, "step": 8325 }, { "epoch": 5.08, "learning_rate": 4.58896551724138e-06, "loss": 0.0607, "step": 8350 }, { "epoch": 5.1, "learning_rate": 4.571724137931035e-06, "loss": 0.0622, "step": 8375 }, { "epoch": 5.11, "learning_rate": 4.55448275862069e-06, "loss": 0.0632, "step": 8400 }, { "epoch": 5.13, "learning_rate": 4.537241379310345e-06, "loss": 0.0654, "step": 8425 }, { "epoch": 5.14, "learning_rate": 4.520000000000001e-06, "loss": 0.0614, "step": 8450 }, { "epoch": 5.16, "learning_rate": 4.502758620689656e-06, "loss": 0.0645, "step": 8475 }, { "epoch": 5.17, "learning_rate": 4.4855172413793105e-06, "loss": 0.0614, "step": 8500 }, { "epoch": 5.19, "learning_rate": 4.468275862068966e-06, "loss": 0.0616, "step": 8525 }, { "epoch": 5.2, "learning_rate": 4.451034482758621e-06, "loss": 0.0621, "step": 8550 }, { "epoch": 5.22, "learning_rate": 4.433793103448276e-06, "loss": 0.0649, "step": 8575 }, { "epoch": 5.23, "learning_rate": 4.416551724137932e-06, "loss": 0.0637, "step": 8600 }, { "epoch": 5.25, "learning_rate": 4.3993103448275866e-06, "loss": 0.0652, "step": 8625 }, { "epoch": 5.26, "learning_rate": 4.3820689655172414e-06, "loss": 0.0665, "step": 8650 }, { "epoch": 5.28, "learning_rate": 4.364827586206897e-06, "loss": 0.0627, "step": 8675 }, { "epoch": 5.3, "learning_rate": 4.347586206896552e-06, "loss": 0.0634, "step": 8700 }, { "epoch": 5.31, "learning_rate": 4.330344827586207e-06, "loss": 0.0633, "step": 8725 }, { "epoch": 5.33, "learning_rate": 4.313103448275863e-06, "loss": 0.0629, "step": 8750 }, { "epoch": 5.34, "learning_rate": 4.2958620689655175e-06, "loss": 0.0634, "step": 8775 }, { "epoch": 5.36, "learning_rate": 4.278620689655173e-06, "loss": 0.0626, "step": 8800 }, { "epoch": 5.37, "learning_rate": 4.261379310344828e-06, "loss": 0.0706, "step": 8825 }, { "epoch": 5.39, "learning_rate": 4.244137931034483e-06, "loss": 0.0637, "step": 8850 }, { "epoch": 5.4, "learning_rate": 4.226896551724139e-06, "loss": 0.0643, "step": 8875 }, { "epoch": 5.42, "learning_rate": 4.2096551724137935e-06, "loss": 0.066, "step": 8900 }, { "epoch": 5.43, "learning_rate": 4.192413793103448e-06, "loss": 0.0608, "step": 8925 }, { "epoch": 5.45, "learning_rate": 4.175172413793104e-06, "loss": 0.0651, "step": 8950 }, { "epoch": 5.46, "learning_rate": 4.157931034482759e-06, "loss": 0.0674, "step": 8975 }, { "epoch": 5.48, "learning_rate": 4.140689655172414e-06, "loss": 0.0648, "step": 9000 }, { "epoch": 5.48, "eval_cer": 10.90301750047229, "eval_loss": 0.30544599890708923, "eval_runtime": 724.481, "eval_samples_per_second": 2.603, "eval_steps_per_second": 0.081, "eval_wer": 30.59864540091763, "step": 9000 }, { "epoch": 5.49, "learning_rate": 4.1234482758620696e-06, "loss": 0.0642, "step": 9025 }, { "epoch": 5.51, "learning_rate": 4.1062068965517244e-06, "loss": 0.0686, "step": 9050 }, { "epoch": 5.52, "learning_rate": 4.088965517241379e-06, "loss": 0.064, "step": 9075 }, { "epoch": 5.54, "learning_rate": 4.071724137931035e-06, "loss": 0.0631, "step": 9100 }, { "epoch": 5.55, "learning_rate": 4.05448275862069e-06, "loss": 0.0678, "step": 9125 }, { "epoch": 5.57, "learning_rate": 4.037241379310346e-06, "loss": 0.0638, "step": 9150 }, { "epoch": 5.58, "learning_rate": 4.0200000000000005e-06, "loss": 0.0626, "step": 9175 }, { "epoch": 5.6, "learning_rate": 4.002758620689655e-06, "loss": 0.066, "step": 9200 }, { "epoch": 5.61, "learning_rate": 3.985517241379311e-06, "loss": 0.0697, "step": 9225 }, { "epoch": 5.63, "learning_rate": 3.968275862068966e-06, "loss": 0.0676, "step": 9250 }, { "epoch": 5.65, "learning_rate": 3.951034482758621e-06, "loss": 0.0661, "step": 9275 }, { "epoch": 5.66, "learning_rate": 3.9337931034482765e-06, "loss": 0.0595, "step": 9300 }, { "epoch": 5.68, "learning_rate": 3.916551724137931e-06, "loss": 0.0643, "step": 9325 }, { "epoch": 5.69, "learning_rate": 3.899310344827586e-06, "loss": 0.067, "step": 9350 }, { "epoch": 5.71, "learning_rate": 3.882068965517242e-06, "loss": 0.0653, "step": 9375 }, { "epoch": 5.72, "learning_rate": 3.864827586206897e-06, "loss": 0.0666, "step": 9400 }, { "epoch": 5.74, "learning_rate": 3.847586206896552e-06, "loss": 0.0661, "step": 9425 }, { "epoch": 5.75, "learning_rate": 3.8303448275862074e-06, "loss": 0.0674, "step": 9450 }, { "epoch": 5.77, "learning_rate": 3.8131034482758623e-06, "loss": 0.0648, "step": 9475 }, { "epoch": 5.78, "learning_rate": 3.795862068965517e-06, "loss": 0.0661, "step": 9500 }, { "epoch": 5.8, "learning_rate": 3.7786206896551725e-06, "loss": 0.0691, "step": 9525 }, { "epoch": 5.81, "learning_rate": 3.7613793103448278e-06, "loss": 0.0648, "step": 9550 }, { "epoch": 5.83, "learning_rate": 3.7441379310344835e-06, "loss": 0.0621, "step": 9575 }, { "epoch": 5.84, "learning_rate": 3.7268965517241383e-06, "loss": 0.0668, "step": 9600 }, { "epoch": 5.86, "learning_rate": 3.7096551724137936e-06, "loss": 0.0684, "step": 9625 }, { "epoch": 5.87, "learning_rate": 3.6924137931034485e-06, "loss": 0.0695, "step": 9650 }, { "epoch": 5.89, "learning_rate": 3.675172413793104e-06, "loss": 0.0619, "step": 9675 }, { "epoch": 5.9, "learning_rate": 3.657931034482759e-06, "loss": 0.0649, "step": 9700 }, { "epoch": 5.92, "learning_rate": 3.640689655172414e-06, "loss": 0.0652, "step": 9725 }, { "epoch": 5.93, "learning_rate": 3.6234482758620693e-06, "loss": 0.0652, "step": 9750 }, { "epoch": 5.95, "learning_rate": 3.6062068965517246e-06, "loss": 0.0672, "step": 9775 }, { "epoch": 5.96, "learning_rate": 3.5889655172413794e-06, "loss": 0.0654, "step": 9800 }, { "epoch": 5.98, "learning_rate": 3.5717241379310347e-06, "loss": 0.0689, "step": 9825 }, { "epoch": 6.0, "learning_rate": 3.55448275862069e-06, "loss": 0.0665, "step": 9850 }, { "epoch": 6.01, "learning_rate": 3.537241379310345e-06, "loss": 0.0526, "step": 9875 }, { "epoch": 6.03, "learning_rate": 3.52e-06, "loss": 0.044, "step": 9900 }, { "epoch": 6.04, "learning_rate": 3.502758620689655e-06, "loss": 0.043, "step": 9925 }, { "epoch": 6.06, "learning_rate": 3.4855172413793103e-06, "loss": 0.042, "step": 9950 }, { "epoch": 6.07, "learning_rate": 3.468275862068966e-06, "loss": 0.0427, "step": 9975 }, { "epoch": 6.09, "learning_rate": 3.4510344827586214e-06, "loss": 0.0436, "step": 10000 }, { "epoch": 6.09, "eval_cer": 10.943376783966201, "eval_loss": 0.3183899223804474, "eval_runtime": 722.5761, "eval_samples_per_second": 2.61, "eval_steps_per_second": 0.082, "eval_wer": 30.52217609788071, "step": 10000 }, { "epoch": 6.1, "learning_rate": 3.4337931034482762e-06, "loss": 0.0402, "step": 10025 }, { "epoch": 6.12, "learning_rate": 3.4172413793103448e-06, "loss": 0.0451, "step": 10050 }, { "epoch": 6.13, "learning_rate": 3.4000000000000005e-06, "loss": 0.0443, "step": 10075 }, { "epoch": 6.15, "learning_rate": 3.3827586206896558e-06, "loss": 0.0439, "step": 10100 }, { "epoch": 6.16, "learning_rate": 3.3655172413793106e-06, "loss": 0.0437, "step": 10125 }, { "epoch": 6.18, "learning_rate": 3.348275862068966e-06, "loss": 0.0481, "step": 10150 }, { "epoch": 6.19, "learning_rate": 3.3310344827586212e-06, "loss": 0.0454, "step": 10175 }, { "epoch": 6.21, "learning_rate": 3.313793103448276e-06, "loss": 0.0446, "step": 10200 }, { "epoch": 6.22, "learning_rate": 3.2965517241379314e-06, "loss": 0.046, "step": 10225 }, { "epoch": 6.24, "learning_rate": 3.2793103448275863e-06, "loss": 0.0428, "step": 10250 }, { "epoch": 6.25, "learning_rate": 3.2620689655172416e-06, "loss": 0.0422, "step": 10275 }, { "epoch": 6.27, "learning_rate": 3.244827586206897e-06, "loss": 0.0406, "step": 10300 }, { "epoch": 6.28, "learning_rate": 3.2275862068965517e-06, "loss": 0.0409, "step": 10325 }, { "epoch": 6.3, "learning_rate": 3.210344827586207e-06, "loss": 0.0472, "step": 10350 }, { "epoch": 6.31, "learning_rate": 3.1931034482758623e-06, "loss": 0.0432, "step": 10375 }, { "epoch": 6.33, "learning_rate": 3.175862068965517e-06, "loss": 0.0449, "step": 10400 }, { "epoch": 6.35, "learning_rate": 3.1586206896551725e-06, "loss": 0.0464, "step": 10425 }, { "epoch": 6.36, "learning_rate": 3.1413793103448278e-06, "loss": 0.0435, "step": 10450 }, { "epoch": 6.38, "learning_rate": 3.124137931034483e-06, "loss": 0.0459, "step": 10475 }, { "epoch": 6.39, "learning_rate": 3.1068965517241384e-06, "loss": 0.0447, "step": 10500 }, { "epoch": 6.41, "learning_rate": 3.0896551724137937e-06, "loss": 0.0451, "step": 10525 }, { "epoch": 6.42, "learning_rate": 3.0724137931034485e-06, "loss": 0.044, "step": 10550 }, { "epoch": 6.44, "learning_rate": 3.055172413793104e-06, "loss": 0.0452, "step": 10575 }, { "epoch": 6.45, "learning_rate": 3.037931034482759e-06, "loss": 0.0446, "step": 10600 }, { "epoch": 6.47, "learning_rate": 3.020689655172414e-06, "loss": 0.043, "step": 10625 }, { "epoch": 6.48, "learning_rate": 3.0034482758620693e-06, "loss": 0.0448, "step": 10650 }, { "epoch": 6.5, "learning_rate": 2.9862068965517246e-06, "loss": 0.0433, "step": 10675 }, { "epoch": 6.51, "learning_rate": 2.9689655172413794e-06, "loss": 0.0449, "step": 10700 }, { "epoch": 6.53, "learning_rate": 2.9517241379310347e-06, "loss": 0.0437, "step": 10725 }, { "epoch": 6.54, "learning_rate": 2.9344827586206896e-06, "loss": 0.0469, "step": 10750 }, { "epoch": 6.56, "learning_rate": 2.917241379310345e-06, "loss": 0.0458, "step": 10775 }, { "epoch": 6.57, "learning_rate": 2.9e-06, "loss": 0.0441, "step": 10800 }, { "epoch": 6.59, "learning_rate": 2.882758620689655e-06, "loss": 0.0439, "step": 10825 }, { "epoch": 6.6, "learning_rate": 2.8655172413793104e-06, "loss": 0.0437, "step": 10850 }, { "epoch": 6.62, "learning_rate": 2.848275862068966e-06, "loss": 0.0454, "step": 10875 }, { "epoch": 6.63, "learning_rate": 2.831034482758621e-06, "loss": 0.0436, "step": 10900 }, { "epoch": 6.65, "learning_rate": 2.8137931034482762e-06, "loss": 0.0446, "step": 10925 }, { "epoch": 6.66, "learning_rate": 2.7965517241379315e-06, "loss": 0.046, "step": 10950 }, { "epoch": 6.68, "learning_rate": 2.7793103448275864e-06, "loss": 0.0428, "step": 10975 }, { "epoch": 6.7, "learning_rate": 2.7620689655172417e-06, "loss": 0.0468, "step": 11000 }, { "epoch": 6.7, "eval_cer": 10.90301750047229, "eval_loss": 0.32771989703178406, "eval_runtime": 722.7705, "eval_samples_per_second": 2.609, "eval_steps_per_second": 0.082, "eval_wer": 30.36923749180686, "step": 11000 }, { "epoch": 6.71, "learning_rate": 2.744827586206897e-06, "loss": 0.0439, "step": 11025 }, { "epoch": 6.73, "learning_rate": 2.727586206896552e-06, "loss": 0.0432, "step": 11050 }, { "epoch": 6.74, "learning_rate": 2.710344827586207e-06, "loss": 0.042, "step": 11075 }, { "epoch": 6.76, "learning_rate": 2.6931034482758624e-06, "loss": 0.0464, "step": 11100 }, { "epoch": 6.77, "learning_rate": 2.6758620689655173e-06, "loss": 0.0446, "step": 11125 }, { "epoch": 6.79, "learning_rate": 2.6586206896551726e-06, "loss": 0.0444, "step": 11150 }, { "epoch": 6.8, "learning_rate": 2.6413793103448275e-06, "loss": 0.044, "step": 11175 }, { "epoch": 6.82, "learning_rate": 2.6241379310344828e-06, "loss": 0.0443, "step": 11200 }, { "epoch": 6.83, "learning_rate": 2.606896551724138e-06, "loss": 0.0451, "step": 11225 }, { "epoch": 6.85, "learning_rate": 2.589655172413793e-06, "loss": 0.0455, "step": 11250 }, { "epoch": 6.86, "learning_rate": 2.5724137931034486e-06, "loss": 0.0437, "step": 11275 }, { "epoch": 6.88, "learning_rate": 2.555172413793104e-06, "loss": 0.0421, "step": 11300 }, { "epoch": 6.89, "learning_rate": 2.537931034482759e-06, "loss": 0.0439, "step": 11325 }, { "epoch": 6.91, "learning_rate": 2.520689655172414e-06, "loss": 0.0396, "step": 11350 }, { "epoch": 6.92, "learning_rate": 2.5034482758620694e-06, "loss": 0.0464, "step": 11375 }, { "epoch": 6.94, "learning_rate": 2.4862068965517243e-06, "loss": 0.0421, "step": 11400 }, { "epoch": 6.95, "learning_rate": 2.4689655172413796e-06, "loss": 0.0417, "step": 11425 }, { "epoch": 6.97, "learning_rate": 2.451724137931035e-06, "loss": 0.046, "step": 11450 }, { "epoch": 6.98, "learning_rate": 2.4344827586206897e-06, "loss": 0.0444, "step": 11475 }, { "epoch": 7.0, "learning_rate": 2.417241379310345e-06, "loss": 0.0444, "step": 11500 }, { "epoch": 7.01, "learning_rate": 2.4000000000000003e-06, "loss": 0.0331, "step": 11525 }, { "epoch": 7.03, "learning_rate": 2.382758620689655e-06, "loss": 0.0275, "step": 11550 }, { "epoch": 7.05, "learning_rate": 2.3655172413793105e-06, "loss": 0.0315, "step": 11575 }, { "epoch": 7.06, "learning_rate": 2.3482758620689658e-06, "loss": 0.0307, "step": 11600 }, { "epoch": 7.08, "learning_rate": 2.331034482758621e-06, "loss": 0.0289, "step": 11625 }, { "epoch": 7.09, "learning_rate": 2.313793103448276e-06, "loss": 0.0288, "step": 11650 }, { "epoch": 7.11, "learning_rate": 2.2965517241379312e-06, "loss": 0.0304, "step": 11675 }, { "epoch": 7.12, "learning_rate": 2.2793103448275865e-06, "loss": 0.0311, "step": 11700 }, { "epoch": 7.14, "learning_rate": 2.2620689655172414e-06, "loss": 0.0274, "step": 11725 }, { "epoch": 7.15, "learning_rate": 2.2448275862068967e-06, "loss": 0.0286, "step": 11750 }, { "epoch": 7.17, "learning_rate": 2.227586206896552e-06, "loss": 0.0293, "step": 11775 }, { "epoch": 7.18, "learning_rate": 2.2103448275862073e-06, "loss": 0.028, "step": 11800 }, { "epoch": 7.2, "learning_rate": 2.193103448275862e-06, "loss": 0.0288, "step": 11825 }, { "epoch": 7.21, "learning_rate": 2.1758620689655174e-06, "loss": 0.0301, "step": 11850 }, { "epoch": 7.23, "learning_rate": 2.1586206896551727e-06, "loss": 0.0295, "step": 11875 }, { "epoch": 7.24, "learning_rate": 2.1413793103448276e-06, "loss": 0.0298, "step": 11900 }, { "epoch": 7.26, "learning_rate": 2.124137931034483e-06, "loss": 0.0269, "step": 11925 }, { "epoch": 7.27, "learning_rate": 2.106896551724138e-06, "loss": 0.0297, "step": 11950 }, { "epoch": 7.29, "learning_rate": 2.0896551724137935e-06, "loss": 0.0278, "step": 11975 }, { "epoch": 7.3, "learning_rate": 2.0724137931034484e-06, "loss": 0.0291, "step": 12000 }, { "epoch": 7.3, "eval_cer": 11.157195115667989, "eval_loss": 0.34108513593673706, "eval_runtime": 719.6449, "eval_samples_per_second": 2.621, "eval_steps_per_second": 0.082, "eval_wer": 30.98099191610225, "step": 12000 }, { "epoch": 7.32, "learning_rate": 2.0551724137931036e-06, "loss": 0.0276, "step": 12025 }, { "epoch": 7.33, "learning_rate": 2.037931034482759e-06, "loss": 0.0291, "step": 12050 }, { "epoch": 7.35, "learning_rate": 2.020689655172414e-06, "loss": 0.0275, "step": 12075 }, { "epoch": 7.36, "learning_rate": 2.003448275862069e-06, "loss": 0.0312, "step": 12100 }, { "epoch": 7.38, "learning_rate": 1.9862068965517244e-06, "loss": 0.0261, "step": 12125 }, { "epoch": 7.4, "learning_rate": 1.9689655172413793e-06, "loss": 0.0295, "step": 12150 }, { "epoch": 7.41, "learning_rate": 1.9517241379310346e-06, "loss": 0.0276, "step": 12175 }, { "epoch": 7.43, "learning_rate": 1.93448275862069e-06, "loss": 0.0328, "step": 12200 }, { "epoch": 7.44, "learning_rate": 1.917241379310345e-06, "loss": 0.0286, "step": 12225 }, { "epoch": 7.46, "learning_rate": 1.9000000000000002e-06, "loss": 0.0292, "step": 12250 }, { "epoch": 7.47, "learning_rate": 1.8827586206896553e-06, "loss": 0.0314, "step": 12275 }, { "epoch": 7.49, "learning_rate": 1.8655172413793104e-06, "loss": 0.031, "step": 12300 }, { "epoch": 7.5, "learning_rate": 1.8482758620689657e-06, "loss": 0.0273, "step": 12325 }, { "epoch": 7.52, "learning_rate": 1.8310344827586208e-06, "loss": 0.0294, "step": 12350 }, { "epoch": 7.53, "learning_rate": 1.813793103448276e-06, "loss": 0.03, "step": 12375 }, { "epoch": 7.55, "learning_rate": 1.7965517241379314e-06, "loss": 0.0292, "step": 12400 }, { "epoch": 7.56, "learning_rate": 1.7793103448275864e-06, "loss": 0.0276, "step": 12425 }, { "epoch": 7.58, "learning_rate": 1.7620689655172415e-06, "loss": 0.03, "step": 12450 }, { "epoch": 7.59, "learning_rate": 1.7448275862068966e-06, "loss": 0.0285, "step": 12475 }, { "epoch": 7.61, "learning_rate": 1.727586206896552e-06, "loss": 0.0297, "step": 12500 }, { "epoch": 7.62, "learning_rate": 1.710344827586207e-06, "loss": 0.0282, "step": 12525 }, { "epoch": 7.64, "learning_rate": 1.693103448275862e-06, "loss": 0.0313, "step": 12550 }, { "epoch": 7.65, "learning_rate": 1.6758620689655176e-06, "loss": 0.0289, "step": 12575 }, { "epoch": 7.67, "learning_rate": 1.6586206896551726e-06, "loss": 0.0292, "step": 12600 }, { "epoch": 7.68, "learning_rate": 1.6420689655172414e-06, "loss": 0.0311, "step": 12625 }, { "epoch": 7.7, "learning_rate": 1.6248275862068965e-06, "loss": 0.0302, "step": 12650 }, { "epoch": 7.71, "learning_rate": 1.607586206896552e-06, "loss": 0.0265, "step": 12675 }, { "epoch": 7.73, "learning_rate": 1.590344827586207e-06, "loss": 0.0295, "step": 12700 }, { "epoch": 7.74, "learning_rate": 1.5731034482758622e-06, "loss": 0.0293, "step": 12725 }, { "epoch": 7.76, "learning_rate": 1.5558620689655174e-06, "loss": 0.0292, "step": 12750 }, { "epoch": 7.78, "learning_rate": 1.5386206896551725e-06, "loss": 0.0295, "step": 12775 }, { "epoch": 7.79, "learning_rate": 1.5213793103448276e-06, "loss": 0.0287, "step": 12800 }, { "epoch": 7.81, "learning_rate": 1.504137931034483e-06, "loss": 0.0288, "step": 12825 }, { "epoch": 7.82, "learning_rate": 1.486896551724138e-06, "loss": 0.03, "step": 12850 }, { "epoch": 7.84, "learning_rate": 1.4696551724137933e-06, "loss": 0.0298, "step": 12875 }, { "epoch": 7.85, "learning_rate": 1.4524137931034486e-06, "loss": 0.0307, "step": 12900 }, { "epoch": 7.87, "learning_rate": 1.4351724137931037e-06, "loss": 0.0295, "step": 12925 }, { "epoch": 7.88, "learning_rate": 1.4179310344827587e-06, "loss": 0.0276, "step": 12950 }, { "epoch": 7.9, "learning_rate": 1.4006896551724138e-06, "loss": 0.0289, "step": 12975 }, { "epoch": 7.91, "learning_rate": 1.3834482758620691e-06, "loss": 0.0275, "step": 13000 }, { "epoch": 7.91, "eval_cer": 11.155477699349099, "eval_loss": 0.3475850820541382, "eval_runtime": 720.0873, "eval_samples_per_second": 2.619, "eval_steps_per_second": 0.082, "eval_wer": 31.06838540528731, "step": 13000 }, { "epoch": 7.93, "learning_rate": 1.3662068965517242e-06, "loss": 0.0284, "step": 13025 }, { "epoch": 7.94, "learning_rate": 1.3489655172413793e-06, "loss": 0.0305, "step": 13050 }, { "epoch": 7.96, "learning_rate": 1.3317241379310348e-06, "loss": 0.0281, "step": 13075 }, { "epoch": 7.97, "learning_rate": 1.3144827586206899e-06, "loss": 0.0307, "step": 13100 }, { "epoch": 7.99, "learning_rate": 1.297241379310345e-06, "loss": 0.0302, "step": 13125 }, { "epoch": 8.0, "learning_rate": 1.28e-06, "loss": 0.0292, "step": 13150 }, { "epoch": 8.02, "learning_rate": 1.2627586206896553e-06, "loss": 0.0193, "step": 13175 }, { "epoch": 8.03, "learning_rate": 1.2455172413793104e-06, "loss": 0.0218, "step": 13200 }, { "epoch": 8.05, "learning_rate": 1.2282758620689657e-06, "loss": 0.0212, "step": 13225 }, { "epoch": 8.06, "learning_rate": 1.2110344827586208e-06, "loss": 0.0212, "step": 13250 }, { "epoch": 8.08, "learning_rate": 1.1937931034482759e-06, "loss": 0.0199, "step": 13275 }, { "epoch": 8.09, "learning_rate": 1.1765517241379312e-06, "loss": 0.0198, "step": 13300 }, { "epoch": 8.11, "learning_rate": 1.1593103448275864e-06, "loss": 0.0194, "step": 13325 }, { "epoch": 8.13, "learning_rate": 1.1420689655172415e-06, "loss": 0.0203, "step": 13350 }, { "epoch": 8.14, "learning_rate": 1.1248275862068966e-06, "loss": 0.0199, "step": 13375 }, { "epoch": 8.16, "learning_rate": 1.1075862068965517e-06, "loss": 0.0197, "step": 13400 }, { "epoch": 8.17, "learning_rate": 1.090344827586207e-06, "loss": 0.0197, "step": 13425 }, { "epoch": 8.19, "learning_rate": 1.0731034482758623e-06, "loss": 0.02, "step": 13450 }, { "epoch": 8.2, "learning_rate": 1.056551724137931e-06, "loss": 0.0214, "step": 13475 }, { "epoch": 8.22, "learning_rate": 1.0393103448275863e-06, "loss": 0.0209, "step": 13500 }, { "epoch": 8.23, "learning_rate": 1.0220689655172416e-06, "loss": 0.0192, "step": 13525 }, { "epoch": 8.25, "learning_rate": 1.0048275862068967e-06, "loss": 0.0184, "step": 13550 }, { "epoch": 8.26, "learning_rate": 9.875862068965518e-07, "loss": 0.0207, "step": 13575 }, { "epoch": 8.28, "learning_rate": 9.703448275862069e-07, "loss": 0.0198, "step": 13600 }, { "epoch": 8.29, "learning_rate": 9.531034482758622e-07, "loss": 0.0198, "step": 13625 }, { "epoch": 8.31, "learning_rate": 9.358620689655173e-07, "loss": 0.02, "step": 13650 }, { "epoch": 8.32, "learning_rate": 9.186206896551725e-07, "loss": 0.021, "step": 13675 }, { "epoch": 8.34, "learning_rate": 9.013793103448276e-07, "loss": 0.0193, "step": 13700 }, { "epoch": 8.35, "learning_rate": 8.841379310344829e-07, "loss": 0.021, "step": 13725 }, { "epoch": 8.37, "learning_rate": 8.66896551724138e-07, "loss": 0.0192, "step": 13750 }, { "epoch": 8.38, "learning_rate": 8.496551724137932e-07, "loss": 0.0187, "step": 13775 }, { "epoch": 8.4, "learning_rate": 8.324137931034483e-07, "loss": 0.0204, "step": 13800 }, { "epoch": 8.41, "learning_rate": 8.151724137931036e-07, "loss": 0.0209, "step": 13825 }, { "epoch": 8.43, "learning_rate": 7.979310344827587e-07, "loss": 0.0194, "step": 13850 }, { "epoch": 8.44, "learning_rate": 7.806896551724138e-07, "loss": 0.0186, "step": 13875 }, { "epoch": 8.46, "learning_rate": 7.63448275862069e-07, "loss": 0.021, "step": 13900 }, { "epoch": 8.48, "learning_rate": 7.462068965517243e-07, "loss": 0.0195, "step": 13925 }, { "epoch": 8.49, "learning_rate": 7.289655172413794e-07, "loss": 0.021, "step": 13950 }, { "epoch": 8.51, "learning_rate": 7.117241379310346e-07, "loss": 0.0199, "step": 13975 }, { "epoch": 8.52, "learning_rate": 6.944827586206897e-07, "loss": 0.0196, "step": 14000 }, { "epoch": 8.52, "eval_cer": 11.106531334260739, "eval_loss": 0.35724931955337524, "eval_runtime": 721.1236, "eval_samples_per_second": 2.615, "eval_steps_per_second": 0.082, "eval_wer": 30.91544679921346, "step": 14000 }, { "epoch": 8.54, "learning_rate": 6.77241379310345e-07, "loss": 0.0218, "step": 14025 }, { "epoch": 8.55, "learning_rate": 6.6e-07, "loss": 0.0202, "step": 14050 }, { "epoch": 8.57, "learning_rate": 6.427586206896552e-07, "loss": 0.0187, "step": 14075 }, { "epoch": 8.58, "learning_rate": 6.255172413793104e-07, "loss": 0.0186, "step": 14100 }, { "epoch": 8.6, "learning_rate": 6.082758620689656e-07, "loss": 0.0197, "step": 14125 }, { "epoch": 8.61, "learning_rate": 5.910344827586208e-07, "loss": 0.0192, "step": 14150 }, { "epoch": 8.63, "learning_rate": 5.737931034482759e-07, "loss": 0.0184, "step": 14175 }, { "epoch": 8.64, "learning_rate": 5.565517241379312e-07, "loss": 0.0194, "step": 14200 }, { "epoch": 8.66, "learning_rate": 5.393103448275862e-07, "loss": 0.0195, "step": 14225 }, { "epoch": 8.67, "learning_rate": 5.220689655172414e-07, "loss": 0.021, "step": 14250 }, { "epoch": 8.69, "learning_rate": 5.048275862068966e-07, "loss": 0.0203, "step": 14275 }, { "epoch": 8.7, "learning_rate": 4.875862068965518e-07, "loss": 0.0199, "step": 14300 }, { "epoch": 8.72, "learning_rate": 4.7034482758620694e-07, "loss": 0.0189, "step": 14325 }, { "epoch": 8.73, "learning_rate": 4.5310344827586213e-07, "loss": 0.0195, "step": 14350 }, { "epoch": 8.75, "learning_rate": 4.3586206896551726e-07, "loss": 0.0182, "step": 14375 }, { "epoch": 8.76, "learning_rate": 4.186206896551725e-07, "loss": 0.021, "step": 14400 }, { "epoch": 8.78, "learning_rate": 4.0137931034482764e-07, "loss": 0.02, "step": 14425 }, { "epoch": 8.79, "learning_rate": 3.841379310344828e-07, "loss": 0.0186, "step": 14450 }, { "epoch": 8.81, "learning_rate": 3.6689655172413796e-07, "loss": 0.0206, "step": 14475 }, { "epoch": 8.83, "learning_rate": 3.496551724137931e-07, "loss": 0.0195, "step": 14500 }, { "epoch": 8.84, "learning_rate": 3.324137931034483e-07, "loss": 0.0174, "step": 14525 }, { "epoch": 8.86, "learning_rate": 3.151724137931034e-07, "loss": 0.0199, "step": 14550 }, { "epoch": 8.87, "learning_rate": 2.9793103448275866e-07, "loss": 0.0202, "step": 14575 }, { "epoch": 8.89, "learning_rate": 2.806896551724138e-07, "loss": 0.0222, "step": 14600 }, { "epoch": 8.9, "learning_rate": 2.63448275862069e-07, "loss": 0.0189, "step": 14625 }, { "epoch": 8.92, "learning_rate": 2.4620689655172417e-07, "loss": 0.019, "step": 14650 }, { "epoch": 8.93, "learning_rate": 2.2896551724137933e-07, "loss": 0.0197, "step": 14675 }, { "epoch": 8.95, "learning_rate": 2.117241379310345e-07, "loss": 0.0177, "step": 14700 }, { "epoch": 8.96, "learning_rate": 1.9448275862068968e-07, "loss": 0.0199, "step": 14725 }, { "epoch": 8.98, "learning_rate": 1.7724137931034484e-07, "loss": 0.02, "step": 14750 }, { "epoch": 8.99, "learning_rate": 1.6e-07, "loss": 0.02, "step": 14775 }, { "epoch": 9.01, "learning_rate": 1.427586206896552e-07, "loss": 0.0165, "step": 14800 }, { "epoch": 9.02, "learning_rate": 1.2551724137931035e-07, "loss": 0.0167, "step": 14825 }, { "epoch": 9.04, "learning_rate": 1.0827586206896553e-07, "loss": 0.0159, "step": 14850 }, { "epoch": 9.05, "learning_rate": 9.10344827586207e-08, "loss": 0.0157, "step": 14875 }, { "epoch": 9.07, "learning_rate": 7.379310344827586e-08, "loss": 0.0162, "step": 14900 }, { "epoch": 9.08, "learning_rate": 5.655172413793104e-08, "loss": 0.0174, "step": 14925 }, { "epoch": 9.1, "learning_rate": 3.9310344827586205e-08, "loss": 0.0156, "step": 14950 }, { "epoch": 9.11, "learning_rate": 2.2068965517241383e-08, "loss": 0.0161, "step": 14975 }, { "epoch": 9.13, "learning_rate": 4.827586206896552e-09, "loss": 0.0159, "step": 15000 }, { "epoch": 9.13, "eval_cer": 11.208717605234684, "eval_loss": 0.36000844836235046, "eval_runtime": 718.6482, "eval_samples_per_second": 2.624, "eval_steps_per_second": 0.082, "eval_wer": 31.03561284684291, "step": 15000 }, { "epoch": 9.13, "step": 15000, "total_flos": 1.3847683858735104e+20, "train_loss": 0.16191740266680718, "train_runtime": 44818.0374, "train_samples_per_second": 10.71, "train_steps_per_second": 0.335 }, { "epoch": 9.13, "eval_cer": 10.90301750047229, "eval_loss": 0.32771989703178406, "eval_runtime": 721.8667, "eval_samples_per_second": 2.613, "eval_steps_per_second": 0.082, "eval_wer": 30.36923749180686, "step": 15000 } ], "max_steps": 15000, "num_train_epochs": 10, "total_flos": 1.3847683858735104e+20, "trial_name": null, "trial_params": null }