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": 12.73741597623886, + "best_model_checkpoint": "./checkpoint-21000", + "epoch": 401.0025062656642, + "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": 4.800000000000001e-07, + "loss": 1.5514, + "step": 25 + }, + { + "epoch": 0.5, + "learning_rate": 9.800000000000001e-07, + "loss": 1.3099, + "step": 50 + }, + { + "epoch": 0.75, + "learning_rate": 1.48e-06, + "loss": 0.9696, + "step": 75 + }, + { + "epoch": 1.0, + "learning_rate": 1.98e-06, + "loss": 0.7451, + "step": 100 + }, + { + "epoch": 1.25, + "learning_rate": 2.4800000000000004e-06, + "loss": 0.6206, + "step": 125 + }, + { + "epoch": 1.5, + "learning_rate": 2.9800000000000003e-06, + "loss": 0.5443, + "step": 150 + }, + { + "epoch": 1.75, + "learning_rate": 3.48e-06, + "loss": 0.4869, + "step": 175 + }, + { + "epoch": 2.01, + "learning_rate": 3.980000000000001e-06, + "loss": 0.4408, + "step": 200 + }, + { + "epoch": 2.26, + "learning_rate": 4.48e-06, + "loss": 0.3833, + "step": 225 + }, + { + "epoch": 2.51, + "learning_rate": 4.980000000000001e-06, + "loss": 0.3499, + "step": 250 + }, + { + "epoch": 2.76, + "learning_rate": 5.480000000000001e-06, + "loss": 0.3299, + "step": 275 + }, + { + "epoch": 3.01, + "learning_rate": 5.98e-06, + "loss": 0.3054, + "step": 300 + }, + { + "epoch": 3.26, + "learning_rate": 6.480000000000001e-06, + "loss": 0.2569, + "step": 325 + }, + { + "epoch": 3.51, + "learning_rate": 6.98e-06, + "loss": 0.243, + "step": 350 + }, + { + "epoch": 3.76, + "learning_rate": 7.48e-06, + "loss": 0.238, + "step": 375 + }, + { + "epoch": 4.01, + "learning_rate": 7.980000000000002e-06, + "loss": 0.2238, + "step": 400 + }, + { + "epoch": 4.26, + "learning_rate": 8.48e-06, + "loss": 0.1751, + "step": 425 + }, + { + "epoch": 4.51, + "learning_rate": 8.98e-06, + "loss": 0.1738, + "step": 450 + }, + { + "epoch": 4.76, + "learning_rate": 9.48e-06, + "loss": 0.1673, + "step": 475 + }, + { + "epoch": 5.01, + "learning_rate": 9.980000000000001e-06, + "loss": 0.1643, + "step": 500 + }, + { + "epoch": 5.26, + "learning_rate": 9.993924050632912e-06, + "loss": 0.1183, + "step": 525 + }, + { + "epoch": 5.51, + "learning_rate": 9.987594936708862e-06, + "loss": 0.1144, + "step": 550 + }, + { + "epoch": 5.76, + "learning_rate": 9.981265822784811e-06, + "loss": 0.1155, + "step": 575 + }, + { + "epoch": 6.02, + "learning_rate": 9.97493670886076e-06, + "loss": 0.1141, + "step": 600 + }, + { + "epoch": 6.27, + "learning_rate": 9.96860759493671e-06, + "loss": 0.0749, + "step": 625 + }, + { + "epoch": 6.52, + "learning_rate": 9.96227848101266e-06, + "loss": 0.0745, + "step": 650 + }, + { + "epoch": 6.77, + "learning_rate": 9.955949367088609e-06, + "loss": 0.0749, + "step": 675 + }, + { + "epoch": 7.02, + "learning_rate": 9.949620253164558e-06, + "loss": 0.0729, + "step": 700 + }, + { + "epoch": 7.27, + "learning_rate": 9.943291139240508e-06, + "loss": 0.0455, + "step": 725 + }, + { + "epoch": 7.52, + "learning_rate": 9.936962025316457e-06, + "loss": 0.0438, + "step": 750 + }, + { + "epoch": 7.77, + "learning_rate": 9.930632911392407e-06, + "loss": 0.0457, + "step": 775 + }, + { + "epoch": 8.02, + "learning_rate": 9.924303797468356e-06, + "loss": 0.0441, + "step": 800 + }, + { + "epoch": 8.27, + "learning_rate": 9.917974683544305e-06, + "loss": 0.0281, + "step": 825 + }, + { + "epoch": 8.52, + "learning_rate": 9.911645569620255e-06, + "loss": 0.0266, + "step": 850 + }, + { + "epoch": 8.77, + "learning_rate": 9.905316455696203e-06, + "loss": 0.027, + "step": 875 + }, + { + "epoch": 9.02, + "learning_rate": 9.898987341772152e-06, + "loss": 0.025, + "step": 900 + }, + { + "epoch": 9.27, + "learning_rate": 9.892658227848101e-06, + "loss": 0.0168, + "step": 925 + }, + { + "epoch": 9.52, + "learning_rate": 9.88632911392405e-06, + "loss": 0.016, + "step": 950 + }, + { + "epoch": 9.77, + "learning_rate": 9.88e-06, + "loss": 0.017, + "step": 975 + }, + { + "epoch": 10.03, + "learning_rate": 9.87367088607595e-06, + "loss": 0.0153, + "step": 1000 + }, + { + "epoch": 10.03, + "eval_loss": 0.26904571056365967, + "eval_runtime": 507.7865, + "eval_samples_per_second": 25.093, + "eval_steps_per_second": 0.786, + "eval_wer": 15.311864936689073, + "step": 1000 + }, + { + "epoch": 10.28, + "learning_rate": 9.867341772151899e-06, + "loss": 0.0107, + "step": 1025 + }, + { + "epoch": 10.53, + "learning_rate": 9.861012658227849e-06, + "loss": 0.0105, + "step": 1050 + }, + { + "epoch": 10.78, + "learning_rate": 9.854683544303798e-06, + "loss": 0.0104, + "step": 1075 + }, + { + "epoch": 11.03, + "learning_rate": 9.848354430379747e-06, + "loss": 0.0102, + "step": 1100 + }, + { + "epoch": 11.28, + "learning_rate": 9.842025316455697e-06, + "loss": 0.0078, + "step": 1125 + }, + { + "epoch": 11.53, + "learning_rate": 9.835696202531646e-06, + "loss": 0.0077, + "step": 1150 + }, + { + "epoch": 11.78, + "learning_rate": 9.829367088607596e-06, + "loss": 0.0083, + "step": 1175 + }, + { + "epoch": 12.03, + "learning_rate": 9.823037974683545e-06, + "loss": 0.0078, + "step": 1200 + }, + { + "epoch": 12.28, + "learning_rate": 9.816708860759495e-06, + "loss": 0.006, + "step": 1225 + }, + { + "epoch": 12.53, + "learning_rate": 9.810379746835444e-06, + "loss": 0.0064, + "step": 1250 + }, + { + "epoch": 12.78, + "learning_rate": 9.804050632911393e-06, + "loss": 0.0065, + "step": 1275 + }, + { + "epoch": 13.03, + "learning_rate": 9.797721518987343e-06, + "loss": 0.0064, + "step": 1300 + }, + { + "epoch": 13.28, + "learning_rate": 9.791392405063292e-06, + "loss": 0.0048, + "step": 1325 + }, + { + "epoch": 13.53, + "learning_rate": 9.785063291139242e-06, + "loss": 0.0048, + "step": 1350 + }, + { + "epoch": 13.78, + "learning_rate": 9.778734177215191e-06, + "loss": 0.0044, + "step": 1375 + }, + { + "epoch": 14.04, + "learning_rate": 9.77240506329114e-06, + "loss": 0.0043, + "step": 1400 + }, + { + "epoch": 14.29, + "learning_rate": 9.76607594936709e-06, + "loss": 0.0037, + "step": 1425 + }, + { + "epoch": 14.54, + "learning_rate": 9.75974683544304e-06, + "loss": 0.0037, + "step": 1450 + }, + { + "epoch": 14.79, + "learning_rate": 9.753417721518989e-06, + "loss": 0.0037, + "step": 1475 + }, + { + "epoch": 15.04, + "learning_rate": 9.747088607594938e-06, + "loss": 0.0039, + "step": 1500 + }, + { + "epoch": 15.29, + "learning_rate": 9.740759493670888e-06, + "loss": 0.0036, + "step": 1525 + }, + { + "epoch": 15.54, + "learning_rate": 9.734430379746837e-06, + "loss": 0.0039, + "step": 1550 + }, + { + "epoch": 15.79, + "learning_rate": 9.728101265822787e-06, + "loss": 0.0034, + "step": 1575 + }, + { + "epoch": 16.04, + "learning_rate": 9.721772151898736e-06, + "loss": 0.0034, + "step": 1600 + }, + { + "epoch": 16.29, + "learning_rate": 9.715443037974684e-06, + "loss": 0.0034, + "step": 1625 + }, + { + "epoch": 16.54, + "learning_rate": 9.709113924050633e-06, + "loss": 0.0034, + "step": 1650 + }, + { + "epoch": 16.79, + "learning_rate": 9.702784810126583e-06, + "loss": 0.0038, + "step": 1675 + }, + { + "epoch": 17.04, + "learning_rate": 9.696455696202532e-06, + "loss": 0.0047, + "step": 1700 + }, + { + "epoch": 17.29, + "learning_rate": 9.690126582278481e-06, + "loss": 0.0042, + "step": 1725 + }, + { + "epoch": 17.54, + "learning_rate": 9.68379746835443e-06, + "loss": 0.0042, + "step": 1750 + }, + { + "epoch": 17.79, + "learning_rate": 9.67746835443038e-06, + "loss": 0.0042, + "step": 1775 + }, + { + "epoch": 18.05, + "learning_rate": 9.67113924050633e-06, + "loss": 0.0043, + "step": 1800 + }, + { + "epoch": 18.3, + "learning_rate": 9.664810126582279e-06, + "loss": 0.0031, + "step": 1825 + }, + { + "epoch": 18.55, + "learning_rate": 9.658481012658229e-06, + "loss": 0.0032, + "step": 1850 + }, + { + "epoch": 18.8, + "learning_rate": 9.652151898734178e-06, + "loss": 0.0036, + "step": 1875 + }, + { + "epoch": 19.05, + "learning_rate": 9.645822784810127e-06, + "loss": 0.0036, + "step": 1900 + }, + { + "epoch": 19.3, + "learning_rate": 9.639493670886077e-06, + "loss": 0.003, + "step": 1925 + }, + { + "epoch": 19.55, + "learning_rate": 9.633164556962026e-06, + "loss": 0.0027, + "step": 1950 + }, + { + "epoch": 19.8, + "learning_rate": 9.626835443037976e-06, + "loss": 0.0026, + "step": 1975 + }, + { + "epoch": 20.05, + "learning_rate": 9.620506329113925e-06, + "loss": 0.0029, + "step": 2000 + }, + { + "epoch": 20.05, + "eval_loss": 0.3131965398788452, + "eval_runtime": 495.4767, + "eval_samples_per_second": 25.717, + "eval_steps_per_second": 0.805, + "eval_wer": 15.033414100359543, + "step": 2000 + }, + { + "epoch": 20.3, + "learning_rate": 9.614177215189875e-06, + "loss": 0.0027, + "step": 2025 + }, + { + "epoch": 20.55, + "learning_rate": 9.607848101265824e-06, + "loss": 0.0024, + "step": 2050 + }, + { + "epoch": 20.8, + "learning_rate": 9.601518987341772e-06, + "loss": 0.0026, + "step": 2075 + }, + { + "epoch": 21.05, + "learning_rate": 9.595189873417721e-06, + "loss": 0.0026, + "step": 2100 + }, + { + "epoch": 21.3, + "learning_rate": 9.58886075949367e-06, + "loss": 0.0027, + "step": 2125 + }, + { + "epoch": 21.55, + "learning_rate": 9.582531645569622e-06, + "loss": 0.0029, + "step": 2150 + }, + { + "epoch": 21.8, + "learning_rate": 9.576202531645571e-06, + "loss": 0.0029, + "step": 2175 + }, + { + "epoch": 22.06, + "learning_rate": 9.56987341772152e-06, + "loss": 0.003, + "step": 2200 + }, + { + "epoch": 22.31, + "learning_rate": 9.56354430379747e-06, + "loss": 0.0025, + "step": 2225 + }, + { + "epoch": 22.56, + "learning_rate": 9.55721518987342e-06, + "loss": 0.0025, + "step": 2250 + }, + { + "epoch": 22.81, + "learning_rate": 9.550886075949369e-06, + "loss": 0.0029, + "step": 2275 + }, + { + "epoch": 23.06, + "learning_rate": 9.544556962025318e-06, + "loss": 0.0028, + "step": 2300 + }, + { + "epoch": 23.31, + "learning_rate": 9.538227848101268e-06, + "loss": 0.0026, + "step": 2325 + }, + { + "epoch": 23.56, + "learning_rate": 9.531898734177215e-06, + "loss": 0.0025, + "step": 2350 + }, + { + "epoch": 23.81, + "learning_rate": 9.525569620253165e-06, + "loss": 0.0027, + "step": 2375 + }, + { + "epoch": 24.06, + "learning_rate": 9.519240506329114e-06, + "loss": 0.0027, + "step": 2400 + }, + { + "epoch": 24.31, + "learning_rate": 9.512911392405064e-06, + "loss": 0.0024, + "step": 2425 + }, + { + "epoch": 24.56, + "learning_rate": 9.506582278481013e-06, + "loss": 0.0021, + "step": 2450 + }, + { + "epoch": 24.81, + "learning_rate": 9.500253164556963e-06, + "loss": 0.002, + "step": 2475 + }, + { + "epoch": 25.06, + "learning_rate": 9.493924050632912e-06, + "loss": 0.002, + "step": 2500 + }, + { + "epoch": 25.31, + "learning_rate": 9.487594936708861e-06, + "loss": 0.0018, + "step": 2525 + }, + { + "epoch": 25.56, + "learning_rate": 9.48126582278481e-06, + "loss": 0.0024, + "step": 2550 + }, + { + "epoch": 25.81, + "learning_rate": 9.47493670886076e-06, + "loss": 0.0023, + "step": 2575 + }, + { + "epoch": 26.07, + "learning_rate": 9.46860759493671e-06, + "loss": 0.0019, + "step": 2600 + }, + { + "epoch": 26.32, + "learning_rate": 9.462278481012659e-06, + "loss": 0.0018, + "step": 2625 + }, + { + "epoch": 26.57, + "learning_rate": 9.455949367088609e-06, + "loss": 0.0018, + "step": 2650 + }, + { + "epoch": 26.82, + "learning_rate": 9.449620253164558e-06, + "loss": 0.0022, + "step": 2675 + }, + { + "epoch": 27.07, + "learning_rate": 9.443291139240507e-06, + "loss": 0.0019, + "step": 2700 + }, + { + "epoch": 27.32, + "learning_rate": 9.436962025316457e-06, + "loss": 0.0019, + "step": 2725 + }, + { + "epoch": 27.57, + "learning_rate": 9.430632911392406e-06, + "loss": 0.0023, + "step": 2750 + }, + { + "epoch": 27.82, + "learning_rate": 9.424303797468356e-06, + "loss": 0.0027, + "step": 2775 + }, + { + "epoch": 28.07, + "learning_rate": 9.417974683544303e-06, + "loss": 0.0022, + "step": 2800 + }, + { + "epoch": 28.32, + "learning_rate": 9.411645569620253e-06, + "loss": 0.0018, + "step": 2825 + }, + { + "epoch": 28.57, + "learning_rate": 9.405316455696202e-06, + "loss": 0.0022, + "step": 2850 + }, + { + "epoch": 28.82, + "learning_rate": 9.398987341772152e-06, + "loss": 0.0024, + "step": 2875 + }, + { + "epoch": 29.07, + "learning_rate": 9.392658227848101e-06, + "loss": 0.0022, + "step": 2900 + }, + { + "epoch": 29.32, + "learning_rate": 9.38632911392405e-06, + "loss": 0.002, + "step": 2925 + }, + { + "epoch": 29.57, + "learning_rate": 9.38e-06, + "loss": 0.0022, + "step": 2950 + }, + { + "epoch": 29.82, + "learning_rate": 9.373670886075951e-06, + "loss": 0.0019, + "step": 2975 + }, + { + "epoch": 30.08, + "learning_rate": 9.3673417721519e-06, + "loss": 0.0018, + "step": 3000 + }, + { + "epoch": 30.08, + "eval_loss": 0.3312050700187683, + "eval_runtime": 495.1693, + "eval_samples_per_second": 25.733, + "eval_steps_per_second": 0.806, + "eval_wer": 14.611341253712679, + "step": 3000 + }, + { + "epoch": 30.33, + "learning_rate": 9.36101265822785e-06, + "loss": 0.0017, + "step": 3025 + }, + { + "epoch": 30.58, + "learning_rate": 9.3546835443038e-06, + "loss": 0.0015, + "step": 3050 + }, + { + "epoch": 30.83, + "learning_rate": 9.348354430379749e-06, + "loss": 0.0017, + "step": 3075 + }, + { + "epoch": 31.08, + "learning_rate": 9.342025316455697e-06, + "loss": 0.0016, + "step": 3100 + }, + { + "epoch": 31.33, + "learning_rate": 9.335696202531646e-06, + "loss": 0.0016, + "step": 3125 + }, + { + "epoch": 31.58, + "learning_rate": 9.329367088607595e-06, + "loss": 0.002, + "step": 3150 + }, + { + "epoch": 31.83, + "learning_rate": 9.323037974683545e-06, + "loss": 0.0019, + "step": 3175 + }, + { + "epoch": 32.08, + "learning_rate": 9.316708860759494e-06, + "loss": 0.0013, + "step": 3200 + }, + { + "epoch": 32.33, + "learning_rate": 9.310379746835444e-06, + "loss": 0.0015, + "step": 3225 + }, + { + "epoch": 32.58, + "learning_rate": 9.304050632911393e-06, + "loss": 0.0013, + "step": 3250 + }, + { + "epoch": 32.83, + "learning_rate": 9.297721518987343e-06, + "loss": 0.0014, + "step": 3275 + }, + { + "epoch": 33.08, + "learning_rate": 9.291392405063292e-06, + "loss": 0.0014, + "step": 3300 + }, + { + "epoch": 33.33, + "learning_rate": 9.285063291139241e-06, + "loss": 0.0012, + "step": 3325 + }, + { + "epoch": 33.58, + "learning_rate": 9.27873417721519e-06, + "loss": 0.001, + "step": 3350 + }, + { + "epoch": 33.83, + "learning_rate": 9.27240506329114e-06, + "loss": 0.0011, + "step": 3375 + }, + { + "epoch": 34.09, + "learning_rate": 9.26607594936709e-06, + "loss": 0.0012, + "step": 3400 + }, + { + "epoch": 34.34, + "learning_rate": 9.259746835443039e-06, + "loss": 0.0015, + "step": 3425 + }, + { + "epoch": 34.59, + "learning_rate": 9.253417721518989e-06, + "loss": 0.0017, + "step": 3450 + }, + { + "epoch": 34.84, + "learning_rate": 9.247088607594938e-06, + "loss": 0.0015, + "step": 3475 + }, + { + "epoch": 35.09, + "learning_rate": 9.240759493670887e-06, + "loss": 0.0017, + "step": 3500 + }, + { + "epoch": 35.34, + "learning_rate": 9.234430379746837e-06, + "loss": 0.0034, + "step": 3525 + }, + { + "epoch": 35.59, + "learning_rate": 9.228101265822785e-06, + "loss": 0.0025, + "step": 3550 + }, + { + "epoch": 35.84, + "learning_rate": 9.221772151898734e-06, + "loss": 0.0023, + "step": 3575 + }, + { + "epoch": 36.09, + "learning_rate": 9.215443037974683e-06, + "loss": 0.0024, + "step": 3600 + }, + { + "epoch": 36.34, + "learning_rate": 9.209113924050633e-06, + "loss": 0.002, + "step": 3625 + }, + { + "epoch": 36.59, + "learning_rate": 9.202784810126582e-06, + "loss": 0.0025, + "step": 3650 + }, + { + "epoch": 36.84, + "learning_rate": 9.196455696202532e-06, + "loss": 0.0035, + "step": 3675 + }, + { + "epoch": 37.09, + "learning_rate": 9.190126582278481e-06, + "loss": 0.0035, + "step": 3700 + }, + { + "epoch": 37.34, + "learning_rate": 9.18379746835443e-06, + "loss": 0.0027, + "step": 3725 + }, + { + "epoch": 37.59, + "learning_rate": 9.17746835443038e-06, + "loss": 0.0026, + "step": 3750 + }, + { + "epoch": 37.84, + "learning_rate": 9.17113924050633e-06, + "loss": 0.0021, + "step": 3775 + }, + { + "epoch": 38.1, + "learning_rate": 9.16481012658228e-06, + "loss": 0.0023, + "step": 3800 + }, + { + "epoch": 38.35, + "learning_rate": 9.158481012658228e-06, + "loss": 0.0019, + "step": 3825 + }, + { + "epoch": 38.6, + "learning_rate": 9.152151898734178e-06, + "loss": 0.0018, + "step": 3850 + }, + { + "epoch": 38.85, + "learning_rate": 9.145822784810127e-06, + "loss": 0.0017, + "step": 3875 + }, + { + "epoch": 39.1, + "learning_rate": 9.139493670886077e-06, + "loss": 0.0014, + "step": 3900 + }, + { + "epoch": 39.35, + "learning_rate": 9.133164556962026e-06, + "loss": 0.0013, + "step": 3925 + }, + { + "epoch": 39.6, + "learning_rate": 9.126835443037975e-06, + "loss": 0.0012, + "step": 3950 + }, + { + "epoch": 39.85, + "learning_rate": 9.120506329113925e-06, + "loss": 0.0012, + "step": 3975 + }, + { + "epoch": 40.1, + "learning_rate": 9.114177215189874e-06, + "loss": 0.0009, + "step": 4000 + }, + { + "epoch": 40.1, + "eval_loss": 0.3375228941440582, + "eval_runtime": 495.2266, + "eval_samples_per_second": 25.73, + "eval_steps_per_second": 0.806, + "eval_wer": 14.091566359230889, + "step": 4000 + }, + { + "epoch": 40.35, + "learning_rate": 9.107848101265824e-06, + "loss": 0.0009, + "step": 4025 + }, + { + "epoch": 40.6, + "learning_rate": 9.101518987341773e-06, + "loss": 0.0007, + "step": 4050 + }, + { + "epoch": 40.85, + "learning_rate": 9.095189873417723e-06, + "loss": 0.0008, + "step": 4075 + }, + { + "epoch": 41.1, + "learning_rate": 9.088860759493672e-06, + "loss": 0.001, + "step": 4100 + }, + { + "epoch": 41.35, + "learning_rate": 9.082531645569621e-06, + "loss": 0.0007, + "step": 4125 + }, + { + "epoch": 41.6, + "learning_rate": 9.07620253164557e-06, + "loss": 0.0007, + "step": 4150 + }, + { + "epoch": 41.85, + "learning_rate": 9.06987341772152e-06, + "loss": 0.0007, + "step": 4175 + }, + { + "epoch": 42.11, + "learning_rate": 9.06354430379747e-06, + "loss": 0.0007, + "step": 4200 + }, + { + "epoch": 42.36, + "learning_rate": 9.057215189873419e-06, + "loss": 0.0005, + "step": 4225 + }, + { + "epoch": 42.61, + "learning_rate": 9.050886075949369e-06, + "loss": 0.0005, + "step": 4250 + }, + { + "epoch": 42.86, + "learning_rate": 9.044556962025316e-06, + "loss": 0.0009, + "step": 4275 + }, + { + "epoch": 43.11, + "learning_rate": 9.038227848101266e-06, + "loss": 0.0009, + "step": 4300 + }, + { + "epoch": 43.36, + "learning_rate": 9.031898734177215e-06, + "loss": 0.0008, + "step": 4325 + }, + { + "epoch": 43.61, + "learning_rate": 9.025569620253165e-06, + "loss": 0.001, + "step": 4350 + }, + { + "epoch": 43.86, + "learning_rate": 9.019240506329114e-06, + "loss": 0.0009, + "step": 4375 + }, + { + "epoch": 44.11, + "learning_rate": 9.012911392405063e-06, + "loss": 0.0009, + "step": 4400 + }, + { + "epoch": 44.36, + "learning_rate": 9.006582278481013e-06, + "loss": 0.0007, + "step": 4425 + }, + { + "epoch": 44.61, + "learning_rate": 9.000253164556962e-06, + "loss": 0.0007, + "step": 4450 + }, + { + "epoch": 44.86, + "learning_rate": 8.993924050632912e-06, + "loss": 0.0006, + "step": 4475 + }, + { + "epoch": 45.11, + "learning_rate": 8.987594936708861e-06, + "loss": 0.0006, + "step": 4500 + }, + { + "epoch": 45.36, + "learning_rate": 8.98126582278481e-06, + "loss": 0.0004, + "step": 4525 + }, + { + "epoch": 45.61, + "learning_rate": 8.97493670886076e-06, + "loss": 0.0004, + "step": 4550 + }, + { + "epoch": 45.86, + "learning_rate": 8.96860759493671e-06, + "loss": 0.0005, + "step": 4575 + }, + { + "epoch": 46.12, + "learning_rate": 8.962278481012659e-06, + "loss": 0.0006, + "step": 4600 + }, + { + "epoch": 46.37, + "learning_rate": 8.955949367088608e-06, + "loss": 0.0004, + "step": 4625 + }, + { + "epoch": 46.62, + "learning_rate": 8.949620253164558e-06, + "loss": 0.0004, + "step": 4650 + }, + { + "epoch": 46.87, + "learning_rate": 8.943291139240507e-06, + "loss": 0.0004, + "step": 4675 + }, + { + "epoch": 47.12, + "learning_rate": 8.936962025316457e-06, + "loss": 0.0004, + "step": 4700 + }, + { + "epoch": 47.37, + "learning_rate": 8.930632911392406e-06, + "loss": 0.0004, + "step": 4725 + }, + { + "epoch": 47.62, + "learning_rate": 8.924303797468355e-06, + "loss": 0.0003, + "step": 4750 + }, + { + "epoch": 47.87, + "learning_rate": 8.917974683544305e-06, + "loss": 0.0004, + "step": 4775 + }, + { + "epoch": 48.12, + "learning_rate": 8.911645569620254e-06, + "loss": 0.0003, + "step": 4800 + }, + { + "epoch": 48.37, + "learning_rate": 8.905316455696204e-06, + "loss": 0.0005, + "step": 4825 + }, + { + "epoch": 48.62, + "learning_rate": 8.898987341772153e-06, + "loss": 0.0008, + "step": 4850 + }, + { + "epoch": 48.87, + "learning_rate": 8.892658227848103e-06, + "loss": 0.0007, + "step": 4875 + }, + { + "epoch": 49.12, + "learning_rate": 8.886329113924052e-06, + "loss": 0.0008, + "step": 4900 + }, + { + "epoch": 49.37, + "learning_rate": 8.880000000000001e-06, + "loss": 0.0009, + "step": 4925 + }, + { + "epoch": 49.62, + "learning_rate": 8.87367088607595e-06, + "loss": 0.0016, + "step": 4950 + }, + { + "epoch": 49.87, + "learning_rate": 8.8673417721519e-06, + "loss": 0.0031, + "step": 4975 + }, + { + "epoch": 50.13, + "learning_rate": 8.86101265822785e-06, + "loss": 0.0037, + "step": 5000 + }, + { + "epoch": 50.13, + "eval_loss": 0.33063799142837524, + "eval_runtime": 495.1244, + "eval_samples_per_second": 25.735, + "eval_steps_per_second": 0.806, + "eval_wer": 14.324097233078007, + "step": 5000 + }, + { + "epoch": 50.38, + "learning_rate": 8.854683544303797e-06, + "loss": 0.0037, + "step": 5025 + }, + { + "epoch": 50.63, + "learning_rate": 8.848354430379747e-06, + "loss": 0.0034, + "step": 5050 + }, + { + "epoch": 50.88, + "learning_rate": 8.842025316455696e-06, + "loss": 0.0039, + "step": 5075 + }, + { + "epoch": 51.13, + "learning_rate": 8.835696202531646e-06, + "loss": 0.004, + "step": 5100 + }, + { + "epoch": 51.38, + "learning_rate": 8.829367088607595e-06, + "loss": 0.0035, + "step": 5125 + }, + { + "epoch": 51.63, + "learning_rate": 8.823037974683545e-06, + "loss": 0.0028, + "step": 5150 + }, + { + "epoch": 51.88, + "learning_rate": 8.816708860759494e-06, + "loss": 0.0029, + "step": 5175 + }, + { + "epoch": 52.13, + "learning_rate": 8.810379746835443e-06, + "loss": 0.0025, + "step": 5200 + }, + { + "epoch": 52.38, + "learning_rate": 8.804050632911393e-06, + "loss": 0.0018, + "step": 5225 + }, + { + "epoch": 52.63, + "learning_rate": 8.797721518987342e-06, + "loss": 0.0016, + "step": 5250 + }, + { + "epoch": 52.88, + "learning_rate": 8.791392405063292e-06, + "loss": 0.0018, + "step": 5275 + }, + { + "epoch": 53.13, + "learning_rate": 8.785063291139241e-06, + "loss": 0.0019, + "step": 5300 + }, + { + "epoch": 53.38, + "learning_rate": 8.77873417721519e-06, + "loss": 0.001, + "step": 5325 + }, + { + "epoch": 53.63, + "learning_rate": 8.77240506329114e-06, + "loss": 0.0014, + "step": 5350 + }, + { + "epoch": 53.88, + "learning_rate": 8.76607594936709e-06, + "loss": 0.001, + "step": 5375 + }, + { + "epoch": 54.14, + "learning_rate": 8.759746835443039e-06, + "loss": 0.0009, + "step": 5400 + }, + { + "epoch": 54.39, + "learning_rate": 8.753417721518988e-06, + "loss": 0.0006, + "step": 5425 + }, + { + "epoch": 54.64, + "learning_rate": 8.747088607594938e-06, + "loss": 0.0008, + "step": 5450 + }, + { + "epoch": 54.89, + "learning_rate": 8.740759493670887e-06, + "loss": 0.0005, + "step": 5475 + }, + { + "epoch": 55.14, + "learning_rate": 8.734430379746837e-06, + "loss": 0.0006, + "step": 5500 + }, + { + "epoch": 55.39, + "learning_rate": 8.728101265822786e-06, + "loss": 0.0004, + "step": 5525 + }, + { + "epoch": 55.64, + "learning_rate": 8.721772151898735e-06, + "loss": 0.0005, + "step": 5550 + }, + { + "epoch": 55.89, + "learning_rate": 8.715443037974685e-06, + "loss": 0.0004, + "step": 5575 + }, + { + "epoch": 56.14, + "learning_rate": 8.709113924050634e-06, + "loss": 0.0003, + "step": 5600 + }, + { + "epoch": 56.39, + "learning_rate": 8.702784810126584e-06, + "loss": 0.0003, + "step": 5625 + }, + { + "epoch": 56.64, + "learning_rate": 8.696455696202533e-06, + "loss": 0.0003, + "step": 5650 + }, + { + "epoch": 56.89, + "learning_rate": 8.690126582278483e-06, + "loss": 0.0003, + "step": 5675 + }, + { + "epoch": 57.14, + "learning_rate": 8.683797468354432e-06, + "loss": 0.0003, + "step": 5700 + }, + { + "epoch": 57.39, + "learning_rate": 8.677468354430381e-06, + "loss": 0.0002, + "step": 5725 + }, + { + "epoch": 57.64, + "learning_rate": 8.671139240506329e-06, + "loss": 0.0002, + "step": 5750 + }, + { + "epoch": 57.89, + "learning_rate": 8.664810126582279e-06, + "loss": 0.0002, + "step": 5775 + }, + { + "epoch": 58.15, + "learning_rate": 8.658481012658228e-06, + "loss": 0.0002, + "step": 5800 + }, + { + "epoch": 58.4, + "learning_rate": 8.652151898734177e-06, + "loss": 0.0002, + "step": 5825 + }, + { + "epoch": 58.65, + "learning_rate": 8.645822784810127e-06, + "loss": 0.0002, + "step": 5850 + }, + { + "epoch": 58.9, + "learning_rate": 8.639493670886076e-06, + "loss": 0.0002, + "step": 5875 + }, + { + "epoch": 59.15, + "learning_rate": 8.633164556962026e-06, + "loss": 0.0002, + "step": 5900 + }, + { + "epoch": 59.4, + "learning_rate": 8.626835443037975e-06, + "loss": 0.0002, + "step": 5925 + }, + { + "epoch": 59.65, + "learning_rate": 8.620506329113925e-06, + "loss": 0.0002, + "step": 5950 + }, + { + "epoch": 59.9, + "learning_rate": 8.614177215189874e-06, + "loss": 0.0002, + "step": 5975 + }, + { + "epoch": 60.15, + "learning_rate": 8.607848101265823e-06, + "loss": 0.0002, + "step": 6000 + }, + { + "epoch": 60.15, + "eval_loss": 0.3628025949001312, + "eval_runtime": 495.3014, + "eval_samples_per_second": 25.726, + "eval_steps_per_second": 0.806, + "eval_wer": 13.54638893231202, + "step": 6000 + }, + { + "epoch": 60.4, + "learning_rate": 8.601518987341773e-06, + "loss": 0.0002, + "step": 6025 + }, + { + "epoch": 60.65, + "learning_rate": 8.595189873417722e-06, + "loss": 0.0002, + "step": 6050 + }, + { + "epoch": 60.9, + "learning_rate": 8.588860759493672e-06, + "loss": 0.0002, + "step": 6075 + }, + { + "epoch": 61.15, + "learning_rate": 8.582531645569621e-06, + "loss": 0.0002, + "step": 6100 + }, + { + "epoch": 61.4, + "learning_rate": 8.57620253164557e-06, + "loss": 0.0002, + "step": 6125 + }, + { + "epoch": 61.65, + "learning_rate": 8.56987341772152e-06, + "loss": 0.0002, + "step": 6150 + }, + { + "epoch": 61.9, + "learning_rate": 8.56354430379747e-06, + "loss": 0.0002, + "step": 6175 + }, + { + "epoch": 62.16, + "learning_rate": 8.557215189873417e-06, + "loss": 0.0002, + "step": 6200 + }, + { + "epoch": 62.41, + "learning_rate": 8.550886075949367e-06, + "loss": 0.0002, + "step": 6225 + }, + { + "epoch": 62.66, + "learning_rate": 8.544556962025318e-06, + "loss": 0.0002, + "step": 6250 + }, + { + "epoch": 62.91, + "learning_rate": 8.538227848101267e-06, + "loss": 0.0002, + "step": 6275 + }, + { + "epoch": 63.16, + "learning_rate": 8.531898734177217e-06, + "loss": 0.0002, + "step": 6300 + }, + { + "epoch": 63.41, + "learning_rate": 8.525569620253166e-06, + "loss": 0.0001, + "step": 6325 + }, + { + "epoch": 63.66, + "learning_rate": 8.519240506329115e-06, + "loss": 0.0001, + "step": 6350 + }, + { + "epoch": 63.91, + "learning_rate": 8.512911392405065e-06, + "loss": 0.0001, + "step": 6375 + }, + { + "epoch": 64.16, + "learning_rate": 8.506582278481014e-06, + "loss": 0.0002, + "step": 6400 + }, + { + "epoch": 64.41, + "learning_rate": 8.500253164556964e-06, + "loss": 0.0001, + "step": 6425 + }, + { + "epoch": 64.66, + "learning_rate": 8.493924050632913e-06, + "loss": 0.0001, + "step": 6450 + }, + { + "epoch": 64.91, + "learning_rate": 8.487594936708863e-06, + "loss": 0.0002, + "step": 6475 + }, + { + "epoch": 65.16, + "learning_rate": 8.48126582278481e-06, + "loss": 0.0001, + "step": 6500 + }, + { + "epoch": 65.41, + "learning_rate": 8.47493670886076e-06, + "loss": 0.0001, + "step": 6525 + }, + { + "epoch": 65.66, + "learning_rate": 8.468607594936709e-06, + "loss": 0.0001, + "step": 6550 + }, + { + "epoch": 65.91, + "learning_rate": 8.462278481012659e-06, + "loss": 0.0001, + "step": 6575 + }, + { + "epoch": 66.17, + "learning_rate": 8.455949367088608e-06, + "loss": 0.0001, + "step": 6600 + }, + { + "epoch": 66.42, + "learning_rate": 8.449620253164557e-06, + "loss": 0.0001, + "step": 6625 + }, + { + "epoch": 66.67, + "learning_rate": 8.443291139240507e-06, + "loss": 0.0001, + "step": 6650 + }, + { + "epoch": 66.92, + "learning_rate": 8.436962025316456e-06, + "loss": 0.0001, + "step": 6675 + }, + { + "epoch": 67.17, + "learning_rate": 8.430632911392406e-06, + "loss": 0.0001, + "step": 6700 + }, + { + "epoch": 67.42, + "learning_rate": 8.424303797468355e-06, + "loss": 0.0001, + "step": 6725 + }, + { + "epoch": 67.67, + "learning_rate": 8.417974683544305e-06, + "loss": 0.0001, + "step": 6750 + }, + { + "epoch": 67.92, + "learning_rate": 8.411645569620254e-06, + "loss": 0.0001, + "step": 6775 + }, + { + "epoch": 68.17, + "learning_rate": 8.405316455696203e-06, + "loss": 0.0001, + "step": 6800 + }, + { + "epoch": 68.42, + "learning_rate": 8.398987341772153e-06, + "loss": 0.0001, + "step": 6825 + }, + { + "epoch": 68.67, + "learning_rate": 8.392658227848102e-06, + "loss": 0.0001, + "step": 6850 + }, + { + "epoch": 68.92, + "learning_rate": 8.386329113924052e-06, + "loss": 0.0001, + "step": 6875 + }, + { + "epoch": 69.17, + "learning_rate": 8.380000000000001e-06, + "loss": 0.0001, + "step": 6900 + }, + { + "epoch": 69.42, + "learning_rate": 8.37367088607595e-06, + "loss": 0.0001, + "step": 6925 + }, + { + "epoch": 69.67, + "learning_rate": 8.367341772151898e-06, + "loss": 0.0001, + "step": 6950 + }, + { + "epoch": 69.92, + "learning_rate": 8.361012658227848e-06, + "loss": 0.0001, + "step": 6975 + }, + { + "epoch": 70.18, + "learning_rate": 8.354683544303797e-06, + "loss": 0.0001, + "step": 7000 + }, + { + "epoch": 70.18, + "eval_loss": 0.3804377615451813, + "eval_runtime": 495.1227, + "eval_samples_per_second": 25.735, + "eval_steps_per_second": 0.806, + "eval_wer": 13.49851492887291, + "step": 7000 + }, + { + "epoch": 70.43, + "learning_rate": 8.348354430379747e-06, + "loss": 0.0001, + "step": 7025 + }, + { + "epoch": 70.68, + "learning_rate": 8.342025316455696e-06, + "loss": 0.0001, + "step": 7050 + }, + { + "epoch": 70.93, + "learning_rate": 8.335696202531647e-06, + "loss": 0.0001, + "step": 7075 + }, + { + "epoch": 71.18, + "learning_rate": 8.329367088607597e-06, + "loss": 0.0001, + "step": 7100 + }, + { + "epoch": 71.43, + "learning_rate": 8.323037974683546e-06, + "loss": 0.0001, + "step": 7125 + }, + { + "epoch": 71.68, + "learning_rate": 8.316708860759495e-06, + "loss": 0.0001, + "step": 7150 + }, + { + "epoch": 71.93, + "learning_rate": 8.310379746835445e-06, + "loss": 0.0001, + "step": 7175 + }, + { + "epoch": 72.18, + "learning_rate": 8.304050632911394e-06, + "loss": 0.0001, + "step": 7200 + }, + { + "epoch": 72.43, + "learning_rate": 8.297721518987342e-06, + "loss": 0.0001, + "step": 7225 + }, + { + "epoch": 72.68, + "learning_rate": 8.291392405063291e-06, + "loss": 0.0001, + "step": 7250 + }, + { + "epoch": 72.93, + "learning_rate": 8.28506329113924e-06, + "loss": 0.0001, + "step": 7275 + }, + { + "epoch": 73.18, + "learning_rate": 8.27873417721519e-06, + "loss": 0.0001, + "step": 7300 + }, + { + "epoch": 73.43, + "learning_rate": 8.27240506329114e-06, + "loss": 0.0001, + "step": 7325 + }, + { + "epoch": 73.68, + "learning_rate": 8.266075949367089e-06, + "loss": 0.0001, + "step": 7350 + }, + { + "epoch": 73.93, + "learning_rate": 8.259746835443039e-06, + "loss": 0.0001, + "step": 7375 + }, + { + "epoch": 74.19, + "learning_rate": 8.253417721518988e-06, + "loss": 0.0001, + "step": 7400 + }, + { + "epoch": 74.44, + "learning_rate": 8.247088607594937e-06, + "loss": 0.0001, + "step": 7425 + }, + { + "epoch": 74.69, + "learning_rate": 8.240759493670887e-06, + "loss": 0.0001, + "step": 7450 + }, + { + "epoch": 74.94, + "learning_rate": 8.234430379746836e-06, + "loss": 0.0001, + "step": 7475 + }, + { + "epoch": 75.19, + "learning_rate": 8.228101265822786e-06, + "loss": 0.0001, + "step": 7500 + }, + { + "epoch": 75.44, + "learning_rate": 8.221772151898735e-06, + "loss": 0.0001, + "step": 7525 + }, + { + "epoch": 75.69, + "learning_rate": 8.215443037974685e-06, + "loss": 0.0001, + "step": 7550 + }, + { + "epoch": 75.94, + "learning_rate": 8.209113924050634e-06, + "loss": 0.0001, + "step": 7575 + }, + { + "epoch": 76.19, + "learning_rate": 8.202784810126583e-06, + "loss": 0.0001, + "step": 7600 + }, + { + "epoch": 76.44, + "learning_rate": 8.196455696202533e-06, + "loss": 0.0001, + "step": 7625 + }, + { + "epoch": 76.69, + "learning_rate": 8.190126582278482e-06, + "loss": 0.0001, + "step": 7650 + }, + { + "epoch": 76.94, + "learning_rate": 8.18379746835443e-06, + "loss": 0.0001, + "step": 7675 + }, + { + "epoch": 77.19, + "learning_rate": 8.17746835443038e-06, + "loss": 0.0001, + "step": 7700 + }, + { + "epoch": 77.44, + "learning_rate": 8.171139240506329e-06, + "loss": 0.0001, + "step": 7725 + }, + { + "epoch": 77.69, + "learning_rate": 8.164810126582278e-06, + "loss": 0.0001, + "step": 7750 + }, + { + "epoch": 77.94, + "learning_rate": 8.158481012658228e-06, + "loss": 0.0001, + "step": 7775 + }, + { + "epoch": 78.2, + "learning_rate": 8.152151898734177e-06, + "loss": 0.0001, + "step": 7800 + }, + { + "epoch": 78.45, + "learning_rate": 8.145822784810127e-06, + "loss": 0.0001, + "step": 7825 + }, + { + "epoch": 78.7, + "learning_rate": 8.139493670886076e-06, + "loss": 0.0001, + "step": 7850 + }, + { + "epoch": 78.95, + "learning_rate": 8.133164556962025e-06, + "loss": 0.0001, + "step": 7875 + }, + { + "epoch": 79.2, + "learning_rate": 8.126835443037976e-06, + "loss": 0.0001, + "step": 7900 + }, + { + "epoch": 79.45, + "learning_rate": 8.120506329113926e-06, + "loss": 0.0001, + "step": 7925 + }, + { + "epoch": 79.7, + "learning_rate": 8.114177215189875e-06, + "loss": 0.0001, + "step": 7950 + }, + { + "epoch": 79.95, + "learning_rate": 8.107848101265823e-06, + "loss": 0.0001, + "step": 7975 + }, + { + "epoch": 80.2, + "learning_rate": 8.101518987341773e-06, + "loss": 0.0001, + "step": 8000 + }, + { + "epoch": 80.2, + "eval_loss": 0.3960939645767212, + "eval_runtime": 495.1327, + "eval_samples_per_second": 25.735, + "eval_steps_per_second": 0.806, + "eval_wer": 13.529779584180085, + "step": 8000 + }, + { + "epoch": 80.45, + "learning_rate": 8.095189873417722e-06, + "loss": 0.0001, + "step": 8025 + }, + { + "epoch": 80.7, + "learning_rate": 8.088860759493671e-06, + "loss": 0.0001, + "step": 8050 + }, + { + "epoch": 80.95, + "learning_rate": 8.08253164556962e-06, + "loss": 0.0001, + "step": 8075 + }, + { + "epoch": 81.2, + "learning_rate": 8.07620253164557e-06, + "loss": 0.0001, + "step": 8100 + }, + { + "epoch": 81.45, + "learning_rate": 8.06987341772152e-06, + "loss": 0.0001, + "step": 8125 + }, + { + "epoch": 81.7, + "learning_rate": 8.063544303797469e-06, + "loss": 0.0001, + "step": 8150 + }, + { + "epoch": 81.95, + "learning_rate": 8.057215189873419e-06, + "loss": 0.0001, + "step": 8175 + }, + { + "epoch": 82.21, + "learning_rate": 8.050886075949368e-06, + "loss": 0.0001, + "step": 8200 + }, + { + "epoch": 82.46, + "learning_rate": 8.044556962025317e-06, + "loss": 0.0001, + "step": 8225 + }, + { + "epoch": 82.71, + "learning_rate": 8.038227848101267e-06, + "loss": 0.0001, + "step": 8250 + }, + { + "epoch": 82.96, + "learning_rate": 8.031898734177216e-06, + "loss": 0.0001, + "step": 8275 + }, + { + "epoch": 83.21, + "learning_rate": 8.025569620253166e-06, + "loss": 0.0001, + "step": 8300 + }, + { + "epoch": 83.46, + "learning_rate": 8.019240506329115e-06, + "loss": 0.0001, + "step": 8325 + }, + { + "epoch": 83.71, + "learning_rate": 8.012911392405064e-06, + "loss": 0.0001, + "step": 8350 + }, + { + "epoch": 83.96, + "learning_rate": 8.006582278481014e-06, + "loss": 0.0001, + "step": 8375 + }, + { + "epoch": 84.21, + "learning_rate": 8.000253164556963e-06, + "loss": 0.0001, + "step": 8400 + }, + { + "epoch": 84.46, + "learning_rate": 7.993924050632911e-06, + "loss": 0.0001, + "step": 8425 + }, + { + "epoch": 84.71, + "learning_rate": 7.98759493670886e-06, + "loss": 0.0001, + "step": 8450 + }, + { + "epoch": 84.96, + "learning_rate": 7.98126582278481e-06, + "loss": 0.0001, + "step": 8475 + }, + { + "epoch": 85.21, + "learning_rate": 7.97493670886076e-06, + "loss": 0.0001, + "step": 8500 + }, + { + "epoch": 85.46, + "learning_rate": 7.968607594936709e-06, + "loss": 0.0001, + "step": 8525 + }, + { + "epoch": 85.71, + "learning_rate": 7.962278481012658e-06, + "loss": 0.0001, + "step": 8550 + }, + { + "epoch": 85.96, + "learning_rate": 7.955949367088608e-06, + "loss": 0.0001, + "step": 8575 + }, + { + "epoch": 86.22, + "learning_rate": 7.949620253164557e-06, + "loss": 0.0001, + "step": 8600 + }, + { + "epoch": 86.47, + "learning_rate": 7.943291139240507e-06, + "loss": 0.0001, + "step": 8625 + }, + { + "epoch": 86.72, + "learning_rate": 7.936962025316456e-06, + "loss": 0.0001, + "step": 8650 + }, + { + "epoch": 86.97, + "learning_rate": 7.930632911392405e-06, + "loss": 0.0001, + "step": 8675 + }, + { + "epoch": 87.22, + "learning_rate": 7.924303797468355e-06, + "loss": 0.0001, + "step": 8700 + }, + { + "epoch": 87.47, + "learning_rate": 7.917974683544304e-06, + "loss": 0.0, + "step": 8725 + }, + { + "epoch": 87.72, + "learning_rate": 7.911645569620254e-06, + "loss": 0.0, + "step": 8750 + }, + { + "epoch": 87.97, + "learning_rate": 7.905316455696203e-06, + "loss": 0.0001, + "step": 8775 + }, + { + "epoch": 88.22, + "learning_rate": 7.898987341772153e-06, + "loss": 0.0, + "step": 8800 + }, + { + "epoch": 88.47, + "learning_rate": 7.892658227848102e-06, + "loss": 0.0, + "step": 8825 + }, + { + "epoch": 88.72, + "learning_rate": 7.886329113924051e-06, + "loss": 0.0, + "step": 8850 + }, + { + "epoch": 88.97, + "learning_rate": 7.88e-06, + "loss": 0.0001, + "step": 8875 + }, + { + "epoch": 89.22, + "learning_rate": 7.87367088607595e-06, + "loss": 0.0001, + "step": 8900 + }, + { + "epoch": 89.47, + "learning_rate": 7.8673417721519e-06, + "loss": 0.0001, + "step": 8925 + }, + { + "epoch": 89.72, + "learning_rate": 7.861012658227849e-06, + "loss": 0.0, + "step": 8950 + }, + { + "epoch": 89.97, + "learning_rate": 7.854683544303798e-06, + "loss": 0.0, + "step": 8975 + }, + { + "epoch": 90.23, + "learning_rate": 7.848354430379748e-06, + "loss": 0.0, + "step": 9000 + }, + { + "epoch": 90.23, + "eval_loss": 0.4116778075695038, + "eval_runtime": 494.4605, + "eval_samples_per_second": 25.769, + "eval_steps_per_second": 0.807, + "eval_wer": 13.564952321400655, + "step": 9000 + }, + { + "epoch": 90.48, + "learning_rate": 7.842025316455697e-06, + "loss": 0.0, + "step": 9025 + }, + { + "epoch": 90.73, + "learning_rate": 7.835696202531647e-06, + "loss": 0.0, + "step": 9050 + }, + { + "epoch": 90.98, + "learning_rate": 7.829367088607596e-06, + "loss": 0.0001, + "step": 9075 + }, + { + "epoch": 91.23, + "learning_rate": 7.823037974683546e-06, + "loss": 0.0, + "step": 9100 + }, + { + "epoch": 91.48, + "learning_rate": 7.816708860759495e-06, + "loss": 0.0, + "step": 9125 + }, + { + "epoch": 91.73, + "learning_rate": 7.810379746835443e-06, + "loss": 0.0, + "step": 9150 + }, + { + "epoch": 91.98, + "learning_rate": 7.804050632911392e-06, + "loss": 0.0, + "step": 9175 + }, + { + "epoch": 92.23, + "learning_rate": 7.797721518987342e-06, + "loss": 0.0001, + "step": 9200 + }, + { + "epoch": 92.48, + "learning_rate": 7.791392405063291e-06, + "loss": 0.0, + "step": 9225 + }, + { + "epoch": 92.73, + "learning_rate": 7.78506329113924e-06, + "loss": 0.0, + "step": 9250 + }, + { + "epoch": 92.98, + "learning_rate": 7.77873417721519e-06, + "loss": 0.0, + "step": 9275 + }, + { + "epoch": 93.23, + "learning_rate": 7.77240506329114e-06, + "loss": 0.0, + "step": 9300 + }, + { + "epoch": 93.48, + "learning_rate": 7.766075949367089e-06, + "loss": 0.0, + "step": 9325 + }, + { + "epoch": 93.73, + "learning_rate": 7.759746835443038e-06, + "loss": 0.0, + "step": 9350 + }, + { + "epoch": 93.98, + "learning_rate": 7.753417721518988e-06, + "loss": 0.0001, + "step": 9375 + }, + { + "epoch": 94.24, + "learning_rate": 7.747088607594937e-06, + "loss": 0.0, + "step": 9400 + }, + { + "epoch": 94.49, + "learning_rate": 7.740759493670887e-06, + "loss": 0.0, + "step": 9425 + }, + { + "epoch": 94.74, + "learning_rate": 7.734430379746836e-06, + "loss": 0.0, + "step": 9450 + }, + { + "epoch": 94.99, + "learning_rate": 7.728101265822785e-06, + "loss": 0.0001, + "step": 9475 + }, + { + "epoch": 95.24, + "learning_rate": 7.721772151898735e-06, + "loss": 0.0, + "step": 9500 + }, + { + "epoch": 95.49, + "learning_rate": 7.715443037974684e-06, + "loss": 0.0, + "step": 9525 + }, + { + "epoch": 95.74, + "learning_rate": 7.709113924050634e-06, + "loss": 0.0, + "step": 9550 + }, + { + "epoch": 95.99, + "learning_rate": 7.702784810126583e-06, + "loss": 0.0, + "step": 9575 + }, + { + "epoch": 96.24, + "learning_rate": 7.696455696202532e-06, + "loss": 0.0, + "step": 9600 + }, + { + "epoch": 96.49, + "learning_rate": 7.690126582278482e-06, + "loss": 0.0, + "step": 9625 + }, + { + "epoch": 96.74, + "learning_rate": 7.683797468354431e-06, + "loss": 0.0, + "step": 9650 + }, + { + "epoch": 96.99, + "learning_rate": 7.67746835443038e-06, + "loss": 0.0, + "step": 9675 + }, + { + "epoch": 97.24, + "learning_rate": 7.67113924050633e-06, + "loss": 0.0, + "step": 9700 + }, + { + "epoch": 97.49, + "learning_rate": 7.66481012658228e-06, + "loss": 0.0, + "step": 9725 + }, + { + "epoch": 97.74, + "learning_rate": 7.658481012658229e-06, + "loss": 0.0, + "step": 9750 + }, + { + "epoch": 97.99, + "learning_rate": 7.652151898734178e-06, + "loss": 0.0, + "step": 9775 + }, + { + "epoch": 98.25, + "learning_rate": 7.645822784810128e-06, + "loss": 0.0, + "step": 9800 + }, + { + "epoch": 98.5, + "learning_rate": 7.639493670886077e-06, + "loss": 0.0, + "step": 9825 + }, + { + "epoch": 98.75, + "learning_rate": 7.633164556962027e-06, + "loss": 0.0, + "step": 9850 + }, + { + "epoch": 99.0, + "learning_rate": 7.626835443037975e-06, + "loss": 0.0, + "step": 9875 + }, + { + "epoch": 99.25, + "learning_rate": 7.620506329113925e-06, + "loss": 0.0, + "step": 9900 + }, + { + "epoch": 99.5, + "learning_rate": 7.614177215189874e-06, + "loss": 0.0, + "step": 9925 + }, + { + "epoch": 99.75, + "learning_rate": 7.607848101265824e-06, + "loss": 0.0, + "step": 9950 + }, + { + "epoch": 100.0, + "learning_rate": 7.601518987341773e-06, + "loss": 0.0, + "step": 9975 + }, + { + "epoch": 100.25, + "learning_rate": 7.5951898734177225e-06, + "loss": 0.0, + "step": 10000 + }, + { + "epoch": 100.25, + "eval_loss": 0.42821693420410156, + "eval_runtime": 494.8072, + "eval_samples_per_second": 25.751, + "eval_steps_per_second": 0.806, + "eval_wer": 13.62455057057996, + "step": 10000 + }, + { + "epoch": 100.5, + "learning_rate": 7.588860759493671e-06, + "loss": 0.0, + "step": 10025 + }, + { + "epoch": 100.75, + "learning_rate": 7.5825316455696205e-06, + "loss": 0.0, + "step": 10050 + }, + { + "epoch": 101.0, + "learning_rate": 7.57620253164557e-06, + "loss": 0.0, + "step": 10075 + }, + { + "epoch": 101.25, + "learning_rate": 7.569873417721519e-06, + "loss": 0.0, + "step": 10100 + }, + { + "epoch": 101.5, + "learning_rate": 7.563544303797469e-06, + "loss": 0.0, + "step": 10125 + }, + { + "epoch": 101.75, + "learning_rate": 7.557215189873418e-06, + "loss": 0.0, + "step": 10150 + }, + { + "epoch": 102.01, + "learning_rate": 7.550886075949368e-06, + "loss": 0.0, + "step": 10175 + }, + { + "epoch": 102.26, + "learning_rate": 7.544556962025317e-06, + "loss": 0.0, + "step": 10200 + }, + { + "epoch": 102.51, + "learning_rate": 7.538227848101266e-06, + "loss": 0.0004, + "step": 10225 + }, + { + "epoch": 102.76, + "learning_rate": 7.532151898734178e-06, + "loss": 0.0133, + "step": 10250 + }, + { + "epoch": 103.01, + "learning_rate": 7.525822784810127e-06, + "loss": 0.0192, + "step": 10275 + }, + { + "epoch": 103.26, + "learning_rate": 7.519493670886077e-06, + "loss": 0.0142, + "step": 10300 + }, + { + "epoch": 103.51, + "learning_rate": 7.513164556962026e-06, + "loss": 0.0118, + "step": 10325 + }, + { + "epoch": 103.76, + "learning_rate": 7.507088607594938e-06, + "loss": 0.0104, + "step": 10350 + }, + { + "epoch": 104.01, + "learning_rate": 7.500759493670887e-06, + "loss": 0.0086, + "step": 10375 + }, + { + "epoch": 104.26, + "learning_rate": 7.4944303797468365e-06, + "loss": 0.0039, + "step": 10400 + }, + { + "epoch": 104.51, + "learning_rate": 7.488101265822786e-06, + "loss": 0.0038, + "step": 10425 + }, + { + "epoch": 104.76, + "learning_rate": 7.481772151898735e-06, + "loss": 0.0036, + "step": 10450 + }, + { + "epoch": 105.01, + "learning_rate": 7.475443037974685e-06, + "loss": 0.0034, + "step": 10475 + }, + { + "epoch": 105.26, + "learning_rate": 7.469113924050634e-06, + "loss": 0.0017, + "step": 10500 + }, + { + "epoch": 105.51, + "learning_rate": 7.462784810126583e-06, + "loss": 0.0014, + "step": 10525 + }, + { + "epoch": 105.76, + "learning_rate": 7.456455696202532e-06, + "loss": 0.0012, + "step": 10550 + }, + { + "epoch": 106.02, + "learning_rate": 7.450126582278482e-06, + "loss": 0.0011, + "step": 10575 + }, + { + "epoch": 106.27, + "learning_rate": 7.443797468354431e-06, + "loss": 0.0006, + "step": 10600 + }, + { + "epoch": 106.52, + "learning_rate": 7.4374683544303805e-06, + "loss": 0.0004, + "step": 10625 + }, + { + "epoch": 106.77, + "learning_rate": 7.43113924050633e-06, + "loss": 0.0004, + "step": 10650 + }, + { + "epoch": 107.02, + "learning_rate": 7.424810126582279e-06, + "loss": 0.0004, + "step": 10675 + }, + { + "epoch": 107.27, + "learning_rate": 7.418481012658229e-06, + "loss": 0.0002, + "step": 10700 + }, + { + "epoch": 107.52, + "learning_rate": 7.412151898734178e-06, + "loss": 0.0003, + "step": 10725 + }, + { + "epoch": 107.77, + "learning_rate": 7.405822784810127e-06, + "loss": 0.0002, + "step": 10750 + }, + { + "epoch": 108.02, + "learning_rate": 7.399493670886076e-06, + "loss": 0.0004, + "step": 10775 + }, + { + "epoch": 108.27, + "learning_rate": 7.393164556962026e-06, + "loss": 0.0003, + "step": 10800 + }, + { + "epoch": 108.52, + "learning_rate": 7.386835443037975e-06, + "loss": 0.0003, + "step": 10825 + }, + { + "epoch": 108.77, + "learning_rate": 7.3805063291139245e-06, + "loss": 0.0002, + "step": 10850 + }, + { + "epoch": 109.02, + "learning_rate": 7.374177215189874e-06, + "loss": 0.0002, + "step": 10875 + }, + { + "epoch": 109.27, + "learning_rate": 7.367848101265823e-06, + "loss": 0.0002, + "step": 10900 + }, + { + "epoch": 109.52, + "learning_rate": 7.361518987341773e-06, + "loss": 0.0002, + "step": 10925 + }, + { + "epoch": 109.77, + "learning_rate": 7.355189873417722e-06, + "loss": 0.0002, + "step": 10950 + }, + { + "epoch": 110.03, + "learning_rate": 7.348860759493671e-06, + "loss": 0.0002, + "step": 10975 + }, + { + "epoch": 110.28, + "learning_rate": 7.34253164556962e-06, + "loss": 0.0001, + "step": 11000 + }, + { + "epoch": 110.28, + "eval_loss": 0.35419365763664246, + "eval_runtime": 494.7916, + "eval_samples_per_second": 25.752, + "eval_steps_per_second": 0.806, + "eval_wer": 13.006096607784901, + "step": 11000 + }, + { + "epoch": 110.53, + "learning_rate": 7.33620253164557e-06, + "loss": 0.0002, + "step": 11025 + }, + { + "epoch": 110.78, + "learning_rate": 7.329873417721519e-06, + "loss": 0.0003, + "step": 11050 + }, + { + "epoch": 111.03, + "learning_rate": 7.3235443037974685e-06, + "loss": 0.0002, + "step": 11075 + }, + { + "epoch": 111.28, + "learning_rate": 7.317215189873418e-06, + "loss": 0.0002, + "step": 11100 + }, + { + "epoch": 111.53, + "learning_rate": 7.310886075949367e-06, + "loss": 0.0002, + "step": 11125 + }, + { + "epoch": 111.78, + "learning_rate": 7.304556962025317e-06, + "loss": 0.0002, + "step": 11150 + }, + { + "epoch": 112.03, + "learning_rate": 7.298227848101267e-06, + "loss": 0.0002, + "step": 11175 + }, + { + "epoch": 112.28, + "learning_rate": 7.2918987341772165e-06, + "loss": 0.0001, + "step": 11200 + }, + { + "epoch": 112.53, + "learning_rate": 7.285569620253166e-06, + "loss": 0.0001, + "step": 11225 + }, + { + "epoch": 112.78, + "learning_rate": 7.2792405063291145e-06, + "loss": 0.0002, + "step": 11250 + }, + { + "epoch": 113.03, + "learning_rate": 7.272911392405064e-06, + "loss": 0.0001, + "step": 11275 + }, + { + "epoch": 113.28, + "learning_rate": 7.266582278481013e-06, + "loss": 0.0002, + "step": 11300 + }, + { + "epoch": 113.53, + "learning_rate": 7.260253164556963e-06, + "loss": 0.0003, + "step": 11325 + }, + { + "epoch": 113.78, + "learning_rate": 7.253924050632912e-06, + "loss": 0.0002, + "step": 11350 + }, + { + "epoch": 114.04, + "learning_rate": 7.247594936708862e-06, + "loss": 0.0004, + "step": 11375 + }, + { + "epoch": 114.29, + "learning_rate": 7.241265822784811e-06, + "loss": 0.0004, + "step": 11400 + }, + { + "epoch": 114.54, + "learning_rate": 7.2349367088607605e-06, + "loss": 0.0005, + "step": 11425 + }, + { + "epoch": 114.79, + "learning_rate": 7.22860759493671e-06, + "loss": 0.0005, + "step": 11450 + }, + { + "epoch": 115.04, + "learning_rate": 7.2222784810126585e-06, + "loss": 0.0004, + "step": 11475 + }, + { + "epoch": 115.29, + "learning_rate": 7.215949367088608e-06, + "loss": 0.0005, + "step": 11500 + }, + { + "epoch": 115.54, + "learning_rate": 7.209620253164557e-06, + "loss": 0.0008, + "step": 11525 + }, + { + "epoch": 115.79, + "learning_rate": 7.203291139240507e-06, + "loss": 0.0006, + "step": 11550 + }, + { + "epoch": 116.04, + "learning_rate": 7.196962025316456e-06, + "loss": 0.0005, + "step": 11575 + }, + { + "epoch": 116.29, + "learning_rate": 7.190632911392406e-06, + "loss": 0.0005, + "step": 11600 + }, + { + "epoch": 116.54, + "learning_rate": 7.184303797468355e-06, + "loss": 0.0004, + "step": 11625 + }, + { + "epoch": 116.79, + "learning_rate": 7.1779746835443045e-06, + "loss": 0.0003, + "step": 11650 + }, + { + "epoch": 117.04, + "learning_rate": 7.171645569620254e-06, + "loss": 0.0003, + "step": 11675 + }, + { + "epoch": 117.29, + "learning_rate": 7.1653164556962025e-06, + "loss": 0.0002, + "step": 11700 + }, + { + "epoch": 117.54, + "learning_rate": 7.158987341772152e-06, + "loss": 0.0002, + "step": 11725 + }, + { + "epoch": 117.79, + "learning_rate": 7.152658227848101e-06, + "loss": 0.0001, + "step": 11750 + }, + { + "epoch": 118.05, + "learning_rate": 7.146329113924051e-06, + "loss": 0.0001, + "step": 11775 + }, + { + "epoch": 118.3, + "learning_rate": 7.14e-06, + "loss": 0.0001, + "step": 11800 + }, + { + "epoch": 118.55, + "learning_rate": 7.13367088607595e-06, + "loss": 0.0001, + "step": 11825 + }, + { + "epoch": 118.8, + "learning_rate": 7.127341772151899e-06, + "loss": 0.0001, + "step": 11850 + }, + { + "epoch": 119.05, + "learning_rate": 7.1210126582278485e-06, + "loss": 0.0001, + "step": 11875 + }, + { + "epoch": 119.3, + "learning_rate": 7.114683544303798e-06, + "loss": 0.0001, + "step": 11900 + }, + { + "epoch": 119.55, + "learning_rate": 7.1083544303797465e-06, + "loss": 0.0001, + "step": 11925 + }, + { + "epoch": 119.8, + "learning_rate": 7.102025316455696e-06, + "loss": 0.0001, + "step": 11950 + }, + { + "epoch": 120.05, + "learning_rate": 7.095696202531647e-06, + "loss": 0.0001, + "step": 11975 + }, + { + "epoch": 120.3, + "learning_rate": 7.089367088607596e-06, + "loss": 0.0001, + "step": 12000 + }, + { + "epoch": 120.3, + "eval_loss": 0.3697241246700287, + "eval_runtime": 497.0151, + "eval_samples_per_second": 25.637, + "eval_steps_per_second": 0.803, + "eval_wer": 13.128224167578553, + "step": 12000 + }, + { + "epoch": 120.55, + "learning_rate": 7.083037974683545e-06, + "loss": 0.0001, + "step": 12025 + }, + { + "epoch": 120.8, + "learning_rate": 7.0767088607594945e-06, + "loss": 0.0001, + "step": 12050 + }, + { + "epoch": 121.05, + "learning_rate": 7.070379746835444e-06, + "loss": 0.0001, + "step": 12075 + }, + { + "epoch": 121.3, + "learning_rate": 7.064050632911393e-06, + "loss": 0.0001, + "step": 12100 + }, + { + "epoch": 121.55, + "learning_rate": 7.057721518987343e-06, + "loss": 0.0001, + "step": 12125 + }, + { + "epoch": 121.8, + "learning_rate": 7.051392405063292e-06, + "loss": 0.0001, + "step": 12150 + }, + { + "epoch": 122.06, + "learning_rate": 7.045063291139242e-06, + "loss": 0.0001, + "step": 12175 + }, + { + "epoch": 122.31, + "learning_rate": 7.038734177215191e-06, + "loss": 0.0001, + "step": 12200 + }, + { + "epoch": 122.56, + "learning_rate": 7.03240506329114e-06, + "loss": 0.0001, + "step": 12225 + }, + { + "epoch": 122.81, + "learning_rate": 7.026075949367089e-06, + "loss": 0.0001, + "step": 12250 + }, + { + "epoch": 123.06, + "learning_rate": 7.0197468354430385e-06, + "loss": 0.0001, + "step": 12275 + }, + { + "epoch": 123.31, + "learning_rate": 7.013417721518988e-06, + "loss": 0.0001, + "step": 12300 + }, + { + "epoch": 123.56, + "learning_rate": 7.007088607594937e-06, + "loss": 0.0001, + "step": 12325 + }, + { + "epoch": 123.81, + "learning_rate": 7.000759493670887e-06, + "loss": 0.0001, + "step": 12350 + }, + { + "epoch": 124.06, + "learning_rate": 6.994430379746836e-06, + "loss": 0.0001, + "step": 12375 + }, + { + "epoch": 124.31, + "learning_rate": 6.988101265822786e-06, + "loss": 0.0001, + "step": 12400 + }, + { + "epoch": 124.56, + "learning_rate": 6.981772151898735e-06, + "loss": 0.0001, + "step": 12425 + }, + { + "epoch": 124.81, + "learning_rate": 6.975443037974684e-06, + "loss": 0.0001, + "step": 12450 + }, + { + "epoch": 125.06, + "learning_rate": 6.969113924050633e-06, + "loss": 0.0001, + "step": 12475 + }, + { + "epoch": 125.31, + "learning_rate": 6.9627848101265825e-06, + "loss": 0.0001, + "step": 12500 + }, + { + "epoch": 125.56, + "learning_rate": 6.956455696202532e-06, + "loss": 0.0001, + "step": 12525 + }, + { + "epoch": 125.81, + "learning_rate": 6.950126582278481e-06, + "loss": 0.0001, + "step": 12550 + }, + { + "epoch": 126.07, + "learning_rate": 6.943797468354431e-06, + "loss": 0.0001, + "step": 12575 + }, + { + "epoch": 126.32, + "learning_rate": 6.93746835443038e-06, + "loss": 0.0001, + "step": 12600 + }, + { + "epoch": 126.57, + "learning_rate": 6.93113924050633e-06, + "loss": 0.0001, + "step": 12625 + }, + { + "epoch": 126.82, + "learning_rate": 6.924810126582279e-06, + "loss": 0.0001, + "step": 12650 + }, + { + "epoch": 127.07, + "learning_rate": 6.918481012658228e-06, + "loss": 0.0001, + "step": 12675 + }, + { + "epoch": 127.32, + "learning_rate": 6.912151898734177e-06, + "loss": 0.0001, + "step": 12700 + }, + { + "epoch": 127.57, + "learning_rate": 6.9058227848101265e-06, + "loss": 0.0001, + "step": 12725 + }, + { + "epoch": 127.82, + "learning_rate": 6.899493670886076e-06, + "loss": 0.0, + "step": 12750 + }, + { + "epoch": 128.07, + "learning_rate": 6.893164556962025e-06, + "loss": 0.0, + "step": 12775 + }, + { + "epoch": 128.32, + "learning_rate": 6.886835443037976e-06, + "loss": 0.0, + "step": 12800 + }, + { + "epoch": 128.57, + "learning_rate": 6.880506329113925e-06, + "loss": 0.0, + "step": 12825 + }, + { + "epoch": 128.82, + "learning_rate": 6.8741772151898745e-06, + "loss": 0.0001, + "step": 12850 + }, + { + "epoch": 129.07, + "learning_rate": 6.867848101265824e-06, + "loss": 0.0, + "step": 12875 + }, + { + "epoch": 129.32, + "learning_rate": 6.861518987341773e-06, + "loss": 0.0, + "step": 12900 + }, + { + "epoch": 129.57, + "learning_rate": 6.855189873417723e-06, + "loss": 0.0, + "step": 12925 + }, + { + "epoch": 129.82, + "learning_rate": 6.848860759493671e-06, + "loss": 0.0001, + "step": 12950 + }, + { + "epoch": 130.08, + "learning_rate": 6.842531645569621e-06, + "loss": 0.0, + "step": 12975 + }, + { + "epoch": 130.33, + "learning_rate": 6.83620253164557e-06, + "loss": 0.0, + "step": 13000 + }, + { + "epoch": 130.33, + "eval_loss": 0.38738957047462463, + "eval_runtime": 494.657, + "eval_samples_per_second": 25.759, + "eval_steps_per_second": 0.807, + "eval_wer": 12.99339534156636, + "step": 13000 + }, + { + "epoch": 130.58, + "learning_rate": 6.82987341772152e-06, + "loss": 0.0, + "step": 13025 + }, + { + "epoch": 130.83, + "learning_rate": 6.823544303797469e-06, + "loss": 0.0001, + "step": 13050 + }, + { + "epoch": 131.08, + "learning_rate": 6.8172151898734185e-06, + "loss": 0.0, + "step": 13075 + }, + { + "epoch": 131.33, + "learning_rate": 6.810886075949368e-06, + "loss": 0.0001, + "step": 13100 + }, + { + "epoch": 131.58, + "learning_rate": 6.804556962025317e-06, + "loss": 0.0, + "step": 13125 + }, + { + "epoch": 131.83, + "learning_rate": 6.798227848101267e-06, + "loss": 0.0, + "step": 13150 + }, + { + "epoch": 132.08, + "learning_rate": 6.791898734177215e-06, + "loss": 0.0, + "step": 13175 + }, + { + "epoch": 132.33, + "learning_rate": 6.785569620253165e-06, + "loss": 0.0, + "step": 13200 + }, + { + "epoch": 132.58, + "learning_rate": 6.779240506329114e-06, + "loss": 0.0, + "step": 13225 + }, + { + "epoch": 132.83, + "learning_rate": 6.772911392405064e-06, + "loss": 0.0, + "step": 13250 + }, + { + "epoch": 133.08, + "learning_rate": 6.766582278481013e-06, + "loss": 0.0, + "step": 13275 + }, + { + "epoch": 133.33, + "learning_rate": 6.7602531645569625e-06, + "loss": 0.0, + "step": 13300 + }, + { + "epoch": 133.58, + "learning_rate": 6.753924050632912e-06, + "loss": 0.0, + "step": 13325 + }, + { + "epoch": 133.83, + "learning_rate": 6.747594936708861e-06, + "loss": 0.0, + "step": 13350 + }, + { + "epoch": 134.09, + "learning_rate": 6.741265822784811e-06, + "loss": 0.0001, + "step": 13375 + }, + { + "epoch": 134.34, + "learning_rate": 6.734936708860759e-06, + "loss": 0.0, + "step": 13400 + }, + { + "epoch": 134.59, + "learning_rate": 6.728607594936709e-06, + "loss": 0.0, + "step": 13425 + }, + { + "epoch": 134.84, + "learning_rate": 6.722278481012658e-06, + "loss": 0.0, + "step": 13450 + }, + { + "epoch": 135.09, + "learning_rate": 6.715949367088608e-06, + "loss": 0.0, + "step": 13475 + }, + { + "epoch": 135.34, + "learning_rate": 6.709620253164557e-06, + "loss": 0.0, + "step": 13500 + }, + { + "epoch": 135.59, + "learning_rate": 6.7032911392405065e-06, + "loss": 0.0, + "step": 13525 + }, + { + "epoch": 135.84, + "learning_rate": 6.696962025316456e-06, + "loss": 0.0, + "step": 13550 + }, + { + "epoch": 136.09, + "learning_rate": 6.690632911392405e-06, + "loss": 0.0, + "step": 13575 + }, + { + "epoch": 136.34, + "learning_rate": 6.684303797468355e-06, + "loss": 0.0, + "step": 13600 + }, + { + "epoch": 136.59, + "learning_rate": 6.677974683544305e-06, + "loss": 0.0, + "step": 13625 + }, + { + "epoch": 136.84, + "learning_rate": 6.6716455696202545e-06, + "loss": 0.0, + "step": 13650 + }, + { + "epoch": 137.09, + "learning_rate": 6.665316455696204e-06, + "loss": 0.0, + "step": 13675 + }, + { + "epoch": 137.34, + "learning_rate": 6.6589873417721525e-06, + "loss": 0.0, + "step": 13700 + }, + { + "epoch": 137.59, + "learning_rate": 6.652658227848102e-06, + "loss": 0.0, + "step": 13725 + }, + { + "epoch": 137.84, + "learning_rate": 6.646329113924051e-06, + "loss": 0.0, + "step": 13750 + }, + { + "epoch": 138.1, + "learning_rate": 6.640000000000001e-06, + "loss": 0.0, + "step": 13775 + }, + { + "epoch": 138.35, + "learning_rate": 6.63367088607595e-06, + "loss": 0.0, + "step": 13800 + }, + { + "epoch": 138.6, + "learning_rate": 6.6273417721519e-06, + "loss": 0.0, + "step": 13825 + }, + { + "epoch": 138.85, + "learning_rate": 6.621012658227849e-06, + "loss": 0.0, + "step": 13850 + }, + { + "epoch": 139.1, + "learning_rate": 6.6146835443037985e-06, + "loss": 0.0, + "step": 13875 + }, + { + "epoch": 139.35, + "learning_rate": 6.608354430379748e-06, + "loss": 0.0, + "step": 13900 + }, + { + "epoch": 139.6, + "learning_rate": 6.6020253164556965e-06, + "loss": 0.0, + "step": 13925 + }, + { + "epoch": 139.85, + "learning_rate": 6.595696202531646e-06, + "loss": 0.0, + "step": 13950 + }, + { + "epoch": 140.1, + "learning_rate": 6.589367088607595e-06, + "loss": 0.0, + "step": 13975 + }, + { + "epoch": 140.35, + "learning_rate": 6.583037974683545e-06, + "loss": 0.0, + "step": 14000 + }, + { + "epoch": 140.35, + "eval_loss": 0.4001910090446472, + "eval_runtime": 494.812, + "eval_samples_per_second": 25.751, + "eval_steps_per_second": 0.806, + "eval_wer": 12.958222604345787, + "step": 14000 + }, + { + "epoch": 140.6, + "learning_rate": 6.576708860759494e-06, + "loss": 0.0, + "step": 14025 + }, + { + "epoch": 140.85, + "learning_rate": 6.570379746835444e-06, + "loss": 0.0, + "step": 14050 + }, + { + "epoch": 141.1, + "learning_rate": 6.564050632911393e-06, + "loss": 0.0, + "step": 14075 + }, + { + "epoch": 141.35, + "learning_rate": 6.5577215189873425e-06, + "loss": 0.0, + "step": 14100 + }, + { + "epoch": 141.6, + "learning_rate": 6.551392405063292e-06, + "loss": 0.0, + "step": 14125 + }, + { + "epoch": 141.85, + "learning_rate": 6.5450632911392405e-06, + "loss": 0.0, + "step": 14150 + }, + { + "epoch": 142.11, + "learning_rate": 6.53873417721519e-06, + "loss": 0.0, + "step": 14175 + }, + { + "epoch": 142.36, + "learning_rate": 6.532405063291139e-06, + "loss": 0.0, + "step": 14200 + }, + { + "epoch": 142.61, + "learning_rate": 6.526075949367089e-06, + "loss": 0.0, + "step": 14225 + }, + { + "epoch": 142.86, + "learning_rate": 6.519746835443038e-06, + "loss": 0.0, + "step": 14250 + }, + { + "epoch": 143.11, + "learning_rate": 6.513417721518988e-06, + "loss": 0.0, + "step": 14275 + }, + { + "epoch": 143.36, + "learning_rate": 6.507088607594937e-06, + "loss": 0.0, + "step": 14300 + }, + { + "epoch": 143.61, + "learning_rate": 6.5007594936708865e-06, + "loss": 0.0, + "step": 14325 + }, + { + "epoch": 143.86, + "learning_rate": 6.494430379746836e-06, + "loss": 0.0, + "step": 14350 + }, + { + "epoch": 144.11, + "learning_rate": 6.4881012658227845e-06, + "loss": 0.0, + "step": 14375 + }, + { + "epoch": 144.36, + "learning_rate": 6.481772151898734e-06, + "loss": 0.0, + "step": 14400 + }, + { + "epoch": 144.61, + "learning_rate": 6.475443037974683e-06, + "loss": 0.0, + "step": 14425 + }, + { + "epoch": 144.86, + "learning_rate": 6.469113924050634e-06, + "loss": 0.0, + "step": 14450 + }, + { + "epoch": 145.11, + "learning_rate": 6.462784810126583e-06, + "loss": 0.0, + "step": 14475 + }, + { + "epoch": 145.36, + "learning_rate": 6.4564556962025325e-06, + "loss": 0.0, + "step": 14500 + }, + { + "epoch": 145.61, + "learning_rate": 6.450126582278482e-06, + "loss": 0.0, + "step": 14525 + }, + { + "epoch": 145.86, + "learning_rate": 6.443797468354431e-06, + "loss": 0.0, + "step": 14550 + }, + { + "epoch": 146.12, + "learning_rate": 6.437468354430381e-06, + "loss": 0.0, + "step": 14575 + }, + { + "epoch": 146.37, + "learning_rate": 6.43113924050633e-06, + "loss": 0.0, + "step": 14600 + }, + { + "epoch": 146.62, + "learning_rate": 6.42481012658228e-06, + "loss": 0.0, + "step": 14625 + }, + { + "epoch": 146.87, + "learning_rate": 6.418481012658228e-06, + "loss": 0.0, + "step": 14650 + }, + { + "epoch": 147.12, + "learning_rate": 6.412151898734178e-06, + "loss": 0.0, + "step": 14675 + }, + { + "epoch": 147.37, + "learning_rate": 6.405822784810127e-06, + "loss": 0.0, + "step": 14700 + }, + { + "epoch": 147.62, + "learning_rate": 6.3994936708860765e-06, + "loss": 0.0, + "step": 14725 + }, + { + "epoch": 147.87, + "learning_rate": 6.393164556962026e-06, + "loss": 0.0, + "step": 14750 + }, + { + "epoch": 148.12, + "learning_rate": 6.386835443037975e-06, + "loss": 0.0, + "step": 14775 + }, + { + "epoch": 148.37, + "learning_rate": 6.380506329113925e-06, + "loss": 0.0, + "step": 14800 + }, + { + "epoch": 148.62, + "learning_rate": 6.374177215189874e-06, + "loss": 0.0, + "step": 14825 + }, + { + "epoch": 148.87, + "learning_rate": 6.367848101265824e-06, + "loss": 0.0, + "step": 14850 + }, + { + "epoch": 149.12, + "learning_rate": 6.361518987341772e-06, + "loss": 0.0, + "step": 14875 + }, + { + "epoch": 149.37, + "learning_rate": 6.355189873417722e-06, + "loss": 0.0, + "step": 14900 + }, + { + "epoch": 149.62, + "learning_rate": 6.348860759493671e-06, + "loss": 0.0, + "step": 14925 + }, + { + "epoch": 149.87, + "learning_rate": 6.3425316455696205e-06, + "loss": 0.0, + "step": 14950 + }, + { + "epoch": 150.13, + "learning_rate": 6.33620253164557e-06, + "loss": 0.0, + "step": 14975 + }, + { + "epoch": 150.38, + "learning_rate": 6.329873417721519e-06, + "loss": 0.0, + "step": 15000 + }, + { + "epoch": 150.38, + "eval_loss": 0.41198477149009705, + "eval_runtime": 494.3468, + "eval_samples_per_second": 25.775, + "eval_steps_per_second": 0.807, + "eval_wer": 12.945521338127246, + "step": 15000 + }, + { + "epoch": 150.63, + "learning_rate": 6.323544303797469e-06, + "loss": 0.0, + "step": 15025 + }, + { + "epoch": 150.88, + "learning_rate": 6.317215189873418e-06, + "loss": 0.0, + "step": 15050 + }, + { + "epoch": 151.13, + "learning_rate": 6.310886075949368e-06, + "loss": 0.0, + "step": 15075 + }, + { + "epoch": 151.38, + "learning_rate": 6.304556962025316e-06, + "loss": 0.0, + "step": 15100 + }, + { + "epoch": 151.63, + "learning_rate": 6.298227848101266e-06, + "loss": 0.0, + "step": 15125 + }, + { + "epoch": 151.88, + "learning_rate": 6.291898734177215e-06, + "loss": 0.0, + "step": 15150 + }, + { + "epoch": 152.13, + "learning_rate": 6.2855696202531645e-06, + "loss": 0.0, + "step": 15175 + }, + { + "epoch": 152.38, + "learning_rate": 6.279240506329114e-06, + "loss": 0.0, + "step": 15200 + }, + { + "epoch": 152.63, + "learning_rate": 6.272911392405063e-06, + "loss": 0.0, + "step": 15225 + }, + { + "epoch": 152.88, + "learning_rate": 6.266582278481013e-06, + "loss": 0.0, + "step": 15250 + }, + { + "epoch": 153.13, + "learning_rate": 6.260253164556963e-06, + "loss": 0.0, + "step": 15275 + }, + { + "epoch": 153.38, + "learning_rate": 6.2539240506329125e-06, + "loss": 0.0, + "step": 15300 + }, + { + "epoch": 153.63, + "learning_rate": 6.247594936708862e-06, + "loss": 0.0, + "step": 15325 + }, + { + "epoch": 153.88, + "learning_rate": 6.241265822784811e-06, + "loss": 0.0, + "step": 15350 + }, + { + "epoch": 154.14, + "learning_rate": 6.234936708860761e-06, + "loss": 0.0, + "step": 15375 + }, + { + "epoch": 154.39, + "learning_rate": 6.228607594936709e-06, + "loss": 0.0, + "step": 15400 + }, + { + "epoch": 154.64, + "learning_rate": 6.222278481012659e-06, + "loss": 0.0, + "step": 15425 + }, + { + "epoch": 154.89, + "learning_rate": 6.215949367088608e-06, + "loss": 0.0, + "step": 15450 + }, + { + "epoch": 155.14, + "learning_rate": 6.209620253164558e-06, + "loss": 0.0, + "step": 15475 + }, + { + "epoch": 155.39, + "learning_rate": 6.203291139240507e-06, + "loss": 0.0, + "step": 15500 + }, + { + "epoch": 155.64, + "learning_rate": 6.1969620253164565e-06, + "loss": 0.0, + "step": 15525 + }, + { + "epoch": 155.89, + "learning_rate": 6.190632911392406e-06, + "loss": 0.0, + "step": 15550 + }, + { + "epoch": 156.14, + "learning_rate": 6.184303797468355e-06, + "loss": 0.0, + "step": 15575 + }, + { + "epoch": 156.39, + "learning_rate": 6.177974683544305e-06, + "loss": 0.0, + "step": 15600 + }, + { + "epoch": 156.64, + "learning_rate": 6.171645569620253e-06, + "loss": 0.0, + "step": 15625 + }, + { + "epoch": 156.89, + "learning_rate": 6.165316455696203e-06, + "loss": 0.0, + "step": 15650 + }, + { + "epoch": 157.14, + "learning_rate": 6.158987341772152e-06, + "loss": 0.0, + "step": 15675 + }, + { + "epoch": 157.39, + "learning_rate": 6.152658227848102e-06, + "loss": 0.0, + "step": 15700 + }, + { + "epoch": 157.64, + "learning_rate": 6.146329113924051e-06, + "loss": 0.0, + "step": 15725 + }, + { + "epoch": 157.89, + "learning_rate": 6.1400000000000005e-06, + "loss": 0.0, + "step": 15750 + }, + { + "epoch": 158.15, + "learning_rate": 6.13367088607595e-06, + "loss": 0.0, + "step": 15775 + }, + { + "epoch": 158.4, + "learning_rate": 6.127341772151899e-06, + "loss": 0.0, + "step": 15800 + }, + { + "epoch": 158.65, + "learning_rate": 6.121012658227849e-06, + "loss": 0.0, + "step": 15825 + }, + { + "epoch": 158.9, + "learning_rate": 6.114683544303797e-06, + "loss": 0.0, + "step": 15850 + }, + { + "epoch": 159.15, + "learning_rate": 6.108354430379747e-06, + "loss": 0.0, + "step": 15875 + }, + { + "epoch": 159.4, + "learning_rate": 6.102025316455696e-06, + "loss": 0.0, + "step": 15900 + }, + { + "epoch": 159.65, + "learning_rate": 6.095696202531646e-06, + "loss": 0.0, + "step": 15925 + }, + { + "epoch": 159.9, + "learning_rate": 6.089367088607595e-06, + "loss": 0.0, + "step": 15950 + }, + { + "epoch": 160.15, + "learning_rate": 6.0830379746835445e-06, + "loss": 0.0, + "step": 15975 + }, + { + "epoch": 160.4, + "learning_rate": 6.076708860759494e-06, + "loss": 0.0, + "step": 16000 + }, + { + "epoch": 160.4, + "eval_loss": 0.42462843656539917, + "eval_runtime": 494.6558, + "eval_samples_per_second": 25.759, + "eval_steps_per_second": 0.807, + "eval_wer": 12.963107706737533, + "step": 16000 + }, + { + "epoch": 160.65, + "learning_rate": 6.070379746835443e-06, + "loss": 0.0, + "step": 16025 + }, + { + "epoch": 160.9, + "learning_rate": 6.064050632911393e-06, + "loss": 0.0, + "step": 16050 + }, + { + "epoch": 161.15, + "learning_rate": 6.057721518987341e-06, + "loss": 0.0, + "step": 16075 + }, + { + "epoch": 161.4, + "learning_rate": 6.0513924050632925e-06, + "loss": 0.0, + "step": 16100 + }, + { + "epoch": 161.65, + "learning_rate": 6.045063291139242e-06, + "loss": 0.0, + "step": 16125 + }, + { + "epoch": 161.9, + "learning_rate": 6.0387341772151905e-06, + "loss": 0.0, + "step": 16150 + }, + { + "epoch": 162.16, + "learning_rate": 6.03240506329114e-06, + "loss": 0.0, + "step": 16175 + }, + { + "epoch": 162.41, + "learning_rate": 6.026075949367089e-06, + "loss": 0.0, + "step": 16200 + }, + { + "epoch": 162.66, + "learning_rate": 6.019746835443039e-06, + "loss": 0.0, + "step": 16225 + }, + { + "epoch": 162.91, + "learning_rate": 6.013417721518988e-06, + "loss": 0.0, + "step": 16250 + }, + { + "epoch": 163.16, + "learning_rate": 6.007088607594938e-06, + "loss": 0.0, + "step": 16275 + }, + { + "epoch": 163.41, + "learning_rate": 6.000759493670887e-06, + "loss": 0.0, + "step": 16300 + }, + { + "epoch": 163.66, + "learning_rate": 5.9944303797468365e-06, + "loss": 0.0, + "step": 16325 + }, + { + "epoch": 163.91, + "learning_rate": 5.988101265822786e-06, + "loss": 0.0, + "step": 16350 + }, + { + "epoch": 164.16, + "learning_rate": 5.9817721518987345e-06, + "loss": 0.0, + "step": 16375 + }, + { + "epoch": 164.41, + "learning_rate": 5.975443037974684e-06, + "loss": 0.0, + "step": 16400 + }, + { + "epoch": 164.66, + "learning_rate": 5.969113924050633e-06, + "loss": 0.0, + "step": 16425 + }, + { + "epoch": 164.91, + "learning_rate": 5.962784810126583e-06, + "loss": 0.0, + "step": 16450 + }, + { + "epoch": 165.16, + "learning_rate": 5.956455696202532e-06, + "loss": 0.0, + "step": 16475 + }, + { + "epoch": 165.41, + "learning_rate": 5.950126582278482e-06, + "loss": 0.0, + "step": 16500 + }, + { + "epoch": 165.66, + "learning_rate": 5.943797468354431e-06, + "loss": 0.0, + "step": 16525 + }, + { + "epoch": 165.91, + "learning_rate": 5.9374683544303805e-06, + "loss": 0.0, + "step": 16550 + }, + { + "epoch": 166.17, + "learning_rate": 5.931139240506329e-06, + "loss": 0.0, + "step": 16575 + }, + { + "epoch": 166.42, + "learning_rate": 5.9248101265822785e-06, + "loss": 0.0, + "step": 16600 + }, + { + "epoch": 166.67, + "learning_rate": 5.918481012658228e-06, + "loss": 0.0, + "step": 16625 + }, + { + "epoch": 166.92, + "learning_rate": 5.912151898734177e-06, + "loss": 0.0, + "step": 16650 + }, + { + "epoch": 167.17, + "learning_rate": 5.905822784810127e-06, + "loss": 0.0, + "step": 16675 + }, + { + "epoch": 167.42, + "learning_rate": 5.899493670886076e-06, + "loss": 0.0, + "step": 16700 + }, + { + "epoch": 167.67, + "learning_rate": 5.893164556962026e-06, + "loss": 0.0, + "step": 16725 + }, + { + "epoch": 167.92, + "learning_rate": 5.886835443037975e-06, + "loss": 0.0, + "step": 16750 + }, + { + "epoch": 168.17, + "learning_rate": 5.8805063291139245e-06, + "loss": 0.0, + "step": 16775 + }, + { + "epoch": 168.42, + "learning_rate": 5.874177215189873e-06, + "loss": 0.0, + "step": 16800 + }, + { + "epoch": 168.67, + "learning_rate": 5.8678481012658225e-06, + "loss": 0.0, + "step": 16825 + }, + { + "epoch": 168.92, + "learning_rate": 5.861518987341772e-06, + "loss": 0.0, + "step": 16850 + }, + { + "epoch": 169.17, + "learning_rate": 5.855189873417721e-06, + "loss": 0.0, + "step": 16875 + }, + { + "epoch": 169.42, + "learning_rate": 5.848860759493671e-06, + "loss": 0.0, + "step": 16900 + }, + { + "epoch": 169.67, + "learning_rate": 5.842531645569621e-06, + "loss": 0.0, + "step": 16925 + }, + { + "epoch": 169.92, + "learning_rate": 5.8362025316455705e-06, + "loss": 0.0, + "step": 16950 + }, + { + "epoch": 170.18, + "learning_rate": 5.82987341772152e-06, + "loss": 0.0, + "step": 16975 + }, + { + "epoch": 170.43, + "learning_rate": 5.823544303797469e-06, + "loss": 0.0, + "step": 17000 + }, + { + "epoch": 170.43, + "eval_loss": 0.436918169260025, + "eval_runtime": 494.3338, + "eval_samples_per_second": 25.776, + "eval_steps_per_second": 0.807, + "eval_wer": 13.00707362826325, + "step": 17000 + }, + { + "epoch": 170.68, + "learning_rate": 5.817215189873419e-06, + "loss": 0.0, + "step": 17025 + }, + { + "epoch": 170.93, + "learning_rate": 5.810886075949368e-06, + "loss": 0.0, + "step": 17050 + }, + { + "epoch": 171.18, + "learning_rate": 5.804556962025318e-06, + "loss": 0.0, + "step": 17075 + }, + { + "epoch": 171.43, + "learning_rate": 5.798227848101266e-06, + "loss": 0.0, + "step": 17100 + }, + { + "epoch": 171.68, + "learning_rate": 5.791898734177216e-06, + "loss": 0.0, + "step": 17125 + }, + { + "epoch": 171.93, + "learning_rate": 5.785569620253165e-06, + "loss": 0.0, + "step": 17150 + }, + { + "epoch": 172.18, + "learning_rate": 5.7792405063291145e-06, + "loss": 0.0, + "step": 17175 + }, + { + "epoch": 172.43, + "learning_rate": 5.772911392405064e-06, + "loss": 0.0, + "step": 17200 + }, + { + "epoch": 172.68, + "learning_rate": 5.766582278481013e-06, + "loss": 0.0, + "step": 17225 + }, + { + "epoch": 172.93, + "learning_rate": 5.760253164556963e-06, + "loss": 0.0, + "step": 17250 + }, + { + "epoch": 173.18, + "learning_rate": 5.753924050632912e-06, + "loss": 0.0, + "step": 17275 + }, + { + "epoch": 173.43, + "learning_rate": 5.747594936708862e-06, + "loss": 0.0, + "step": 17300 + }, + { + "epoch": 173.68, + "learning_rate": 5.74126582278481e-06, + "loss": 0.0, + "step": 17325 + }, + { + "epoch": 173.93, + "learning_rate": 5.73493670886076e-06, + "loss": 0.0, + "step": 17350 + }, + { + "epoch": 174.19, + "learning_rate": 5.728607594936709e-06, + "loss": 0.0, + "step": 17375 + }, + { + "epoch": 174.44, + "learning_rate": 5.7222784810126585e-06, + "loss": 0.0, + "step": 17400 + }, + { + "epoch": 174.69, + "learning_rate": 5.715949367088608e-06, + "loss": 0.0, + "step": 17425 + }, + { + "epoch": 174.94, + "learning_rate": 5.709620253164557e-06, + "loss": 0.0, + "step": 17450 + }, + { + "epoch": 175.19, + "learning_rate": 5.703291139240507e-06, + "loss": 0.0, + "step": 17475 + }, + { + "epoch": 175.44, + "learning_rate": 5.696962025316456e-06, + "loss": 0.0, + "step": 17500 + }, + { + "epoch": 175.69, + "learning_rate": 5.690632911392406e-06, + "loss": 0.0, + "step": 17525 + }, + { + "epoch": 175.94, + "learning_rate": 5.684303797468354e-06, + "loss": 0.0, + "step": 17550 + }, + { + "epoch": 176.19, + "learning_rate": 5.677974683544304e-06, + "loss": 0.0, + "step": 17575 + }, + { + "epoch": 176.44, + "learning_rate": 5.671645569620253e-06, + "loss": 0.0, + "step": 17600 + }, + { + "epoch": 176.69, + "learning_rate": 5.6653164556962025e-06, + "loss": 0.0, + "step": 17625 + }, + { + "epoch": 176.94, + "learning_rate": 5.658987341772152e-06, + "loss": 0.0, + "step": 17650 + }, + { + "epoch": 177.19, + "learning_rate": 5.652658227848101e-06, + "loss": 0.0, + "step": 17675 + }, + { + "epoch": 177.44, + "learning_rate": 5.646329113924051e-06, + "loss": 0.0, + "step": 17700 + }, + { + "epoch": 177.69, + "learning_rate": 5.64e-06, + "loss": 0.0, + "step": 17725 + }, + { + "epoch": 177.94, + "learning_rate": 5.6336708860759505e-06, + "loss": 0.0, + "step": 17750 + }, + { + "epoch": 178.2, + "learning_rate": 5.6273417721519e-06, + "loss": 0.0, + "step": 17775 + }, + { + "epoch": 178.45, + "learning_rate": 5.621012658227849e-06, + "loss": 0.0, + "step": 17800 + }, + { + "epoch": 178.7, + "learning_rate": 5.614683544303799e-06, + "loss": 0.0, + "step": 17825 + }, + { + "epoch": 178.95, + "learning_rate": 5.608354430379747e-06, + "loss": 0.0, + "step": 17850 + }, + { + "epoch": 179.2, + "learning_rate": 5.602025316455697e-06, + "loss": 0.0, + "step": 17875 + }, + { + "epoch": 179.45, + "learning_rate": 5.595696202531646e-06, + "loss": 0.0, + "step": 17900 + }, + { + "epoch": 179.7, + "learning_rate": 5.589367088607596e-06, + "loss": 0.0, + "step": 17925 + }, + { + "epoch": 179.95, + "learning_rate": 5.583037974683545e-06, + "loss": 0.0, + "step": 17950 + }, + { + "epoch": 180.2, + "learning_rate": 5.5767088607594945e-06, + "loss": 0.0, + "step": 17975 + }, + { + "epoch": 180.45, + "learning_rate": 5.570379746835444e-06, + "loss": 0.0, + "step": 18000 + }, + { + "epoch": 180.45, + "eval_loss": 0.4500702917575836, + "eval_runtime": 494.3421, + "eval_samples_per_second": 25.776, + "eval_steps_per_second": 0.807, + "eval_wer": 13.036384242613725, + "step": 18000 + }, + { + "epoch": 180.7, + "learning_rate": 5.564050632911393e-06, + "loss": 0.0, + "step": 18025 + }, + { + "epoch": 180.95, + "learning_rate": 5.557721518987343e-06, + "loss": 0.0, + "step": 18050 + }, + { + "epoch": 181.2, + "learning_rate": 5.551392405063291e-06, + "loss": 0.0, + "step": 18075 + }, + { + "epoch": 181.45, + "learning_rate": 5.545063291139241e-06, + "loss": 0.0, + "step": 18100 + }, + { + "epoch": 181.7, + "learning_rate": 5.53873417721519e-06, + "loss": 0.0, + "step": 18125 + }, + { + "epoch": 181.95, + "learning_rate": 5.53240506329114e-06, + "loss": 0.0, + "step": 18150 + }, + { + "epoch": 182.21, + "learning_rate": 5.526075949367089e-06, + "loss": 0.0, + "step": 18175 + }, + { + "epoch": 182.46, + "learning_rate": 5.5197468354430385e-06, + "loss": 0.0, + "step": 18200 + }, + { + "epoch": 182.71, + "learning_rate": 5.513417721518988e-06, + "loss": 0.0, + "step": 18225 + }, + { + "epoch": 182.96, + "learning_rate": 5.507088607594937e-06, + "loss": 0.0, + "step": 18250 + }, + { + "epoch": 183.21, + "learning_rate": 5.500759493670887e-06, + "loss": 0.0, + "step": 18275 + }, + { + "epoch": 183.46, + "learning_rate": 5.494430379746835e-06, + "loss": 0.0, + "step": 18300 + }, + { + "epoch": 183.71, + "learning_rate": 5.488101265822785e-06, + "loss": 0.0, + "step": 18325 + }, + { + "epoch": 183.96, + "learning_rate": 5.481772151898734e-06, + "loss": 0.0, + "step": 18350 + }, + { + "epoch": 184.21, + "learning_rate": 5.475443037974684e-06, + "loss": 0.0, + "step": 18375 + }, + { + "epoch": 184.46, + "learning_rate": 5.469367088607596e-06, + "loss": 0.0, + "step": 18400 + }, + { + "epoch": 184.71, + "learning_rate": 5.4630379746835445e-06, + "loss": 0.0, + "step": 18425 + }, + { + "epoch": 184.96, + "learning_rate": 5.456708860759494e-06, + "loss": 0.0, + "step": 18450 + }, + { + "epoch": 185.21, + "learning_rate": 5.450379746835443e-06, + "loss": 0.0, + "step": 18475 + }, + { + "epoch": 185.46, + "learning_rate": 5.444050632911393e-06, + "loss": 0.0, + "step": 18500 + }, + { + "epoch": 185.71, + "learning_rate": 5.437721518987342e-06, + "loss": 0.0, + "step": 18525 + }, + { + "epoch": 185.96, + "learning_rate": 5.431392405063292e-06, + "loss": 0.0, + "step": 18550 + }, + { + "epoch": 186.22, + "learning_rate": 5.425063291139241e-06, + "loss": 0.0, + "step": 18575 + }, + { + "epoch": 186.47, + "learning_rate": 5.4187341772151905e-06, + "loss": 0.0, + "step": 18600 + }, + { + "epoch": 186.72, + "learning_rate": 5.41240506329114e-06, + "loss": 0.0, + "step": 18625 + }, + { + "epoch": 186.97, + "learning_rate": 5.4060759493670885e-06, + "loss": 0.0, + "step": 18650 + }, + { + "epoch": 187.22, + "learning_rate": 5.399746835443038e-06, + "loss": 0.0, + "step": 18675 + }, + { + "epoch": 187.47, + "learning_rate": 5.393417721518987e-06, + "loss": 0.0, + "step": 18700 + }, + { + "epoch": 187.72, + "learning_rate": 5.387088607594937e-06, + "loss": 0.0, + "step": 18725 + }, + { + "epoch": 187.97, + "learning_rate": 5.380759493670886e-06, + "loss": 0.0, + "step": 18750 + }, + { + "epoch": 188.22, + "learning_rate": 5.374430379746836e-06, + "loss": 0.0, + "step": 18775 + }, + { + "epoch": 188.47, + "learning_rate": 5.368101265822785e-06, + "loss": 0.0, + "step": 18800 + }, + { + "epoch": 188.72, + "learning_rate": 5.3617721518987345e-06, + "loss": 0.0, + "step": 18825 + }, + { + "epoch": 188.97, + "learning_rate": 5.355443037974684e-06, + "loss": 0.0, + "step": 18850 + }, + { + "epoch": 189.22, + "learning_rate": 5.3491139240506325e-06, + "loss": 0.0, + "step": 18875 + }, + { + "epoch": 189.47, + "learning_rate": 5.342784810126582e-06, + "loss": 0.0, + "step": 18900 + }, + { + "epoch": 189.72, + "learning_rate": 5.336455696202531e-06, + "loss": 0.0, + "step": 18925 + }, + { + "epoch": 189.97, + "learning_rate": 5.330126582278481e-06, + "loss": 0.0, + "step": 18950 + }, + { + "epoch": 190.23, + "learning_rate": 5.323797468354431e-06, + "loss": 0.0, + "step": 18975 + }, + { + "epoch": 190.48, + "learning_rate": 5.3174683544303805e-06, + "loss": 0.0, + "step": 19000 + }, + { + "epoch": 190.48, + "eval_loss": 0.46377480030059814, + "eval_runtime": 494.2458, + "eval_samples_per_second": 25.781, + "eval_steps_per_second": 0.807, + "eval_wer": 13.037361263092073, + "step": 19000 + }, + { + "epoch": 190.73, + "learning_rate": 5.31113924050633e-06, + "loss": 0.0, + "step": 19025 + }, + { + "epoch": 190.98, + "learning_rate": 5.304810126582279e-06, + "loss": 0.0, + "step": 19050 + }, + { + "epoch": 191.23, + "learning_rate": 5.298481012658229e-06, + "loss": 0.0, + "step": 19075 + }, + { + "epoch": 191.48, + "learning_rate": 5.292151898734178e-06, + "loss": 0.0, + "step": 19100 + }, + { + "epoch": 191.73, + "learning_rate": 5.285822784810128e-06, + "loss": 0.0, + "step": 19125 + }, + { + "epoch": 191.98, + "learning_rate": 5.279493670886077e-06, + "loss": 0.0, + "step": 19150 + }, + { + "epoch": 192.23, + "learning_rate": 5.273164556962026e-06, + "loss": 0.0, + "step": 19175 + }, + { + "epoch": 192.48, + "learning_rate": 5.266835443037975e-06, + "loss": 0.0, + "step": 19200 + }, + { + "epoch": 192.73, + "learning_rate": 5.2605063291139245e-06, + "loss": 0.0, + "step": 19225 + }, + { + "epoch": 192.98, + "learning_rate": 5.254177215189874e-06, + "loss": 0.0, + "step": 19250 + }, + { + "epoch": 193.23, + "learning_rate": 5.247848101265823e-06, + "loss": 0.0, + "step": 19275 + }, + { + "epoch": 193.48, + "learning_rate": 5.241518987341773e-06, + "loss": 0.0, + "step": 19300 + }, + { + "epoch": 193.73, + "learning_rate": 5.235189873417722e-06, + "loss": 0.0, + "step": 19325 + }, + { + "epoch": 193.98, + "learning_rate": 5.228860759493672e-06, + "loss": 0.0, + "step": 19350 + }, + { + "epoch": 194.24, + "learning_rate": 5.222531645569621e-06, + "loss": 0.0, + "step": 19375 + }, + { + "epoch": 194.49, + "learning_rate": 5.21620253164557e-06, + "loss": 0.0, + "step": 19400 + }, + { + "epoch": 194.74, + "learning_rate": 5.209873417721519e-06, + "loss": 0.0, + "step": 19425 + }, + { + "epoch": 194.99, + "learning_rate": 5.2035443037974685e-06, + "loss": 0.0, + "step": 19450 + }, + { + "epoch": 195.24, + "learning_rate": 5.197215189873418e-06, + "loss": 0.0, + "step": 19475 + }, + { + "epoch": 195.49, + "learning_rate": 5.190886075949367e-06, + "loss": 0.0, + "step": 19500 + }, + { + "epoch": 195.74, + "learning_rate": 5.184556962025317e-06, + "loss": 0.0, + "step": 19525 + }, + { + "epoch": 195.99, + "learning_rate": 5.178227848101266e-06, + "loss": 0.0, + "step": 19550 + }, + { + "epoch": 196.24, + "learning_rate": 5.171898734177216e-06, + "loss": 0.0, + "step": 19575 + }, + { + "epoch": 196.49, + "learning_rate": 5.165569620253165e-06, + "loss": 0.0, + "step": 19600 + }, + { + "epoch": 196.74, + "learning_rate": 5.159240506329114e-06, + "loss": 0.0, + "step": 19625 + }, + { + "epoch": 196.99, + "learning_rate": 5.152911392405063e-06, + "loss": 0.0, + "step": 19650 + }, + { + "epoch": 197.24, + "learning_rate": 5.1465822784810125e-06, + "loss": 0.0, + "step": 19675 + }, + { + "epoch": 197.49, + "learning_rate": 5.140253164556962e-06, + "loss": 0.0, + "step": 19700 + }, + { + "epoch": 197.74, + "learning_rate": 5.133924050632911e-06, + "loss": 0.0, + "step": 19725 + }, + { + "epoch": 197.99, + "learning_rate": 5.127594936708861e-06, + "loss": 0.0, + "step": 19750 + }, + { + "epoch": 198.25, + "learning_rate": 5.12126582278481e-06, + "loss": 0.0, + "step": 19775 + }, + { + "epoch": 198.5, + "learning_rate": 5.1149367088607605e-06, + "loss": 0.0, + "step": 19800 + }, + { + "epoch": 198.75, + "learning_rate": 5.10860759493671e-06, + "loss": 0.0, + "step": 19825 + }, + { + "epoch": 199.0, + "learning_rate": 5.102278481012659e-06, + "loss": 0.0, + "step": 19850 + }, + { + "epoch": 199.25, + "learning_rate": 5.095949367088609e-06, + "loss": 0.0, + "step": 19875 + }, + { + "epoch": 199.5, + "learning_rate": 5.089620253164557e-06, + "loss": 0.0, + "step": 19900 + }, + { + "epoch": 199.75, + "learning_rate": 5.083291139240507e-06, + "loss": 0.0, + "step": 19925 + }, + { + "epoch": 200.0, + "learning_rate": 5.076962025316456e-06, + "loss": 0.0, + "step": 19950 + }, + { + "epoch": 200.25, + "learning_rate": 5.070632911392406e-06, + "loss": 0.0, + "step": 19975 + }, + { + "epoch": 200.5, + "learning_rate": 5.064303797468355e-06, + "loss": 0.0, + "step": 20000 + }, + { + "epoch": 200.5, + "eval_loss": 0.47855982184410095, + "eval_runtime": 494.9169, + "eval_samples_per_second": 25.746, + "eval_steps_per_second": 0.806, + "eval_wer": 13.089143348444585, + "step": 20000 + }, + { + "epoch": 200.75, + "learning_rate": 5.0579746835443045e-06, + "loss": 0.0, + "step": 20025 + }, + { + "epoch": 201.0, + "learning_rate": 5.051645569620254e-06, + "loss": 0.0, + "step": 20050 + }, + { + "epoch": 201.25, + "learning_rate": 5.045316455696203e-06, + "loss": 0.0, + "step": 20075 + }, + { + "epoch": 201.5, + "learning_rate": 5.038987341772153e-06, + "loss": 0.0, + "step": 20100 + }, + { + "epoch": 201.75, + "learning_rate": 5.032658227848101e-06, + "loss": 0.0, + "step": 20125 + }, + { + "epoch": 202.01, + "learning_rate": 5.026329113924051e-06, + "loss": 0.0, + "step": 20150 + }, + { + "epoch": 202.26, + "learning_rate": 5.02e-06, + "loss": 0.0, + "step": 20175 + }, + { + "epoch": 202.51, + "learning_rate": 5.01367088607595e-06, + "loss": 0.0, + "step": 20200 + }, + { + "epoch": 202.76, + "learning_rate": 5.007341772151899e-06, + "loss": 0.0, + "step": 20225 + }, + { + "epoch": 203.01, + "learning_rate": 5.0010126582278485e-06, + "loss": 0.0, + "step": 20250 + }, + { + "epoch": 203.26, + "learning_rate": 4.994683544303798e-06, + "loss": 0.0, + "step": 20275 + }, + { + "epoch": 203.51, + "learning_rate": 4.988354430379747e-06, + "loss": 0.0, + "step": 20300 + }, + { + "epoch": 203.76, + "learning_rate": 4.982025316455697e-06, + "loss": 0.0, + "step": 20325 + }, + { + "epoch": 204.01, + "learning_rate": 4.975696202531645e-06, + "loss": 0.0, + "step": 20350 + }, + { + "epoch": 204.26, + "learning_rate": 4.969367088607595e-06, + "loss": 0.0, + "step": 20375 + }, + { + "epoch": 204.51, + "learning_rate": 4.963037974683545e-06, + "loss": 0.0, + "step": 20400 + }, + { + "epoch": 204.76, + "learning_rate": 4.9567088607594945e-06, + "loss": 0.0, + "step": 20425 + }, + { + "epoch": 205.01, + "learning_rate": 4.950632911392406e-06, + "loss": 0.0, + "step": 20450 + }, + { + "epoch": 205.26, + "learning_rate": 4.944556962025317e-06, + "loss": 0.0, + "step": 20475 + }, + { + "epoch": 205.51, + "learning_rate": 4.938227848101266e-06, + "loss": 0.0, + "step": 20500 + }, + { + "epoch": 205.76, + "learning_rate": 4.931898734177215e-06, + "loss": 0.0, + "step": 20525 + }, + { + "epoch": 206.02, + "learning_rate": 4.925569620253165e-06, + "loss": 0.0, + "step": 20550 + }, + { + "epoch": 206.27, + "learning_rate": 4.919240506329114e-06, + "loss": 0.0, + "step": 20575 + }, + { + "epoch": 206.52, + "learning_rate": 4.912911392405064e-06, + "loss": 0.0, + "step": 20600 + }, + { + "epoch": 206.77, + "learning_rate": 4.906582278481013e-06, + "loss": 0.0025, + "step": 20625 + }, + { + "epoch": 207.02, + "learning_rate": 4.9002531645569625e-06, + "loss": 0.0077, + "step": 20650 + }, + { + "epoch": 207.27, + "learning_rate": 4.894177215189874e-06, + "loss": 0.006, + "step": 20675 + }, + { + "epoch": 207.52, + "learning_rate": 4.8878481012658226e-06, + "loss": 0.0041, + "step": 20700 + }, + { + "epoch": 207.77, + "learning_rate": 4.881518987341772e-06, + "loss": 0.0031, + "step": 20725 + }, + { + "epoch": 208.02, + "learning_rate": 4.8751898734177214e-06, + "loss": 0.0029, + "step": 20750 + }, + { + "epoch": 208.27, + "learning_rate": 4.868860759493672e-06, + "loss": 0.0012, + "step": 20775 + }, + { + "epoch": 208.52, + "learning_rate": 4.862531645569621e-06, + "loss": 0.0009, + "step": 20800 + }, + { + "epoch": 208.77, + "learning_rate": 4.8562025316455706e-06, + "loss": 0.001, + "step": 20825 + }, + { + "epoch": 209.02, + "learning_rate": 4.849873417721519e-06, + "loss": 0.0008, + "step": 20850 + }, + { + "epoch": 209.27, + "learning_rate": 4.8435443037974686e-06, + "loss": 0.0002, + "step": 20875 + }, + { + "epoch": 209.52, + "learning_rate": 4.837215189873418e-06, + "loss": 0.0002, + "step": 20900 + }, + { + "epoch": 209.77, + "learning_rate": 4.830886075949367e-06, + "loss": 0.0002, + "step": 20925 + }, + { + "epoch": 210.03, + "learning_rate": 4.824556962025317e-06, + "loss": 0.0001, + "step": 20950 + }, + { + "epoch": 210.28, + "learning_rate": 4.818227848101266e-06, + "loss": 0.0001, + "step": 20975 + }, + { + "epoch": 210.53, + "learning_rate": 4.811898734177216e-06, + "loss": 0.0001, + "step": 21000 + }, + { + "epoch": 210.53, + "eval_loss": 0.3784871995449066, + "eval_runtime": 494.6044, + "eval_samples_per_second": 25.762, + "eval_steps_per_second": 0.807, + "eval_wer": 12.73741597623886, + "step": 21000 + }, + { + "epoch": 210.78, + "learning_rate": 4.805569620253165e-06, + "loss": 0.0001, + "step": 21025 + }, + { + "epoch": 211.03, + "learning_rate": 4.7992405063291146e-06, + "loss": 0.0, + "step": 21050 + }, + { + "epoch": 211.28, + "learning_rate": 4.792911392405063e-06, + "loss": 0.0, + "step": 21075 + }, + { + "epoch": 211.53, + "learning_rate": 4.7865822784810126e-06, + "loss": 0.0, + "step": 21100 + }, + { + "epoch": 211.78, + "learning_rate": 4.780253164556962e-06, + "loss": 0.0, + "step": 21125 + }, + { + "epoch": 212.03, + "learning_rate": 4.7739240506329114e-06, + "loss": 0.0, + "step": 21150 + }, + { + "epoch": 212.28, + "learning_rate": 4.767594936708862e-06, + "loss": 0.0, + "step": 21175 + }, + { + "epoch": 212.53, + "learning_rate": 4.761265822784811e-06, + "loss": 0.0, + "step": 21200 + }, + { + "epoch": 212.78, + "learning_rate": 4.75493670886076e-06, + "loss": 0.0, + "step": 21225 + }, + { + "epoch": 213.03, + "learning_rate": 4.748607594936709e-06, + "loss": 0.0, + "step": 21250 + }, + { + "epoch": 213.28, + "learning_rate": 4.7422784810126586e-06, + "loss": 0.0, + "step": 21275 + }, + { + "epoch": 213.53, + "learning_rate": 4.735949367088608e-06, + "loss": 0.0, + "step": 21300 + }, + { + "epoch": 213.78, + "learning_rate": 4.729620253164557e-06, + "loss": 0.0, + "step": 21325 + }, + { + "epoch": 214.04, + "learning_rate": 4.723291139240507e-06, + "loss": 0.0, + "step": 21350 + }, + { + "epoch": 214.29, + "learning_rate": 4.716962025316456e-06, + "loss": 0.0, + "step": 21375 + }, + { + "epoch": 214.54, + "learning_rate": 4.710632911392406e-06, + "loss": 0.0, + "step": 21400 + }, + { + "epoch": 214.79, + "learning_rate": 4.704303797468355e-06, + "loss": 0.0, + "step": 21425 + }, + { + "epoch": 215.04, + "learning_rate": 4.697974683544304e-06, + "loss": 0.0, + "step": 21450 + }, + { + "epoch": 215.29, + "learning_rate": 4.691645569620253e-06, + "loss": 0.0, + "step": 21475 + }, + { + "epoch": 215.54, + "learning_rate": 4.6853164556962026e-06, + "loss": 0.0, + "step": 21500 + }, + { + "epoch": 215.79, + "learning_rate": 4.678987341772152e-06, + "loss": 0.0, + "step": 21525 + }, + { + "epoch": 216.04, + "learning_rate": 4.672658227848101e-06, + "loss": 0.0, + "step": 21550 + }, + { + "epoch": 216.29, + "learning_rate": 4.666329113924051e-06, + "loss": 0.0, + "step": 21575 + }, + { + "epoch": 216.54, + "learning_rate": 4.66e-06, + "loss": 0.0, + "step": 21600 + }, + { + "epoch": 216.79, + "learning_rate": 4.65367088607595e-06, + "loss": 0.0, + "step": 21625 + }, + { + "epoch": 217.04, + "learning_rate": 4.647341772151899e-06, + "loss": 0.0, + "step": 21650 + }, + { + "epoch": 217.29, + "learning_rate": 4.6410126582278486e-06, + "loss": 0.0, + "step": 21675 + }, + { + "epoch": 217.54, + "learning_rate": 4.634683544303798e-06, + "loss": 0.0, + "step": 21700 + }, + { + "epoch": 217.79, + "learning_rate": 4.628354430379747e-06, + "loss": 0.0, + "step": 21725 + }, + { + "epoch": 218.05, + "learning_rate": 4.622025316455697e-06, + "loss": 0.0, + "step": 21750 + }, + { + "epoch": 218.3, + "learning_rate": 4.615696202531646e-06, + "loss": 0.0, + "step": 21775 + }, + { + "epoch": 218.55, + "learning_rate": 4.609367088607596e-06, + "loss": 0.0, + "step": 21800 + }, + { + "epoch": 218.8, + "learning_rate": 4.603037974683544e-06, + "loss": 0.0, + "step": 21825 + }, + { + "epoch": 219.05, + "learning_rate": 4.596708860759494e-06, + "loss": 0.0, + "step": 21850 + }, + { + "epoch": 219.3, + "learning_rate": 4.590379746835443e-06, + "loss": 0.0, + "step": 21875 + }, + { + "epoch": 219.55, + "learning_rate": 4.5840506329113926e-06, + "loss": 0.0, + "step": 21900 + }, + { + "epoch": 219.8, + "learning_rate": 4.577721518987342e-06, + "loss": 0.0, + "step": 21925 + }, + { + "epoch": 220.05, + "learning_rate": 4.571392405063291e-06, + "loss": 0.0, + "step": 21950 + }, + { + "epoch": 220.3, + "learning_rate": 4.565063291139241e-06, + "loss": 0.0, + "step": 21975 + }, + { + "epoch": 220.55, + "learning_rate": 4.55873417721519e-06, + "loss": 0.0, + "step": 22000 + }, + { + "epoch": 220.55, + "eval_loss": 0.4097413122653961, + "eval_runtime": 494.7302, + "eval_samples_per_second": 25.755, + "eval_steps_per_second": 0.807, + "eval_wer": 12.81655463498515, + "step": 22000 + }, + { + "epoch": 220.8, + "learning_rate": 4.55240506329114e-06, + "loss": 0.0, + "step": 22025 + }, + { + "epoch": 221.05, + "learning_rate": 4.546075949367089e-06, + "loss": 0.0, + "step": 22050 + }, + { + "epoch": 221.3, + "learning_rate": 4.5397468354430386e-06, + "loss": 0.0, + "step": 22075 + }, + { + "epoch": 221.55, + "learning_rate": 4.533417721518988e-06, + "loss": 0.0, + "step": 22100 + }, + { + "epoch": 221.8, + "learning_rate": 4.527088607594937e-06, + "loss": 0.0, + "step": 22125 + }, + { + "epoch": 222.06, + "learning_rate": 4.520759493670887e-06, + "loss": 0.0, + "step": 22150 + }, + { + "epoch": 222.31, + "learning_rate": 4.514430379746836e-06, + "loss": 0.0, + "step": 22175 + }, + { + "epoch": 222.56, + "learning_rate": 4.508101265822785e-06, + "loss": 0.0, + "step": 22200 + }, + { + "epoch": 222.81, + "learning_rate": 4.501772151898734e-06, + "loss": 0.0, + "step": 22225 + }, + { + "epoch": 223.06, + "learning_rate": 4.495443037974684e-06, + "loss": 0.0, + "step": 22250 + }, + { + "epoch": 223.31, + "learning_rate": 4.489113924050633e-06, + "loss": 0.0, + "step": 22275 + }, + { + "epoch": 223.56, + "learning_rate": 4.4827848101265826e-06, + "loss": 0.0, + "step": 22300 + }, + { + "epoch": 223.81, + "learning_rate": 4.476455696202532e-06, + "loss": 0.0, + "step": 22325 + }, + { + "epoch": 224.06, + "learning_rate": 4.470126582278481e-06, + "loss": 0.0, + "step": 22350 + }, + { + "epoch": 224.31, + "learning_rate": 4.463797468354431e-06, + "loss": 0.0, + "step": 22375 + }, + { + "epoch": 224.56, + "learning_rate": 4.457468354430379e-06, + "loss": 0.0, + "step": 22400 + }, + { + "epoch": 224.81, + "learning_rate": 4.45113924050633e-06, + "loss": 0.0, + "step": 22425 + }, + { + "epoch": 225.06, + "learning_rate": 4.444810126582279e-06, + "loss": 0.0, + "step": 22450 + }, + { + "epoch": 225.31, + "learning_rate": 4.4384810126582286e-06, + "loss": 0.0, + "step": 22475 + }, + { + "epoch": 225.56, + "learning_rate": 4.432151898734178e-06, + "loss": 0.0, + "step": 22500 + }, + { + "epoch": 225.81, + "learning_rate": 4.425822784810127e-06, + "loss": 0.0, + "step": 22525 + }, + { + "epoch": 226.07, + "learning_rate": 4.419493670886076e-06, + "loss": 0.0, + "step": 22550 + }, + { + "epoch": 226.32, + "learning_rate": 4.413164556962025e-06, + "loss": 0.0, + "step": 22575 + }, + { + "epoch": 226.57, + "learning_rate": 4.406835443037975e-06, + "loss": 0.0, + "step": 22600 + }, + { + "epoch": 226.82, + "learning_rate": 4.400506329113924e-06, + "loss": 0.0, + "step": 22625 + }, + { + "epoch": 227.07, + "learning_rate": 4.394177215189874e-06, + "loss": 0.0, + "step": 22650 + }, + { + "epoch": 227.32, + "learning_rate": 4.387848101265823e-06, + "loss": 0.0, + "step": 22675 + }, + { + "epoch": 227.57, + "learning_rate": 4.3815189873417726e-06, + "loss": 0.0, + "step": 22700 + }, + { + "epoch": 227.82, + "learning_rate": 4.375189873417722e-06, + "loss": 0.0, + "step": 22725 + }, + { + "epoch": 228.07, + "learning_rate": 4.368860759493671e-06, + "loss": 0.0, + "step": 22750 + }, + { + "epoch": 228.32, + "learning_rate": 4.36253164556962e-06, + "loss": 0.0, + "step": 22775 + }, + { + "epoch": 228.57, + "learning_rate": 4.356202531645569e-06, + "loss": 0.0, + "step": 22800 + }, + { + "epoch": 228.82, + "learning_rate": 4.34987341772152e-06, + "loss": 0.0, + "step": 22825 + }, + { + "epoch": 229.07, + "learning_rate": 4.343544303797469e-06, + "loss": 0.0, + "step": 22850 + }, + { + "epoch": 229.32, + "learning_rate": 4.3372151898734185e-06, + "loss": 0.0, + "step": 22875 + }, + { + "epoch": 229.57, + "learning_rate": 4.330886075949368e-06, + "loss": 0.0, + "step": 22900 + }, + { + "epoch": 229.82, + "learning_rate": 4.3245569620253166e-06, + "loss": 0.0, + "step": 22925 + }, + { + "epoch": 230.08, + "learning_rate": 4.318227848101266e-06, + "loss": 0.0, + "step": 22950 + }, + { + "epoch": 230.33, + "learning_rate": 4.311898734177215e-06, + "loss": 0.0, + "step": 22975 + }, + { + "epoch": 230.58, + "learning_rate": 4.305569620253165e-06, + "loss": 0.0, + "step": 23000 + }, + { + "epoch": 230.58, + "eval_loss": 0.42355358600616455, + "eval_runtime": 494.9457, + "eval_samples_per_second": 25.744, + "eval_steps_per_second": 0.806, + "eval_wer": 12.8175316554635, + "step": 23000 + }, + { + "epoch": 230.83, + "learning_rate": 4.299240506329114e-06, + "loss": 0.0, + "step": 23025 + }, + { + "epoch": 231.08, + "learning_rate": 4.292911392405064e-06, + "loss": 0.0, + "step": 23050 + }, + { + "epoch": 231.33, + "learning_rate": 4.286582278481013e-06, + "loss": 0.0, + "step": 23075 + }, + { + "epoch": 231.58, + "learning_rate": 4.2802531645569625e-06, + "loss": 0.0, + "step": 23100 + }, + { + "epoch": 231.83, + "learning_rate": 4.273924050632912e-06, + "loss": 0.0, + "step": 23125 + }, + { + "epoch": 232.08, + "learning_rate": 4.2675949367088606e-06, + "loss": 0.0, + "step": 23150 + }, + { + "epoch": 232.33, + "learning_rate": 4.26126582278481e-06, + "loss": 0.0, + "step": 23175 + }, + { + "epoch": 232.58, + "learning_rate": 4.254936708860759e-06, + "loss": 0.0, + "step": 23200 + }, + { + "epoch": 232.83, + "learning_rate": 4.248607594936709e-06, + "loss": 0.0, + "step": 23225 + }, + { + "epoch": 233.08, + "learning_rate": 4.242278481012659e-06, + "loss": 0.0, + "step": 23250 + }, + { + "epoch": 233.33, + "learning_rate": 4.2359493670886085e-06, + "loss": 0.0, + "step": 23275 + }, + { + "epoch": 233.58, + "learning_rate": 4.229620253164557e-06, + "loss": 0.0, + "step": 23300 + }, + { + "epoch": 233.83, + "learning_rate": 4.2232911392405066e-06, + "loss": 0.0, + "step": 23325 + }, + { + "epoch": 234.09, + "learning_rate": 4.216962025316456e-06, + "loss": 0.0, + "step": 23350 + }, + { + "epoch": 234.34, + "learning_rate": 4.210632911392405e-06, + "loss": 0.0, + "step": 23375 + }, + { + "epoch": 234.59, + "learning_rate": 4.204303797468355e-06, + "loss": 0.0, + "step": 23400 + }, + { + "epoch": 234.84, + "learning_rate": 4.197974683544304e-06, + "loss": 0.0, + "step": 23425 + }, + { + "epoch": 235.09, + "learning_rate": 4.191645569620254e-06, + "loss": 0.0, + "step": 23450 + }, + { + "epoch": 235.34, + "learning_rate": 4.185316455696203e-06, + "loss": 0.0, + "step": 23475 + }, + { + "epoch": 235.59, + "learning_rate": 4.1789873417721525e-06, + "loss": 0.0, + "step": 23500 + }, + { + "epoch": 235.84, + "learning_rate": 4.172658227848101e-06, + "loss": 0.0, + "step": 23525 + }, + { + "epoch": 236.09, + "learning_rate": 4.1663291139240506e-06, + "loss": 0.0, + "step": 23550 + }, + { + "epoch": 236.34, + "learning_rate": 4.16e-06, + "loss": 0.0, + "step": 23575 + }, + { + "epoch": 236.59, + "learning_rate": 4.153670886075949e-06, + "loss": 0.0, + "step": 23600 + }, + { + "epoch": 236.84, + "learning_rate": 4.147341772151899e-06, + "loss": 0.0, + "step": 23625 + }, + { + "epoch": 237.09, + "learning_rate": 4.141012658227849e-06, + "loss": 0.0, + "step": 23650 + }, + { + "epoch": 237.34, + "learning_rate": 4.134683544303798e-06, + "loss": 0.0, + "step": 23675 + }, + { + "epoch": 237.59, + "learning_rate": 4.128354430379747e-06, + "loss": 0.0, + "step": 23700 + }, + { + "epoch": 237.84, + "learning_rate": 4.1220253164556965e-06, + "loss": 0.0, + "step": 23725 + }, + { + "epoch": 238.1, + "learning_rate": 4.115696202531646e-06, + "loss": 0.0, + "step": 23750 + }, + { + "epoch": 238.35, + "learning_rate": 4.109367088607595e-06, + "loss": 0.0, + "step": 23775 + }, + { + "epoch": 238.6, + "learning_rate": 4.103037974683545e-06, + "loss": 0.0, + "step": 23800 + }, + { + "epoch": 238.85, + "learning_rate": 4.096708860759494e-06, + "loss": 0.0, + "step": 23825 + }, + { + "epoch": 239.1, + "learning_rate": 4.090379746835444e-06, + "loss": 0.0, + "step": 23850 + }, + { + "epoch": 239.35, + "learning_rate": 4.084050632911393e-06, + "loss": 0.0, + "step": 23875 + }, + { + "epoch": 239.6, + "learning_rate": 4.077721518987342e-06, + "loss": 0.0, + "step": 23900 + }, + { + "epoch": 239.85, + "learning_rate": 4.071392405063291e-06, + "loss": 0.0, + "step": 23925 + }, + { + "epoch": 240.1, + "learning_rate": 4.0650632911392405e-06, + "loss": 0.0, + "step": 23950 + }, + { + "epoch": 240.35, + "learning_rate": 4.05873417721519e-06, + "loss": 0.0, + "step": 23975 + }, + { + "epoch": 240.6, + "learning_rate": 4.052405063291139e-06, + "loss": 0.0, + "step": 24000 + }, + { + "epoch": 240.6, + "eval_loss": 0.43396899104118347, + "eval_runtime": 494.1294, + "eval_samples_per_second": 25.787, + "eval_steps_per_second": 0.807, + "eval_wer": 12.80385336876661, + "step": 24000 + }, + { + "epoch": 240.85, + "learning_rate": 4.046075949367089e-06, + "loss": 0.0, + "step": 24025 + }, + { + "epoch": 241.1, + "learning_rate": 4.039746835443038e-06, + "loss": 0.0, + "step": 24050 + }, + { + "epoch": 241.35, + "learning_rate": 4.033417721518988e-06, + "loss": 0.0, + "step": 24075 + }, + { + "epoch": 241.6, + "learning_rate": 4.027088607594937e-06, + "loss": 0.0, + "step": 24100 + }, + { + "epoch": 241.85, + "learning_rate": 4.0207594936708865e-06, + "loss": 0.0, + "step": 24125 + }, + { + "epoch": 242.11, + "learning_rate": 4.014430379746836e-06, + "loss": 0.0, + "step": 24150 + }, + { + "epoch": 242.36, + "learning_rate": 4.008101265822785e-06, + "loss": 0.0, + "step": 24175 + }, + { + "epoch": 242.61, + "learning_rate": 4.001772151898735e-06, + "loss": 0.0, + "step": 24200 + }, + { + "epoch": 242.86, + "learning_rate": 3.995443037974684e-06, + "loss": 0.0, + "step": 24225 + }, + { + "epoch": 243.11, + "learning_rate": 3.989113924050633e-06, + "loss": 0.0, + "step": 24250 + }, + { + "epoch": 243.36, + "learning_rate": 3.982784810126582e-06, + "loss": 0.0, + "step": 24275 + }, + { + "epoch": 243.61, + "learning_rate": 3.976455696202532e-06, + "loss": 0.0, + "step": 24300 + }, + { + "epoch": 243.86, + "learning_rate": 3.970126582278481e-06, + "loss": 0.0, + "step": 24325 + }, + { + "epoch": 244.11, + "learning_rate": 3.9637974683544305e-06, + "loss": 0.0, + "step": 24350 + }, + { + "epoch": 244.36, + "learning_rate": 3.95746835443038e-06, + "loss": 0.0, + "step": 24375 + }, + { + "epoch": 244.61, + "learning_rate": 3.951139240506329e-06, + "loss": 0.0, + "step": 24400 + }, + { + "epoch": 244.86, + "learning_rate": 3.944810126582279e-06, + "loss": 0.0, + "step": 24425 + }, + { + "epoch": 245.11, + "learning_rate": 3.938481012658228e-06, + "loss": 0.0, + "step": 24450 + }, + { + "epoch": 245.36, + "learning_rate": 3.932151898734178e-06, + "loss": 0.0, + "step": 24475 + }, + { + "epoch": 245.61, + "learning_rate": 3.925822784810127e-06, + "loss": 0.0, + "step": 24500 + }, + { + "epoch": 245.86, + "learning_rate": 3.9194936708860765e-06, + "loss": 0.0, + "step": 24525 + }, + { + "epoch": 246.12, + "learning_rate": 3.913164556962026e-06, + "loss": 0.0, + "step": 24550 + }, + { + "epoch": 246.37, + "learning_rate": 3.906835443037975e-06, + "loss": 0.0, + "step": 24575 + }, + { + "epoch": 246.62, + "learning_rate": 3.900506329113925e-06, + "loss": 0.0, + "step": 24600 + }, + { + "epoch": 246.87, + "learning_rate": 3.894177215189873e-06, + "loss": 0.0, + "step": 24625 + }, + { + "epoch": 247.12, + "learning_rate": 3.887848101265823e-06, + "loss": 0.0, + "step": 24650 + }, + { + "epoch": 247.37, + "learning_rate": 3.881518987341772e-06, + "loss": 0.0, + "step": 24675 + }, + { + "epoch": 247.62, + "learning_rate": 3.875189873417722e-06, + "loss": 0.0, + "step": 24700 + }, + { + "epoch": 247.87, + "learning_rate": 3.868860759493671e-06, + "loss": 0.0, + "step": 24725 + }, + { + "epoch": 248.12, + "learning_rate": 3.8625316455696205e-06, + "loss": 0.0, + "step": 24750 + }, + { + "epoch": 248.37, + "learning_rate": 3.85620253164557e-06, + "loss": 0.0, + "step": 24775 + }, + { + "epoch": 248.62, + "learning_rate": 3.849873417721519e-06, + "loss": 0.0, + "step": 24800 + }, + { + "epoch": 248.87, + "learning_rate": 3.843544303797469e-06, + "loss": 0.0, + "step": 24825 + }, + { + "epoch": 249.12, + "learning_rate": 3.837215189873417e-06, + "loss": 0.0, + "step": 24850 + }, + { + "epoch": 249.37, + "learning_rate": 3.830886075949368e-06, + "loss": 0.0, + "step": 24875 + }, + { + "epoch": 249.62, + "learning_rate": 3.824556962025317e-06, + "loss": 0.0, + "step": 24900 + }, + { + "epoch": 249.87, + "learning_rate": 3.8182278481012665e-06, + "loss": 0.0, + "step": 24925 + }, + { + "epoch": 250.13, + "learning_rate": 3.8118987341772155e-06, + "loss": 0.0, + "step": 24950 + }, + { + "epoch": 250.38, + "learning_rate": 3.805569620253165e-06, + "loss": 0.0, + "step": 24975 + }, + { + "epoch": 250.63, + "learning_rate": 3.7992405063291144e-06, + "loss": 0.0, + "step": 25000 + }, + { + "epoch": 250.63, + "eval_loss": 0.4431443512439728, + "eval_runtime": 494.1601, + "eval_samples_per_second": 25.785, + "eval_steps_per_second": 0.807, + "eval_wer": 12.8155776145068, + "step": 25000 + }, + { + "epoch": 250.88, + "learning_rate": 3.792911392405064e-06, + "loss": 0.0, + "step": 25025 + }, + { + "epoch": 251.13, + "learning_rate": 3.786582278481013e-06, + "loss": 0.0, + "step": 25050 + }, + { + "epoch": 251.38, + "learning_rate": 3.7802531645569623e-06, + "loss": 0.0, + "step": 25075 + }, + { + "epoch": 251.63, + "learning_rate": 3.7739240506329117e-06, + "loss": 0.0, + "step": 25100 + }, + { + "epoch": 251.88, + "learning_rate": 3.767594936708861e-06, + "loss": 0.0, + "step": 25125 + }, + { + "epoch": 252.13, + "learning_rate": 3.7612658227848105e-06, + "loss": 0.0, + "step": 25150 + }, + { + "epoch": 252.38, + "learning_rate": 3.7549367088607595e-06, + "loss": 0.0, + "step": 25175 + }, + { + "epoch": 252.63, + "learning_rate": 3.748607594936709e-06, + "loss": 0.0, + "step": 25200 + }, + { + "epoch": 252.88, + "learning_rate": 3.7422784810126584e-06, + "loss": 0.0, + "step": 25225 + }, + { + "epoch": 253.13, + "learning_rate": 3.735949367088608e-06, + "loss": 0.0, + "step": 25250 + }, + { + "epoch": 253.38, + "learning_rate": 3.729620253164557e-06, + "loss": 0.0, + "step": 25275 + }, + { + "epoch": 253.63, + "learning_rate": 3.7232911392405067e-06, + "loss": 0.0, + "step": 25300 + }, + { + "epoch": 253.88, + "learning_rate": 3.716962025316456e-06, + "loss": 0.0, + "step": 25325 + }, + { + "epoch": 254.14, + "learning_rate": 3.7106329113924055e-06, + "loss": 0.0, + "step": 25350 + }, + { + "epoch": 254.39, + "learning_rate": 3.704303797468355e-06, + "loss": 0.0, + "step": 25375 + }, + { + "epoch": 254.64, + "learning_rate": 3.6979746835443044e-06, + "loss": 0.0, + "step": 25400 + }, + { + "epoch": 254.89, + "learning_rate": 3.6916455696202534e-06, + "loss": 0.0, + "step": 25425 + }, + { + "epoch": 255.14, + "learning_rate": 3.685316455696203e-06, + "loss": 0.0, + "step": 25450 + }, + { + "epoch": 255.39, + "learning_rate": 3.6789873417721523e-06, + "loss": 0.0, + "step": 25475 + }, + { + "epoch": 255.64, + "learning_rate": 3.6726582278481017e-06, + "loss": 0.0, + "step": 25500 + }, + { + "epoch": 255.89, + "learning_rate": 3.6663291139240507e-06, + "loss": 0.0, + "step": 25525 + }, + { + "epoch": 256.14, + "learning_rate": 3.66e-06, + "loss": 0.0, + "step": 25550 + }, + { + "epoch": 256.39, + "learning_rate": 3.6536708860759495e-06, + "loss": 0.0, + "step": 25575 + }, + { + "epoch": 256.64, + "learning_rate": 3.647341772151899e-06, + "loss": 0.0, + "step": 25600 + }, + { + "epoch": 256.89, + "learning_rate": 3.6410126582278484e-06, + "loss": 0.0, + "step": 25625 + }, + { + "epoch": 257.14, + "learning_rate": 3.6346835443037974e-06, + "loss": 0.0, + "step": 25650 + }, + { + "epoch": 257.39, + "learning_rate": 3.628354430379747e-06, + "loss": 0.0, + "step": 25675 + }, + { + "epoch": 257.64, + "learning_rate": 3.6220253164556967e-06, + "loss": 0.0, + "step": 25700 + }, + { + "epoch": 257.89, + "learning_rate": 3.615696202531646e-06, + "loss": 0.0, + "step": 25725 + }, + { + "epoch": 258.15, + "learning_rate": 3.6093670886075955e-06, + "loss": 0.0, + "step": 25750 + }, + { + "epoch": 258.4, + "learning_rate": 3.603037974683545e-06, + "loss": 0.0, + "step": 25775 + }, + { + "epoch": 258.65, + "learning_rate": 3.596708860759494e-06, + "loss": 0.0, + "step": 25800 + }, + { + "epoch": 258.9, + "learning_rate": 3.5903797468354434e-06, + "loss": 0.0, + "step": 25825 + }, + { + "epoch": 259.15, + "learning_rate": 3.584050632911393e-06, + "loss": 0.0, + "step": 25850 + }, + { + "epoch": 259.4, + "learning_rate": 3.5777215189873422e-06, + "loss": 0.0, + "step": 25875 + }, + { + "epoch": 259.65, + "learning_rate": 3.5713924050632913e-06, + "loss": 0.0, + "step": 25900 + }, + { + "epoch": 259.9, + "learning_rate": 3.5650632911392407e-06, + "loss": 0.0, + "step": 25925 + }, + { + "epoch": 260.15, + "learning_rate": 3.55873417721519e-06, + "loss": 0.0, + "step": 25950 + }, + { + "epoch": 260.4, + "learning_rate": 3.5524050632911395e-06, + "loss": 0.0, + "step": 25975 + }, + { + "epoch": 260.65, + "learning_rate": 3.546075949367089e-06, + "loss": 0.0, + "step": 26000 + }, + { + "epoch": 260.65, + "eval_loss": 0.4517054259777069, + "eval_runtime": 494.2671, + "eval_samples_per_second": 25.78, + "eval_steps_per_second": 0.807, + "eval_wer": 12.805807409723307, + "step": 26000 + }, + { + "epoch": 260.9, + "learning_rate": 3.539746835443038e-06, + "loss": 0.0, + "step": 26025 + }, + { + "epoch": 261.15, + "learning_rate": 3.5334177215189874e-06, + "loss": 0.0, + "step": 26050 + }, + { + "epoch": 261.4, + "learning_rate": 3.527088607594937e-06, + "loss": 0.0, + "step": 26075 + }, + { + "epoch": 261.65, + "learning_rate": 3.5207594936708862e-06, + "loss": 0.0, + "step": 26100 + }, + { + "epoch": 261.9, + "learning_rate": 3.514430379746836e-06, + "loss": 0.0, + "step": 26125 + }, + { + "epoch": 262.16, + "learning_rate": 3.5081012658227855e-06, + "loss": 0.0, + "step": 26150 + }, + { + "epoch": 262.41, + "learning_rate": 3.5017721518987345e-06, + "loss": 0.0, + "step": 26175 + }, + { + "epoch": 262.66, + "learning_rate": 3.495443037974684e-06, + "loss": 0.0, + "step": 26200 + }, + { + "epoch": 262.91, + "learning_rate": 3.4891139240506334e-06, + "loss": 0.0, + "step": 26225 + }, + { + "epoch": 263.16, + "learning_rate": 3.482784810126583e-06, + "loss": 0.0, + "step": 26250 + }, + { + "epoch": 263.41, + "learning_rate": 3.476455696202532e-06, + "loss": 0.0, + "step": 26275 + }, + { + "epoch": 263.66, + "learning_rate": 3.4701265822784812e-06, + "loss": 0.0, + "step": 26300 + }, + { + "epoch": 263.91, + "learning_rate": 3.4637974683544307e-06, + "loss": 0.0, + "step": 26325 + }, + { + "epoch": 264.16, + "learning_rate": 3.45746835443038e-06, + "loss": 0.0, + "step": 26350 + }, + { + "epoch": 264.41, + "learning_rate": 3.451139240506329e-06, + "loss": 0.0, + "step": 26375 + }, + { + "epoch": 264.66, + "learning_rate": 3.4448101265822785e-06, + "loss": 0.0, + "step": 26400 + }, + { + "epoch": 264.91, + "learning_rate": 3.438481012658228e-06, + "loss": 0.0, + "step": 26425 + }, + { + "epoch": 265.16, + "learning_rate": 3.4321518987341774e-06, + "loss": 0.0, + "step": 26450 + }, + { + "epoch": 265.41, + "learning_rate": 3.425822784810127e-06, + "loss": 0.0, + "step": 26475 + }, + { + "epoch": 265.66, + "learning_rate": 3.419493670886076e-06, + "loss": 0.0, + "step": 26500 + }, + { + "epoch": 265.91, + "learning_rate": 3.4131645569620257e-06, + "loss": 0.0, + "step": 26525 + }, + { + "epoch": 266.17, + "learning_rate": 3.406835443037975e-06, + "loss": 0.0, + "step": 26550 + }, + { + "epoch": 266.42, + "learning_rate": 3.4005063291139245e-06, + "loss": 0.0, + "step": 26575 + }, + { + "epoch": 266.67, + "learning_rate": 3.394177215189874e-06, + "loss": 0.0, + "step": 26600 + }, + { + "epoch": 266.92, + "learning_rate": 3.3878481012658234e-06, + "loss": 0.0, + "step": 26625 + }, + { + "epoch": 267.17, + "learning_rate": 3.3815189873417724e-06, + "loss": 0.0, + "step": 26650 + }, + { + "epoch": 267.42, + "learning_rate": 3.375189873417722e-06, + "loss": 0.0, + "step": 26675 + }, + { + "epoch": 267.67, + "learning_rate": 3.3688607594936712e-06, + "loss": 0.0, + "step": 26700 + }, + { + "epoch": 267.92, + "learning_rate": 3.3625316455696207e-06, + "loss": 0.0, + "step": 26725 + }, + { + "epoch": 268.17, + "learning_rate": 3.3562025316455697e-06, + "loss": 0.0, + "step": 26750 + }, + { + "epoch": 268.42, + "learning_rate": 3.349873417721519e-06, + "loss": 0.0, + "step": 26775 + }, + { + "epoch": 268.67, + "learning_rate": 3.3435443037974685e-06, + "loss": 0.0, + "step": 26800 + }, + { + "epoch": 268.92, + "learning_rate": 3.33746835443038e-06, + "loss": 0.0, + "step": 26825 + }, + { + "epoch": 269.17, + "learning_rate": 3.3311392405063294e-06, + "loss": 0.0, + "step": 26850 + }, + { + "epoch": 269.42, + "learning_rate": 3.3248101265822784e-06, + "loss": 0.0, + "step": 26875 + }, + { + "epoch": 269.67, + "learning_rate": 3.318481012658228e-06, + "loss": 0.0, + "step": 26900 + }, + { + "epoch": 269.92, + "learning_rate": 3.3121518987341777e-06, + "loss": 0.0, + "step": 26925 + }, + { + "epoch": 270.18, + "learning_rate": 3.305822784810127e-06, + "loss": 0.0, + "step": 26950 + }, + { + "epoch": 270.43, + "learning_rate": 3.2994936708860765e-06, + "loss": 0.0, + "step": 26975 + }, + { + "epoch": 270.68, + "learning_rate": 3.2931645569620256e-06, + "loss": 0.0, + "step": 27000 + }, + { + "epoch": 270.68, + "eval_loss": 0.4601159989833832, + "eval_runtime": 494.0636, + "eval_samples_per_second": 25.79, + "eval_steps_per_second": 0.808, + "eval_wer": 12.792129123026418, + "step": 27000 + }, + { + "epoch": 270.93, + "learning_rate": 3.286835443037975e-06, + "loss": 0.0, + "step": 27025 + }, + { + "epoch": 271.18, + "learning_rate": 3.2805063291139244e-06, + "loss": 0.0, + "step": 27050 + }, + { + "epoch": 271.43, + "learning_rate": 3.274177215189874e-06, + "loss": 0.0, + "step": 27075 + }, + { + "epoch": 271.68, + "learning_rate": 3.2678481012658233e-06, + "loss": 0.0, + "step": 27100 + }, + { + "epoch": 271.93, + "learning_rate": 3.2615189873417723e-06, + "loss": 0.0, + "step": 27125 + }, + { + "epoch": 272.18, + "learning_rate": 3.2551898734177217e-06, + "loss": 0.0, + "step": 27150 + }, + { + "epoch": 272.43, + "learning_rate": 3.248860759493671e-06, + "loss": 0.0, + "step": 27175 + }, + { + "epoch": 272.68, + "learning_rate": 3.2425316455696205e-06, + "loss": 0.0, + "step": 27200 + }, + { + "epoch": 272.93, + "learning_rate": 3.2362025316455696e-06, + "loss": 0.0, + "step": 27225 + }, + { + "epoch": 273.18, + "learning_rate": 3.229873417721519e-06, + "loss": 0.0, + "step": 27250 + }, + { + "epoch": 273.43, + "learning_rate": 3.2235443037974684e-06, + "loss": 0.0, + "step": 27275 + }, + { + "epoch": 273.68, + "learning_rate": 3.217215189873418e-06, + "loss": 0.0, + "step": 27300 + }, + { + "epoch": 273.93, + "learning_rate": 3.2108860759493673e-06, + "loss": 0.0, + "step": 27325 + }, + { + "epoch": 274.19, + "learning_rate": 3.204556962025317e-06, + "loss": 0.0, + "step": 27350 + }, + { + "epoch": 274.44, + "learning_rate": 3.198227848101266e-06, + "loss": 0.0, + "step": 27375 + }, + { + "epoch": 274.69, + "learning_rate": 3.1918987341772155e-06, + "loss": 0.0, + "step": 27400 + }, + { + "epoch": 274.94, + "learning_rate": 3.185569620253165e-06, + "loss": 0.0, + "step": 27425 + }, + { + "epoch": 275.19, + "learning_rate": 3.1792405063291144e-06, + "loss": 0.0, + "step": 27450 + }, + { + "epoch": 275.44, + "learning_rate": 3.172911392405064e-06, + "loss": 0.0, + "step": 27475 + }, + { + "epoch": 275.69, + "learning_rate": 3.166582278481013e-06, + "loss": 0.0, + "step": 27500 + }, + { + "epoch": 275.94, + "learning_rate": 3.1602531645569623e-06, + "loss": 0.0, + "step": 27525 + }, + { + "epoch": 276.19, + "learning_rate": 3.1539240506329117e-06, + "loss": 0.0, + "step": 27550 + }, + { + "epoch": 276.44, + "learning_rate": 3.147594936708861e-06, + "loss": 0.0, + "step": 27575 + }, + { + "epoch": 276.69, + "learning_rate": 3.14126582278481e-06, + "loss": 0.0, + "step": 27600 + }, + { + "epoch": 276.94, + "learning_rate": 3.1349367088607595e-06, + "loss": 0.0, + "step": 27625 + }, + { + "epoch": 277.19, + "learning_rate": 3.128607594936709e-06, + "loss": 0.0, + "step": 27650 + }, + { + "epoch": 277.44, + "learning_rate": 3.1222784810126584e-06, + "loss": 0.0, + "step": 27675 + }, + { + "epoch": 277.69, + "learning_rate": 3.115949367088608e-06, + "loss": 0.0, + "step": 27700 + }, + { + "epoch": 277.94, + "learning_rate": 3.109620253164557e-06, + "loss": 0.0, + "step": 27725 + }, + { + "epoch": 278.2, + "learning_rate": 3.1032911392405067e-06, + "loss": 0.0, + "step": 27750 + }, + { + "epoch": 278.45, + "learning_rate": 3.096962025316456e-06, + "loss": 0.0, + "step": 27775 + }, + { + "epoch": 278.7, + "learning_rate": 3.0906329113924055e-06, + "loss": 0.0, + "step": 27800 + }, + { + "epoch": 278.95, + "learning_rate": 3.084303797468355e-06, + "loss": 0.0, + "step": 27825 + }, + { + "epoch": 279.2, + "learning_rate": 3.077974683544304e-06, + "loss": 0.0, + "step": 27850 + }, + { + "epoch": 279.45, + "learning_rate": 3.0716455696202534e-06, + "loss": 0.0, + "step": 27875 + }, + { + "epoch": 279.7, + "learning_rate": 3.065316455696203e-06, + "loss": 0.0, + "step": 27900 + }, + { + "epoch": 279.95, + "learning_rate": 3.0589873417721523e-06, + "loss": 0.0, + "step": 27925 + }, + { + "epoch": 280.2, + "learning_rate": 3.0526582278481017e-06, + "loss": 0.0, + "step": 27950 + }, + { + "epoch": 280.45, + "learning_rate": 3.0463291139240507e-06, + "loss": 0.0, + "step": 27975 + }, + { + "epoch": 280.7, + "learning_rate": 3.04e-06, + "loss": 0.0, + "step": 28000 + }, + { + "epoch": 280.7, + "eval_loss": 0.46891674399375916, + "eval_runtime": 494.5419, + "eval_samples_per_second": 25.765, + "eval_steps_per_second": 0.807, + "eval_wer": 12.802876348288262, + "step": 28000 + }, + { + "epoch": 280.95, + "learning_rate": 3.0336708860759495e-06, + "loss": 0.0, + "step": 28025 + }, + { + "epoch": 281.2, + "learning_rate": 3.027341772151899e-06, + "loss": 0.0, + "step": 28050 + }, + { + "epoch": 281.45, + "learning_rate": 3.021012658227848e-06, + "loss": 0.0, + "step": 28075 + }, + { + "epoch": 281.7, + "learning_rate": 3.0146835443037974e-06, + "loss": 0.0, + "step": 28100 + }, + { + "epoch": 281.95, + "learning_rate": 3.008354430379747e-06, + "loss": 0.0, + "step": 28125 + }, + { + "epoch": 282.21, + "learning_rate": 3.0020253164556967e-06, + "loss": 0.0, + "step": 28150 + }, + { + "epoch": 282.46, + "learning_rate": 2.995696202531646e-06, + "loss": 0.0, + "step": 28175 + }, + { + "epoch": 282.71, + "learning_rate": 2.9893670886075955e-06, + "loss": 0.0, + "step": 28200 + }, + { + "epoch": 282.96, + "learning_rate": 2.9830379746835445e-06, + "loss": 0.0, + "step": 28225 + }, + { + "epoch": 283.21, + "learning_rate": 2.976708860759494e-06, + "loss": 0.0, + "step": 28250 + }, + { + "epoch": 283.46, + "learning_rate": 2.9703797468354434e-06, + "loss": 0.0, + "step": 28275 + }, + { + "epoch": 283.71, + "learning_rate": 2.964050632911393e-06, + "loss": 0.0, + "step": 28300 + }, + { + "epoch": 283.96, + "learning_rate": 2.9577215189873423e-06, + "loss": 0.0, + "step": 28325 + }, + { + "epoch": 284.21, + "learning_rate": 2.9513924050632913e-06, + "loss": 0.0, + "step": 28350 + }, + { + "epoch": 284.46, + "learning_rate": 2.9450632911392407e-06, + "loss": 0.0, + "step": 28375 + }, + { + "epoch": 284.71, + "learning_rate": 2.93873417721519e-06, + "loss": 0.0, + "step": 28400 + }, + { + "epoch": 284.96, + "learning_rate": 2.9324050632911395e-06, + "loss": 0.0, + "step": 28425 + }, + { + "epoch": 285.21, + "learning_rate": 2.9260759493670885e-06, + "loss": 0.0, + "step": 28450 + }, + { + "epoch": 285.46, + "learning_rate": 2.919746835443038e-06, + "loss": 0.0, + "step": 28475 + }, + { + "epoch": 285.71, + "learning_rate": 2.9134177215189874e-06, + "loss": 0.0, + "step": 28500 + }, + { + "epoch": 285.96, + "learning_rate": 2.907088607594937e-06, + "loss": 0.0, + "step": 28525 + }, + { + "epoch": 286.22, + "learning_rate": 2.9007594936708863e-06, + "loss": 0.0, + "step": 28550 + }, + { + "epoch": 286.47, + "learning_rate": 2.894430379746836e-06, + "loss": 0.0, + "step": 28575 + }, + { + "epoch": 286.72, + "learning_rate": 2.888101265822785e-06, + "loss": 0.0, + "step": 28600 + }, + { + "epoch": 286.97, + "learning_rate": 2.8817721518987345e-06, + "loss": 0.0, + "step": 28625 + }, + { + "epoch": 287.22, + "learning_rate": 2.875443037974684e-06, + "loss": 0.0, + "step": 28650 + }, + { + "epoch": 287.47, + "learning_rate": 2.8691139240506334e-06, + "loss": 0.0, + "step": 28675 + }, + { + "epoch": 287.72, + "learning_rate": 2.8627848101265824e-06, + "loss": 0.0, + "step": 28700 + }, + { + "epoch": 287.97, + "learning_rate": 2.856455696202532e-06, + "loss": 0.0, + "step": 28725 + }, + { + "epoch": 288.22, + "learning_rate": 2.8501265822784813e-06, + "loss": 0.0, + "step": 28750 + }, + { + "epoch": 288.47, + "learning_rate": 2.8437974683544307e-06, + "loss": 0.0, + "step": 28775 + }, + { + "epoch": 288.72, + "learning_rate": 2.83746835443038e-06, + "loss": 0.0, + "step": 28800 + }, + { + "epoch": 288.97, + "learning_rate": 2.831139240506329e-06, + "loss": 0.0, + "step": 28825 + }, + { + "epoch": 289.22, + "learning_rate": 2.8248101265822785e-06, + "loss": 0.0, + "step": 28850 + }, + { + "epoch": 289.47, + "learning_rate": 2.81873417721519e-06, + "loss": 0.0, + "step": 28875 + }, + { + "epoch": 289.72, + "learning_rate": 2.8124050632911394e-06, + "loss": 0.0, + "step": 28900 + }, + { + "epoch": 289.97, + "learning_rate": 2.8060759493670884e-06, + "loss": 0.0, + "step": 28925 + }, + { + "epoch": 290.23, + "learning_rate": 2.799746835443038e-06, + "loss": 0.0, + "step": 28950 + }, + { + "epoch": 290.48, + "learning_rate": 2.7934177215189877e-06, + "loss": 0.0, + "step": 28975 + }, + { + "epoch": 290.73, + "learning_rate": 2.787088607594937e-06, + "loss": 0.0, + "step": 29000 + }, + { + "epoch": 290.73, + "eval_loss": 0.4773997366428375, + "eval_runtime": 494.3805, + "eval_samples_per_second": 25.774, + "eval_steps_per_second": 0.807, + "eval_wer": 12.80385336876661, + "step": 29000 + }, + { + "epoch": 290.98, + "learning_rate": 2.7807594936708866e-06, + "loss": 0.0, + "step": 29025 + }, + { + "epoch": 291.23, + "learning_rate": 2.774430379746836e-06, + "loss": 0.0, + "step": 29050 + }, + { + "epoch": 291.48, + "learning_rate": 2.768101265822785e-06, + "loss": 0.0, + "step": 29075 + }, + { + "epoch": 291.73, + "learning_rate": 2.7617721518987344e-06, + "loss": 0.0, + "step": 29100 + }, + { + "epoch": 291.98, + "learning_rate": 2.755443037974684e-06, + "loss": 0.0, + "step": 29125 + }, + { + "epoch": 292.23, + "learning_rate": 2.7491139240506333e-06, + "loss": 0.0, + "step": 29150 + }, + { + "epoch": 292.48, + "learning_rate": 2.7427848101265823e-06, + "loss": 0.0, + "step": 29175 + }, + { + "epoch": 292.73, + "learning_rate": 2.7364556962025317e-06, + "loss": 0.0, + "step": 29200 + }, + { + "epoch": 292.98, + "learning_rate": 2.730126582278481e-06, + "loss": 0.0, + "step": 29225 + }, + { + "epoch": 293.23, + "learning_rate": 2.7237974683544306e-06, + "loss": 0.0, + "step": 29250 + }, + { + "epoch": 293.48, + "learning_rate": 2.71746835443038e-06, + "loss": 0.0, + "step": 29275 + }, + { + "epoch": 293.73, + "learning_rate": 2.711139240506329e-06, + "loss": 0.0, + "step": 29300 + }, + { + "epoch": 293.98, + "learning_rate": 2.7048101265822784e-06, + "loss": 0.0, + "step": 29325 + }, + { + "epoch": 294.24, + "learning_rate": 2.698481012658228e-06, + "loss": 0.0, + "step": 29350 + }, + { + "epoch": 294.49, + "learning_rate": 2.6921518987341777e-06, + "loss": 0.0, + "step": 29375 + }, + { + "epoch": 294.74, + "learning_rate": 2.685822784810127e-06, + "loss": 0.0, + "step": 29400 + }, + { + "epoch": 294.99, + "learning_rate": 2.6794936708860766e-06, + "loss": 0.0, + "step": 29425 + }, + { + "epoch": 295.24, + "learning_rate": 2.6731645569620256e-06, + "loss": 0.0, + "step": 29450 + }, + { + "epoch": 295.49, + "learning_rate": 2.666835443037975e-06, + "loss": 0.0, + "step": 29475 + }, + { + "epoch": 295.74, + "learning_rate": 2.6605063291139244e-06, + "loss": 0.0, + "step": 29500 + }, + { + "epoch": 295.99, + "learning_rate": 2.654177215189874e-06, + "loss": 0.0, + "step": 29525 + }, + { + "epoch": 296.24, + "learning_rate": 2.647848101265823e-06, + "loss": 0.0, + "step": 29550 + }, + { + "epoch": 296.49, + "learning_rate": 2.6415189873417723e-06, + "loss": 0.0, + "step": 29575 + }, + { + "epoch": 296.74, + "learning_rate": 2.6351898734177217e-06, + "loss": 0.0, + "step": 29600 + }, + { + "epoch": 296.99, + "learning_rate": 2.628860759493671e-06, + "loss": 0.0, + "step": 29625 + }, + { + "epoch": 297.24, + "learning_rate": 2.6225316455696206e-06, + "loss": 0.0, + "step": 29650 + }, + { + "epoch": 297.49, + "learning_rate": 2.6162025316455696e-06, + "loss": 0.0, + "step": 29675 + }, + { + "epoch": 297.74, + "learning_rate": 2.609873417721519e-06, + "loss": 0.0, + "step": 29700 + }, + { + "epoch": 297.99, + "learning_rate": 2.6035443037974684e-06, + "loss": 0.0, + "step": 29725 + }, + { + "epoch": 298.25, + "learning_rate": 2.597215189873418e-06, + "loss": 0.0, + "step": 29750 + }, + { + "epoch": 298.5, + "learning_rate": 2.590886075949367e-06, + "loss": 0.0, + "step": 29775 + }, + { + "epoch": 298.75, + "learning_rate": 2.584556962025317e-06, + "loss": 0.0, + "step": 29800 + }, + { + "epoch": 299.0, + "learning_rate": 2.578227848101266e-06, + "loss": 0.0, + "step": 29825 + }, + { + "epoch": 299.25, + "learning_rate": 2.5718987341772156e-06, + "loss": 0.0, + "step": 29850 + }, + { + "epoch": 299.5, + "learning_rate": 2.565569620253165e-06, + "loss": 0.0, + "step": 29875 + }, + { + "epoch": 299.75, + "learning_rate": 2.5592405063291144e-06, + "loss": 0.0, + "step": 29900 + }, + { + "epoch": 300.0, + "learning_rate": 2.5529113924050634e-06, + "loss": 0.0, + "step": 29925 + }, + { + "epoch": 300.25, + "learning_rate": 2.546582278481013e-06, + "loss": 0.0, + "step": 29950 + }, + { + "epoch": 300.5, + "learning_rate": 2.5402531645569623e-06, + "loss": 0.0, + "step": 29975 + }, + { + "epoch": 300.75, + "learning_rate": 2.5339240506329117e-06, + "loss": 0.0, + "step": 30000 + }, + { + "epoch": 300.75, + "eval_loss": 0.4863458573818207, + "eval_runtime": 493.8881, + "eval_samples_per_second": 25.799, + "eval_steps_per_second": 0.808, + "eval_wer": 12.796037204939816, + "step": 30000 + }, + { + "epoch": 301.0, + "learning_rate": 2.527594936708861e-06, + "loss": 0.0, + "step": 30025 + }, + { + "epoch": 301.25, + "learning_rate": 2.52126582278481e-06, + "loss": 0.0, + "step": 30050 + }, + { + "epoch": 301.5, + "learning_rate": 2.5149367088607596e-06, + "loss": 0.0, + "step": 30075 + }, + { + "epoch": 301.75, + "learning_rate": 2.508607594936709e-06, + "loss": 0.0, + "step": 30100 + }, + { + "epoch": 302.01, + "learning_rate": 2.5022784810126584e-06, + "loss": 0.0, + "step": 30125 + }, + { + "epoch": 302.26, + "learning_rate": 2.495949367088608e-06, + "loss": 0.0, + "step": 30150 + }, + { + "epoch": 302.51, + "learning_rate": 2.4896202531645573e-06, + "loss": 0.0, + "step": 30175 + }, + { + "epoch": 302.76, + "learning_rate": 2.4832911392405067e-06, + "loss": 0.0, + "step": 30200 + }, + { + "epoch": 303.01, + "learning_rate": 2.4769620253164557e-06, + "loss": 0.0, + "step": 30225 + }, + { + "epoch": 303.26, + "learning_rate": 2.470632911392405e-06, + "loss": 0.0, + "step": 30250 + }, + { + "epoch": 303.51, + "learning_rate": 2.4643037974683546e-06, + "loss": 0.0, + "step": 30275 + }, + { + "epoch": 303.76, + "learning_rate": 2.457974683544304e-06, + "loss": 0.0, + "step": 30300 + }, + { + "epoch": 304.01, + "learning_rate": 2.4516455696202534e-06, + "loss": 0.0, + "step": 30325 + }, + { + "epoch": 304.26, + "learning_rate": 2.445316455696203e-06, + "loss": 0.0, + "step": 30350 + }, + { + "epoch": 304.51, + "learning_rate": 2.4389873417721523e-06, + "loss": 0.0, + "step": 30375 + }, + { + "epoch": 304.76, + "learning_rate": 2.4326582278481013e-06, + "loss": 0.0, + "step": 30400 + }, + { + "epoch": 305.01, + "learning_rate": 2.4263291139240507e-06, + "loss": 0.0, + "step": 30425 + }, + { + "epoch": 305.26, + "learning_rate": 2.42e-06, + "loss": 0.0, + "step": 30450 + }, + { + "epoch": 305.51, + "learning_rate": 2.4136708860759496e-06, + "loss": 0.0, + "step": 30475 + }, + { + "epoch": 305.76, + "learning_rate": 2.407341772151899e-06, + "loss": 0.0, + "step": 30500 + }, + { + "epoch": 306.02, + "learning_rate": 2.4010126582278484e-06, + "loss": 0.0, + "step": 30525 + }, + { + "epoch": 306.27, + "learning_rate": 2.394683544303798e-06, + "loss": 0.0, + "step": 30550 + }, + { + "epoch": 306.52, + "learning_rate": 2.3883544303797473e-06, + "loss": 0.0, + "step": 30575 + }, + { + "epoch": 306.77, + "learning_rate": 2.3820253164556963e-06, + "loss": 0.0, + "step": 30600 + }, + { + "epoch": 307.02, + "learning_rate": 2.3756962025316457e-06, + "loss": 0.0, + "step": 30625 + }, + { + "epoch": 307.27, + "learning_rate": 2.369367088607595e-06, + "loss": 0.0, + "step": 30650 + }, + { + "epoch": 307.52, + "learning_rate": 2.3630379746835445e-06, + "loss": 0.0, + "step": 30675 + }, + { + "epoch": 307.77, + "learning_rate": 2.3567088607594936e-06, + "loss": 0.0, + "step": 30700 + }, + { + "epoch": 308.02, + "learning_rate": 2.3503797468354434e-06, + "loss": 0.0, + "step": 30725 + }, + { + "epoch": 308.27, + "learning_rate": 2.344050632911393e-06, + "loss": 0.0, + "step": 30750 + }, + { + "epoch": 308.52, + "learning_rate": 2.337721518987342e-06, + "loss": 0.0, + "step": 30775 + }, + { + "epoch": 308.77, + "learning_rate": 2.3313924050632913e-06, + "loss": 0.0, + "step": 30800 + }, + { + "epoch": 309.02, + "learning_rate": 2.3250632911392407e-06, + "loss": 0.0, + "step": 30825 + }, + { + "epoch": 309.27, + "learning_rate": 2.31873417721519e-06, + "loss": 0.0, + "step": 30850 + }, + { + "epoch": 309.52, + "learning_rate": 2.3124050632911395e-06, + "loss": 0.0, + "step": 30875 + }, + { + "epoch": 309.77, + "learning_rate": 2.3060759493670886e-06, + "loss": 0.0, + "step": 30900 + }, + { + "epoch": 310.03, + "learning_rate": 2.3000000000000004e-06, + "loss": 0.0, + "step": 30925 + }, + { + "epoch": 310.28, + "learning_rate": 2.2936708860759494e-06, + "loss": 0.0, + "step": 30950 + }, + { + "epoch": 310.53, + "learning_rate": 2.287341772151899e-06, + "loss": 0.0, + "step": 30975 + }, + { + "epoch": 310.78, + "learning_rate": 2.2810126582278483e-06, + "loss": 0.0, + "step": 31000 + }, + { + "epoch": 310.78, + "eval_loss": 0.4948866367340088, + "eval_runtime": 493.9318, + "eval_samples_per_second": 25.797, + "eval_steps_per_second": 0.808, + "eval_wer": 12.79115210254807, + "step": 31000 + }, + { + "epoch": 311.03, + "learning_rate": 2.2746835443037977e-06, + "loss": 0.0, + "step": 31025 + }, + { + "epoch": 311.28, + "learning_rate": 2.268354430379747e-06, + "loss": 0.0, + "step": 31050 + }, + { + "epoch": 311.53, + "learning_rate": 2.262025316455696e-06, + "loss": 0.0, + "step": 31075 + }, + { + "epoch": 311.78, + "learning_rate": 2.2556962025316456e-06, + "loss": 0.0, + "step": 31100 + }, + { + "epoch": 312.03, + "learning_rate": 2.2493670886075954e-06, + "loss": 0.0, + "step": 31125 + }, + { + "epoch": 312.28, + "learning_rate": 2.2430379746835444e-06, + "loss": 0.0, + "step": 31150 + }, + { + "epoch": 312.53, + "learning_rate": 2.236708860759494e-06, + "loss": 0.0, + "step": 31175 + }, + { + "epoch": 312.78, + "learning_rate": 2.2303797468354433e-06, + "loss": 0.0, + "step": 31200 + }, + { + "epoch": 313.03, + "learning_rate": 2.2240506329113927e-06, + "loss": 0.0, + "step": 31225 + }, + { + "epoch": 313.28, + "learning_rate": 2.2177215189873417e-06, + "loss": 0.0, + "step": 31250 + }, + { + "epoch": 313.53, + "learning_rate": 2.211392405063291e-06, + "loss": 0.0, + "step": 31275 + }, + { + "epoch": 313.78, + "learning_rate": 2.2050632911392406e-06, + "loss": 0.0, + "step": 31300 + }, + { + "epoch": 314.04, + "learning_rate": 2.19873417721519e-06, + "loss": 0.0, + "step": 31325 + }, + { + "epoch": 314.29, + "learning_rate": 2.1924050632911394e-06, + "loss": 0.0, + "step": 31350 + }, + { + "epoch": 314.54, + "learning_rate": 2.186075949367089e-06, + "loss": 0.0, + "step": 31375 + }, + { + "epoch": 314.79, + "learning_rate": 2.1797468354430383e-06, + "loss": 0.0, + "step": 31400 + }, + { + "epoch": 315.04, + "learning_rate": 2.1734177215189877e-06, + "loss": 0.0, + "step": 31425 + }, + { + "epoch": 315.29, + "learning_rate": 2.1670886075949367e-06, + "loss": 0.0, + "step": 31450 + }, + { + "epoch": 315.54, + "learning_rate": 2.160759493670886e-06, + "loss": 0.0, + "step": 31475 + }, + { + "epoch": 315.79, + "learning_rate": 2.1544303797468356e-06, + "loss": 0.0, + "step": 31500 + }, + { + "epoch": 316.04, + "learning_rate": 2.148101265822785e-06, + "loss": 0.0, + "step": 31525 + }, + { + "epoch": 316.29, + "learning_rate": 2.1417721518987344e-06, + "loss": 0.0, + "step": 31550 + }, + { + "epoch": 316.54, + "learning_rate": 2.135443037974684e-06, + "loss": 0.0, + "step": 31575 + }, + { + "epoch": 316.79, + "learning_rate": 2.1291139240506333e-06, + "loss": 0.0, + "step": 31600 + }, + { + "epoch": 317.04, + "learning_rate": 2.1227848101265823e-06, + "loss": 0.0, + "step": 31625 + }, + { + "epoch": 317.29, + "learning_rate": 2.1164556962025317e-06, + "loss": 0.0, + "step": 31650 + }, + { + "epoch": 317.54, + "learning_rate": 2.110126582278481e-06, + "loss": 0.0, + "step": 31675 + }, + { + "epoch": 317.79, + "learning_rate": 2.1037974683544306e-06, + "loss": 0.0, + "step": 31700 + }, + { + "epoch": 318.05, + "learning_rate": 2.09746835443038e-06, + "loss": 0.0, + "step": 31725 + }, + { + "epoch": 318.3, + "learning_rate": 2.0911392405063294e-06, + "loss": 0.0, + "step": 31750 + }, + { + "epoch": 318.55, + "learning_rate": 2.084810126582279e-06, + "loss": 0.0, + "step": 31775 + }, + { + "epoch": 318.8, + "learning_rate": 2.078481012658228e-06, + "loss": 0.0, + "step": 31800 + }, + { + "epoch": 319.05, + "learning_rate": 2.0721518987341773e-06, + "loss": 0.0, + "step": 31825 + }, + { + "epoch": 319.3, + "learning_rate": 2.0658227848101267e-06, + "loss": 0.0, + "step": 31850 + }, + { + "epoch": 319.55, + "learning_rate": 2.059493670886076e-06, + "loss": 0.0, + "step": 31875 + }, + { + "epoch": 319.8, + "learning_rate": 2.0531645569620256e-06, + "loss": 0.0, + "step": 31900 + }, + { + "epoch": 320.05, + "learning_rate": 2.046835443037975e-06, + "loss": 0.0, + "step": 31925 + }, + { + "epoch": 320.3, + "learning_rate": 2.0405063291139244e-06, + "loss": 0.0, + "step": 31950 + }, + { + "epoch": 320.55, + "learning_rate": 2.034177215189874e-06, + "loss": 0.0, + "step": 31975 + }, + { + "epoch": 320.8, + "learning_rate": 2.027848101265823e-06, + "loss": 0.0, + "step": 32000 + }, + { + "epoch": 320.8, + "eval_loss": 0.5037282705307007, + "eval_runtime": 493.6303, + "eval_samples_per_second": 25.813, + "eval_steps_per_second": 0.808, + "eval_wer": 12.810692512115054, + "step": 32000 + }, + { + "epoch": 321.05, + "learning_rate": 2.0215189873417723e-06, + "loss": 0.0, + "step": 32025 + }, + { + "epoch": 321.3, + "learning_rate": 2.0151898734177217e-06, + "loss": 0.0, + "step": 32050 + }, + { + "epoch": 321.55, + "learning_rate": 2.008860759493671e-06, + "loss": 0.0, + "step": 32075 + }, + { + "epoch": 321.8, + "learning_rate": 2.00253164556962e-06, + "loss": 0.0, + "step": 32100 + }, + { + "epoch": 322.06, + "learning_rate": 1.9962025316455696e-06, + "loss": 0.0, + "step": 32125 + }, + { + "epoch": 322.31, + "learning_rate": 1.9898734177215194e-06, + "loss": 0.0, + "step": 32150 + }, + { + "epoch": 322.56, + "learning_rate": 1.9835443037974684e-06, + "loss": 0.0, + "step": 32175 + }, + { + "epoch": 322.81, + "learning_rate": 1.977215189873418e-06, + "loss": 0.0, + "step": 32200 + }, + { + "epoch": 323.06, + "learning_rate": 1.9708860759493673e-06, + "loss": 0.0, + "step": 32225 + }, + { + "epoch": 323.31, + "learning_rate": 1.9645569620253167e-06, + "loss": 0.0, + "step": 32250 + }, + { + "epoch": 323.56, + "learning_rate": 1.958227848101266e-06, + "loss": 0.0, + "step": 32275 + }, + { + "epoch": 323.81, + "learning_rate": 1.951898734177215e-06, + "loss": 0.0, + "step": 32300 + }, + { + "epoch": 324.06, + "learning_rate": 1.9455696202531646e-06, + "loss": 0.0, + "step": 32325 + }, + { + "epoch": 324.31, + "learning_rate": 1.939240506329114e-06, + "loss": 0.0, + "step": 32350 + }, + { + "epoch": 324.56, + "learning_rate": 1.9329113924050634e-06, + "loss": 0.0, + "step": 32375 + }, + { + "epoch": 324.81, + "learning_rate": 1.926582278481013e-06, + "loss": 0.0, + "step": 32400 + }, + { + "epoch": 325.06, + "learning_rate": 1.9202531645569623e-06, + "loss": 0.0, + "step": 32425 + }, + { + "epoch": 325.31, + "learning_rate": 1.9139240506329117e-06, + "loss": 0.0, + "step": 32450 + }, + { + "epoch": 325.56, + "learning_rate": 1.9075949367088607e-06, + "loss": 0.0, + "step": 32475 + }, + { + "epoch": 325.81, + "learning_rate": 1.9012658227848101e-06, + "loss": 0.0, + "step": 32500 + }, + { + "epoch": 326.07, + "learning_rate": 1.8949367088607596e-06, + "loss": 0.0, + "step": 32525 + }, + { + "epoch": 326.32, + "learning_rate": 1.8886075949367092e-06, + "loss": 0.0, + "step": 32550 + }, + { + "epoch": 326.57, + "learning_rate": 1.8822784810126584e-06, + "loss": 0.0, + "step": 32575 + }, + { + "epoch": 326.82, + "learning_rate": 1.8759493670886078e-06, + "loss": 0.0, + "step": 32600 + }, + { + "epoch": 327.07, + "learning_rate": 1.869620253164557e-06, + "loss": 0.0, + "step": 32625 + }, + { + "epoch": 327.32, + "learning_rate": 1.8632911392405065e-06, + "loss": 0.0, + "step": 32650 + }, + { + "epoch": 327.57, + "learning_rate": 1.8569620253164557e-06, + "loss": 0.0, + "step": 32675 + }, + { + "epoch": 327.82, + "learning_rate": 1.8506329113924051e-06, + "loss": 0.0, + "step": 32700 + }, + { + "epoch": 328.07, + "learning_rate": 1.8443037974683546e-06, + "loss": 0.0, + "step": 32725 + }, + { + "epoch": 328.32, + "learning_rate": 1.837974683544304e-06, + "loss": 0.0, + "step": 32750 + }, + { + "epoch": 328.57, + "learning_rate": 1.8316455696202534e-06, + "loss": 0.0, + "step": 32775 + }, + { + "epoch": 328.82, + "learning_rate": 1.8253164556962028e-06, + "loss": 0.0, + "step": 32800 + }, + { + "epoch": 329.07, + "learning_rate": 1.818987341772152e-06, + "loss": 0.0, + "step": 32825 + }, + { + "epoch": 329.32, + "learning_rate": 1.8126582278481015e-06, + "loss": 0.0, + "step": 32850 + }, + { + "epoch": 329.57, + "learning_rate": 1.8063291139240507e-06, + "loss": 0.0, + "step": 32875 + }, + { + "epoch": 329.82, + "learning_rate": 1.8000000000000001e-06, + "loss": 0.0, + "step": 32900 + }, + { + "epoch": 330.08, + "learning_rate": 1.7936708860759493e-06, + "loss": 0.0, + "step": 32925 + }, + { + "epoch": 330.33, + "learning_rate": 1.787594936708861e-06, + "loss": 0.0, + "step": 32950 + }, + { + "epoch": 330.58, + "learning_rate": 1.7812658227848104e-06, + "loss": 0.0, + "step": 32975 + }, + { + "epoch": 330.83, + "learning_rate": 1.7749367088607597e-06, + "loss": 0.0, + "step": 33000 + }, + { + "epoch": 330.83, + "eval_loss": 0.5114565491676331, + "eval_runtime": 493.8947, + "eval_samples_per_second": 25.799, + "eval_steps_per_second": 0.808, + "eval_wer": 12.808738471158357, + "step": 33000 + }, + { + "epoch": 331.08, + "learning_rate": 1.768607594936709e-06, + "loss": 0.0, + "step": 33025 + }, + { + "epoch": 331.33, + "learning_rate": 1.7622784810126583e-06, + "loss": 0.0, + "step": 33050 + }, + { + "epoch": 331.58, + "learning_rate": 1.7559493670886077e-06, + "loss": 0.0, + "step": 33075 + }, + { + "epoch": 331.83, + "learning_rate": 1.749620253164557e-06, + "loss": 0.0, + "step": 33100 + }, + { + "epoch": 332.08, + "learning_rate": 1.7432911392405064e-06, + "loss": 0.0, + "step": 33125 + }, + { + "epoch": 332.33, + "learning_rate": 1.736962025316456e-06, + "loss": 0.0, + "step": 33150 + }, + { + "epoch": 332.58, + "learning_rate": 1.7306329113924052e-06, + "loss": 0.0, + "step": 33175 + }, + { + "epoch": 332.83, + "learning_rate": 1.7243037974683547e-06, + "loss": 0.0, + "step": 33200 + }, + { + "epoch": 333.08, + "learning_rate": 1.7179746835443039e-06, + "loss": 0.0, + "step": 33225 + }, + { + "epoch": 333.33, + "learning_rate": 1.7116455696202533e-06, + "loss": 0.0, + "step": 33250 + }, + { + "epoch": 333.58, + "learning_rate": 1.7053164556962027e-06, + "loss": 0.0, + "step": 33275 + }, + { + "epoch": 333.83, + "learning_rate": 1.698987341772152e-06, + "loss": 0.0, + "step": 33300 + }, + { + "epoch": 334.09, + "learning_rate": 1.6926582278481014e-06, + "loss": 0.0, + "step": 33325 + }, + { + "epoch": 334.34, + "learning_rate": 1.686329113924051e-06, + "loss": 0.0, + "step": 33350 + }, + { + "epoch": 334.59, + "learning_rate": 1.6800000000000002e-06, + "loss": 0.0, + "step": 33375 + }, + { + "epoch": 334.84, + "learning_rate": 1.6736708860759496e-06, + "loss": 0.0, + "step": 33400 + }, + { + "epoch": 335.09, + "learning_rate": 1.6673417721518989e-06, + "loss": 0.0, + "step": 33425 + }, + { + "epoch": 335.34, + "learning_rate": 1.6610126582278483e-06, + "loss": 0.0, + "step": 33450 + }, + { + "epoch": 335.59, + "learning_rate": 1.6546835443037975e-06, + "loss": 0.0, + "step": 33475 + }, + { + "epoch": 335.84, + "learning_rate": 1.648354430379747e-06, + "loss": 0.0, + "step": 33500 + }, + { + "epoch": 336.09, + "learning_rate": 1.6420253164556962e-06, + "loss": 0.0, + "step": 33525 + }, + { + "epoch": 336.34, + "learning_rate": 1.6356962025316456e-06, + "loss": 0.0, + "step": 33550 + }, + { + "epoch": 336.59, + "learning_rate": 1.6293670886075952e-06, + "loss": 0.0, + "step": 33575 + }, + { + "epoch": 336.84, + "learning_rate": 1.6230379746835444e-06, + "loss": 0.0, + "step": 33600 + }, + { + "epoch": 337.09, + "learning_rate": 1.6167088607594939e-06, + "loss": 0.0, + "step": 33625 + }, + { + "epoch": 337.34, + "learning_rate": 1.610379746835443e-06, + "loss": 0.0, + "step": 33650 + }, + { + "epoch": 337.59, + "learning_rate": 1.6040506329113925e-06, + "loss": 0.0, + "step": 33675 + }, + { + "epoch": 337.84, + "learning_rate": 1.597721518987342e-06, + "loss": 0.0, + "step": 33700 + }, + { + "epoch": 338.1, + "learning_rate": 1.5913924050632911e-06, + "loss": 0.0, + "step": 33725 + }, + { + "epoch": 338.35, + "learning_rate": 1.5850632911392406e-06, + "loss": 0.0, + "step": 33750 + }, + { + "epoch": 338.6, + "learning_rate": 1.5787341772151902e-06, + "loss": 0.0, + "step": 33775 + }, + { + "epoch": 338.85, + "learning_rate": 1.5724050632911394e-06, + "loss": 0.0, + "step": 33800 + }, + { + "epoch": 339.1, + "learning_rate": 1.5660759493670889e-06, + "loss": 0.0, + "step": 33825 + }, + { + "epoch": 339.35, + "learning_rate": 1.559746835443038e-06, + "loss": 0.0, + "step": 33850 + }, + { + "epoch": 339.6, + "learning_rate": 1.5534177215189875e-06, + "loss": 0.0, + "step": 33875 + }, + { + "epoch": 339.85, + "learning_rate": 1.5470886075949367e-06, + "loss": 0.0, + "step": 33900 + }, + { + "epoch": 340.1, + "learning_rate": 1.5407594936708861e-06, + "loss": 0.0, + "step": 33925 + }, + { + "epoch": 340.35, + "learning_rate": 1.5344303797468354e-06, + "loss": 0.0, + "step": 33950 + }, + { + "epoch": 340.6, + "learning_rate": 1.528101265822785e-06, + "loss": 0.0, + "step": 33975 + }, + { + "epoch": 340.85, + "learning_rate": 1.5217721518987344e-06, + "loss": 0.0, + "step": 34000 + }, + { + "epoch": 340.85, + "eval_loss": 0.5190522074699402, + "eval_runtime": 493.7266, + "eval_samples_per_second": 25.808, + "eval_steps_per_second": 0.808, + "eval_wer": 12.82925590120369, + "step": 34000 + }, + { + "epoch": 341.1, + "learning_rate": 1.5154430379746836e-06, + "loss": 0.0, + "step": 34025 + }, + { + "epoch": 341.35, + "learning_rate": 1.509113924050633e-06, + "loss": 0.0, + "step": 34050 + }, + { + "epoch": 341.6, + "learning_rate": 1.5027848101265823e-06, + "loss": 0.0, + "step": 34075 + }, + { + "epoch": 341.85, + "learning_rate": 1.4964556962025317e-06, + "loss": 0.0, + "step": 34100 + }, + { + "epoch": 342.11, + "learning_rate": 1.4901265822784811e-06, + "loss": 0.0, + "step": 34125 + }, + { + "epoch": 342.36, + "learning_rate": 1.4837974683544304e-06, + "loss": 0.0, + "step": 34150 + }, + { + "epoch": 342.61, + "learning_rate": 1.47746835443038e-06, + "loss": 0.0, + "step": 34175 + }, + { + "epoch": 342.86, + "learning_rate": 1.4711392405063294e-06, + "loss": 0.0, + "step": 34200 + }, + { + "epoch": 343.11, + "learning_rate": 1.4648101265822786e-06, + "loss": 0.0, + "step": 34225 + }, + { + "epoch": 343.36, + "learning_rate": 1.458481012658228e-06, + "loss": 0.0, + "step": 34250 + }, + { + "epoch": 343.61, + "learning_rate": 1.4521518987341773e-06, + "loss": 0.0, + "step": 34275 + }, + { + "epoch": 343.86, + "learning_rate": 1.4458227848101267e-06, + "loss": 0.0, + "step": 34300 + }, + { + "epoch": 344.11, + "learning_rate": 1.439493670886076e-06, + "loss": 0.0, + "step": 34325 + }, + { + "epoch": 344.36, + "learning_rate": 1.4331645569620254e-06, + "loss": 0.0, + "step": 34350 + }, + { + "epoch": 344.61, + "learning_rate": 1.4268354430379746e-06, + "loss": 0.0, + "step": 34375 + }, + { + "epoch": 344.86, + "learning_rate": 1.4205063291139242e-06, + "loss": 0.0, + "step": 34400 + }, + { + "epoch": 345.11, + "learning_rate": 1.4141772151898736e-06, + "loss": 0.0, + "step": 34425 + }, + { + "epoch": 345.36, + "learning_rate": 1.4078481012658229e-06, + "loss": 0.0, + "step": 34450 + }, + { + "epoch": 345.61, + "learning_rate": 1.4015189873417723e-06, + "loss": 0.0, + "step": 34475 + }, + { + "epoch": 345.86, + "learning_rate": 1.3951898734177215e-06, + "loss": 0.0, + "step": 34500 + }, + { + "epoch": 346.12, + "learning_rate": 1.388860759493671e-06, + "loss": 0.0, + "step": 34525 + }, + { + "epoch": 346.37, + "learning_rate": 1.3825316455696204e-06, + "loss": 0.0, + "step": 34550 + }, + { + "epoch": 346.62, + "learning_rate": 1.3762025316455696e-06, + "loss": 0.0, + "step": 34575 + }, + { + "epoch": 346.87, + "learning_rate": 1.3698734177215192e-06, + "loss": 0.0, + "step": 34600 + }, + { + "epoch": 347.12, + "learning_rate": 1.3635443037974686e-06, + "loss": 0.0, + "step": 34625 + }, + { + "epoch": 347.37, + "learning_rate": 1.3572151898734179e-06, + "loss": 0.0, + "step": 34650 + }, + { + "epoch": 347.62, + "learning_rate": 1.3508860759493673e-06, + "loss": 0.0, + "step": 34675 + }, + { + "epoch": 347.87, + "learning_rate": 1.3445569620253165e-06, + "loss": 0.0, + "step": 34700 + }, + { + "epoch": 348.12, + "learning_rate": 1.338227848101266e-06, + "loss": 0.0, + "step": 34725 + }, + { + "epoch": 348.37, + "learning_rate": 1.3318987341772151e-06, + "loss": 0.0, + "step": 34750 + }, + { + "epoch": 348.62, + "learning_rate": 1.3255696202531646e-06, + "loss": 0.0, + "step": 34775 + }, + { + "epoch": 348.87, + "learning_rate": 1.3192405063291142e-06, + "loss": 0.0, + "step": 34800 + }, + { + "epoch": 349.12, + "learning_rate": 1.3129113924050634e-06, + "loss": 0.0, + "step": 34825 + }, + { + "epoch": 349.37, + "learning_rate": 1.3065822784810129e-06, + "loss": 0.0, + "step": 34850 + }, + { + "epoch": 349.62, + "learning_rate": 1.300253164556962e-06, + "loss": 0.0, + "step": 34875 + }, + { + "epoch": 349.87, + "learning_rate": 1.2939240506329115e-06, + "loss": 0.0, + "step": 34900 + }, + { + "epoch": 350.13, + "learning_rate": 1.287594936708861e-06, + "loss": 0.0, + "step": 34925 + }, + { + "epoch": 350.38, + "learning_rate": 1.2812658227848101e-06, + "loss": 0.0, + "step": 34950 + }, + { + "epoch": 350.63, + "learning_rate": 1.2749367088607596e-06, + "loss": 0.0, + "step": 34975 + }, + { + "epoch": 350.88, + "learning_rate": 1.268607594936709e-06, + "loss": 0.0, + "step": 35000 + }, + { + "epoch": 350.88, + "eval_loss": 0.5256416201591492, + "eval_runtime": 493.6908, + "eval_samples_per_second": 25.81, + "eval_steps_per_second": 0.808, + "eval_wer": 12.891785211818041, + "step": 35000 + }, + { + "epoch": 351.13, + "learning_rate": 1.2622784810126584e-06, + "loss": 0.0, + "step": 35025 + }, + { + "epoch": 351.38, + "learning_rate": 1.2562025316455697e-06, + "loss": 0.0, + "step": 35050 + }, + { + "epoch": 351.63, + "learning_rate": 1.249873417721519e-06, + "loss": 0.0, + "step": 35075 + }, + { + "epoch": 351.88, + "learning_rate": 1.2435443037974685e-06, + "loss": 0.0, + "step": 35100 + }, + { + "epoch": 352.13, + "learning_rate": 1.2372151898734177e-06, + "loss": 0.0, + "step": 35125 + }, + { + "epoch": 352.38, + "learning_rate": 1.2308860759493672e-06, + "loss": 0.0, + "step": 35150 + }, + { + "epoch": 352.63, + "learning_rate": 1.2245569620253166e-06, + "loss": 0.0, + "step": 35175 + }, + { + "epoch": 352.88, + "learning_rate": 1.2182278481012658e-06, + "loss": 0.0, + "step": 35200 + }, + { + "epoch": 353.13, + "learning_rate": 1.2118987341772152e-06, + "loss": 0.0, + "step": 35225 + }, + { + "epoch": 353.38, + "learning_rate": 1.2055696202531647e-06, + "loss": 0.0, + "step": 35250 + }, + { + "epoch": 353.63, + "learning_rate": 1.199240506329114e-06, + "loss": 0.0, + "step": 35275 + }, + { + "epoch": 353.88, + "learning_rate": 1.1929113924050633e-06, + "loss": 0.0, + "step": 35300 + }, + { + "epoch": 354.14, + "learning_rate": 1.1865822784810127e-06, + "loss": 0.0, + "step": 35325 + }, + { + "epoch": 354.39, + "learning_rate": 1.1802531645569622e-06, + "loss": 0.0, + "step": 35350 + }, + { + "epoch": 354.64, + "learning_rate": 1.1739240506329116e-06, + "loss": 0.0, + "step": 35375 + }, + { + "epoch": 354.89, + "learning_rate": 1.1675949367088608e-06, + "loss": 0.0, + "step": 35400 + }, + { + "epoch": 355.14, + "learning_rate": 1.1612658227848102e-06, + "loss": 0.0, + "step": 35425 + }, + { + "epoch": 355.39, + "learning_rate": 1.1549367088607597e-06, + "loss": 0.0, + "step": 35450 + }, + { + "epoch": 355.64, + "learning_rate": 1.1486075949367089e-06, + "loss": 0.0, + "step": 35475 + }, + { + "epoch": 355.89, + "learning_rate": 1.1422784810126583e-06, + "loss": 0.0, + "step": 35500 + }, + { + "epoch": 356.14, + "learning_rate": 1.1359493670886077e-06, + "loss": 0.0, + "step": 35525 + }, + { + "epoch": 356.39, + "learning_rate": 1.1296202531645572e-06, + "loss": 0.0, + "step": 35550 + }, + { + "epoch": 356.64, + "learning_rate": 1.1232911392405064e-06, + "loss": 0.0, + "step": 35575 + }, + { + "epoch": 356.89, + "learning_rate": 1.1169620253164558e-06, + "loss": 0.0, + "step": 35600 + }, + { + "epoch": 357.14, + "learning_rate": 1.110632911392405e-06, + "loss": 0.0, + "step": 35625 + }, + { + "epoch": 357.39, + "learning_rate": 1.1043037974683547e-06, + "loss": 0.0, + "step": 35650 + }, + { + "epoch": 357.64, + "learning_rate": 1.0979746835443039e-06, + "loss": 0.0, + "step": 35675 + }, + { + "epoch": 357.89, + "learning_rate": 1.0916455696202533e-06, + "loss": 0.0, + "step": 35700 + }, + { + "epoch": 358.15, + "learning_rate": 1.0853164556962025e-06, + "loss": 0.0, + "step": 35725 + }, + { + "epoch": 358.4, + "learning_rate": 1.078987341772152e-06, + "loss": 0.0, + "step": 35750 + }, + { + "epoch": 358.65, + "learning_rate": 1.0726582278481014e-06, + "loss": 0.0, + "step": 35775 + }, + { + "epoch": 358.9, + "learning_rate": 1.0663291139240508e-06, + "loss": 0.0, + "step": 35800 + }, + { + "epoch": 359.15, + "learning_rate": 1.06e-06, + "loss": 0.0, + "step": 35825 + }, + { + "epoch": 359.4, + "learning_rate": 1.0536708860759494e-06, + "loss": 0.0, + "step": 35850 + }, + { + "epoch": 359.65, + "learning_rate": 1.0473417721518989e-06, + "loss": 0.0, + "step": 35875 + }, + { + "epoch": 359.9, + "learning_rate": 1.0410126582278483e-06, + "loss": 0.0, + "step": 35900 + }, + { + "epoch": 360.15, + "learning_rate": 1.0346835443037975e-06, + "loss": 0.0, + "step": 35925 + }, + { + "epoch": 360.4, + "learning_rate": 1.028354430379747e-06, + "loss": 0.0, + "step": 35950 + }, + { + "epoch": 360.65, + "learning_rate": 1.0220253164556964e-06, + "loss": 0.0, + "step": 35975 + }, + { + "epoch": 360.9, + "learning_rate": 1.0156962025316456e-06, + "loss": 0.0, + "step": 36000 + }, + { + "epoch": 360.9, + "eval_loss": 0.5312889814376831, + "eval_runtime": 493.6997, + "eval_samples_per_second": 25.809, + "eval_steps_per_second": 0.808, + "eval_wer": 12.881037986556198, + "step": 36000 + }, + { + "epoch": 361.15, + "learning_rate": 1.009367088607595e-06, + "loss": 0.0, + "step": 36025 + }, + { + "epoch": 361.4, + "learning_rate": 1.0030379746835442e-06, + "loss": 0.0, + "step": 36050 + }, + { + "epoch": 361.65, + "learning_rate": 9.967088607594939e-07, + "loss": 0.0, + "step": 36075 + }, + { + "epoch": 361.9, + "learning_rate": 9.90379746835443e-07, + "loss": 0.0, + "step": 36100 + }, + { + "epoch": 362.16, + "learning_rate": 9.840506329113925e-07, + "loss": 0.0, + "step": 36125 + }, + { + "epoch": 362.41, + "learning_rate": 9.777215189873417e-07, + "loss": 0.0, + "step": 36150 + }, + { + "epoch": 362.66, + "learning_rate": 9.713924050632914e-07, + "loss": 0.0, + "step": 36175 + }, + { + "epoch": 362.91, + "learning_rate": 9.650632911392406e-07, + "loss": 0.0, + "step": 36200 + }, + { + "epoch": 363.16, + "learning_rate": 9.5873417721519e-07, + "loss": 0.0, + "step": 36225 + }, + { + "epoch": 363.41, + "learning_rate": 9.524050632911392e-07, + "loss": 0.0, + "step": 36250 + }, + { + "epoch": 363.66, + "learning_rate": 9.460759493670888e-07, + "loss": 0.0, + "step": 36275 + }, + { + "epoch": 363.91, + "learning_rate": 9.397468354430381e-07, + "loss": 0.0, + "step": 36300 + }, + { + "epoch": 364.16, + "learning_rate": 9.334177215189874e-07, + "loss": 0.0, + "step": 36325 + }, + { + "epoch": 364.41, + "learning_rate": 9.270886075949367e-07, + "loss": 0.0, + "step": 36350 + }, + { + "epoch": 364.66, + "learning_rate": 9.207594936708863e-07, + "loss": 0.0, + "step": 36375 + }, + { + "epoch": 364.91, + "learning_rate": 9.144303797468356e-07, + "loss": 0.0, + "step": 36400 + }, + { + "epoch": 365.16, + "learning_rate": 9.081012658227849e-07, + "loss": 0.0, + "step": 36425 + }, + { + "epoch": 365.41, + "learning_rate": 9.017721518987342e-07, + "loss": 0.0, + "step": 36450 + }, + { + "epoch": 365.66, + "learning_rate": 8.954430379746837e-07, + "loss": 0.0, + "step": 36475 + }, + { + "epoch": 365.91, + "learning_rate": 8.89113924050633e-07, + "loss": 0.0, + "step": 36500 + }, + { + "epoch": 366.17, + "learning_rate": 8.827848101265823e-07, + "loss": 0.0, + "step": 36525 + }, + { + "epoch": 366.42, + "learning_rate": 8.764556962025317e-07, + "loss": 0.0, + "step": 36550 + }, + { + "epoch": 366.67, + "learning_rate": 8.701265822784812e-07, + "loss": 0.0, + "step": 36575 + }, + { + "epoch": 366.92, + "learning_rate": 8.637974683544305e-07, + "loss": 0.0, + "step": 36600 + }, + { + "epoch": 367.17, + "learning_rate": 8.574683544303798e-07, + "loss": 0.0, + "step": 36625 + }, + { + "epoch": 367.42, + "learning_rate": 8.511392405063291e-07, + "loss": 0.0, + "step": 36650 + }, + { + "epoch": 367.67, + "learning_rate": 8.448101265822784e-07, + "loss": 0.0, + "step": 36675 + }, + { + "epoch": 367.92, + "learning_rate": 8.38481012658228e-07, + "loss": 0.0, + "step": 36700 + }, + { + "epoch": 368.17, + "learning_rate": 8.321518987341773e-07, + "loss": 0.0, + "step": 36725 + }, + { + "epoch": 368.42, + "learning_rate": 8.258227848101266e-07, + "loss": 0.0, + "step": 36750 + }, + { + "epoch": 368.67, + "learning_rate": 8.194936708860759e-07, + "loss": 0.0, + "step": 36775 + }, + { + "epoch": 368.92, + "learning_rate": 8.131645569620255e-07, + "loss": 0.0, + "step": 36800 + }, + { + "epoch": 369.17, + "learning_rate": 8.068354430379748e-07, + "loss": 0.0, + "step": 36825 + }, + { + "epoch": 369.42, + "learning_rate": 8.005063291139241e-07, + "loss": 0.0, + "step": 36850 + }, + { + "epoch": 369.67, + "learning_rate": 7.941772151898734e-07, + "loss": 0.0, + "step": 36875 + }, + { + "epoch": 369.92, + "learning_rate": 7.878481012658229e-07, + "loss": 0.0, + "step": 36900 + }, + { + "epoch": 370.18, + "learning_rate": 7.815189873417722e-07, + "loss": 0.0, + "step": 36925 + }, + { + "epoch": 370.43, + "learning_rate": 7.751898734177215e-07, + "loss": 0.0, + "step": 36950 + }, + { + "epoch": 370.68, + "learning_rate": 7.688607594936709e-07, + "loss": 0.0, + "step": 36975 + }, + { + "epoch": 370.93, + "learning_rate": 7.625316455696204e-07, + "loss": 0.0, + "step": 37000 + }, + { + "epoch": 370.93, + "eval_loss": 0.536057710647583, + "eval_runtime": 494.572, + "eval_samples_per_second": 25.764, + "eval_steps_per_second": 0.807, + "eval_wer": 12.904486478036581, + "step": 37000 + }, + { + "epoch": 371.18, + "learning_rate": 7.562025316455697e-07, + "loss": 0.0, + "step": 37025 + }, + { + "epoch": 371.43, + "learning_rate": 7.49873417721519e-07, + "loss": 0.0, + "step": 37050 + }, + { + "epoch": 371.68, + "learning_rate": 7.437974683544305e-07, + "loss": 0.0, + "step": 37075 + }, + { + "epoch": 371.93, + "learning_rate": 7.374683544303799e-07, + "loss": 0.0, + "step": 37100 + }, + { + "epoch": 372.18, + "learning_rate": 7.311392405063292e-07, + "loss": 0.0, + "step": 37125 + }, + { + "epoch": 372.43, + "learning_rate": 7.248101265822785e-07, + "loss": 0.0, + "step": 37150 + }, + { + "epoch": 372.68, + "learning_rate": 7.184810126582279e-07, + "loss": 0.0, + "step": 37175 + }, + { + "epoch": 372.93, + "learning_rate": 7.121518987341773e-07, + "loss": 0.0, + "step": 37200 + }, + { + "epoch": 373.18, + "learning_rate": 7.058227848101266e-07, + "loss": 0.0, + "step": 37225 + }, + { + "epoch": 373.43, + "learning_rate": 6.99493670886076e-07, + "loss": 0.0, + "step": 37250 + }, + { + "epoch": 373.68, + "learning_rate": 6.931645569620254e-07, + "loss": 0.0, + "step": 37275 + }, + { + "epoch": 373.93, + "learning_rate": 6.868354430379748e-07, + "loss": 0.0, + "step": 37300 + }, + { + "epoch": 374.19, + "learning_rate": 6.805063291139241e-07, + "loss": 0.0, + "step": 37325 + }, + { + "epoch": 374.44, + "learning_rate": 6.741772151898734e-07, + "loss": 0.0, + "step": 37350 + }, + { + "epoch": 374.69, + "learning_rate": 6.678481012658227e-07, + "loss": 0.0, + "step": 37375 + }, + { + "epoch": 374.94, + "learning_rate": 6.615189873417723e-07, + "loss": 0.0, + "step": 37400 + }, + { + "epoch": 375.19, + "learning_rate": 6.551898734177216e-07, + "loss": 0.0, + "step": 37425 + }, + { + "epoch": 375.44, + "learning_rate": 6.488607594936709e-07, + "loss": 0.0, + "step": 37450 + }, + { + "epoch": 375.69, + "learning_rate": 6.425316455696202e-07, + "loss": 0.0, + "step": 37475 + }, + { + "epoch": 375.94, + "learning_rate": 6.362025316455697e-07, + "loss": 0.0, + "step": 37500 + }, + { + "epoch": 376.19, + "learning_rate": 6.298734177215191e-07, + "loss": 0.0, + "step": 37525 + }, + { + "epoch": 376.44, + "learning_rate": 6.235443037974684e-07, + "loss": 0.0, + "step": 37550 + }, + { + "epoch": 376.69, + "learning_rate": 6.172151898734177e-07, + "loss": 0.0, + "step": 37575 + }, + { + "epoch": 376.94, + "learning_rate": 6.108860759493672e-07, + "loss": 0.0, + "step": 37600 + }, + { + "epoch": 377.19, + "learning_rate": 6.045569620253165e-07, + "loss": 0.0, + "step": 37625 + }, + { + "epoch": 377.44, + "learning_rate": 5.982278481012658e-07, + "loss": 0.0, + "step": 37650 + }, + { + "epoch": 377.69, + "learning_rate": 5.918987341772152e-07, + "loss": 0.0, + "step": 37675 + }, + { + "epoch": 377.94, + "learning_rate": 5.855696202531646e-07, + "loss": 0.0, + "step": 37700 + }, + { + "epoch": 378.2, + "learning_rate": 5.79240506329114e-07, + "loss": 0.0, + "step": 37725 + }, + { + "epoch": 378.45, + "learning_rate": 5.729113924050633e-07, + "loss": 0.0, + "step": 37750 + }, + { + "epoch": 378.7, + "learning_rate": 5.665822784810127e-07, + "loss": 0.0, + "step": 37775 + }, + { + "epoch": 378.95, + "learning_rate": 5.602531645569621e-07, + "loss": 0.0, + "step": 37800 + }, + { + "epoch": 379.2, + "learning_rate": 5.539240506329115e-07, + "loss": 0.0, + "step": 37825 + }, + { + "epoch": 379.45, + "learning_rate": 5.475949367088608e-07, + "loss": 0.0, + "step": 37850 + }, + { + "epoch": 379.7, + "learning_rate": 5.412658227848102e-07, + "loss": 0.0, + "step": 37875 + }, + { + "epoch": 379.95, + "learning_rate": 5.349367088607596e-07, + "loss": 0.0, + "step": 37900 + }, + { + "epoch": 380.2, + "learning_rate": 5.286075949367089e-07, + "loss": 0.0, + "step": 37925 + }, + { + "epoch": 380.45, + "learning_rate": 5.222784810126583e-07, + "loss": 0.0, + "step": 37950 + }, + { + "epoch": 380.7, + "learning_rate": 5.159493670886076e-07, + "loss": 0.0, + "step": 37975 + }, + { + "epoch": 380.95, + "learning_rate": 5.09620253164557e-07, + "loss": 0.0, + "step": 38000 + }, + { + "epoch": 380.95, + "eval_loss": 0.5394255518913269, + "eval_runtime": 495.5391, + "eval_samples_per_second": 25.713, + "eval_steps_per_second": 0.805, + "eval_wer": 12.899601375644835, + "step": 38000 + }, + { + "epoch": 381.2, + "learning_rate": 5.032911392405064e-07, + "loss": 0.0, + "step": 38025 + }, + { + "epoch": 381.45, + "learning_rate": 4.969620253164557e-07, + "loss": 0.0, + "step": 38050 + }, + { + "epoch": 381.7, + "learning_rate": 4.90632911392405e-07, + "loss": 0.0, + "step": 38075 + }, + { + "epoch": 381.95, + "learning_rate": 4.843037974683545e-07, + "loss": 0.0, + "step": 38100 + }, + { + "epoch": 382.21, + "learning_rate": 4.779746835443038e-07, + "loss": 0.0, + "step": 38125 + }, + { + "epoch": 382.46, + "learning_rate": 4.716455696202532e-07, + "loss": 0.0, + "step": 38150 + }, + { + "epoch": 382.71, + "learning_rate": 4.653164556962025e-07, + "loss": 0.0, + "step": 38175 + }, + { + "epoch": 382.96, + "learning_rate": 4.5898734177215195e-07, + "loss": 0.0, + "step": 38200 + }, + { + "epoch": 383.21, + "learning_rate": 4.5265822784810127e-07, + "loss": 0.0, + "step": 38225 + }, + { + "epoch": 383.46, + "learning_rate": 4.463291139240507e-07, + "loss": 0.0, + "step": 38250 + }, + { + "epoch": 383.71, + "learning_rate": 4.4e-07, + "loss": 0.0, + "step": 38275 + }, + { + "epoch": 383.96, + "learning_rate": 4.336708860759494e-07, + "loss": 0.0, + "step": 38300 + }, + { + "epoch": 384.21, + "learning_rate": 4.2734177215189877e-07, + "loss": 0.0, + "step": 38325 + }, + { + "epoch": 384.46, + "learning_rate": 4.2101265822784815e-07, + "loss": 0.0, + "step": 38350 + }, + { + "epoch": 384.71, + "learning_rate": 4.1468354430379747e-07, + "loss": 0.0, + "step": 38375 + }, + { + "epoch": 384.96, + "learning_rate": 4.083544303797469e-07, + "loss": 0.0, + "step": 38400 + }, + { + "epoch": 385.21, + "learning_rate": 4.020253164556962e-07, + "loss": 0.0, + "step": 38425 + }, + { + "epoch": 385.46, + "learning_rate": 3.956962025316456e-07, + "loss": 0.0, + "step": 38450 + }, + { + "epoch": 385.71, + "learning_rate": 3.8936708860759497e-07, + "loss": 0.0, + "step": 38475 + }, + { + "epoch": 385.96, + "learning_rate": 3.8303797468354434e-07, + "loss": 0.0, + "step": 38500 + }, + { + "epoch": 386.22, + "learning_rate": 3.7670886075949366e-07, + "loss": 0.0, + "step": 38525 + }, + { + "epoch": 386.47, + "learning_rate": 3.703797468354431e-07, + "loss": 0.0, + "step": 38550 + }, + { + "epoch": 386.72, + "learning_rate": 3.640506329113924e-07, + "loss": 0.0, + "step": 38575 + }, + { + "epoch": 386.97, + "learning_rate": 3.5772151898734184e-07, + "loss": 0.0, + "step": 38600 + }, + { + "epoch": 387.22, + "learning_rate": 3.5139240506329116e-07, + "loss": 0.0, + "step": 38625 + }, + { + "epoch": 387.47, + "learning_rate": 3.4506329113924054e-07, + "loss": 0.0, + "step": 38650 + }, + { + "epoch": 387.72, + "learning_rate": 3.387341772151899e-07, + "loss": 0.0, + "step": 38675 + }, + { + "epoch": 387.97, + "learning_rate": 3.324050632911393e-07, + "loss": 0.0, + "step": 38700 + }, + { + "epoch": 388.22, + "learning_rate": 3.260759493670886e-07, + "loss": 0.0, + "step": 38725 + }, + { + "epoch": 388.47, + "learning_rate": 3.1974683544303804e-07, + "loss": 0.0, + "step": 38750 + }, + { + "epoch": 388.72, + "learning_rate": 3.1341772151898736e-07, + "loss": 0.0, + "step": 38775 + }, + { + "epoch": 388.97, + "learning_rate": 3.0708860759493673e-07, + "loss": 0.0, + "step": 38800 + }, + { + "epoch": 389.22, + "learning_rate": 3.007594936708861e-07, + "loss": 0.0, + "step": 38825 + }, + { + "epoch": 389.47, + "learning_rate": 2.944303797468355e-07, + "loss": 0.0, + "step": 38850 + }, + { + "epoch": 389.72, + "learning_rate": 2.8810126582278486e-07, + "loss": 0.0, + "step": 38875 + }, + { + "epoch": 389.97, + "learning_rate": 2.817721518987342e-07, + "loss": 0.0, + "step": 38900 + }, + { + "epoch": 390.23, + "learning_rate": 2.7544303797468355e-07, + "loss": 0.0, + "step": 38925 + }, + { + "epoch": 390.48, + "learning_rate": 2.6911392405063293e-07, + "loss": 0.0, + "step": 38950 + }, + { + "epoch": 390.73, + "learning_rate": 2.627848101265823e-07, + "loss": 0.0, + "step": 38975 + }, + { + "epoch": 390.98, + "learning_rate": 2.564556962025317e-07, + "loss": 0.0, + "step": 39000 + }, + { + "epoch": 390.98, + "eval_loss": 0.5416933298110962, + "eval_runtime": 495.0613, + "eval_samples_per_second": 25.738, + "eval_steps_per_second": 0.806, + "eval_wer": 12.912302641863374, + "step": 39000 + }, + { + "epoch": 391.23, + "learning_rate": 2.5012658227848105e-07, + "loss": 0.0, + "step": 39025 + }, + { + "epoch": 391.48, + "learning_rate": 2.4379746835443043e-07, + "loss": 0.0, + "step": 39050 + }, + { + "epoch": 391.73, + "learning_rate": 2.3746835443037978e-07, + "loss": 0.0, + "step": 39075 + }, + { + "epoch": 391.98, + "learning_rate": 2.3139240506329115e-07, + "loss": 0.0, + "step": 39100 + }, + { + "epoch": 392.23, + "learning_rate": 2.2506329113924052e-07, + "loss": 0.0, + "step": 39125 + }, + { + "epoch": 392.48, + "learning_rate": 2.1873417721518987e-07, + "loss": 0.0, + "step": 39150 + }, + { + "epoch": 392.73, + "learning_rate": 2.1240506329113925e-07, + "loss": 0.0, + "step": 39175 + }, + { + "epoch": 392.98, + "learning_rate": 2.0607594936708862e-07, + "loss": 0.0, + "step": 39200 + }, + { + "epoch": 393.23, + "learning_rate": 1.99746835443038e-07, + "loss": 0.0, + "step": 39225 + }, + { + "epoch": 393.48, + "learning_rate": 1.9341772151898734e-07, + "loss": 0.0, + "step": 39250 + }, + { + "epoch": 393.73, + "learning_rate": 1.8708860759493672e-07, + "loss": 0.0, + "step": 39275 + }, + { + "epoch": 393.98, + "learning_rate": 1.807594936708861e-07, + "loss": 0.0, + "step": 39300 + }, + { + "epoch": 394.24, + "learning_rate": 1.7443037974683544e-07, + "loss": 0.0, + "step": 39325 + }, + { + "epoch": 394.49, + "learning_rate": 1.6810126582278482e-07, + "loss": 0.0, + "step": 39350 + }, + { + "epoch": 394.74, + "learning_rate": 1.617721518987342e-07, + "loss": 0.0, + "step": 39375 + }, + { + "epoch": 394.99, + "learning_rate": 1.5544303797468356e-07, + "loss": 0.0, + "step": 39400 + }, + { + "epoch": 395.24, + "learning_rate": 1.491139240506329e-07, + "loss": 0.0, + "step": 39425 + }, + { + "epoch": 395.49, + "learning_rate": 1.427848101265823e-07, + "loss": 0.0, + "step": 39450 + }, + { + "epoch": 395.74, + "learning_rate": 1.3645569620253166e-07, + "loss": 0.0, + "step": 39475 + }, + { + "epoch": 395.99, + "learning_rate": 1.3012658227848104e-07, + "loss": 0.0, + "step": 39500 + }, + { + "epoch": 396.24, + "learning_rate": 1.2379746835443039e-07, + "loss": 0.0, + "step": 39525 + }, + { + "epoch": 396.49, + "learning_rate": 1.1746835443037976e-07, + "loss": 0.0, + "step": 39550 + }, + { + "epoch": 396.74, + "learning_rate": 1.1113924050632912e-07, + "loss": 0.0, + "step": 39575 + }, + { + "epoch": 396.99, + "learning_rate": 1.048101265822785e-07, + "loss": 0.0, + "step": 39600 + }, + { + "epoch": 397.24, + "learning_rate": 9.848101265822786e-08, + "loss": 0.0, + "step": 39625 + }, + { + "epoch": 397.49, + "learning_rate": 9.215189873417723e-08, + "loss": 0.0, + "step": 39650 + }, + { + "epoch": 397.74, + "learning_rate": 8.58227848101266e-08, + "loss": 0.0, + "step": 39675 + }, + { + "epoch": 397.99, + "learning_rate": 7.949367088607594e-08, + "loss": 0.0, + "step": 39700 + }, + { + "epoch": 398.25, + "learning_rate": 7.316455696202532e-08, + "loss": 0.0, + "step": 39725 + }, + { + "epoch": 398.5, + "learning_rate": 6.683544303797469e-08, + "loss": 0.0, + "step": 39750 + }, + { + "epoch": 398.75, + "learning_rate": 6.050632911392405e-08, + "loss": 0.0, + "step": 39775 + }, + { + "epoch": 399.0, + "learning_rate": 5.417721518987343e-08, + "loss": 0.0, + "step": 39800 + }, + { + "epoch": 399.25, + "learning_rate": 4.784810126582278e-08, + "loss": 0.0, + "step": 39825 + }, + { + "epoch": 399.5, + "learning_rate": 4.151898734177215e-08, + "loss": 0.0, + "step": 39850 + }, + { + "epoch": 399.75, + "learning_rate": 3.518987341772152e-08, + "loss": 0.0, + "step": 39875 + }, + { + "epoch": 400.0, + "learning_rate": 2.8860759493670888e-08, + "loss": 0.0, + "step": 39900 + }, + { + "epoch": 400.25, + "learning_rate": 2.2531645569620256e-08, + "loss": 0.0, + "step": 39925 + }, + { + "epoch": 400.5, + "learning_rate": 1.6202531645569624e-08, + "loss": 0.0, + "step": 39950 + }, + { + "epoch": 400.75, + "learning_rate": 9.873417721518989e-09, + "loss": 0.0, + "step": 39975 + }, + { + "epoch": 401.0, + "learning_rate": 3.544303797468355e-09, + "loss": 0.0, + "step": 40000 + }, + { + "epoch": 401.0, + "eval_loss": 0.5424726009368896, + "eval_runtime": 494.402, + "eval_samples_per_second": 25.773, + "eval_steps_per_second": 0.807, + "eval_wer": 12.912302641863374, + "step": 40000 + }, + { + "epoch": 401.0, + "step": 40000, + "total_flos": 2.9490969165093274e+21, + "train_loss": 0.007129912708719667, + "train_runtime": 186833.5774, + "train_samples_per_second": 54.808, + "train_steps_per_second": 0.214 + } + ], + "logging_steps": 25, + "max_steps": 40000, + "num_input_tokens_seen": 0, + "num_train_epochs": 405, + "save_steps": 1000, + "total_flos": 2.9490969165093274e+21, + "train_batch_size": 64, + "trial_name": null, + "trial_params": null +}