diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,9990 @@ +{ + "best_metric": 7.451735188369549, + "best_model_checkpoint": "./checkpoint-13000", + "epoch": 401.50564617314933, + "eval_steps": 1000, + "global_step": 40000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.25, + "learning_rate": 5.000000000000001e-07, + "loss": 1.0482, + "step": 25 + }, + { + "epoch": 0.5, + "learning_rate": 9.800000000000001e-07, + "loss": 0.6567, + "step": 50 + }, + { + "epoch": 0.75, + "learning_rate": 1.48e-06, + "loss": 0.477, + "step": 75 + }, + { + "epoch": 1.0, + "learning_rate": 1.98e-06, + "loss": 0.4014, + "step": 100 + }, + { + "epoch": 1.25, + "learning_rate": 2.4800000000000004e-06, + "loss": 0.3432, + "step": 125 + }, + { + "epoch": 1.51, + "learning_rate": 2.9800000000000003e-06, + "loss": 0.301, + "step": 150 + }, + { + "epoch": 1.76, + "learning_rate": 3.48e-06, + "loss": 0.2734, + "step": 175 + }, + { + "epoch": 2.01, + "learning_rate": 3.980000000000001e-06, + "loss": 0.2445, + "step": 200 + }, + { + "epoch": 2.26, + "learning_rate": 4.48e-06, + "loss": 0.1973, + "step": 225 + }, + { + "epoch": 2.51, + "learning_rate": 4.960000000000001e-06, + "loss": 0.1906, + "step": 250 + }, + { + "epoch": 2.76, + "learning_rate": 5.460000000000001e-06, + "loss": 0.1791, + "step": 275 + }, + { + "epoch": 3.01, + "learning_rate": 5.9600000000000005e-06, + "loss": 0.1715, + "step": 300 + }, + { + "epoch": 3.26, + "learning_rate": 6.460000000000001e-06, + "loss": 0.1248, + "step": 325 + }, + { + "epoch": 3.51, + "learning_rate": 6.96e-06, + "loss": 0.1223, + "step": 350 + }, + { + "epoch": 3.76, + "learning_rate": 7.4600000000000006e-06, + "loss": 0.1242, + "step": 375 + }, + { + "epoch": 4.02, + "learning_rate": 7.960000000000002e-06, + "loss": 0.1217, + "step": 400 + }, + { + "epoch": 4.27, + "learning_rate": 8.46e-06, + "loss": 0.0816, + "step": 425 + }, + { + "epoch": 4.52, + "learning_rate": 8.96e-06, + "loss": 0.0817, + "step": 450 + }, + { + "epoch": 4.77, + "learning_rate": 9.460000000000001e-06, + "loss": 0.0858, + "step": 475 + }, + { + "epoch": 5.02, + "learning_rate": 9.960000000000001e-06, + "loss": 0.0818, + "step": 500 + }, + { + "epoch": 5.27, + "learning_rate": 9.994177215189874e-06, + "loss": 0.0518, + "step": 525 + }, + { + "epoch": 5.52, + "learning_rate": 9.987848101265824e-06, + "loss": 0.054, + "step": 550 + }, + { + "epoch": 5.77, + "learning_rate": 9.981518987341773e-06, + "loss": 0.0552, + "step": 575 + }, + { + "epoch": 6.02, + "learning_rate": 9.975189873417723e-06, + "loss": 0.0543, + "step": 600 + }, + { + "epoch": 6.27, + "learning_rate": 9.968860759493672e-06, + "loss": 0.0344, + "step": 625 + }, + { + "epoch": 6.52, + "learning_rate": 9.962531645569621e-06, + "loss": 0.0353, + "step": 650 + }, + { + "epoch": 6.78, + "learning_rate": 9.956202531645571e-06, + "loss": 0.0365, + "step": 675 + }, + { + "epoch": 7.03, + "learning_rate": 9.94987341772152e-06, + "loss": 0.0332, + "step": 700 + }, + { + "epoch": 7.28, + "learning_rate": 9.943544303797468e-06, + "loss": 0.021, + "step": 725 + }, + { + "epoch": 7.53, + "learning_rate": 9.937215189873417e-06, + "loss": 0.0227, + "step": 750 + }, + { + "epoch": 7.78, + "learning_rate": 9.930886075949367e-06, + "loss": 0.0216, + "step": 775 + }, + { + "epoch": 8.03, + "learning_rate": 9.924556962025316e-06, + "loss": 0.0223, + "step": 800 + }, + { + "epoch": 8.28, + "learning_rate": 9.918227848101266e-06, + "loss": 0.0145, + "step": 825 + }, + { + "epoch": 8.53, + "learning_rate": 9.911898734177215e-06, + "loss": 0.0146, + "step": 850 + }, + { + "epoch": 8.78, + "learning_rate": 9.905569620253165e-06, + "loss": 0.0157, + "step": 875 + }, + { + "epoch": 9.03, + "learning_rate": 9.899240506329116e-06, + "loss": 0.0157, + "step": 900 + }, + { + "epoch": 9.28, + "learning_rate": 9.892911392405065e-06, + "loss": 0.0105, + "step": 925 + }, + { + "epoch": 9.54, + "learning_rate": 9.886582278481015e-06, + "loss": 0.0101, + "step": 950 + }, + { + "epoch": 9.79, + "learning_rate": 9.880253164556964e-06, + "loss": 0.0109, + "step": 975 + }, + { + "epoch": 10.04, + "learning_rate": 9.873924050632913e-06, + "loss": 0.0112, + "step": 1000 + }, + { + "epoch": 10.04, + "eval_loss": 0.21815913915634155, + "eval_runtime": 2849.5314, + "eval_samples_per_second": 4.472, + "eval_steps_per_second": 0.28, + "eval_wer": 10.157104892918555, + "step": 1000 + }, + { + "epoch": 10.29, + "learning_rate": 9.867594936708861e-06, + "loss": 0.008, + "step": 1025 + }, + { + "epoch": 10.54, + "learning_rate": 9.86126582278481e-06, + "loss": 0.0089, + "step": 1050 + }, + { + "epoch": 10.79, + "learning_rate": 9.85493670886076e-06, + "loss": 0.0098, + "step": 1075 + }, + { + "epoch": 11.04, + "learning_rate": 9.84860759493671e-06, + "loss": 0.008, + "step": 1100 + }, + { + "epoch": 11.29, + "learning_rate": 9.842278481012659e-06, + "loss": 0.0055, + "step": 1125 + }, + { + "epoch": 11.54, + "learning_rate": 9.835949367088608e-06, + "loss": 0.0063, + "step": 1150 + }, + { + "epoch": 11.79, + "learning_rate": 9.829620253164558e-06, + "loss": 0.0067, + "step": 1175 + }, + { + "epoch": 12.05, + "learning_rate": 9.823291139240507e-06, + "loss": 0.0069, + "step": 1200 + }, + { + "epoch": 12.3, + "learning_rate": 9.816962025316457e-06, + "loss": 0.006, + "step": 1225 + }, + { + "epoch": 12.55, + "learning_rate": 9.810632911392406e-06, + "loss": 0.0067, + "step": 1250 + }, + { + "epoch": 12.8, + "learning_rate": 9.804303797468355e-06, + "loss": 0.0067, + "step": 1275 + }, + { + "epoch": 13.05, + "learning_rate": 9.797974683544305e-06, + "loss": 0.0061, + "step": 1300 + }, + { + "epoch": 13.3, + "learning_rate": 9.791645569620254e-06, + "loss": 0.0049, + "step": 1325 + }, + { + "epoch": 13.55, + "learning_rate": 9.785316455696204e-06, + "loss": 0.0043, + "step": 1350 + }, + { + "epoch": 13.8, + "learning_rate": 9.778987341772153e-06, + "loss": 0.0048, + "step": 1375 + }, + { + "epoch": 14.05, + "learning_rate": 9.772658227848103e-06, + "loss": 0.0057, + "step": 1400 + }, + { + "epoch": 14.3, + "learning_rate": 9.766329113924052e-06, + "loss": 0.0045, + "step": 1425 + }, + { + "epoch": 14.55, + "learning_rate": 9.760000000000001e-06, + "loss": 0.0047, + "step": 1450 + }, + { + "epoch": 14.81, + "learning_rate": 9.75367088607595e-06, + "loss": 0.0057, + "step": 1475 + }, + { + "epoch": 15.06, + "learning_rate": 9.747341772151899e-06, + "loss": 0.0061, + "step": 1500 + }, + { + "epoch": 15.31, + "learning_rate": 9.741012658227848e-06, + "loss": 0.0046, + "step": 1525 + }, + { + "epoch": 15.56, + "learning_rate": 9.734683544303797e-06, + "loss": 0.005, + "step": 1550 + }, + { + "epoch": 15.81, + "learning_rate": 9.728354430379747e-06, + "loss": 0.0047, + "step": 1575 + }, + { + "epoch": 16.06, + "learning_rate": 9.722025316455696e-06, + "loss": 0.0053, + "step": 1600 + }, + { + "epoch": 16.31, + "learning_rate": 9.715696202531646e-06, + "loss": 0.0042, + "step": 1625 + }, + { + "epoch": 16.56, + "learning_rate": 9.709367088607595e-06, + "loss": 0.0035, + "step": 1650 + }, + { + "epoch": 16.81, + "learning_rate": 9.703037974683545e-06, + "loss": 0.0034, + "step": 1675 + }, + { + "epoch": 17.06, + "learning_rate": 9.696708860759494e-06, + "loss": 0.0037, + "step": 1700 + }, + { + "epoch": 17.31, + "learning_rate": 9.690379746835445e-06, + "loss": 0.0033, + "step": 1725 + }, + { + "epoch": 17.57, + "learning_rate": 9.684050632911393e-06, + "loss": 0.0034, + "step": 1750 + }, + { + "epoch": 17.82, + "learning_rate": 9.677721518987342e-06, + "loss": 0.0033, + "step": 1775 + }, + { + "epoch": 18.07, + "learning_rate": 9.671392405063292e-06, + "loss": 0.0033, + "step": 1800 + }, + { + "epoch": 18.32, + "learning_rate": 9.665063291139241e-06, + "loss": 0.0033, + "step": 1825 + }, + { + "epoch": 18.57, + "learning_rate": 9.65873417721519e-06, + "loss": 0.0035, + "step": 1850 + }, + { + "epoch": 18.82, + "learning_rate": 9.65240506329114e-06, + "loss": 0.0036, + "step": 1875 + }, + { + "epoch": 19.07, + "learning_rate": 9.64607594936709e-06, + "loss": 0.0045, + "step": 1900 + }, + { + "epoch": 19.32, + "learning_rate": 9.639746835443039e-06, + "loss": 0.0041, + "step": 1925 + }, + { + "epoch": 19.57, + "learning_rate": 9.633417721518988e-06, + "loss": 0.0041, + "step": 1950 + }, + { + "epoch": 19.82, + "learning_rate": 9.627088607594938e-06, + "loss": 0.0039, + "step": 1975 + }, + { + "epoch": 20.08, + "learning_rate": 9.620759493670887e-06, + "loss": 0.0052, + "step": 2000 + }, + { + "epoch": 20.08, + "eval_loss": 0.23722027242183685, + "eval_runtime": 2841.2175, + "eval_samples_per_second": 4.485, + "eval_steps_per_second": 0.281, + "eval_wer": 9.627559793653274, + "step": 2000 + }, + { + "epoch": 20.33, + "learning_rate": 9.614430379746837e-06, + "loss": 0.0039, + "step": 2025 + }, + { + "epoch": 20.58, + "learning_rate": 9.608101265822786e-06, + "loss": 0.0037, + "step": 2050 + }, + { + "epoch": 20.83, + "learning_rate": 9.601772151898735e-06, + "loss": 0.004, + "step": 2075 + }, + { + "epoch": 21.08, + "learning_rate": 9.595443037974685e-06, + "loss": 0.0044, + "step": 2100 + }, + { + "epoch": 21.33, + "learning_rate": 9.589113924050634e-06, + "loss": 0.0032, + "step": 2125 + }, + { + "epoch": 21.58, + "learning_rate": 9.582784810126584e-06, + "loss": 0.0036, + "step": 2150 + }, + { + "epoch": 21.83, + "learning_rate": 9.576455696202533e-06, + "loss": 0.0038, + "step": 2175 + }, + { + "epoch": 22.08, + "learning_rate": 9.570126582278481e-06, + "loss": 0.0036, + "step": 2200 + }, + { + "epoch": 22.33, + "learning_rate": 9.56379746835443e-06, + "loss": 0.0035, + "step": 2225 + }, + { + "epoch": 22.58, + "learning_rate": 9.55746835443038e-06, + "loss": 0.0042, + "step": 2250 + }, + { + "epoch": 22.84, + "learning_rate": 9.551139240506329e-06, + "loss": 0.0039, + "step": 2275 + }, + { + "epoch": 23.09, + "learning_rate": 9.544810126582279e-06, + "loss": 0.0038, + "step": 2300 + }, + { + "epoch": 23.34, + "learning_rate": 9.538481012658228e-06, + "loss": 0.0038, + "step": 2325 + }, + { + "epoch": 23.59, + "learning_rate": 9.532151898734177e-06, + "loss": 0.0037, + "step": 2350 + }, + { + "epoch": 23.84, + "learning_rate": 9.525822784810127e-06, + "loss": 0.005, + "step": 2375 + }, + { + "epoch": 24.09, + "learning_rate": 9.519493670886076e-06, + "loss": 0.0043, + "step": 2400 + }, + { + "epoch": 24.34, + "learning_rate": 9.513164556962026e-06, + "loss": 0.0035, + "step": 2425 + }, + { + "epoch": 24.59, + "learning_rate": 9.506835443037975e-06, + "loss": 0.0032, + "step": 2450 + }, + { + "epoch": 24.84, + "learning_rate": 9.500506329113925e-06, + "loss": 0.0041, + "step": 2475 + }, + { + "epoch": 25.09, + "learning_rate": 9.494177215189874e-06, + "loss": 0.0029, + "step": 2500 + }, + { + "epoch": 25.35, + "learning_rate": 9.487848101265823e-06, + "loss": 0.0023, + "step": 2525 + }, + { + "epoch": 25.6, + "learning_rate": 9.481518987341773e-06, + "loss": 0.0022, + "step": 2550 + }, + { + "epoch": 25.85, + "learning_rate": 9.475189873417722e-06, + "loss": 0.003, + "step": 2575 + }, + { + "epoch": 26.1, + "learning_rate": 9.468860759493672e-06, + "loss": 0.0037, + "step": 2600 + }, + { + "epoch": 26.35, + "learning_rate": 9.462531645569621e-06, + "loss": 0.0027, + "step": 2625 + }, + { + "epoch": 26.6, + "learning_rate": 9.45620253164557e-06, + "loss": 0.0032, + "step": 2650 + }, + { + "epoch": 26.85, + "learning_rate": 9.44987341772152e-06, + "loss": 0.0033, + "step": 2675 + }, + { + "epoch": 27.1, + "learning_rate": 9.44354430379747e-06, + "loss": 0.0037, + "step": 2700 + }, + { + "epoch": 27.35, + "learning_rate": 9.437215189873419e-06, + "loss": 0.0031, + "step": 2725 + }, + { + "epoch": 27.6, + "learning_rate": 9.430886075949368e-06, + "loss": 0.0027, + "step": 2750 + }, + { + "epoch": 27.85, + "learning_rate": 9.424556962025318e-06, + "loss": 0.003, + "step": 2775 + }, + { + "epoch": 28.11, + "learning_rate": 9.418227848101267e-06, + "loss": 0.003, + "step": 2800 + }, + { + "epoch": 28.36, + "learning_rate": 9.411898734177217e-06, + "loss": 0.0025, + "step": 2825 + }, + { + "epoch": 28.61, + "learning_rate": 9.405569620253166e-06, + "loss": 0.0023, + "step": 2850 + }, + { + "epoch": 28.86, + "learning_rate": 9.399240506329115e-06, + "loss": 0.0027, + "step": 2875 + }, + { + "epoch": 29.11, + "learning_rate": 9.392911392405065e-06, + "loss": 0.0023, + "step": 2900 + }, + { + "epoch": 29.36, + "learning_rate": 9.386582278481014e-06, + "loss": 0.0022, + "step": 2925 + }, + { + "epoch": 29.61, + "learning_rate": 9.380253164556962e-06, + "loss": 0.0018, + "step": 2950 + }, + { + "epoch": 29.86, + "learning_rate": 9.373924050632911e-06, + "loss": 0.0018, + "step": 2975 + }, + { + "epoch": 30.11, + "learning_rate": 9.367594936708861e-06, + "loss": 0.0017, + "step": 3000 + }, + { + "epoch": 30.11, + "eval_loss": 0.24174295365810394, + "eval_runtime": 2833.8733, + "eval_samples_per_second": 4.496, + "eval_steps_per_second": 0.281, + "eval_wer": 9.014967953728311, + "step": 3000 + }, + { + "epoch": 30.36, + "learning_rate": 9.36126582278481e-06, + "loss": 0.0024, + "step": 3025 + }, + { + "epoch": 30.61, + "learning_rate": 9.355189873417723e-06, + "loss": 0.002, + "step": 3050 + }, + { + "epoch": 30.87, + "learning_rate": 9.348860759493671e-06, + "loss": 0.0025, + "step": 3075 + }, + { + "epoch": 31.12, + "learning_rate": 9.34253164556962e-06, + "loss": 0.0026, + "step": 3100 + }, + { + "epoch": 31.37, + "learning_rate": 9.33620253164557e-06, + "loss": 0.0024, + "step": 3125 + }, + { + "epoch": 31.62, + "learning_rate": 9.32987341772152e-06, + "loss": 0.0019, + "step": 3150 + }, + { + "epoch": 31.87, + "learning_rate": 9.323544303797469e-06, + "loss": 0.003, + "step": 3175 + }, + { + "epoch": 32.12, + "learning_rate": 9.317215189873418e-06, + "loss": 0.0026, + "step": 3200 + }, + { + "epoch": 32.37, + "learning_rate": 9.310886075949368e-06, + "loss": 0.003, + "step": 3225 + }, + { + "epoch": 32.62, + "learning_rate": 9.304556962025317e-06, + "loss": 0.0027, + "step": 3250 + }, + { + "epoch": 32.87, + "learning_rate": 9.298227848101267e-06, + "loss": 0.0032, + "step": 3275 + }, + { + "epoch": 33.12, + "learning_rate": 9.291898734177216e-06, + "loss": 0.0022, + "step": 3300 + }, + { + "epoch": 33.38, + "learning_rate": 9.285569620253165e-06, + "loss": 0.0018, + "step": 3325 + }, + { + "epoch": 33.63, + "learning_rate": 9.279240506329115e-06, + "loss": 0.0019, + "step": 3350 + }, + { + "epoch": 33.88, + "learning_rate": 9.272911392405064e-06, + "loss": 0.0019, + "step": 3375 + }, + { + "epoch": 34.13, + "learning_rate": 9.266582278481014e-06, + "loss": 0.0017, + "step": 3400 + }, + { + "epoch": 34.38, + "learning_rate": 9.260253164556963e-06, + "loss": 0.0015, + "step": 3425 + }, + { + "epoch": 34.63, + "learning_rate": 9.253924050632913e-06, + "loss": 0.0011, + "step": 3450 + }, + { + "epoch": 34.88, + "learning_rate": 9.247594936708862e-06, + "loss": 0.0015, + "step": 3475 + }, + { + "epoch": 35.13, + "learning_rate": 9.241265822784811e-06, + "loss": 0.0018, + "step": 3500 + }, + { + "epoch": 35.38, + "learning_rate": 9.23493670886076e-06, + "loss": 0.0014, + "step": 3525 + }, + { + "epoch": 35.63, + "learning_rate": 9.228607594936709e-06, + "loss": 0.002, + "step": 3550 + }, + { + "epoch": 35.88, + "learning_rate": 9.222278481012658e-06, + "loss": 0.0024, + "step": 3575 + }, + { + "epoch": 36.14, + "learning_rate": 9.215949367088607e-06, + "loss": 0.0026, + "step": 3600 + }, + { + "epoch": 36.39, + "learning_rate": 9.209620253164557e-06, + "loss": 0.0029, + "step": 3625 + }, + { + "epoch": 36.64, + "learning_rate": 9.203291139240506e-06, + "loss": 0.0032, + "step": 3650 + }, + { + "epoch": 36.89, + "learning_rate": 9.196962025316456e-06, + "loss": 0.0033, + "step": 3675 + }, + { + "epoch": 37.14, + "learning_rate": 9.190632911392405e-06, + "loss": 0.0036, + "step": 3700 + }, + { + "epoch": 37.39, + "learning_rate": 9.184303797468355e-06, + "loss": 0.0032, + "step": 3725 + }, + { + "epoch": 37.64, + "learning_rate": 9.177974683544306e-06, + "loss": 0.0027, + "step": 3750 + }, + { + "epoch": 37.89, + "learning_rate": 9.171645569620255e-06, + "loss": 0.0033, + "step": 3775 + }, + { + "epoch": 38.14, + "learning_rate": 9.165316455696205e-06, + "loss": 0.0032, + "step": 3800 + }, + { + "epoch": 38.39, + "learning_rate": 9.158987341772152e-06, + "loss": 0.0023, + "step": 3825 + }, + { + "epoch": 38.64, + "learning_rate": 9.152658227848102e-06, + "loss": 0.0024, + "step": 3850 + }, + { + "epoch": 38.9, + "learning_rate": 9.146329113924051e-06, + "loss": 0.003, + "step": 3875 + }, + { + "epoch": 39.15, + "learning_rate": 9.14e-06, + "loss": 0.0024, + "step": 3900 + }, + { + "epoch": 39.4, + "learning_rate": 9.13367088607595e-06, + "loss": 0.0018, + "step": 3925 + }, + { + "epoch": 39.65, + "learning_rate": 9.1273417721519e-06, + "loss": 0.0023, + "step": 3950 + }, + { + "epoch": 39.9, + "learning_rate": 9.121012658227849e-06, + "loss": 0.0023, + "step": 3975 + }, + { + "epoch": 40.15, + "learning_rate": 9.114683544303798e-06, + "loss": 0.0022, + "step": 4000 + }, + { + "epoch": 40.15, + "eval_loss": 0.2340633124113083, + "eval_runtime": 2834.4887, + "eval_samples_per_second": 4.495, + "eval_steps_per_second": 0.281, + "eval_wer": 8.893817414413006, + "step": 4000 + }, + { + "epoch": 40.4, + "learning_rate": 9.108354430379748e-06, + "loss": 0.0021, + "step": 4025 + }, + { + "epoch": 40.65, + "learning_rate": 9.102025316455697e-06, + "loss": 0.002, + "step": 4050 + }, + { + "epoch": 40.9, + "learning_rate": 9.095696202531647e-06, + "loss": 0.0022, + "step": 4075 + }, + { + "epoch": 41.15, + "learning_rate": 9.089367088607596e-06, + "loss": 0.0021, + "step": 4100 + }, + { + "epoch": 41.41, + "learning_rate": 9.083037974683545e-06, + "loss": 0.0023, + "step": 4125 + }, + { + "epoch": 41.66, + "learning_rate": 9.076708860759495e-06, + "loss": 0.0025, + "step": 4150 + }, + { + "epoch": 41.91, + "learning_rate": 9.070379746835444e-06, + "loss": 0.0028, + "step": 4175 + }, + { + "epoch": 42.16, + "learning_rate": 9.064050632911394e-06, + "loss": 0.0028, + "step": 4200 + }, + { + "epoch": 42.41, + "learning_rate": 9.057721518987343e-06, + "loss": 0.0022, + "step": 4225 + }, + { + "epoch": 42.66, + "learning_rate": 9.051392405063293e-06, + "loss": 0.0017, + "step": 4250 + }, + { + "epoch": 42.91, + "learning_rate": 9.04506329113924e-06, + "loss": 0.0022, + "step": 4275 + }, + { + "epoch": 43.16, + "learning_rate": 9.03873417721519e-06, + "loss": 0.0019, + "step": 4300 + }, + { + "epoch": 43.41, + "learning_rate": 9.03240506329114e-06, + "loss": 0.0014, + "step": 4325 + }, + { + "epoch": 43.66, + "learning_rate": 9.026075949367089e-06, + "loss": 0.0025, + "step": 4350 + }, + { + "epoch": 43.91, + "learning_rate": 9.019746835443038e-06, + "loss": 0.0021, + "step": 4375 + }, + { + "epoch": 44.17, + "learning_rate": 9.013417721518987e-06, + "loss": 0.0021, + "step": 4400 + }, + { + "epoch": 44.42, + "learning_rate": 9.007088607594937e-06, + "loss": 0.0022, + "step": 4425 + }, + { + "epoch": 44.67, + "learning_rate": 9.000759493670886e-06, + "loss": 0.0019, + "step": 4450 + }, + { + "epoch": 44.92, + "learning_rate": 8.994430379746836e-06, + "loss": 0.0023, + "step": 4475 + }, + { + "epoch": 45.17, + "learning_rate": 8.988101265822785e-06, + "loss": 0.002, + "step": 4500 + }, + { + "epoch": 45.42, + "learning_rate": 8.981772151898735e-06, + "loss": 0.0018, + "step": 4525 + }, + { + "epoch": 45.67, + "learning_rate": 8.975443037974684e-06, + "loss": 0.0025, + "step": 4550 + }, + { + "epoch": 45.92, + "learning_rate": 8.969113924050633e-06, + "loss": 0.0021, + "step": 4575 + }, + { + "epoch": 46.17, + "learning_rate": 8.962784810126583e-06, + "loss": 0.0018, + "step": 4600 + }, + { + "epoch": 46.42, + "learning_rate": 8.956455696202532e-06, + "loss": 0.0021, + "step": 4625 + }, + { + "epoch": 46.68, + "learning_rate": 8.950126582278482e-06, + "loss": 0.0019, + "step": 4650 + }, + { + "epoch": 46.93, + "learning_rate": 8.943797468354431e-06, + "loss": 0.0024, + "step": 4675 + }, + { + "epoch": 47.18, + "learning_rate": 8.93746835443038e-06, + "loss": 0.0019, + "step": 4700 + }, + { + "epoch": 47.43, + "learning_rate": 8.93113924050633e-06, + "loss": 0.0021, + "step": 4725 + }, + { + "epoch": 47.68, + "learning_rate": 8.92481012658228e-06, + "loss": 0.0026, + "step": 4750 + }, + { + "epoch": 47.93, + "learning_rate": 8.918481012658229e-06, + "loss": 0.002, + "step": 4775 + }, + { + "epoch": 48.18, + "learning_rate": 8.912151898734178e-06, + "loss": 0.0023, + "step": 4800 + }, + { + "epoch": 48.43, + "learning_rate": 8.905822784810128e-06, + "loss": 0.002, + "step": 4825 + }, + { + "epoch": 48.68, + "learning_rate": 8.899493670886077e-06, + "loss": 0.0021, + "step": 4850 + }, + { + "epoch": 48.93, + "learning_rate": 8.893164556962027e-06, + "loss": 0.0022, + "step": 4875 + }, + { + "epoch": 49.18, + "learning_rate": 8.886835443037976e-06, + "loss": 0.0015, + "step": 4900 + }, + { + "epoch": 49.44, + "learning_rate": 8.880506329113925e-06, + "loss": 0.0014, + "step": 4925 + }, + { + "epoch": 49.69, + "learning_rate": 8.874177215189875e-06, + "loss": 0.002, + "step": 4950 + }, + { + "epoch": 49.94, + "learning_rate": 8.867848101265824e-06, + "loss": 0.0023, + "step": 4975 + }, + { + "epoch": 50.19, + "learning_rate": 8.861518987341772e-06, + "loss": 0.0023, + "step": 5000 + }, + { + "epoch": 50.19, + "eval_loss": 0.24509792029857635, + "eval_runtime": 2843.0455, + "eval_samples_per_second": 4.482, + "eval_steps_per_second": 0.28, + "eval_wer": 8.93876035641707, + "step": 5000 + }, + { + "epoch": 50.44, + "learning_rate": 8.855189873417721e-06, + "loss": 0.002, + "step": 5025 + }, + { + "epoch": 50.69, + "learning_rate": 8.848860759493671e-06, + "loss": 0.002, + "step": 5050 + }, + { + "epoch": 50.94, + "learning_rate": 8.84253164556962e-06, + "loss": 0.0021, + "step": 5075 + }, + { + "epoch": 51.19, + "learning_rate": 8.83620253164557e-06, + "loss": 0.0018, + "step": 5100 + }, + { + "epoch": 51.44, + "learning_rate": 8.829873417721519e-06, + "loss": 0.0017, + "step": 5125 + }, + { + "epoch": 51.69, + "learning_rate": 8.823544303797469e-06, + "loss": 0.0023, + "step": 5150 + }, + { + "epoch": 51.94, + "learning_rate": 8.817215189873418e-06, + "loss": 0.0023, + "step": 5175 + }, + { + "epoch": 52.2, + "learning_rate": 8.810886075949367e-06, + "loss": 0.0015, + "step": 5200 + }, + { + "epoch": 52.45, + "learning_rate": 8.804556962025317e-06, + "loss": 0.0023, + "step": 5225 + }, + { + "epoch": 52.7, + "learning_rate": 8.798227848101266e-06, + "loss": 0.0022, + "step": 5250 + }, + { + "epoch": 52.95, + "learning_rate": 8.791898734177216e-06, + "loss": 0.0028, + "step": 5275 + }, + { + "epoch": 53.2, + "learning_rate": 8.785569620253165e-06, + "loss": 0.0019, + "step": 5300 + }, + { + "epoch": 53.45, + "learning_rate": 8.779240506329115e-06, + "loss": 0.002, + "step": 5325 + }, + { + "epoch": 53.7, + "learning_rate": 8.772911392405064e-06, + "loss": 0.002, + "step": 5350 + }, + { + "epoch": 53.95, + "learning_rate": 8.766582278481013e-06, + "loss": 0.0019, + "step": 5375 + }, + { + "epoch": 54.2, + "learning_rate": 8.760253164556963e-06, + "loss": 0.0012, + "step": 5400 + }, + { + "epoch": 54.45, + "learning_rate": 8.753924050632912e-06, + "loss": 0.0012, + "step": 5425 + }, + { + "epoch": 54.71, + "learning_rate": 8.747594936708862e-06, + "loss": 0.0013, + "step": 5450 + }, + { + "epoch": 54.96, + "learning_rate": 8.741265822784811e-06, + "loss": 0.0018, + "step": 5475 + }, + { + "epoch": 55.21, + "learning_rate": 8.73493670886076e-06, + "loss": 0.0014, + "step": 5500 + }, + { + "epoch": 55.46, + "learning_rate": 8.72860759493671e-06, + "loss": 0.0017, + "step": 5525 + }, + { + "epoch": 55.71, + "learning_rate": 8.72227848101266e-06, + "loss": 0.0015, + "step": 5550 + }, + { + "epoch": 55.96, + "learning_rate": 8.715949367088609e-06, + "loss": 0.0017, + "step": 5575 + }, + { + "epoch": 56.21, + "learning_rate": 8.709620253164558e-06, + "loss": 0.0009, + "step": 5600 + }, + { + "epoch": 56.46, + "learning_rate": 8.703291139240508e-06, + "loss": 0.0008, + "step": 5625 + }, + { + "epoch": 56.71, + "learning_rate": 8.696962025316457e-06, + "loss": 0.0012, + "step": 5650 + }, + { + "epoch": 56.96, + "learning_rate": 8.690632911392407e-06, + "loss": 0.0014, + "step": 5675 + }, + { + "epoch": 57.21, + "learning_rate": 8.684303797468356e-06, + "loss": 0.0012, + "step": 5700 + }, + { + "epoch": 57.47, + "learning_rate": 8.677974683544305e-06, + "loss": 0.0007, + "step": 5725 + }, + { + "epoch": 57.72, + "learning_rate": 8.671645569620253e-06, + "loss": 0.0008, + "step": 5750 + }, + { + "epoch": 57.97, + "learning_rate": 8.665316455696203e-06, + "loss": 0.0014, + "step": 5775 + }, + { + "epoch": 58.22, + "learning_rate": 8.658987341772152e-06, + "loss": 0.0011, + "step": 5800 + }, + { + "epoch": 58.47, + "learning_rate": 8.652658227848101e-06, + "loss": 0.0009, + "step": 5825 + }, + { + "epoch": 58.72, + "learning_rate": 8.646329113924051e-06, + "loss": 0.0012, + "step": 5850 + }, + { + "epoch": 58.97, + "learning_rate": 8.64e-06, + "loss": 0.0011, + "step": 5875 + }, + { + "epoch": 59.22, + "learning_rate": 8.63367088607595e-06, + "loss": 0.0008, + "step": 5900 + }, + { + "epoch": 59.47, + "learning_rate": 8.627341772151899e-06, + "loss": 0.0006, + "step": 5925 + }, + { + "epoch": 59.72, + "learning_rate": 8.621012658227849e-06, + "loss": 0.0006, + "step": 5950 + }, + { + "epoch": 59.97, + "learning_rate": 8.614683544303798e-06, + "loss": 0.0007, + "step": 5975 + }, + { + "epoch": 60.23, + "learning_rate": 8.608354430379747e-06, + "loss": 0.0006, + "step": 6000 + }, + { + "epoch": 60.23, + "eval_loss": 0.25170740485191345, + "eval_runtime": 2836.0485, + "eval_samples_per_second": 4.493, + "eval_steps_per_second": 0.281, + "eval_wer": 8.416054400500235, + "step": 6000 + }, + { + "epoch": 60.48, + "learning_rate": 8.602025316455697e-06, + "loss": 0.0008, + "step": 6025 + }, + { + "epoch": 60.73, + "learning_rate": 8.595696202531646e-06, + "loss": 0.0007, + "step": 6050 + }, + { + "epoch": 60.98, + "learning_rate": 8.589367088607596e-06, + "loss": 0.001, + "step": 6075 + }, + { + "epoch": 61.23, + "learning_rate": 8.583037974683545e-06, + "loss": 0.0007, + "step": 6100 + }, + { + "epoch": 61.48, + "learning_rate": 8.576708860759495e-06, + "loss": 0.0009, + "step": 6125 + }, + { + "epoch": 61.73, + "learning_rate": 8.570379746835444e-06, + "loss": 0.0015, + "step": 6150 + }, + { + "epoch": 61.98, + "learning_rate": 8.564050632911393e-06, + "loss": 0.0012, + "step": 6175 + }, + { + "epoch": 62.23, + "learning_rate": 8.557721518987341e-06, + "loss": 0.0012, + "step": 6200 + }, + { + "epoch": 62.48, + "learning_rate": 8.551392405063292e-06, + "loss": 0.0016, + "step": 6225 + }, + { + "epoch": 62.74, + "learning_rate": 8.545063291139242e-06, + "loss": 0.0013, + "step": 6250 + }, + { + "epoch": 62.99, + "learning_rate": 8.538734177215191e-06, + "loss": 0.0019, + "step": 6275 + }, + { + "epoch": 63.24, + "learning_rate": 8.53240506329114e-06, + "loss": 0.002, + "step": 6300 + }, + { + "epoch": 63.49, + "learning_rate": 8.52607594936709e-06, + "loss": 0.0026, + "step": 6325 + }, + { + "epoch": 63.74, + "learning_rate": 8.51974683544304e-06, + "loss": 0.0024, + "step": 6350 + }, + { + "epoch": 63.99, + "learning_rate": 8.513417721518989e-06, + "loss": 0.0027, + "step": 6375 + }, + { + "epoch": 64.24, + "learning_rate": 8.507088607594938e-06, + "loss": 0.0018, + "step": 6400 + }, + { + "epoch": 64.49, + "learning_rate": 8.500759493670888e-06, + "loss": 0.0017, + "step": 6425 + }, + { + "epoch": 64.74, + "learning_rate": 8.494430379746837e-06, + "loss": 0.0021, + "step": 6450 + }, + { + "epoch": 64.99, + "learning_rate": 8.488101265822787e-06, + "loss": 0.002, + "step": 6475 + }, + { + "epoch": 65.24, + "learning_rate": 8.481772151898734e-06, + "loss": 0.0017, + "step": 6500 + }, + { + "epoch": 65.5, + "learning_rate": 8.475443037974684e-06, + "loss": 0.002, + "step": 6525 + }, + { + "epoch": 65.75, + "learning_rate": 8.469113924050633e-06, + "loss": 0.0022, + "step": 6550 + }, + { + "epoch": 66.0, + "learning_rate": 8.462784810126583e-06, + "loss": 0.0018, + "step": 6575 + }, + { + "epoch": 66.25, + "learning_rate": 8.456455696202532e-06, + "loss": 0.0015, + "step": 6600 + }, + { + "epoch": 66.5, + "learning_rate": 8.450126582278481e-06, + "loss": 0.0011, + "step": 6625 + }, + { + "epoch": 66.75, + "learning_rate": 8.443797468354431e-06, + "loss": 0.0019, + "step": 6650 + }, + { + "epoch": 67.0, + "learning_rate": 8.43746835443038e-06, + "loss": 0.0018, + "step": 6675 + }, + { + "epoch": 67.25, + "learning_rate": 8.43113924050633e-06, + "loss": 0.0021, + "step": 6700 + }, + { + "epoch": 67.5, + "learning_rate": 8.424810126582279e-06, + "loss": 0.0016, + "step": 6725 + }, + { + "epoch": 67.75, + "learning_rate": 8.418481012658229e-06, + "loss": 0.0016, + "step": 6750 + }, + { + "epoch": 68.01, + "learning_rate": 8.412151898734178e-06, + "loss": 0.0025, + "step": 6775 + }, + { + "epoch": 68.26, + "learning_rate": 8.405822784810127e-06, + "loss": 0.0016, + "step": 6800 + }, + { + "epoch": 68.51, + "learning_rate": 8.399493670886077e-06, + "loss": 0.0017, + "step": 6825 + }, + { + "epoch": 68.76, + "learning_rate": 8.393164556962026e-06, + "loss": 0.0012, + "step": 6850 + }, + { + "epoch": 69.01, + "learning_rate": 8.386835443037976e-06, + "loss": 0.0014, + "step": 6875 + }, + { + "epoch": 69.26, + "learning_rate": 8.380506329113925e-06, + "loss": 0.0008, + "step": 6900 + }, + { + "epoch": 69.51, + "learning_rate": 8.374177215189873e-06, + "loss": 0.0007, + "step": 6925 + }, + { + "epoch": 69.76, + "learning_rate": 8.367848101265822e-06, + "loss": 0.0009, + "step": 6950 + }, + { + "epoch": 70.01, + "learning_rate": 8.361518987341772e-06, + "loss": 0.0009, + "step": 6975 + }, + { + "epoch": 70.26, + "learning_rate": 8.355189873417721e-06, + "loss": 0.0006, + "step": 7000 + }, + { + "epoch": 70.26, + "eval_loss": 0.24993743002414703, + "eval_runtime": 2839.6991, + "eval_samples_per_second": 4.487, + "eval_steps_per_second": 0.281, + "eval_wer": 8.098522745036735, + "step": 7000 + }, + { + "epoch": 70.51, + "learning_rate": 8.34886075949367e-06, + "loss": 0.0006, + "step": 7025 + }, + { + "epoch": 70.77, + "learning_rate": 8.342531645569622e-06, + "loss": 0.0006, + "step": 7050 + }, + { + "epoch": 71.02, + "learning_rate": 8.336202531645571e-06, + "loss": 0.0004, + "step": 7075 + }, + { + "epoch": 71.27, + "learning_rate": 8.32987341772152e-06, + "loss": 0.0005, + "step": 7100 + }, + { + "epoch": 71.52, + "learning_rate": 8.32354430379747e-06, + "loss": 0.0004, + "step": 7125 + }, + { + "epoch": 71.77, + "learning_rate": 8.31721518987342e-06, + "loss": 0.0004, + "step": 7150 + }, + { + "epoch": 72.02, + "learning_rate": 8.310886075949369e-06, + "loss": 0.0005, + "step": 7175 + }, + { + "epoch": 72.27, + "learning_rate": 8.304556962025318e-06, + "loss": 0.001, + "step": 7200 + }, + { + "epoch": 72.52, + "learning_rate": 8.298227848101266e-06, + "loss": 0.0015, + "step": 7225 + }, + { + "epoch": 72.77, + "learning_rate": 8.291898734177215e-06, + "loss": 0.0015, + "step": 7250 + }, + { + "epoch": 73.02, + "learning_rate": 8.285569620253165e-06, + "loss": 0.0015, + "step": 7275 + }, + { + "epoch": 73.27, + "learning_rate": 8.279240506329114e-06, + "loss": 0.0016, + "step": 7300 + }, + { + "epoch": 73.53, + "learning_rate": 8.272911392405064e-06, + "loss": 0.001, + "step": 7325 + }, + { + "epoch": 73.78, + "learning_rate": 8.266582278481013e-06, + "loss": 0.0015, + "step": 7350 + }, + { + "epoch": 74.03, + "learning_rate": 8.260253164556963e-06, + "loss": 0.0023, + "step": 7375 + }, + { + "epoch": 74.28, + "learning_rate": 8.253924050632912e-06, + "loss": 0.0016, + "step": 7400 + }, + { + "epoch": 74.53, + "learning_rate": 8.247594936708861e-06, + "loss": 0.0017, + "step": 7425 + }, + { + "epoch": 74.78, + "learning_rate": 8.24126582278481e-06, + "loss": 0.0016, + "step": 7450 + }, + { + "epoch": 75.03, + "learning_rate": 8.23493670886076e-06, + "loss": 0.0019, + "step": 7475 + }, + { + "epoch": 75.28, + "learning_rate": 8.22860759493671e-06, + "loss": 0.0022, + "step": 7500 + }, + { + "epoch": 75.53, + "learning_rate": 8.222278481012659e-06, + "loss": 0.0017, + "step": 7525 + }, + { + "epoch": 75.78, + "learning_rate": 8.215949367088609e-06, + "loss": 0.0016, + "step": 7550 + }, + { + "epoch": 76.04, + "learning_rate": 8.209620253164558e-06, + "loss": 0.0014, + "step": 7575 + }, + { + "epoch": 76.29, + "learning_rate": 8.203291139240507e-06, + "loss": 0.0013, + "step": 7600 + }, + { + "epoch": 76.54, + "learning_rate": 8.196962025316457e-06, + "loss": 0.0013, + "step": 7625 + }, + { + "epoch": 76.79, + "learning_rate": 8.190632911392406e-06, + "loss": 0.0014, + "step": 7650 + }, + { + "epoch": 77.04, + "learning_rate": 8.184303797468354e-06, + "loss": 0.0017, + "step": 7675 + }, + { + "epoch": 77.29, + "learning_rate": 8.177974683544303e-06, + "loss": 0.0016, + "step": 7700 + }, + { + "epoch": 77.54, + "learning_rate": 8.171645569620253e-06, + "loss": 0.0013, + "step": 7725 + }, + { + "epoch": 77.79, + "learning_rate": 8.165316455696202e-06, + "loss": 0.0014, + "step": 7750 + }, + { + "epoch": 78.04, + "learning_rate": 8.158987341772152e-06, + "loss": 0.0009, + "step": 7775 + }, + { + "epoch": 78.29, + "learning_rate": 8.152658227848101e-06, + "loss": 0.0007, + "step": 7800 + }, + { + "epoch": 78.54, + "learning_rate": 8.14632911392405e-06, + "loss": 0.0009, + "step": 7825 + }, + { + "epoch": 78.8, + "learning_rate": 8.14e-06, + "loss": 0.0005, + "step": 7850 + }, + { + "epoch": 79.05, + "learning_rate": 8.133670886075951e-06, + "loss": 0.0012, + "step": 7875 + }, + { + "epoch": 79.3, + "learning_rate": 8.1273417721519e-06, + "loss": 0.0009, + "step": 7900 + }, + { + "epoch": 79.55, + "learning_rate": 8.12101265822785e-06, + "loss": 0.0006, + "step": 7925 + }, + { + "epoch": 79.8, + "learning_rate": 8.1146835443038e-06, + "loss": 0.0008, + "step": 7950 + }, + { + "epoch": 80.05, + "learning_rate": 8.108354430379747e-06, + "loss": 0.0008, + "step": 7975 + }, + { + "epoch": 80.3, + "learning_rate": 8.102025316455697e-06, + "loss": 0.0008, + "step": 8000 + }, + { + "epoch": 80.3, + "eval_loss": 0.25484558939933777, + "eval_runtime": 2843.0839, + "eval_samples_per_second": 4.482, + "eval_steps_per_second": 0.28, + "eval_wer": 8.34668594653744, + "step": 8000 + }, + { + "epoch": 80.55, + "learning_rate": 8.095696202531646e-06, + "loss": 0.0014, + "step": 8025 + }, + { + "epoch": 80.8, + "learning_rate": 8.089367088607595e-06, + "loss": 0.0012, + "step": 8050 + }, + { + "epoch": 81.05, + "learning_rate": 8.083037974683545e-06, + "loss": 0.0018, + "step": 8075 + }, + { + "epoch": 81.3, + "learning_rate": 8.076708860759494e-06, + "loss": 0.0017, + "step": 8100 + }, + { + "epoch": 81.56, + "learning_rate": 8.070379746835444e-06, + "loss": 0.0011, + "step": 8125 + }, + { + "epoch": 81.81, + "learning_rate": 8.064050632911393e-06, + "loss": 0.0011, + "step": 8150 + }, + { + "epoch": 82.06, + "learning_rate": 8.057721518987343e-06, + "loss": 0.0012, + "step": 8175 + }, + { + "epoch": 82.31, + "learning_rate": 8.051392405063292e-06, + "loss": 0.0009, + "step": 8200 + }, + { + "epoch": 82.56, + "learning_rate": 8.045063291139241e-06, + "loss": 0.0011, + "step": 8225 + }, + { + "epoch": 82.81, + "learning_rate": 8.03873417721519e-06, + "loss": 0.0013, + "step": 8250 + }, + { + "epoch": 83.06, + "learning_rate": 8.03240506329114e-06, + "loss": 0.0012, + "step": 8275 + }, + { + "epoch": 83.31, + "learning_rate": 8.02607594936709e-06, + "loss": 0.0009, + "step": 8300 + }, + { + "epoch": 83.56, + "learning_rate": 8.019746835443039e-06, + "loss": 0.0012, + "step": 8325 + }, + { + "epoch": 83.81, + "learning_rate": 8.013417721518989e-06, + "loss": 0.001, + "step": 8350 + }, + { + "epoch": 84.07, + "learning_rate": 8.007088607594938e-06, + "loss": 0.001, + "step": 8375 + }, + { + "epoch": 84.32, + "learning_rate": 8.000759493670887e-06, + "loss": 0.0006, + "step": 8400 + }, + { + "epoch": 84.57, + "learning_rate": 7.994430379746835e-06, + "loss": 0.0006, + "step": 8425 + }, + { + "epoch": 84.82, + "learning_rate": 7.988101265822785e-06, + "loss": 0.0007, + "step": 8450 + }, + { + "epoch": 85.07, + "learning_rate": 7.981772151898734e-06, + "loss": 0.0008, + "step": 8475 + }, + { + "epoch": 85.32, + "learning_rate": 7.975443037974683e-06, + "loss": 0.0009, + "step": 8500 + }, + { + "epoch": 85.57, + "learning_rate": 7.969113924050633e-06, + "loss": 0.0015, + "step": 8525 + }, + { + "epoch": 85.82, + "learning_rate": 7.962784810126582e-06, + "loss": 0.0013, + "step": 8550 + }, + { + "epoch": 86.07, + "learning_rate": 7.956455696202532e-06, + "loss": 0.0014, + "step": 8575 + }, + { + "epoch": 86.32, + "learning_rate": 7.950126582278481e-06, + "loss": 0.0016, + "step": 8600 + }, + { + "epoch": 86.57, + "learning_rate": 7.94379746835443e-06, + "loss": 0.0013, + "step": 8625 + }, + { + "epoch": 86.83, + "learning_rate": 7.93746835443038e-06, + "loss": 0.0014, + "step": 8650 + }, + { + "epoch": 87.08, + "learning_rate": 7.93113924050633e-06, + "loss": 0.0015, + "step": 8675 + }, + { + "epoch": 87.33, + "learning_rate": 7.924810126582279e-06, + "loss": 0.0017, + "step": 8700 + }, + { + "epoch": 87.58, + "learning_rate": 7.918481012658228e-06, + "loss": 0.0018, + "step": 8725 + }, + { + "epoch": 87.83, + "learning_rate": 7.912151898734178e-06, + "loss": 0.0016, + "step": 8750 + }, + { + "epoch": 88.08, + "learning_rate": 7.905822784810127e-06, + "loss": 0.0015, + "step": 8775 + }, + { + "epoch": 88.33, + "learning_rate": 7.899493670886077e-06, + "loss": 0.0014, + "step": 8800 + }, + { + "epoch": 88.58, + "learning_rate": 7.893164556962026e-06, + "loss": 0.0011, + "step": 8825 + }, + { + "epoch": 88.83, + "learning_rate": 7.886835443037975e-06, + "loss": 0.0011, + "step": 8850 + }, + { + "epoch": 89.08, + "learning_rate": 7.880506329113925e-06, + "loss": 0.0011, + "step": 8875 + }, + { + "epoch": 89.34, + "learning_rate": 7.874177215189874e-06, + "loss": 0.0009, + "step": 8900 + }, + { + "epoch": 89.59, + "learning_rate": 7.867848101265824e-06, + "loss": 0.001, + "step": 8925 + }, + { + "epoch": 89.84, + "learning_rate": 7.861518987341773e-06, + "loss": 0.0008, + "step": 8950 + }, + { + "epoch": 90.09, + "learning_rate": 7.855189873417723e-06, + "loss": 0.0005, + "step": 8975 + }, + { + "epoch": 90.34, + "learning_rate": 7.848860759493672e-06, + "loss": 0.0004, + "step": 9000 + }, + { + "epoch": 90.34, + "eval_loss": 0.2497812658548355, + "eval_runtime": 2842.9475, + "eval_samples_per_second": 4.482, + "eval_steps_per_second": 0.28, + "eval_wer": 7.955877755197749, + "step": 9000 + }, + { + "epoch": 90.59, + "learning_rate": 7.842531645569621e-06, + "loss": 0.0003, + "step": 9025 + }, + { + "epoch": 90.84, + "learning_rate": 7.83620253164557e-06, + "loss": 0.0005, + "step": 9050 + }, + { + "epoch": 91.09, + "learning_rate": 7.82987341772152e-06, + "loss": 0.0007, + "step": 9075 + }, + { + "epoch": 91.34, + "learning_rate": 7.82354430379747e-06, + "loss": 0.0012, + "step": 9100 + }, + { + "epoch": 91.59, + "learning_rate": 7.817215189873419e-06, + "loss": 0.0014, + "step": 9125 + }, + { + "epoch": 91.84, + "learning_rate": 7.810886075949367e-06, + "loss": 0.0012, + "step": 9150 + }, + { + "epoch": 92.1, + "learning_rate": 7.804556962025316e-06, + "loss": 0.001, + "step": 9175 + }, + { + "epoch": 92.35, + "learning_rate": 7.798227848101266e-06, + "loss": 0.001, + "step": 9200 + }, + { + "epoch": 92.6, + "learning_rate": 7.791898734177215e-06, + "loss": 0.0009, + "step": 9225 + }, + { + "epoch": 92.85, + "learning_rate": 7.785569620253165e-06, + "loss": 0.0008, + "step": 9250 + }, + { + "epoch": 93.1, + "learning_rate": 7.779240506329114e-06, + "loss": 0.0012, + "step": 9275 + }, + { + "epoch": 93.35, + "learning_rate": 7.772911392405063e-06, + "loss": 0.0008, + "step": 9300 + }, + { + "epoch": 93.6, + "learning_rate": 7.766582278481013e-06, + "loss": 0.001, + "step": 9325 + }, + { + "epoch": 93.85, + "learning_rate": 7.760253164556962e-06, + "loss": 0.0017, + "step": 9350 + }, + { + "epoch": 94.1, + "learning_rate": 7.753924050632912e-06, + "loss": 0.0017, + "step": 9375 + }, + { + "epoch": 94.35, + "learning_rate": 7.747594936708861e-06, + "loss": 0.0011, + "step": 9400 + }, + { + "epoch": 94.6, + "learning_rate": 7.74126582278481e-06, + "loss": 0.0015, + "step": 9425 + }, + { + "epoch": 94.86, + "learning_rate": 7.73493670886076e-06, + "loss": 0.0012, + "step": 9450 + }, + { + "epoch": 95.11, + "learning_rate": 7.72860759493671e-06, + "loss": 0.001, + "step": 9475 + }, + { + "epoch": 95.36, + "learning_rate": 7.722278481012659e-06, + "loss": 0.001, + "step": 9500 + }, + { + "epoch": 95.61, + "learning_rate": 7.715949367088608e-06, + "loss": 0.0011, + "step": 9525 + }, + { + "epoch": 95.86, + "learning_rate": 7.709620253164558e-06, + "loss": 0.0011, + "step": 9550 + }, + { + "epoch": 96.11, + "learning_rate": 7.703291139240507e-06, + "loss": 0.0009, + "step": 9575 + }, + { + "epoch": 96.36, + "learning_rate": 7.696962025316457e-06, + "loss": 0.0014, + "step": 9600 + }, + { + "epoch": 96.61, + "learning_rate": 7.690632911392406e-06, + "loss": 0.0017, + "step": 9625 + }, + { + "epoch": 96.86, + "learning_rate": 7.684303797468355e-06, + "loss": 0.0013, + "step": 9650 + }, + { + "epoch": 97.11, + "learning_rate": 7.677974683544305e-06, + "loss": 0.001, + "step": 9675 + }, + { + "epoch": 97.37, + "learning_rate": 7.671645569620254e-06, + "loss": 0.0009, + "step": 9700 + }, + { + "epoch": 97.62, + "learning_rate": 7.665316455696204e-06, + "loss": 0.0007, + "step": 9725 + }, + { + "epoch": 97.87, + "learning_rate": 7.658987341772153e-06, + "loss": 0.0008, + "step": 9750 + }, + { + "epoch": 98.12, + "learning_rate": 7.652658227848103e-06, + "loss": 0.0007, + "step": 9775 + }, + { + "epoch": 98.37, + "learning_rate": 7.646329113924052e-06, + "loss": 0.0007, + "step": 9800 + }, + { + "epoch": 98.62, + "learning_rate": 7.640000000000001e-06, + "loss": 0.0005, + "step": 9825 + }, + { + "epoch": 98.87, + "learning_rate": 7.63367088607595e-06, + "loss": 0.0005, + "step": 9850 + }, + { + "epoch": 99.12, + "learning_rate": 7.627341772151899e-06, + "loss": 0.0002, + "step": 9875 + }, + { + "epoch": 99.37, + "learning_rate": 7.621012658227849e-06, + "loss": 0.0003, + "step": 9900 + }, + { + "epoch": 99.62, + "learning_rate": 7.614683544303798e-06, + "loss": 0.0003, + "step": 9925 + }, + { + "epoch": 99.87, + "learning_rate": 7.608354430379748e-06, + "loss": 0.0003, + "step": 9950 + }, + { + "epoch": 100.13, + "learning_rate": 7.602025316455697e-06, + "loss": 0.0002, + "step": 9975 + }, + { + "epoch": 100.38, + "learning_rate": 7.595696202531646e-06, + "loss": 0.0003, + "step": 10000 + }, + { + "epoch": 100.38, + "eval_loss": 0.24893002212047577, + "eval_runtime": 2830.2725, + "eval_samples_per_second": 4.502, + "eval_steps_per_second": 0.282, + "eval_wer": 7.694036267000157, + "step": 10000 + }, + { + "epoch": 100.63, + "learning_rate": 7.589367088607595e-06, + "loss": 0.0002, + "step": 10025 + }, + { + "epoch": 100.88, + "learning_rate": 7.5830379746835445e-06, + "loss": 0.0001, + "step": 10050 + }, + { + "epoch": 101.13, + "learning_rate": 7.576708860759494e-06, + "loss": 0.0001, + "step": 10075 + }, + { + "epoch": 101.38, + "learning_rate": 7.570379746835443e-06, + "loss": 0.0001, + "step": 10100 + }, + { + "epoch": 101.63, + "learning_rate": 7.564050632911393e-06, + "loss": 0.0001, + "step": 10125 + }, + { + "epoch": 101.88, + "learning_rate": 7.557721518987342e-06, + "loss": 0.0001, + "step": 10150 + }, + { + "epoch": 102.13, + "learning_rate": 7.551392405063292e-06, + "loss": 0.0001, + "step": 10175 + }, + { + "epoch": 102.38, + "learning_rate": 7.545063291139241e-06, + "loss": 0.0001, + "step": 10200 + }, + { + "epoch": 102.63, + "learning_rate": 7.53873417721519e-06, + "loss": 0.0001, + "step": 10225 + }, + { + "epoch": 102.89, + "learning_rate": 7.532405063291139e-06, + "loss": 0.0001, + "step": 10250 + }, + { + "epoch": 103.14, + "learning_rate": 7.5260759493670885e-06, + "loss": 0.0001, + "step": 10275 + }, + { + "epoch": 103.39, + "learning_rate": 7.519746835443038e-06, + "loss": 0.0, + "step": 10300 + }, + { + "epoch": 103.64, + "learning_rate": 7.513417721518987e-06, + "loss": 0.0, + "step": 10325 + }, + { + "epoch": 103.89, + "learning_rate": 7.507088607594938e-06, + "loss": 0.0, + "step": 10350 + }, + { + "epoch": 104.14, + "learning_rate": 7.500759493670887e-06, + "loss": 0.0001, + "step": 10375 + }, + { + "epoch": 104.39, + "learning_rate": 7.4944303797468365e-06, + "loss": 0.0, + "step": 10400 + }, + { + "epoch": 104.64, + "learning_rate": 7.488101265822786e-06, + "loss": 0.0, + "step": 10425 + }, + { + "epoch": 104.89, + "learning_rate": 7.481772151898735e-06, + "loss": 0.0001, + "step": 10450 + }, + { + "epoch": 105.14, + "learning_rate": 7.475443037974685e-06, + "loss": 0.0, + "step": 10475 + }, + { + "epoch": 105.4, + "learning_rate": 7.469113924050634e-06, + "loss": 0.0, + "step": 10500 + }, + { + "epoch": 105.65, + "learning_rate": 7.462784810126583e-06, + "loss": 0.0, + "step": 10525 + }, + { + "epoch": 105.9, + "learning_rate": 7.456455696202532e-06, + "loss": 0.0001, + "step": 10550 + }, + { + "epoch": 106.15, + "learning_rate": 7.450126582278482e-06, + "loss": 0.0, + "step": 10575 + }, + { + "epoch": 106.4, + "learning_rate": 7.443797468354431e-06, + "loss": 0.0, + "step": 10600 + }, + { + "epoch": 106.65, + "learning_rate": 7.4374683544303805e-06, + "loss": 0.0, + "step": 10625 + }, + { + "epoch": 106.9, + "learning_rate": 7.43113924050633e-06, + "loss": 0.0, + "step": 10650 + }, + { + "epoch": 107.15, + "learning_rate": 7.424810126582279e-06, + "loss": 0.0, + "step": 10675 + }, + { + "epoch": 107.4, + "learning_rate": 7.418481012658229e-06, + "loss": 0.0, + "step": 10700 + }, + { + "epoch": 107.65, + "learning_rate": 7.412151898734178e-06, + "loss": 0.0, + "step": 10725 + }, + { + "epoch": 107.9, + "learning_rate": 7.405822784810127e-06, + "loss": 0.0, + "step": 10750 + }, + { + "epoch": 108.16, + "learning_rate": 7.399493670886076e-06, + "loss": 0.0, + "step": 10775 + }, + { + "epoch": 108.41, + "learning_rate": 7.393164556962026e-06, + "loss": 0.0, + "step": 10800 + }, + { + "epoch": 108.66, + "learning_rate": 7.386835443037975e-06, + "loss": 0.0, + "step": 10825 + }, + { + "epoch": 108.91, + "learning_rate": 7.3805063291139245e-06, + "loss": 0.0, + "step": 10850 + }, + { + "epoch": 109.16, + "learning_rate": 7.374177215189874e-06, + "loss": 0.0, + "step": 10875 + }, + { + "epoch": 109.41, + "learning_rate": 7.367848101265823e-06, + "loss": 0.0, + "step": 10900 + }, + { + "epoch": 109.66, + "learning_rate": 7.361518987341773e-06, + "loss": 0.0, + "step": 10925 + }, + { + "epoch": 109.91, + "learning_rate": 7.355189873417722e-06, + "loss": 0.0, + "step": 10950 + }, + { + "epoch": 110.16, + "learning_rate": 7.348860759493671e-06, + "loss": 0.0, + "step": 10975 + }, + { + "epoch": 110.41, + "learning_rate": 7.34253164556962e-06, + "loss": 0.0, + "step": 11000 + }, + { + "epoch": 110.41, + "eval_loss": 0.2905968129634857, + "eval_runtime": 2826.8699, + "eval_samples_per_second": 4.507, + "eval_steps_per_second": 0.282, + "eval_wer": 7.545529154291073, + "step": 11000 + }, + { + "epoch": 110.66, + "learning_rate": 7.33620253164557e-06, + "loss": 0.0, + "step": 11025 + }, + { + "epoch": 110.92, + "learning_rate": 7.329873417721519e-06, + "loss": 0.0, + "step": 11050 + }, + { + "epoch": 111.17, + "learning_rate": 7.3235443037974685e-06, + "loss": 0.0, + "step": 11075 + }, + { + "epoch": 111.42, + "learning_rate": 7.317215189873418e-06, + "loss": 0.0, + "step": 11100 + }, + { + "epoch": 111.67, + "learning_rate": 7.310886075949367e-06, + "loss": 0.0, + "step": 11125 + }, + { + "epoch": 111.92, + "learning_rate": 7.304556962025317e-06, + "loss": 0.0, + "step": 11150 + }, + { + "epoch": 112.17, + "learning_rate": 7.298227848101267e-06, + "loss": 0.0, + "step": 11175 + }, + { + "epoch": 112.42, + "learning_rate": 7.2918987341772165e-06, + "loss": 0.0, + "step": 11200 + }, + { + "epoch": 112.67, + "learning_rate": 7.285569620253166e-06, + "loss": 0.0, + "step": 11225 + }, + { + "epoch": 112.92, + "learning_rate": 7.2792405063291145e-06, + "loss": 0.0, + "step": 11250 + }, + { + "epoch": 113.17, + "learning_rate": 7.272911392405064e-06, + "loss": 0.0, + "step": 11275 + }, + { + "epoch": 113.43, + "learning_rate": 7.266582278481013e-06, + "loss": 0.0, + "step": 11300 + }, + { + "epoch": 113.68, + "learning_rate": 7.260253164556963e-06, + "loss": 0.0, + "step": 11325 + }, + { + "epoch": 113.93, + "learning_rate": 7.253924050632912e-06, + "loss": 0.0, + "step": 11350 + }, + { + "epoch": 114.18, + "learning_rate": 7.247594936708862e-06, + "loss": 0.0, + "step": 11375 + }, + { + "epoch": 114.43, + "learning_rate": 7.241265822784811e-06, + "loss": 0.0, + "step": 11400 + }, + { + "epoch": 114.68, + "learning_rate": 7.2349367088607605e-06, + "loss": 0.0, + "step": 11425 + }, + { + "epoch": 114.93, + "learning_rate": 7.22860759493671e-06, + "loss": 0.0, + "step": 11450 + }, + { + "epoch": 115.18, + "learning_rate": 7.2222784810126585e-06, + "loss": 0.0, + "step": 11475 + }, + { + "epoch": 115.43, + "learning_rate": 7.215949367088608e-06, + "loss": 0.0, + "step": 11500 + }, + { + "epoch": 115.68, + "learning_rate": 7.209620253164557e-06, + "loss": 0.0, + "step": 11525 + }, + { + "epoch": 115.93, + "learning_rate": 7.203291139240507e-06, + "loss": 0.0, + "step": 11550 + }, + { + "epoch": 116.19, + "learning_rate": 7.196962025316456e-06, + "loss": 0.0, + "step": 11575 + }, + { + "epoch": 116.44, + "learning_rate": 7.190632911392406e-06, + "loss": 0.0, + "step": 11600 + }, + { + "epoch": 116.69, + "learning_rate": 7.184303797468355e-06, + "loss": 0.0, + "step": 11625 + }, + { + "epoch": 116.94, + "learning_rate": 7.1779746835443045e-06, + "loss": 0.0, + "step": 11650 + }, + { + "epoch": 117.19, + "learning_rate": 7.171645569620254e-06, + "loss": 0.0, + "step": 11675 + }, + { + "epoch": 117.44, + "learning_rate": 7.1653164556962025e-06, + "loss": 0.0, + "step": 11700 + }, + { + "epoch": 117.69, + "learning_rate": 7.158987341772152e-06, + "loss": 0.0, + "step": 11725 + }, + { + "epoch": 117.94, + "learning_rate": 7.152658227848101e-06, + "loss": 0.0, + "step": 11750 + }, + { + "epoch": 118.19, + "learning_rate": 7.146329113924051e-06, + "loss": 0.0, + "step": 11775 + }, + { + "epoch": 118.44, + "learning_rate": 7.14e-06, + "loss": 0.0, + "step": 11800 + }, + { + "epoch": 118.7, + "learning_rate": 7.13367088607595e-06, + "loss": 0.0, + "step": 11825 + }, + { + "epoch": 118.95, + "learning_rate": 7.127341772151899e-06, + "loss": 0.0, + "step": 11850 + }, + { + "epoch": 119.2, + "learning_rate": 7.1210126582278485e-06, + "loss": 0.0, + "step": 11875 + }, + { + "epoch": 119.45, + "learning_rate": 7.114683544303798e-06, + "loss": 0.0, + "step": 11900 + }, + { + "epoch": 119.7, + "learning_rate": 7.1083544303797465e-06, + "loss": 0.0, + "step": 11925 + }, + { + "epoch": 119.95, + "learning_rate": 7.102025316455696e-06, + "loss": 0.0, + "step": 11950 + }, + { + "epoch": 120.2, + "learning_rate": 7.095696202531647e-06, + "loss": 0.0, + "step": 11975 + }, + { + "epoch": 120.45, + "learning_rate": 7.089367088607596e-06, + "loss": 0.0, + "step": 12000 + }, + { + "epoch": 120.45, + "eval_loss": 0.3027211129665375, + "eval_runtime": 2829.6888, + "eval_samples_per_second": 4.503, + "eval_steps_per_second": 0.282, + "eval_wer": 7.459551352196342, + "step": 12000 + }, + { + "epoch": 120.7, + "learning_rate": 7.083037974683545e-06, + "loss": 0.0, + "step": 12025 + }, + { + "epoch": 120.95, + "learning_rate": 7.0767088607594945e-06, + "loss": 0.0, + "step": 12050 + }, + { + "epoch": 121.2, + "learning_rate": 7.070379746835444e-06, + "loss": 0.0, + "step": 12075 + }, + { + "epoch": 121.46, + "learning_rate": 7.064050632911393e-06, + "loss": 0.0, + "step": 12100 + }, + { + "epoch": 121.71, + "learning_rate": 7.057721518987343e-06, + "loss": 0.0, + "step": 12125 + }, + { + "epoch": 121.96, + "learning_rate": 7.051392405063292e-06, + "loss": 0.0, + "step": 12150 + }, + { + "epoch": 122.21, + "learning_rate": 7.045063291139242e-06, + "loss": 0.0, + "step": 12175 + }, + { + "epoch": 122.46, + "learning_rate": 7.038734177215191e-06, + "loss": 0.0, + "step": 12200 + }, + { + "epoch": 122.71, + "learning_rate": 7.03240506329114e-06, + "loss": 0.0, + "step": 12225 + }, + { + "epoch": 122.96, + "learning_rate": 7.026075949367089e-06, + "loss": 0.0, + "step": 12250 + }, + { + "epoch": 123.21, + "learning_rate": 7.0197468354430385e-06, + "loss": 0.0, + "step": 12275 + }, + { + "epoch": 123.46, + "learning_rate": 7.013417721518988e-06, + "loss": 0.0, + "step": 12300 + }, + { + "epoch": 123.71, + "learning_rate": 7.007088607594937e-06, + "loss": 0.0, + "step": 12325 + }, + { + "epoch": 123.96, + "learning_rate": 7.000759493670887e-06, + "loss": 0.0, + "step": 12350 + }, + { + "epoch": 124.22, + "learning_rate": 6.994430379746836e-06, + "loss": 0.0, + "step": 12375 + }, + { + "epoch": 124.47, + "learning_rate": 6.988101265822786e-06, + "loss": 0.0, + "step": 12400 + }, + { + "epoch": 124.72, + "learning_rate": 6.981772151898735e-06, + "loss": 0.0, + "step": 12425 + }, + { + "epoch": 124.97, + "learning_rate": 6.975443037974684e-06, + "loss": 0.0, + "step": 12450 + }, + { + "epoch": 125.22, + "learning_rate": 6.969113924050633e-06, + "loss": 0.0, + "step": 12475 + }, + { + "epoch": 125.47, + "learning_rate": 6.9627848101265825e-06, + "loss": 0.0, + "step": 12500 + }, + { + "epoch": 125.72, + "learning_rate": 6.956455696202532e-06, + "loss": 0.0, + "step": 12525 + }, + { + "epoch": 125.97, + "learning_rate": 6.950126582278481e-06, + "loss": 0.0, + "step": 12550 + }, + { + "epoch": 126.22, + "learning_rate": 6.943797468354431e-06, + "loss": 0.0, + "step": 12575 + }, + { + "epoch": 126.47, + "learning_rate": 6.93746835443038e-06, + "loss": 0.0, + "step": 12600 + }, + { + "epoch": 126.73, + "learning_rate": 6.93113924050633e-06, + "loss": 0.0, + "step": 12625 + }, + { + "epoch": 126.98, + "learning_rate": 6.924810126582279e-06, + "loss": 0.0, + "step": 12650 + }, + { + "epoch": 127.23, + "learning_rate": 6.918481012658228e-06, + "loss": 0.0, + "step": 12675 + }, + { + "epoch": 127.48, + "learning_rate": 6.912151898734177e-06, + "loss": 0.0, + "step": 12700 + }, + { + "epoch": 127.73, + "learning_rate": 6.9058227848101265e-06, + "loss": 0.0, + "step": 12725 + }, + { + "epoch": 127.98, + "learning_rate": 6.899493670886076e-06, + "loss": 0.0, + "step": 12750 + }, + { + "epoch": 128.23, + "learning_rate": 6.893164556962025e-06, + "loss": 0.0, + "step": 12775 + }, + { + "epoch": 128.48, + "learning_rate": 6.886835443037976e-06, + "loss": 0.0, + "step": 12800 + }, + { + "epoch": 128.73, + "learning_rate": 6.880506329113925e-06, + "loss": 0.0, + "step": 12825 + }, + { + "epoch": 128.98, + "learning_rate": 6.8741772151898745e-06, + "loss": 0.0, + "step": 12850 + }, + { + "epoch": 129.23, + "learning_rate": 6.867848101265824e-06, + "loss": 0.0, + "step": 12875 + }, + { + "epoch": 129.49, + "learning_rate": 6.861518987341773e-06, + "loss": 0.0, + "step": 12900 + }, + { + "epoch": 129.74, + "learning_rate": 6.855189873417723e-06, + "loss": 0.0, + "step": 12925 + }, + { + "epoch": 129.99, + "learning_rate": 6.848860759493671e-06, + "loss": 0.0, + "step": 12950 + }, + { + "epoch": 130.24, + "learning_rate": 6.842531645569621e-06, + "loss": 0.0, + "step": 12975 + }, + { + "epoch": 130.49, + "learning_rate": 6.83620253164557e-06, + "loss": 0.0, + "step": 13000 + }, + { + "epoch": 130.49, + "eval_loss": 0.31367388367652893, + "eval_runtime": 2826.8322, + "eval_samples_per_second": 4.508, + "eval_steps_per_second": 0.282, + "eval_wer": 7.451735188369549, + "step": 13000 + }, + { + "epoch": 130.74, + "learning_rate": 6.82987341772152e-06, + "loss": 0.0, + "step": 13025 + }, + { + "epoch": 130.99, + "learning_rate": 6.823544303797469e-06, + "loss": 0.0, + "step": 13050 + }, + { + "epoch": 131.24, + "learning_rate": 6.8172151898734185e-06, + "loss": 0.0, + "step": 13075 + }, + { + "epoch": 131.49, + "learning_rate": 6.810886075949368e-06, + "loss": 0.0, + "step": 13100 + }, + { + "epoch": 131.74, + "learning_rate": 6.804556962025317e-06, + "loss": 0.0, + "step": 13125 + }, + { + "epoch": 131.99, + "learning_rate": 6.798227848101267e-06, + "loss": 0.0, + "step": 13150 + }, + { + "epoch": 132.25, + "learning_rate": 6.791898734177215e-06, + "loss": 0.0, + "step": 13175 + }, + { + "epoch": 132.5, + "learning_rate": 6.785569620253165e-06, + "loss": 0.0, + "step": 13200 + }, + { + "epoch": 132.75, + "learning_rate": 6.779240506329114e-06, + "loss": 0.0, + "step": 13225 + }, + { + "epoch": 133.0, + "learning_rate": 6.772911392405064e-06, + "loss": 0.0, + "step": 13250 + }, + { + "epoch": 133.25, + "learning_rate": 6.766582278481013e-06, + "loss": 0.0, + "step": 13275 + }, + { + "epoch": 133.5, + "learning_rate": 6.7602531645569625e-06, + "loss": 0.0, + "step": 13300 + }, + { + "epoch": 133.75, + "learning_rate": 6.753924050632912e-06, + "loss": 0.0, + "step": 13325 + }, + { + "epoch": 134.0, + "learning_rate": 6.747594936708861e-06, + "loss": 0.0, + "step": 13350 + }, + { + "epoch": 134.25, + "learning_rate": 6.741265822784811e-06, + "loss": 0.0, + "step": 13375 + }, + { + "epoch": 134.5, + "learning_rate": 6.734936708860759e-06, + "loss": 0.0, + "step": 13400 + }, + { + "epoch": 134.76, + "learning_rate": 6.728607594936709e-06, + "loss": 0.0, + "step": 13425 + }, + { + "epoch": 135.01, + "learning_rate": 6.722278481012658e-06, + "loss": 0.0, + "step": 13450 + }, + { + "epoch": 135.26, + "learning_rate": 6.715949367088608e-06, + "loss": 0.0, + "step": 13475 + }, + { + "epoch": 135.51, + "learning_rate": 6.709620253164557e-06, + "loss": 0.0, + "step": 13500 + }, + { + "epoch": 135.76, + "learning_rate": 6.7032911392405065e-06, + "loss": 0.0, + "step": 13525 + }, + { + "epoch": 136.01, + "learning_rate": 6.696962025316456e-06, + "loss": 0.0, + "step": 13550 + }, + { + "epoch": 136.26, + "learning_rate": 6.690632911392405e-06, + "loss": 0.0, + "step": 13575 + }, + { + "epoch": 136.51, + "learning_rate": 6.684303797468355e-06, + "loss": 0.0, + "step": 13600 + }, + { + "epoch": 136.76, + "learning_rate": 6.677974683544305e-06, + "loss": 0.0, + "step": 13625 + }, + { + "epoch": 137.01, + "learning_rate": 6.6716455696202545e-06, + "loss": 0.0, + "step": 13650 + }, + { + "epoch": 137.26, + "learning_rate": 6.665316455696204e-06, + "loss": 0.0, + "step": 13675 + }, + { + "epoch": 137.52, + "learning_rate": 6.6589873417721525e-06, + "loss": 0.0, + "step": 13700 + }, + { + "epoch": 137.77, + "learning_rate": 6.652658227848102e-06, + "loss": 0.0, + "step": 13725 + }, + { + "epoch": 138.02, + "learning_rate": 6.646329113924051e-06, + "loss": 0.0, + "step": 13750 + }, + { + "epoch": 138.27, + "learning_rate": 6.640000000000001e-06, + "loss": 0.0, + "step": 13775 + }, + { + "epoch": 138.52, + "learning_rate": 6.63367088607595e-06, + "loss": 0.0, + "step": 13800 + }, + { + "epoch": 138.77, + "learning_rate": 6.6273417721519e-06, + "loss": 0.0, + "step": 13825 + }, + { + "epoch": 139.02, + "learning_rate": 6.621012658227849e-06, + "loss": 0.0, + "step": 13850 + }, + { + "epoch": 139.27, + "learning_rate": 6.6146835443037985e-06, + "loss": 0.0, + "step": 13875 + }, + { + "epoch": 139.52, + "learning_rate": 6.608354430379748e-06, + "loss": 0.0, + "step": 13900 + }, + { + "epoch": 139.77, + "learning_rate": 6.6020253164556965e-06, + "loss": 0.0, + "step": 13925 + }, + { + "epoch": 140.03, + "learning_rate": 6.595696202531646e-06, + "loss": 0.0, + "step": 13950 + }, + { + "epoch": 140.28, + "learning_rate": 6.589367088607595e-06, + "loss": 0.0, + "step": 13975 + }, + { + "epoch": 140.53, + "learning_rate": 6.583037974683545e-06, + "loss": 0.0, + "step": 14000 + }, + { + "epoch": 140.53, + "eval_loss": 0.32425281405448914, + "eval_runtime": 2828.1904, + "eval_samples_per_second": 4.505, + "eval_steps_per_second": 0.282, + "eval_wer": 7.464436454588089, + "step": 14000 + }, + { + "epoch": 140.78, + "learning_rate": 6.576708860759494e-06, + "loss": 0.0, + "step": 14025 + }, + { + "epoch": 141.03, + "learning_rate": 6.570379746835444e-06, + "loss": 0.0, + "step": 14050 + }, + { + "epoch": 141.28, + "learning_rate": 6.564050632911393e-06, + "loss": 0.0, + "step": 14075 + }, + { + "epoch": 141.53, + "learning_rate": 6.5577215189873425e-06, + "loss": 0.0, + "step": 14100 + }, + { + "epoch": 141.78, + "learning_rate": 6.551392405063292e-06, + "loss": 0.0, + "step": 14125 + }, + { + "epoch": 142.03, + "learning_rate": 6.5450632911392405e-06, + "loss": 0.0, + "step": 14150 + }, + { + "epoch": 142.28, + "learning_rate": 6.53873417721519e-06, + "loss": 0.0, + "step": 14175 + }, + { + "epoch": 142.53, + "learning_rate": 6.532405063291139e-06, + "loss": 0.0, + "step": 14200 + }, + { + "epoch": 142.79, + "learning_rate": 6.526075949367089e-06, + "loss": 0.0, + "step": 14225 + }, + { + "epoch": 143.04, + "learning_rate": 6.519746835443038e-06, + "loss": 0.0, + "step": 14250 + }, + { + "epoch": 143.29, + "learning_rate": 6.513417721518988e-06, + "loss": 0.0, + "step": 14275 + }, + { + "epoch": 143.54, + "learning_rate": 6.507088607594937e-06, + "loss": 0.0, + "step": 14300 + }, + { + "epoch": 143.79, + "learning_rate": 6.5007594936708865e-06, + "loss": 0.0, + "step": 14325 + }, + { + "epoch": 144.04, + "learning_rate": 6.494430379746836e-06, + "loss": 0.0, + "step": 14350 + }, + { + "epoch": 144.29, + "learning_rate": 6.4881012658227845e-06, + "loss": 0.0, + "step": 14375 + }, + { + "epoch": 144.54, + "learning_rate": 6.481772151898734e-06, + "loss": 0.0, + "step": 14400 + }, + { + "epoch": 144.79, + "learning_rate": 6.475443037974683e-06, + "loss": 0.0, + "step": 14425 + }, + { + "epoch": 145.04, + "learning_rate": 6.469113924050634e-06, + "loss": 0.0, + "step": 14450 + }, + { + "epoch": 145.29, + "learning_rate": 6.462784810126583e-06, + "loss": 0.0, + "step": 14475 + }, + { + "epoch": 145.55, + "learning_rate": 6.4564556962025325e-06, + "loss": 0.0, + "step": 14500 + }, + { + "epoch": 145.8, + "learning_rate": 6.450126582278482e-06, + "loss": 0.0, + "step": 14525 + }, + { + "epoch": 146.05, + "learning_rate": 6.443797468354431e-06, + "loss": 0.0, + "step": 14550 + }, + { + "epoch": 146.3, + "learning_rate": 6.437468354430381e-06, + "loss": 0.0, + "step": 14575 + }, + { + "epoch": 146.55, + "learning_rate": 6.43113924050633e-06, + "loss": 0.0, + "step": 14600 + }, + { + "epoch": 146.8, + "learning_rate": 6.42481012658228e-06, + "loss": 0.0, + "step": 14625 + }, + { + "epoch": 147.05, + "learning_rate": 6.418481012658228e-06, + "loss": 0.0, + "step": 14650 + }, + { + "epoch": 147.3, + "learning_rate": 6.412151898734178e-06, + "loss": 0.0, + "step": 14675 + }, + { + "epoch": 147.55, + "learning_rate": 6.405822784810127e-06, + "loss": 0.0, + "step": 14700 + }, + { + "epoch": 147.8, + "learning_rate": 6.3994936708860765e-06, + "loss": 0.0, + "step": 14725 + }, + { + "epoch": 148.06, + "learning_rate": 6.393164556962026e-06, + "loss": 0.0, + "step": 14750 + }, + { + "epoch": 148.31, + "learning_rate": 6.386835443037975e-06, + "loss": 0.0, + "step": 14775 + }, + { + "epoch": 148.56, + "learning_rate": 6.380506329113925e-06, + "loss": 0.0, + "step": 14800 + }, + { + "epoch": 148.81, + "learning_rate": 6.374177215189874e-06, + "loss": 0.0, + "step": 14825 + }, + { + "epoch": 149.06, + "learning_rate": 6.367848101265824e-06, + "loss": 0.0, + "step": 14850 + }, + { + "epoch": 149.31, + "learning_rate": 6.361518987341772e-06, + "loss": 0.0, + "step": 14875 + }, + { + "epoch": 149.56, + "learning_rate": 6.355189873417722e-06, + "loss": 0.0, + "step": 14900 + }, + { + "epoch": 149.81, + "learning_rate": 6.348860759493671e-06, + "loss": 0.0, + "step": 14925 + }, + { + "epoch": 150.06, + "learning_rate": 6.3425316455696205e-06, + "loss": 0.0, + "step": 14950 + }, + { + "epoch": 150.31, + "learning_rate": 6.33620253164557e-06, + "loss": 0.0, + "step": 14975 + }, + { + "epoch": 150.56, + "learning_rate": 6.329873417721519e-06, + "loss": 0.0, + "step": 15000 + }, + { + "epoch": 150.56, + "eval_loss": 0.3350587785243988, + "eval_runtime": 2827.6479, + "eval_samples_per_second": 4.506, + "eval_steps_per_second": 0.282, + "eval_wer": 7.476160700328278, + "step": 15000 + }, + { + "epoch": 150.82, + "learning_rate": 6.323544303797469e-06, + "loss": 0.0, + "step": 15025 + }, + { + "epoch": 151.07, + "learning_rate": 6.317215189873418e-06, + "loss": 0.0, + "step": 15050 + }, + { + "epoch": 151.32, + "learning_rate": 6.310886075949368e-06, + "loss": 0.0, + "step": 15075 + }, + { + "epoch": 151.57, + "learning_rate": 6.304556962025316e-06, + "loss": 0.0, + "step": 15100 + }, + { + "epoch": 151.82, + "learning_rate": 6.298227848101266e-06, + "loss": 0.0, + "step": 15125 + }, + { + "epoch": 152.07, + "learning_rate": 6.291898734177215e-06, + "loss": 0.0, + "step": 15150 + }, + { + "epoch": 152.32, + "learning_rate": 6.2855696202531645e-06, + "loss": 0.0, + "step": 15175 + }, + { + "epoch": 152.57, + "learning_rate": 6.279240506329114e-06, + "loss": 0.0, + "step": 15200 + }, + { + "epoch": 152.82, + "learning_rate": 6.272911392405063e-06, + "loss": 0.0, + "step": 15225 + }, + { + "epoch": 153.07, + "learning_rate": 6.266582278481013e-06, + "loss": 0.0, + "step": 15250 + }, + { + "epoch": 153.32, + "learning_rate": 6.260253164556963e-06, + "loss": 0.0, + "step": 15275 + }, + { + "epoch": 153.58, + "learning_rate": 6.2539240506329125e-06, + "loss": 0.0, + "step": 15300 + }, + { + "epoch": 153.83, + "learning_rate": 6.247594936708862e-06, + "loss": 0.0, + "step": 15325 + }, + { + "epoch": 154.08, + "learning_rate": 6.241265822784811e-06, + "loss": 0.0, + "step": 15350 + }, + { + "epoch": 154.33, + "learning_rate": 6.234936708860761e-06, + "loss": 0.0, + "step": 15375 + }, + { + "epoch": 154.58, + "learning_rate": 6.228607594936709e-06, + "loss": 0.0, + "step": 15400 + }, + { + "epoch": 154.83, + "learning_rate": 6.222278481012659e-06, + "loss": 0.0, + "step": 15425 + }, + { + "epoch": 155.08, + "learning_rate": 6.215949367088608e-06, + "loss": 0.0, + "step": 15450 + }, + { + "epoch": 155.33, + "learning_rate": 6.209620253164558e-06, + "loss": 0.0, + "step": 15475 + }, + { + "epoch": 155.58, + "learning_rate": 6.203291139240507e-06, + "loss": 0.0, + "step": 15500 + }, + { + "epoch": 155.83, + "learning_rate": 6.1969620253164565e-06, + "loss": 0.0, + "step": 15525 + }, + { + "epoch": 156.09, + "learning_rate": 6.190632911392406e-06, + "loss": 0.0, + "step": 15550 + }, + { + "epoch": 156.34, + "learning_rate": 6.184303797468355e-06, + "loss": 0.0, + "step": 15575 + }, + { + "epoch": 156.59, + "learning_rate": 6.177974683544305e-06, + "loss": 0.0, + "step": 15600 + }, + { + "epoch": 156.84, + "learning_rate": 6.171645569620253e-06, + "loss": 0.0, + "step": 15625 + }, + { + "epoch": 157.09, + "learning_rate": 6.165316455696203e-06, + "loss": 0.0, + "step": 15650 + }, + { + "epoch": 157.34, + "learning_rate": 6.158987341772152e-06, + "loss": 0.0, + "step": 15675 + }, + { + "epoch": 157.59, + "learning_rate": 6.152658227848102e-06, + "loss": 0.0, + "step": 15700 + }, + { + "epoch": 157.84, + "learning_rate": 6.146329113924051e-06, + "loss": 0.0, + "step": 15725 + }, + { + "epoch": 158.09, + "learning_rate": 6.1400000000000005e-06, + "loss": 0.0, + "step": 15750 + }, + { + "epoch": 158.34, + "learning_rate": 6.13367088607595e-06, + "loss": 0.0, + "step": 15775 + }, + { + "epoch": 158.59, + "learning_rate": 6.127341772151899e-06, + "loss": 0.0, + "step": 15800 + }, + { + "epoch": 158.85, + "learning_rate": 6.121012658227849e-06, + "loss": 0.0, + "step": 15825 + }, + { + "epoch": 159.1, + "learning_rate": 6.114683544303797e-06, + "loss": 0.0, + "step": 15850 + }, + { + "epoch": 159.35, + "learning_rate": 6.108354430379747e-06, + "loss": 0.0, + "step": 15875 + }, + { + "epoch": 159.6, + "learning_rate": 6.102025316455696e-06, + "loss": 0.0, + "step": 15900 + }, + { + "epoch": 159.85, + "learning_rate": 6.095696202531646e-06, + "loss": 0.0, + "step": 15925 + }, + { + "epoch": 160.1, + "learning_rate": 6.089367088607595e-06, + "loss": 0.0, + "step": 15950 + }, + { + "epoch": 160.35, + "learning_rate": 6.0830379746835445e-06, + "loss": 0.0, + "step": 15975 + }, + { + "epoch": 160.6, + "learning_rate": 6.076708860759494e-06, + "loss": 0.0, + "step": 16000 + }, + { + "epoch": 160.6, + "eval_loss": 0.3459363281726837, + "eval_runtime": 2825.2813, + "eval_samples_per_second": 4.51, + "eval_steps_per_second": 0.282, + "eval_wer": 7.455643270282946, + "step": 16000 + }, + { + "epoch": 160.85, + "learning_rate": 6.070379746835443e-06, + "loss": 0.0, + "step": 16025 + }, + { + "epoch": 161.1, + "learning_rate": 6.064050632911393e-06, + "loss": 0.0, + "step": 16050 + }, + { + "epoch": 161.36, + "learning_rate": 6.057721518987341e-06, + "loss": 0.0, + "step": 16075 + }, + { + "epoch": 161.61, + "learning_rate": 6.0513924050632925e-06, + "loss": 0.0, + "step": 16100 + }, + { + "epoch": 161.86, + "learning_rate": 6.045063291139242e-06, + "loss": 0.0, + "step": 16125 + }, + { + "epoch": 162.11, + "learning_rate": 6.0387341772151905e-06, + "loss": 0.0, + "step": 16150 + }, + { + "epoch": 162.36, + "learning_rate": 6.03240506329114e-06, + "loss": 0.0, + "step": 16175 + }, + { + "epoch": 162.61, + "learning_rate": 6.026075949367089e-06, + "loss": 0.0, + "step": 16200 + }, + { + "epoch": 162.86, + "learning_rate": 6.019746835443039e-06, + "loss": 0.0, + "step": 16225 + }, + { + "epoch": 163.11, + "learning_rate": 6.013417721518988e-06, + "loss": 0.0, + "step": 16250 + }, + { + "epoch": 163.36, + "learning_rate": 6.007088607594938e-06, + "loss": 0.0, + "step": 16275 + }, + { + "epoch": 163.61, + "learning_rate": 6.000759493670887e-06, + "loss": 0.0, + "step": 16300 + }, + { + "epoch": 163.86, + "learning_rate": 5.9944303797468365e-06, + "loss": 0.0, + "step": 16325 + }, + { + "epoch": 164.12, + "learning_rate": 5.988101265822786e-06, + "loss": 0.0, + "step": 16350 + }, + { + "epoch": 164.37, + "learning_rate": 5.9817721518987345e-06, + "loss": 0.0, + "step": 16375 + }, + { + "epoch": 164.62, + "learning_rate": 5.975443037974684e-06, + "loss": 0.0, + "step": 16400 + }, + { + "epoch": 164.87, + "learning_rate": 5.969113924050633e-06, + "loss": 0.0, + "step": 16425 + }, + { + "epoch": 165.12, + "learning_rate": 5.962784810126583e-06, + "loss": 0.0, + "step": 16450 + }, + { + "epoch": 165.37, + "learning_rate": 5.956455696202532e-06, + "loss": 0.0, + "step": 16475 + }, + { + "epoch": 165.62, + "learning_rate": 5.950126582278482e-06, + "loss": 0.0, + "step": 16500 + }, + { + "epoch": 165.87, + "learning_rate": 5.943797468354431e-06, + "loss": 0.0, + "step": 16525 + }, + { + "epoch": 166.12, + "learning_rate": 5.9374683544303805e-06, + "loss": 0.0, + "step": 16550 + }, + { + "epoch": 166.37, + "learning_rate": 5.931139240506329e-06, + "loss": 0.0, + "step": 16575 + }, + { + "epoch": 166.62, + "learning_rate": 5.9248101265822785e-06, + "loss": 0.0, + "step": 16600 + }, + { + "epoch": 166.88, + "learning_rate": 5.918481012658228e-06, + "loss": 0.0, + "step": 16625 + }, + { + "epoch": 167.13, + "learning_rate": 5.912151898734177e-06, + "loss": 0.0, + "step": 16650 + }, + { + "epoch": 167.38, + "learning_rate": 5.905822784810127e-06, + "loss": 0.0, + "step": 16675 + }, + { + "epoch": 167.63, + "learning_rate": 5.899493670886076e-06, + "loss": 0.0, + "step": 16700 + }, + { + "epoch": 167.88, + "learning_rate": 5.893164556962026e-06, + "loss": 0.0, + "step": 16725 + }, + { + "epoch": 168.13, + "learning_rate": 5.886835443037975e-06, + "loss": 0.0, + "step": 16750 + }, + { + "epoch": 168.38, + "learning_rate": 5.8805063291139245e-06, + "loss": 0.0, + "step": 16775 + }, + { + "epoch": 168.63, + "learning_rate": 5.874177215189873e-06, + "loss": 0.0, + "step": 16800 + }, + { + "epoch": 168.88, + "learning_rate": 5.8678481012658225e-06, + "loss": 0.0, + "step": 16825 + }, + { + "epoch": 169.13, + "learning_rate": 5.861518987341772e-06, + "loss": 0.0, + "step": 16850 + }, + { + "epoch": 169.39, + "learning_rate": 5.855189873417721e-06, + "loss": 0.0, + "step": 16875 + }, + { + "epoch": 169.64, + "learning_rate": 5.848860759493671e-06, + "loss": 0.0, + "step": 16900 + }, + { + "epoch": 169.89, + "learning_rate": 5.842531645569621e-06, + "loss": 0.0, + "step": 16925 + }, + { + "epoch": 170.14, + "learning_rate": 5.8362025316455705e-06, + "loss": 0.0, + "step": 16950 + }, + { + "epoch": 170.39, + "learning_rate": 5.82987341772152e-06, + "loss": 0.0, + "step": 16975 + }, + { + "epoch": 170.64, + "learning_rate": 5.823544303797469e-06, + "loss": 0.0, + "step": 17000 + }, + { + "epoch": 170.64, + "eval_loss": 0.35653552412986755, + "eval_runtime": 2825.8367, + "eval_samples_per_second": 4.509, + "eval_steps_per_second": 0.282, + "eval_wer": 7.460528372674691, + "step": 17000 + }, + { + "epoch": 170.89, + "learning_rate": 5.817215189873419e-06, + "loss": 0.0, + "step": 17025 + }, + { + "epoch": 171.14, + "learning_rate": 5.810886075949368e-06, + "loss": 0.0, + "step": 17050 + }, + { + "epoch": 171.39, + "learning_rate": 5.804556962025318e-06, + "loss": 0.0, + "step": 17075 + }, + { + "epoch": 171.64, + "learning_rate": 5.798227848101266e-06, + "loss": 0.0, + "step": 17100 + }, + { + "epoch": 171.89, + "learning_rate": 5.791898734177216e-06, + "loss": 0.0, + "step": 17125 + }, + { + "epoch": 172.15, + "learning_rate": 5.785569620253165e-06, + "loss": 0.0, + "step": 17150 + }, + { + "epoch": 172.4, + "learning_rate": 5.7792405063291145e-06, + "loss": 0.0, + "step": 17175 + }, + { + "epoch": 172.65, + "learning_rate": 5.772911392405064e-06, + "loss": 0.0, + "step": 17200 + }, + { + "epoch": 172.9, + "learning_rate": 5.766582278481013e-06, + "loss": 0.0, + "step": 17225 + }, + { + "epoch": 173.15, + "learning_rate": 5.760253164556963e-06, + "loss": 0.0, + "step": 17250 + }, + { + "epoch": 173.4, + "learning_rate": 5.753924050632912e-06, + "loss": 0.0, + "step": 17275 + }, + { + "epoch": 173.65, + "learning_rate": 5.747594936708862e-06, + "loss": 0.0, + "step": 17300 + }, + { + "epoch": 173.9, + "learning_rate": 5.74126582278481e-06, + "loss": 0.0, + "step": 17325 + }, + { + "epoch": 174.15, + "learning_rate": 5.73493670886076e-06, + "loss": 0.0, + "step": 17350 + }, + { + "epoch": 174.4, + "learning_rate": 5.728607594936709e-06, + "loss": 0.0, + "step": 17375 + }, + { + "epoch": 174.65, + "learning_rate": 5.7222784810126585e-06, + "loss": 0.0, + "step": 17400 + }, + { + "epoch": 174.91, + "learning_rate": 5.715949367088608e-06, + "loss": 0.0, + "step": 17425 + }, + { + "epoch": 175.16, + "learning_rate": 5.709620253164557e-06, + "loss": 0.0, + "step": 17450 + }, + { + "epoch": 175.41, + "learning_rate": 5.703291139240507e-06, + "loss": 0.0, + "step": 17475 + }, + { + "epoch": 175.66, + "learning_rate": 5.696962025316456e-06, + "loss": 0.0, + "step": 17500 + }, + { + "epoch": 175.91, + "learning_rate": 5.690632911392406e-06, + "loss": 0.0, + "step": 17525 + }, + { + "epoch": 176.16, + "learning_rate": 5.684303797468354e-06, + "loss": 0.0, + "step": 17550 + }, + { + "epoch": 176.41, + "learning_rate": 5.677974683544304e-06, + "loss": 0.0, + "step": 17575 + }, + { + "epoch": 176.66, + "learning_rate": 5.671645569620253e-06, + "loss": 0.0, + "step": 17600 + }, + { + "epoch": 176.91, + "learning_rate": 5.6653164556962025e-06, + "loss": 0.0, + "step": 17625 + }, + { + "epoch": 177.16, + "learning_rate": 5.658987341772152e-06, + "loss": 0.0, + "step": 17650 + }, + { + "epoch": 177.42, + "learning_rate": 5.652658227848101e-06, + "loss": 0.0, + "step": 17675 + }, + { + "epoch": 177.67, + "learning_rate": 5.646329113924051e-06, + "loss": 0.0, + "step": 17700 + }, + { + "epoch": 177.92, + "learning_rate": 5.64e-06, + "loss": 0.0, + "step": 17725 + }, + { + "epoch": 178.17, + "learning_rate": 5.6336708860759505e-06, + "loss": 0.0, + "step": 17750 + }, + { + "epoch": 178.42, + "learning_rate": 5.6273417721519e-06, + "loss": 0.0, + "step": 17775 + }, + { + "epoch": 178.67, + "learning_rate": 5.621012658227849e-06, + "loss": 0.0, + "step": 17800 + }, + { + "epoch": 178.92, + "learning_rate": 5.614683544303799e-06, + "loss": 0.0, + "step": 17825 + }, + { + "epoch": 179.17, + "learning_rate": 5.608354430379747e-06, + "loss": 0.0, + "step": 17850 + }, + { + "epoch": 179.42, + "learning_rate": 5.602025316455697e-06, + "loss": 0.0, + "step": 17875 + }, + { + "epoch": 179.67, + "learning_rate": 5.595696202531646e-06, + "loss": 0.0, + "step": 17900 + }, + { + "epoch": 179.92, + "learning_rate": 5.589367088607596e-06, + "loss": 0.0, + "step": 17925 + }, + { + "epoch": 180.18, + "learning_rate": 5.583037974683545e-06, + "loss": 0.0, + "step": 17950 + }, + { + "epoch": 180.43, + "learning_rate": 5.5767088607594945e-06, + "loss": 0.0, + "step": 17975 + }, + { + "epoch": 180.68, + "learning_rate": 5.570379746835444e-06, + "loss": 0.0, + "step": 18000 + }, + { + "epoch": 180.68, + "eval_loss": 0.36885330080986023, + "eval_runtime": 2834.4693, + "eval_samples_per_second": 4.495, + "eval_steps_per_second": 0.281, + "eval_wer": 7.499609191808661, + "step": 18000 + }, + { + "epoch": 180.93, + "learning_rate": 5.564050632911393e-06, + "loss": 0.0, + "step": 18025 + }, + { + "epoch": 181.18, + "learning_rate": 5.557721518987343e-06, + "loss": 0.0, + "step": 18050 + }, + { + "epoch": 181.43, + "learning_rate": 5.551392405063291e-06, + "loss": 0.0, + "step": 18075 + }, + { + "epoch": 181.68, + "learning_rate": 5.545063291139241e-06, + "loss": 0.0, + "step": 18100 + }, + { + "epoch": 181.93, + "learning_rate": 5.53873417721519e-06, + "loss": 0.0, + "step": 18125 + }, + { + "epoch": 182.18, + "learning_rate": 5.53240506329114e-06, + "loss": 0.0, + "step": 18150 + }, + { + "epoch": 182.43, + "learning_rate": 5.526075949367089e-06, + "loss": 0.0, + "step": 18175 + }, + { + "epoch": 182.69, + "learning_rate": 5.5197468354430385e-06, + "loss": 0.0, + "step": 18200 + }, + { + "epoch": 182.94, + "learning_rate": 5.513417721518988e-06, + "loss": 0.0, + "step": 18225 + }, + { + "epoch": 183.19, + "learning_rate": 5.507088607594937e-06, + "loss": 0.0, + "step": 18250 + }, + { + "epoch": 183.44, + "learning_rate": 5.500759493670887e-06, + "loss": 0.0, + "step": 18275 + }, + { + "epoch": 183.69, + "learning_rate": 5.494430379746835e-06, + "loss": 0.0, + "step": 18300 + }, + { + "epoch": 183.94, + "learning_rate": 5.488101265822785e-06, + "loss": 0.0, + "step": 18325 + }, + { + "epoch": 184.19, + "learning_rate": 5.481772151898734e-06, + "loss": 0.0, + "step": 18350 + }, + { + "epoch": 184.44, + "learning_rate": 5.475443037974684e-06, + "loss": 0.0, + "step": 18375 + }, + { + "epoch": 184.69, + "learning_rate": 5.469113924050633e-06, + "loss": 0.0, + "step": 18400 + }, + { + "epoch": 184.94, + "learning_rate": 5.4627848101265825e-06, + "loss": 0.0, + "step": 18425 + }, + { + "epoch": 185.19, + "learning_rate": 5.456455696202532e-06, + "loss": 0.0, + "step": 18450 + }, + { + "epoch": 185.45, + "learning_rate": 5.450126582278481e-06, + "loss": 0.0, + "step": 18475 + }, + { + "epoch": 185.7, + "learning_rate": 5.44379746835443e-06, + "loss": 0.0, + "step": 18500 + }, + { + "epoch": 185.95, + "learning_rate": 5.437468354430379e-06, + "loss": 0.0, + "step": 18525 + }, + { + "epoch": 186.2, + "learning_rate": 5.431139240506329e-06, + "loss": 0.0, + "step": 18550 + }, + { + "epoch": 186.45, + "learning_rate": 5.424810126582279e-06, + "loss": 0.0, + "step": 18575 + }, + { + "epoch": 186.7, + "learning_rate": 5.4184810126582285e-06, + "loss": 0.0, + "step": 18600 + }, + { + "epoch": 186.95, + "learning_rate": 5.412151898734178e-06, + "loss": 0.0, + "step": 18625 + }, + { + "epoch": 187.2, + "learning_rate": 5.405822784810127e-06, + "loss": 0.0, + "step": 18650 + }, + { + "epoch": 187.45, + "learning_rate": 5.399493670886077e-06, + "loss": 0.0, + "step": 18675 + }, + { + "epoch": 187.7, + "learning_rate": 5.393164556962026e-06, + "loss": 0.0, + "step": 18700 + }, + { + "epoch": 187.95, + "learning_rate": 5.386835443037976e-06, + "loss": 0.0, + "step": 18725 + }, + { + "epoch": 188.21, + "learning_rate": 5.380506329113925e-06, + "loss": 0.0, + "step": 18750 + }, + { + "epoch": 188.46, + "learning_rate": 5.3741772151898745e-06, + "loss": 0.0, + "step": 18775 + }, + { + "epoch": 188.71, + "learning_rate": 5.367848101265823e-06, + "loss": 0.0, + "step": 18800 + }, + { + "epoch": 188.96, + "learning_rate": 5.3615189873417725e-06, + "loss": 0.0, + "step": 18825 + }, + { + "epoch": 189.21, + "learning_rate": 5.355189873417722e-06, + "loss": 0.0, + "step": 18850 + }, + { + "epoch": 189.46, + "learning_rate": 5.348860759493671e-06, + "loss": 0.0, + "step": 18875 + }, + { + "epoch": 189.71, + "learning_rate": 5.342531645569621e-06, + "loss": 0.0, + "step": 18900 + }, + { + "epoch": 189.96, + "learning_rate": 5.33620253164557e-06, + "loss": 0.0, + "step": 18925 + }, + { + "epoch": 190.21, + "learning_rate": 5.32987341772152e-06, + "loss": 0.0, + "step": 18950 + }, + { + "epoch": 190.46, + "learning_rate": 5.323544303797469e-06, + "loss": 0.0, + "step": 18975 + }, + { + "epoch": 190.72, + "learning_rate": 5.3172151898734185e-06, + "loss": 0.0, + "step": 19000 + }, + { + "epoch": 190.72, + "eval_loss": 0.38055649399757385, + "eval_runtime": 2838.3594, + "eval_samples_per_second": 4.489, + "eval_steps_per_second": 0.281, + "eval_wer": 7.593403157730186, + "step": 19000 + }, + { + "epoch": 190.97, + "learning_rate": 5.310886075949367e-06, + "loss": 0.0, + "step": 19025 + }, + { + "epoch": 191.22, + "learning_rate": 5.3045569620253165e-06, + "loss": 0.0, + "step": 19050 + }, + { + "epoch": 191.47, + "learning_rate": 5.298227848101266e-06, + "loss": 0.0, + "step": 19075 + }, + { + "epoch": 191.72, + "learning_rate": 5.291898734177215e-06, + "loss": 0.0, + "step": 19100 + }, + { + "epoch": 191.97, + "learning_rate": 5.285569620253165e-06, + "loss": 0.0, + "step": 19125 + }, + { + "epoch": 192.22, + "learning_rate": 5.279240506329114e-06, + "loss": 0.0, + "step": 19150 + }, + { + "epoch": 192.47, + "learning_rate": 5.272911392405064e-06, + "loss": 0.0, + "step": 19175 + }, + { + "epoch": 192.72, + "learning_rate": 5.266582278481013e-06, + "loss": 0.0, + "step": 19200 + }, + { + "epoch": 192.97, + "learning_rate": 5.2602531645569625e-06, + "loss": 0.0, + "step": 19225 + }, + { + "epoch": 193.22, + "learning_rate": 5.253924050632911e-06, + "loss": 0.0, + "step": 19250 + }, + { + "epoch": 193.48, + "learning_rate": 5.2475949367088605e-06, + "loss": 0.0, + "step": 19275 + }, + { + "epoch": 193.73, + "learning_rate": 5.24126582278481e-06, + "loss": 0.0, + "step": 19300 + }, + { + "epoch": 193.98, + "learning_rate": 5.234936708860759e-06, + "loss": 0.0, + "step": 19325 + }, + { + "epoch": 194.23, + "learning_rate": 5.228607594936709e-06, + "loss": 0.0, + "step": 19350 + }, + { + "epoch": 194.48, + "learning_rate": 5.222278481012658e-06, + "loss": 0.0, + "step": 19375 + }, + { + "epoch": 194.73, + "learning_rate": 5.2159493670886085e-06, + "loss": 0.0, + "step": 19400 + }, + { + "epoch": 194.98, + "learning_rate": 5.209620253164558e-06, + "loss": 0.0, + "step": 19425 + }, + { + "epoch": 195.23, + "learning_rate": 5.203291139240507e-06, + "loss": 0.0, + "step": 19450 + }, + { + "epoch": 195.48, + "learning_rate": 5.196962025316457e-06, + "loss": 0.0, + "step": 19475 + }, + { + "epoch": 195.73, + "learning_rate": 5.190632911392406e-06, + "loss": 0.0, + "step": 19500 + }, + { + "epoch": 195.98, + "learning_rate": 5.184303797468356e-06, + "loss": 0.0, + "step": 19525 + }, + { + "epoch": 196.24, + "learning_rate": 5.177974683544304e-06, + "loss": 0.0, + "step": 19550 + }, + { + "epoch": 196.49, + "learning_rate": 5.171645569620254e-06, + "loss": 0.0, + "step": 19575 + }, + { + "epoch": 196.74, + "learning_rate": 5.165316455696203e-06, + "loss": 0.0, + "step": 19600 + }, + { + "epoch": 196.99, + "learning_rate": 5.1589873417721525e-06, + "loss": 0.0, + "step": 19625 + }, + { + "epoch": 197.24, + "learning_rate": 5.152911392405063e-06, + "loss": 0.0, + "step": 19650 + }, + { + "epoch": 197.49, + "learning_rate": 5.1465822784810125e-06, + "loss": 0.0, + "step": 19675 + }, + { + "epoch": 197.74, + "learning_rate": 5.140253164556962e-06, + "loss": 0.0, + "step": 19700 + }, + { + "epoch": 197.99, + "learning_rate": 5.133924050632911e-06, + "loss": 0.0, + "step": 19725 + }, + { + "epoch": 198.24, + "learning_rate": 5.127594936708861e-06, + "loss": 0.0, + "step": 19750 + }, + { + "epoch": 198.49, + "learning_rate": 5.12126582278481e-06, + "loss": 0.0, + "step": 19775 + }, + { + "epoch": 198.75, + "learning_rate": 5.1149367088607605e-06, + "loss": 0.0, + "step": 19800 + }, + { + "epoch": 199.0, + "learning_rate": 5.10860759493671e-06, + "loss": 0.0, + "step": 19825 + }, + { + "epoch": 199.25, + "learning_rate": 5.102278481012659e-06, + "loss": 0.0, + "step": 19850 + }, + { + "epoch": 199.5, + "learning_rate": 5.095949367088609e-06, + "loss": 0.0, + "step": 19875 + }, + { + "epoch": 199.75, + "learning_rate": 5.089620253164557e-06, + "loss": 0.0, + "step": 19900 + }, + { + "epoch": 200.0, + "learning_rate": 5.083291139240507e-06, + "loss": 0.0, + "step": 19925 + }, + { + "epoch": 200.25, + "learning_rate": 5.076962025316456e-06, + "loss": 0.0, + "step": 19950 + }, + { + "epoch": 200.5, + "learning_rate": 5.070632911392406e-06, + "loss": 0.0, + "step": 19975 + }, + { + "epoch": 200.75, + "learning_rate": 5.064303797468355e-06, + "loss": 0.0, + "step": 20000 + }, + { + "epoch": 200.75, + "eval_loss": 0.3912152349948883, + "eval_runtime": 2832.4168, + "eval_samples_per_second": 4.499, + "eval_steps_per_second": 0.281, + "eval_wer": 7.6344380178208535, + "step": 20000 + }, + { + "epoch": 201.0, + "learning_rate": 5.0579746835443045e-06, + "loss": 0.0, + "step": 20025 + }, + { + "epoch": 201.25, + "learning_rate": 5.051645569620254e-06, + "loss": 0.0, + "step": 20050 + }, + { + "epoch": 201.51, + "learning_rate": 5.045316455696203e-06, + "loss": 0.0, + "step": 20075 + }, + { + "epoch": 201.76, + "learning_rate": 5.038987341772153e-06, + "loss": 0.0, + "step": 20100 + }, + { + "epoch": 202.01, + "learning_rate": 5.032658227848101e-06, + "loss": 0.0, + "step": 20125 + }, + { + "epoch": 202.26, + "learning_rate": 5.026329113924051e-06, + "loss": 0.0, + "step": 20150 + }, + { + "epoch": 202.51, + "learning_rate": 5.02e-06, + "loss": 0.0, + "step": 20175 + }, + { + "epoch": 202.76, + "learning_rate": 5.01367088607595e-06, + "loss": 0.0, + "step": 20200 + }, + { + "epoch": 203.01, + "learning_rate": 5.007341772151899e-06, + "loss": 0.0, + "step": 20225 + }, + { + "epoch": 203.26, + "learning_rate": 5.0010126582278485e-06, + "loss": 0.0, + "step": 20250 + }, + { + "epoch": 203.51, + "learning_rate": 4.994683544303798e-06, + "loss": 0.0, + "step": 20275 + }, + { + "epoch": 203.76, + "learning_rate": 4.988354430379747e-06, + "loss": 0.0, + "step": 20300 + }, + { + "epoch": 204.02, + "learning_rate": 4.982025316455697e-06, + "loss": 0.0, + "step": 20325 + }, + { + "epoch": 204.27, + "learning_rate": 4.975696202531645e-06, + "loss": 0.0, + "step": 20350 + }, + { + "epoch": 204.52, + "learning_rate": 4.969367088607595e-06, + "loss": 0.0, + "step": 20375 + }, + { + "epoch": 204.77, + "learning_rate": 4.963037974683545e-06, + "loss": 0.0, + "step": 20400 + }, + { + "epoch": 205.02, + "learning_rate": 4.9567088607594945e-06, + "loss": 0.0, + "step": 20425 + }, + { + "epoch": 205.27, + "learning_rate": 4.950379746835444e-06, + "loss": 0.0, + "step": 20450 + }, + { + "epoch": 205.52, + "learning_rate": 4.944050632911393e-06, + "loss": 0.0, + "step": 20475 + }, + { + "epoch": 205.77, + "learning_rate": 4.937721518987342e-06, + "loss": 0.0, + "step": 20500 + }, + { + "epoch": 206.02, + "learning_rate": 4.931392405063291e-06, + "loss": 0.0, + "step": 20525 + }, + { + "epoch": 206.27, + "learning_rate": 4.925063291139241e-06, + "loss": 0.0, + "step": 20550 + }, + { + "epoch": 206.52, + "learning_rate": 4.91873417721519e-06, + "loss": 0.0, + "step": 20575 + }, + { + "epoch": 206.78, + "learning_rate": 4.91240506329114e-06, + "loss": 0.0, + "step": 20600 + }, + { + "epoch": 207.03, + "learning_rate": 4.906075949367089e-06, + "loss": 0.0, + "step": 20625 + }, + { + "epoch": 207.28, + "learning_rate": 4.8997468354430385e-06, + "loss": 0.0, + "step": 20650 + }, + { + "epoch": 207.53, + "learning_rate": 4.893417721518988e-06, + "loss": 0.0, + "step": 20675 + }, + { + "epoch": 207.78, + "learning_rate": 4.887088607594937e-06, + "loss": 0.0, + "step": 20700 + }, + { + "epoch": 208.03, + "learning_rate": 4.880759493670886e-06, + "loss": 0.0, + "step": 20725 + }, + { + "epoch": 208.28, + "learning_rate": 4.874430379746835e-06, + "loss": 0.0, + "step": 20750 + }, + { + "epoch": 208.53, + "learning_rate": 4.868101265822785e-06, + "loss": 0.0, + "step": 20775 + }, + { + "epoch": 208.78, + "learning_rate": 4.861772151898734e-06, + "loss": 0.0, + "step": 20800 + }, + { + "epoch": 209.03, + "learning_rate": 4.8554430379746845e-06, + "loss": 0.0, + "step": 20825 + }, + { + "epoch": 209.28, + "learning_rate": 4.849113924050634e-06, + "loss": 0.0, + "step": 20850 + }, + { + "epoch": 209.54, + "learning_rate": 4.8427848101265825e-06, + "loss": 0.0, + "step": 20875 + }, + { + "epoch": 209.79, + "learning_rate": 4.836455696202532e-06, + "loss": 0.0, + "step": 20900 + }, + { + "epoch": 210.04, + "learning_rate": 4.830126582278481e-06, + "loss": 0.0, + "step": 20925 + }, + { + "epoch": 210.29, + "learning_rate": 4.823797468354431e-06, + "loss": 0.0, + "step": 20950 + }, + { + "epoch": 210.54, + "learning_rate": 4.81746835443038e-06, + "loss": 0.0, + "step": 20975 + }, + { + "epoch": 210.79, + "learning_rate": 4.81113924050633e-06, + "loss": 0.0, + "step": 21000 + }, + { + "epoch": 210.79, + "eval_loss": 0.400534451007843, + "eval_runtime": 2829.8565, + "eval_samples_per_second": 4.503, + "eval_steps_per_second": 0.282, + "eval_wer": 7.548460215726122, + "step": 21000 + }, + { + "epoch": 211.04, + "learning_rate": 4.804810126582279e-06, + "loss": 0.0, + "step": 21025 + }, + { + "epoch": 211.29, + "learning_rate": 4.7984810126582285e-06, + "loss": 0.0, + "step": 21050 + }, + { + "epoch": 211.54, + "learning_rate": 4.792151898734178e-06, + "loss": 0.0, + "step": 21075 + }, + { + "epoch": 211.79, + "learning_rate": 4.7858227848101265e-06, + "loss": 0.0, + "step": 21100 + }, + { + "epoch": 212.05, + "learning_rate": 4.779493670886076e-06, + "loss": 0.0, + "step": 21125 + }, + { + "epoch": 212.3, + "learning_rate": 4.773164556962025e-06, + "loss": 0.0, + "step": 21150 + }, + { + "epoch": 212.55, + "learning_rate": 4.766835443037975e-06, + "loss": 0.0, + "step": 21175 + }, + { + "epoch": 212.8, + "learning_rate": 4.760506329113924e-06, + "loss": 0.0, + "step": 21200 + }, + { + "epoch": 213.05, + "learning_rate": 4.7541772151898745e-06, + "loss": 0.0, + "step": 21225 + }, + { + "epoch": 213.3, + "learning_rate": 4.747848101265823e-06, + "loss": 0.0, + "step": 21250 + }, + { + "epoch": 213.55, + "learning_rate": 4.7415189873417725e-06, + "loss": 0.0, + "step": 21275 + }, + { + "epoch": 213.8, + "learning_rate": 4.735189873417722e-06, + "loss": 0.0, + "step": 21300 + }, + { + "epoch": 214.05, + "learning_rate": 4.728860759493671e-06, + "loss": 0.0, + "step": 21325 + }, + { + "epoch": 214.3, + "learning_rate": 4.722531645569621e-06, + "loss": 0.0, + "step": 21350 + }, + { + "epoch": 214.55, + "learning_rate": 4.71620253164557e-06, + "loss": 0.0, + "step": 21375 + }, + { + "epoch": 214.81, + "learning_rate": 4.70987341772152e-06, + "loss": 0.0, + "step": 21400 + }, + { + "epoch": 215.06, + "learning_rate": 4.703544303797469e-06, + "loss": 0.0, + "step": 21425 + }, + { + "epoch": 215.31, + "learning_rate": 4.6972151898734185e-06, + "loss": 0.0, + "step": 21450 + }, + { + "epoch": 215.56, + "learning_rate": 4.690886075949367e-06, + "loss": 0.0, + "step": 21475 + }, + { + "epoch": 215.81, + "learning_rate": 4.6845569620253165e-06, + "loss": 0.0, + "step": 21500 + }, + { + "epoch": 216.06, + "learning_rate": 4.678227848101266e-06, + "loss": 0.0, + "step": 21525 + }, + { + "epoch": 216.31, + "learning_rate": 4.671898734177215e-06, + "loss": 0.0, + "step": 21550 + }, + { + "epoch": 216.56, + "learning_rate": 4.665569620253165e-06, + "loss": 0.0, + "step": 21575 + }, + { + "epoch": 216.81, + "learning_rate": 4.659240506329114e-06, + "loss": 0.0, + "step": 21600 + }, + { + "epoch": 217.06, + "learning_rate": 4.652911392405064e-06, + "loss": 0.0, + "step": 21625 + }, + { + "epoch": 217.31, + "learning_rate": 4.646582278481013e-06, + "loss": 0.0, + "step": 21650 + }, + { + "epoch": 217.57, + "learning_rate": 4.6402531645569625e-06, + "loss": 0.0, + "step": 21675 + }, + { + "epoch": 217.82, + "learning_rate": 4.633924050632912e-06, + "loss": 0.0, + "step": 21700 + }, + { + "epoch": 218.07, + "learning_rate": 4.627594936708861e-06, + "loss": 0.0, + "step": 21725 + }, + { + "epoch": 218.32, + "learning_rate": 4.621265822784811e-06, + "loss": 0.0, + "step": 21750 + }, + { + "epoch": 218.57, + "learning_rate": 4.61493670886076e-06, + "loss": 0.0, + "step": 21775 + }, + { + "epoch": 218.82, + "learning_rate": 4.60860759493671e-06, + "loss": 0.0, + "step": 21800 + }, + { + "epoch": 219.07, + "learning_rate": 4.602278481012658e-06, + "loss": 0.0, + "step": 21825 + }, + { + "epoch": 219.32, + "learning_rate": 4.595949367088608e-06, + "loss": 0.0, + "step": 21850 + }, + { + "epoch": 219.57, + "learning_rate": 4.589620253164557e-06, + "loss": 0.0, + "step": 21875 + }, + { + "epoch": 219.82, + "learning_rate": 4.5832911392405065e-06, + "loss": 0.0, + "step": 21900 + }, + { + "epoch": 220.08, + "learning_rate": 4.576962025316456e-06, + "loss": 0.0, + "step": 21925 + }, + { + "epoch": 220.33, + "learning_rate": 4.570632911392405e-06, + "loss": 0.0, + "step": 21950 + }, + { + "epoch": 220.58, + "learning_rate": 4.564303797468355e-06, + "loss": 0.0, + "step": 21975 + }, + { + "epoch": 220.83, + "learning_rate": 4.557974683544304e-06, + "loss": 0.0, + "step": 22000 + }, + { + "epoch": 220.83, + "eval_loss": 0.4101511240005493, + "eval_runtime": 2839.3882, + "eval_samples_per_second": 4.488, + "eval_steps_per_second": 0.281, + "eval_wer": 7.6266218539940605, + "step": 22000 + }, + { + "epoch": 221.08, + "learning_rate": 4.551645569620254e-06, + "loss": 0.0, + "step": 22025 + }, + { + "epoch": 221.33, + "learning_rate": 4.545316455696203e-06, + "loss": 0.0, + "step": 22050 + }, + { + "epoch": 221.58, + "learning_rate": 4.5389873417721525e-06, + "loss": 0.0, + "step": 22075 + }, + { + "epoch": 221.83, + "learning_rate": 4.532658227848102e-06, + "loss": 0.0, + "step": 22100 + }, + { + "epoch": 222.08, + "learning_rate": 4.526329113924051e-06, + "loss": 0.0, + "step": 22125 + }, + { + "epoch": 222.33, + "learning_rate": 4.520000000000001e-06, + "loss": 0.0, + "step": 22150 + }, + { + "epoch": 222.58, + "learning_rate": 4.51367088607595e-06, + "loss": 0.0, + "step": 22175 + }, + { + "epoch": 222.84, + "learning_rate": 4.507341772151899e-06, + "loss": 0.0, + "step": 22200 + }, + { + "epoch": 223.09, + "learning_rate": 4.501012658227848e-06, + "loss": 0.0, + "step": 22225 + }, + { + "epoch": 223.34, + "learning_rate": 4.494683544303798e-06, + "loss": 0.0, + "step": 22250 + }, + { + "epoch": 223.59, + "learning_rate": 4.488354430379747e-06, + "loss": 0.0, + "step": 22275 + }, + { + "epoch": 223.84, + "learning_rate": 4.4820253164556965e-06, + "loss": 0.0, + "step": 22300 + }, + { + "epoch": 224.09, + "learning_rate": 4.475696202531646e-06, + "loss": 0.0, + "step": 22325 + }, + { + "epoch": 224.34, + "learning_rate": 4.469367088607595e-06, + "loss": 0.0, + "step": 22350 + }, + { + "epoch": 224.59, + "learning_rate": 4.463037974683545e-06, + "loss": 0.0, + "step": 22375 + }, + { + "epoch": 224.84, + "learning_rate": 4.456708860759494e-06, + "loss": 0.0, + "step": 22400 + }, + { + "epoch": 225.09, + "learning_rate": 4.450379746835443e-06, + "loss": 0.0, + "step": 22425 + }, + { + "epoch": 225.35, + "learning_rate": 4.444050632911392e-06, + "loss": 0.0, + "step": 22450 + }, + { + "epoch": 225.6, + "learning_rate": 4.4377215189873425e-06, + "loss": 0.0, + "step": 22475 + }, + { + "epoch": 225.85, + "learning_rate": 4.431392405063292e-06, + "loss": 0.0, + "step": 22500 + }, + { + "epoch": 226.1, + "learning_rate": 4.425063291139241e-06, + "loss": 0.0, + "step": 22525 + }, + { + "epoch": 226.35, + "learning_rate": 4.418734177215191e-06, + "loss": 0.0, + "step": 22550 + }, + { + "epoch": 226.6, + "learning_rate": 4.412405063291139e-06, + "loss": 0.0, + "step": 22575 + }, + { + "epoch": 226.85, + "learning_rate": 4.406075949367089e-06, + "loss": 0.0, + "step": 22600 + }, + { + "epoch": 227.1, + "learning_rate": 4.399746835443038e-06, + "loss": 0.0, + "step": 22625 + }, + { + "epoch": 227.35, + "learning_rate": 4.393417721518988e-06, + "loss": 0.0, + "step": 22650 + }, + { + "epoch": 227.6, + "learning_rate": 4.387088607594937e-06, + "loss": 0.0, + "step": 22675 + }, + { + "epoch": 227.85, + "learning_rate": 4.3807594936708865e-06, + "loss": 0.0, + "step": 22700 + }, + { + "epoch": 228.11, + "learning_rate": 4.374430379746836e-06, + "loss": 0.0, + "step": 22725 + }, + { + "epoch": 228.36, + "learning_rate": 4.368101265822785e-06, + "loss": 0.0, + "step": 22750 + }, + { + "epoch": 228.61, + "learning_rate": 4.361772151898735e-06, + "loss": 0.0, + "step": 22775 + }, + { + "epoch": 228.86, + "learning_rate": 4.355443037974683e-06, + "loss": 0.0, + "step": 22800 + }, + { + "epoch": 229.11, + "learning_rate": 4.349113924050633e-06, + "loss": 0.0, + "step": 22825 + }, + { + "epoch": 229.36, + "learning_rate": 4.342784810126582e-06, + "loss": 0.0, + "step": 22850 + }, + { + "epoch": 229.61, + "learning_rate": 4.3364556962025325e-06, + "loss": 0.0, + "step": 22875 + }, + { + "epoch": 229.86, + "learning_rate": 4.330126582278482e-06, + "loss": 0.0, + "step": 22900 + }, + { + "epoch": 230.11, + "learning_rate": 4.323797468354431e-06, + "loss": 0.0, + "step": 22925 + }, + { + "epoch": 230.36, + "learning_rate": 4.31746835443038e-06, + "loss": 0.0, + "step": 22950 + }, + { + "epoch": 230.61, + "learning_rate": 4.311392405063291e-06, + "loss": 0.0074, + "step": 22975 + }, + { + "epoch": 230.87, + "learning_rate": 4.305063291139241e-06, + "loss": 0.0079, + "step": 23000 + }, + { + "epoch": 230.87, + "eval_loss": 0.24673040211200714, + "eval_runtime": 2817.1734, + "eval_samples_per_second": 4.523, + "eval_steps_per_second": 0.283, + "eval_wer": 9.165429107394091, + "step": 23000 + }, + { + "epoch": 231.12, + "learning_rate": 4.298987341772152e-06, + "loss": 0.0049, + "step": 23025 + }, + { + "epoch": 231.37, + "learning_rate": 4.292658227848102e-06, + "loss": 0.0041, + "step": 23050 + }, + { + "epoch": 231.62, + "learning_rate": 4.286329113924051e-06, + "loss": 0.0025, + "step": 23075 + }, + { + "epoch": 231.87, + "learning_rate": 4.2800000000000005e-06, + "loss": 0.0028, + "step": 23100 + }, + { + "epoch": 232.12, + "learning_rate": 4.27367088607595e-06, + "loss": 0.0018, + "step": 23125 + }, + { + "epoch": 232.37, + "learning_rate": 4.2673417721518985e-06, + "loss": 0.0009, + "step": 23150 + }, + { + "epoch": 232.62, + "learning_rate": 4.261012658227848e-06, + "loss": 0.0007, + "step": 23175 + }, + { + "epoch": 232.87, + "learning_rate": 4.254683544303797e-06, + "loss": 0.0005, + "step": 23200 + }, + { + "epoch": 233.12, + "learning_rate": 4.248354430379747e-06, + "loss": 0.0003, + "step": 23225 + }, + { + "epoch": 233.38, + "learning_rate": 4.242025316455697e-06, + "loss": 0.0003, + "step": 23250 + }, + { + "epoch": 233.63, + "learning_rate": 4.2356962025316465e-06, + "loss": 0.0002, + "step": 23275 + }, + { + "epoch": 233.88, + "learning_rate": 4.229367088607595e-06, + "loss": 0.0003, + "step": 23300 + }, + { + "epoch": 234.13, + "learning_rate": 4.2230379746835445e-06, + "loss": 0.0003, + "step": 23325 + }, + { + "epoch": 234.38, + "learning_rate": 4.216708860759494e-06, + "loss": 0.0001, + "step": 23350 + }, + { + "epoch": 234.63, + "learning_rate": 4.210379746835443e-06, + "loss": 0.0001, + "step": 23375 + }, + { + "epoch": 234.88, + "learning_rate": 4.204050632911393e-06, + "loss": 0.0001, + "step": 23400 + }, + { + "epoch": 235.13, + "learning_rate": 4.197721518987342e-06, + "loss": 0.0001, + "step": 23425 + }, + { + "epoch": 235.38, + "learning_rate": 4.191392405063292e-06, + "loss": 0.0001, + "step": 23450 + }, + { + "epoch": 235.63, + "learning_rate": 4.185063291139241e-06, + "loss": 0.0001, + "step": 23475 + }, + { + "epoch": 235.88, + "learning_rate": 4.1787341772151905e-06, + "loss": 0.0001, + "step": 23500 + }, + { + "epoch": 236.14, + "learning_rate": 4.172405063291139e-06, + "loss": 0.0001, + "step": 23525 + }, + { + "epoch": 236.39, + "learning_rate": 4.1660759493670885e-06, + "loss": 0.0, + "step": 23550 + }, + { + "epoch": 236.64, + "learning_rate": 4.159746835443038e-06, + "loss": 0.0, + "step": 23575 + }, + { + "epoch": 236.89, + "learning_rate": 4.153417721518987e-06, + "loss": 0.0, + "step": 23600 + }, + { + "epoch": 237.14, + "learning_rate": 4.147088607594937e-06, + "loss": 0.0001, + "step": 23625 + }, + { + "epoch": 237.39, + "learning_rate": 4.140759493670886e-06, + "loss": 0.0, + "step": 23650 + }, + { + "epoch": 237.64, + "learning_rate": 4.134430379746836e-06, + "loss": 0.0001, + "step": 23675 + }, + { + "epoch": 237.89, + "learning_rate": 4.128101265822785e-06, + "loss": 0.0, + "step": 23700 + }, + { + "epoch": 238.14, + "learning_rate": 4.1217721518987345e-06, + "loss": 0.0, + "step": 23725 + }, + { + "epoch": 238.39, + "learning_rate": 4.115443037974684e-06, + "loss": 0.0, + "step": 23750 + }, + { + "epoch": 238.64, + "learning_rate": 4.109113924050633e-06, + "loss": 0.0001, + "step": 23775 + }, + { + "epoch": 238.9, + "learning_rate": 4.102784810126583e-06, + "loss": 0.0, + "step": 23800 + }, + { + "epoch": 239.15, + "learning_rate": 4.096455696202532e-06, + "loss": 0.0, + "step": 23825 + }, + { + "epoch": 239.4, + "learning_rate": 4.090126582278482e-06, + "loss": 0.0, + "step": 23850 + }, + { + "epoch": 239.65, + "learning_rate": 4.083797468354431e-06, + "loss": 0.0, + "step": 23875 + }, + { + "epoch": 239.9, + "learning_rate": 4.07746835443038e-06, + "loss": 0.0001, + "step": 23900 + }, + { + "epoch": 240.15, + "learning_rate": 4.071139240506329e-06, + "loss": 0.0001, + "step": 23925 + }, + { + "epoch": 240.4, + "learning_rate": 4.0648101265822785e-06, + "loss": 0.0, + "step": 23950 + }, + { + "epoch": 240.65, + "learning_rate": 4.058481012658228e-06, + "loss": 0.0, + "step": 23975 + }, + { + "epoch": 240.9, + "learning_rate": 4.052151898734177e-06, + "loss": 0.0, + "step": 24000 + }, + { + "epoch": 240.9, + "eval_loss": 0.30968427658081055, + "eval_runtime": 2825.8546, + "eval_samples_per_second": 4.509, + "eval_steps_per_second": 0.282, + "eval_wer": 7.761450680006253, + "step": 24000 + }, + { + "epoch": 241.15, + "learning_rate": 4.045822784810127e-06, + "loss": 0.0, + "step": 24025 + }, + { + "epoch": 241.41, + "learning_rate": 4.039493670886076e-06, + "loss": 0.0, + "step": 24050 + }, + { + "epoch": 241.66, + "learning_rate": 4.033164556962026e-06, + "loss": 0.0, + "step": 24075 + }, + { + "epoch": 241.91, + "learning_rate": 4.026835443037975e-06, + "loss": 0.0, + "step": 24100 + }, + { + "epoch": 242.16, + "learning_rate": 4.0205063291139245e-06, + "loss": 0.0, + "step": 24125 + }, + { + "epoch": 242.41, + "learning_rate": 4.014177215189874e-06, + "loss": 0.0, + "step": 24150 + }, + { + "epoch": 242.66, + "learning_rate": 4.007848101265823e-06, + "loss": 0.0, + "step": 24175 + }, + { + "epoch": 242.91, + "learning_rate": 4.001518987341773e-06, + "loss": 0.0, + "step": 24200 + }, + { + "epoch": 243.16, + "learning_rate": 3.995189873417722e-06, + "loss": 0.0, + "step": 24225 + }, + { + "epoch": 243.41, + "learning_rate": 3.988860759493672e-06, + "loss": 0.0, + "step": 24250 + }, + { + "epoch": 243.66, + "learning_rate": 3.98253164556962e-06, + "loss": 0.0, + "step": 24275 + }, + { + "epoch": 243.91, + "learning_rate": 3.97620253164557e-06, + "loss": 0.0, + "step": 24300 + }, + { + "epoch": 244.17, + "learning_rate": 3.969873417721519e-06, + "loss": 0.0, + "step": 24325 + }, + { + "epoch": 244.42, + "learning_rate": 3.9635443037974685e-06, + "loss": 0.0, + "step": 24350 + }, + { + "epoch": 244.67, + "learning_rate": 3.957215189873418e-06, + "loss": 0.0, + "step": 24375 + }, + { + "epoch": 244.92, + "learning_rate": 3.950886075949367e-06, + "loss": 0.0, + "step": 24400 + }, + { + "epoch": 245.17, + "learning_rate": 3.944556962025317e-06, + "loss": 0.0, + "step": 24425 + }, + { + "epoch": 245.42, + "learning_rate": 3.938227848101266e-06, + "loss": 0.0, + "step": 24450 + }, + { + "epoch": 245.67, + "learning_rate": 3.931898734177216e-06, + "loss": 0.0, + "step": 24475 + }, + { + "epoch": 245.92, + "learning_rate": 3.925569620253165e-06, + "loss": 0.0, + "step": 24500 + }, + { + "epoch": 246.17, + "learning_rate": 3.9192405063291145e-06, + "loss": 0.0, + "step": 24525 + }, + { + "epoch": 246.42, + "learning_rate": 3.912911392405064e-06, + "loss": 0.0, + "step": 24550 + }, + { + "epoch": 246.68, + "learning_rate": 3.906582278481013e-06, + "loss": 0.0, + "step": 24575 + }, + { + "epoch": 246.93, + "learning_rate": 3.900253164556963e-06, + "loss": 0.0, + "step": 24600 + }, + { + "epoch": 247.18, + "learning_rate": 3.893924050632911e-06, + "loss": 0.0, + "step": 24625 + }, + { + "epoch": 247.43, + "learning_rate": 3.887594936708861e-06, + "loss": 0.0, + "step": 24650 + }, + { + "epoch": 247.68, + "learning_rate": 3.88126582278481e-06, + "loss": 0.0, + "step": 24675 + }, + { + "epoch": 247.93, + "learning_rate": 3.87493670886076e-06, + "loss": 0.0, + "step": 24700 + }, + { + "epoch": 248.18, + "learning_rate": 3.868607594936709e-06, + "loss": 0.0, + "step": 24725 + }, + { + "epoch": 248.43, + "learning_rate": 3.8622784810126585e-06, + "loss": 0.0, + "step": 24750 + }, + { + "epoch": 248.68, + "learning_rate": 3.855949367088608e-06, + "loss": 0.0, + "step": 24775 + }, + { + "epoch": 248.93, + "learning_rate": 3.849620253164557e-06, + "loss": 0.0, + "step": 24800 + }, + { + "epoch": 249.18, + "learning_rate": 3.843291139240507e-06, + "loss": 0.0, + "step": 24825 + }, + { + "epoch": 249.44, + "learning_rate": 3.836962025316455e-06, + "loss": 0.0, + "step": 24850 + }, + { + "epoch": 249.69, + "learning_rate": 3.830632911392405e-06, + "loss": 0.0, + "step": 24875 + }, + { + "epoch": 249.94, + "learning_rate": 3.824303797468355e-06, + "loss": 0.0, + "step": 24900 + }, + { + "epoch": 250.19, + "learning_rate": 3.8179746835443045e-06, + "loss": 0.0, + "step": 24925 + }, + { + "epoch": 250.44, + "learning_rate": 3.8116455696202535e-06, + "loss": 0.0, + "step": 24950 + }, + { + "epoch": 250.69, + "learning_rate": 3.805316455696203e-06, + "loss": 0.0, + "step": 24975 + }, + { + "epoch": 250.94, + "learning_rate": 3.7989873417721524e-06, + "loss": 0.0, + "step": 25000 + }, + { + "epoch": 250.94, + "eval_loss": 0.3311292827129364, + "eval_runtime": 2833.7855, + "eval_samples_per_second": 4.496, + "eval_steps_per_second": 0.281, + "eval_wer": 7.724323901828982, + "step": 25000 + }, + { + "epoch": 251.19, + "learning_rate": 3.792658227848102e-06, + "loss": 0.0, + "step": 25025 + }, + { + "epoch": 251.44, + "learning_rate": 3.786329113924051e-06, + "loss": 0.0, + "step": 25050 + }, + { + "epoch": 251.69, + "learning_rate": 3.7800000000000002e-06, + "loss": 0.0, + "step": 25075 + }, + { + "epoch": 251.94, + "learning_rate": 3.7736708860759497e-06, + "loss": 0.0, + "step": 25100 + }, + { + "epoch": 252.2, + "learning_rate": 3.767341772151899e-06, + "loss": 0.0, + "step": 25125 + }, + { + "epoch": 252.45, + "learning_rate": 3.7610126582278485e-06, + "loss": 0.0, + "step": 25150 + }, + { + "epoch": 252.7, + "learning_rate": 3.7546835443037975e-06, + "loss": 0.0, + "step": 25175 + }, + { + "epoch": 252.95, + "learning_rate": 3.748354430379747e-06, + "loss": 0.0, + "step": 25200 + }, + { + "epoch": 253.2, + "learning_rate": 3.7420253164556964e-06, + "loss": 0.0, + "step": 25225 + }, + { + "epoch": 253.45, + "learning_rate": 3.735696202531646e-06, + "loss": 0.0, + "step": 25250 + }, + { + "epoch": 253.7, + "learning_rate": 3.729367088607595e-06, + "loss": 0.0, + "step": 25275 + }, + { + "epoch": 253.95, + "learning_rate": 3.7230379746835442e-06, + "loss": 0.0, + "step": 25300 + }, + { + "epoch": 254.2, + "learning_rate": 3.716708860759494e-06, + "loss": 0.0, + "step": 25325 + }, + { + "epoch": 254.45, + "learning_rate": 3.7103797468354435e-06, + "loss": 0.0, + "step": 25350 + }, + { + "epoch": 254.71, + "learning_rate": 3.704050632911393e-06, + "loss": 0.0, + "step": 25375 + }, + { + "epoch": 254.96, + "learning_rate": 3.6977215189873424e-06, + "loss": 0.0, + "step": 25400 + }, + { + "epoch": 255.21, + "learning_rate": 3.6913924050632914e-06, + "loss": 0.0, + "step": 25425 + }, + { + "epoch": 255.46, + "learning_rate": 3.685063291139241e-06, + "loss": 0.0, + "step": 25450 + }, + { + "epoch": 255.71, + "learning_rate": 3.6787341772151902e-06, + "loss": 0.0, + "step": 25475 + }, + { + "epoch": 255.96, + "learning_rate": 3.6724050632911397e-06, + "loss": 0.0, + "step": 25500 + }, + { + "epoch": 256.21, + "learning_rate": 3.666075949367089e-06, + "loss": 0.0, + "step": 25525 + }, + { + "epoch": 256.46, + "learning_rate": 3.659746835443038e-06, + "loss": 0.0, + "step": 25550 + }, + { + "epoch": 256.71, + "learning_rate": 3.6534177215189875e-06, + "loss": 0.0, + "step": 25575 + }, + { + "epoch": 256.96, + "learning_rate": 3.647088607594937e-06, + "loss": 0.0, + "step": 25600 + }, + { + "epoch": 257.21, + "learning_rate": 3.6407594936708864e-06, + "loss": 0.0, + "step": 25625 + }, + { + "epoch": 257.47, + "learning_rate": 3.6344303797468354e-06, + "loss": 0.0, + "step": 25650 + }, + { + "epoch": 257.72, + "learning_rate": 3.628101265822785e-06, + "loss": 0.0, + "step": 25675 + }, + { + "epoch": 257.97, + "learning_rate": 3.6217721518987342e-06, + "loss": 0.0, + "step": 25700 + }, + { + "epoch": 258.22, + "learning_rate": 3.615443037974684e-06, + "loss": 0.0, + "step": 25725 + }, + { + "epoch": 258.47, + "learning_rate": 3.6091139240506335e-06, + "loss": 0.0, + "step": 25750 + }, + { + "epoch": 258.72, + "learning_rate": 3.602784810126583e-06, + "loss": 0.0, + "step": 25775 + }, + { + "epoch": 258.97, + "learning_rate": 3.596455696202532e-06, + "loss": 0.0, + "step": 25800 + }, + { + "epoch": 259.22, + "learning_rate": 3.5901265822784814e-06, + "loss": 0.0, + "step": 25825 + }, + { + "epoch": 259.47, + "learning_rate": 3.583797468354431e-06, + "loss": 0.0, + "step": 25850 + }, + { + "epoch": 259.72, + "learning_rate": 3.5774683544303802e-06, + "loss": 0.0, + "step": 25875 + }, + { + "epoch": 259.97, + "learning_rate": 3.5711392405063292e-06, + "loss": 0.0, + "step": 25900 + }, + { + "epoch": 260.23, + "learning_rate": 3.5648101265822787e-06, + "loss": 0.0, + "step": 25925 + }, + { + "epoch": 260.48, + "learning_rate": 3.558481012658228e-06, + "loss": 0.0, + "step": 25950 + }, + { + "epoch": 260.73, + "learning_rate": 3.5521518987341775e-06, + "loss": 0.0, + "step": 25975 + }, + { + "epoch": 260.98, + "learning_rate": 3.545822784810127e-06, + "loss": 0.0, + "step": 26000 + }, + { + "epoch": 260.98, + "eval_loss": 0.3445892930030823, + "eval_runtime": 2827.7825, + "eval_samples_per_second": 4.506, + "eval_steps_per_second": 0.282, + "eval_wer": 7.7028294513053, + "step": 26000 + }, + { + "epoch": 261.23, + "learning_rate": 3.539493670886076e-06, + "loss": 0.0, + "step": 26025 + }, + { + "epoch": 261.48, + "learning_rate": 3.5331645569620254e-06, + "loss": 0.0, + "step": 26050 + }, + { + "epoch": 261.73, + "learning_rate": 3.526835443037975e-06, + "loss": 0.0, + "step": 26075 + }, + { + "epoch": 261.98, + "learning_rate": 3.5205063291139242e-06, + "loss": 0.0, + "step": 26100 + }, + { + "epoch": 262.23, + "learning_rate": 3.514177215189874e-06, + "loss": 0.0, + "step": 26125 + }, + { + "epoch": 262.48, + "learning_rate": 3.5078481012658235e-06, + "loss": 0.0, + "step": 26150 + }, + { + "epoch": 262.74, + "learning_rate": 3.5015189873417725e-06, + "loss": 0.0, + "step": 26175 + }, + { + "epoch": 262.99, + "learning_rate": 3.495189873417722e-06, + "loss": 0.0, + "step": 26200 + }, + { + "epoch": 263.24, + "learning_rate": 3.4888607594936714e-06, + "loss": 0.0, + "step": 26225 + }, + { + "epoch": 263.49, + "learning_rate": 3.482531645569621e-06, + "loss": 0.0, + "step": 26250 + }, + { + "epoch": 263.74, + "learning_rate": 3.47620253164557e-06, + "loss": 0.0, + "step": 26275 + }, + { + "epoch": 263.99, + "learning_rate": 3.4698734177215192e-06, + "loss": 0.0, + "step": 26300 + }, + { + "epoch": 264.24, + "learning_rate": 3.4635443037974687e-06, + "loss": 0.0, + "step": 26325 + }, + { + "epoch": 264.49, + "learning_rate": 3.457215189873418e-06, + "loss": 0.0, + "step": 26350 + }, + { + "epoch": 264.74, + "learning_rate": 3.4508860759493675e-06, + "loss": 0.0, + "step": 26375 + }, + { + "epoch": 264.99, + "learning_rate": 3.4445569620253165e-06, + "loss": 0.0, + "step": 26400 + }, + { + "epoch": 265.24, + "learning_rate": 3.438227848101266e-06, + "loss": 0.0, + "step": 26425 + }, + { + "epoch": 265.5, + "learning_rate": 3.4318987341772154e-06, + "loss": 0.0, + "step": 26450 + }, + { + "epoch": 265.75, + "learning_rate": 3.425569620253165e-06, + "loss": 0.0, + "step": 26475 + }, + { + "epoch": 266.0, + "learning_rate": 3.419240506329114e-06, + "loss": 0.0, + "step": 26500 + }, + { + "epoch": 266.25, + "learning_rate": 3.4129113924050632e-06, + "loss": 0.0, + "step": 26525 + }, + { + "epoch": 266.5, + "learning_rate": 3.406582278481013e-06, + "loss": 0.0, + "step": 26550 + }, + { + "epoch": 266.75, + "learning_rate": 3.4002531645569625e-06, + "loss": 0.0, + "step": 26575 + }, + { + "epoch": 267.0, + "learning_rate": 3.393924050632912e-06, + "loss": 0.0, + "step": 26600 + }, + { + "epoch": 267.25, + "learning_rate": 3.3875949367088614e-06, + "loss": 0.0, + "step": 26625 + }, + { + "epoch": 267.5, + "learning_rate": 3.3812658227848104e-06, + "loss": 0.0, + "step": 26650 + }, + { + "epoch": 267.75, + "learning_rate": 3.37493670886076e-06, + "loss": 0.0, + "step": 26675 + }, + { + "epoch": 268.01, + "learning_rate": 3.3686075949367092e-06, + "loss": 0.0, + "step": 26700 + }, + { + "epoch": 268.26, + "learning_rate": 3.3622784810126586e-06, + "loss": 0.0, + "step": 26725 + }, + { + "epoch": 268.51, + "learning_rate": 3.3559493670886077e-06, + "loss": 0.0, + "step": 26750 + }, + { + "epoch": 268.76, + "learning_rate": 3.349620253164557e-06, + "loss": 0.0, + "step": 26775 + }, + { + "epoch": 269.01, + "learning_rate": 3.3432911392405065e-06, + "loss": 0.0, + "step": 26800 + }, + { + "epoch": 269.26, + "learning_rate": 3.336962025316456e-06, + "loss": 0.0, + "step": 26825 + }, + { + "epoch": 269.51, + "learning_rate": 3.3306329113924054e-06, + "loss": 0.0, + "step": 26850 + }, + { + "epoch": 269.76, + "learning_rate": 3.3243037974683544e-06, + "loss": 0.0, + "step": 26875 + }, + { + "epoch": 270.01, + "learning_rate": 3.317974683544304e-06, + "loss": 0.0, + "step": 26900 + }, + { + "epoch": 270.26, + "learning_rate": 3.3116455696202532e-06, + "loss": 0.0, + "step": 26925 + }, + { + "epoch": 270.51, + "learning_rate": 3.305316455696203e-06, + "loss": 0.0, + "step": 26950 + }, + { + "epoch": 270.77, + "learning_rate": 3.2989873417721525e-06, + "loss": 0.0, + "step": 26975 + }, + { + "epoch": 271.02, + "learning_rate": 3.292658227848102e-06, + "loss": 0.0, + "step": 27000 + }, + { + "epoch": 271.02, + "eval_loss": 0.355086088180542, + "eval_runtime": 2825.2162, + "eval_samples_per_second": 4.51, + "eval_steps_per_second": 0.282, + "eval_wer": 7.754611536657809, + "step": 27000 + }, + { + "epoch": 271.27, + "learning_rate": 3.286329113924051e-06, + "loss": 0.0, + "step": 27025 + }, + { + "epoch": 271.52, + "learning_rate": 3.2800000000000004e-06, + "loss": 0.0, + "step": 27050 + }, + { + "epoch": 271.77, + "learning_rate": 3.2736708860759498e-06, + "loss": 0.0, + "step": 27075 + }, + { + "epoch": 272.02, + "learning_rate": 3.2673417721518992e-06, + "loss": 0.0, + "step": 27100 + }, + { + "epoch": 272.27, + "learning_rate": 3.2610126582278482e-06, + "loss": 0.0, + "step": 27125 + }, + { + "epoch": 272.52, + "learning_rate": 3.2546835443037976e-06, + "loss": 0.0, + "step": 27150 + }, + { + "epoch": 272.77, + "learning_rate": 3.248354430379747e-06, + "loss": 0.0, + "step": 27175 + }, + { + "epoch": 273.02, + "learning_rate": 3.2420253164556965e-06, + "loss": 0.0, + "step": 27200 + }, + { + "epoch": 273.27, + "learning_rate": 3.235696202531646e-06, + "loss": 0.0, + "step": 27225 + }, + { + "epoch": 273.53, + "learning_rate": 3.229367088607595e-06, + "loss": 0.0, + "step": 27250 + }, + { + "epoch": 273.78, + "learning_rate": 3.2230379746835444e-06, + "loss": 0.0, + "step": 27275 + }, + { + "epoch": 274.03, + "learning_rate": 3.216708860759494e-06, + "loss": 0.0, + "step": 27300 + }, + { + "epoch": 274.28, + "learning_rate": 3.2103797468354432e-06, + "loss": 0.0, + "step": 27325 + }, + { + "epoch": 274.53, + "learning_rate": 3.2040506329113922e-06, + "loss": 0.0, + "step": 27350 + }, + { + "epoch": 274.78, + "learning_rate": 3.1977215189873425e-06, + "loss": 0.0, + "step": 27375 + }, + { + "epoch": 275.03, + "learning_rate": 3.1913924050632915e-06, + "loss": 0.0, + "step": 27400 + }, + { + "epoch": 275.28, + "learning_rate": 3.185063291139241e-06, + "loss": 0.0, + "step": 27425 + }, + { + "epoch": 275.53, + "learning_rate": 3.1787341772151904e-06, + "loss": 0.0, + "step": 27450 + }, + { + "epoch": 275.78, + "learning_rate": 3.1724050632911398e-06, + "loss": 0.0, + "step": 27475 + }, + { + "epoch": 276.04, + "learning_rate": 3.1660759493670888e-06, + "loss": 0.0, + "step": 27500 + }, + { + "epoch": 276.29, + "learning_rate": 3.1597468354430382e-06, + "loss": 0.0, + "step": 27525 + }, + { + "epoch": 276.54, + "learning_rate": 3.1534177215189876e-06, + "loss": 0.0, + "step": 27550 + }, + { + "epoch": 276.79, + "learning_rate": 3.147088607594937e-06, + "loss": 0.0, + "step": 27575 + }, + { + "epoch": 277.04, + "learning_rate": 3.140759493670886e-06, + "loss": 0.0, + "step": 27600 + }, + { + "epoch": 277.29, + "learning_rate": 3.1344303797468355e-06, + "loss": 0.0, + "step": 27625 + }, + { + "epoch": 277.54, + "learning_rate": 3.128101265822785e-06, + "loss": 0.0, + "step": 27650 + }, + { + "epoch": 277.79, + "learning_rate": 3.1217721518987344e-06, + "loss": 0.0, + "step": 27675 + }, + { + "epoch": 278.04, + "learning_rate": 3.1154430379746838e-06, + "loss": 0.0, + "step": 27700 + }, + { + "epoch": 278.29, + "learning_rate": 3.109113924050633e-06, + "loss": 0.0, + "step": 27725 + }, + { + "epoch": 278.54, + "learning_rate": 3.1027848101265822e-06, + "loss": 0.0, + "step": 27750 + }, + { + "epoch": 278.8, + "learning_rate": 3.096455696202532e-06, + "loss": 0.0, + "step": 27775 + }, + { + "epoch": 279.05, + "learning_rate": 3.0901265822784815e-06, + "loss": 0.0, + "step": 27800 + }, + { + "epoch": 279.3, + "learning_rate": 3.083797468354431e-06, + "loss": 0.0, + "step": 27825 + }, + { + "epoch": 279.55, + "learning_rate": 3.0774683544303804e-06, + "loss": 0.0, + "step": 27850 + }, + { + "epoch": 279.8, + "learning_rate": 3.0711392405063294e-06, + "loss": 0.0, + "step": 27875 + }, + { + "epoch": 280.05, + "learning_rate": 3.0648101265822788e-06, + "loss": 0.0, + "step": 27900 + }, + { + "epoch": 280.3, + "learning_rate": 3.058481012658228e-06, + "loss": 0.0, + "step": 27925 + }, + { + "epoch": 280.55, + "learning_rate": 3.0521518987341776e-06, + "loss": 0.0, + "step": 27950 + }, + { + "epoch": 280.8, + "learning_rate": 3.0458227848101266e-06, + "loss": 0.0, + "step": 27975 + }, + { + "epoch": 281.05, + "learning_rate": 3.039493670886076e-06, + "loss": 0.0, + "step": 28000 + }, + { + "epoch": 281.05, + "eval_loss": 0.36458390951156616, + "eval_runtime": 2827.7596, + "eval_samples_per_second": 4.506, + "eval_steps_per_second": 0.282, + "eval_wer": 7.798577458183524, + "step": 28000 + }, + { + "epoch": 281.3, + "learning_rate": 3.0331645569620255e-06, + "loss": 0.0, + "step": 28025 + }, + { + "epoch": 281.56, + "learning_rate": 3.026835443037975e-06, + "loss": 0.0, + "step": 28050 + }, + { + "epoch": 281.81, + "learning_rate": 3.0205063291139244e-06, + "loss": 0.0, + "step": 28075 + }, + { + "epoch": 282.06, + "learning_rate": 3.0141772151898734e-06, + "loss": 0.0, + "step": 28100 + }, + { + "epoch": 282.31, + "learning_rate": 3.0078481012658228e-06, + "loss": 0.0, + "step": 28125 + }, + { + "epoch": 282.56, + "learning_rate": 3.0015189873417722e-06, + "loss": 0.0, + "step": 28150 + }, + { + "epoch": 282.81, + "learning_rate": 2.9951898734177216e-06, + "loss": 0.0, + "step": 28175 + }, + { + "epoch": 283.06, + "learning_rate": 2.9888607594936715e-06, + "loss": 0.0, + "step": 28200 + }, + { + "epoch": 283.31, + "learning_rate": 2.982531645569621e-06, + "loss": 0.0, + "step": 28225 + }, + { + "epoch": 283.56, + "learning_rate": 2.97620253164557e-06, + "loss": 0.0, + "step": 28250 + }, + { + "epoch": 283.81, + "learning_rate": 2.9698734177215194e-06, + "loss": 0.0, + "step": 28275 + }, + { + "epoch": 284.07, + "learning_rate": 2.9635443037974688e-06, + "loss": 0.0, + "step": 28300 + }, + { + "epoch": 284.32, + "learning_rate": 2.957215189873418e-06, + "loss": 0.0, + "step": 28325 + }, + { + "epoch": 284.57, + "learning_rate": 2.950886075949367e-06, + "loss": 0.0, + "step": 28350 + }, + { + "epoch": 284.82, + "learning_rate": 2.9445569620253166e-06, + "loss": 0.0, + "step": 28375 + }, + { + "epoch": 285.07, + "learning_rate": 2.938227848101266e-06, + "loss": 0.0, + "step": 28400 + }, + { + "epoch": 285.32, + "learning_rate": 2.9318987341772155e-06, + "loss": 0.0, + "step": 28425 + }, + { + "epoch": 285.57, + "learning_rate": 2.925569620253165e-06, + "loss": 0.0, + "step": 28450 + }, + { + "epoch": 285.82, + "learning_rate": 2.919240506329114e-06, + "loss": 0.0, + "step": 28475 + }, + { + "epoch": 286.07, + "learning_rate": 2.9129113924050634e-06, + "loss": 0.0, + "step": 28500 + }, + { + "epoch": 286.32, + "learning_rate": 2.9065822784810128e-06, + "loss": 0.0, + "step": 28525 + }, + { + "epoch": 286.57, + "learning_rate": 2.900253164556962e-06, + "loss": 0.0, + "step": 28550 + }, + { + "epoch": 286.83, + "learning_rate": 2.8939240506329112e-06, + "loss": 0.0, + "step": 28575 + }, + { + "epoch": 287.08, + "learning_rate": 2.887594936708861e-06, + "loss": 0.0, + "step": 28600 + }, + { + "epoch": 287.33, + "learning_rate": 2.8812658227848105e-06, + "loss": 0.0, + "step": 28625 + }, + { + "epoch": 287.58, + "learning_rate": 2.87493670886076e-06, + "loss": 0.0, + "step": 28650 + }, + { + "epoch": 287.83, + "learning_rate": 2.8686075949367093e-06, + "loss": 0.0, + "step": 28675 + }, + { + "epoch": 288.08, + "learning_rate": 2.8622784810126588e-06, + "loss": 0.0, + "step": 28700 + }, + { + "epoch": 288.33, + "learning_rate": 2.8559493670886078e-06, + "loss": 0.0, + "step": 28725 + }, + { + "epoch": 288.58, + "learning_rate": 2.849620253164557e-06, + "loss": 0.0, + "step": 28750 + }, + { + "epoch": 288.83, + "learning_rate": 2.8432911392405066e-06, + "loss": 0.0, + "step": 28775 + }, + { + "epoch": 289.08, + "learning_rate": 2.836962025316456e-06, + "loss": 0.0, + "step": 28800 + }, + { + "epoch": 289.34, + "learning_rate": 2.830632911392405e-06, + "loss": 0.0, + "step": 28825 + }, + { + "epoch": 289.59, + "learning_rate": 2.8243037974683545e-06, + "loss": 0.0, + "step": 28850 + }, + { + "epoch": 289.84, + "learning_rate": 2.817974683544304e-06, + "loss": 0.0, + "step": 28875 + }, + { + "epoch": 290.09, + "learning_rate": 2.8116455696202533e-06, + "loss": 0.0, + "step": 28900 + }, + { + "epoch": 290.34, + "learning_rate": 2.8053164556962028e-06, + "loss": 0.0, + "step": 28925 + }, + { + "epoch": 290.59, + "learning_rate": 2.7989873417721518e-06, + "loss": 0.0, + "step": 28950 + }, + { + "epoch": 290.84, + "learning_rate": 2.792658227848101e-06, + "loss": 0.0, + "step": 28975 + }, + { + "epoch": 291.09, + "learning_rate": 2.7863291139240506e-06, + "loss": 0.0, + "step": 29000 + }, + { + "epoch": 291.09, + "eval_loss": 0.3728525936603546, + "eval_runtime": 2833.2163, + "eval_samples_per_second": 4.497, + "eval_steps_per_second": 0.281, + "eval_wer": 7.7780600281381895, + "step": 29000 + }, + { + "epoch": 291.34, + "learning_rate": 2.7800000000000005e-06, + "loss": 0.0, + "step": 29025 + }, + { + "epoch": 291.59, + "learning_rate": 2.773924050632912e-06, + "loss": 0.0, + "step": 29050 + }, + { + "epoch": 291.84, + "learning_rate": 2.767594936708861e-06, + "loss": 0.0, + "step": 29075 + }, + { + "epoch": 292.1, + "learning_rate": 2.7612658227848104e-06, + "loss": 0.0, + "step": 29100 + }, + { + "epoch": 292.35, + "learning_rate": 2.75493670886076e-06, + "loss": 0.0, + "step": 29125 + }, + { + "epoch": 292.6, + "learning_rate": 2.7486075949367092e-06, + "loss": 0.0, + "step": 29150 + }, + { + "epoch": 292.85, + "learning_rate": 2.7422784810126587e-06, + "loss": 0.0, + "step": 29175 + }, + { + "epoch": 293.1, + "learning_rate": 2.7359493670886077e-06, + "loss": 0.0, + "step": 29200 + }, + { + "epoch": 293.35, + "learning_rate": 2.729620253164557e-06, + "loss": 0.0, + "step": 29225 + }, + { + "epoch": 293.6, + "learning_rate": 2.7232911392405065e-06, + "loss": 0.0, + "step": 29250 + }, + { + "epoch": 293.85, + "learning_rate": 2.716962025316456e-06, + "loss": 0.0, + "step": 29275 + }, + { + "epoch": 294.1, + "learning_rate": 2.710632911392405e-06, + "loss": 0.0, + "step": 29300 + }, + { + "epoch": 294.35, + "learning_rate": 2.7043037974683544e-06, + "loss": 0.0, + "step": 29325 + }, + { + "epoch": 294.6, + "learning_rate": 2.697974683544304e-06, + "loss": 0.0, + "step": 29350 + }, + { + "epoch": 294.86, + "learning_rate": 2.6916455696202532e-06, + "loss": 0.0, + "step": 29375 + }, + { + "epoch": 295.11, + "learning_rate": 2.685316455696203e-06, + "loss": 0.0, + "step": 29400 + }, + { + "epoch": 295.36, + "learning_rate": 2.6789873417721525e-06, + "loss": 0.0, + "step": 29425 + }, + { + "epoch": 295.61, + "learning_rate": 2.6726582278481015e-06, + "loss": 0.0, + "step": 29450 + }, + { + "epoch": 295.86, + "learning_rate": 2.666329113924051e-06, + "loss": 0.0, + "step": 29475 + }, + { + "epoch": 296.11, + "learning_rate": 2.6600000000000004e-06, + "loss": 0.0, + "step": 29500 + }, + { + "epoch": 296.36, + "learning_rate": 2.65367088607595e-06, + "loss": 0.0, + "step": 29525 + }, + { + "epoch": 296.61, + "learning_rate": 2.6473417721518992e-06, + "loss": 0.0, + "step": 29550 + }, + { + "epoch": 296.86, + "learning_rate": 2.6410126582278482e-06, + "loss": 0.0, + "step": 29575 + }, + { + "epoch": 297.11, + "learning_rate": 2.6346835443037977e-06, + "loss": 0.0, + "step": 29600 + }, + { + "epoch": 297.37, + "learning_rate": 2.628354430379747e-06, + "loss": 0.0, + "step": 29625 + }, + { + "epoch": 297.62, + "learning_rate": 2.6220253164556965e-06, + "loss": 0.0, + "step": 29650 + }, + { + "epoch": 297.87, + "learning_rate": 2.6156962025316455e-06, + "loss": 0.0, + "step": 29675 + }, + { + "epoch": 298.12, + "learning_rate": 2.609367088607595e-06, + "loss": 0.0, + "step": 29700 + }, + { + "epoch": 298.37, + "learning_rate": 2.6030379746835444e-06, + "loss": 0.0, + "step": 29725 + }, + { + "epoch": 298.62, + "learning_rate": 2.596708860759494e-06, + "loss": 0.0, + "step": 29750 + }, + { + "epoch": 298.87, + "learning_rate": 2.5903797468354432e-06, + "loss": 0.0, + "step": 29775 + }, + { + "epoch": 299.12, + "learning_rate": 2.5840506329113922e-06, + "loss": 0.0, + "step": 29800 + }, + { + "epoch": 299.37, + "learning_rate": 2.577721518987342e-06, + "loss": 0.0, + "step": 29825 + }, + { + "epoch": 299.62, + "learning_rate": 2.5713924050632915e-06, + "loss": 0.0, + "step": 29850 + }, + { + "epoch": 299.87, + "learning_rate": 2.565063291139241e-06, + "loss": 0.0, + "step": 29875 + }, + { + "epoch": 300.13, + "learning_rate": 2.5587341772151904e-06, + "loss": 0.0, + "step": 29900 + }, + { + "epoch": 300.38, + "learning_rate": 2.5524050632911394e-06, + "loss": 0.0, + "step": 29925 + }, + { + "epoch": 300.63, + "learning_rate": 2.546075949367089e-06, + "loss": 0.0, + "step": 29950 + }, + { + "epoch": 300.88, + "learning_rate": 2.5397468354430382e-06, + "loss": 0.0, + "step": 29975 + }, + { + "epoch": 301.13, + "learning_rate": 2.5334177215189876e-06, + "loss": 0.0, + "step": 30000 + }, + { + "epoch": 301.13, + "eval_loss": 0.38112595677375793, + "eval_runtime": 2830.2491, + "eval_samples_per_second": 4.502, + "eval_steps_per_second": 0.282, + "eval_wer": 7.763404720962952, + "step": 30000 + }, + { + "epoch": 301.38, + "learning_rate": 2.527088607594937e-06, + "loss": 0.0, + "step": 30025 + }, + { + "epoch": 301.63, + "learning_rate": 2.520759493670886e-06, + "loss": 0.0, + "step": 30050 + }, + { + "epoch": 301.88, + "learning_rate": 2.5144303797468355e-06, + "loss": 0.0, + "step": 30075 + }, + { + "epoch": 302.13, + "learning_rate": 2.508101265822785e-06, + "loss": 0.0, + "step": 30100 + }, + { + "epoch": 302.38, + "learning_rate": 2.5017721518987344e-06, + "loss": 0.0, + "step": 30125 + }, + { + "epoch": 302.63, + "learning_rate": 2.495443037974684e-06, + "loss": 0.0, + "step": 30150 + }, + { + "epoch": 302.89, + "learning_rate": 2.4891139240506332e-06, + "loss": 0.0, + "step": 30175 + }, + { + "epoch": 303.14, + "learning_rate": 2.4827848101265826e-06, + "loss": 0.0, + "step": 30200 + }, + { + "epoch": 303.39, + "learning_rate": 2.4764556962025317e-06, + "loss": 0.0, + "step": 30225 + }, + { + "epoch": 303.64, + "learning_rate": 2.470126582278481e-06, + "loss": 0.0, + "step": 30250 + }, + { + "epoch": 303.89, + "learning_rate": 2.4637974683544305e-06, + "loss": 0.0, + "step": 30275 + }, + { + "epoch": 304.14, + "learning_rate": 2.45746835443038e-06, + "loss": 0.0, + "step": 30300 + }, + { + "epoch": 304.39, + "learning_rate": 2.4511392405063294e-06, + "loss": 0.0, + "step": 30325 + }, + { + "epoch": 304.64, + "learning_rate": 2.444810126582279e-06, + "loss": 0.0, + "step": 30350 + }, + { + "epoch": 304.89, + "learning_rate": 2.4384810126582282e-06, + "loss": 0.0, + "step": 30375 + }, + { + "epoch": 305.14, + "learning_rate": 2.4321518987341776e-06, + "loss": 0.0, + "step": 30400 + }, + { + "epoch": 305.4, + "learning_rate": 2.4258227848101266e-06, + "loss": 0.0, + "step": 30425 + }, + { + "epoch": 305.65, + "learning_rate": 2.419493670886076e-06, + "loss": 0.0, + "step": 30450 + }, + { + "epoch": 305.9, + "learning_rate": 2.4131645569620255e-06, + "loss": 0.0, + "step": 30475 + }, + { + "epoch": 306.15, + "learning_rate": 2.406835443037975e-06, + "loss": 0.0, + "step": 30500 + }, + { + "epoch": 306.4, + "learning_rate": 2.400506329113924e-06, + "loss": 0.0, + "step": 30525 + }, + { + "epoch": 306.65, + "learning_rate": 2.394177215189874e-06, + "loss": 0.0, + "step": 30550 + }, + { + "epoch": 306.9, + "learning_rate": 2.3878481012658232e-06, + "loss": 0.0, + "step": 30575 + }, + { + "epoch": 307.15, + "learning_rate": 2.3815189873417722e-06, + "loss": 0.0, + "step": 30600 + }, + { + "epoch": 307.4, + "learning_rate": 2.3751898734177216e-06, + "loss": 0.0, + "step": 30625 + }, + { + "epoch": 307.65, + "learning_rate": 2.368860759493671e-06, + "loss": 0.0, + "step": 30650 + }, + { + "epoch": 307.9, + "learning_rate": 2.3625316455696205e-06, + "loss": 0.0, + "step": 30675 + }, + { + "epoch": 308.16, + "learning_rate": 2.35620253164557e-06, + "loss": 0.0, + "step": 30700 + }, + { + "epoch": 308.41, + "learning_rate": 2.349873417721519e-06, + "loss": 0.0, + "step": 30725 + }, + { + "epoch": 308.66, + "learning_rate": 2.3435443037974688e-06, + "loss": 0.0, + "step": 30750 + }, + { + "epoch": 308.91, + "learning_rate": 2.337215189873418e-06, + "loss": 0.0, + "step": 30775 + }, + { + "epoch": 309.16, + "learning_rate": 2.3308860759493672e-06, + "loss": 0.0, + "step": 30800 + }, + { + "epoch": 309.41, + "learning_rate": 2.3245569620253166e-06, + "loss": 0.0, + "step": 30825 + }, + { + "epoch": 309.66, + "learning_rate": 2.318227848101266e-06, + "loss": 0.0, + "step": 30850 + }, + { + "epoch": 309.91, + "learning_rate": 2.3118987341772155e-06, + "loss": 0.0, + "step": 30875 + }, + { + "epoch": 310.16, + "learning_rate": 2.3055696202531645e-06, + "loss": 0.0, + "step": 30900 + }, + { + "epoch": 310.41, + "learning_rate": 2.299240506329114e-06, + "loss": 0.0, + "step": 30925 + }, + { + "epoch": 310.66, + "learning_rate": 2.2929113924050638e-06, + "loss": 0.0, + "step": 30950 + }, + { + "epoch": 310.92, + "learning_rate": 2.286582278481013e-06, + "loss": 0.0, + "step": 30975 + }, + { + "epoch": 311.17, + "learning_rate": 2.2802531645569622e-06, + "loss": 0.0, + "step": 31000 + }, + { + "epoch": 311.17, + "eval_loss": 0.38778597116470337, + "eval_runtime": 2828.9196, + "eval_samples_per_second": 4.504, + "eval_steps_per_second": 0.282, + "eval_wer": 7.7702438643113965, + "step": 31000 + }, + { + "epoch": 311.42, + "learning_rate": 2.2739240506329116e-06, + "loss": 0.0, + "step": 31025 + }, + { + "epoch": 311.67, + "learning_rate": 2.267594936708861e-06, + "loss": 0.0, + "step": 31050 + }, + { + "epoch": 311.92, + "learning_rate": 2.261518987341772e-06, + "loss": 0.0, + "step": 31075 + }, + { + "epoch": 312.17, + "learning_rate": 2.2551898734177215e-06, + "loss": 0.0, + "step": 31100 + }, + { + "epoch": 312.42, + "learning_rate": 2.248860759493671e-06, + "loss": 0.0, + "step": 31125 + }, + { + "epoch": 312.67, + "learning_rate": 2.2425316455696204e-06, + "loss": 0.0, + "step": 31150 + }, + { + "epoch": 312.92, + "learning_rate": 2.23620253164557e-06, + "loss": 0.0, + "step": 31175 + }, + { + "epoch": 313.17, + "learning_rate": 2.2298734177215192e-06, + "loss": 0.0, + "step": 31200 + }, + { + "epoch": 313.43, + "learning_rate": 2.2235443037974687e-06, + "loss": 0.0, + "step": 31225 + }, + { + "epoch": 313.68, + "learning_rate": 2.217215189873418e-06, + "loss": 0.0, + "step": 31250 + }, + { + "epoch": 313.93, + "learning_rate": 2.210886075949367e-06, + "loss": 0.0, + "step": 31275 + }, + { + "epoch": 314.18, + "learning_rate": 2.2045569620253165e-06, + "loss": 0.0, + "step": 31300 + }, + { + "epoch": 314.43, + "learning_rate": 2.198227848101266e-06, + "loss": 0.0, + "step": 31325 + }, + { + "epoch": 314.68, + "learning_rate": 2.1918987341772154e-06, + "loss": 0.0, + "step": 31350 + }, + { + "epoch": 314.93, + "learning_rate": 2.185569620253165e-06, + "loss": 0.0, + "step": 31375 + }, + { + "epoch": 315.18, + "learning_rate": 2.1792405063291142e-06, + "loss": 0.0, + "step": 31400 + }, + { + "epoch": 315.43, + "learning_rate": 2.1729113924050637e-06, + "loss": 0.0, + "step": 31425 + }, + { + "epoch": 315.68, + "learning_rate": 2.1665822784810127e-06, + "loss": 0.0, + "step": 31450 + }, + { + "epoch": 315.93, + "learning_rate": 2.160253164556962e-06, + "loss": 0.0, + "step": 31475 + }, + { + "epoch": 316.19, + "learning_rate": 2.1539240506329115e-06, + "loss": 0.0, + "step": 31500 + }, + { + "epoch": 316.44, + "learning_rate": 2.147594936708861e-06, + "loss": 0.0, + "step": 31525 + }, + { + "epoch": 316.69, + "learning_rate": 2.1412658227848104e-06, + "loss": 0.0, + "step": 31550 + }, + { + "epoch": 316.94, + "learning_rate": 2.13493670886076e-06, + "loss": 0.0, + "step": 31575 + }, + { + "epoch": 317.19, + "learning_rate": 2.1286075949367092e-06, + "loss": 0.0, + "step": 31600 + }, + { + "epoch": 317.44, + "learning_rate": 2.1222784810126582e-06, + "loss": 0.0, + "step": 31625 + }, + { + "epoch": 317.69, + "learning_rate": 2.1159493670886077e-06, + "loss": 0.0, + "step": 31650 + }, + { + "epoch": 317.94, + "learning_rate": 2.109620253164557e-06, + "loss": 0.0, + "step": 31675 + }, + { + "epoch": 318.19, + "learning_rate": 2.1032911392405065e-06, + "loss": 0.0, + "step": 31700 + }, + { + "epoch": 318.44, + "learning_rate": 2.096962025316456e-06, + "loss": 0.0, + "step": 31725 + }, + { + "epoch": 318.7, + "learning_rate": 2.090632911392405e-06, + "loss": 0.0, + "step": 31750 + }, + { + "epoch": 318.95, + "learning_rate": 2.084303797468355e-06, + "loss": 0.0, + "step": 31775 + }, + { + "epoch": 319.2, + "learning_rate": 2.0779746835443042e-06, + "loss": 0.0, + "step": 31800 + }, + { + "epoch": 319.45, + "learning_rate": 2.0716455696202532e-06, + "loss": 0.0, + "step": 31825 + }, + { + "epoch": 319.7, + "learning_rate": 2.0653164556962027e-06, + "loss": 0.0, + "step": 31850 + }, + { + "epoch": 319.95, + "learning_rate": 2.058987341772152e-06, + "loss": 0.0, + "step": 31875 + }, + { + "epoch": 320.2, + "learning_rate": 2.0526582278481015e-06, + "loss": 0.0, + "step": 31900 + }, + { + "epoch": 320.45, + "learning_rate": 2.0463291139240505e-06, + "loss": 0.0, + "step": 31925 + }, + { + "epoch": 320.7, + "learning_rate": 2.04e-06, + "loss": 0.0, + "step": 31950 + }, + { + "epoch": 320.95, + "learning_rate": 2.03367088607595e-06, + "loss": 0.0, + "step": 31975 + }, + { + "epoch": 321.2, + "learning_rate": 2.027341772151899e-06, + "loss": 0.0, + "step": 32000 + }, + { + "epoch": 321.2, + "eval_loss": 0.39481475949287415, + "eval_runtime": 2829.7401, + "eval_samples_per_second": 4.503, + "eval_steps_per_second": 0.282, + "eval_wer": 7.772197905268094, + "step": 32000 + }, + { + "epoch": 321.46, + "learning_rate": 2.0210126582278482e-06, + "loss": 0.0, + "step": 32025 + }, + { + "epoch": 321.71, + "learning_rate": 2.0146835443037977e-06, + "loss": 0.0, + "step": 32050 + }, + { + "epoch": 321.96, + "learning_rate": 2.008354430379747e-06, + "loss": 0.0, + "step": 32075 + }, + { + "epoch": 322.21, + "learning_rate": 2.0020253164556965e-06, + "loss": 0.0, + "step": 32100 + }, + { + "epoch": 322.46, + "learning_rate": 1.9956962025316455e-06, + "loss": 0.0, + "step": 32125 + }, + { + "epoch": 322.71, + "learning_rate": 1.989367088607595e-06, + "loss": 0.0, + "step": 32150 + }, + { + "epoch": 322.96, + "learning_rate": 1.9830379746835444e-06, + "loss": 0.0, + "step": 32175 + }, + { + "epoch": 323.21, + "learning_rate": 1.976708860759494e-06, + "loss": 0.0, + "step": 32200 + }, + { + "epoch": 323.46, + "learning_rate": 1.9703797468354432e-06, + "loss": 0.0, + "step": 32225 + }, + { + "epoch": 323.71, + "learning_rate": 1.9640506329113927e-06, + "loss": 0.0, + "step": 32250 + }, + { + "epoch": 323.96, + "learning_rate": 1.957721518987342e-06, + "loss": 0.0, + "step": 32275 + }, + { + "epoch": 324.22, + "learning_rate": 1.951392405063291e-06, + "loss": 0.0, + "step": 32300 + }, + { + "epoch": 324.47, + "learning_rate": 1.9450632911392405e-06, + "loss": 0.0, + "step": 32325 + }, + { + "epoch": 324.72, + "learning_rate": 1.93873417721519e-06, + "loss": 0.0, + "step": 32350 + }, + { + "epoch": 324.97, + "learning_rate": 1.9324050632911394e-06, + "loss": 0.0, + "step": 32375 + }, + { + "epoch": 325.22, + "learning_rate": 1.926075949367089e-06, + "loss": 0.0, + "step": 32400 + }, + { + "epoch": 325.47, + "learning_rate": 1.9197468354430382e-06, + "loss": 0.0, + "step": 32425 + }, + { + "epoch": 325.72, + "learning_rate": 1.9134177215189877e-06, + "loss": 0.0, + "step": 32450 + }, + { + "epoch": 325.97, + "learning_rate": 1.9070886075949369e-06, + "loss": 0.0, + "step": 32475 + }, + { + "epoch": 326.22, + "learning_rate": 1.900759493670886e-06, + "loss": 0.0, + "step": 32500 + }, + { + "epoch": 326.47, + "learning_rate": 1.8944303797468355e-06, + "loss": 0.0, + "step": 32525 + }, + { + "epoch": 326.73, + "learning_rate": 1.888101265822785e-06, + "loss": 0.0, + "step": 32550 + }, + { + "epoch": 326.98, + "learning_rate": 1.8817721518987344e-06, + "loss": 0.0, + "step": 32575 + }, + { + "epoch": 327.23, + "learning_rate": 1.8754430379746838e-06, + "loss": 0.0, + "step": 32600 + }, + { + "epoch": 327.48, + "learning_rate": 1.8691139240506332e-06, + "loss": 0.0, + "step": 32625 + }, + { + "epoch": 327.73, + "learning_rate": 1.8627848101265824e-06, + "loss": 0.0, + "step": 32650 + }, + { + "epoch": 327.98, + "learning_rate": 1.8564556962025319e-06, + "loss": 0.0, + "step": 32675 + }, + { + "epoch": 328.23, + "learning_rate": 1.850126582278481e-06, + "loss": 0.0, + "step": 32700 + }, + { + "epoch": 328.48, + "learning_rate": 1.8437974683544305e-06, + "loss": 0.0, + "step": 32725 + }, + { + "epoch": 328.73, + "learning_rate": 1.8374683544303797e-06, + "loss": 0.0, + "step": 32750 + }, + { + "epoch": 328.98, + "learning_rate": 1.8311392405063292e-06, + "loss": 0.0, + "step": 32775 + }, + { + "epoch": 329.23, + "learning_rate": 1.8248101265822788e-06, + "loss": 0.0, + "step": 32800 + }, + { + "epoch": 329.49, + "learning_rate": 1.818481012658228e-06, + "loss": 0.0, + "step": 32825 + }, + { + "epoch": 329.74, + "learning_rate": 1.8121518987341774e-06, + "loss": 0.0, + "step": 32850 + }, + { + "epoch": 329.99, + "learning_rate": 1.8058227848101267e-06, + "loss": 0.0, + "step": 32875 + }, + { + "epoch": 330.24, + "learning_rate": 1.799493670886076e-06, + "loss": 0.0, + "step": 32900 + }, + { + "epoch": 330.49, + "learning_rate": 1.7931645569620253e-06, + "loss": 0.0, + "step": 32925 + }, + { + "epoch": 330.74, + "learning_rate": 1.7868354430379747e-06, + "loss": 0.0, + "step": 32950 + }, + { + "epoch": 330.99, + "learning_rate": 1.7805063291139242e-06, + "loss": 0.0, + "step": 32975 + }, + { + "epoch": 331.24, + "learning_rate": 1.7741772151898736e-06, + "loss": 0.0, + "step": 33000 + }, + { + "epoch": 331.24, + "eval_loss": 0.4002886712551117, + "eval_runtime": 2835.03, + "eval_samples_per_second": 4.494, + "eval_steps_per_second": 0.281, + "eval_wer": 7.730186024699077, + "step": 33000 + }, + { + "epoch": 331.49, + "learning_rate": 1.767848101265823e-06, + "loss": 0.0, + "step": 33025 + }, + { + "epoch": 331.74, + "learning_rate": 1.7615189873417724e-06, + "loss": 0.0, + "step": 33050 + }, + { + "epoch": 331.99, + "learning_rate": 1.7551898734177217e-06, + "loss": 0.0, + "step": 33075 + }, + { + "epoch": 332.25, + "learning_rate": 1.748860759493671e-06, + "loss": 0.0, + "step": 33100 + }, + { + "epoch": 332.5, + "learning_rate": 1.7427848101265823e-06, + "loss": 0.0, + "step": 33125 + }, + { + "epoch": 332.75, + "learning_rate": 1.7364556962025317e-06, + "loss": 0.0, + "step": 33150 + }, + { + "epoch": 333.0, + "learning_rate": 1.7301265822784814e-06, + "loss": 0.0, + "step": 33175 + }, + { + "epoch": 333.25, + "learning_rate": 1.7237974683544306e-06, + "loss": 0.0, + "step": 33200 + }, + { + "epoch": 333.5, + "learning_rate": 1.71746835443038e-06, + "loss": 0.0, + "step": 33225 + }, + { + "epoch": 333.75, + "learning_rate": 1.7111392405063292e-06, + "loss": 0.0, + "step": 33250 + }, + { + "epoch": 334.0, + "learning_rate": 1.7048101265822787e-06, + "loss": 0.0, + "step": 33275 + }, + { + "epoch": 334.25, + "learning_rate": 1.6984810126582279e-06, + "loss": 0.0, + "step": 33300 + }, + { + "epoch": 334.5, + "learning_rate": 1.6921518987341773e-06, + "loss": 0.0, + "step": 33325 + }, + { + "epoch": 334.76, + "learning_rate": 1.6858227848101265e-06, + "loss": 0.0, + "step": 33350 + }, + { + "epoch": 335.01, + "learning_rate": 1.6797468354430382e-06, + "loss": 0.0, + "step": 33375 + }, + { + "epoch": 335.26, + "learning_rate": 1.6734177215189876e-06, + "loss": 0.0, + "step": 33400 + }, + { + "epoch": 335.51, + "learning_rate": 1.6670886075949368e-06, + "loss": 0.0, + "step": 33425 + }, + { + "epoch": 335.76, + "learning_rate": 1.6607594936708863e-06, + "loss": 0.0, + "step": 33450 + }, + { + "epoch": 336.01, + "learning_rate": 1.6544303797468355e-06, + "loss": 0.0, + "step": 33475 + }, + { + "epoch": 336.26, + "learning_rate": 1.648101265822785e-06, + "loss": 0.0, + "step": 33500 + }, + { + "epoch": 336.51, + "learning_rate": 1.6417721518987343e-06, + "loss": 0.0, + "step": 33525 + }, + { + "epoch": 336.76, + "learning_rate": 1.6354430379746836e-06, + "loss": 0.0, + "step": 33550 + }, + { + "epoch": 337.01, + "learning_rate": 1.6291139240506332e-06, + "loss": 0.0, + "step": 33575 + }, + { + "epoch": 337.26, + "learning_rate": 1.6227848101265824e-06, + "loss": 0.0, + "step": 33600 + }, + { + "epoch": 337.52, + "learning_rate": 1.6164556962025318e-06, + "loss": 0.0, + "step": 33625 + }, + { + "epoch": 337.77, + "learning_rate": 1.6101265822784813e-06, + "loss": 0.0, + "step": 33650 + }, + { + "epoch": 338.02, + "learning_rate": 1.6037974683544305e-06, + "loss": 0.0, + "step": 33675 + }, + { + "epoch": 338.27, + "learning_rate": 1.59746835443038e-06, + "loss": 0.0, + "step": 33700 + }, + { + "epoch": 338.52, + "learning_rate": 1.5911392405063291e-06, + "loss": 0.0, + "step": 33725 + }, + { + "epoch": 338.77, + "learning_rate": 1.5848101265822786e-06, + "loss": 0.0, + "step": 33750 + }, + { + "epoch": 339.02, + "learning_rate": 1.5784810126582278e-06, + "loss": 0.0, + "step": 33775 + }, + { + "epoch": 339.27, + "learning_rate": 1.5721518987341774e-06, + "loss": 0.0, + "step": 33800 + }, + { + "epoch": 339.52, + "learning_rate": 1.5658227848101268e-06, + "loss": 0.0, + "step": 33825 + }, + { + "epoch": 339.77, + "learning_rate": 1.559493670886076e-06, + "loss": 0.0, + "step": 33850 + }, + { + "epoch": 340.03, + "learning_rate": 1.5531645569620255e-06, + "loss": 0.0, + "step": 33875 + }, + { + "epoch": 340.28, + "learning_rate": 1.5468354430379747e-06, + "loss": 0.0, + "step": 33900 + }, + { + "epoch": 340.53, + "learning_rate": 1.5405063291139241e-06, + "loss": 0.0, + "step": 33925 + }, + { + "epoch": 340.78, + "learning_rate": 1.5341772151898736e-06, + "loss": 0.0, + "step": 33950 + }, + { + "epoch": 341.03, + "learning_rate": 1.5278481012658228e-06, + "loss": 0.0, + "step": 33975 + }, + { + "epoch": 341.28, + "learning_rate": 1.5215189873417724e-06, + "loss": 0.0, + "step": 34000 + }, + { + "epoch": 341.28, + "eval_loss": 0.4057931900024414, + "eval_runtime": 2832.6042, + "eval_samples_per_second": 4.498, + "eval_steps_per_second": 0.281, + "eval_wer": 7.731163045177427, + "step": 34000 + }, + { + "epoch": 341.53, + "learning_rate": 1.5151898734177216e-06, + "loss": 0.0, + "step": 34025 + }, + { + "epoch": 341.78, + "learning_rate": 1.508860759493671e-06, + "loss": 0.0, + "step": 34050 + }, + { + "epoch": 342.03, + "learning_rate": 1.5025316455696205e-06, + "loss": 0.0, + "step": 34075 + }, + { + "epoch": 342.28, + "learning_rate": 1.4962025316455697e-06, + "loss": 0.0, + "step": 34100 + }, + { + "epoch": 342.53, + "learning_rate": 1.4898734177215191e-06, + "loss": 0.0, + "step": 34125 + }, + { + "epoch": 342.79, + "learning_rate": 1.4835443037974683e-06, + "loss": 0.0, + "step": 34150 + }, + { + "epoch": 343.04, + "learning_rate": 1.4772151898734178e-06, + "loss": 0.0, + "step": 34175 + }, + { + "epoch": 343.29, + "learning_rate": 1.4708860759493674e-06, + "loss": 0.0, + "step": 34200 + }, + { + "epoch": 343.54, + "learning_rate": 1.4645569620253166e-06, + "loss": 0.0, + "step": 34225 + }, + { + "epoch": 343.79, + "learning_rate": 1.458227848101266e-06, + "loss": 0.0, + "step": 34250 + }, + { + "epoch": 344.04, + "learning_rate": 1.4518987341772153e-06, + "loss": 0.0, + "step": 34275 + }, + { + "epoch": 344.29, + "learning_rate": 1.4455696202531647e-06, + "loss": 0.0, + "step": 34300 + }, + { + "epoch": 344.54, + "learning_rate": 1.439240506329114e-06, + "loss": 0.0, + "step": 34325 + }, + { + "epoch": 344.79, + "learning_rate": 1.4329113924050633e-06, + "loss": 0.0, + "step": 34350 + }, + { + "epoch": 345.04, + "learning_rate": 1.4265822784810128e-06, + "loss": 0.0, + "step": 34375 + }, + { + "epoch": 345.29, + "learning_rate": 1.4202531645569622e-06, + "loss": 0.0, + "step": 34400 + }, + { + "epoch": 345.55, + "learning_rate": 1.4139240506329116e-06, + "loss": 0.0, + "step": 34425 + }, + { + "epoch": 345.8, + "learning_rate": 1.4075949367088608e-06, + "loss": 0.0, + "step": 34450 + }, + { + "epoch": 346.05, + "learning_rate": 1.4012658227848103e-06, + "loss": 0.0, + "step": 34475 + }, + { + "epoch": 346.3, + "learning_rate": 1.3949367088607597e-06, + "loss": 0.0, + "step": 34500 + }, + { + "epoch": 346.55, + "learning_rate": 1.388607594936709e-06, + "loss": 0.0, + "step": 34525 + }, + { + "epoch": 346.8, + "learning_rate": 1.3822784810126583e-06, + "loss": 0.0, + "step": 34550 + }, + { + "epoch": 347.05, + "learning_rate": 1.3759493670886076e-06, + "loss": 0.0, + "step": 34575 + }, + { + "epoch": 347.3, + "learning_rate": 1.3696202531645572e-06, + "loss": 0.0, + "step": 34600 + }, + { + "epoch": 347.55, + "learning_rate": 1.3632911392405066e-06, + "loss": 0.0, + "step": 34625 + }, + { + "epoch": 347.8, + "learning_rate": 1.3569620253164558e-06, + "loss": 0.0, + "step": 34650 + }, + { + "epoch": 348.06, + "learning_rate": 1.3506329113924053e-06, + "loss": 0.0, + "step": 34675 + }, + { + "epoch": 348.31, + "learning_rate": 1.3443037974683545e-06, + "loss": 0.0, + "step": 34700 + }, + { + "epoch": 348.56, + "learning_rate": 1.337974683544304e-06, + "loss": 0.0, + "step": 34725 + }, + { + "epoch": 348.81, + "learning_rate": 1.3316455696202531e-06, + "loss": 0.0, + "step": 34750 + }, + { + "epoch": 349.06, + "learning_rate": 1.3253164556962025e-06, + "loss": 0.0, + "step": 34775 + }, + { + "epoch": 349.31, + "learning_rate": 1.318987341772152e-06, + "loss": 0.0, + "step": 34800 + }, + { + "epoch": 349.56, + "learning_rate": 1.3126582278481014e-06, + "loss": 0.0, + "step": 34825 + }, + { + "epoch": 349.81, + "learning_rate": 1.3063291139240508e-06, + "loss": 0.0, + "step": 34850 + }, + { + "epoch": 350.06, + "learning_rate": 1.3e-06, + "loss": 0.0, + "step": 34875 + }, + { + "epoch": 350.31, + "learning_rate": 1.2936708860759495e-06, + "loss": 0.0, + "step": 34900 + }, + { + "epoch": 350.56, + "learning_rate": 1.287341772151899e-06, + "loss": 0.0, + "step": 34925 + }, + { + "epoch": 350.82, + "learning_rate": 1.2810126582278481e-06, + "loss": 0.0, + "step": 34950 + }, + { + "epoch": 351.07, + "learning_rate": 1.2746835443037975e-06, + "loss": 0.0, + "step": 34975 + }, + { + "epoch": 351.32, + "learning_rate": 1.2683544303797468e-06, + "loss": 0.0, + "step": 35000 + }, + { + "epoch": 351.32, + "eval_loss": 0.41078662872314453, + "eval_runtime": 2834.2574, + "eval_samples_per_second": 4.496, + "eval_steps_per_second": 0.281, + "eval_wer": 7.729209004220729, + "step": 35000 + }, + { + "epoch": 351.57, + "learning_rate": 1.2620253164556964e-06, + "loss": 0.0, + "step": 35025 + }, + { + "epoch": 351.82, + "learning_rate": 1.2556962025316458e-06, + "loss": 0.0, + "step": 35050 + }, + { + "epoch": 352.07, + "learning_rate": 1.249367088607595e-06, + "loss": 0.0, + "step": 35075 + }, + { + "epoch": 352.32, + "learning_rate": 1.2430379746835445e-06, + "loss": 0.0, + "step": 35100 + }, + { + "epoch": 352.57, + "learning_rate": 1.2367088607594937e-06, + "loss": 0.0, + "step": 35125 + }, + { + "epoch": 352.82, + "learning_rate": 1.2303797468354431e-06, + "loss": 0.0, + "step": 35150 + }, + { + "epoch": 353.07, + "learning_rate": 1.2240506329113925e-06, + "loss": 0.0, + "step": 35175 + }, + { + "epoch": 353.32, + "learning_rate": 1.217721518987342e-06, + "loss": 0.0, + "step": 35200 + }, + { + "epoch": 353.58, + "learning_rate": 1.2113924050632912e-06, + "loss": 0.0, + "step": 35225 + }, + { + "epoch": 353.83, + "learning_rate": 1.2050632911392406e-06, + "loss": 0.0, + "step": 35250 + }, + { + "epoch": 354.08, + "learning_rate": 1.19873417721519e-06, + "loss": 0.0, + "step": 35275 + }, + { + "epoch": 354.33, + "learning_rate": 1.1924050632911393e-06, + "loss": 0.0, + "step": 35300 + }, + { + "epoch": 354.58, + "learning_rate": 1.1860759493670887e-06, + "loss": 0.0, + "step": 35325 + }, + { + "epoch": 354.83, + "learning_rate": 1.1797468354430381e-06, + "loss": 0.0, + "step": 35350 + }, + { + "epoch": 355.08, + "learning_rate": 1.1734177215189875e-06, + "loss": 0.0, + "step": 35375 + }, + { + "epoch": 355.33, + "learning_rate": 1.1670886075949368e-06, + "loss": 0.0, + "step": 35400 + }, + { + "epoch": 355.58, + "learning_rate": 1.1607594936708862e-06, + "loss": 0.0, + "step": 35425 + }, + { + "epoch": 355.83, + "learning_rate": 1.1544303797468354e-06, + "loss": 0.0, + "step": 35450 + }, + { + "epoch": 356.09, + "learning_rate": 1.148101265822785e-06, + "loss": 0.0, + "step": 35475 + }, + { + "epoch": 356.34, + "learning_rate": 1.1417721518987343e-06, + "loss": 0.0, + "step": 35500 + }, + { + "epoch": 356.59, + "learning_rate": 1.1354430379746837e-06, + "loss": 0.0, + "step": 35525 + }, + { + "epoch": 356.84, + "learning_rate": 1.129113924050633e-06, + "loss": 0.0, + "step": 35550 + }, + { + "epoch": 357.09, + "learning_rate": 1.1227848101265823e-06, + "loss": 0.0, + "step": 35575 + }, + { + "epoch": 357.34, + "learning_rate": 1.1164556962025318e-06, + "loss": 0.0, + "step": 35600 + }, + { + "epoch": 357.59, + "learning_rate": 1.1101265822784812e-06, + "loss": 0.0, + "step": 35625 + }, + { + "epoch": 357.84, + "learning_rate": 1.1037974683544304e-06, + "loss": 0.0, + "step": 35650 + }, + { + "epoch": 358.09, + "learning_rate": 1.0974683544303798e-06, + "loss": 0.0, + "step": 35675 + }, + { + "epoch": 358.34, + "learning_rate": 1.0911392405063293e-06, + "loss": 0.0, + "step": 35700 + }, + { + "epoch": 358.59, + "learning_rate": 1.0848101265822787e-06, + "loss": 0.0, + "step": 35725 + }, + { + "epoch": 358.85, + "learning_rate": 1.078481012658228e-06, + "loss": 0.0, + "step": 35750 + }, + { + "epoch": 359.1, + "learning_rate": 1.0721518987341773e-06, + "loss": 0.0, + "step": 35775 + }, + { + "epoch": 359.35, + "learning_rate": 1.0658227848101268e-06, + "loss": 0.0, + "step": 35800 + }, + { + "epoch": 359.6, + "learning_rate": 1.059493670886076e-06, + "loss": 0.0, + "step": 35825 + }, + { + "epoch": 359.85, + "learning_rate": 1.0531645569620254e-06, + "loss": 0.0, + "step": 35850 + }, + { + "epoch": 360.1, + "learning_rate": 1.0468354430379746e-06, + "loss": 0.0, + "step": 35875 + }, + { + "epoch": 360.35, + "learning_rate": 1.0405063291139243e-06, + "loss": 0.0, + "step": 35900 + }, + { + "epoch": 360.6, + "learning_rate": 1.0341772151898735e-06, + "loss": 0.0, + "step": 35925 + }, + { + "epoch": 360.85, + "learning_rate": 1.027848101265823e-06, + "loss": 0.0, + "step": 35950 + }, + { + "epoch": 361.1, + "learning_rate": 1.0215189873417721e-06, + "loss": 0.0, + "step": 35975 + }, + { + "epoch": 361.36, + "learning_rate": 1.0151898734177218e-06, + "loss": 0.0, + "step": 36000 + }, + { + "epoch": 361.36, + "eval_loss": 0.4141576886177063, + "eval_runtime": 2833.1664, + "eval_samples_per_second": 4.497, + "eval_steps_per_second": 0.281, + "eval_wer": 7.732140065655776, + "step": 36000 + }, + { + "epoch": 361.61, + "learning_rate": 1.008860759493671e-06, + "loss": 0.0, + "step": 36025 + }, + { + "epoch": 361.86, + "learning_rate": 1.0025316455696204e-06, + "loss": 0.0, + "step": 36050 + }, + { + "epoch": 362.11, + "learning_rate": 9.962025316455696e-07, + "loss": 0.0, + "step": 36075 + }, + { + "epoch": 362.36, + "learning_rate": 9.89873417721519e-07, + "loss": 0.0, + "step": 36100 + }, + { + "epoch": 362.61, + "learning_rate": 9.835443037974685e-07, + "loss": 0.0, + "step": 36125 + }, + { + "epoch": 362.86, + "learning_rate": 9.77215189873418e-07, + "loss": 0.0, + "step": 36150 + }, + { + "epoch": 363.11, + "learning_rate": 9.708860759493671e-07, + "loss": 0.0, + "step": 36175 + }, + { + "epoch": 363.36, + "learning_rate": 9.645569620253165e-07, + "loss": 0.0, + "step": 36200 + }, + { + "epoch": 363.61, + "learning_rate": 9.58227848101266e-07, + "loss": 0.0, + "step": 36225 + }, + { + "epoch": 363.86, + "learning_rate": 9.518987341772153e-07, + "loss": 0.0, + "step": 36250 + }, + { + "epoch": 364.12, + "learning_rate": 9.455696202531646e-07, + "loss": 0.0, + "step": 36275 + }, + { + "epoch": 364.37, + "learning_rate": 9.39240506329114e-07, + "loss": 0.0, + "step": 36300 + }, + { + "epoch": 364.62, + "learning_rate": 9.329113924050634e-07, + "loss": 0.0, + "step": 36325 + }, + { + "epoch": 364.87, + "learning_rate": 9.265822784810127e-07, + "loss": 0.0, + "step": 36350 + }, + { + "epoch": 365.12, + "learning_rate": 9.202531645569621e-07, + "loss": 0.0, + "step": 36375 + }, + { + "epoch": 365.37, + "learning_rate": 9.139240506329115e-07, + "loss": 0.0, + "step": 36400 + }, + { + "epoch": 365.62, + "learning_rate": 9.075949367088609e-07, + "loss": 0.0, + "step": 36425 + }, + { + "epoch": 365.87, + "learning_rate": 9.012658227848102e-07, + "loss": 0.0, + "step": 36450 + }, + { + "epoch": 366.12, + "learning_rate": 8.949367088607595e-07, + "loss": 0.0, + "step": 36475 + }, + { + "epoch": 366.37, + "learning_rate": 8.886075949367088e-07, + "loss": 0.0, + "step": 36500 + }, + { + "epoch": 366.62, + "learning_rate": 8.822784810126584e-07, + "loss": 0.0, + "step": 36525 + }, + { + "epoch": 366.88, + "learning_rate": 8.759493670886077e-07, + "loss": 0.0, + "step": 36550 + }, + { + "epoch": 367.13, + "learning_rate": 8.69620253164557e-07, + "loss": 0.0, + "step": 36575 + }, + { + "epoch": 367.38, + "learning_rate": 8.632911392405063e-07, + "loss": 0.0, + "step": 36600 + }, + { + "epoch": 367.63, + "learning_rate": 8.569620253164559e-07, + "loss": 0.0, + "step": 36625 + }, + { + "epoch": 367.88, + "learning_rate": 8.506329113924052e-07, + "loss": 0.0, + "step": 36650 + }, + { + "epoch": 368.13, + "learning_rate": 8.443037974683545e-07, + "loss": 0.0, + "step": 36675 + }, + { + "epoch": 368.38, + "learning_rate": 8.379746835443038e-07, + "loss": 0.0, + "step": 36700 + }, + { + "epoch": 368.63, + "learning_rate": 8.316455696202532e-07, + "loss": 0.0, + "step": 36725 + }, + { + "epoch": 368.88, + "learning_rate": 8.253164556962026e-07, + "loss": 0.0, + "step": 36750 + }, + { + "epoch": 369.13, + "learning_rate": 8.189873417721519e-07, + "loss": 0.0, + "step": 36775 + }, + { + "epoch": 369.39, + "learning_rate": 8.126582278481013e-07, + "loss": 0.0, + "step": 36800 + }, + { + "epoch": 369.64, + "learning_rate": 8.063291139240507e-07, + "loss": 0.0, + "step": 36825 + }, + { + "epoch": 369.89, + "learning_rate": 8.000000000000001e-07, + "loss": 0.0, + "step": 36850 + }, + { + "epoch": 370.14, + "learning_rate": 7.936708860759494e-07, + "loss": 0.0, + "step": 36875 + }, + { + "epoch": 370.39, + "learning_rate": 7.873417721518987e-07, + "loss": 0.0, + "step": 36900 + }, + { + "epoch": 370.64, + "learning_rate": 7.810126582278482e-07, + "loss": 0.0, + "step": 36925 + }, + { + "epoch": 370.89, + "learning_rate": 7.746835443037976e-07, + "loss": 0.0, + "step": 36950 + }, + { + "epoch": 371.14, + "learning_rate": 7.683544303797469e-07, + "loss": 0.0, + "step": 36975 + }, + { + "epoch": 371.39, + "learning_rate": 7.620253164556962e-07, + "loss": 0.0, + "step": 37000 + }, + { + "epoch": 371.39, + "eval_loss": 0.4170127809047699, + "eval_runtime": 2836.3086, + "eval_samples_per_second": 4.492, + "eval_steps_per_second": 0.281, + "eval_wer": 7.720415819915585, + "step": 37000 + }, + { + "epoch": 371.64, + "learning_rate": 7.556962025316456e-07, + "loss": 0.0, + "step": 37025 + }, + { + "epoch": 371.89, + "learning_rate": 7.493670886075951e-07, + "loss": 0.0, + "step": 37050 + }, + { + "epoch": 372.15, + "learning_rate": 7.430379746835444e-07, + "loss": 0.0, + "step": 37075 + }, + { + "epoch": 372.4, + "learning_rate": 7.367088607594937e-07, + "loss": 0.0, + "step": 37100 + }, + { + "epoch": 372.65, + "learning_rate": 7.303797468354431e-07, + "loss": 0.0, + "step": 37125 + }, + { + "epoch": 372.9, + "learning_rate": 7.240506329113925e-07, + "loss": 0.0, + "step": 37150 + }, + { + "epoch": 373.15, + "learning_rate": 7.177215189873418e-07, + "loss": 0.0, + "step": 37175 + }, + { + "epoch": 373.4, + "learning_rate": 7.113924050632912e-07, + "loss": 0.0, + "step": 37200 + }, + { + "epoch": 373.65, + "learning_rate": 7.050632911392406e-07, + "loss": 0.0, + "step": 37225 + }, + { + "epoch": 373.9, + "learning_rate": 6.9873417721519e-07, + "loss": 0.0, + "step": 37250 + }, + { + "epoch": 374.15, + "learning_rate": 6.924050632911393e-07, + "loss": 0.0, + "step": 37275 + }, + { + "epoch": 374.4, + "learning_rate": 6.860759493670886e-07, + "loss": 0.0, + "step": 37300 + }, + { + "epoch": 374.65, + "learning_rate": 6.797468354430379e-07, + "loss": 0.0, + "step": 37325 + }, + { + "epoch": 374.91, + "learning_rate": 6.734177215189875e-07, + "loss": 0.0, + "step": 37350 + }, + { + "epoch": 375.16, + "learning_rate": 6.670886075949368e-07, + "loss": 0.0, + "step": 37375 + }, + { + "epoch": 375.41, + "learning_rate": 6.607594936708861e-07, + "loss": 0.0, + "step": 37400 + }, + { + "epoch": 375.66, + "learning_rate": 6.544303797468354e-07, + "loss": 0.0, + "step": 37425 + }, + { + "epoch": 375.91, + "learning_rate": 6.483544303797469e-07, + "loss": 0.0, + "step": 37450 + }, + { + "epoch": 376.16, + "learning_rate": 6.420253164556962e-07, + "loss": 0.0, + "step": 37475 + }, + { + "epoch": 376.41, + "learning_rate": 6.356962025316456e-07, + "loss": 0.0, + "step": 37500 + }, + { + "epoch": 376.66, + "learning_rate": 6.29367088607595e-07, + "loss": 0.0, + "step": 37525 + }, + { + "epoch": 376.91, + "learning_rate": 6.230379746835444e-07, + "loss": 0.0, + "step": 37550 + }, + { + "epoch": 377.16, + "learning_rate": 6.167088607594937e-07, + "loss": 0.0, + "step": 37575 + }, + { + "epoch": 377.42, + "learning_rate": 6.103797468354431e-07, + "loss": 0.0, + "step": 37600 + }, + { + "epoch": 377.67, + "learning_rate": 6.040506329113924e-07, + "loss": 0.0, + "step": 37625 + }, + { + "epoch": 377.92, + "learning_rate": 5.977215189873419e-07, + "loss": 0.0, + "step": 37650 + }, + { + "epoch": 378.17, + "learning_rate": 5.913924050632912e-07, + "loss": 0.0, + "step": 37675 + }, + { + "epoch": 378.42, + "learning_rate": 5.850632911392406e-07, + "loss": 0.0, + "step": 37700 + }, + { + "epoch": 378.67, + "learning_rate": 5.787341772151899e-07, + "loss": 0.0, + "step": 37725 + }, + { + "epoch": 378.92, + "learning_rate": 5.724050632911393e-07, + "loss": 0.0, + "step": 37750 + }, + { + "epoch": 379.17, + "learning_rate": 5.660759493670887e-07, + "loss": 0.0, + "step": 37775 + }, + { + "epoch": 379.42, + "learning_rate": 5.59746835443038e-07, + "loss": 0.0, + "step": 37800 + }, + { + "epoch": 379.67, + "learning_rate": 5.534177215189873e-07, + "loss": 0.0, + "step": 37825 + }, + { + "epoch": 379.92, + "learning_rate": 5.470886075949368e-07, + "loss": 0.0, + "step": 37850 + }, + { + "epoch": 380.18, + "learning_rate": 5.407594936708861e-07, + "loss": 0.0, + "step": 37875 + }, + { + "epoch": 380.43, + "learning_rate": 5.344303797468354e-07, + "loss": 0.0, + "step": 37900 + }, + { + "epoch": 380.68, + "learning_rate": 5.281012658227848e-07, + "loss": 0.0, + "step": 37925 + }, + { + "epoch": 380.93, + "learning_rate": 5.217721518987342e-07, + "loss": 0.0, + "step": 37950 + }, + { + "epoch": 381.18, + "learning_rate": 5.154430379746836e-07, + "loss": 0.0, + "step": 37975 + }, + { + "epoch": 381.43, + "learning_rate": 5.091139240506329e-07, + "loss": 0.0, + "step": 38000 + }, + { + "epoch": 381.43, + "eval_loss": 0.4188634753227234, + "eval_runtime": 2829.9673, + "eval_samples_per_second": 4.503, + "eval_steps_per_second": 0.282, + "eval_wer": 7.725300922307332, + "step": 38000 + }, + { + "epoch": 381.68, + "learning_rate": 5.027848101265823e-07, + "loss": 0.0, + "step": 38025 + }, + { + "epoch": 381.93, + "learning_rate": 4.964556962025317e-07, + "loss": 0.0, + "step": 38050 + }, + { + "epoch": 382.18, + "learning_rate": 4.901265822784811e-07, + "loss": 0.0, + "step": 38075 + }, + { + "epoch": 382.43, + "learning_rate": 4.837974683544304e-07, + "loss": 0.0, + "step": 38100 + }, + { + "epoch": 382.69, + "learning_rate": 4.774683544303798e-07, + "loss": 0.0, + "step": 38125 + }, + { + "epoch": 382.94, + "learning_rate": 4.711392405063291e-07, + "loss": 0.0, + "step": 38150 + }, + { + "epoch": 383.19, + "learning_rate": 4.6481012658227853e-07, + "loss": 0.0, + "step": 38175 + }, + { + "epoch": 383.44, + "learning_rate": 4.5848101265822785e-07, + "loss": 0.0, + "step": 38200 + }, + { + "epoch": 383.69, + "learning_rate": 4.521518987341773e-07, + "loss": 0.0, + "step": 38225 + }, + { + "epoch": 383.94, + "learning_rate": 4.458227848101266e-07, + "loss": 0.0, + "step": 38250 + }, + { + "epoch": 384.19, + "learning_rate": 4.39493670886076e-07, + "loss": 0.0, + "step": 38275 + }, + { + "epoch": 384.44, + "learning_rate": 4.3316455696202535e-07, + "loss": 0.0, + "step": 38300 + }, + { + "epoch": 384.69, + "learning_rate": 4.268354430379747e-07, + "loss": 0.0, + "step": 38325 + }, + { + "epoch": 384.94, + "learning_rate": 4.2050632911392405e-07, + "loss": 0.0, + "step": 38350 + }, + { + "epoch": 385.19, + "learning_rate": 4.141772151898735e-07, + "loss": 0.0, + "step": 38375 + }, + { + "epoch": 385.45, + "learning_rate": 4.078481012658228e-07, + "loss": 0.0, + "step": 38400 + }, + { + "epoch": 385.7, + "learning_rate": 4.015189873417722e-07, + "loss": 0.0, + "step": 38425 + }, + { + "epoch": 385.95, + "learning_rate": 3.9518987341772155e-07, + "loss": 0.0, + "step": 38450 + }, + { + "epoch": 386.2, + "learning_rate": 3.888607594936709e-07, + "loss": 0.0, + "step": 38475 + }, + { + "epoch": 386.45, + "learning_rate": 3.825316455696203e-07, + "loss": 0.0, + "step": 38500 + }, + { + "epoch": 386.7, + "learning_rate": 3.7620253164556967e-07, + "loss": 0.0, + "step": 38525 + }, + { + "epoch": 386.95, + "learning_rate": 3.69873417721519e-07, + "loss": 0.0, + "step": 38550 + }, + { + "epoch": 387.2, + "learning_rate": 3.635443037974684e-07, + "loss": 0.0, + "step": 38575 + }, + { + "epoch": 387.45, + "learning_rate": 3.5721518987341774e-07, + "loss": 0.0, + "step": 38600 + }, + { + "epoch": 387.7, + "learning_rate": 3.5088607594936717e-07, + "loss": 0.0, + "step": 38625 + }, + { + "epoch": 387.95, + "learning_rate": 3.445569620253165e-07, + "loss": 0.0, + "step": 38650 + }, + { + "epoch": 388.21, + "learning_rate": 3.3822784810126587e-07, + "loss": 0.0, + "step": 38675 + }, + { + "epoch": 388.46, + "learning_rate": 3.318987341772152e-07, + "loss": 0.0, + "step": 38700 + }, + { + "epoch": 388.71, + "learning_rate": 3.255696202531646e-07, + "loss": 0.0, + "step": 38725 + }, + { + "epoch": 388.96, + "learning_rate": 3.1924050632911394e-07, + "loss": 0.0, + "step": 38750 + }, + { + "epoch": 389.21, + "learning_rate": 3.1291139240506326e-07, + "loss": 0.0, + "step": 38775 + }, + { + "epoch": 389.46, + "learning_rate": 3.065822784810127e-07, + "loss": 0.0, + "step": 38800 + }, + { + "epoch": 389.71, + "learning_rate": 3.0025316455696206e-07, + "loss": 0.0, + "step": 38825 + }, + { + "epoch": 389.96, + "learning_rate": 2.9392405063291144e-07, + "loss": 0.0, + "step": 38850 + }, + { + "epoch": 390.21, + "learning_rate": 2.875949367088608e-07, + "loss": 0.0, + "step": 38875 + }, + { + "epoch": 390.46, + "learning_rate": 2.8126582278481013e-07, + "loss": 0.0, + "step": 38900 + }, + { + "epoch": 390.72, + "learning_rate": 2.749367088607595e-07, + "loss": 0.0, + "step": 38925 + }, + { + "epoch": 390.97, + "learning_rate": 2.686075949367089e-07, + "loss": 0.0, + "step": 38950 + }, + { + "epoch": 391.22, + "learning_rate": 2.6227848101265826e-07, + "loss": 0.0, + "step": 38975 + }, + { + "epoch": 391.47, + "learning_rate": 2.5594936708860763e-07, + "loss": 0.0, + "step": 39000 + }, + { + "epoch": 391.47, + "eval_loss": 0.4201821982860565, + "eval_runtime": 2852.7738, + "eval_samples_per_second": 4.467, + "eval_steps_per_second": 0.279, + "eval_wer": 7.726277942785681, + "step": 39000 + }, + { + "epoch": 391.72, + "learning_rate": 2.49620253164557e-07, + "loss": 0.0, + "step": 39025 + }, + { + "epoch": 391.97, + "learning_rate": 2.432911392405064e-07, + "loss": 0.0, + "step": 39050 + }, + { + "epoch": 392.22, + "learning_rate": 2.3696202531645573e-07, + "loss": 0.0, + "step": 39075 + }, + { + "epoch": 392.47, + "learning_rate": 2.306329113924051e-07, + "loss": 0.0, + "step": 39100 + }, + { + "epoch": 392.72, + "learning_rate": 2.2430379746835445e-07, + "loss": 0.0, + "step": 39125 + }, + { + "epoch": 392.97, + "learning_rate": 2.1797468354430383e-07, + "loss": 0.0, + "step": 39150 + }, + { + "epoch": 393.22, + "learning_rate": 2.116455696202532e-07, + "loss": 0.0, + "step": 39175 + }, + { + "epoch": 393.48, + "learning_rate": 2.0531645569620258e-07, + "loss": 0.0, + "step": 39200 + }, + { + "epoch": 393.73, + "learning_rate": 1.989873417721519e-07, + "loss": 0.0, + "step": 39225 + }, + { + "epoch": 393.98, + "learning_rate": 1.9265822784810127e-07, + "loss": 0.0, + "step": 39250 + }, + { + "epoch": 394.23, + "learning_rate": 1.8632911392405062e-07, + "loss": 0.0, + "step": 39275 + }, + { + "epoch": 394.48, + "learning_rate": 1.8e-07, + "loss": 0.0, + "step": 39300 + }, + { + "epoch": 394.73, + "learning_rate": 1.7367088607594937e-07, + "loss": 0.0, + "step": 39325 + }, + { + "epoch": 394.98, + "learning_rate": 1.6734177215189875e-07, + "loss": 0.0, + "step": 39350 + }, + { + "epoch": 395.23, + "learning_rate": 1.610126582278481e-07, + "loss": 0.0, + "step": 39375 + }, + { + "epoch": 395.48, + "learning_rate": 1.546835443037975e-07, + "loss": 0.0, + "step": 39400 + }, + { + "epoch": 395.73, + "learning_rate": 1.4835443037974684e-07, + "loss": 0.0, + "step": 39425 + }, + { + "epoch": 395.98, + "learning_rate": 1.4202531645569622e-07, + "loss": 0.0, + "step": 39450 + }, + { + "epoch": 396.24, + "learning_rate": 1.3569620253164557e-07, + "loss": 0.0, + "step": 39475 + }, + { + "epoch": 396.49, + "learning_rate": 1.2936708860759494e-07, + "loss": 0.0, + "step": 39500 + }, + { + "epoch": 396.74, + "learning_rate": 1.2329113924050634e-07, + "loss": 0.0, + "step": 39525 + }, + { + "epoch": 396.99, + "learning_rate": 1.169620253164557e-07, + "loss": 0.0, + "step": 39550 + }, + { + "epoch": 397.24, + "learning_rate": 1.1063291139240506e-07, + "loss": 0.0, + "step": 39575 + }, + { + "epoch": 397.49, + "learning_rate": 1.0430379746835444e-07, + "loss": 0.0, + "step": 39600 + }, + { + "epoch": 397.74, + "learning_rate": 9.79746835443038e-08, + "loss": 0.0, + "step": 39625 + }, + { + "epoch": 397.99, + "learning_rate": 9.164556962025317e-08, + "loss": 0.0, + "step": 39650 + }, + { + "epoch": 398.24, + "learning_rate": 8.531645569620254e-08, + "loss": 0.0, + "step": 39675 + }, + { + "epoch": 398.49, + "learning_rate": 7.898734177215191e-08, + "loss": 0.0, + "step": 39700 + }, + { + "epoch": 398.75, + "learning_rate": 7.265822784810127e-08, + "loss": 0.0, + "step": 39725 + }, + { + "epoch": 399.0, + "learning_rate": 6.632911392405063e-08, + "loss": 0.0, + "step": 39750 + }, + { + "epoch": 399.25, + "learning_rate": 6.000000000000001e-08, + "loss": 0.0, + "step": 39775 + }, + { + "epoch": 399.5, + "learning_rate": 5.3670886075949376e-08, + "loss": 0.0, + "step": 39800 + }, + { + "epoch": 399.75, + "learning_rate": 4.7341772151898744e-08, + "loss": 0.0, + "step": 39825 + }, + { + "epoch": 400.0, + "learning_rate": 4.1012658227848106e-08, + "loss": 0.0, + "step": 39850 + }, + { + "epoch": 400.25, + "learning_rate": 3.4683544303797474e-08, + "loss": 0.0, + "step": 39875 + }, + { + "epoch": 400.5, + "learning_rate": 2.835443037974684e-08, + "loss": 0.0, + "step": 39900 + }, + { + "epoch": 400.75, + "learning_rate": 2.2025316455696203e-08, + "loss": 0.0, + "step": 39925 + }, + { + "epoch": 401.0, + "learning_rate": 1.569620253164557e-08, + "loss": 0.0, + "step": 39950 + }, + { + "epoch": 401.25, + "learning_rate": 9.367088607594938e-09, + "loss": 0.0, + "step": 39975 + }, + { + "epoch": 401.51, + "learning_rate": 3.037974683544304e-09, + "loss": 0.0, + "step": 40000 + }, + { + "epoch": 401.51, + "eval_loss": 0.42060527205467224, + "eval_runtime": 2833.1072, + "eval_samples_per_second": 4.498, + "eval_steps_per_second": 0.281, + "eval_wer": 7.720415819915585, + "step": 40000 + }, + { + "epoch": 401.51, + "step": 40000, + "total_flos": 2.172424342140513e+22, + "train_loss": 0.004165954868968765, + "train_runtime": 1001494.6755, + "train_samples_per_second": 10.225, + "train_steps_per_second": 0.04 + } + ], + "logging_steps": 25, + "max_steps": 40000, + "num_input_tokens_seen": 0, + "num_train_epochs": 405, + "save_steps": 1000, + "total_flos": 2.172424342140513e+22, + "train_batch_size": 32, + "trial_name": null, + "trial_params": null +}