{ "best_metric": 26.518461874590344, "best_model_checkpoint": "whisper-small-mn-8/checkpoint-15000", "epoch": 5.197505197505198, "global_step": 15000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.01, "learning_rate": 4.2000000000000006e-07, "loss": 5.7521, "step": 25 }, { "epoch": 0.02, "learning_rate": 9.200000000000001e-07, "loss": 4.2433, "step": 50 }, { "epoch": 0.03, "learning_rate": 1.42e-06, "loss": 3.3083, "step": 75 }, { "epoch": 0.03, "learning_rate": 1.9200000000000003e-06, "loss": 2.6035, "step": 100 }, { "epoch": 0.04, "learning_rate": 2.42e-06, "loss": 2.098, "step": 125 }, { "epoch": 0.05, "learning_rate": 2.92e-06, "loss": 1.7045, "step": 150 }, { "epoch": 0.06, "learning_rate": 3.4200000000000007e-06, "loss": 1.4407, "step": 175 }, { "epoch": 0.07, "learning_rate": 3.920000000000001e-06, "loss": 1.2631, "step": 200 }, { "epoch": 0.08, "learning_rate": 4.42e-06, "loss": 1.1279, "step": 225 }, { "epoch": 0.09, "learning_rate": 4.92e-06, "loss": 1.0251, "step": 250 }, { "epoch": 0.1, "learning_rate": 5.420000000000001e-06, "loss": 0.935, "step": 275 }, { "epoch": 0.1, "learning_rate": 5.92e-06, "loss": 0.8578, "step": 300 }, { "epoch": 0.11, "learning_rate": 6.42e-06, "loss": 0.8026, "step": 325 }, { "epoch": 0.12, "learning_rate": 6.92e-06, "loss": 0.7708, "step": 350 }, { "epoch": 0.13, "learning_rate": 7.420000000000001e-06, "loss": 0.7159, "step": 375 }, { "epoch": 0.14, "learning_rate": 7.92e-06, "loss": 0.6616, "step": 400 }, { "epoch": 0.15, "learning_rate": 8.42e-06, "loss": 0.6103, "step": 425 }, { "epoch": 0.16, "learning_rate": 8.920000000000001e-06, "loss": 0.58, "step": 450 }, { "epoch": 0.16, "learning_rate": 9.42e-06, "loss": 0.5592, "step": 475 }, { "epoch": 0.17, "learning_rate": 9.920000000000002e-06, "loss": 0.5538, "step": 500 }, { "epoch": 0.18, "learning_rate": 9.98551724137931e-06, "loss": 0.5301, "step": 525 }, { "epoch": 0.19, "learning_rate": 9.968275862068967e-06, "loss": 0.5256, "step": 550 }, { "epoch": 0.2, "learning_rate": 9.951034482758622e-06, "loss": 0.4933, "step": 575 }, { "epoch": 0.21, "learning_rate": 9.933793103448277e-06, "loss": 0.4914, "step": 600 }, { "epoch": 0.22, "learning_rate": 9.916551724137932e-06, "loss": 0.4576, "step": 625 }, { "epoch": 0.23, "learning_rate": 9.899310344827588e-06, "loss": 0.4671, "step": 650 }, { "epoch": 0.23, "learning_rate": 9.882068965517241e-06, "loss": 0.4616, "step": 675 }, { "epoch": 0.24, "learning_rate": 9.864827586206898e-06, "loss": 0.4608, "step": 700 }, { "epoch": 0.25, "learning_rate": 9.847586206896553e-06, "loss": 0.4375, "step": 725 }, { "epoch": 0.26, "learning_rate": 9.830344827586208e-06, "loss": 0.4422, "step": 750 }, { "epoch": 0.27, "learning_rate": 9.813103448275862e-06, "loss": 0.4206, "step": 775 }, { "epoch": 0.28, "learning_rate": 9.795862068965517e-06, "loss": 0.4086, "step": 800 }, { "epoch": 0.29, "learning_rate": 9.778620689655172e-06, "loss": 0.4152, "step": 825 }, { "epoch": 0.29, "learning_rate": 9.761379310344829e-06, "loss": 0.4028, "step": 850 }, { "epoch": 0.3, "learning_rate": 9.744137931034484e-06, "loss": 0.4038, "step": 875 }, { "epoch": 0.31, "learning_rate": 9.726896551724139e-06, "loss": 0.4014, "step": 900 }, { "epoch": 0.32, "learning_rate": 9.709655172413795e-06, "loss": 0.3884, "step": 925 }, { "epoch": 0.33, "learning_rate": 9.692413793103448e-06, "loss": 0.3931, "step": 950 }, { "epoch": 0.34, "learning_rate": 9.675172413793105e-06, "loss": 0.3773, "step": 975 }, { "epoch": 0.35, "learning_rate": 9.65793103448276e-06, "loss": 0.3717, "step": 1000 }, { "epoch": 0.35, "eval_cer": 16.966355814312948, "eval_loss": 0.4004291296005249, "eval_runtime": 765.5183, "eval_samples_per_second": 2.464, "eval_steps_per_second": 0.077, "eval_wer": 46.95761415774525, "step": 1000 }, { "epoch": 0.36, "learning_rate": 9.640689655172415e-06, "loss": 0.3748, "step": 1025 }, { "epoch": 0.36, "learning_rate": 9.62344827586207e-06, "loss": 0.3679, "step": 1050 }, { "epoch": 0.37, "learning_rate": 9.606206896551726e-06, "loss": 0.3767, "step": 1075 }, { "epoch": 0.38, "learning_rate": 9.58896551724138e-06, "loss": 0.3593, "step": 1100 }, { "epoch": 0.39, "learning_rate": 9.571724137931036e-06, "loss": 0.3752, "step": 1125 }, { "epoch": 0.4, "learning_rate": 9.55448275862069e-06, "loss": 0.3604, "step": 1150 }, { "epoch": 0.41, "learning_rate": 9.537241379310345e-06, "loss": 0.3416, "step": 1175 }, { "epoch": 0.42, "learning_rate": 9.52e-06, "loss": 0.3397, "step": 1200 }, { "epoch": 0.42, "learning_rate": 9.502758620689655e-06, "loss": 0.3516, "step": 1225 }, { "epoch": 0.43, "learning_rate": 9.48551724137931e-06, "loss": 0.3509, "step": 1250 }, { "epoch": 0.44, "learning_rate": 9.468275862068967e-06, "loss": 0.3346, "step": 1275 }, { "epoch": 0.45, "learning_rate": 9.451034482758622e-06, "loss": 0.348, "step": 1300 }, { "epoch": 0.46, "learning_rate": 9.433793103448276e-06, "loss": 0.3333, "step": 1325 }, { "epoch": 0.47, "learning_rate": 9.416551724137933e-06, "loss": 0.3296, "step": 1350 }, { "epoch": 0.48, "learning_rate": 9.399310344827586e-06, "loss": 0.3275, "step": 1375 }, { "epoch": 0.49, "learning_rate": 9.382068965517243e-06, "loss": 0.3277, "step": 1400 }, { "epoch": 0.49, "learning_rate": 9.364827586206898e-06, "loss": 0.3284, "step": 1425 }, { "epoch": 0.5, "learning_rate": 9.347586206896552e-06, "loss": 0.3171, "step": 1450 }, { "epoch": 0.51, "learning_rate": 9.330344827586207e-06, "loss": 0.3232, "step": 1475 }, { "epoch": 0.52, "learning_rate": 9.313103448275864e-06, "loss": 0.3207, "step": 1500 }, { "epoch": 0.53, "learning_rate": 9.295862068965517e-06, "loss": 0.3199, "step": 1525 }, { "epoch": 0.54, "learning_rate": 9.278620689655174e-06, "loss": 0.3219, "step": 1550 }, { "epoch": 0.55, "learning_rate": 9.261379310344828e-06, "loss": 0.309, "step": 1575 }, { "epoch": 0.55, "learning_rate": 9.244137931034483e-06, "loss": 0.322, "step": 1600 }, { "epoch": 0.56, "learning_rate": 9.226896551724138e-06, "loss": 0.3089, "step": 1625 }, { "epoch": 0.57, "learning_rate": 9.209655172413793e-06, "loss": 0.3119, "step": 1650 }, { "epoch": 0.58, "learning_rate": 9.192413793103448e-06, "loss": 0.302, "step": 1675 }, { "epoch": 0.59, "learning_rate": 9.175172413793105e-06, "loss": 0.3096, "step": 1700 }, { "epoch": 0.6, "learning_rate": 9.15793103448276e-06, "loss": 0.3076, "step": 1725 }, { "epoch": 0.61, "learning_rate": 9.140689655172414e-06, "loss": 0.3081, "step": 1750 }, { "epoch": 0.62, "learning_rate": 9.12344827586207e-06, "loss": 0.2998, "step": 1775 }, { "epoch": 0.62, "learning_rate": 9.106206896551724e-06, "loss": 0.3047, "step": 1800 }, { "epoch": 0.63, "learning_rate": 9.08896551724138e-06, "loss": 0.2942, "step": 1825 }, { "epoch": 0.64, "learning_rate": 9.071724137931035e-06, "loss": 0.291, "step": 1850 }, { "epoch": 0.65, "learning_rate": 9.05448275862069e-06, "loss": 0.3023, "step": 1875 }, { "epoch": 0.66, "learning_rate": 9.037241379310345e-06, "loss": 0.3062, "step": 1900 }, { "epoch": 0.67, "learning_rate": 9.020000000000002e-06, "loss": 0.2873, "step": 1925 }, { "epoch": 0.68, "learning_rate": 9.002758620689655e-06, "loss": 0.2826, "step": 1950 }, { "epoch": 0.68, "learning_rate": 8.985517241379311e-06, "loss": 0.291, "step": 1975 }, { "epoch": 0.69, "learning_rate": 8.968275862068966e-06, "loss": 0.286, "step": 2000 }, { "epoch": 0.69, "eval_cer": 13.550414756041013, "eval_loss": 0.31291720271110535, "eval_runtime": 730.0577, "eval_samples_per_second": 2.583, "eval_steps_per_second": 0.081, "eval_wer": 37.393489185055714, "step": 2000 }, { "epoch": 0.7, "learning_rate": 8.951034482758621e-06, "loss": 0.2999, "step": 2025 }, { "epoch": 0.71, "learning_rate": 8.933793103448276e-06, "loss": 0.2938, "step": 2050 }, { "epoch": 0.72, "learning_rate": 8.916551724137931e-06, "loss": 0.2769, "step": 2075 }, { "epoch": 0.73, "learning_rate": 8.899310344827588e-06, "loss": 0.287, "step": 2100 }, { "epoch": 0.74, "learning_rate": 8.882068965517242e-06, "loss": 0.2841, "step": 2125 }, { "epoch": 0.74, "learning_rate": 8.864827586206897e-06, "loss": 0.2818, "step": 2150 }, { "epoch": 0.75, "learning_rate": 8.847586206896552e-06, "loss": 0.2996, "step": 2175 }, { "epoch": 0.76, "learning_rate": 8.830344827586209e-06, "loss": 0.2787, "step": 2200 }, { "epoch": 0.77, "learning_rate": 8.813103448275862e-06, "loss": 0.2781, "step": 2225 }, { "epoch": 0.78, "learning_rate": 8.795862068965518e-06, "loss": 0.2734, "step": 2250 }, { "epoch": 0.79, "learning_rate": 8.778620689655173e-06, "loss": 0.2811, "step": 2275 }, { "epoch": 0.8, "learning_rate": 8.761379310344828e-06, "loss": 0.2775, "step": 2300 }, { "epoch": 0.81, "learning_rate": 8.744137931034483e-06, "loss": 0.2821, "step": 2325 }, { "epoch": 0.81, "learning_rate": 8.72689655172414e-06, "loss": 0.2799, "step": 2350 }, { "epoch": 0.82, "learning_rate": 8.709655172413793e-06, "loss": 0.2777, "step": 2375 }, { "epoch": 0.83, "learning_rate": 8.69241379310345e-06, "loss": 0.2756, "step": 2400 }, { "epoch": 0.84, "learning_rate": 8.675172413793104e-06, "loss": 0.2793, "step": 2425 }, { "epoch": 0.85, "learning_rate": 8.657931034482759e-06, "loss": 0.2769, "step": 2450 }, { "epoch": 0.86, "learning_rate": 8.640689655172414e-06, "loss": 0.2755, "step": 2475 }, { "epoch": 0.87, "learning_rate": 8.623448275862069e-06, "loss": 0.2876, "step": 2500 }, { "epoch": 0.87, "learning_rate": 8.606206896551725e-06, "loss": 0.2634, "step": 2525 }, { "epoch": 0.88, "learning_rate": 8.58896551724138e-06, "loss": 0.278, "step": 2550 }, { "epoch": 0.89, "learning_rate": 8.571724137931035e-06, "loss": 0.2636, "step": 2575 }, { "epoch": 0.9, "learning_rate": 8.55448275862069e-06, "loss": 0.2657, "step": 2600 }, { "epoch": 0.91, "learning_rate": 8.537241379310347e-06, "loss": 0.2765, "step": 2625 }, { "epoch": 0.92, "learning_rate": 8.52e-06, "loss": 0.2816, "step": 2650 }, { "epoch": 0.93, "learning_rate": 8.502758620689656e-06, "loss": 0.266, "step": 2675 }, { "epoch": 0.94, "learning_rate": 8.485517241379311e-06, "loss": 0.271, "step": 2700 }, { "epoch": 0.94, "learning_rate": 8.468275862068966e-06, "loss": 0.263, "step": 2725 }, { "epoch": 0.95, "learning_rate": 8.451034482758621e-06, "loss": 0.2665, "step": 2750 }, { "epoch": 0.96, "learning_rate": 8.433793103448277e-06, "loss": 0.2628, "step": 2775 }, { "epoch": 0.97, "learning_rate": 8.41655172413793e-06, "loss": 0.2622, "step": 2800 }, { "epoch": 0.98, "learning_rate": 8.399310344827587e-06, "loss": 0.2628, "step": 2825 }, { "epoch": 0.99, "learning_rate": 8.382068965517242e-06, "loss": 0.2615, "step": 2850 }, { "epoch": 1.0, "learning_rate": 8.364827586206897e-06, "loss": 0.2603, "step": 2875 }, { "epoch": 1.0, "learning_rate": 8.347586206896552e-06, "loss": 0.2416, "step": 2900 }, { "epoch": 1.01, "learning_rate": 8.330344827586208e-06, "loss": 0.2158, "step": 2925 }, { "epoch": 1.02, "learning_rate": 8.313103448275863e-06, "loss": 0.2291, "step": 2950 }, { "epoch": 1.03, "learning_rate": 8.295862068965518e-06, "loss": 0.2348, "step": 2975 }, { "epoch": 1.04, "learning_rate": 8.278620689655173e-06, "loss": 0.2287, "step": 3000 }, { "epoch": 1.04, "eval_cer": 11.780617239425009, "eval_loss": 0.27681875228881836, "eval_runtime": 722.4474, "eval_samples_per_second": 2.611, "eval_steps_per_second": 0.082, "eval_wer": 33.19313961109897, "step": 3000 }, { "epoch": 1.05, "learning_rate": 8.261379310344828e-06, "loss": 0.2303, "step": 3025 }, { "epoch": 1.06, "learning_rate": 8.244137931034484e-06, "loss": 0.2153, "step": 3050 }, { "epoch": 1.07, "learning_rate": 8.226896551724138e-06, "loss": 0.2276, "step": 3075 }, { "epoch": 1.07, "learning_rate": 8.209655172413794e-06, "loss": 0.223, "step": 3100 }, { "epoch": 1.08, "learning_rate": 8.192413793103449e-06, "loss": 0.2163, "step": 3125 }, { "epoch": 1.09, "learning_rate": 8.175172413793104e-06, "loss": 0.223, "step": 3150 }, { "epoch": 1.1, "learning_rate": 8.157931034482759e-06, "loss": 0.2245, "step": 3175 }, { "epoch": 1.11, "learning_rate": 8.140689655172415e-06, "loss": 0.2192, "step": 3200 }, { "epoch": 1.12, "learning_rate": 8.123448275862069e-06, "loss": 0.2197, "step": 3225 }, { "epoch": 1.13, "learning_rate": 8.106206896551725e-06, "loss": 0.22, "step": 3250 }, { "epoch": 1.13, "learning_rate": 8.08896551724138e-06, "loss": 0.2209, "step": 3275 }, { "epoch": 1.14, "learning_rate": 8.071724137931035e-06, "loss": 0.2143, "step": 3300 }, { "epoch": 1.15, "learning_rate": 8.054482758620691e-06, "loss": 0.2254, "step": 3325 }, { "epoch": 1.16, "learning_rate": 8.037241379310346e-06, "loss": 0.2261, "step": 3350 }, { "epoch": 1.17, "learning_rate": 8.020000000000001e-06, "loss": 0.2231, "step": 3375 }, { "epoch": 1.18, "learning_rate": 8.002758620689656e-06, "loss": 0.2227, "step": 3400 }, { "epoch": 1.19, "learning_rate": 7.985517241379311e-06, "loss": 0.2233, "step": 3425 }, { "epoch": 1.2, "learning_rate": 7.968275862068966e-06, "loss": 0.2184, "step": 3450 }, { "epoch": 1.2, "learning_rate": 7.951034482758622e-06, "loss": 0.208, "step": 3475 }, { "epoch": 1.21, "learning_rate": 7.933793103448275e-06, "loss": 0.2262, "step": 3500 }, { "epoch": 1.22, "learning_rate": 7.916551724137932e-06, "loss": 0.2215, "step": 3525 }, { "epoch": 1.23, "learning_rate": 7.899310344827587e-06, "loss": 0.2175, "step": 3550 }, { "epoch": 1.24, "learning_rate": 7.882068965517242e-06, "loss": 0.2258, "step": 3575 }, { "epoch": 1.25, "learning_rate": 7.864827586206897e-06, "loss": 0.2237, "step": 3600 }, { "epoch": 1.26, "learning_rate": 7.847586206896553e-06, "loss": 0.2168, "step": 3625 }, { "epoch": 1.26, "learning_rate": 7.830344827586206e-06, "loss": 0.2263, "step": 3650 }, { "epoch": 1.27, "learning_rate": 7.813103448275863e-06, "loss": 0.2224, "step": 3675 }, { "epoch": 1.28, "learning_rate": 7.795862068965518e-06, "loss": 0.2238, "step": 3700 }, { "epoch": 1.29, "learning_rate": 7.778620689655173e-06, "loss": 0.2184, "step": 3725 }, { "epoch": 1.3, "learning_rate": 7.76137931034483e-06, "loss": 0.2268, "step": 3750 }, { "epoch": 1.31, "learning_rate": 7.744137931034484e-06, "loss": 0.214, "step": 3775 }, { "epoch": 1.32, "learning_rate": 7.726896551724139e-06, "loss": 0.2189, "step": 3800 }, { "epoch": 1.33, "learning_rate": 7.709655172413794e-06, "loss": 0.2194, "step": 3825 }, { "epoch": 1.33, "learning_rate": 7.692413793103449e-06, "loss": 0.2232, "step": 3850 }, { "epoch": 1.34, "learning_rate": 7.675172413793104e-06, "loss": 0.2197, "step": 3875 }, { "epoch": 1.35, "learning_rate": 7.65793103448276e-06, "loss": 0.2154, "step": 3900 }, { "epoch": 1.36, "learning_rate": 7.640689655172413e-06, "loss": 0.2138, "step": 3925 }, { "epoch": 1.37, "learning_rate": 7.62344827586207e-06, "loss": 0.2052, "step": 3950 }, { "epoch": 1.38, "learning_rate": 7.606206896551725e-06, "loss": 0.2183, "step": 3975 }, { "epoch": 1.39, "learning_rate": 7.58896551724138e-06, "loss": 0.2257, "step": 4000 }, { "epoch": 1.39, "eval_cer": 11.023236642794581, "eval_loss": 0.2590450048446655, "eval_runtime": 729.8283, "eval_samples_per_second": 2.584, "eval_steps_per_second": 0.081, "eval_wer": 30.72427354162115, "step": 4000 }, { "epoch": 1.39, "learning_rate": 7.5717241379310345e-06, "loss": 0.2154, "step": 4025 }, { "epoch": 1.4, "learning_rate": 7.55448275862069e-06, "loss": 0.2262, "step": 4050 }, { "epoch": 1.41, "learning_rate": 7.537241379310345e-06, "loss": 0.2083, "step": 4075 }, { "epoch": 1.42, "learning_rate": 7.520000000000001e-06, "loss": 0.206, "step": 4100 }, { "epoch": 1.43, "learning_rate": 7.5027586206896566e-06, "loss": 0.2101, "step": 4125 }, { "epoch": 1.44, "learning_rate": 7.485517241379311e-06, "loss": 0.2114, "step": 4150 }, { "epoch": 1.45, "learning_rate": 7.468275862068966e-06, "loss": 0.2115, "step": 4175 }, { "epoch": 1.46, "learning_rate": 7.451034482758621e-06, "loss": 0.2086, "step": 4200 }, { "epoch": 1.46, "learning_rate": 7.433793103448277e-06, "loss": 0.22, "step": 4225 }, { "epoch": 1.47, "learning_rate": 7.416551724137932e-06, "loss": 0.2175, "step": 4250 }, { "epoch": 1.48, "learning_rate": 7.3993103448275875e-06, "loss": 0.2069, "step": 4275 }, { "epoch": 1.49, "learning_rate": 7.3820689655172415e-06, "loss": 0.2133, "step": 4300 }, { "epoch": 1.5, "learning_rate": 7.364827586206897e-06, "loss": 0.2135, "step": 4325 }, { "epoch": 1.51, "learning_rate": 7.347586206896552e-06, "loss": 0.2148, "step": 4350 }, { "epoch": 1.52, "learning_rate": 7.330344827586208e-06, "loss": 0.2138, "step": 4375 }, { "epoch": 1.52, "learning_rate": 7.313103448275863e-06, "loss": 0.2156, "step": 4400 }, { "epoch": 1.53, "learning_rate": 7.2958620689655175e-06, "loss": 0.2082, "step": 4425 }, { "epoch": 1.54, "learning_rate": 7.278620689655172e-06, "loss": 0.2084, "step": 4450 }, { "epoch": 1.55, "learning_rate": 7.261379310344828e-06, "loss": 0.2119, "step": 4475 }, { "epoch": 1.56, "learning_rate": 7.244137931034483e-06, "loss": 0.2192, "step": 4500 }, { "epoch": 1.57, "learning_rate": 7.226896551724139e-06, "loss": 0.2085, "step": 4525 }, { "epoch": 1.58, "learning_rate": 7.2096551724137944e-06, "loss": 0.2031, "step": 4550 }, { "epoch": 1.59, "learning_rate": 7.1924137931034485e-06, "loss": 0.2177, "step": 4575 }, { "epoch": 1.59, "learning_rate": 7.175172413793104e-06, "loss": 0.2093, "step": 4600 }, { "epoch": 1.6, "learning_rate": 7.157931034482759e-06, "loss": 0.2084, "step": 4625 }, { "epoch": 1.61, "learning_rate": 7.140689655172415e-06, "loss": 0.2045, "step": 4650 }, { "epoch": 1.62, "learning_rate": 7.12344827586207e-06, "loss": 0.2078, "step": 4675 }, { "epoch": 1.63, "learning_rate": 7.106206896551725e-06, "loss": 0.2041, "step": 4700 }, { "epoch": 1.64, "learning_rate": 7.088965517241379e-06, "loss": 0.2081, "step": 4725 }, { "epoch": 1.65, "learning_rate": 7.071724137931035e-06, "loss": 0.2102, "step": 4750 }, { "epoch": 1.65, "learning_rate": 7.05448275862069e-06, "loss": 0.2063, "step": 4775 }, { "epoch": 1.66, "learning_rate": 7.037241379310346e-06, "loss": 0.2041, "step": 4800 }, { "epoch": 1.67, "learning_rate": 7.0200000000000006e-06, "loss": 0.1989, "step": 4825 }, { "epoch": 1.68, "learning_rate": 7.002758620689655e-06, "loss": 0.2058, "step": 4850 }, { "epoch": 1.69, "learning_rate": 6.98551724137931e-06, "loss": 0.2166, "step": 4875 }, { "epoch": 1.7, "learning_rate": 6.968275862068966e-06, "loss": 0.1983, "step": 4900 }, { "epoch": 1.71, "learning_rate": 6.951034482758622e-06, "loss": 0.2099, "step": 4925 }, { "epoch": 1.72, "learning_rate": 6.933793103448277e-06, "loss": 0.2048, "step": 4950 }, { "epoch": 1.72, "learning_rate": 6.916551724137932e-06, "loss": 0.206, "step": 4975 }, { "epoch": 1.73, "learning_rate": 6.899310344827586e-06, "loss": 0.2029, "step": 5000 }, { "epoch": 1.73, "eval_cer": 10.414412557748124, "eval_loss": 0.24278895556926727, "eval_runtime": 721.577, "eval_samples_per_second": 2.614, "eval_steps_per_second": 0.082, "eval_wer": 29.200349573956743, "step": 5000 }, { "epoch": 1.74, "learning_rate": 6.882068965517242e-06, "loss": 0.1993, "step": 5025 }, { "epoch": 1.75, "learning_rate": 6.864827586206897e-06, "loss": 0.1981, "step": 5050 }, { "epoch": 1.76, "learning_rate": 6.847586206896553e-06, "loss": 0.2036, "step": 5075 }, { "epoch": 1.77, "learning_rate": 6.8303448275862075e-06, "loss": 0.2076, "step": 5100 }, { "epoch": 1.78, "learning_rate": 6.813103448275863e-06, "loss": 0.2014, "step": 5125 }, { "epoch": 1.78, "learning_rate": 6.795862068965517e-06, "loss": 0.2109, "step": 5150 }, { "epoch": 1.79, "learning_rate": 6.778620689655173e-06, "loss": 0.209, "step": 5175 }, { "epoch": 1.8, "learning_rate": 6.761379310344828e-06, "loss": 0.2149, "step": 5200 }, { "epoch": 1.81, "learning_rate": 6.7441379310344836e-06, "loss": 0.2038, "step": 5225 }, { "epoch": 1.82, "learning_rate": 6.7268965517241384e-06, "loss": 0.2136, "step": 5250 }, { "epoch": 1.83, "learning_rate": 6.709655172413793e-06, "loss": 0.2065, "step": 5275 }, { "epoch": 1.84, "learning_rate": 6.692413793103448e-06, "loss": 0.2078, "step": 5300 }, { "epoch": 1.85, "learning_rate": 6.675172413793104e-06, "loss": 0.2102, "step": 5325 }, { "epoch": 1.85, "learning_rate": 6.65793103448276e-06, "loss": 0.2021, "step": 5350 }, { "epoch": 1.86, "learning_rate": 6.6406896551724145e-06, "loss": 0.2034, "step": 5375 }, { "epoch": 1.87, "learning_rate": 6.62344827586207e-06, "loss": 0.2108, "step": 5400 }, { "epoch": 1.88, "learning_rate": 6.606206896551724e-06, "loss": 0.2059, "step": 5425 }, { "epoch": 1.89, "learning_rate": 6.58896551724138e-06, "loss": 0.2059, "step": 5450 }, { "epoch": 1.9, "learning_rate": 6.571724137931035e-06, "loss": 0.2065, "step": 5475 }, { "epoch": 1.91, "learning_rate": 6.5544827586206905e-06, "loss": 0.2063, "step": 5500 }, { "epoch": 1.91, "learning_rate": 6.537241379310345e-06, "loss": 0.2014, "step": 5525 }, { "epoch": 1.92, "learning_rate": 6.520000000000001e-06, "loss": 0.192, "step": 5550 }, { "epoch": 1.93, "learning_rate": 6.502758620689655e-06, "loss": 0.1984, "step": 5575 }, { "epoch": 1.94, "learning_rate": 6.485517241379311e-06, "loss": 0.2068, "step": 5600 }, { "epoch": 1.95, "learning_rate": 6.468275862068966e-06, "loss": 0.2098, "step": 5625 }, { "epoch": 1.96, "learning_rate": 6.4510344827586214e-06, "loss": 0.1944, "step": 5650 }, { "epoch": 1.97, "learning_rate": 6.433793103448276e-06, "loss": 0.2155, "step": 5675 }, { "epoch": 1.98, "learning_rate": 6.416551724137931e-06, "loss": 0.1972, "step": 5700 }, { "epoch": 1.98, "learning_rate": 6.399310344827587e-06, "loss": 0.2016, "step": 5725 }, { "epoch": 1.99, "learning_rate": 6.382068965517242e-06, "loss": 0.2045, "step": 5750 }, { "epoch": 2.0, "learning_rate": 6.3648275862068975e-06, "loss": 0.1958, "step": 5775 }, { "epoch": 2.01, "learning_rate": 6.347586206896552e-06, "loss": 0.1609, "step": 5800 }, { "epoch": 2.02, "learning_rate": 6.330344827586208e-06, "loss": 0.1631, "step": 5825 }, { "epoch": 2.03, "learning_rate": 6.313103448275862e-06, "loss": 0.1684, "step": 5850 }, { "epoch": 2.04, "learning_rate": 6.295862068965518e-06, "loss": 0.1639, "step": 5875 }, { "epoch": 2.04, "learning_rate": 6.278620689655173e-06, "loss": 0.161, "step": 5900 }, { "epoch": 2.05, "learning_rate": 6.261379310344828e-06, "loss": 0.1581, "step": 5925 }, { "epoch": 2.06, "learning_rate": 6.244137931034483e-06, "loss": 0.1671, "step": 5950 }, { "epoch": 2.07, "learning_rate": 6.226896551724139e-06, "loss": 0.1695, "step": 5975 }, { "epoch": 2.08, "learning_rate": 6.209655172413793e-06, "loss": 0.1691, "step": 6000 }, { "epoch": 2.08, "eval_cer": 10.03057001047624, "eval_loss": 0.2407660186290741, "eval_runtime": 718.5256, "eval_samples_per_second": 2.625, "eval_steps_per_second": 0.082, "eval_wer": 28.435656543587502, "step": 6000 }, { "epoch": 2.09, "learning_rate": 6.192413793103449e-06, "loss": 0.1592, "step": 6025 }, { "epoch": 2.1, "learning_rate": 6.175172413793104e-06, "loss": 0.163, "step": 6050 }, { "epoch": 2.1, "learning_rate": 6.157931034482759e-06, "loss": 0.1653, "step": 6075 }, { "epoch": 2.11, "learning_rate": 6.140689655172414e-06, "loss": 0.164, "step": 6100 }, { "epoch": 2.12, "learning_rate": 6.123448275862069e-06, "loss": 0.1694, "step": 6125 }, { "epoch": 2.13, "learning_rate": 6.106206896551725e-06, "loss": 0.1703, "step": 6150 }, { "epoch": 2.14, "learning_rate": 6.08896551724138e-06, "loss": 0.1633, "step": 6175 }, { "epoch": 2.15, "learning_rate": 6.071724137931035e-06, "loss": 0.1578, "step": 6200 }, { "epoch": 2.16, "learning_rate": 6.05448275862069e-06, "loss": 0.1669, "step": 6225 }, { "epoch": 2.17, "learning_rate": 6.037241379310346e-06, "loss": 0.1542, "step": 6250 }, { "epoch": 2.17, "learning_rate": 6.02e-06, "loss": 0.1595, "step": 6275 }, { "epoch": 2.18, "learning_rate": 6.002758620689656e-06, "loss": 0.1637, "step": 6300 }, { "epoch": 2.19, "learning_rate": 5.9855172413793105e-06, "loss": 0.1617, "step": 6325 }, { "epoch": 2.2, "learning_rate": 5.968275862068966e-06, "loss": 0.1636, "step": 6350 }, { "epoch": 2.21, "learning_rate": 5.951034482758621e-06, "loss": 0.1669, "step": 6375 }, { "epoch": 2.22, "learning_rate": 5.933793103448277e-06, "loss": 0.1706, "step": 6400 }, { "epoch": 2.23, "learning_rate": 5.916551724137931e-06, "loss": 0.1609, "step": 6425 }, { "epoch": 2.23, "learning_rate": 5.899310344827587e-06, "loss": 0.1658, "step": 6450 }, { "epoch": 2.24, "learning_rate": 5.8820689655172415e-06, "loss": 0.165, "step": 6475 }, { "epoch": 2.25, "learning_rate": 5.864827586206897e-06, "loss": 0.1644, "step": 6500 }, { "epoch": 2.26, "learning_rate": 5.847586206896552e-06, "loss": 0.169, "step": 6525 }, { "epoch": 2.27, "learning_rate": 5.830344827586208e-06, "loss": 0.1754, "step": 6550 }, { "epoch": 2.28, "learning_rate": 5.813103448275863e-06, "loss": 0.1669, "step": 6575 }, { "epoch": 2.29, "learning_rate": 5.7958620689655175e-06, "loss": 0.1755, "step": 6600 }, { "epoch": 2.3, "learning_rate": 5.778620689655173e-06, "loss": 0.1649, "step": 6625 }, { "epoch": 2.3, "learning_rate": 5.761379310344828e-06, "loss": 0.1663, "step": 6650 }, { "epoch": 2.31, "learning_rate": 5.744137931034484e-06, "loss": 0.1733, "step": 6675 }, { "epoch": 2.32, "learning_rate": 5.726896551724138e-06, "loss": 0.1639, "step": 6700 }, { "epoch": 2.33, "learning_rate": 5.7096551724137936e-06, "loss": 0.1704, "step": 6725 }, { "epoch": 2.34, "learning_rate": 5.692413793103448e-06, "loss": 0.1668, "step": 6750 }, { "epoch": 2.35, "learning_rate": 5.675172413793104e-06, "loss": 0.1636, "step": 6775 }, { "epoch": 2.36, "learning_rate": 5.657931034482759e-06, "loss": 0.1598, "step": 6800 }, { "epoch": 2.36, "learning_rate": 5.640689655172415e-06, "loss": 0.171, "step": 6825 }, { "epoch": 2.37, "learning_rate": 5.623448275862069e-06, "loss": 0.1668, "step": 6850 }, { "epoch": 2.38, "learning_rate": 5.6062068965517245e-06, "loss": 0.1716, "step": 6875 }, { "epoch": 2.39, "learning_rate": 5.588965517241379e-06, "loss": 0.1655, "step": 6900 }, { "epoch": 2.4, "learning_rate": 5.571724137931035e-06, "loss": 0.1622, "step": 6925 }, { "epoch": 2.41, "learning_rate": 5.554482758620691e-06, "loss": 0.1629, "step": 6950 }, { "epoch": 2.42, "learning_rate": 5.537241379310346e-06, "loss": 0.1639, "step": 6975 }, { "epoch": 2.43, "learning_rate": 5.5200000000000005e-06, "loss": 0.1626, "step": 7000 }, { "epoch": 2.43, "eval_cer": 10.048602881824584, "eval_loss": 0.23688766360282898, "eval_runtime": 725.1364, "eval_samples_per_second": 2.601, "eval_steps_per_second": 0.081, "eval_wer": 28.05877212147695, "step": 7000 }, { "epoch": 2.43, "learning_rate": 5.502758620689655e-06, "loss": 0.166, "step": 7025 }, { "epoch": 2.44, "learning_rate": 5.485517241379311e-06, "loss": 0.1691, "step": 7050 }, { "epoch": 2.45, "learning_rate": 5.468275862068966e-06, "loss": 0.1695, "step": 7075 }, { "epoch": 2.46, "learning_rate": 5.451034482758622e-06, "loss": 0.1685, "step": 7100 }, { "epoch": 2.47, "learning_rate": 5.433793103448276e-06, "loss": 0.1694, "step": 7125 }, { "epoch": 2.48, "learning_rate": 5.4165517241379314e-06, "loss": 0.1585, "step": 7150 }, { "epoch": 2.49, "learning_rate": 5.399310344827586e-06, "loss": 0.1686, "step": 7175 }, { "epoch": 2.49, "learning_rate": 5.382068965517242e-06, "loss": 0.1667, "step": 7200 }, { "epoch": 2.5, "learning_rate": 5.364827586206897e-06, "loss": 0.158, "step": 7225 }, { "epoch": 2.51, "learning_rate": 5.347586206896553e-06, "loss": 0.1675, "step": 7250 }, { "epoch": 2.52, "learning_rate": 5.330344827586207e-06, "loss": 0.1614, "step": 7275 }, { "epoch": 2.53, "learning_rate": 5.313103448275862e-06, "loss": 0.1698, "step": 7300 }, { "epoch": 2.54, "learning_rate": 5.295862068965517e-06, "loss": 0.1613, "step": 7325 }, { "epoch": 2.55, "learning_rate": 5.278620689655173e-06, "loss": 0.1631, "step": 7350 }, { "epoch": 2.56, "learning_rate": 5.261379310344829e-06, "loss": 0.1697, "step": 7375 }, { "epoch": 2.56, "learning_rate": 5.2441379310344835e-06, "loss": 0.1645, "step": 7400 }, { "epoch": 2.57, "learning_rate": 5.226896551724138e-06, "loss": 0.1657, "step": 7425 }, { "epoch": 2.58, "learning_rate": 5.209655172413793e-06, "loss": 0.1645, "step": 7450 }, { "epoch": 2.59, "learning_rate": 5.192413793103449e-06, "loss": 0.1638, "step": 7475 }, { "epoch": 2.6, "learning_rate": 5.175172413793104e-06, "loss": 0.162, "step": 7500 }, { "epoch": 2.61, "learning_rate": 5.1579310344827596e-06, "loss": 0.1567, "step": 7525 }, { "epoch": 2.62, "learning_rate": 5.140689655172414e-06, "loss": 0.1619, "step": 7550 }, { "epoch": 2.62, "learning_rate": 5.123448275862069e-06, "loss": 0.1589, "step": 7575 }, { "epoch": 2.63, "learning_rate": 5.106206896551724e-06, "loss": 0.1608, "step": 7600 }, { "epoch": 2.64, "learning_rate": 5.08896551724138e-06, "loss": 0.1623, "step": 7625 }, { "epoch": 2.65, "learning_rate": 5.071724137931035e-06, "loss": 0.1662, "step": 7650 }, { "epoch": 2.66, "learning_rate": 5.0544827586206905e-06, "loss": 0.1609, "step": 7675 }, { "epoch": 2.67, "learning_rate": 5.0372413793103445e-06, "loss": 0.1579, "step": 7700 }, { "epoch": 2.68, "learning_rate": 5.02e-06, "loss": 0.1649, "step": 7725 }, { "epoch": 2.69, "learning_rate": 5.002758620689656e-06, "loss": 0.1604, "step": 7750 }, { "epoch": 2.69, "learning_rate": 4.985517241379311e-06, "loss": 0.1646, "step": 7775 }, { "epoch": 2.7, "learning_rate": 4.968275862068966e-06, "loss": 0.1633, "step": 7800 }, { "epoch": 2.71, "learning_rate": 4.951034482758621e-06, "loss": 0.1611, "step": 7825 }, { "epoch": 2.72, "learning_rate": 4.933793103448276e-06, "loss": 0.1604, "step": 7850 }, { "epoch": 2.73, "learning_rate": 4.916551724137931e-06, "loss": 0.1595, "step": 7875 }, { "epoch": 2.74, "learning_rate": 4.899310344827586e-06, "loss": 0.1556, "step": 7900 }, { "epoch": 2.75, "learning_rate": 4.882068965517242e-06, "loss": 0.1658, "step": 7925 }, { "epoch": 2.75, "learning_rate": 4.8648275862068974e-06, "loss": 0.1618, "step": 7950 }, { "epoch": 2.76, "learning_rate": 4.847586206896552e-06, "loss": 0.1589, "step": 7975 }, { "epoch": 2.77, "learning_rate": 4.830344827586207e-06, "loss": 0.1588, "step": 8000 }, { "epoch": 2.77, "eval_cer": 9.681934497741597, "eval_loss": 0.2320903092622757, "eval_runtime": 725.3496, "eval_samples_per_second": 2.6, "eval_steps_per_second": 0.081, "eval_wer": 27.23399606729299, "step": 8000 }, { "epoch": 2.78, "learning_rate": 4.813103448275863e-06, "loss": 0.1632, "step": 8025 }, { "epoch": 2.79, "learning_rate": 4.795862068965518e-06, "loss": 0.157, "step": 8050 }, { "epoch": 2.8, "learning_rate": 4.778620689655173e-06, "loss": 0.1617, "step": 8075 }, { "epoch": 2.81, "learning_rate": 4.761379310344828e-06, "loss": 0.1652, "step": 8100 }, { "epoch": 2.82, "learning_rate": 4.744137931034483e-06, "loss": 0.1524, "step": 8125 }, { "epoch": 2.82, "learning_rate": 4.726896551724138e-06, "loss": 0.1547, "step": 8150 }, { "epoch": 2.83, "learning_rate": 4.709655172413794e-06, "loss": 0.1644, "step": 8175 }, { "epoch": 2.84, "learning_rate": 4.692413793103449e-06, "loss": 0.162, "step": 8200 }, { "epoch": 2.85, "learning_rate": 4.6751724137931035e-06, "loss": 0.1707, "step": 8225 }, { "epoch": 2.86, "learning_rate": 4.657931034482759e-06, "loss": 0.1657, "step": 8250 }, { "epoch": 2.87, "learning_rate": 4.640689655172414e-06, "loss": 0.1638, "step": 8275 }, { "epoch": 2.88, "learning_rate": 4.623448275862069e-06, "loss": 0.1583, "step": 8300 }, { "epoch": 2.88, "learning_rate": 4.606206896551724e-06, "loss": 0.1571, "step": 8325 }, { "epoch": 2.89, "learning_rate": 4.58896551724138e-06, "loss": 0.1578, "step": 8350 }, { "epoch": 2.9, "learning_rate": 4.571724137931035e-06, "loss": 0.1704, "step": 8375 }, { "epoch": 2.91, "learning_rate": 4.55448275862069e-06, "loss": 0.1651, "step": 8400 }, { "epoch": 2.92, "learning_rate": 4.537241379310345e-06, "loss": 0.1592, "step": 8425 }, { "epoch": 2.93, "learning_rate": 4.520000000000001e-06, "loss": 0.1527, "step": 8450 }, { "epoch": 2.94, "learning_rate": 4.502758620689656e-06, "loss": 0.1601, "step": 8475 }, { "epoch": 2.95, "learning_rate": 4.4855172413793105e-06, "loss": 0.1584, "step": 8500 }, { "epoch": 2.95, "learning_rate": 4.468275862068966e-06, "loss": 0.1627, "step": 8525 }, { "epoch": 2.96, "learning_rate": 4.451034482758621e-06, "loss": 0.1588, "step": 8550 }, { "epoch": 2.97, "learning_rate": 4.433793103448276e-06, "loss": 0.1628, "step": 8575 }, { "epoch": 2.98, "learning_rate": 4.416551724137932e-06, "loss": 0.1599, "step": 8600 }, { "epoch": 2.99, "learning_rate": 4.3993103448275866e-06, "loss": 0.1626, "step": 8625 }, { "epoch": 3.0, "learning_rate": 4.3820689655172414e-06, "loss": 0.1591, "step": 8650 }, { "epoch": 3.01, "learning_rate": 4.364827586206897e-06, "loss": 0.1473, "step": 8675 }, { "epoch": 3.01, "learning_rate": 4.347586206896552e-06, "loss": 0.1287, "step": 8700 }, { "epoch": 3.02, "learning_rate": 4.330344827586207e-06, "loss": 0.129, "step": 8725 }, { "epoch": 3.03, "learning_rate": 4.313103448275863e-06, "loss": 0.1339, "step": 8750 }, { "epoch": 3.04, "learning_rate": 4.2958620689655175e-06, "loss": 0.1287, "step": 8775 }, { "epoch": 3.05, "learning_rate": 4.278620689655173e-06, "loss": 0.1326, "step": 8800 }, { "epoch": 3.06, "learning_rate": 4.261379310344828e-06, "loss": 0.1284, "step": 8825 }, { "epoch": 3.07, "learning_rate": 4.244137931034483e-06, "loss": 0.1244, "step": 8850 }, { "epoch": 3.08, "learning_rate": 4.226896551724139e-06, "loss": 0.1198, "step": 8875 }, { "epoch": 3.08, "learning_rate": 4.2096551724137935e-06, "loss": 0.134, "step": 8900 }, { "epoch": 3.09, "learning_rate": 4.192413793103448e-06, "loss": 0.1323, "step": 8925 }, { "epoch": 3.1, "learning_rate": 4.175172413793104e-06, "loss": 0.1315, "step": 8950 }, { "epoch": 3.11, "learning_rate": 4.157931034482759e-06, "loss": 0.1248, "step": 8975 }, { "epoch": 3.12, "learning_rate": 4.140689655172414e-06, "loss": 0.1271, "step": 9000 }, { "epoch": 3.12, "eval_cer": 9.557421814622083, "eval_loss": 0.23485350608825684, "eval_runtime": 723.0189, "eval_samples_per_second": 2.609, "eval_steps_per_second": 0.082, "eval_wer": 26.84072536596024, "step": 9000 }, { "epoch": 3.13, "learning_rate": 4.1234482758620696e-06, "loss": 0.1291, "step": 9025 }, { "epoch": 3.14, "learning_rate": 4.1062068965517244e-06, "loss": 0.1324, "step": 9050 }, { "epoch": 3.14, "learning_rate": 4.088965517241379e-06, "loss": 0.1294, "step": 9075 }, { "epoch": 3.15, "learning_rate": 4.071724137931035e-06, "loss": 0.1334, "step": 9100 }, { "epoch": 3.16, "learning_rate": 4.05448275862069e-06, "loss": 0.1325, "step": 9125 }, { "epoch": 3.17, "learning_rate": 4.037241379310346e-06, "loss": 0.1343, "step": 9150 }, { "epoch": 3.18, "learning_rate": 4.0200000000000005e-06, "loss": 0.1313, "step": 9175 }, { "epoch": 3.19, "learning_rate": 4.002758620689655e-06, "loss": 0.1331, "step": 9200 }, { "epoch": 3.2, "learning_rate": 3.985517241379311e-06, "loss": 0.135, "step": 9225 }, { "epoch": 3.21, "learning_rate": 3.968275862068966e-06, "loss": 0.1359, "step": 9250 }, { "epoch": 3.21, "learning_rate": 3.951034482758621e-06, "loss": 0.13, "step": 9275 }, { "epoch": 3.22, "learning_rate": 3.9337931034482765e-06, "loss": 0.1335, "step": 9300 }, { "epoch": 3.23, "learning_rate": 3.916551724137931e-06, "loss": 0.1269, "step": 9325 }, { "epoch": 3.24, "learning_rate": 3.899310344827586e-06, "loss": 0.1296, "step": 9350 }, { "epoch": 3.25, "learning_rate": 3.882068965517242e-06, "loss": 0.1247, "step": 9375 }, { "epoch": 3.26, "learning_rate": 3.864827586206897e-06, "loss": 0.1249, "step": 9400 }, { "epoch": 3.27, "learning_rate": 3.847586206896552e-06, "loss": 0.1332, "step": 9425 }, { "epoch": 3.27, "learning_rate": 3.8303448275862074e-06, "loss": 0.1297, "step": 9450 }, { "epoch": 3.28, "learning_rate": 3.8131034482758623e-06, "loss": 0.1358, "step": 9475 }, { "epoch": 3.29, "learning_rate": 3.795862068965517e-06, "loss": 0.1275, "step": 9500 }, { "epoch": 3.3, "learning_rate": 3.7786206896551725e-06, "loss": 0.1374, "step": 9525 }, { "epoch": 3.31, "learning_rate": 3.7613793103448278e-06, "loss": 0.143, "step": 9550 }, { "epoch": 3.32, "learning_rate": 3.7441379310344835e-06, "loss": 0.1259, "step": 9575 }, { "epoch": 3.33, "learning_rate": 3.7268965517241383e-06, "loss": 0.1267, "step": 9600 }, { "epoch": 3.34, "learning_rate": 3.7096551724137936e-06, "loss": 0.1352, "step": 9625 }, { "epoch": 3.34, "learning_rate": 3.6924137931034485e-06, "loss": 0.1419, "step": 9650 }, { "epoch": 3.35, "learning_rate": 3.675172413793104e-06, "loss": 0.1292, "step": 9675 }, { "epoch": 3.36, "learning_rate": 3.657931034482759e-06, "loss": 0.1289, "step": 9700 }, { "epoch": 3.37, "learning_rate": 3.640689655172414e-06, "loss": 0.13, "step": 9725 }, { "epoch": 3.38, "learning_rate": 3.6234482758620693e-06, "loss": 0.1299, "step": 9750 }, { "epoch": 3.39, "learning_rate": 3.6062068965517246e-06, "loss": 0.1378, "step": 9775 }, { "epoch": 3.4, "learning_rate": 3.5889655172413794e-06, "loss": 0.1326, "step": 9800 }, { "epoch": 3.4, "learning_rate": 3.5717241379310347e-06, "loss": 0.1339, "step": 9825 }, { "epoch": 3.41, "learning_rate": 3.55448275862069e-06, "loss": 0.1299, "step": 9850 }, { "epoch": 3.42, "learning_rate": 3.537241379310345e-06, "loss": 0.131, "step": 9875 }, { "epoch": 3.43, "learning_rate": 3.52e-06, "loss": 0.133, "step": 9900 }, { "epoch": 3.44, "learning_rate": 3.502758620689655e-06, "loss": 0.1278, "step": 9925 }, { "epoch": 3.45, "learning_rate": 3.4855172413793103e-06, "loss": 0.137, "step": 9950 }, { "epoch": 3.46, "learning_rate": 3.468275862068966e-06, "loss": 0.1299, "step": 9975 }, { "epoch": 3.47, "learning_rate": 3.4510344827586214e-06, "loss": 0.1263, "step": 10000 }, { "epoch": 3.47, "eval_cer": 9.651879712161024, "eval_loss": 0.2355508804321289, "eval_runtime": 724.1513, "eval_samples_per_second": 2.604, "eval_steps_per_second": 0.081, "eval_wer": 27.162988857330127, "step": 10000 }, { "epoch": 3.47, "learning_rate": 3.4337931034482762e-06, "loss": 0.1355, "step": 10025 }, { "epoch": 3.48, "learning_rate": 3.4165517241379315e-06, "loss": 0.1272, "step": 10050 }, { "epoch": 3.49, "learning_rate": 3.3993103448275864e-06, "loss": 0.1357, "step": 10075 }, { "epoch": 3.5, "learning_rate": 3.3820689655172417e-06, "loss": 0.1312, "step": 10100 }, { "epoch": 3.51, "learning_rate": 3.364827586206897e-06, "loss": 0.1263, "step": 10125 }, { "epoch": 3.52, "learning_rate": 3.347586206896552e-06, "loss": 0.1307, "step": 10150 }, { "epoch": 3.53, "learning_rate": 3.330344827586207e-06, "loss": 0.1271, "step": 10175 }, { "epoch": 3.53, "learning_rate": 3.3131034482758624e-06, "loss": 0.1275, "step": 10200 }, { "epoch": 3.54, "learning_rate": 3.2958620689655173e-06, "loss": 0.1233, "step": 10225 }, { "epoch": 3.55, "learning_rate": 3.2786206896551726e-06, "loss": 0.13, "step": 10250 }, { "epoch": 3.56, "learning_rate": 3.261379310344828e-06, "loss": 0.1323, "step": 10275 }, { "epoch": 3.57, "learning_rate": 3.2441379310344828e-06, "loss": 0.1274, "step": 10300 }, { "epoch": 3.58, "learning_rate": 3.226896551724138e-06, "loss": 0.1331, "step": 10325 }, { "epoch": 3.59, "learning_rate": 3.209655172413793e-06, "loss": 0.1272, "step": 10350 }, { "epoch": 3.59, "learning_rate": 3.1924137931034486e-06, "loss": 0.1338, "step": 10375 }, { "epoch": 3.6, "learning_rate": 3.175172413793104e-06, "loss": 0.1242, "step": 10400 }, { "epoch": 3.61, "learning_rate": 3.1579310344827592e-06, "loss": 0.1313, "step": 10425 }, { "epoch": 3.62, "learning_rate": 3.140689655172414e-06, "loss": 0.1259, "step": 10450 }, { "epoch": 3.63, "learning_rate": 3.1234482758620694e-06, "loss": 0.1262, "step": 10475 }, { "epoch": 3.64, "learning_rate": 3.1068965517241384e-06, "loss": 0.1347, "step": 10500 }, { "epoch": 3.65, "learning_rate": 3.0896551724137937e-06, "loss": 0.1334, "step": 10525 }, { "epoch": 3.66, "learning_rate": 3.0724137931034485e-06, "loss": 0.1258, "step": 10550 }, { "epoch": 3.66, "learning_rate": 3.055172413793104e-06, "loss": 0.1338, "step": 10575 }, { "epoch": 3.67, "learning_rate": 3.037931034482759e-06, "loss": 0.1279, "step": 10600 }, { "epoch": 3.68, "learning_rate": 3.020689655172414e-06, "loss": 0.1248, "step": 10625 }, { "epoch": 3.69, "learning_rate": 3.0034482758620693e-06, "loss": 0.1341, "step": 10650 }, { "epoch": 3.7, "learning_rate": 2.9862068965517246e-06, "loss": 0.1296, "step": 10675 }, { "epoch": 3.71, "learning_rate": 2.9689655172413794e-06, "loss": 0.1291, "step": 10700 }, { "epoch": 3.72, "learning_rate": 2.9517241379310347e-06, "loss": 0.1327, "step": 10725 }, { "epoch": 3.72, "learning_rate": 2.9344827586206896e-06, "loss": 0.1295, "step": 10750 }, { "epoch": 3.73, "learning_rate": 2.917241379310345e-06, "loss": 0.1351, "step": 10775 }, { "epoch": 3.74, "learning_rate": 2.9e-06, "loss": 0.1246, "step": 10800 }, { "epoch": 3.75, "learning_rate": 2.882758620689655e-06, "loss": 0.1282, "step": 10825 }, { "epoch": 3.76, "learning_rate": 2.8655172413793104e-06, "loss": 0.1241, "step": 10850 }, { "epoch": 3.77, "learning_rate": 2.848275862068966e-06, "loss": 0.1247, "step": 10875 }, { "epoch": 3.78, "learning_rate": 2.831034482758621e-06, "loss": 0.132, "step": 10900 }, { "epoch": 3.79, "learning_rate": 2.8137931034482762e-06, "loss": 0.1192, "step": 10925 }, { "epoch": 3.79, "learning_rate": 2.7965517241379315e-06, "loss": 0.1295, "step": 10950 }, { "epoch": 3.8, "learning_rate": 2.7793103448275864e-06, "loss": 0.127, "step": 10975 }, { "epoch": 3.81, "learning_rate": 2.7620689655172417e-06, "loss": 0.1314, "step": 11000 }, { "epoch": 3.81, "eval_cer": 9.427756882545898, "eval_loss": 0.2339998185634613, "eval_runtime": 719.7681, "eval_samples_per_second": 2.62, "eval_steps_per_second": 0.082, "eval_wer": 26.5566965261088, "step": 11000 }, { "epoch": 3.82, "learning_rate": 2.744827586206897e-06, "loss": 0.1256, "step": 11025 }, { "epoch": 3.83, "learning_rate": 2.727586206896552e-06, "loss": 0.1311, "step": 11050 }, { "epoch": 3.84, "learning_rate": 2.710344827586207e-06, "loss": 0.1216, "step": 11075 }, { "epoch": 3.85, "learning_rate": 2.6931034482758624e-06, "loss": 0.1311, "step": 11100 }, { "epoch": 3.85, "learning_rate": 2.6758620689655173e-06, "loss": 0.1337, "step": 11125 }, { "epoch": 3.86, "learning_rate": 2.6586206896551726e-06, "loss": 0.1293, "step": 11150 }, { "epoch": 3.87, "learning_rate": 2.6413793103448275e-06, "loss": 0.1262, "step": 11175 }, { "epoch": 3.88, "learning_rate": 2.6241379310344828e-06, "loss": 0.1348, "step": 11200 }, { "epoch": 3.89, "learning_rate": 2.606896551724138e-06, "loss": 0.1307, "step": 11225 }, { "epoch": 3.9, "learning_rate": 2.589655172413793e-06, "loss": 0.131, "step": 11250 }, { "epoch": 3.91, "learning_rate": 2.5724137931034486e-06, "loss": 0.1343, "step": 11275 }, { "epoch": 3.92, "learning_rate": 2.555172413793104e-06, "loss": 0.1303, "step": 11300 }, { "epoch": 3.92, "learning_rate": 2.537931034482759e-06, "loss": 0.1397, "step": 11325 }, { "epoch": 3.93, "learning_rate": 2.520689655172414e-06, "loss": 0.1341, "step": 11350 }, { "epoch": 3.94, "learning_rate": 2.5034482758620694e-06, "loss": 0.1336, "step": 11375 }, { "epoch": 3.95, "learning_rate": 2.4862068965517243e-06, "loss": 0.133, "step": 11400 }, { "epoch": 3.96, "learning_rate": 2.4689655172413796e-06, "loss": 0.1305, "step": 11425 }, { "epoch": 3.97, "learning_rate": 2.451724137931035e-06, "loss": 0.1297, "step": 11450 }, { "epoch": 3.98, "learning_rate": 2.4344827586206897e-06, "loss": 0.1267, "step": 11475 }, { "epoch": 3.98, "learning_rate": 2.417241379310345e-06, "loss": 0.1281, "step": 11500 }, { "epoch": 3.99, "learning_rate": 2.4000000000000003e-06, "loss": 0.1252, "step": 11525 }, { "epoch": 4.0, "learning_rate": 2.382758620689655e-06, "loss": 0.1225, "step": 11550 }, { "epoch": 4.01, "learning_rate": 2.3655172413793105e-06, "loss": 0.1077, "step": 11575 }, { "epoch": 4.02, "learning_rate": 2.3482758620689658e-06, "loss": 0.1032, "step": 11600 }, { "epoch": 4.03, "learning_rate": 2.331034482758621e-06, "loss": 0.1026, "step": 11625 }, { "epoch": 4.04, "learning_rate": 2.313793103448276e-06, "loss": 0.105, "step": 11650 }, { "epoch": 4.05, "learning_rate": 2.2965517241379312e-06, "loss": 0.1035, "step": 11675 }, { "epoch": 4.05, "learning_rate": 2.2793103448275865e-06, "loss": 0.1058, "step": 11700 }, { "epoch": 4.06, "learning_rate": 2.2620689655172414e-06, "loss": 0.1063, "step": 11725 }, { "epoch": 4.07, "learning_rate": 2.2448275862068967e-06, "loss": 0.1079, "step": 11750 }, { "epoch": 4.08, "learning_rate": 2.227586206896552e-06, "loss": 0.1039, "step": 11775 }, { "epoch": 4.09, "learning_rate": 2.2103448275862073e-06, "loss": 0.1052, "step": 11800 }, { "epoch": 4.1, "learning_rate": 2.193103448275862e-06, "loss": 0.1087, "step": 11825 }, { "epoch": 4.11, "learning_rate": 2.1758620689655174e-06, "loss": 0.1102, "step": 11850 }, { "epoch": 4.11, "learning_rate": 2.1586206896551727e-06, "loss": 0.1053, "step": 11875 }, { "epoch": 4.12, "learning_rate": 2.1413793103448276e-06, "loss": 0.1098, "step": 11900 }, { "epoch": 4.13, "learning_rate": 2.124137931034483e-06, "loss": 0.1047, "step": 11925 }, { "epoch": 4.14, "learning_rate": 2.106896551724138e-06, "loss": 0.1063, "step": 11950 }, { "epoch": 4.15, "learning_rate": 2.0896551724137935e-06, "loss": 0.1134, "step": 11975 }, { "epoch": 4.16, "learning_rate": 2.0724137931034484e-06, "loss": 0.1062, "step": 12000 }, { "epoch": 4.16, "eval_cer": 9.516203822968725, "eval_loss": 0.23904471099376678, "eval_runtime": 723.1604, "eval_samples_per_second": 2.608, "eval_steps_per_second": 0.082, "eval_wer": 26.633165829145728, "step": 12000 }, { "epoch": 4.17, "learning_rate": 2.0551724137931036e-06, "loss": 0.1107, "step": 12025 }, { "epoch": 4.18, "learning_rate": 2.037931034482759e-06, "loss": 0.1055, "step": 12050 }, { "epoch": 4.18, "learning_rate": 2.020689655172414e-06, "loss": 0.1064, "step": 12075 }, { "epoch": 4.19, "learning_rate": 2.003448275862069e-06, "loss": 0.11, "step": 12100 }, { "epoch": 4.2, "learning_rate": 1.9862068965517244e-06, "loss": 0.102, "step": 12125 }, { "epoch": 4.21, "learning_rate": 1.9689655172413793e-06, "loss": 0.1046, "step": 12150 }, { "epoch": 4.22, "learning_rate": 1.9517241379310346e-06, "loss": 0.1057, "step": 12175 }, { "epoch": 4.23, "learning_rate": 1.93448275862069e-06, "loss": 0.1101, "step": 12200 }, { "epoch": 4.24, "learning_rate": 1.917241379310345e-06, "loss": 0.1097, "step": 12225 }, { "epoch": 4.24, "learning_rate": 1.9000000000000002e-06, "loss": 0.1064, "step": 12250 }, { "epoch": 4.25, "learning_rate": 1.8827586206896553e-06, "loss": 0.1071, "step": 12275 }, { "epoch": 4.26, "learning_rate": 1.8655172413793104e-06, "loss": 0.1057, "step": 12300 }, { "epoch": 4.27, "learning_rate": 1.8482758620689657e-06, "loss": 0.104, "step": 12325 }, { "epoch": 4.28, "learning_rate": 1.8310344827586208e-06, "loss": 0.1044, "step": 12350 }, { "epoch": 4.29, "learning_rate": 1.813793103448276e-06, "loss": 0.106, "step": 12375 }, { "epoch": 4.3, "learning_rate": 1.7965517241379314e-06, "loss": 0.1069, "step": 12400 }, { "epoch": 4.31, "learning_rate": 1.7793103448275864e-06, "loss": 0.1086, "step": 12425 }, { "epoch": 4.31, "learning_rate": 1.7620689655172415e-06, "loss": 0.1131, "step": 12450 }, { "epoch": 4.32, "learning_rate": 1.7448275862068966e-06, "loss": 0.1035, "step": 12475 }, { "epoch": 4.33, "learning_rate": 1.727586206896552e-06, "loss": 0.1091, "step": 12500 }, { "epoch": 4.34, "learning_rate": 1.710344827586207e-06, "loss": 0.1063, "step": 12525 }, { "epoch": 4.35, "learning_rate": 1.693103448275862e-06, "loss": 0.1028, "step": 12550 }, { "epoch": 4.36, "learning_rate": 1.6758620689655176e-06, "loss": 0.0989, "step": 12575 }, { "epoch": 4.37, "learning_rate": 1.6586206896551726e-06, "loss": 0.1048, "step": 12600 }, { "epoch": 4.37, "learning_rate": 1.6413793103448277e-06, "loss": 0.109, "step": 12625 }, { "epoch": 4.38, "learning_rate": 1.6241379310344828e-06, "loss": 0.1084, "step": 12650 }, { "epoch": 4.39, "learning_rate": 1.6068965517241381e-06, "loss": 0.1056, "step": 12675 }, { "epoch": 4.4, "learning_rate": 1.5896551724137932e-06, "loss": 0.1017, "step": 12700 }, { "epoch": 4.41, "learning_rate": 1.5724137931034483e-06, "loss": 0.1036, "step": 12725 }, { "epoch": 4.42, "learning_rate": 1.5551724137931036e-06, "loss": 0.1142, "step": 12750 }, { "epoch": 4.43, "learning_rate": 1.5386206896551725e-06, "loss": 0.1141, "step": 12775 }, { "epoch": 4.44, "learning_rate": 1.5213793103448276e-06, "loss": 0.1064, "step": 12800 }, { "epoch": 4.44, "learning_rate": 1.504137931034483e-06, "loss": 0.1019, "step": 12825 }, { "epoch": 4.45, "learning_rate": 1.486896551724138e-06, "loss": 0.1052, "step": 12850 }, { "epoch": 4.46, "learning_rate": 1.4696551724137933e-06, "loss": 0.107, "step": 12875 }, { "epoch": 4.47, "learning_rate": 1.4524137931034486e-06, "loss": 0.1069, "step": 12900 }, { "epoch": 4.48, "learning_rate": 1.4351724137931037e-06, "loss": 0.1105, "step": 12925 }, { "epoch": 4.49, "learning_rate": 1.4179310344827587e-06, "loss": 0.1057, "step": 12950 }, { "epoch": 4.5, "learning_rate": 1.4006896551724138e-06, "loss": 0.108, "step": 12975 }, { "epoch": 4.5, "learning_rate": 1.3834482758620691e-06, "loss": 0.1081, "step": 13000 }, { "epoch": 4.5, "eval_cer": 9.508475449533721, "eval_loss": 0.2397749125957489, "eval_runtime": 725.5304, "eval_samples_per_second": 2.599, "eval_steps_per_second": 0.081, "eval_wer": 26.584006991479136, "step": 13000 }, { "epoch": 4.51, "learning_rate": 1.3662068965517242e-06, "loss": 0.1056, "step": 13025 }, { "epoch": 4.52, "learning_rate": 1.3489655172413793e-06, "loss": 0.1052, "step": 13050 }, { "epoch": 4.53, "learning_rate": 1.3317241379310348e-06, "loss": 0.1099, "step": 13075 }, { "epoch": 4.54, "learning_rate": 1.3144827586206899e-06, "loss": 0.1121, "step": 13100 }, { "epoch": 4.55, "learning_rate": 1.297241379310345e-06, "loss": 0.1055, "step": 13125 }, { "epoch": 4.56, "learning_rate": 1.28e-06, "loss": 0.1046, "step": 13150 }, { "epoch": 4.57, "learning_rate": 1.2627586206896553e-06, "loss": 0.1089, "step": 13175 }, { "epoch": 4.57, "learning_rate": 1.2455172413793104e-06, "loss": 0.1113, "step": 13200 }, { "epoch": 4.58, "learning_rate": 1.2282758620689657e-06, "loss": 0.1089, "step": 13225 }, { "epoch": 4.59, "learning_rate": 1.2110344827586208e-06, "loss": 0.1064, "step": 13250 }, { "epoch": 4.6, "learning_rate": 1.1937931034482759e-06, "loss": 0.1094, "step": 13275 }, { "epoch": 4.61, "learning_rate": 1.1765517241379312e-06, "loss": 0.1073, "step": 13300 }, { "epoch": 4.62, "learning_rate": 1.1593103448275864e-06, "loss": 0.1068, "step": 13325 }, { "epoch": 4.63, "learning_rate": 1.1420689655172415e-06, "loss": 0.1088, "step": 13350 }, { "epoch": 4.63, "learning_rate": 1.1248275862068966e-06, "loss": 0.1032, "step": 13375 }, { "epoch": 4.64, "learning_rate": 1.1075862068965517e-06, "loss": 0.105, "step": 13400 }, { "epoch": 4.65, "learning_rate": 1.090344827586207e-06, "loss": 0.1013, "step": 13425 }, { "epoch": 4.66, "learning_rate": 1.0731034482758623e-06, "loss": 0.1087, "step": 13450 }, { "epoch": 4.67, "learning_rate": 1.0558620689655174e-06, "loss": 0.1017, "step": 13475 }, { "epoch": 4.68, "learning_rate": 1.0386206896551724e-06, "loss": 0.1039, "step": 13500 }, { "epoch": 4.69, "learning_rate": 1.0213793103448277e-06, "loss": 0.1092, "step": 13525 }, { "epoch": 4.7, "learning_rate": 1.0041379310344828e-06, "loss": 0.109, "step": 13550 }, { "epoch": 4.7, "learning_rate": 9.868965517241381e-07, "loss": 0.1039, "step": 13575 }, { "epoch": 4.71, "learning_rate": 9.696551724137932e-07, "loss": 0.1065, "step": 13600 }, { "epoch": 4.72, "learning_rate": 9.524137931034484e-07, "loss": 0.1074, "step": 13625 }, { "epoch": 4.73, "learning_rate": 9.351724137931036e-07, "loss": 0.1126, "step": 13650 }, { "epoch": 4.74, "learning_rate": 9.179310344827587e-07, "loss": 0.1046, "step": 13675 }, { "epoch": 4.75, "learning_rate": 9.006896551724138e-07, "loss": 0.1041, "step": 13700 }, { "epoch": 4.76, "learning_rate": 8.834482758620691e-07, "loss": 0.1054, "step": 13725 }, { "epoch": 4.76, "learning_rate": 8.662068965517242e-07, "loss": 0.1049, "step": 13750 }, { "epoch": 4.77, "learning_rate": 8.489655172413794e-07, "loss": 0.1088, "step": 13775 }, { "epoch": 4.78, "learning_rate": 8.317241379310345e-07, "loss": 0.1006, "step": 13800 }, { "epoch": 4.79, "learning_rate": 8.144827586206898e-07, "loss": 0.1044, "step": 13825 }, { "epoch": 4.8, "learning_rate": 7.97241379310345e-07, "loss": 0.1073, "step": 13850 }, { "epoch": 4.81, "learning_rate": 7.8e-07, "loss": 0.0954, "step": 13875 }, { "epoch": 4.82, "learning_rate": 7.627586206896552e-07, "loss": 0.1054, "step": 13900 }, { "epoch": 4.83, "learning_rate": 7.455172413793104e-07, "loss": 0.1045, "step": 13925 }, { "epoch": 4.83, "learning_rate": 7.282758620689656e-07, "loss": 0.1054, "step": 13950 }, { "epoch": 4.84, "learning_rate": 7.110344827586208e-07, "loss": 0.1069, "step": 13975 }, { "epoch": 4.85, "learning_rate": 6.937931034482759e-07, "loss": 0.1033, "step": 14000 }, { "epoch": 4.85, "eval_cer": 9.480138080272038, "eval_loss": 0.24019023776054382, "eval_runtime": 721.8637, "eval_samples_per_second": 2.613, "eval_steps_per_second": 0.082, "eval_wer": 26.70963513218265, "step": 14000 }, { "epoch": 4.86, "learning_rate": 6.765517241379311e-07, "loss": 0.1075, "step": 14025 }, { "epoch": 4.87, "learning_rate": 6.593103448275863e-07, "loss": 0.1036, "step": 14050 }, { "epoch": 4.88, "learning_rate": 6.420689655172414e-07, "loss": 0.1085, "step": 14075 }, { "epoch": 4.89, "learning_rate": 6.248275862068965e-07, "loss": 0.1035, "step": 14100 }, { "epoch": 4.89, "learning_rate": 6.075862068965518e-07, "loss": 0.1008, "step": 14125 }, { "epoch": 4.9, "learning_rate": 5.903448275862069e-07, "loss": 0.1072, "step": 14150 }, { "epoch": 4.91, "learning_rate": 5.731034482758621e-07, "loss": 0.111, "step": 14175 }, { "epoch": 4.92, "learning_rate": 5.558620689655173e-07, "loss": 0.1123, "step": 14200 }, { "epoch": 4.93, "learning_rate": 5.386206896551725e-07, "loss": 0.1086, "step": 14225 }, { "epoch": 4.94, "learning_rate": 5.213793103448275e-07, "loss": 0.1051, "step": 14250 }, { "epoch": 4.95, "learning_rate": 5.041379310344828e-07, "loss": 0.1044, "step": 14275 }, { "epoch": 4.95, "learning_rate": 4.868965517241379e-07, "loss": 0.1096, "step": 14300 }, { "epoch": 4.96, "learning_rate": 4.6965517241379317e-07, "loss": 0.1079, "step": 14325 }, { "epoch": 4.97, "learning_rate": 4.524137931034483e-07, "loss": 0.1071, "step": 14350 }, { "epoch": 4.98, "learning_rate": 4.351724137931035e-07, "loss": 0.1099, "step": 14375 }, { "epoch": 4.99, "learning_rate": 4.179310344827586e-07, "loss": 0.1056, "step": 14400 }, { "epoch": 5.0, "learning_rate": 4.006896551724138e-07, "loss": 0.1058, "step": 14425 }, { "epoch": 5.01, "learning_rate": 3.8344827586206895e-07, "loss": 0.092, "step": 14450 }, { "epoch": 5.02, "learning_rate": 3.662068965517242e-07, "loss": 0.0914, "step": 14475 }, { "epoch": 5.02, "learning_rate": 3.489655172413793e-07, "loss": 0.0928, "step": 14500 }, { "epoch": 5.03, "learning_rate": 3.317241379310345e-07, "loss": 0.0895, "step": 14525 }, { "epoch": 5.04, "learning_rate": 3.1448275862068964e-07, "loss": 0.0953, "step": 14550 }, { "epoch": 5.05, "learning_rate": 2.9724137931034483e-07, "loss": 0.0949, "step": 14575 }, { "epoch": 5.06, "learning_rate": 2.8e-07, "loss": 0.0941, "step": 14600 }, { "epoch": 5.07, "learning_rate": 2.627586206896552e-07, "loss": 0.0933, "step": 14625 }, { "epoch": 5.08, "learning_rate": 2.455172413793104e-07, "loss": 0.0991, "step": 14650 }, { "epoch": 5.08, "learning_rate": 2.2827586206896553e-07, "loss": 0.0933, "step": 14675 }, { "epoch": 5.09, "learning_rate": 2.1103448275862072e-07, "loss": 0.094, "step": 14700 }, { "epoch": 5.1, "learning_rate": 1.9379310344827588e-07, "loss": 0.0927, "step": 14725 }, { "epoch": 5.11, "learning_rate": 1.7655172413793107e-07, "loss": 0.0874, "step": 14750 }, { "epoch": 5.12, "learning_rate": 1.5931034482758623e-07, "loss": 0.0942, "step": 14775 }, { "epoch": 5.13, "learning_rate": 1.420689655172414e-07, "loss": 0.0921, "step": 14800 }, { "epoch": 5.14, "learning_rate": 1.2482758620689658e-07, "loss": 0.0914, "step": 14825 }, { "epoch": 5.15, "learning_rate": 1.0827586206896553e-07, "loss": 0.0993, "step": 14850 }, { "epoch": 5.15, "learning_rate": 9.10344827586207e-08, "loss": 0.0934, "step": 14875 }, { "epoch": 5.16, "learning_rate": 7.379310344827586e-08, "loss": 0.0889, "step": 14900 }, { "epoch": 5.17, "learning_rate": 5.655172413793104e-08, "loss": 0.0944, "step": 14925 }, { "epoch": 5.18, "learning_rate": 3.9310344827586205e-08, "loss": 0.0941, "step": 14950 }, { "epoch": 5.19, "learning_rate": 2.2068965517241383e-08, "loss": 0.0973, "step": 14975 }, { "epoch": 5.2, "learning_rate": 4.827586206896552e-09, "loss": 0.097, "step": 15000 }, { "epoch": 5.2, "eval_cer": 9.46811616603981, "eval_loss": 0.24208466708660126, "eval_runtime": 723.4063, "eval_samples_per_second": 2.607, "eval_steps_per_second": 0.082, "eval_wer": 26.518461874590344, "step": 15000 }, { "epoch": 5.2, "step": 15000, "total_flos": 1.385080058105856e+20, "train_loss": 0.2266951851050059, "train_runtime": 44464.3629, "train_samples_per_second": 10.795, "train_steps_per_second": 0.337 }, { "epoch": 5.2, "eval_cer": 9.46811616603981, "eval_loss": 0.24208466708660126, "eval_runtime": 722.7359, "eval_samples_per_second": 2.61, "eval_steps_per_second": 0.082, "eval_wer": 26.518461874590344, "step": 15000 } ], "max_steps": 15000, "num_train_epochs": 6, "total_flos": 1.385080058105856e+20, "trial_name": null, "trial_params": null }