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": 19.094888228857275, + "best_model_checkpoint": "./checkpoint-19000", + "epoch": 400.0, + "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": 1.6499999999999997e-06, + "loss": 2.4538, + "step": 25 + }, + { + "epoch": 0.5, + "learning_rate": 3.5249999999999997e-06, + "loss": 1.87, + "step": 50 + }, + { + "epoch": 0.75, + "learning_rate": 5.399999999999999e-06, + "loss": 1.3171, + "step": 75 + }, + { + "epoch": 1.0, + "learning_rate": 7.275e-06, + "loss": 1.016, + "step": 100 + }, + { + "epoch": 1.25, + "learning_rate": 9.149999999999999e-06, + "loss": 0.834, + "step": 125 + }, + { + "epoch": 1.5, + "learning_rate": 1.1024999999999999e-05, + "loss": 0.7356, + "step": 150 + }, + { + "epoch": 1.75, + "learning_rate": 1.2899999999999998e-05, + "loss": 0.6513, + "step": 175 + }, + { + "epoch": 2.0, + "learning_rate": 1.4775e-05, + "loss": 0.5926, + "step": 200 + }, + { + "epoch": 2.25, + "learning_rate": 1.6649999999999998e-05, + "loss": 0.5086, + "step": 225 + }, + { + "epoch": 2.5, + "learning_rate": 1.8525e-05, + "loss": 0.4757, + "step": 250 + }, + { + "epoch": 2.75, + "learning_rate": 2.04e-05, + "loss": 0.4501, + "step": 275 + }, + { + "epoch": 3.0, + "learning_rate": 2.2274999999999996e-05, + "loss": 0.4266, + "step": 300 + }, + { + "epoch": 3.25, + "learning_rate": 2.4149999999999997e-05, + "loss": 0.357, + "step": 325 + }, + { + "epoch": 3.5, + "learning_rate": 2.6024999999999996e-05, + "loss": 0.3415, + "step": 350 + }, + { + "epoch": 3.75, + "learning_rate": 2.7899999999999997e-05, + "loss": 0.3319, + "step": 375 + }, + { + "epoch": 4.0, + "learning_rate": 2.9775e-05, + "loss": 0.3189, + "step": 400 + }, + { + "epoch": 4.25, + "learning_rate": 3.165e-05, + "loss": 0.253, + "step": 425 + }, + { + "epoch": 4.5, + "learning_rate": 3.3524999999999995e-05, + "loss": 0.2526, + "step": 450 + }, + { + "epoch": 4.75, + "learning_rate": 3.539999999999999e-05, + "loss": 0.2463, + "step": 475 + }, + { + "epoch": 5.0, + "learning_rate": 3.7275e-05, + "loss": 0.2471, + "step": 500 + }, + { + "epoch": 5.25, + "learning_rate": 3.747911392405063e-05, + "loss": 0.184, + "step": 525 + }, + { + "epoch": 5.5, + "learning_rate": 3.7455379746835444e-05, + "loss": 0.1809, + "step": 550 + }, + { + "epoch": 5.75, + "learning_rate": 3.743164556962025e-05, + "loss": 0.1773, + "step": 575 + }, + { + "epoch": 6.0, + "learning_rate": 3.7407911392405063e-05, + "loss": 0.1852, + "step": 600 + }, + { + "epoch": 6.25, + "learning_rate": 3.738417721518987e-05, + "loss": 0.1259, + "step": 625 + }, + { + "epoch": 6.5, + "learning_rate": 3.736044303797468e-05, + "loss": 0.1287, + "step": 650 + }, + { + "epoch": 6.75, + "learning_rate": 3.733670886075949e-05, + "loss": 0.1289, + "step": 675 + }, + { + "epoch": 7.0, + "learning_rate": 3.73129746835443e-05, + "loss": 0.1312, + "step": 700 + }, + { + "epoch": 7.25, + "learning_rate": 3.7289240506329114e-05, + "loss": 0.0865, + "step": 725 + }, + { + "epoch": 7.5, + "learning_rate": 3.726550632911392e-05, + "loss": 0.0904, + "step": 750 + }, + { + "epoch": 7.75, + "learning_rate": 3.724177215189873e-05, + "loss": 0.0907, + "step": 775 + }, + { + "epoch": 8.0, + "learning_rate": 3.721803797468354e-05, + "loss": 0.0926, + "step": 800 + }, + { + "epoch": 8.25, + "learning_rate": 3.719430379746835e-05, + "loss": 0.0587, + "step": 825 + }, + { + "epoch": 8.5, + "learning_rate": 3.717056962025316e-05, + "loss": 0.061, + "step": 850 + }, + { + "epoch": 8.75, + "learning_rate": 3.714683544303797e-05, + "loss": 0.0625, + "step": 875 + }, + { + "epoch": 9.0, + "learning_rate": 3.7123101265822784e-05, + "loss": 0.0624, + "step": 900 + }, + { + "epoch": 9.25, + "learning_rate": 3.709936708860759e-05, + "loss": 0.0386, + "step": 925 + }, + { + "epoch": 9.5, + "learning_rate": 3.70756329113924e-05, + "loss": 0.0405, + "step": 950 + }, + { + "epoch": 9.75, + "learning_rate": 3.705189873417721e-05, + "loss": 0.0415, + "step": 975 + }, + { + "epoch": 10.0, + "learning_rate": 3.702816455696202e-05, + "loss": 0.0426, + "step": 1000 + }, + { + "epoch": 10.0, + "eval_loss": 0.34511512517929077, + "eval_runtime": 91.8649, + "eval_samples_per_second": 138.704, + "eval_steps_per_second": 1.089, + "eval_wer": 23.200328278880725, + "step": 1000 + }, + { + "epoch": 10.25, + "learning_rate": 3.700443037974683e-05, + "loss": 0.0259, + "step": 1025 + }, + { + "epoch": 10.5, + "learning_rate": 3.698069620253164e-05, + "loss": 0.0259, + "step": 1050 + }, + { + "epoch": 10.75, + "learning_rate": 3.6956962025316454e-05, + "loss": 0.0277, + "step": 1075 + }, + { + "epoch": 11.0, + "learning_rate": 3.693322784810126e-05, + "loss": 0.0279, + "step": 1100 + }, + { + "epoch": 11.25, + "learning_rate": 3.690949367088607e-05, + "loss": 0.0177, + "step": 1125 + }, + { + "epoch": 11.5, + "learning_rate": 3.688575949367088e-05, + "loss": 0.0184, + "step": 1150 + }, + { + "epoch": 11.75, + "learning_rate": 3.686202531645569e-05, + "loss": 0.018, + "step": 1175 + }, + { + "epoch": 12.0, + "learning_rate": 3.68382911392405e-05, + "loss": 0.019, + "step": 1200 + }, + { + "epoch": 12.25, + "learning_rate": 3.681455696202531e-05, + "loss": 0.0122, + "step": 1225 + }, + { + "epoch": 12.5, + "learning_rate": 3.6790822784810124e-05, + "loss": 0.0136, + "step": 1250 + }, + { + "epoch": 12.75, + "learning_rate": 3.676708860759493e-05, + "loss": 0.0136, + "step": 1275 + }, + { + "epoch": 13.0, + "learning_rate": 3.674335443037974e-05, + "loss": 0.014, + "step": 1300 + }, + { + "epoch": 13.25, + "learning_rate": 3.6719620253164556e-05, + "loss": 0.0097, + "step": 1325 + }, + { + "epoch": 13.5, + "learning_rate": 3.669588607594937e-05, + "loss": 0.0096, + "step": 1350 + }, + { + "epoch": 13.75, + "learning_rate": 3.6672151898734175e-05, + "loss": 0.0087, + "step": 1375 + }, + { + "epoch": 14.0, + "learning_rate": 3.664841772151899e-05, + "loss": 0.0088, + "step": 1400 + }, + { + "epoch": 14.25, + "learning_rate": 3.6624683544303794e-05, + "loss": 0.0064, + "step": 1425 + }, + { + "epoch": 14.5, + "learning_rate": 3.660094936708861e-05, + "loss": 0.0064, + "step": 1450 + }, + { + "epoch": 14.75, + "learning_rate": 3.657721518987341e-05, + "loss": 0.0063, + "step": 1475 + }, + { + "epoch": 15.0, + "learning_rate": 3.6553481012658226e-05, + "loss": 0.0063, + "step": 1500 + }, + { + "epoch": 15.25, + "learning_rate": 3.652974683544304e-05, + "loss": 0.0052, + "step": 1525 + }, + { + "epoch": 15.5, + "learning_rate": 3.6506012658227845e-05, + "loss": 0.0057, + "step": 1550 + }, + { + "epoch": 15.75, + "learning_rate": 3.648227848101266e-05, + "loss": 0.0059, + "step": 1575 + }, + { + "epoch": 16.0, + "learning_rate": 3.6458544303797464e-05, + "loss": 0.0057, + "step": 1600 + }, + { + "epoch": 16.25, + "learning_rate": 3.643481012658228e-05, + "loss": 0.0044, + "step": 1625 + }, + { + "epoch": 16.5, + "learning_rate": 3.641107594936708e-05, + "loss": 0.0044, + "step": 1650 + }, + { + "epoch": 16.75, + "learning_rate": 3.6387341772151896e-05, + "loss": 0.0047, + "step": 1675 + }, + { + "epoch": 17.0, + "learning_rate": 3.636360759493671e-05, + "loss": 0.0051, + "step": 1700 + }, + { + "epoch": 17.25, + "learning_rate": 3.6339873417721515e-05, + "loss": 0.0047, + "step": 1725 + }, + { + "epoch": 17.5, + "learning_rate": 3.631613924050633e-05, + "loss": 0.0052, + "step": 1750 + }, + { + "epoch": 17.75, + "learning_rate": 3.6292405063291134e-05, + "loss": 0.006, + "step": 1775 + }, + { + "epoch": 18.0, + "learning_rate": 3.626867088607595e-05, + "loss": 0.0079, + "step": 1800 + }, + { + "epoch": 18.25, + "learning_rate": 3.624493670886075e-05, + "loss": 0.0061, + "step": 1825 + }, + { + "epoch": 18.5, + "learning_rate": 3.6221202531645566e-05, + "loss": 0.0067, + "step": 1850 + }, + { + "epoch": 18.75, + "learning_rate": 3.619746835443038e-05, + "loss": 0.007, + "step": 1875 + }, + { + "epoch": 19.0, + "learning_rate": 3.6173734177215185e-05, + "loss": 0.0083, + "step": 1900 + }, + { + "epoch": 19.25, + "learning_rate": 3.615e-05, + "loss": 0.0072, + "step": 1925 + }, + { + "epoch": 19.5, + "learning_rate": 3.6126265822784804e-05, + "loss": 0.0079, + "step": 1950 + }, + { + "epoch": 19.75, + "learning_rate": 3.610253164556962e-05, + "loss": 0.0078, + "step": 1975 + }, + { + "epoch": 20.0, + "learning_rate": 3.607879746835442e-05, + "loss": 0.0077, + "step": 2000 + }, + { + "epoch": 20.0, + "eval_loss": 0.41232848167419434, + "eval_runtime": 81.6367, + "eval_samples_per_second": 156.082, + "eval_steps_per_second": 1.225, + "eval_wer": 22.605322807566047, + "step": 2000 + }, + { + "epoch": 20.25, + "learning_rate": 3.6055063291139236e-05, + "loss": 0.0059, + "step": 2025 + }, + { + "epoch": 20.5, + "learning_rate": 3.603132911392405e-05, + "loss": 0.0063, + "step": 2050 + }, + { + "epoch": 20.75, + "learning_rate": 3.6007594936708855e-05, + "loss": 0.0078, + "step": 2075 + }, + { + "epoch": 21.0, + "learning_rate": 3.598386075949367e-05, + "loss": 0.0081, + "step": 2100 + }, + { + "epoch": 21.25, + "learning_rate": 3.596012658227848e-05, + "loss": 0.0065, + "step": 2125 + }, + { + "epoch": 21.5, + "learning_rate": 3.593639240506329e-05, + "loss": 0.0078, + "step": 2150 + }, + { + "epoch": 21.75, + "learning_rate": 3.59126582278481e-05, + "loss": 0.008, + "step": 2175 + }, + { + "epoch": 22.0, + "learning_rate": 3.588892405063291e-05, + "loss": 0.0081, + "step": 2200 + }, + { + "epoch": 22.25, + "learning_rate": 3.586518987341772e-05, + "loss": 0.0071, + "step": 2225 + }, + { + "epoch": 22.5, + "learning_rate": 3.584145569620253e-05, + "loss": 0.007, + "step": 2250 + }, + { + "epoch": 22.75, + "learning_rate": 3.581772151898734e-05, + "loss": 0.0068, + "step": 2275 + }, + { + "epoch": 23.0, + "learning_rate": 3.579398734177215e-05, + "loss": 0.0073, + "step": 2300 + }, + { + "epoch": 23.25, + "learning_rate": 3.577025316455696e-05, + "loss": 0.0052, + "step": 2325 + }, + { + "epoch": 23.5, + "learning_rate": 3.574651898734177e-05, + "loss": 0.0051, + "step": 2350 + }, + { + "epoch": 23.75, + "learning_rate": 3.572278481012658e-05, + "loss": 0.0054, + "step": 2375 + }, + { + "epoch": 24.0, + "learning_rate": 3.569905063291139e-05, + "loss": 0.0053, + "step": 2400 + }, + { + "epoch": 24.25, + "learning_rate": 3.56753164556962e-05, + "loss": 0.0045, + "step": 2425 + }, + { + "epoch": 24.5, + "learning_rate": 3.565158227848101e-05, + "loss": 0.0044, + "step": 2450 + }, + { + "epoch": 24.75, + "learning_rate": 3.562784810126582e-05, + "loss": 0.004, + "step": 2475 + }, + { + "epoch": 25.0, + "learning_rate": 3.5604113924050627e-05, + "loss": 0.0044, + "step": 2500 + }, + { + "epoch": 25.25, + "learning_rate": 3.558037974683544e-05, + "loss": 0.0033, + "step": 2525 + }, + { + "epoch": 25.5, + "learning_rate": 3.555664556962025e-05, + "loss": 0.0029, + "step": 2550 + }, + { + "epoch": 25.75, + "learning_rate": 3.553291139240506e-05, + "loss": 0.0028, + "step": 2575 + }, + { + "epoch": 26.0, + "learning_rate": 3.550917721518987e-05, + "loss": 0.0033, + "step": 2600 + }, + { + "epoch": 26.25, + "learning_rate": 3.548544303797468e-05, + "loss": 0.0031, + "step": 2625 + }, + { + "epoch": 26.5, + "learning_rate": 3.546170886075949e-05, + "loss": 0.0028, + "step": 2650 + }, + { + "epoch": 26.75, + "learning_rate": 3.5437974683544297e-05, + "loss": 0.0028, + "step": 2675 + }, + { + "epoch": 27.0, + "learning_rate": 3.541424050632911e-05, + "loss": 0.0027, + "step": 2700 + }, + { + "epoch": 27.25, + "learning_rate": 3.539050632911392e-05, + "loss": 0.0025, + "step": 2725 + }, + { + "epoch": 27.5, + "learning_rate": 3.536677215189873e-05, + "loss": 0.0028, + "step": 2750 + }, + { + "epoch": 27.75, + "learning_rate": 3.534303797468354e-05, + "loss": 0.0025, + "step": 2775 + }, + { + "epoch": 28.0, + "learning_rate": 3.531930379746835e-05, + "loss": 0.0024, + "step": 2800 + }, + { + "epoch": 28.25, + "learning_rate": 3.529556962025316e-05, + "loss": 0.002, + "step": 2825 + }, + { + "epoch": 28.5, + "learning_rate": 3.527183544303797e-05, + "loss": 0.0021, + "step": 2850 + }, + { + "epoch": 28.75, + "learning_rate": 3.524810126582278e-05, + "loss": 0.0019, + "step": 2875 + }, + { + "epoch": 29.0, + "learning_rate": 3.522436708860759e-05, + "loss": 0.002, + "step": 2900 + }, + { + "epoch": 29.25, + "learning_rate": 3.5200632911392405e-05, + "loss": 0.0015, + "step": 2925 + }, + { + "epoch": 29.5, + "learning_rate": 3.517689873417721e-05, + "loss": 0.0014, + "step": 2950 + }, + { + "epoch": 29.75, + "learning_rate": 3.5153164556962024e-05, + "loss": 0.0015, + "step": 2975 + }, + { + "epoch": 30.0, + "learning_rate": 3.512943037974684e-05, + "loss": 0.0013, + "step": 3000 + }, + { + "epoch": 30.0, + "eval_loss": 0.4287857711315155, + "eval_runtime": 80.2692, + "eval_samples_per_second": 158.741, + "eval_steps_per_second": 1.246, + "eval_wer": 21.19645927778646, + "step": 3000 + }, + { + "epoch": 30.25, + "learning_rate": 3.510569620253164e-05, + "loss": 0.001, + "step": 3025 + }, + { + "epoch": 30.5, + "learning_rate": 3.5081962025316456e-05, + "loss": 0.0009, + "step": 3050 + }, + { + "epoch": 30.75, + "learning_rate": 3.505822784810126e-05, + "loss": 0.0009, + "step": 3075 + }, + { + "epoch": 31.0, + "learning_rate": 3.5034493670886075e-05, + "loss": 0.0009, + "step": 3100 + }, + { + "epoch": 31.25, + "learning_rate": 3.501075949367088e-05, + "loss": 0.0008, + "step": 3125 + }, + { + "epoch": 31.5, + "learning_rate": 3.4987025316455694e-05, + "loss": 0.0008, + "step": 3150 + }, + { + "epoch": 31.75, + "learning_rate": 3.496329113924051e-05, + "loss": 0.0008, + "step": 3175 + }, + { + "epoch": 32.0, + "learning_rate": 3.493955696202531e-05, + "loss": 0.0008, + "step": 3200 + }, + { + "epoch": 32.25, + "learning_rate": 3.4915822784810126e-05, + "loss": 0.0007, + "step": 3225 + }, + { + "epoch": 32.5, + "learning_rate": 3.489208860759493e-05, + "loss": 0.0007, + "step": 3250 + }, + { + "epoch": 32.75, + "learning_rate": 3.4868354430379745e-05, + "loss": 0.0007, + "step": 3275 + }, + { + "epoch": 33.0, + "learning_rate": 3.484462025316455e-05, + "loss": 0.0007, + "step": 3300 + }, + { + "epoch": 33.25, + "learning_rate": 3.4820886075949364e-05, + "loss": 0.0006, + "step": 3325 + }, + { + "epoch": 33.5, + "learning_rate": 3.479715189873418e-05, + "loss": 0.0006, + "step": 3350 + }, + { + "epoch": 33.75, + "learning_rate": 3.477341772151898e-05, + "loss": 0.0006, + "step": 3375 + }, + { + "epoch": 34.0, + "learning_rate": 3.4749683544303796e-05, + "loss": 0.0006, + "step": 3400 + }, + { + "epoch": 34.25, + "learning_rate": 3.47259493670886e-05, + "loss": 0.0006, + "step": 3425 + }, + { + "epoch": 34.5, + "learning_rate": 3.4702215189873415e-05, + "loss": 0.0006, + "step": 3450 + }, + { + "epoch": 34.75, + "learning_rate": 3.467848101265822e-05, + "loss": 0.0006, + "step": 3475 + }, + { + "epoch": 35.0, + "learning_rate": 3.4654746835443034e-05, + "loss": 0.0006, + "step": 3500 + }, + { + "epoch": 35.25, + "learning_rate": 3.463101265822785e-05, + "loss": 0.0005, + "step": 3525 + }, + { + "epoch": 35.5, + "learning_rate": 3.460727848101265e-05, + "loss": 0.0005, + "step": 3550 + }, + { + "epoch": 35.75, + "learning_rate": 3.4583544303797466e-05, + "loss": 0.0005, + "step": 3575 + }, + { + "epoch": 36.0, + "learning_rate": 3.455981012658227e-05, + "loss": 0.0005, + "step": 3600 + }, + { + "epoch": 36.25, + "learning_rate": 3.4536075949367085e-05, + "loss": 0.0005, + "step": 3625 + }, + { + "epoch": 36.5, + "learning_rate": 3.451234177215189e-05, + "loss": 0.0005, + "step": 3650 + }, + { + "epoch": 36.75, + "learning_rate": 3.4488607594936704e-05, + "loss": 0.0005, + "step": 3675 + }, + { + "epoch": 37.0, + "learning_rate": 3.446487341772152e-05, + "loss": 0.0005, + "step": 3700 + }, + { + "epoch": 37.25, + "learning_rate": 3.444113924050632e-05, + "loss": 0.0005, + "step": 3725 + }, + { + "epoch": 37.5, + "learning_rate": 3.4417405063291136e-05, + "loss": 0.0005, + "step": 3750 + }, + { + "epoch": 37.75, + "learning_rate": 3.439367088607595e-05, + "loss": 0.0005, + "step": 3775 + }, + { + "epoch": 38.0, + "learning_rate": 3.436993670886076e-05, + "loss": 0.0005, + "step": 3800 + }, + { + "epoch": 38.25, + "learning_rate": 3.434620253164557e-05, + "loss": 0.0004, + "step": 3825 + }, + { + "epoch": 38.5, + "learning_rate": 3.432246835443038e-05, + "loss": 0.0004, + "step": 3850 + }, + { + "epoch": 38.75, + "learning_rate": 3.429873417721519e-05, + "loss": 0.0004, + "step": 3875 + }, + { + "epoch": 39.0, + "learning_rate": 3.4275e-05, + "loss": 0.0004, + "step": 3900 + }, + { + "epoch": 39.25, + "learning_rate": 3.4251265822784806e-05, + "loss": 0.0004, + "step": 3925 + }, + { + "epoch": 39.5, + "learning_rate": 3.422753164556962e-05, + "loss": 0.0004, + "step": 3950 + }, + { + "epoch": 39.75, + "learning_rate": 3.420379746835443e-05, + "loss": 0.0004, + "step": 3975 + }, + { + "epoch": 40.0, + "learning_rate": 3.418006329113924e-05, + "loss": 0.0004, + "step": 4000 + }, + { + "epoch": 40.0, + "eval_loss": 0.45376476645469666, + "eval_runtime": 81.5739, + "eval_samples_per_second": 156.202, + "eval_steps_per_second": 1.226, + "eval_wer": 21.192551195873065, + "step": 4000 + }, + { + "epoch": 40.25, + "learning_rate": 3.415632911392405e-05, + "loss": 0.0004, + "step": 4025 + }, + { + "epoch": 40.5, + "learning_rate": 3.413259493670886e-05, + "loss": 0.0004, + "step": 4050 + }, + { + "epoch": 40.75, + "learning_rate": 3.410886075949367e-05, + "loss": 0.0004, + "step": 4075 + }, + { + "epoch": 41.0, + "learning_rate": 3.4085126582278476e-05, + "loss": 0.0004, + "step": 4100 + }, + { + "epoch": 41.25, + "learning_rate": 3.406139240506329e-05, + "loss": 0.0004, + "step": 4125 + }, + { + "epoch": 41.5, + "learning_rate": 3.40376582278481e-05, + "loss": 0.0004, + "step": 4150 + }, + { + "epoch": 41.75, + "learning_rate": 3.401392405063291e-05, + "loss": 0.0004, + "step": 4175 + }, + { + "epoch": 42.0, + "learning_rate": 3.399018987341772e-05, + "loss": 0.0004, + "step": 4200 + }, + { + "epoch": 42.25, + "learning_rate": 3.396645569620253e-05, + "loss": 0.0003, + "step": 4225 + }, + { + "epoch": 42.5, + "learning_rate": 3.394272151898734e-05, + "loss": 0.0003, + "step": 4250 + }, + { + "epoch": 42.75, + "learning_rate": 3.3918987341772146e-05, + "loss": 0.0004, + "step": 4275 + }, + { + "epoch": 43.0, + "learning_rate": 3.389525316455696e-05, + "loss": 0.0004, + "step": 4300 + }, + { + "epoch": 43.25, + "learning_rate": 3.387151898734177e-05, + "loss": 0.0003, + "step": 4325 + }, + { + "epoch": 43.5, + "learning_rate": 3.384778481012658e-05, + "loss": 0.0003, + "step": 4350 + }, + { + "epoch": 43.75, + "learning_rate": 3.382405063291139e-05, + "loss": 0.0003, + "step": 4375 + }, + { + "epoch": 44.0, + "learning_rate": 3.3800316455696197e-05, + "loss": 0.0003, + "step": 4400 + }, + { + "epoch": 44.25, + "learning_rate": 3.377658227848101e-05, + "loss": 0.0003, + "step": 4425 + }, + { + "epoch": 44.5, + "learning_rate": 3.3752848101265816e-05, + "loss": 0.0003, + "step": 4450 + }, + { + "epoch": 44.75, + "learning_rate": 3.372911392405063e-05, + "loss": 0.0003, + "step": 4475 + }, + { + "epoch": 45.0, + "learning_rate": 3.370537974683544e-05, + "loss": 0.0003, + "step": 4500 + }, + { + "epoch": 45.25, + "learning_rate": 3.368164556962025e-05, + "loss": 0.0003, + "step": 4525 + }, + { + "epoch": 45.5, + "learning_rate": 3.365791139240506e-05, + "loss": 0.0003, + "step": 4550 + }, + { + "epoch": 45.75, + "learning_rate": 3.363417721518987e-05, + "loss": 0.0003, + "step": 4575 + }, + { + "epoch": 46.0, + "learning_rate": 3.361044303797468e-05, + "loss": 0.0003, + "step": 4600 + }, + { + "epoch": 46.25, + "learning_rate": 3.358670886075949e-05, + "loss": 0.0003, + "step": 4625 + }, + { + "epoch": 46.5, + "learning_rate": 3.3562974683544305e-05, + "loss": 0.0003, + "step": 4650 + }, + { + "epoch": 46.75, + "learning_rate": 3.353924050632911e-05, + "loss": 0.0003, + "step": 4675 + }, + { + "epoch": 47.0, + "learning_rate": 3.3515506329113924e-05, + "loss": 0.0003, + "step": 4700 + }, + { + "epoch": 47.25, + "learning_rate": 3.349177215189873e-05, + "loss": 0.0003, + "step": 4725 + }, + { + "epoch": 47.5, + "learning_rate": 3.346803797468354e-05, + "loss": 0.0003, + "step": 4750 + }, + { + "epoch": 47.75, + "learning_rate": 3.344430379746835e-05, + "loss": 0.0003, + "step": 4775 + }, + { + "epoch": 48.0, + "learning_rate": 3.342056962025316e-05, + "loss": 0.0003, + "step": 4800 + }, + { + "epoch": 48.25, + "learning_rate": 3.3396835443037975e-05, + "loss": 0.0002, + "step": 4825 + }, + { + "epoch": 48.5, + "learning_rate": 3.337310126582278e-05, + "loss": 0.0002, + "step": 4850 + }, + { + "epoch": 48.75, + "learning_rate": 3.3349367088607594e-05, + "loss": 0.0002, + "step": 4875 + }, + { + "epoch": 49.0, + "learning_rate": 3.33256329113924e-05, + "loss": 0.0003, + "step": 4900 + }, + { + "epoch": 49.25, + "learning_rate": 3.330189873417721e-05, + "loss": 0.0002, + "step": 4925 + }, + { + "epoch": 49.5, + "learning_rate": 3.327816455696202e-05, + "loss": 0.0002, + "step": 4950 + }, + { + "epoch": 49.75, + "learning_rate": 3.325443037974683e-05, + "loss": 0.0002, + "step": 4975 + }, + { + "epoch": 50.0, + "learning_rate": 3.3230696202531645e-05, + "loss": 0.0003, + "step": 5000 + }, + { + "epoch": 50.0, + "eval_loss": 0.475699245929718, + "eval_runtime": 80.4553, + "eval_samples_per_second": 158.374, + "eval_steps_per_second": 1.243, + "eval_wer": 21.180826950132875, + "step": 5000 + }, + { + "epoch": 50.25, + "learning_rate": 3.320696202531645e-05, + "loss": 0.0002, + "step": 5025 + }, + { + "epoch": 50.5, + "learning_rate": 3.3183227848101264e-05, + "loss": 0.0002, + "step": 5050 + }, + { + "epoch": 50.75, + "learning_rate": 3.315949367088607e-05, + "loss": 0.0002, + "step": 5075 + }, + { + "epoch": 51.0, + "learning_rate": 3.313575949367088e-05, + "loss": 0.0002, + "step": 5100 + }, + { + "epoch": 51.25, + "learning_rate": 3.311202531645569e-05, + "loss": 0.0002, + "step": 5125 + }, + { + "epoch": 51.5, + "learning_rate": 3.30882911392405e-05, + "loss": 0.0002, + "step": 5150 + }, + { + "epoch": 51.75, + "learning_rate": 3.3064556962025315e-05, + "loss": 0.0002, + "step": 5175 + }, + { + "epoch": 52.0, + "learning_rate": 3.304082278481012e-05, + "loss": 0.0002, + "step": 5200 + }, + { + "epoch": 52.25, + "learning_rate": 3.3017088607594934e-05, + "loss": 0.0002, + "step": 5225 + }, + { + "epoch": 52.5, + "learning_rate": 3.299335443037974e-05, + "loss": 0.0002, + "step": 5250 + }, + { + "epoch": 52.75, + "learning_rate": 3.296962025316455e-05, + "loss": 0.0002, + "step": 5275 + }, + { + "epoch": 53.0, + "learning_rate": 3.2945886075949366e-05, + "loss": 0.0002, + "step": 5300 + }, + { + "epoch": 53.25, + "learning_rate": 3.292215189873417e-05, + "loss": 0.0002, + "step": 5325 + }, + { + "epoch": 53.5, + "learning_rate": 3.2898417721518985e-05, + "loss": 0.0002, + "step": 5350 + }, + { + "epoch": 53.75, + "learning_rate": 3.287468354430379e-05, + "loss": 0.0002, + "step": 5375 + }, + { + "epoch": 54.0, + "learning_rate": 3.2850949367088604e-05, + "loss": 0.0002, + "step": 5400 + }, + { + "epoch": 54.25, + "learning_rate": 3.282721518987342e-05, + "loss": 0.0002, + "step": 5425 + }, + { + "epoch": 54.5, + "learning_rate": 3.280348101265823e-05, + "loss": 0.0002, + "step": 5450 + }, + { + "epoch": 54.75, + "learning_rate": 3.2779746835443036e-05, + "loss": 0.0002, + "step": 5475 + }, + { + "epoch": 55.0, + "learning_rate": 3.275601265822785e-05, + "loss": 0.0002, + "step": 5500 + }, + { + "epoch": 55.25, + "learning_rate": 3.2732278481012655e-05, + "loss": 0.0002, + "step": 5525 + }, + { + "epoch": 55.5, + "learning_rate": 3.270854430379747e-05, + "loss": 0.0002, + "step": 5550 + }, + { + "epoch": 55.75, + "learning_rate": 3.2684810126582274e-05, + "loss": 0.0002, + "step": 5575 + }, + { + "epoch": 56.0, + "learning_rate": 3.266107594936709e-05, + "loss": 0.0002, + "step": 5600 + }, + { + "epoch": 56.25, + "learning_rate": 3.26373417721519e-05, + "loss": 0.0002, + "step": 5625 + }, + { + "epoch": 56.5, + "learning_rate": 3.2613607594936706e-05, + "loss": 0.0002, + "step": 5650 + }, + { + "epoch": 56.75, + "learning_rate": 3.258987341772152e-05, + "loss": 0.0002, + "step": 5675 + }, + { + "epoch": 57.0, + "learning_rate": 3.2566139240506325e-05, + "loss": 0.0002, + "step": 5700 + }, + { + "epoch": 57.25, + "learning_rate": 3.254240506329114e-05, + "loss": 0.0002, + "step": 5725 + }, + { + "epoch": 57.5, + "learning_rate": 3.2518670886075944e-05, + "loss": 0.0002, + "step": 5750 + }, + { + "epoch": 57.75, + "learning_rate": 3.249493670886076e-05, + "loss": 0.0002, + "step": 5775 + }, + { + "epoch": 58.0, + "learning_rate": 3.247120253164557e-05, + "loss": 0.0054, + "step": 5800 + }, + { + "epoch": 58.25, + "learning_rate": 3.2447468354430376e-05, + "loss": 0.0387, + "step": 5825 + }, + { + "epoch": 58.5, + "learning_rate": 3.242373417721519e-05, + "loss": 0.0443, + "step": 5850 + }, + { + "epoch": 58.75, + "learning_rate": 3.2399999999999995e-05, + "loss": 0.0406, + "step": 5875 + }, + { + "epoch": 59.0, + "learning_rate": 3.237626582278481e-05, + "loss": 0.0352, + "step": 5900 + }, + { + "epoch": 59.25, + "learning_rate": 3.2352531645569614e-05, + "loss": 0.0224, + "step": 5925 + }, + { + "epoch": 59.5, + "learning_rate": 3.232879746835443e-05, + "loss": 0.0213, + "step": 5950 + }, + { + "epoch": 59.75, + "learning_rate": 3.230506329113924e-05, + "loss": 0.0218, + "step": 5975 + }, + { + "epoch": 60.0, + "learning_rate": 3.2281329113924046e-05, + "loss": 0.0206, + "step": 6000 + }, + { + "epoch": 60.0, + "eval_loss": 0.41719070076942444, + "eval_runtime": 86.1367, + "eval_samples_per_second": 147.928, + "eval_steps_per_second": 1.161, + "eval_wer": 22.275089885884007, + "step": 6000 + }, + { + "epoch": 60.25, + "learning_rate": 3.225759493670886e-05, + "loss": 0.0123, + "step": 6025 + }, + { + "epoch": 60.5, + "learning_rate": 3.2233860759493665e-05, + "loss": 0.0125, + "step": 6050 + }, + { + "epoch": 60.75, + "learning_rate": 3.221012658227848e-05, + "loss": 0.0126, + "step": 6075 + }, + { + "epoch": 61.0, + "learning_rate": 3.2186392405063284e-05, + "loss": 0.0126, + "step": 6100 + }, + { + "epoch": 61.25, + "learning_rate": 3.2162658227848097e-05, + "loss": 0.0078, + "step": 6125 + }, + { + "epoch": 61.5, + "learning_rate": 3.213892405063291e-05, + "loss": 0.0072, + "step": 6150 + }, + { + "epoch": 61.75, + "learning_rate": 3.2115189873417716e-05, + "loss": 0.0068, + "step": 6175 + }, + { + "epoch": 62.0, + "learning_rate": 3.209145569620253e-05, + "loss": 0.0073, + "step": 6200 + }, + { + "epoch": 62.25, + "learning_rate": 3.206772151898734e-05, + "loss": 0.0041, + "step": 6225 + }, + { + "epoch": 62.5, + "learning_rate": 3.2043987341772154e-05, + "loss": 0.004, + "step": 6250 + }, + { + "epoch": 62.75, + "learning_rate": 3.202025316455696e-05, + "loss": 0.0039, + "step": 6275 + }, + { + "epoch": 63.0, + "learning_rate": 3.199651898734177e-05, + "loss": 0.0036, + "step": 6300 + }, + { + "epoch": 63.25, + "learning_rate": 3.197278481012658e-05, + "loss": 0.0022, + "step": 6325 + }, + { + "epoch": 63.5, + "learning_rate": 3.194905063291139e-05, + "loss": 0.0016, + "step": 6350 + }, + { + "epoch": 63.75, + "learning_rate": 3.19253164556962e-05, + "loss": 0.0019, + "step": 6375 + }, + { + "epoch": 64.0, + "learning_rate": 3.190158227848101e-05, + "loss": 0.0017, + "step": 6400 + }, + { + "epoch": 64.25, + "learning_rate": 3.1877848101265824e-05, + "loss": 0.001, + "step": 6425 + }, + { + "epoch": 64.5, + "learning_rate": 3.185411392405063e-05, + "loss": 0.0015, + "step": 6450 + }, + { + "epoch": 64.75, + "learning_rate": 3.183037974683544e-05, + "loss": 0.001, + "step": 6475 + }, + { + "epoch": 65.0, + "learning_rate": 3.180664556962025e-05, + "loss": 0.0013, + "step": 6500 + }, + { + "epoch": 65.25, + "learning_rate": 3.178291139240506e-05, + "loss": 0.0008, + "step": 6525 + }, + { + "epoch": 65.5, + "learning_rate": 3.175917721518987e-05, + "loss": 0.0008, + "step": 6550 + }, + { + "epoch": 65.75, + "learning_rate": 3.173544303797468e-05, + "loss": 0.001, + "step": 6575 + }, + { + "epoch": 66.0, + "learning_rate": 3.1711708860759494e-05, + "loss": 0.0008, + "step": 6600 + }, + { + "epoch": 66.25, + "learning_rate": 3.16879746835443e-05, + "loss": 0.0005, + "step": 6625 + }, + { + "epoch": 66.5, + "learning_rate": 3.166424050632911e-05, + "loss": 0.0007, + "step": 6650 + }, + { + "epoch": 66.75, + "learning_rate": 3.164050632911392e-05, + "loss": 0.0007, + "step": 6675 + }, + { + "epoch": 67.0, + "learning_rate": 3.161677215189873e-05, + "loss": 0.0006, + "step": 6700 + }, + { + "epoch": 67.25, + "learning_rate": 3.159303797468354e-05, + "loss": 0.0004, + "step": 6725 + }, + { + "epoch": 67.5, + "learning_rate": 3.156930379746835e-05, + "loss": 0.0004, + "step": 6750 + }, + { + "epoch": 67.75, + "learning_rate": 3.1545569620253164e-05, + "loss": 0.0004, + "step": 6775 + }, + { + "epoch": 68.0, + "learning_rate": 3.152183544303797e-05, + "loss": 0.0004, + "step": 6800 + }, + { + "epoch": 68.25, + "learning_rate": 3.149810126582278e-05, + "loss": 0.0003, + "step": 6825 + }, + { + "epoch": 68.5, + "learning_rate": 3.147436708860759e-05, + "loss": 0.0003, + "step": 6850 + }, + { + "epoch": 68.75, + "learning_rate": 3.14506329113924e-05, + "loss": 0.0003, + "step": 6875 + }, + { + "epoch": 69.0, + "learning_rate": 3.142689873417721e-05, + "loss": 0.0003, + "step": 6900 + }, + { + "epoch": 69.25, + "learning_rate": 3.140316455696202e-05, + "loss": 0.0003, + "step": 6925 + }, + { + "epoch": 69.5, + "learning_rate": 3.1379430379746834e-05, + "loss": 0.0003, + "step": 6950 + }, + { + "epoch": 69.75, + "learning_rate": 3.135569620253164e-05, + "loss": 0.0003, + "step": 6975 + }, + { + "epoch": 70.0, + "learning_rate": 3.133196202531645e-05, + "loss": 0.0003, + "step": 7000 + }, + { + "epoch": 70.0, + "eval_loss": 0.43740585446357727, + "eval_runtime": 80.2604, + "eval_samples_per_second": 158.758, + "eval_steps_per_second": 1.246, + "eval_wer": 19.513052993590748, + "step": 7000 + }, + { + "epoch": 70.25, + "learning_rate": 3.130822784810126e-05, + "loss": 0.0003, + "step": 7025 + }, + { + "epoch": 70.5, + "learning_rate": 3.128449367088607e-05, + "loss": 0.0003, + "step": 7050 + }, + { + "epoch": 70.75, + "learning_rate": 3.1260759493670885e-05, + "loss": 0.0003, + "step": 7075 + }, + { + "epoch": 71.0, + "learning_rate": 3.12370253164557e-05, + "loss": 0.0003, + "step": 7100 + }, + { + "epoch": 71.25, + "learning_rate": 3.1213291139240504e-05, + "loss": 0.0002, + "step": 7125 + }, + { + "epoch": 71.5, + "learning_rate": 3.118955696202532e-05, + "loss": 0.0002, + "step": 7150 + }, + { + "epoch": 71.75, + "learning_rate": 3.116582278481012e-05, + "loss": 0.0003, + "step": 7175 + }, + { + "epoch": 72.0, + "learning_rate": 3.1142088607594936e-05, + "loss": 0.0002, + "step": 7200 + }, + { + "epoch": 72.25, + "learning_rate": 3.111835443037974e-05, + "loss": 0.0002, + "step": 7225 + }, + { + "epoch": 72.5, + "learning_rate": 3.1094620253164555e-05, + "loss": 0.0002, + "step": 7250 + }, + { + "epoch": 72.75, + "learning_rate": 3.107088607594937e-05, + "loss": 0.0002, + "step": 7275 + }, + { + "epoch": 73.0, + "learning_rate": 3.1047151898734174e-05, + "loss": 0.0002, + "step": 7300 + }, + { + "epoch": 73.25, + "learning_rate": 3.102341772151899e-05, + "loss": 0.0002, + "step": 7325 + }, + { + "epoch": 73.5, + "learning_rate": 3.099968354430379e-05, + "loss": 0.0002, + "step": 7350 + }, + { + "epoch": 73.75, + "learning_rate": 3.0975949367088606e-05, + "loss": 0.0002, + "step": 7375 + }, + { + "epoch": 74.0, + "learning_rate": 3.095221518987341e-05, + "loss": 0.0002, + "step": 7400 + }, + { + "epoch": 74.25, + "learning_rate": 3.0928481012658225e-05, + "loss": 0.0002, + "step": 7425 + }, + { + "epoch": 74.5, + "learning_rate": 3.090474683544304e-05, + "loss": 0.0002, + "step": 7450 + }, + { + "epoch": 74.75, + "learning_rate": 3.0881012658227844e-05, + "loss": 0.0002, + "step": 7475 + }, + { + "epoch": 75.0, + "learning_rate": 3.085727848101266e-05, + "loss": 0.0002, + "step": 7500 + }, + { + "epoch": 75.25, + "learning_rate": 3.083354430379746e-05, + "loss": 0.0002, + "step": 7525 + }, + { + "epoch": 75.5, + "learning_rate": 3.0809810126582276e-05, + "loss": 0.0002, + "step": 7550 + }, + { + "epoch": 75.75, + "learning_rate": 3.078607594936709e-05, + "loss": 0.0002, + "step": 7575 + }, + { + "epoch": 76.0, + "learning_rate": 3.0762341772151895e-05, + "loss": 0.0002, + "step": 7600 + }, + { + "epoch": 76.25, + "learning_rate": 3.073860759493671e-05, + "loss": 0.0002, + "step": 7625 + }, + { + "epoch": 76.5, + "learning_rate": 3.0714873417721514e-05, + "loss": 0.0002, + "step": 7650 + }, + { + "epoch": 76.75, + "learning_rate": 3.069113924050633e-05, + "loss": 0.0002, + "step": 7675 + }, + { + "epoch": 77.0, + "learning_rate": 3.066740506329113e-05, + "loss": 0.0002, + "step": 7700 + }, + { + "epoch": 77.25, + "learning_rate": 3.0643670886075946e-05, + "loss": 0.0002, + "step": 7725 + }, + { + "epoch": 77.5, + "learning_rate": 3.061993670886076e-05, + "loss": 0.0002, + "step": 7750 + }, + { + "epoch": 77.75, + "learning_rate": 3.0596202531645565e-05, + "loss": 0.0002, + "step": 7775 + }, + { + "epoch": 78.0, + "learning_rate": 3.057246835443038e-05, + "loss": 0.0002, + "step": 7800 + }, + { + "epoch": 78.25, + "learning_rate": 3.0548734177215184e-05, + "loss": 0.0002, + "step": 7825 + }, + { + "epoch": 78.5, + "learning_rate": 3.0524999999999997e-05, + "loss": 0.0002, + "step": 7850 + }, + { + "epoch": 78.75, + "learning_rate": 3.050126582278481e-05, + "loss": 0.0002, + "step": 7875 + }, + { + "epoch": 79.0, + "learning_rate": 3.047753164556962e-05, + "loss": 0.0002, + "step": 7900 + }, + { + "epoch": 79.25, + "learning_rate": 3.045379746835443e-05, + "loss": 0.0001, + "step": 7925 + }, + { + "epoch": 79.5, + "learning_rate": 3.0430063291139238e-05, + "loss": 0.0001, + "step": 7950 + }, + { + "epoch": 79.75, + "learning_rate": 3.040632911392405e-05, + "loss": 0.0001, + "step": 7975 + }, + { + "epoch": 80.0, + "learning_rate": 3.038259493670886e-05, + "loss": 0.0002, + "step": 8000 + }, + { + "epoch": 80.0, + "eval_loss": 0.4546888768672943, + "eval_runtime": 80.4032, + "eval_samples_per_second": 158.476, + "eval_steps_per_second": 1.244, + "eval_wer": 19.50914491167735, + "step": 8000 + }, + { + "epoch": 80.25, + "learning_rate": 3.035886075949367e-05, + "loss": 0.0001, + "step": 8025 + }, + { + "epoch": 80.5, + "learning_rate": 3.033512658227848e-05, + "loss": 0.0001, + "step": 8050 + }, + { + "epoch": 80.75, + "learning_rate": 3.031139240506329e-05, + "loss": 0.0001, + "step": 8075 + }, + { + "epoch": 81.0, + "learning_rate": 3.02876582278481e-05, + "loss": 0.0001, + "step": 8100 + }, + { + "epoch": 81.25, + "learning_rate": 3.0263924050632908e-05, + "loss": 0.0001, + "step": 8125 + }, + { + "epoch": 81.5, + "learning_rate": 3.024018987341772e-05, + "loss": 0.0001, + "step": 8150 + }, + { + "epoch": 81.75, + "learning_rate": 3.021645569620253e-05, + "loss": 0.0001, + "step": 8175 + }, + { + "epoch": 82.0, + "learning_rate": 3.019272151898734e-05, + "loss": 0.0001, + "step": 8200 + }, + { + "epoch": 82.25, + "learning_rate": 3.016898734177215e-05, + "loss": 0.0001, + "step": 8225 + }, + { + "epoch": 82.5, + "learning_rate": 3.014525316455696e-05, + "loss": 0.0001, + "step": 8250 + }, + { + "epoch": 82.75, + "learning_rate": 3.012151898734177e-05, + "loss": 0.0001, + "step": 8275 + }, + { + "epoch": 83.0, + "learning_rate": 3.0097784810126578e-05, + "loss": 0.0001, + "step": 8300 + }, + { + "epoch": 83.25, + "learning_rate": 3.007405063291139e-05, + "loss": 0.0001, + "step": 8325 + }, + { + "epoch": 83.5, + "learning_rate": 3.00503164556962e-05, + "loss": 0.0001, + "step": 8350 + }, + { + "epoch": 83.75, + "learning_rate": 3.002658227848101e-05, + "loss": 0.0001, + "step": 8375 + }, + { + "epoch": 84.0, + "learning_rate": 3.000284810126582e-05, + "loss": 0.0001, + "step": 8400 + }, + { + "epoch": 84.25, + "learning_rate": 2.997911392405063e-05, + "loss": 0.0001, + "step": 8425 + }, + { + "epoch": 84.5, + "learning_rate": 2.9955379746835438e-05, + "loss": 0.0001, + "step": 8450 + }, + { + "epoch": 84.75, + "learning_rate": 2.993164556962025e-05, + "loss": 0.0001, + "step": 8475 + }, + { + "epoch": 85.0, + "learning_rate": 2.990791139240506e-05, + "loss": 0.0001, + "step": 8500 + }, + { + "epoch": 85.25, + "learning_rate": 2.988417721518987e-05, + "loss": 0.0001, + "step": 8525 + }, + { + "epoch": 85.5, + "learning_rate": 2.986044303797468e-05, + "loss": 0.0001, + "step": 8550 + }, + { + "epoch": 85.75, + "learning_rate": 2.983670886075949e-05, + "loss": 0.0001, + "step": 8575 + }, + { + "epoch": 86.0, + "learning_rate": 2.98129746835443e-05, + "loss": 0.0001, + "step": 8600 + }, + { + "epoch": 86.25, + "learning_rate": 2.9789240506329108e-05, + "loss": 0.0001, + "step": 8625 + }, + { + "epoch": 86.5, + "learning_rate": 2.976550632911392e-05, + "loss": 0.0001, + "step": 8650 + }, + { + "epoch": 86.75, + "learning_rate": 2.974177215189873e-05, + "loss": 0.0001, + "step": 8675 + }, + { + "epoch": 87.0, + "learning_rate": 2.9718037974683544e-05, + "loss": 0.0001, + "step": 8700 + }, + { + "epoch": 87.25, + "learning_rate": 2.9694303797468353e-05, + "loss": 0.0001, + "step": 8725 + }, + { + "epoch": 87.5, + "learning_rate": 2.9670569620253163e-05, + "loss": 0.0001, + "step": 8750 + }, + { + "epoch": 87.75, + "learning_rate": 2.9646835443037972e-05, + "loss": 0.0001, + "step": 8775 + }, + { + "epoch": 88.0, + "learning_rate": 2.9623101265822785e-05, + "loss": 0.0001, + "step": 8800 + }, + { + "epoch": 88.25, + "learning_rate": 2.9599367088607594e-05, + "loss": 0.0001, + "step": 8825 + }, + { + "epoch": 88.5, + "learning_rate": 2.9575632911392404e-05, + "loss": 0.0001, + "step": 8850 + }, + { + "epoch": 88.75, + "learning_rate": 2.9551898734177213e-05, + "loss": 0.0001, + "step": 8875 + }, + { + "epoch": 89.0, + "learning_rate": 2.9528164556962023e-05, + "loss": 0.0001, + "step": 8900 + }, + { + "epoch": 89.25, + "learning_rate": 2.9504430379746832e-05, + "loss": 0.0001, + "step": 8925 + }, + { + "epoch": 89.5, + "learning_rate": 2.9480696202531645e-05, + "loss": 0.0001, + "step": 8950 + }, + { + "epoch": 89.75, + "learning_rate": 2.9456962025316455e-05, + "loss": 0.0001, + "step": 8975 + }, + { + "epoch": 90.0, + "learning_rate": 2.9433227848101264e-05, + "loss": 0.0001, + "step": 9000 + }, + { + "epoch": 90.0, + "eval_loss": 0.4697396755218506, + "eval_runtime": 80.422, + "eval_samples_per_second": 158.439, + "eval_steps_per_second": 1.243, + "eval_wer": 19.5062138502423, + "step": 9000 + }, + { + "epoch": 90.25, + "learning_rate": 2.9409493670886074e-05, + "loss": 0.0001, + "step": 9025 + }, + { + "epoch": 90.5, + "learning_rate": 2.9385759493670883e-05, + "loss": 0.0001, + "step": 9050 + }, + { + "epoch": 90.75, + "learning_rate": 2.9362025316455693e-05, + "loss": 0.0001, + "step": 9075 + }, + { + "epoch": 91.0, + "learning_rate": 2.9338291139240502e-05, + "loss": 0.0001, + "step": 9100 + }, + { + "epoch": 91.25, + "learning_rate": 2.9314556962025315e-05, + "loss": 0.0001, + "step": 9125 + }, + { + "epoch": 91.5, + "learning_rate": 2.9290822784810125e-05, + "loss": 0.0001, + "step": 9150 + }, + { + "epoch": 91.75, + "learning_rate": 2.9267088607594934e-05, + "loss": 0.0001, + "step": 9175 + }, + { + "epoch": 92.0, + "learning_rate": 2.9243354430379744e-05, + "loss": 0.0001, + "step": 9200 + }, + { + "epoch": 92.25, + "learning_rate": 2.9219620253164553e-05, + "loss": 0.0001, + "step": 9225 + }, + { + "epoch": 92.5, + "learning_rate": 2.9195886075949363e-05, + "loss": 0.0001, + "step": 9250 + }, + { + "epoch": 92.75, + "learning_rate": 2.9172151898734172e-05, + "loss": 0.0001, + "step": 9275 + }, + { + "epoch": 93.0, + "learning_rate": 2.9148417721518985e-05, + "loss": 0.0001, + "step": 9300 + }, + { + "epoch": 93.25, + "learning_rate": 2.9124683544303795e-05, + "loss": 0.0001, + "step": 9325 + }, + { + "epoch": 93.5, + "learning_rate": 2.9100949367088604e-05, + "loss": 0.0001, + "step": 9350 + }, + { + "epoch": 93.75, + "learning_rate": 2.9077215189873414e-05, + "loss": 0.0001, + "step": 9375 + }, + { + "epoch": 94.0, + "learning_rate": 2.9053481012658223e-05, + "loss": 0.0001, + "step": 9400 + }, + { + "epoch": 94.25, + "learning_rate": 2.9029746835443033e-05, + "loss": 0.0001, + "step": 9425 + }, + { + "epoch": 94.5, + "learning_rate": 2.9006012658227842e-05, + "loss": 0.0001, + "step": 9450 + }, + { + "epoch": 94.75, + "learning_rate": 2.8982278481012655e-05, + "loss": 0.0001, + "step": 9475 + }, + { + "epoch": 95.0, + "learning_rate": 2.8958544303797465e-05, + "loss": 0.0001, + "step": 9500 + }, + { + "epoch": 95.25, + "learning_rate": 2.8934810126582278e-05, + "loss": 0.0001, + "step": 9525 + }, + { + "epoch": 95.5, + "learning_rate": 2.8911075949367087e-05, + "loss": 0.0001, + "step": 9550 + }, + { + "epoch": 95.75, + "learning_rate": 2.8887341772151897e-05, + "loss": 0.0001, + "step": 9575 + }, + { + "epoch": 96.0, + "learning_rate": 2.886360759493671e-05, + "loss": 0.0001, + "step": 9600 + }, + { + "epoch": 96.25, + "learning_rate": 2.883987341772152e-05, + "loss": 0.0001, + "step": 9625 + }, + { + "epoch": 96.5, + "learning_rate": 2.881613924050633e-05, + "loss": 0.0001, + "step": 9650 + }, + { + "epoch": 96.75, + "learning_rate": 2.8792405063291138e-05, + "loss": 0.0001, + "step": 9675 + }, + { + "epoch": 97.0, + "learning_rate": 2.8768670886075948e-05, + "loss": 0.0001, + "step": 9700 + }, + { + "epoch": 97.25, + "learning_rate": 2.8744936708860757e-05, + "loss": 0.0001, + "step": 9725 + }, + { + "epoch": 97.5, + "learning_rate": 2.8721202531645567e-05, + "loss": 0.0001, + "step": 9750 + }, + { + "epoch": 97.75, + "learning_rate": 2.869746835443038e-05, + "loss": 0.0001, + "step": 9775 + }, + { + "epoch": 98.0, + "learning_rate": 2.867373417721519e-05, + "loss": 0.0001, + "step": 9800 + }, + { + "epoch": 98.25, + "learning_rate": 2.865e-05, + "loss": 0.0001, + "step": 9825 + }, + { + "epoch": 98.5, + "learning_rate": 2.8626265822784808e-05, + "loss": 0.0001, + "step": 9850 + }, + { + "epoch": 98.75, + "learning_rate": 2.8602531645569617e-05, + "loss": 0.0001, + "step": 9875 + }, + { + "epoch": 99.0, + "learning_rate": 2.8578797468354427e-05, + "loss": 0.0001, + "step": 9900 + }, + { + "epoch": 99.25, + "learning_rate": 2.8555063291139236e-05, + "loss": 0.0001, + "step": 9925 + }, + { + "epoch": 99.5, + "learning_rate": 2.853132911392405e-05, + "loss": 0.0001, + "step": 9950 + }, + { + "epoch": 99.75, + "learning_rate": 2.850759493670886e-05, + "loss": 0.0001, + "step": 9975 + }, + { + "epoch": 100.0, + "learning_rate": 2.848386075949367e-05, + "loss": 0.0001, + "step": 10000 + }, + { + "epoch": 100.0, + "eval_loss": 0.48529908061027527, + "eval_runtime": 80.5352, + "eval_samples_per_second": 158.217, + "eval_steps_per_second": 1.242, + "eval_wer": 19.51989213693919, + "step": 10000 + }, + { + "epoch": 100.25, + "learning_rate": 2.8460126582278478e-05, + "loss": 0.0001, + "step": 10025 + }, + { + "epoch": 100.5, + "learning_rate": 2.8436392405063287e-05, + "loss": 0.0001, + "step": 10050 + }, + { + "epoch": 100.75, + "learning_rate": 2.8412658227848097e-05, + "loss": 0.0001, + "step": 10075 + }, + { + "epoch": 101.0, + "learning_rate": 2.8388924050632906e-05, + "loss": 0.0001, + "step": 10100 + }, + { + "epoch": 101.25, + "learning_rate": 2.836518987341772e-05, + "loss": 0.0001, + "step": 10125 + }, + { + "epoch": 101.5, + "learning_rate": 2.834145569620253e-05, + "loss": 0.0001, + "step": 10150 + }, + { + "epoch": 101.75, + "learning_rate": 2.831772151898734e-05, + "loss": 0.0001, + "step": 10175 + }, + { + "epoch": 102.0, + "learning_rate": 2.8293987341772148e-05, + "loss": 0.0001, + "step": 10200 + }, + { + "epoch": 102.25, + "learning_rate": 2.8270253164556957e-05, + "loss": 0.0001, + "step": 10225 + }, + { + "epoch": 102.5, + "learning_rate": 2.8246518987341767e-05, + "loss": 0.0001, + "step": 10250 + }, + { + "epoch": 102.75, + "learning_rate": 2.8222784810126576e-05, + "loss": 0.0001, + "step": 10275 + }, + { + "epoch": 103.0, + "learning_rate": 2.819905063291139e-05, + "loss": 0.0001, + "step": 10300 + }, + { + "epoch": 103.25, + "learning_rate": 2.81753164556962e-05, + "loss": 0.0001, + "step": 10325 + }, + { + "epoch": 103.5, + "learning_rate": 2.815158227848101e-05, + "loss": 0.0001, + "step": 10350 + }, + { + "epoch": 103.75, + "learning_rate": 2.812784810126582e-05, + "loss": 0.0001, + "step": 10375 + }, + { + "epoch": 104.0, + "learning_rate": 2.810411392405063e-05, + "loss": 0.0, + "step": 10400 + }, + { + "epoch": 104.25, + "learning_rate": 2.8080379746835444e-05, + "loss": 0.0, + "step": 10425 + }, + { + "epoch": 104.5, + "learning_rate": 2.8056645569620253e-05, + "loss": 0.0, + "step": 10450 + }, + { + "epoch": 104.75, + "learning_rate": 2.8032911392405063e-05, + "loss": 0.0, + "step": 10475 + }, + { + "epoch": 105.0, + "learning_rate": 2.8009177215189872e-05, + "loss": 0.0001, + "step": 10500 + }, + { + "epoch": 105.25, + "learning_rate": 2.798544303797468e-05, + "loss": 0.0, + "step": 10525 + }, + { + "epoch": 105.5, + "learning_rate": 2.796170886075949e-05, + "loss": 0.0, + "step": 10550 + }, + { + "epoch": 105.75, + "learning_rate": 2.79379746835443e-05, + "loss": 0.0001, + "step": 10575 + }, + { + "epoch": 106.0, + "learning_rate": 2.7914240506329113e-05, + "loss": 0.0001, + "step": 10600 + }, + { + "epoch": 106.25, + "learning_rate": 2.7890506329113923e-05, + "loss": 0.0, + "step": 10625 + }, + { + "epoch": 106.5, + "learning_rate": 2.7866772151898732e-05, + "loss": 0.0, + "step": 10650 + }, + { + "epoch": 106.75, + "learning_rate": 2.7843037974683542e-05, + "loss": 0.0001, + "step": 10675 + }, + { + "epoch": 107.0, + "learning_rate": 2.781930379746835e-05, + "loss": 0.0, + "step": 10700 + }, + { + "epoch": 107.25, + "learning_rate": 2.779556962025316e-05, + "loss": 0.0, + "step": 10725 + }, + { + "epoch": 107.5, + "learning_rate": 2.777183544303797e-05, + "loss": 0.0, + "step": 10750 + }, + { + "epoch": 107.75, + "learning_rate": 2.7748101265822783e-05, + "loss": 0.0001, + "step": 10775 + }, + { + "epoch": 108.0, + "learning_rate": 2.7724367088607593e-05, + "loss": 0.0001, + "step": 10800 + }, + { + "epoch": 108.25, + "learning_rate": 2.7700632911392402e-05, + "loss": 0.0001, + "step": 10825 + }, + { + "epoch": 108.5, + "learning_rate": 2.7676898734177212e-05, + "loss": 0.0, + "step": 10850 + }, + { + "epoch": 108.75, + "learning_rate": 2.765316455696202e-05, + "loss": 0.0001, + "step": 10875 + }, + { + "epoch": 109.0, + "learning_rate": 2.762943037974683e-05, + "loss": 0.0, + "step": 10900 + }, + { + "epoch": 109.25, + "learning_rate": 2.7605696202531644e-05, + "loss": 0.0, + "step": 10925 + }, + { + "epoch": 109.5, + "learning_rate": 2.7581962025316453e-05, + "loss": 0.0, + "step": 10950 + }, + { + "epoch": 109.75, + "learning_rate": 2.7558227848101263e-05, + "loss": 0.0, + "step": 10975 + }, + { + "epoch": 110.0, + "learning_rate": 2.7534493670886072e-05, + "loss": 0.0001, + "step": 11000 + }, + { + "epoch": 110.0, + "eval_loss": 0.5008699297904968, + "eval_runtime": 80.5626, + "eval_samples_per_second": 158.163, + "eval_steps_per_second": 1.241, + "eval_wer": 19.568743160856652, + "step": 11000 + }, + { + "epoch": 110.25, + "learning_rate": 2.7510759493670882e-05, + "loss": 0.0, + "step": 11025 + }, + { + "epoch": 110.5, + "learning_rate": 2.748702531645569e-05, + "loss": 0.0, + "step": 11050 + }, + { + "epoch": 110.75, + "learning_rate": 2.74632911392405e-05, + "loss": 0.0, + "step": 11075 + }, + { + "epoch": 111.0, + "learning_rate": 2.7439556962025314e-05, + "loss": 0.0, + "step": 11100 + }, + { + "epoch": 111.25, + "learning_rate": 2.7415822784810123e-05, + "loss": 0.0, + "step": 11125 + }, + { + "epoch": 111.5, + "learning_rate": 2.7392088607594933e-05, + "loss": 0.0, + "step": 11150 + }, + { + "epoch": 111.75, + "learning_rate": 2.7368354430379746e-05, + "loss": 0.0, + "step": 11175 + }, + { + "epoch": 112.0, + "learning_rate": 2.7344620253164555e-05, + "loss": 0.0, + "step": 11200 + }, + { + "epoch": 112.25, + "learning_rate": 2.7320886075949368e-05, + "loss": 0.0, + "step": 11225 + }, + { + "epoch": 112.5, + "learning_rate": 2.7297151898734178e-05, + "loss": 0.0, + "step": 11250 + }, + { + "epoch": 112.75, + "learning_rate": 2.7273417721518987e-05, + "loss": 0.0, + "step": 11275 + }, + { + "epoch": 113.0, + "learning_rate": 2.7249683544303797e-05, + "loss": 0.0, + "step": 11300 + }, + { + "epoch": 113.25, + "learning_rate": 2.7225949367088606e-05, + "loss": 0.0, + "step": 11325 + }, + { + "epoch": 113.5, + "learning_rate": 2.7202215189873416e-05, + "loss": 0.0, + "step": 11350 + }, + { + "epoch": 113.75, + "learning_rate": 2.7178481012658225e-05, + "loss": 0.0, + "step": 11375 + }, + { + "epoch": 114.0, + "learning_rate": 2.7154746835443038e-05, + "loss": 0.0, + "step": 11400 + }, + { + "epoch": 114.25, + "learning_rate": 2.7131012658227848e-05, + "loss": 0.0, + "step": 11425 + }, + { + "epoch": 114.5, + "learning_rate": 2.7107278481012657e-05, + "loss": 0.0, + "step": 11450 + }, + { + "epoch": 114.75, + "learning_rate": 2.7083544303797467e-05, + "loss": 0.0, + "step": 11475 + }, + { + "epoch": 115.0, + "learning_rate": 2.7059810126582276e-05, + "loss": 0.0001, + "step": 11500 + }, + { + "epoch": 115.25, + "learning_rate": 2.7036075949367086e-05, + "loss": 0.0, + "step": 11525 + }, + { + "epoch": 115.5, + "learning_rate": 2.7012341772151895e-05, + "loss": 0.0001, + "step": 11550 + }, + { + "epoch": 115.75, + "learning_rate": 2.6988607594936708e-05, + "loss": 0.0, + "step": 11575 + }, + { + "epoch": 116.0, + "learning_rate": 2.6964873417721517e-05, + "loss": 0.0, + "step": 11600 + }, + { + "epoch": 116.25, + "learning_rate": 2.6941139240506327e-05, + "loss": 0.0, + "step": 11625 + }, + { + "epoch": 116.5, + "learning_rate": 2.6917405063291136e-05, + "loss": 0.0, + "step": 11650 + }, + { + "epoch": 116.75, + "learning_rate": 2.6893670886075946e-05, + "loss": 0.0, + "step": 11675 + }, + { + "epoch": 117.0, + "learning_rate": 2.6869936708860755e-05, + "loss": 0.0, + "step": 11700 + }, + { + "epoch": 117.25, + "learning_rate": 2.6846202531645565e-05, + "loss": 0.0, + "step": 11725 + }, + { + "epoch": 117.5, + "learning_rate": 2.6822468354430378e-05, + "loss": 0.0, + "step": 11750 + }, + { + "epoch": 117.75, + "learning_rate": 2.6798734177215187e-05, + "loss": 0.0, + "step": 11775 + }, + { + "epoch": 118.0, + "learning_rate": 2.6774999999999997e-05, + "loss": 0.0, + "step": 11800 + }, + { + "epoch": 118.25, + "learning_rate": 2.6751265822784806e-05, + "loss": 0.0, + "step": 11825 + }, + { + "epoch": 118.5, + "learning_rate": 2.6727531645569616e-05, + "loss": 0.0, + "step": 11850 + }, + { + "epoch": 118.75, + "learning_rate": 2.6703797468354425e-05, + "loss": 0.0, + "step": 11875 + }, + { + "epoch": 119.0, + "learning_rate": 2.6680063291139235e-05, + "loss": 0.0, + "step": 11900 + }, + { + "epoch": 119.25, + "learning_rate": 2.6656329113924048e-05, + "loss": 0.0, + "step": 11925 + }, + { + "epoch": 119.5, + "learning_rate": 2.6632594936708857e-05, + "loss": 0.0, + "step": 11950 + }, + { + "epoch": 119.75, + "learning_rate": 2.660886075949367e-05, + "loss": 0.0, + "step": 11975 + }, + { + "epoch": 120.0, + "learning_rate": 2.658512658227848e-05, + "loss": 0.0, + "step": 12000 + }, + { + "epoch": 120.0, + "eval_loss": 0.5174862146377563, + "eval_runtime": 80.4061, + "eval_samples_per_second": 158.471, + "eval_steps_per_second": 1.244, + "eval_wer": 19.65862904486478, + "step": 12000 + }, + { + "epoch": 120.25, + "learning_rate": 2.656139240506329e-05, + "loss": 0.0, + "step": 12025 + }, + { + "epoch": 120.5, + "learning_rate": 2.6537658227848102e-05, + "loss": 0.0, + "step": 12050 + }, + { + "epoch": 120.75, + "learning_rate": 2.651392405063291e-05, + "loss": 0.0, + "step": 12075 + }, + { + "epoch": 121.0, + "learning_rate": 2.649018987341772e-05, + "loss": 0.0, + "step": 12100 + }, + { + "epoch": 121.25, + "learning_rate": 2.646645569620253e-05, + "loss": 0.0, + "step": 12125 + }, + { + "epoch": 121.5, + "learning_rate": 2.644272151898734e-05, + "loss": 0.0, + "step": 12150 + }, + { + "epoch": 121.75, + "learning_rate": 2.641898734177215e-05, + "loss": 0.0, + "step": 12175 + }, + { + "epoch": 122.0, + "learning_rate": 2.639525316455696e-05, + "loss": 0.0, + "step": 12200 + }, + { + "epoch": 122.25, + "learning_rate": 2.6371518987341772e-05, + "loss": 0.0, + "step": 12225 + }, + { + "epoch": 122.5, + "learning_rate": 2.634778481012658e-05, + "loss": 0.0, + "step": 12250 + }, + { + "epoch": 122.75, + "learning_rate": 2.632405063291139e-05, + "loss": 0.0, + "step": 12275 + }, + { + "epoch": 123.0, + "learning_rate": 2.63003164556962e-05, + "loss": 0.0, + "step": 12300 + }, + { + "epoch": 123.25, + "learning_rate": 2.627658227848101e-05, + "loss": 0.0, + "step": 12325 + }, + { + "epoch": 123.5, + "learning_rate": 2.625284810126582e-05, + "loss": 0.0, + "step": 12350 + }, + { + "epoch": 123.75, + "learning_rate": 2.622911392405063e-05, + "loss": 0.0, + "step": 12375 + }, + { + "epoch": 124.0, + "learning_rate": 2.6205379746835442e-05, + "loss": 0.0, + "step": 12400 + }, + { + "epoch": 124.25, + "learning_rate": 2.618164556962025e-05, + "loss": 0.0, + "step": 12425 + }, + { + "epoch": 124.5, + "learning_rate": 2.615791139240506e-05, + "loss": 0.0, + "step": 12450 + }, + { + "epoch": 124.75, + "learning_rate": 2.613417721518987e-05, + "loss": 0.0, + "step": 12475 + }, + { + "epoch": 125.0, + "learning_rate": 2.611044303797468e-05, + "loss": 0.0, + "step": 12500 + }, + { + "epoch": 125.25, + "learning_rate": 2.608670886075949e-05, + "loss": 0.0, + "step": 12525 + }, + { + "epoch": 125.5, + "learning_rate": 2.60629746835443e-05, + "loss": 0.0, + "step": 12550 + }, + { + "epoch": 125.75, + "learning_rate": 2.6039240506329112e-05, + "loss": 0.0, + "step": 12575 + }, + { + "epoch": 126.0, + "learning_rate": 2.601550632911392e-05, + "loss": 0.0, + "step": 12600 + }, + { + "epoch": 126.25, + "learning_rate": 2.599177215189873e-05, + "loss": 0.0, + "step": 12625 + }, + { + "epoch": 126.5, + "learning_rate": 2.596803797468354e-05, + "loss": 0.0, + "step": 12650 + }, + { + "epoch": 126.75, + "learning_rate": 2.594430379746835e-05, + "loss": 0.0, + "step": 12675 + }, + { + "epoch": 127.0, + "learning_rate": 2.592056962025316e-05, + "loss": 0.0, + "step": 12700 + }, + { + "epoch": 127.25, + "learning_rate": 2.5896835443037972e-05, + "loss": 0.0, + "step": 12725 + }, + { + "epoch": 127.5, + "learning_rate": 2.5873101265822782e-05, + "loss": 0.0, + "step": 12750 + }, + { + "epoch": 127.75, + "learning_rate": 2.584936708860759e-05, + "loss": 0.0, + "step": 12775 + }, + { + "epoch": 128.0, + "learning_rate": 2.5825632911392404e-05, + "loss": 0.0, + "step": 12800 + }, + { + "epoch": 128.25, + "learning_rate": 2.5801898734177214e-05, + "loss": 0.0, + "step": 12825 + }, + { + "epoch": 128.5, + "learning_rate": 2.5778164556962023e-05, + "loss": 0.0, + "step": 12850 + }, + { + "epoch": 128.75, + "learning_rate": 2.5754430379746836e-05, + "loss": 0.0, + "step": 12875 + }, + { + "epoch": 129.0, + "learning_rate": 2.5730696202531646e-05, + "loss": 0.0, + "step": 12900 + }, + { + "epoch": 129.25, + "learning_rate": 2.5706962025316455e-05, + "loss": 0.0, + "step": 12925 + }, + { + "epoch": 129.5, + "learning_rate": 2.5683227848101265e-05, + "loss": 0.0, + "step": 12950 + }, + { + "epoch": 129.75, + "learning_rate": 2.5659493670886074e-05, + "loss": 0.0, + "step": 12975 + }, + { + "epoch": 130.0, + "learning_rate": 2.5635759493670884e-05, + "loss": 0.0, + "step": 13000 + }, + { + "epoch": 130.0, + "eval_loss": 0.5348048806190491, + "eval_runtime": 80.5003, + "eval_samples_per_second": 158.285, + "eval_steps_per_second": 1.242, + "eval_wer": 19.77294044083164, + "step": 13000 + }, + { + "epoch": 130.25, + "learning_rate": 2.5612025316455693e-05, + "loss": 0.0, + "step": 13025 + }, + { + "epoch": 130.5, + "learning_rate": 2.5588291139240506e-05, + "loss": 0.0, + "step": 13050 + }, + { + "epoch": 130.75, + "learning_rate": 2.5564556962025316e-05, + "loss": 0.0, + "step": 13075 + }, + { + "epoch": 131.0, + "learning_rate": 2.5540822784810125e-05, + "loss": 0.0, + "step": 13100 + }, + { + "epoch": 131.25, + "learning_rate": 2.5517088607594935e-05, + "loss": 0.0, + "step": 13125 + }, + { + "epoch": 131.5, + "learning_rate": 2.5493354430379744e-05, + "loss": 0.0, + "step": 13150 + }, + { + "epoch": 131.75, + "learning_rate": 2.5469620253164554e-05, + "loss": 0.0, + "step": 13175 + }, + { + "epoch": 132.0, + "learning_rate": 2.5445886075949367e-05, + "loss": 0.0, + "step": 13200 + }, + { + "epoch": 132.25, + "learning_rate": 2.5422151898734176e-05, + "loss": 0.0, + "step": 13225 + }, + { + "epoch": 132.5, + "learning_rate": 2.5398417721518986e-05, + "loss": 0.0, + "step": 13250 + }, + { + "epoch": 132.75, + "learning_rate": 2.5374683544303795e-05, + "loss": 0.0, + "step": 13275 + }, + { + "epoch": 133.0, + "learning_rate": 2.5350949367088605e-05, + "loss": 0.0, + "step": 13300 + }, + { + "epoch": 133.25, + "learning_rate": 2.5327215189873414e-05, + "loss": 0.0, + "step": 13325 + }, + { + "epoch": 133.5, + "learning_rate": 2.5303481012658224e-05, + "loss": 0.0, + "step": 13350 + }, + { + "epoch": 133.75, + "learning_rate": 2.5279746835443036e-05, + "loss": 0.0, + "step": 13375 + }, + { + "epoch": 134.0, + "learning_rate": 2.5256012658227846e-05, + "loss": 0.0, + "step": 13400 + }, + { + "epoch": 134.25, + "learning_rate": 2.5232278481012656e-05, + "loss": 0.0, + "step": 13425 + }, + { + "epoch": 134.5, + "learning_rate": 2.5208544303797465e-05, + "loss": 0.0, + "step": 13450 + }, + { + "epoch": 134.75, + "learning_rate": 2.5184810126582275e-05, + "loss": 0.0, + "step": 13475 + }, + { + "epoch": 135.0, + "learning_rate": 2.5161075949367084e-05, + "loss": 0.0, + "step": 13500 + }, + { + "epoch": 135.25, + "learning_rate": 2.5137341772151894e-05, + "loss": 0.0, + "step": 13525 + }, + { + "epoch": 135.5, + "learning_rate": 2.5113607594936706e-05, + "loss": 0.0, + "step": 13550 + }, + { + "epoch": 135.75, + "learning_rate": 2.5089873417721516e-05, + "loss": 0.0, + "step": 13575 + }, + { + "epoch": 136.0, + "learning_rate": 2.5066139240506325e-05, + "loss": 0.0, + "step": 13600 + }, + { + "epoch": 136.25, + "learning_rate": 2.504240506329114e-05, + "loss": 0.0, + "step": 13625 + }, + { + "epoch": 136.5, + "learning_rate": 2.5018670886075948e-05, + "loss": 0.0, + "step": 13650 + }, + { + "epoch": 136.75, + "learning_rate": 2.499493670886076e-05, + "loss": 0.0, + "step": 13675 + }, + { + "epoch": 137.0, + "learning_rate": 2.497120253164557e-05, + "loss": 0.0, + "step": 13700 + }, + { + "epoch": 137.25, + "learning_rate": 2.494746835443038e-05, + "loss": 0.0, + "step": 13725 + }, + { + "epoch": 137.5, + "learning_rate": 2.492373417721519e-05, + "loss": 0.0, + "step": 13750 + }, + { + "epoch": 137.75, + "learning_rate": 2.49e-05, + "loss": 0.0, + "step": 13775 + }, + { + "epoch": 138.0, + "learning_rate": 2.4876265822784808e-05, + "loss": 0.0, + "step": 13800 + }, + { + "epoch": 138.25, + "learning_rate": 2.4852531645569618e-05, + "loss": 0.0, + "step": 13825 + }, + { + "epoch": 138.5, + "learning_rate": 2.482879746835443e-05, + "loss": 0.0, + "step": 13850 + }, + { + "epoch": 138.75, + "learning_rate": 2.480506329113924e-05, + "loss": 0.0, + "step": 13875 + }, + { + "epoch": 139.0, + "learning_rate": 2.478132911392405e-05, + "loss": 0.0, + "step": 13900 + }, + { + "epoch": 139.25, + "learning_rate": 2.475759493670886e-05, + "loss": 0.0, + "step": 13925 + }, + { + "epoch": 139.5, + "learning_rate": 2.473386075949367e-05, + "loss": 0.0, + "step": 13950 + }, + { + "epoch": 139.75, + "learning_rate": 2.4710126582278478e-05, + "loss": 0.0, + "step": 13975 + }, + { + "epoch": 140.0, + "learning_rate": 2.4686392405063288e-05, + "loss": 0.0, + "step": 14000 + }, + { + "epoch": 140.0, + "eval_loss": 0.553089439868927, + "eval_runtime": 80.4755, + "eval_samples_per_second": 158.334, + "eval_steps_per_second": 1.243, + "eval_wer": 19.784664686571833, + "step": 14000 + }, + { + "epoch": 140.25, + "learning_rate": 2.46626582278481e-05, + "loss": 0.0, + "step": 14025 + }, + { + "epoch": 140.5, + "learning_rate": 2.463892405063291e-05, + "loss": 0.0, + "step": 14050 + }, + { + "epoch": 140.75, + "learning_rate": 2.461518987341772e-05, + "loss": 0.0, + "step": 14075 + }, + { + "epoch": 141.0, + "learning_rate": 2.459145569620253e-05, + "loss": 0.0, + "step": 14100 + }, + { + "epoch": 141.25, + "learning_rate": 2.456772151898734e-05, + "loss": 0.0, + "step": 14125 + }, + { + "epoch": 141.5, + "learning_rate": 2.4543987341772148e-05, + "loss": 0.0, + "step": 14150 + }, + { + "epoch": 141.75, + "learning_rate": 2.4520253164556958e-05, + "loss": 0.0, + "step": 14175 + }, + { + "epoch": 142.0, + "learning_rate": 2.449651898734177e-05, + "loss": 0.0, + "step": 14200 + }, + { + "epoch": 142.25, + "learning_rate": 2.447278481012658e-05, + "loss": 0.0, + "step": 14225 + }, + { + "epoch": 142.5, + "learning_rate": 2.444905063291139e-05, + "loss": 0.0, + "step": 14250 + }, + { + "epoch": 142.75, + "learning_rate": 2.44253164556962e-05, + "loss": 0.0, + "step": 14275 + }, + { + "epoch": 143.0, + "learning_rate": 2.440158227848101e-05, + "loss": 0.0, + "step": 14300 + }, + { + "epoch": 143.25, + "learning_rate": 2.4377848101265818e-05, + "loss": 0.0, + "step": 14325 + }, + { + "epoch": 143.5, + "learning_rate": 2.4354113924050628e-05, + "loss": 0.0, + "step": 14350 + }, + { + "epoch": 143.75, + "learning_rate": 2.4333227848101262e-05, + "loss": 0.0107, + "step": 14375 + }, + { + "epoch": 144.0, + "learning_rate": 2.4309493670886075e-05, + "loss": 0.0398, + "step": 14400 + }, + { + "epoch": 144.25, + "learning_rate": 2.4285759493670885e-05, + "loss": 0.0325, + "step": 14425 + }, + { + "epoch": 144.5, + "learning_rate": 2.4262025316455694e-05, + "loss": 0.024, + "step": 14450 + }, + { + "epoch": 144.75, + "learning_rate": 2.4238291139240504e-05, + "loss": 0.0209, + "step": 14475 + }, + { + "epoch": 145.0, + "learning_rate": 2.4214556962025313e-05, + "loss": 0.0178, + "step": 14500 + }, + { + "epoch": 145.25, + "learning_rate": 2.4190822784810123e-05, + "loss": 0.0087, + "step": 14525 + }, + { + "epoch": 145.5, + "learning_rate": 2.4167088607594932e-05, + "loss": 0.0102, + "step": 14550 + }, + { + "epoch": 145.75, + "learning_rate": 2.4143354430379745e-05, + "loss": 0.0085, + "step": 14575 + }, + { + "epoch": 146.0, + "learning_rate": 2.4119620253164555e-05, + "loss": 0.007, + "step": 14600 + }, + { + "epoch": 146.25, + "learning_rate": 2.4095886075949364e-05, + "loss": 0.0046, + "step": 14625 + }, + { + "epoch": 146.5, + "learning_rate": 2.4072151898734174e-05, + "loss": 0.005, + "step": 14650 + }, + { + "epoch": 146.75, + "learning_rate": 2.4048417721518983e-05, + "loss": 0.0035, + "step": 14675 + }, + { + "epoch": 147.0, + "learning_rate": 2.4024683544303793e-05, + "loss": 0.0035, + "step": 14700 + }, + { + "epoch": 147.25, + "learning_rate": 2.4000949367088602e-05, + "loss": 0.0021, + "step": 14725 + }, + { + "epoch": 147.5, + "learning_rate": 2.3977215189873415e-05, + "loss": 0.002, + "step": 14750 + }, + { + "epoch": 147.75, + "learning_rate": 2.3953481012658225e-05, + "loss": 0.0016, + "step": 14775 + }, + { + "epoch": 148.0, + "learning_rate": 2.3929746835443034e-05, + "loss": 0.0015, + "step": 14800 + }, + { + "epoch": 148.25, + "learning_rate": 2.3906012658227847e-05, + "loss": 0.0006, + "step": 14825 + }, + { + "epoch": 148.5, + "learning_rate": 2.3882278481012657e-05, + "loss": 0.0005, + "step": 14850 + }, + { + "epoch": 148.75, + "learning_rate": 2.385854430379747e-05, + "loss": 0.0004, + "step": 14875 + }, + { + "epoch": 149.0, + "learning_rate": 2.383481012658228e-05, + "loss": 0.0006, + "step": 14900 + }, + { + "epoch": 149.25, + "learning_rate": 2.381107594936709e-05, + "loss": 0.0003, + "step": 14925 + }, + { + "epoch": 149.5, + "learning_rate": 2.3787341772151898e-05, + "loss": 0.0003, + "step": 14950 + }, + { + "epoch": 149.75, + "learning_rate": 2.3763607594936707e-05, + "loss": 0.0003, + "step": 14975 + }, + { + "epoch": 150.0, + "learning_rate": 2.3739873417721517e-05, + "loss": 0.0002, + "step": 15000 + }, + { + "epoch": 150.0, + "eval_loss": 0.4625629484653473, + "eval_runtime": 81.6484, + "eval_samples_per_second": 156.059, + "eval_steps_per_second": 1.225, + "eval_wer": 19.472995153978427, + "step": 15000 + }, + { + "epoch": 150.25, + "learning_rate": 2.3716139240506327e-05, + "loss": 0.0002, + "step": 15025 + }, + { + "epoch": 150.5, + "learning_rate": 2.369240506329114e-05, + "loss": 0.0002, + "step": 15050 + }, + { + "epoch": 150.75, + "learning_rate": 2.366867088607595e-05, + "loss": 0.0002, + "step": 15075 + }, + { + "epoch": 151.0, + "learning_rate": 2.364493670886076e-05, + "loss": 0.0002, + "step": 15100 + }, + { + "epoch": 151.25, + "learning_rate": 2.3621202531645568e-05, + "loss": 0.0001, + "step": 15125 + }, + { + "epoch": 151.5, + "learning_rate": 2.3597468354430377e-05, + "loss": 0.0001, + "step": 15150 + }, + { + "epoch": 151.75, + "learning_rate": 2.3573734177215187e-05, + "loss": 0.0001, + "step": 15175 + }, + { + "epoch": 152.0, + "learning_rate": 2.3549999999999996e-05, + "loss": 0.0002, + "step": 15200 + }, + { + "epoch": 152.25, + "learning_rate": 2.352626582278481e-05, + "loss": 0.0001, + "step": 15225 + }, + { + "epoch": 152.5, + "learning_rate": 2.350253164556962e-05, + "loss": 0.0001, + "step": 15250 + }, + { + "epoch": 152.75, + "learning_rate": 2.347879746835443e-05, + "loss": 0.0001, + "step": 15275 + }, + { + "epoch": 153.0, + "learning_rate": 2.3455063291139238e-05, + "loss": 0.0001, + "step": 15300 + }, + { + "epoch": 153.25, + "learning_rate": 2.3431329113924047e-05, + "loss": 0.0001, + "step": 15325 + }, + { + "epoch": 153.5, + "learning_rate": 2.3407594936708857e-05, + "loss": 0.0001, + "step": 15350 + }, + { + "epoch": 153.75, + "learning_rate": 2.3383860759493666e-05, + "loss": 0.0001, + "step": 15375 + }, + { + "epoch": 154.0, + "learning_rate": 2.336012658227848e-05, + "loss": 0.0001, + "step": 15400 + }, + { + "epoch": 154.25, + "learning_rate": 2.333639240506329e-05, + "loss": 0.0001, + "step": 15425 + }, + { + "epoch": 154.5, + "learning_rate": 2.3312658227848098e-05, + "loss": 0.0001, + "step": 15450 + }, + { + "epoch": 154.75, + "learning_rate": 2.3288924050632908e-05, + "loss": 0.0001, + "step": 15475 + }, + { + "epoch": 155.0, + "learning_rate": 2.3265189873417717e-05, + "loss": 0.0001, + "step": 15500 + }, + { + "epoch": 155.25, + "learning_rate": 2.3241455696202527e-05, + "loss": 0.0001, + "step": 15525 + }, + { + "epoch": 155.5, + "learning_rate": 2.321772151898734e-05, + "loss": 0.0001, + "step": 15550 + }, + { + "epoch": 155.75, + "learning_rate": 2.319398734177215e-05, + "loss": 0.0001, + "step": 15575 + }, + { + "epoch": 156.0, + "learning_rate": 2.317025316455696e-05, + "loss": 0.0001, + "step": 15600 + }, + { + "epoch": 156.25, + "learning_rate": 2.3146518987341768e-05, + "loss": 0.0001, + "step": 15625 + }, + { + "epoch": 156.5, + "learning_rate": 2.312278481012658e-05, + "loss": 0.0001, + "step": 15650 + }, + { + "epoch": 156.75, + "learning_rate": 2.309905063291139e-05, + "loss": 0.0001, + "step": 15675 + }, + { + "epoch": 157.0, + "learning_rate": 2.3075316455696204e-05, + "loss": 0.0001, + "step": 15700 + }, + { + "epoch": 157.25, + "learning_rate": 2.3051582278481013e-05, + "loss": 0.0001, + "step": 15725 + }, + { + "epoch": 157.5, + "learning_rate": 2.3027848101265823e-05, + "loss": 0.0001, + "step": 15750 + }, + { + "epoch": 157.75, + "learning_rate": 2.3004113924050632e-05, + "loss": 0.0001, + "step": 15775 + }, + { + "epoch": 158.0, + "learning_rate": 2.298037974683544e-05, + "loss": 0.0001, + "step": 15800 + }, + { + "epoch": 158.25, + "learning_rate": 2.295664556962025e-05, + "loss": 0.0001, + "step": 15825 + }, + { + "epoch": 158.5, + "learning_rate": 2.293291139240506e-05, + "loss": 0.0001, + "step": 15850 + }, + { + "epoch": 158.75, + "learning_rate": 2.2909177215189873e-05, + "loss": 0.0001, + "step": 15875 + }, + { + "epoch": 159.0, + "learning_rate": 2.2885443037974683e-05, + "loss": 0.0001, + "step": 15900 + }, + { + "epoch": 159.25, + "learning_rate": 2.2861708860759492e-05, + "loss": 0.0001, + "step": 15925 + }, + { + "epoch": 159.5, + "learning_rate": 2.2837974683544302e-05, + "loss": 0.0001, + "step": 15950 + }, + { + "epoch": 159.75, + "learning_rate": 2.281424050632911e-05, + "loss": 0.0001, + "step": 15975 + }, + { + "epoch": 160.0, + "learning_rate": 2.279050632911392e-05, + "loss": 0.0001, + "step": 16000 + }, + { + "epoch": 160.0, + "eval_loss": 0.4813273549079895, + "eval_runtime": 80.403, + "eval_samples_per_second": 158.477, + "eval_steps_per_second": 1.244, + "eval_wer": 19.21994685008598, + "step": 16000 + }, + { + "epoch": 160.25, + "learning_rate": 2.2766772151898734e-05, + "loss": 0.0001, + "step": 16025 + }, + { + "epoch": 160.5, + "learning_rate": 2.2743037974683543e-05, + "loss": 0.0001, + "step": 16050 + }, + { + "epoch": 160.75, + "learning_rate": 2.2719303797468353e-05, + "loss": 0.0001, + "step": 16075 + }, + { + "epoch": 161.0, + "learning_rate": 2.2695569620253162e-05, + "loss": 0.0001, + "step": 16100 + }, + { + "epoch": 161.25, + "learning_rate": 2.2671835443037972e-05, + "loss": 0.0001, + "step": 16125 + }, + { + "epoch": 161.5, + "learning_rate": 2.264810126582278e-05, + "loss": 0.0001, + "step": 16150 + }, + { + "epoch": 161.75, + "learning_rate": 2.262436708860759e-05, + "loss": 0.0001, + "step": 16175 + }, + { + "epoch": 162.0, + "learning_rate": 2.2600632911392404e-05, + "loss": 0.0001, + "step": 16200 + }, + { + "epoch": 162.25, + "learning_rate": 2.2576898734177213e-05, + "loss": 0.0001, + "step": 16225 + }, + { + "epoch": 162.5, + "learning_rate": 2.2553164556962023e-05, + "loss": 0.0001, + "step": 16250 + }, + { + "epoch": 162.75, + "learning_rate": 2.2529430379746832e-05, + "loss": 0.0001, + "step": 16275 + }, + { + "epoch": 163.0, + "learning_rate": 2.2505696202531642e-05, + "loss": 0.0001, + "step": 16300 + }, + { + "epoch": 163.25, + "learning_rate": 2.248196202531645e-05, + "loss": 0.0001, + "step": 16325 + }, + { + "epoch": 163.5, + "learning_rate": 2.245822784810126e-05, + "loss": 0.0001, + "step": 16350 + }, + { + "epoch": 163.75, + "learning_rate": 2.2434493670886074e-05, + "loss": 0.0001, + "step": 16375 + }, + { + "epoch": 164.0, + "learning_rate": 2.2410759493670883e-05, + "loss": 0.0001, + "step": 16400 + }, + { + "epoch": 164.25, + "learning_rate": 2.2387025316455693e-05, + "loss": 0.0001, + "step": 16425 + }, + { + "epoch": 164.5, + "learning_rate": 2.2363291139240502e-05, + "loss": 0.0001, + "step": 16450 + }, + { + "epoch": 164.75, + "learning_rate": 2.2339556962025315e-05, + "loss": 0.0001, + "step": 16475 + }, + { + "epoch": 165.0, + "learning_rate": 2.2315822784810128e-05, + "loss": 0.0001, + "step": 16500 + }, + { + "epoch": 165.25, + "learning_rate": 2.2292088607594938e-05, + "loss": 0.0001, + "step": 16525 + }, + { + "epoch": 165.5, + "learning_rate": 2.2268354430379747e-05, + "loss": 0.0001, + "step": 16550 + }, + { + "epoch": 165.75, + "learning_rate": 2.2244620253164557e-05, + "loss": 0.0001, + "step": 16575 + }, + { + "epoch": 166.0, + "learning_rate": 2.2220886075949366e-05, + "loss": 0.0001, + "step": 16600 + }, + { + "epoch": 166.25, + "learning_rate": 2.2197151898734176e-05, + "loss": 0.0001, + "step": 16625 + }, + { + "epoch": 166.5, + "learning_rate": 2.2173417721518985e-05, + "loss": 0.0001, + "step": 16650 + }, + { + "epoch": 166.75, + "learning_rate": 2.2149683544303798e-05, + "loss": 0.0001, + "step": 16675 + }, + { + "epoch": 167.0, + "learning_rate": 2.2125949367088608e-05, + "loss": 0.0001, + "step": 16700 + }, + { + "epoch": 167.25, + "learning_rate": 2.2102215189873417e-05, + "loss": 0.0001, + "step": 16725 + }, + { + "epoch": 167.5, + "learning_rate": 2.2078481012658227e-05, + "loss": 0.0001, + "step": 16750 + }, + { + "epoch": 167.75, + "learning_rate": 2.2054746835443036e-05, + "loss": 0.0001, + "step": 16775 + }, + { + "epoch": 168.0, + "learning_rate": 2.2031012658227846e-05, + "loss": 0.0001, + "step": 16800 + }, + { + "epoch": 168.25, + "learning_rate": 2.2007278481012655e-05, + "loss": 0.0, + "step": 16825 + }, + { + "epoch": 168.5, + "learning_rate": 2.1983544303797468e-05, + "loss": 0.0, + "step": 16850 + }, + { + "epoch": 168.75, + "learning_rate": 2.1959810126582277e-05, + "loss": 0.0, + "step": 16875 + }, + { + "epoch": 169.0, + "learning_rate": 2.1936075949367087e-05, + "loss": 0.0001, + "step": 16900 + }, + { + "epoch": 169.25, + "learning_rate": 2.1912341772151896e-05, + "loss": 0.0, + "step": 16925 + }, + { + "epoch": 169.5, + "learning_rate": 2.1888607594936706e-05, + "loss": 0.0001, + "step": 16950 + }, + { + "epoch": 169.75, + "learning_rate": 2.1864873417721515e-05, + "loss": 0.0, + "step": 16975 + }, + { + "epoch": 170.0, + "learning_rate": 2.1841139240506325e-05, + "loss": 0.0, + "step": 17000 + }, + { + "epoch": 170.0, + "eval_loss": 0.49322769045829773, + "eval_runtime": 80.3155, + "eval_samples_per_second": 158.649, + "eval_steps_per_second": 1.245, + "eval_wer": 19.169141785211817, + "step": 17000 + }, + { + "epoch": 170.25, + "learning_rate": 2.1817405063291138e-05, + "loss": 0.0, + "step": 17025 + }, + { + "epoch": 170.5, + "learning_rate": 2.1793670886075947e-05, + "loss": 0.0, + "step": 17050 + }, + { + "epoch": 170.75, + "learning_rate": 2.1769936708860757e-05, + "loss": 0.0, + "step": 17075 + }, + { + "epoch": 171.0, + "learning_rate": 2.1746202531645566e-05, + "loss": 0.0001, + "step": 17100 + }, + { + "epoch": 171.25, + "learning_rate": 2.1722468354430376e-05, + "loss": 0.0, + "step": 17125 + }, + { + "epoch": 171.5, + "learning_rate": 2.1698734177215185e-05, + "loss": 0.0001, + "step": 17150 + }, + { + "epoch": 171.75, + "learning_rate": 2.1674999999999995e-05, + "loss": 0.0, + "step": 17175 + }, + { + "epoch": 172.0, + "learning_rate": 2.1651265822784808e-05, + "loss": 0.0, + "step": 17200 + }, + { + "epoch": 172.25, + "learning_rate": 2.1627531645569617e-05, + "loss": 0.0, + "step": 17225 + }, + { + "epoch": 172.5, + "learning_rate": 2.1603797468354427e-05, + "loss": 0.0, + "step": 17250 + }, + { + "epoch": 172.75, + "learning_rate": 2.1580063291139236e-05, + "loss": 0.0, + "step": 17275 + }, + { + "epoch": 173.0, + "learning_rate": 2.155632911392405e-05, + "loss": 0.0, + "step": 17300 + }, + { + "epoch": 173.25, + "learning_rate": 2.1532594936708862e-05, + "loss": 0.0, + "step": 17325 + }, + { + "epoch": 173.5, + "learning_rate": 2.150886075949367e-05, + "loss": 0.0, + "step": 17350 + }, + { + "epoch": 173.75, + "learning_rate": 2.148512658227848e-05, + "loss": 0.0, + "step": 17375 + }, + { + "epoch": 174.0, + "learning_rate": 2.146139240506329e-05, + "loss": 0.0, + "step": 17400 + }, + { + "epoch": 174.25, + "learning_rate": 2.14376582278481e-05, + "loss": 0.0, + "step": 17425 + }, + { + "epoch": 174.5, + "learning_rate": 2.141392405063291e-05, + "loss": 0.0, + "step": 17450 + }, + { + "epoch": 174.75, + "learning_rate": 2.139018987341772e-05, + "loss": 0.0, + "step": 17475 + }, + { + "epoch": 175.0, + "learning_rate": 2.1366455696202532e-05, + "loss": 0.0, + "step": 17500 + }, + { + "epoch": 175.25, + "learning_rate": 2.134272151898734e-05, + "loss": 0.0, + "step": 17525 + }, + { + "epoch": 175.5, + "learning_rate": 2.131898734177215e-05, + "loss": 0.0, + "step": 17550 + }, + { + "epoch": 175.75, + "learning_rate": 2.129525316455696e-05, + "loss": 0.0, + "step": 17575 + }, + { + "epoch": 176.0, + "learning_rate": 2.127151898734177e-05, + "loss": 0.0, + "step": 17600 + }, + { + "epoch": 176.25, + "learning_rate": 2.124778481012658e-05, + "loss": 0.0, + "step": 17625 + }, + { + "epoch": 176.5, + "learning_rate": 2.122405063291139e-05, + "loss": 0.0, + "step": 17650 + }, + { + "epoch": 176.75, + "learning_rate": 2.1200316455696202e-05, + "loss": 0.0, + "step": 17675 + }, + { + "epoch": 177.0, + "learning_rate": 2.117658227848101e-05, + "loss": 0.0, + "step": 17700 + }, + { + "epoch": 177.25, + "learning_rate": 2.115284810126582e-05, + "loss": 0.0, + "step": 17725 + }, + { + "epoch": 177.5, + "learning_rate": 2.112911392405063e-05, + "loss": 0.0, + "step": 17750 + }, + { + "epoch": 177.75, + "learning_rate": 2.110537974683544e-05, + "loss": 0.0, + "step": 17775 + }, + { + "epoch": 178.0, + "learning_rate": 2.108164556962025e-05, + "loss": 0.0, + "step": 17800 + }, + { + "epoch": 178.25, + "learning_rate": 2.1057911392405062e-05, + "loss": 0.0, + "step": 17825 + }, + { + "epoch": 178.5, + "learning_rate": 2.1034177215189872e-05, + "loss": 0.0, + "step": 17850 + }, + { + "epoch": 178.75, + "learning_rate": 2.101044303797468e-05, + "loss": 0.0, + "step": 17875 + }, + { + "epoch": 179.0, + "learning_rate": 2.098670886075949e-05, + "loss": 0.0, + "step": 17900 + }, + { + "epoch": 179.25, + "learning_rate": 2.09629746835443e-05, + "loss": 0.0, + "step": 17925 + }, + { + "epoch": 179.5, + "learning_rate": 2.093924050632911e-05, + "loss": 0.0, + "step": 17950 + }, + { + "epoch": 179.75, + "learning_rate": 2.091550632911392e-05, + "loss": 0.0, + "step": 17975 + }, + { + "epoch": 180.0, + "learning_rate": 2.0891772151898732e-05, + "loss": 0.0, + "step": 18000 + }, + { + "epoch": 180.0, + "eval_loss": 0.5040577054023743, + "eval_runtime": 80.2662, + "eval_samples_per_second": 158.747, + "eval_steps_per_second": 1.246, + "eval_wer": 19.1290839455995, + "step": 18000 + }, + { + "epoch": 180.25, + "learning_rate": 2.0868037974683542e-05, + "loss": 0.0, + "step": 18025 + }, + { + "epoch": 180.5, + "learning_rate": 2.084430379746835e-05, + "loss": 0.0, + "step": 18050 + }, + { + "epoch": 180.75, + "learning_rate": 2.082056962025316e-05, + "loss": 0.0, + "step": 18075 + }, + { + "epoch": 181.0, + "learning_rate": 2.0796835443037974e-05, + "loss": 0.0, + "step": 18100 + }, + { + "epoch": 181.25, + "learning_rate": 2.0773101265822783e-05, + "loss": 0.0, + "step": 18125 + }, + { + "epoch": 181.5, + "learning_rate": 2.0749367088607596e-05, + "loss": 0.0, + "step": 18150 + }, + { + "epoch": 181.75, + "learning_rate": 2.0725632911392406e-05, + "loss": 0.0, + "step": 18175 + }, + { + "epoch": 182.0, + "learning_rate": 2.0701898734177215e-05, + "loss": 0.0, + "step": 18200 + }, + { + "epoch": 182.25, + "learning_rate": 2.0678164556962025e-05, + "loss": 0.0, + "step": 18225 + }, + { + "epoch": 182.5, + "learning_rate": 2.0654430379746834e-05, + "loss": 0.0, + "step": 18250 + }, + { + "epoch": 182.75, + "learning_rate": 2.0630696202531644e-05, + "loss": 0.0, + "step": 18275 + }, + { + "epoch": 183.0, + "learning_rate": 2.0606962025316457e-05, + "loss": 0.0, + "step": 18300 + }, + { + "epoch": 183.25, + "learning_rate": 2.0583227848101266e-05, + "loss": 0.0, + "step": 18325 + }, + { + "epoch": 183.5, + "learning_rate": 2.0559493670886076e-05, + "loss": 0.0, + "step": 18350 + }, + { + "epoch": 183.75, + "learning_rate": 2.0535759493670885e-05, + "loss": 0.0, + "step": 18375 + }, + { + "epoch": 184.0, + "learning_rate": 2.0512025316455695e-05, + "loss": 0.0, + "step": 18400 + }, + { + "epoch": 184.25, + "learning_rate": 2.0488291139240504e-05, + "loss": 0.0, + "step": 18425 + }, + { + "epoch": 184.5, + "learning_rate": 2.0464556962025314e-05, + "loss": 0.0, + "step": 18450 + }, + { + "epoch": 184.75, + "learning_rate": 2.0440822784810127e-05, + "loss": 0.0, + "step": 18475 + }, + { + "epoch": 185.0, + "learning_rate": 2.0417088607594936e-05, + "loss": 0.0, + "step": 18500 + }, + { + "epoch": 185.25, + "learning_rate": 2.0393354430379746e-05, + "loss": 0.0, + "step": 18525 + }, + { + "epoch": 185.5, + "learning_rate": 2.0369620253164555e-05, + "loss": 0.0, + "step": 18550 + }, + { + "epoch": 185.75, + "learning_rate": 2.0345886075949365e-05, + "loss": 0.0, + "step": 18575 + }, + { + "epoch": 186.0, + "learning_rate": 2.0322151898734174e-05, + "loss": 0.0, + "step": 18600 + }, + { + "epoch": 186.25, + "learning_rate": 2.0298417721518984e-05, + "loss": 0.0, + "step": 18625 + }, + { + "epoch": 186.5, + "learning_rate": 2.0274683544303796e-05, + "loss": 0.0, + "step": 18650 + }, + { + "epoch": 186.75, + "learning_rate": 2.0250949367088606e-05, + "loss": 0.0, + "step": 18675 + }, + { + "epoch": 187.0, + "learning_rate": 2.0227215189873415e-05, + "loss": 0.0, + "step": 18700 + }, + { + "epoch": 187.25, + "learning_rate": 2.0203481012658225e-05, + "loss": 0.0, + "step": 18725 + }, + { + "epoch": 187.5, + "learning_rate": 2.0179746835443035e-05, + "loss": 0.0, + "step": 18750 + }, + { + "epoch": 187.75, + "learning_rate": 2.0156012658227844e-05, + "loss": 0.0, + "step": 18775 + }, + { + "epoch": 188.0, + "learning_rate": 2.0132278481012654e-05, + "loss": 0.0, + "step": 18800 + }, + { + "epoch": 188.25, + "learning_rate": 2.0108544303797466e-05, + "loss": 0.0, + "step": 18825 + }, + { + "epoch": 188.5, + "learning_rate": 2.0084810126582276e-05, + "loss": 0.0, + "step": 18850 + }, + { + "epoch": 188.75, + "learning_rate": 2.0061075949367085e-05, + "loss": 0.0, + "step": 18875 + }, + { + "epoch": 189.0, + "learning_rate": 2.0037341772151895e-05, + "loss": 0.0, + "step": 18900 + }, + { + "epoch": 189.25, + "learning_rate": 2.0013607594936708e-05, + "loss": 0.0, + "step": 18925 + }, + { + "epoch": 189.5, + "learning_rate": 1.998987341772152e-05, + "loss": 0.0, + "step": 18950 + }, + { + "epoch": 189.75, + "learning_rate": 1.996613924050633e-05, + "loss": 0.0, + "step": 18975 + }, + { + "epoch": 190.0, + "learning_rate": 1.994240506329114e-05, + "loss": 0.0, + "step": 19000 + }, + { + "epoch": 190.0, + "eval_loss": 0.5146323442459106, + "eval_runtime": 80.5291, + "eval_samples_per_second": 158.229, + "eval_steps_per_second": 1.242, + "eval_wer": 19.094888228857275, + "step": 19000 + }, + { + "epoch": 190.25, + "learning_rate": 1.991867088607595e-05, + "loss": 0.0, + "step": 19025 + }, + { + "epoch": 190.5, + "learning_rate": 1.989493670886076e-05, + "loss": 0.0, + "step": 19050 + }, + { + "epoch": 190.75, + "learning_rate": 1.9871202531645568e-05, + "loss": 0.0, + "step": 19075 + }, + { + "epoch": 191.0, + "learning_rate": 1.9847468354430378e-05, + "loss": 0.0, + "step": 19100 + }, + { + "epoch": 191.25, + "learning_rate": 1.982373417721519e-05, + "loss": 0.0, + "step": 19125 + }, + { + "epoch": 191.5, + "learning_rate": 1.98e-05, + "loss": 0.0, + "step": 19150 + }, + { + "epoch": 191.75, + "learning_rate": 1.977626582278481e-05, + "loss": 0.0, + "step": 19175 + }, + { + "epoch": 192.0, + "learning_rate": 1.975253164556962e-05, + "loss": 0.0, + "step": 19200 + }, + { + "epoch": 192.25, + "learning_rate": 1.972879746835443e-05, + "loss": 0.0, + "step": 19225 + }, + { + "epoch": 192.5, + "learning_rate": 1.9705063291139238e-05, + "loss": 0.0, + "step": 19250 + }, + { + "epoch": 192.75, + "learning_rate": 1.9681329113924048e-05, + "loss": 0.0, + "step": 19275 + }, + { + "epoch": 193.0, + "learning_rate": 1.965759493670886e-05, + "loss": 0.0, + "step": 19300 + }, + { + "epoch": 193.25, + "learning_rate": 1.963386075949367e-05, + "loss": 0.0, + "step": 19325 + }, + { + "epoch": 193.5, + "learning_rate": 1.961012658227848e-05, + "loss": 0.0, + "step": 19350 + }, + { + "epoch": 193.75, + "learning_rate": 1.958639240506329e-05, + "loss": 0.0, + "step": 19375 + }, + { + "epoch": 194.0, + "learning_rate": 1.95626582278481e-05, + "loss": 0.0, + "step": 19400 + }, + { + "epoch": 194.25, + "learning_rate": 1.9538924050632908e-05, + "loss": 0.0, + "step": 19425 + }, + { + "epoch": 194.5, + "learning_rate": 1.9515189873417718e-05, + "loss": 0.0, + "step": 19450 + }, + { + "epoch": 194.75, + "learning_rate": 1.949145569620253e-05, + "loss": 0.0, + "step": 19475 + }, + { + "epoch": 195.0, + "learning_rate": 1.946772151898734e-05, + "loss": 0.0, + "step": 19500 + }, + { + "epoch": 195.25, + "learning_rate": 1.944398734177215e-05, + "loss": 0.0, + "step": 19525 + }, + { + "epoch": 195.5, + "learning_rate": 1.942025316455696e-05, + "loss": 0.0, + "step": 19550 + }, + { + "epoch": 195.75, + "learning_rate": 1.939651898734177e-05, + "loss": 0.0, + "step": 19575 + }, + { + "epoch": 196.0, + "learning_rate": 1.9372784810126578e-05, + "loss": 0.0, + "step": 19600 + }, + { + "epoch": 196.25, + "learning_rate": 1.9349050632911388e-05, + "loss": 0.0, + "step": 19625 + }, + { + "epoch": 196.5, + "learning_rate": 1.93253164556962e-05, + "loss": 0.0, + "step": 19650 + }, + { + "epoch": 196.75, + "learning_rate": 1.930158227848101e-05, + "loss": 0.0, + "step": 19675 + }, + { + "epoch": 197.0, + "learning_rate": 1.927784810126582e-05, + "loss": 0.0, + "step": 19700 + }, + { + "epoch": 197.25, + "learning_rate": 1.925411392405063e-05, + "loss": 0.0, + "step": 19725 + }, + { + "epoch": 197.5, + "learning_rate": 1.9230379746835442e-05, + "loss": 0.0, + "step": 19750 + }, + { + "epoch": 197.75, + "learning_rate": 1.9206645569620255e-05, + "loss": 0.0, + "step": 19775 + }, + { + "epoch": 198.0, + "learning_rate": 1.9182911392405064e-05, + "loss": 0.0, + "step": 19800 + }, + { + "epoch": 198.25, + "learning_rate": 1.9159177215189874e-05, + "loss": 0.0, + "step": 19825 + }, + { + "epoch": 198.5, + "learning_rate": 1.9135443037974683e-05, + "loss": 0.0, + "step": 19850 + }, + { + "epoch": 198.75, + "learning_rate": 1.9111708860759493e-05, + "loss": 0.0, + "step": 19875 + }, + { + "epoch": 199.0, + "learning_rate": 1.9087974683544302e-05, + "loss": 0.0, + "step": 19900 + }, + { + "epoch": 199.25, + "learning_rate": 1.9064240506329112e-05, + "loss": 0.0, + "step": 19925 + }, + { + "epoch": 199.5, + "learning_rate": 1.9040506329113925e-05, + "loss": 0.0, + "step": 19950 + }, + { + "epoch": 199.75, + "learning_rate": 1.9016772151898734e-05, + "loss": 0.0, + "step": 19975 + }, + { + "epoch": 200.0, + "learning_rate": 1.8993037974683544e-05, + "loss": 0.0, + "step": 20000 + }, + { + "epoch": 200.0, + "eval_loss": 0.5253657698631287, + "eval_runtime": 80.492, + "eval_samples_per_second": 158.301, + "eval_steps_per_second": 1.242, + "eval_wer": 19.123221822729406, + "step": 20000 + }, + { + "epoch": 200.25, + "learning_rate": 1.8969303797468353e-05, + "loss": 0.0, + "step": 20025 + }, + { + "epoch": 200.5, + "learning_rate": 1.8945569620253163e-05, + "loss": 0.0, + "step": 20050 + }, + { + "epoch": 200.75, + "learning_rate": 1.8921835443037972e-05, + "loss": 0.0, + "step": 20075 + }, + { + "epoch": 201.0, + "learning_rate": 1.8898101265822785e-05, + "loss": 0.0, + "step": 20100 + }, + { + "epoch": 201.25, + "learning_rate": 1.8874367088607595e-05, + "loss": 0.0, + "step": 20125 + }, + { + "epoch": 201.5, + "learning_rate": 1.8850632911392404e-05, + "loss": 0.0, + "step": 20150 + }, + { + "epoch": 201.75, + "learning_rate": 1.8826898734177214e-05, + "loss": 0.0, + "step": 20175 + }, + { + "epoch": 202.0, + "learning_rate": 1.8803164556962023e-05, + "loss": 0.0, + "step": 20200 + }, + { + "epoch": 202.25, + "learning_rate": 1.8779430379746833e-05, + "loss": 0.0, + "step": 20225 + }, + { + "epoch": 202.5, + "learning_rate": 1.8755696202531642e-05, + "loss": 0.0, + "step": 20250 + }, + { + "epoch": 202.75, + "learning_rate": 1.8731962025316455e-05, + "loss": 0.0, + "step": 20275 + }, + { + "epoch": 203.0, + "learning_rate": 1.8708227848101265e-05, + "loss": 0.0, + "step": 20300 + }, + { + "epoch": 203.25, + "learning_rate": 1.8684493670886074e-05, + "loss": 0.0, + "step": 20325 + }, + { + "epoch": 203.5, + "learning_rate": 1.8660759493670884e-05, + "loss": 0.0, + "step": 20350 + }, + { + "epoch": 203.75, + "learning_rate": 1.8637025316455696e-05, + "loss": 0.0, + "step": 20375 + }, + { + "epoch": 204.0, + "learning_rate": 1.8613291139240506e-05, + "loss": 0.0, + "step": 20400 + }, + { + "epoch": 204.25, + "learning_rate": 1.8589556962025316e-05, + "loss": 0.0, + "step": 20425 + }, + { + "epoch": 204.5, + "learning_rate": 1.8565822784810125e-05, + "loss": 0.0, + "step": 20450 + }, + { + "epoch": 204.75, + "learning_rate": 1.8542088607594935e-05, + "loss": 0.0, + "step": 20475 + }, + { + "epoch": 205.0, + "learning_rate": 1.8518354430379744e-05, + "loss": 0.0, + "step": 20500 + }, + { + "epoch": 205.25, + "learning_rate": 1.8494620253164557e-05, + "loss": 0.0, + "step": 20525 + }, + { + "epoch": 205.5, + "learning_rate": 1.8470886075949366e-05, + "loss": 0.0, + "step": 20550 + }, + { + "epoch": 205.75, + "learning_rate": 1.8447151898734176e-05, + "loss": 0.0, + "step": 20575 + }, + { + "epoch": 206.0, + "learning_rate": 1.8423417721518985e-05, + "loss": 0.0, + "step": 20600 + }, + { + "epoch": 206.25, + "learning_rate": 1.8399683544303795e-05, + "loss": 0.0, + "step": 20625 + }, + { + "epoch": 206.5, + "learning_rate": 1.8375949367088604e-05, + "loss": 0.0, + "step": 20650 + }, + { + "epoch": 206.75, + "learning_rate": 1.8352215189873414e-05, + "loss": 0.0, + "step": 20675 + }, + { + "epoch": 207.0, + "learning_rate": 1.8328481012658227e-05, + "loss": 0.0, + "step": 20700 + }, + { + "epoch": 207.25, + "learning_rate": 1.8304746835443036e-05, + "loss": 0.0, + "step": 20725 + }, + { + "epoch": 207.5, + "learning_rate": 1.8281012658227846e-05, + "loss": 0.0, + "step": 20750 + }, + { + "epoch": 207.75, + "learning_rate": 1.825727848101266e-05, + "loss": 0.0, + "step": 20775 + }, + { + "epoch": 208.0, + "learning_rate": 1.8233544303797468e-05, + "loss": 0.0, + "step": 20800 + }, + { + "epoch": 208.25, + "learning_rate": 1.8209810126582278e-05, + "loss": 0.0, + "step": 20825 + }, + { + "epoch": 208.5, + "learning_rate": 1.8186075949367087e-05, + "loss": 0.0, + "step": 20850 + }, + { + "epoch": 208.75, + "learning_rate": 1.8162341772151897e-05, + "loss": 0.0, + "step": 20875 + }, + { + "epoch": 209.0, + "learning_rate": 1.8138607594936706e-05, + "loss": 0.0, + "step": 20900 + }, + { + "epoch": 209.25, + "learning_rate": 1.811487341772152e-05, + "loss": 0.0, + "step": 20925 + }, + { + "epoch": 209.5, + "learning_rate": 1.809113924050633e-05, + "loss": 0.0, + "step": 20950 + }, + { + "epoch": 209.75, + "learning_rate": 1.8067405063291138e-05, + "loss": 0.0, + "step": 20975 + }, + { + "epoch": 210.0, + "learning_rate": 1.8043670886075948e-05, + "loss": 0.0, + "step": 21000 + }, + { + "epoch": 210.0, + "eval_loss": 0.5368518233299255, + "eval_runtime": 80.6917, + "eval_samples_per_second": 157.91, + "eval_steps_per_second": 1.239, + "eval_wer": 19.136900109426293, + "step": 21000 + }, + { + "epoch": 210.25, + "learning_rate": 1.8019936708860757e-05, + "loss": 0.0, + "step": 21025 + }, + { + "epoch": 210.5, + "learning_rate": 1.7996202531645567e-05, + "loss": 0.0, + "step": 21050 + }, + { + "epoch": 210.75, + "learning_rate": 1.7972468354430376e-05, + "loss": 0.0, + "step": 21075 + }, + { + "epoch": 211.0, + "learning_rate": 1.794873417721519e-05, + "loss": 0.0, + "step": 21100 + }, + { + "epoch": 211.25, + "learning_rate": 1.7925e-05, + "loss": 0.0, + "step": 21125 + }, + { + "epoch": 211.5, + "learning_rate": 1.7901265822784808e-05, + "loss": 0.0, + "step": 21150 + }, + { + "epoch": 211.75, + "learning_rate": 1.7877531645569618e-05, + "loss": 0.0, + "step": 21175 + }, + { + "epoch": 212.0, + "learning_rate": 1.785379746835443e-05, + "loss": 0.0, + "step": 21200 + }, + { + "epoch": 212.25, + "learning_rate": 1.783006329113924e-05, + "loss": 0.0, + "step": 21225 + }, + { + "epoch": 212.5, + "learning_rate": 1.780632911392405e-05, + "loss": 0.0, + "step": 21250 + }, + { + "epoch": 212.75, + "learning_rate": 1.778259493670886e-05, + "loss": 0.0, + "step": 21275 + }, + { + "epoch": 213.0, + "learning_rate": 1.775886075949367e-05, + "loss": 0.0, + "step": 21300 + }, + { + "epoch": 213.25, + "learning_rate": 1.773512658227848e-05, + "loss": 0.0, + "step": 21325 + }, + { + "epoch": 213.5, + "learning_rate": 1.771139240506329e-05, + "loss": 0.0, + "step": 21350 + }, + { + "epoch": 213.75, + "learning_rate": 1.76876582278481e-05, + "loss": 0.0, + "step": 21375 + }, + { + "epoch": 214.0, + "learning_rate": 1.766392405063291e-05, + "loss": 0.0, + "step": 21400 + }, + { + "epoch": 214.25, + "learning_rate": 1.764018987341772e-05, + "loss": 0.0, + "step": 21425 + }, + { + "epoch": 214.5, + "learning_rate": 1.761645569620253e-05, + "loss": 0.0, + "step": 21450 + }, + { + "epoch": 214.75, + "learning_rate": 1.759272151898734e-05, + "loss": 0.0, + "step": 21475 + }, + { + "epoch": 215.0, + "learning_rate": 1.756898734177215e-05, + "loss": 0.0, + "step": 21500 + }, + { + "epoch": 215.25, + "learning_rate": 1.754525316455696e-05, + "loss": 0.0, + "step": 21525 + }, + { + "epoch": 215.5, + "learning_rate": 1.752151898734177e-05, + "loss": 0.0, + "step": 21550 + }, + { + "epoch": 215.75, + "learning_rate": 1.749778481012658e-05, + "loss": 0.0, + "step": 21575 + }, + { + "epoch": 216.0, + "learning_rate": 1.7474050632911393e-05, + "loss": 0.0, + "step": 21600 + }, + { + "epoch": 216.25, + "learning_rate": 1.7450316455696202e-05, + "loss": 0.0, + "step": 21625 + }, + { + "epoch": 216.5, + "learning_rate": 1.7426582278481012e-05, + "loss": 0.0, + "step": 21650 + }, + { + "epoch": 216.75, + "learning_rate": 1.740284810126582e-05, + "loss": 0.0, + "step": 21675 + }, + { + "epoch": 217.0, + "learning_rate": 1.737911392405063e-05, + "loss": 0.0, + "step": 21700 + }, + { + "epoch": 217.25, + "learning_rate": 1.735537974683544e-05, + "loss": 0.0, + "step": 21725 + }, + { + "epoch": 217.5, + "learning_rate": 1.7331645569620253e-05, + "loss": 0.0, + "step": 21750 + }, + { + "epoch": 217.75, + "learning_rate": 1.7307911392405063e-05, + "loss": 0.0, + "step": 21775 + }, + { + "epoch": 218.0, + "learning_rate": 1.7284177215189872e-05, + "loss": 0.0, + "step": 21800 + }, + { + "epoch": 218.25, + "learning_rate": 1.7260443037974682e-05, + "loss": 0.0, + "step": 21825 + }, + { + "epoch": 218.5, + "learning_rate": 1.723670886075949e-05, + "loss": 0.0, + "step": 21850 + }, + { + "epoch": 218.75, + "learning_rate": 1.72129746835443e-05, + "loss": 0.0, + "step": 21875 + }, + { + "epoch": 219.0, + "learning_rate": 1.718924050632911e-05, + "loss": 0.0, + "step": 21900 + }, + { + "epoch": 219.25, + "learning_rate": 1.7165506329113923e-05, + "loss": 0.0, + "step": 21925 + }, + { + "epoch": 219.5, + "learning_rate": 1.7141772151898733e-05, + "loss": 0.0, + "step": 21950 + }, + { + "epoch": 219.75, + "learning_rate": 1.7118037974683542e-05, + "loss": 0.0, + "step": 21975 + }, + { + "epoch": 220.0, + "learning_rate": 1.7094303797468352e-05, + "loss": 0.0, + "step": 22000 + }, + { + "epoch": 220.0, + "eval_loss": 0.5483614802360535, + "eval_runtime": 80.6444, + "eval_samples_per_second": 158.002, + "eval_steps_per_second": 1.24, + "eval_wer": 19.112474597467564, + "step": 22000 + }, + { + "epoch": 220.25, + "learning_rate": 1.7070569620253165e-05, + "loss": 0.0, + "step": 22025 + }, + { + "epoch": 220.5, + "learning_rate": 1.7046835443037974e-05, + "loss": 0.0, + "step": 22050 + }, + { + "epoch": 220.75, + "learning_rate": 1.7023101265822784e-05, + "loss": 0.0, + "step": 22075 + }, + { + "epoch": 221.0, + "learning_rate": 1.6999367088607593e-05, + "loss": 0.0, + "step": 22100 + }, + { + "epoch": 221.25, + "learning_rate": 1.6975632911392403e-05, + "loss": 0.0, + "step": 22125 + }, + { + "epoch": 221.5, + "learning_rate": 1.6951898734177216e-05, + "loss": 0.0, + "step": 22150 + }, + { + "epoch": 221.75, + "learning_rate": 1.6928164556962025e-05, + "loss": 0.0, + "step": 22175 + }, + { + "epoch": 222.0, + "learning_rate": 1.6904430379746835e-05, + "loss": 0.0, + "step": 22200 + }, + { + "epoch": 222.25, + "learning_rate": 1.6880696202531644e-05, + "loss": 0.0, + "step": 22225 + }, + { + "epoch": 222.5, + "learning_rate": 1.6856962025316454e-05, + "loss": 0.0, + "step": 22250 + }, + { + "epoch": 222.75, + "learning_rate": 1.6833227848101263e-05, + "loss": 0.0, + "step": 22275 + }, + { + "epoch": 223.0, + "learning_rate": 1.6809493670886073e-05, + "loss": 0.0, + "step": 22300 + }, + { + "epoch": 223.25, + "learning_rate": 1.6785759493670885e-05, + "loss": 0.0, + "step": 22325 + }, + { + "epoch": 223.5, + "learning_rate": 1.6762025316455695e-05, + "loss": 0.0, + "step": 22350 + }, + { + "epoch": 223.75, + "learning_rate": 1.6738291139240504e-05, + "loss": 0.0, + "step": 22375 + }, + { + "epoch": 224.0, + "learning_rate": 1.6714556962025314e-05, + "loss": 0.0, + "step": 22400 + }, + { + "epoch": 224.25, + "learning_rate": 1.6690822784810127e-05, + "loss": 0.0, + "step": 22425 + }, + { + "epoch": 224.5, + "learning_rate": 1.6667088607594936e-05, + "loss": 0.0, + "step": 22450 + }, + { + "epoch": 224.75, + "learning_rate": 1.6643354430379746e-05, + "loss": 0.0, + "step": 22475 + }, + { + "epoch": 225.0, + "learning_rate": 1.6619620253164555e-05, + "loss": 0.0, + "step": 22500 + }, + { + "epoch": 225.25, + "learning_rate": 1.6595886075949365e-05, + "loss": 0.0, + "step": 22525 + }, + { + "epoch": 225.5, + "learning_rate": 1.6572151898734178e-05, + "loss": 0.0, + "step": 22550 + }, + { + "epoch": 225.75, + "learning_rate": 1.6548417721518987e-05, + "loss": 0.0, + "step": 22575 + }, + { + "epoch": 226.0, + "learning_rate": 1.6524683544303797e-05, + "loss": 0.0, + "step": 22600 + }, + { + "epoch": 226.25, + "learning_rate": 1.6500949367088606e-05, + "loss": 0.0, + "step": 22625 + }, + { + "epoch": 226.5, + "learning_rate": 1.6477215189873416e-05, + "loss": 0.0, + "step": 22650 + }, + { + "epoch": 226.75, + "learning_rate": 1.6453481012658225e-05, + "loss": 0.0, + "step": 22675 + }, + { + "epoch": 227.0, + "learning_rate": 1.6429746835443035e-05, + "loss": 0.0, + "step": 22700 + }, + { + "epoch": 227.25, + "learning_rate": 1.6406012658227848e-05, + "loss": 0.0, + "step": 22725 + }, + { + "epoch": 227.5, + "learning_rate": 1.6382278481012657e-05, + "loss": 0.0, + "step": 22750 + }, + { + "epoch": 227.75, + "learning_rate": 1.6358544303797467e-05, + "loss": 0.0, + "step": 22775 + }, + { + "epoch": 228.0, + "learning_rate": 1.6334810126582276e-05, + "loss": 0.0, + "step": 22800 + }, + { + "epoch": 228.25, + "learning_rate": 1.6311075949367086e-05, + "loss": 0.0, + "step": 22825 + }, + { + "epoch": 228.5, + "learning_rate": 1.62873417721519e-05, + "loss": 0.0, + "step": 22850 + }, + { + "epoch": 228.75, + "learning_rate": 1.6263607594936708e-05, + "loss": 0.0, + "step": 22875 + }, + { + "epoch": 229.0, + "learning_rate": 1.6239873417721518e-05, + "loss": 0.0, + "step": 22900 + }, + { + "epoch": 229.25, + "learning_rate": 1.6216139240506327e-05, + "loss": 0.0, + "step": 22925 + }, + { + "epoch": 229.5, + "learning_rate": 1.6192405063291137e-05, + "loss": 0.0, + "step": 22950 + }, + { + "epoch": 229.75, + "learning_rate": 1.616867088607595e-05, + "loss": 0.0, + "step": 22975 + }, + { + "epoch": 230.0, + "learning_rate": 1.614493670886076e-05, + "loss": 0.0, + "step": 23000 + }, + { + "epoch": 230.0, + "eval_loss": 0.5606202483177185, + "eval_runtime": 80.6021, + "eval_samples_per_second": 158.085, + "eval_steps_per_second": 1.241, + "eval_wer": 19.132992027512895, + "step": 23000 + }, + { + "epoch": 230.25, + "learning_rate": 1.612120253164557e-05, + "loss": 0.0, + "step": 23025 + }, + { + "epoch": 230.5, + "learning_rate": 1.6097468354430378e-05, + "loss": 0.0, + "step": 23050 + }, + { + "epoch": 230.75, + "learning_rate": 1.6073734177215188e-05, + "loss": 0.0, + "step": 23075 + }, + { + "epoch": 231.0, + "learning_rate": 1.6049999999999997e-05, + "loss": 0.0, + "step": 23100 + }, + { + "epoch": 231.25, + "learning_rate": 1.6026265822784807e-05, + "loss": 0.0, + "step": 23125 + }, + { + "epoch": 231.5, + "learning_rate": 1.600253164556962e-05, + "loss": 0.0, + "step": 23150 + }, + { + "epoch": 231.75, + "learning_rate": 1.597879746835443e-05, + "loss": 0.0, + "step": 23175 + }, + { + "epoch": 232.0, + "learning_rate": 1.595506329113924e-05, + "loss": 0.0, + "step": 23200 + }, + { + "epoch": 232.25, + "learning_rate": 1.5931329113924048e-05, + "loss": 0.0, + "step": 23225 + }, + { + "epoch": 232.5, + "learning_rate": 1.590759493670886e-05, + "loss": 0.0, + "step": 23250 + }, + { + "epoch": 232.75, + "learning_rate": 1.588386075949367e-05, + "loss": 0.0, + "step": 23275 + }, + { + "epoch": 233.0, + "learning_rate": 1.586012658227848e-05, + "loss": 0.0, + "step": 23300 + }, + { + "epoch": 233.25, + "learning_rate": 1.583639240506329e-05, + "loss": 0.0, + "step": 23325 + }, + { + "epoch": 233.5, + "learning_rate": 1.58126582278481e-05, + "loss": 0.0, + "step": 23350 + }, + { + "epoch": 233.75, + "learning_rate": 1.5788924050632912e-05, + "loss": 0.0, + "step": 23375 + }, + { + "epoch": 234.0, + "learning_rate": 1.576518987341772e-05, + "loss": 0.0, + "step": 23400 + }, + { + "epoch": 234.25, + "learning_rate": 1.574145569620253e-05, + "loss": 0.0, + "step": 23425 + }, + { + "epoch": 234.5, + "learning_rate": 1.571772151898734e-05, + "loss": 0.0, + "step": 23450 + }, + { + "epoch": 234.75, + "learning_rate": 1.569398734177215e-05, + "loss": 0.0, + "step": 23475 + }, + { + "epoch": 235.0, + "learning_rate": 1.567025316455696e-05, + "loss": 0.0, + "step": 23500 + }, + { + "epoch": 235.25, + "learning_rate": 1.564651898734177e-05, + "loss": 0.0, + "step": 23525 + }, + { + "epoch": 235.5, + "learning_rate": 1.5622784810126582e-05, + "loss": 0.0, + "step": 23550 + }, + { + "epoch": 235.75, + "learning_rate": 1.559905063291139e-05, + "loss": 0.0, + "step": 23575 + }, + { + "epoch": 236.0, + "learning_rate": 1.55753164556962e-05, + "loss": 0.0, + "step": 23600 + }, + { + "epoch": 236.25, + "learning_rate": 1.555158227848101e-05, + "loss": 0.0, + "step": 23625 + }, + { + "epoch": 236.5, + "learning_rate": 1.552784810126582e-05, + "loss": 0.0, + "step": 23650 + }, + { + "epoch": 236.75, + "learning_rate": 1.5504113924050633e-05, + "loss": 0.0, + "step": 23675 + }, + { + "epoch": 237.0, + "learning_rate": 1.5480379746835442e-05, + "loss": 0.0, + "step": 23700 + }, + { + "epoch": 237.25, + "learning_rate": 1.5456645569620252e-05, + "loss": 0.0, + "step": 23725 + }, + { + "epoch": 237.5, + "learning_rate": 1.543291139240506e-05, + "loss": 0.0, + "step": 23750 + }, + { + "epoch": 237.75, + "learning_rate": 1.5409177215189874e-05, + "loss": 0.0, + "step": 23775 + }, + { + "epoch": 238.0, + "learning_rate": 1.5385443037974684e-05, + "loss": 0.0, + "step": 23800 + }, + { + "epoch": 238.25, + "learning_rate": 1.5361708860759493e-05, + "loss": 0.0, + "step": 23825 + }, + { + "epoch": 238.5, + "learning_rate": 1.5337974683544303e-05, + "loss": 0.0, + "step": 23850 + }, + { + "epoch": 238.75, + "learning_rate": 1.5314240506329112e-05, + "loss": 0.0, + "step": 23875 + }, + { + "epoch": 239.0, + "learning_rate": 1.529050632911392e-05, + "loss": 0.0, + "step": 23900 + }, + { + "epoch": 239.25, + "learning_rate": 1.526677215189873e-05, + "loss": 0.0, + "step": 23925 + }, + { + "epoch": 239.5, + "learning_rate": 1.5243037974683542e-05, + "loss": 0.0, + "step": 23950 + }, + { + "epoch": 239.75, + "learning_rate": 1.5219303797468352e-05, + "loss": 0.0, + "step": 23975 + }, + { + "epoch": 240.0, + "learning_rate": 1.5195569620253163e-05, + "loss": 0.0, + "step": 24000 + }, + { + "epoch": 240.0, + "eval_loss": 0.5731757879257202, + "eval_runtime": 80.5462, + "eval_samples_per_second": 158.195, + "eval_steps_per_second": 1.242, + "eval_wer": 19.196498358605595, + "step": 24000 + }, + { + "epoch": 240.25, + "learning_rate": 1.5171835443037973e-05, + "loss": 0.0, + "step": 24025 + }, + { + "epoch": 240.5, + "learning_rate": 1.5148101265822782e-05, + "loss": 0.0, + "step": 24050 + }, + { + "epoch": 240.75, + "learning_rate": 1.5124367088607595e-05, + "loss": 0.0, + "step": 24075 + }, + { + "epoch": 241.0, + "learning_rate": 1.5100632911392404e-05, + "loss": 0.0, + "step": 24100 + }, + { + "epoch": 241.25, + "learning_rate": 1.5076898734177214e-05, + "loss": 0.0, + "step": 24125 + }, + { + "epoch": 241.5, + "learning_rate": 1.5053164556962025e-05, + "loss": 0.0, + "step": 24150 + }, + { + "epoch": 241.75, + "learning_rate": 1.5029430379746835e-05, + "loss": 0.0, + "step": 24175 + }, + { + "epoch": 242.0, + "learning_rate": 1.5005696202531644e-05, + "loss": 0.0, + "step": 24200 + }, + { + "epoch": 242.25, + "learning_rate": 1.4981962025316455e-05, + "loss": 0.0, + "step": 24225 + }, + { + "epoch": 242.5, + "learning_rate": 1.4958227848101265e-05, + "loss": 0.0, + "step": 24250 + }, + { + "epoch": 242.75, + "learning_rate": 1.4934493670886074e-05, + "loss": 0.0, + "step": 24275 + }, + { + "epoch": 243.0, + "learning_rate": 1.4910759493670884e-05, + "loss": 0.0, + "step": 24300 + }, + { + "epoch": 243.25, + "learning_rate": 1.4887025316455695e-05, + "loss": 0.0, + "step": 24325 + }, + { + "epoch": 243.5, + "learning_rate": 1.4863291139240505e-05, + "loss": 0.0, + "step": 24350 + }, + { + "epoch": 243.75, + "learning_rate": 1.4839556962025314e-05, + "loss": 0.0, + "step": 24375 + }, + { + "epoch": 244.0, + "learning_rate": 1.4816772151898732e-05, + "loss": 0.0, + "step": 24400 + }, + { + "epoch": 244.25, + "learning_rate": 1.4793037974683542e-05, + "loss": 0.0, + "step": 24425 + }, + { + "epoch": 244.5, + "learning_rate": 1.4769303797468353e-05, + "loss": 0.0, + "step": 24450 + }, + { + "epoch": 244.75, + "learning_rate": 1.4745569620253164e-05, + "loss": 0.0, + "step": 24475 + }, + { + "epoch": 245.0, + "learning_rate": 1.4721835443037973e-05, + "loss": 0.0, + "step": 24500 + }, + { + "epoch": 245.25, + "learning_rate": 1.4698101265822785e-05, + "loss": 0.0, + "step": 24525 + }, + { + "epoch": 245.5, + "learning_rate": 1.4674367088607594e-05, + "loss": 0.0, + "step": 24550 + }, + { + "epoch": 245.75, + "learning_rate": 1.4650632911392404e-05, + "loss": 0.0, + "step": 24575 + }, + { + "epoch": 246.0, + "learning_rate": 1.4626898734177215e-05, + "loss": 0.0, + "step": 24600 + }, + { + "epoch": 246.25, + "learning_rate": 1.4603164556962024e-05, + "loss": 0.0, + "step": 24625 + }, + { + "epoch": 246.5, + "learning_rate": 1.4579430379746834e-05, + "loss": 0.0, + "step": 24650 + }, + { + "epoch": 246.75, + "learning_rate": 1.4555696202531643e-05, + "loss": 0.0, + "step": 24675 + }, + { + "epoch": 247.0, + "learning_rate": 1.4531962025316455e-05, + "loss": 0.0, + "step": 24700 + }, + { + "epoch": 247.25, + "learning_rate": 1.4508227848101264e-05, + "loss": 0.0, + "step": 24725 + }, + { + "epoch": 247.5, + "learning_rate": 1.4484493670886074e-05, + "loss": 0.0, + "step": 24750 + }, + { + "epoch": 247.75, + "learning_rate": 1.4460759493670885e-05, + "loss": 0.0, + "step": 24775 + }, + { + "epoch": 248.0, + "learning_rate": 1.4437025316455694e-05, + "loss": 0.0, + "step": 24800 + }, + { + "epoch": 248.25, + "learning_rate": 1.4413291139240504e-05, + "loss": 0.0, + "step": 24825 + }, + { + "epoch": 248.5, + "learning_rate": 1.4389556962025315e-05, + "loss": 0.0, + "step": 24850 + }, + { + "epoch": 248.75, + "learning_rate": 1.4365822784810126e-05, + "loss": 0.0, + "step": 24875 + }, + { + "epoch": 249.0, + "learning_rate": 1.4342088607594936e-05, + "loss": 0.0, + "step": 24900 + }, + { + "epoch": 249.25, + "learning_rate": 1.4318354430379747e-05, + "loss": 0.0, + "step": 24925 + }, + { + "epoch": 249.5, + "learning_rate": 1.4294620253164556e-05, + "loss": 0.0, + "step": 24950 + }, + { + "epoch": 249.75, + "learning_rate": 1.4270886075949366e-05, + "loss": 0.0, + "step": 24975 + }, + { + "epoch": 250.0, + "learning_rate": 1.4247151898734177e-05, + "loss": 0.0, + "step": 25000 + }, + { + "epoch": 250.0, + "eval_loss": 0.5863945484161377, + "eval_runtime": 80.5107, + "eval_samples_per_second": 158.265, + "eval_steps_per_second": 1.242, + "eval_wer": 19.221900891042676, + "step": 25000 + }, + { + "epoch": 250.25, + "learning_rate": 1.4223417721518987e-05, + "loss": 0.0, + "step": 25025 + }, + { + "epoch": 250.5, + "learning_rate": 1.4199683544303796e-05, + "loss": 0.0, + "step": 25050 + }, + { + "epoch": 250.75, + "learning_rate": 1.4175949367088606e-05, + "loss": 0.0, + "step": 25075 + }, + { + "epoch": 251.0, + "learning_rate": 1.4152215189873417e-05, + "loss": 0.0, + "step": 25100 + }, + { + "epoch": 251.25, + "learning_rate": 1.4128481012658226e-05, + "loss": 0.0, + "step": 25125 + }, + { + "epoch": 251.5, + "learning_rate": 1.4104746835443036e-05, + "loss": 0.0, + "step": 25150 + }, + { + "epoch": 251.75, + "learning_rate": 1.4081012658227847e-05, + "loss": 0.0, + "step": 25175 + }, + { + "epoch": 252.0, + "learning_rate": 1.4057278481012657e-05, + "loss": 0.0, + "step": 25200 + }, + { + "epoch": 252.25, + "learning_rate": 1.4033544303797466e-05, + "loss": 0.0, + "step": 25225 + }, + { + "epoch": 252.5, + "learning_rate": 1.4009810126582276e-05, + "loss": 0.0, + "step": 25250 + }, + { + "epoch": 252.75, + "learning_rate": 1.3986075949367087e-05, + "loss": 0.0, + "step": 25275 + }, + { + "epoch": 253.0, + "learning_rate": 1.3962341772151898e-05, + "loss": 0.0, + "step": 25300 + }, + { + "epoch": 253.25, + "learning_rate": 1.393860759493671e-05, + "loss": 0.0, + "step": 25325 + }, + { + "epoch": 253.5, + "learning_rate": 1.3914873417721519e-05, + "loss": 0.0, + "step": 25350 + }, + { + "epoch": 253.75, + "learning_rate": 1.3891139240506328e-05, + "loss": 0.0, + "step": 25375 + }, + { + "epoch": 254.0, + "learning_rate": 1.3867405063291138e-05, + "loss": 0.0, + "step": 25400 + }, + { + "epoch": 254.25, + "learning_rate": 1.3843670886075949e-05, + "loss": 0.0, + "step": 25425 + }, + { + "epoch": 254.5, + "learning_rate": 1.3819936708860758e-05, + "loss": 0.0, + "step": 25450 + }, + { + "epoch": 254.75, + "learning_rate": 1.3796202531645568e-05, + "loss": 0.0, + "step": 25475 + }, + { + "epoch": 255.0, + "learning_rate": 1.377246835443038e-05, + "loss": 0.0, + "step": 25500 + }, + { + "epoch": 255.25, + "learning_rate": 1.3748734177215189e-05, + "loss": 0.0, + "step": 25525 + }, + { + "epoch": 255.5, + "learning_rate": 1.3724999999999998e-05, + "loss": 0.0, + "step": 25550 + }, + { + "epoch": 255.75, + "learning_rate": 1.3701265822784808e-05, + "loss": 0.0, + "step": 25575 + }, + { + "epoch": 256.0, + "learning_rate": 1.3677531645569619e-05, + "loss": 0.0, + "step": 25600 + }, + { + "epoch": 256.25, + "learning_rate": 1.3653797468354428e-05, + "loss": 0.0, + "step": 25625 + }, + { + "epoch": 256.5, + "learning_rate": 1.3630063291139238e-05, + "loss": 0.0, + "step": 25650 + }, + { + "epoch": 256.75, + "learning_rate": 1.3606329113924049e-05, + "loss": 0.0, + "step": 25675 + }, + { + "epoch": 257.0, + "learning_rate": 1.358259493670886e-05, + "loss": 0.0, + "step": 25700 + }, + { + "epoch": 257.25, + "learning_rate": 1.355886075949367e-05, + "loss": 0.0, + "step": 25725 + }, + { + "epoch": 257.5, + "learning_rate": 1.3535126582278481e-05, + "loss": 0.0, + "step": 25750 + }, + { + "epoch": 257.75, + "learning_rate": 1.351139240506329e-05, + "loss": 0.0, + "step": 25775 + }, + { + "epoch": 258.0, + "learning_rate": 1.34876582278481e-05, + "loss": 0.0, + "step": 25800 + }, + { + "epoch": 258.25, + "learning_rate": 1.3463924050632911e-05, + "loss": 0.0, + "step": 25825 + }, + { + "epoch": 258.5, + "learning_rate": 1.344018987341772e-05, + "loss": 0.0, + "step": 25850 + }, + { + "epoch": 258.75, + "learning_rate": 1.341645569620253e-05, + "loss": 0.0, + "step": 25875 + }, + { + "epoch": 259.0, + "learning_rate": 1.3392721518987341e-05, + "loss": 0.0, + "step": 25900 + }, + { + "epoch": 259.25, + "learning_rate": 1.3368987341772151e-05, + "loss": 0.0, + "step": 25925 + }, + { + "epoch": 259.5, + "learning_rate": 1.334525316455696e-05, + "loss": 0.0, + "step": 25950 + }, + { + "epoch": 259.75, + "learning_rate": 1.332151898734177e-05, + "loss": 0.0, + "step": 25975 + }, + { + "epoch": 260.0, + "learning_rate": 1.3297784810126581e-05, + "loss": 0.0, + "step": 26000 + }, + { + "epoch": 260.0, + "eval_loss": 0.6003339886665344, + "eval_runtime": 80.5514, + "eval_samples_per_second": 158.185, + "eval_steps_per_second": 1.241, + "eval_wer": 19.310809754572457, + "step": 26000 + }, + { + "epoch": 260.25, + "learning_rate": 1.327405063291139e-05, + "loss": 0.0, + "step": 26025 + }, + { + "epoch": 260.5, + "learning_rate": 1.32503164556962e-05, + "loss": 0.0, + "step": 26050 + }, + { + "epoch": 260.75, + "learning_rate": 1.3226582278481011e-05, + "loss": 0.0, + "step": 26075 + }, + { + "epoch": 261.0, + "learning_rate": 1.3202848101265821e-05, + "loss": 0.0, + "step": 26100 + }, + { + "epoch": 261.25, + "learning_rate": 1.3179113924050632e-05, + "loss": 0.0, + "step": 26125 + }, + { + "epoch": 261.5, + "learning_rate": 1.3155379746835443e-05, + "loss": 0.0, + "step": 26150 + }, + { + "epoch": 261.75, + "learning_rate": 1.3131645569620253e-05, + "loss": 0.0, + "step": 26175 + }, + { + "epoch": 262.0, + "learning_rate": 1.3107911392405062e-05, + "loss": 0.0, + "step": 26200 + }, + { + "epoch": 262.25, + "learning_rate": 1.3084177215189874e-05, + "loss": 0.0, + "step": 26225 + }, + { + "epoch": 262.5, + "learning_rate": 1.3060443037974683e-05, + "loss": 0.0, + "step": 26250 + }, + { + "epoch": 262.75, + "learning_rate": 1.3036708860759493e-05, + "loss": 0.0, + "step": 26275 + }, + { + "epoch": 263.0, + "learning_rate": 1.3012974683544302e-05, + "loss": 0.0, + "step": 26300 + }, + { + "epoch": 263.25, + "learning_rate": 1.2989240506329113e-05, + "loss": 0.0, + "step": 26325 + }, + { + "epoch": 263.5, + "learning_rate": 1.2965506329113923e-05, + "loss": 0.0, + "step": 26350 + }, + { + "epoch": 263.75, + "learning_rate": 1.2941772151898732e-05, + "loss": 0.0, + "step": 26375 + }, + { + "epoch": 264.0, + "learning_rate": 1.2918037974683543e-05, + "loss": 0.0, + "step": 26400 + }, + { + "epoch": 264.25, + "learning_rate": 1.2894303797468353e-05, + "loss": 0.0, + "step": 26425 + }, + { + "epoch": 264.5, + "learning_rate": 1.2870569620253162e-05, + "loss": 0.0, + "step": 26450 + }, + { + "epoch": 264.75, + "learning_rate": 1.284778481012658e-05, + "loss": 0.0, + "step": 26475 + }, + { + "epoch": 265.0, + "learning_rate": 1.282405063291139e-05, + "loss": 0.0, + "step": 26500 + }, + { + "epoch": 265.25, + "learning_rate": 1.2800316455696203e-05, + "loss": 0.0, + "step": 26525 + }, + { + "epoch": 265.5, + "learning_rate": 1.2776582278481012e-05, + "loss": 0.0, + "step": 26550 + }, + { + "epoch": 265.75, + "learning_rate": 1.2752848101265822e-05, + "loss": 0.0, + "step": 26575 + }, + { + "epoch": 266.0, + "learning_rate": 1.2729113924050633e-05, + "loss": 0.0, + "step": 26600 + }, + { + "epoch": 266.25, + "learning_rate": 1.2705379746835443e-05, + "loss": 0.0, + "step": 26625 + }, + { + "epoch": 266.5, + "learning_rate": 1.2681645569620252e-05, + "loss": 0.0, + "step": 26650 + }, + { + "epoch": 266.75, + "learning_rate": 1.2657911392405062e-05, + "loss": 0.0, + "step": 26675 + }, + { + "epoch": 267.0, + "learning_rate": 1.2634177215189873e-05, + "loss": 0.0, + "step": 26700 + }, + { + "epoch": 267.25, + "learning_rate": 1.2610443037974682e-05, + "loss": 0.0, + "step": 26725 + }, + { + "epoch": 267.5, + "learning_rate": 1.2586708860759492e-05, + "loss": 0.0, + "step": 26750 + }, + { + "epoch": 267.75, + "learning_rate": 1.2562974683544303e-05, + "loss": 0.0, + "step": 26775 + }, + { + "epoch": 268.0, + "learning_rate": 1.2539240506329112e-05, + "loss": 0.0, + "step": 26800 + }, + { + "epoch": 268.25, + "learning_rate": 1.2515506329113922e-05, + "loss": 0.0, + "step": 26825 + }, + { + "epoch": 268.5, + "learning_rate": 1.2491772151898733e-05, + "loss": 0.0, + "step": 26850 + }, + { + "epoch": 268.75, + "learning_rate": 1.2468037974683543e-05, + "loss": 0.0, + "step": 26875 + }, + { + "epoch": 269.0, + "learning_rate": 1.2444303797468352e-05, + "loss": 0.0, + "step": 26900 + }, + { + "epoch": 269.25, + "learning_rate": 1.2420569620253165e-05, + "loss": 0.0, + "step": 26925 + }, + { + "epoch": 269.5, + "learning_rate": 1.2396835443037975e-05, + "loss": 0.0, + "step": 26950 + }, + { + "epoch": 269.75, + "learning_rate": 1.2373101265822784e-05, + "loss": 0.0, + "step": 26975 + }, + { + "epoch": 270.0, + "learning_rate": 1.2349367088607595e-05, + "loss": 0.0, + "step": 27000 + }, + { + "epoch": 270.0, + "eval_loss": 0.6140475273132324, + "eval_runtime": 80.6529, + "eval_samples_per_second": 157.986, + "eval_steps_per_second": 1.24, + "eval_wer": 19.371385024230108, + "step": 27000 + }, + { + "epoch": 270.25, + "learning_rate": 1.2325632911392405e-05, + "loss": 0.0, + "step": 27025 + }, + { + "epoch": 270.5, + "learning_rate": 1.2301898734177214e-05, + "loss": 0.0, + "step": 27050 + }, + { + "epoch": 270.75, + "learning_rate": 1.2278164556962024e-05, + "loss": 0.0, + "step": 27075 + }, + { + "epoch": 271.0, + "learning_rate": 1.2254430379746835e-05, + "loss": 0.0, + "step": 27100 + }, + { + "epoch": 271.25, + "learning_rate": 1.2230696202531645e-05, + "loss": 0.0, + "step": 27125 + }, + { + "epoch": 271.5, + "learning_rate": 1.2206962025316454e-05, + "loss": 0.0, + "step": 27150 + }, + { + "epoch": 271.75, + "learning_rate": 1.2183227848101265e-05, + "loss": 0.0, + "step": 27175 + }, + { + "epoch": 272.0, + "learning_rate": 1.2159493670886075e-05, + "loss": 0.0, + "step": 27200 + }, + { + "epoch": 272.25, + "learning_rate": 1.2135759493670884e-05, + "loss": 0.0, + "step": 27225 + }, + { + "epoch": 272.5, + "learning_rate": 1.2112025316455694e-05, + "loss": 0.0, + "step": 27250 + }, + { + "epoch": 272.75, + "learning_rate": 1.2088291139240505e-05, + "loss": 0.0, + "step": 27275 + }, + { + "epoch": 273.0, + "learning_rate": 1.2064556962025314e-05, + "loss": 0.0, + "step": 27300 + }, + { + "epoch": 273.25, + "learning_rate": 1.2040822784810124e-05, + "loss": 0.0, + "step": 27325 + }, + { + "epoch": 273.5, + "learning_rate": 1.2017088607594937e-05, + "loss": 0.0, + "step": 27350 + }, + { + "epoch": 273.75, + "learning_rate": 1.1993354430379746e-05, + "loss": 0.0, + "step": 27375 + }, + { + "epoch": 274.0, + "learning_rate": 1.1969620253164556e-05, + "loss": 0.0, + "step": 27400 + }, + { + "epoch": 274.25, + "learning_rate": 1.1945886075949367e-05, + "loss": 0.0, + "step": 27425 + }, + { + "epoch": 274.5, + "learning_rate": 1.1922151898734177e-05, + "loss": 0.0, + "step": 27450 + }, + { + "epoch": 274.75, + "learning_rate": 1.1898417721518986e-05, + "loss": 0.0, + "step": 27475 + }, + { + "epoch": 275.0, + "learning_rate": 1.1874683544303797e-05, + "loss": 0.0, + "step": 27500 + }, + { + "epoch": 275.25, + "learning_rate": 1.1850949367088607e-05, + "loss": 0.0, + "step": 27525 + }, + { + "epoch": 275.5, + "learning_rate": 1.1827215189873416e-05, + "loss": 0.0, + "step": 27550 + }, + { + "epoch": 275.75, + "learning_rate": 1.1803481012658226e-05, + "loss": 0.0, + "step": 27575 + }, + { + "epoch": 276.0, + "learning_rate": 1.1779746835443037e-05, + "loss": 0.0, + "step": 27600 + }, + { + "epoch": 276.25, + "learning_rate": 1.1756012658227847e-05, + "loss": 0.0, + "step": 27625 + }, + { + "epoch": 276.5, + "learning_rate": 1.1732278481012656e-05, + "loss": 0.0, + "step": 27650 + }, + { + "epoch": 276.75, + "learning_rate": 1.1708544303797467e-05, + "loss": 0.0, + "step": 27675 + }, + { + "epoch": 277.0, + "learning_rate": 1.1684810126582277e-05, + "loss": 0.0, + "step": 27700 + }, + { + "epoch": 277.25, + "learning_rate": 1.1661075949367086e-05, + "loss": 0.0, + "step": 27725 + }, + { + "epoch": 277.5, + "learning_rate": 1.1637341772151899e-05, + "loss": 0.0, + "step": 27750 + }, + { + "epoch": 277.75, + "learning_rate": 1.1613607594936709e-05, + "loss": 0.0, + "step": 27775 + }, + { + "epoch": 278.0, + "learning_rate": 1.1589873417721518e-05, + "loss": 0.0, + "step": 27800 + }, + { + "epoch": 278.25, + "learning_rate": 1.156613924050633e-05, + "loss": 0.0, + "step": 27825 + }, + { + "epoch": 278.5, + "learning_rate": 1.1542405063291139e-05, + "loss": 0.0, + "step": 27850 + }, + { + "epoch": 278.75, + "learning_rate": 1.1518670886075948e-05, + "loss": 0.0, + "step": 27875 + }, + { + "epoch": 279.0, + "learning_rate": 1.1496835443037973e-05, + "loss": 0.0033, + "step": 27900 + }, + { + "epoch": 279.25, + "learning_rate": 1.1474050632911391e-05, + "loss": 0.0072, + "step": 27925 + }, + { + "epoch": 279.5, + "learning_rate": 1.14503164556962e-05, + "loss": 0.0052, + "step": 27950 + }, + { + "epoch": 279.75, + "learning_rate": 1.1426582278481012e-05, + "loss": 0.004, + "step": 27975 + }, + { + "epoch": 280.0, + "learning_rate": 1.1402848101265821e-05, + "loss": 0.0034, + "step": 28000 + }, + { + "epoch": 280.0, + "eval_loss": 0.5535955429077148, + "eval_runtime": 81.9868, + "eval_samples_per_second": 155.415, + "eval_steps_per_second": 1.22, + "eval_wer": 20.663006096607788, + "step": 28000 + }, + { + "epoch": 280.25, + "learning_rate": 1.137911392405063e-05, + "loss": 0.0018, + "step": 28025 + }, + { + "epoch": 280.5, + "learning_rate": 1.1355379746835442e-05, + "loss": 0.0018, + "step": 28050 + }, + { + "epoch": 280.75, + "learning_rate": 1.1331645569620251e-05, + "loss": 0.0015, + "step": 28075 + }, + { + "epoch": 281.0, + "learning_rate": 1.1307911392405061e-05, + "loss": 0.0015, + "step": 28100 + }, + { + "epoch": 281.25, + "learning_rate": 1.1284177215189874e-05, + "loss": 0.0003, + "step": 28125 + }, + { + "epoch": 281.5, + "learning_rate": 1.1260443037974683e-05, + "loss": 0.0004, + "step": 28150 + }, + { + "epoch": 281.75, + "learning_rate": 1.1236708860759493e-05, + "loss": 0.0004, + "step": 28175 + }, + { + "epoch": 282.0, + "learning_rate": 1.1212974683544304e-05, + "loss": 0.0003, + "step": 28200 + }, + { + "epoch": 282.25, + "learning_rate": 1.1189240506329114e-05, + "loss": 0.0001, + "step": 28225 + }, + { + "epoch": 282.5, + "learning_rate": 1.1165506329113923e-05, + "loss": 0.0001, + "step": 28250 + }, + { + "epoch": 282.75, + "learning_rate": 1.1141772151898733e-05, + "loss": 0.0001, + "step": 28275 + }, + { + "epoch": 283.0, + "learning_rate": 1.1118037974683544e-05, + "loss": 0.0001, + "step": 28300 + }, + { + "epoch": 283.25, + "learning_rate": 1.1094303797468353e-05, + "loss": 0.0001, + "step": 28325 + }, + { + "epoch": 283.5, + "learning_rate": 1.1070569620253163e-05, + "loss": 0.0001, + "step": 28350 + }, + { + "epoch": 283.75, + "learning_rate": 1.1046835443037974e-05, + "loss": 0.0, + "step": 28375 + }, + { + "epoch": 284.0, + "learning_rate": 1.1023101265822783e-05, + "loss": 0.0003, + "step": 28400 + }, + { + "epoch": 284.25, + "learning_rate": 1.0999367088607593e-05, + "loss": 0.0, + "step": 28425 + }, + { + "epoch": 284.5, + "learning_rate": 1.0975632911392404e-05, + "loss": 0.0, + "step": 28450 + }, + { + "epoch": 284.75, + "learning_rate": 1.0951898734177214e-05, + "loss": 0.0, + "step": 28475 + }, + { + "epoch": 285.0, + "learning_rate": 1.0928164556962023e-05, + "loss": 0.0, + "step": 28500 + }, + { + "epoch": 285.25, + "learning_rate": 1.0904430379746836e-05, + "loss": 0.0, + "step": 28525 + }, + { + "epoch": 285.5, + "learning_rate": 1.0880696202531646e-05, + "loss": 0.0, + "step": 28550 + }, + { + "epoch": 285.75, + "learning_rate": 1.0856962025316455e-05, + "loss": 0.0, + "step": 28575 + }, + { + "epoch": 286.0, + "learning_rate": 1.0833227848101266e-05, + "loss": 0.0, + "step": 28600 + }, + { + "epoch": 286.25, + "learning_rate": 1.0809493670886076e-05, + "loss": 0.0, + "step": 28625 + }, + { + "epoch": 286.5, + "learning_rate": 1.0785759493670885e-05, + "loss": 0.0, + "step": 28650 + }, + { + "epoch": 286.75, + "learning_rate": 1.0762025316455695e-05, + "loss": 0.0, + "step": 28675 + }, + { + "epoch": 287.0, + "learning_rate": 1.0738291139240506e-05, + "loss": 0.0, + "step": 28700 + }, + { + "epoch": 287.25, + "learning_rate": 1.0714556962025316e-05, + "loss": 0.0, + "step": 28725 + }, + { + "epoch": 287.5, + "learning_rate": 1.0690822784810125e-05, + "loss": 0.0, + "step": 28750 + }, + { + "epoch": 287.75, + "learning_rate": 1.0667088607594936e-05, + "loss": 0.0, + "step": 28775 + }, + { + "epoch": 288.0, + "learning_rate": 1.0643354430379746e-05, + "loss": 0.0, + "step": 28800 + }, + { + "epoch": 288.25, + "learning_rate": 1.0619620253164555e-05, + "loss": 0.0, + "step": 28825 + }, + { + "epoch": 288.5, + "learning_rate": 1.0595886075949365e-05, + "loss": 0.0, + "step": 28850 + }, + { + "epoch": 288.75, + "learning_rate": 1.0572151898734176e-05, + "loss": 0.0, + "step": 28875 + }, + { + "epoch": 289.0, + "learning_rate": 1.0548417721518985e-05, + "loss": 0.0, + "step": 28900 + }, + { + "epoch": 289.25, + "learning_rate": 1.0524683544303795e-05, + "loss": 0.0, + "step": 28925 + }, + { + "epoch": 289.5, + "learning_rate": 1.0500949367088608e-05, + "loss": 0.0, + "step": 28950 + }, + { + "epoch": 289.75, + "learning_rate": 1.0477215189873417e-05, + "loss": 0.0, + "step": 28975 + }, + { + "epoch": 290.0, + "learning_rate": 1.0453481012658227e-05, + "loss": 0.0, + "step": 29000 + }, + { + "epoch": 290.0, + "eval_loss": 0.5485877394676208, + "eval_runtime": 80.6015, + "eval_samples_per_second": 158.086, + "eval_steps_per_second": 1.241, + "eval_wer": 19.339143348444583, + "step": 29000 + }, + { + "epoch": 290.25, + "learning_rate": 1.0429746835443038e-05, + "loss": 0.0, + "step": 29025 + }, + { + "epoch": 290.5, + "learning_rate": 1.0406012658227848e-05, + "loss": 0.0, + "step": 29050 + }, + { + "epoch": 290.75, + "learning_rate": 1.0382278481012657e-05, + "loss": 0.0, + "step": 29075 + }, + { + "epoch": 291.0, + "learning_rate": 1.0358544303797468e-05, + "loss": 0.0, + "step": 29100 + }, + { + "epoch": 291.25, + "learning_rate": 1.0334810126582278e-05, + "loss": 0.0, + "step": 29125 + }, + { + "epoch": 291.5, + "learning_rate": 1.0311075949367087e-05, + "loss": 0.0, + "step": 29150 + }, + { + "epoch": 291.75, + "learning_rate": 1.0287341772151897e-05, + "loss": 0.0, + "step": 29175 + }, + { + "epoch": 292.0, + "learning_rate": 1.0263607594936708e-05, + "loss": 0.0, + "step": 29200 + }, + { + "epoch": 292.25, + "learning_rate": 1.0239873417721518e-05, + "loss": 0.0, + "step": 29225 + }, + { + "epoch": 292.5, + "learning_rate": 1.0216139240506327e-05, + "loss": 0.0, + "step": 29250 + }, + { + "epoch": 292.75, + "learning_rate": 1.0192405063291138e-05, + "loss": 0.0, + "step": 29275 + }, + { + "epoch": 293.0, + "learning_rate": 1.0168670886075948e-05, + "loss": 0.0, + "step": 29300 + }, + { + "epoch": 293.25, + "learning_rate": 1.0144936708860757e-05, + "loss": 0.0, + "step": 29325 + }, + { + "epoch": 293.5, + "learning_rate": 1.012120253164557e-05, + "loss": 0.0, + "step": 29350 + }, + { + "epoch": 293.75, + "learning_rate": 1.009746835443038e-05, + "loss": 0.0, + "step": 29375 + }, + { + "epoch": 294.0, + "learning_rate": 1.0073734177215189e-05, + "loss": 0.0, + "step": 29400 + }, + { + "epoch": 294.25, + "learning_rate": 1.005e-05, + "loss": 0.0, + "step": 29425 + }, + { + "epoch": 294.5, + "learning_rate": 1.002626582278481e-05, + "loss": 0.0, + "step": 29450 + }, + { + "epoch": 294.75, + "learning_rate": 1.000253164556962e-05, + "loss": 0.0, + "step": 29475 + }, + { + "epoch": 295.0, + "learning_rate": 9.97879746835443e-06, + "loss": 0.0, + "step": 29500 + }, + { + "epoch": 295.25, + "learning_rate": 9.95506329113924e-06, + "loss": 0.0, + "step": 29525 + }, + { + "epoch": 295.5, + "learning_rate": 9.93132911392405e-06, + "loss": 0.0, + "step": 29550 + }, + { + "epoch": 295.75, + "learning_rate": 9.907594936708859e-06, + "loss": 0.0, + "step": 29575 + }, + { + "epoch": 296.0, + "learning_rate": 9.88386075949367e-06, + "loss": 0.0, + "step": 29600 + }, + { + "epoch": 296.25, + "learning_rate": 9.86012658227848e-06, + "loss": 0.0, + "step": 29625 + }, + { + "epoch": 296.5, + "learning_rate": 9.83639240506329e-06, + "loss": 0.0, + "step": 29650 + }, + { + "epoch": 296.75, + "learning_rate": 9.8126582278481e-06, + "loss": 0.0, + "step": 29675 + }, + { + "epoch": 297.0, + "learning_rate": 9.78892405063291e-06, + "loss": 0.0, + "step": 29700 + }, + { + "epoch": 297.25, + "learning_rate": 9.76518987341772e-06, + "loss": 0.0, + "step": 29725 + }, + { + "epoch": 297.5, + "learning_rate": 9.741455696202532e-06, + "loss": 0.0, + "step": 29750 + }, + { + "epoch": 297.75, + "learning_rate": 9.717721518987342e-06, + "loss": 0.0, + "step": 29775 + }, + { + "epoch": 298.0, + "learning_rate": 9.693987341772151e-06, + "loss": 0.0, + "step": 29800 + }, + { + "epoch": 298.25, + "learning_rate": 9.670253164556963e-06, + "loss": 0.0, + "step": 29825 + }, + { + "epoch": 298.5, + "learning_rate": 9.646518987341772e-06, + "loss": 0.0, + "step": 29850 + }, + { + "epoch": 298.75, + "learning_rate": 9.622784810126582e-06, + "loss": 0.0, + "step": 29875 + }, + { + "epoch": 299.0, + "learning_rate": 9.599050632911391e-06, + "loss": 0.0, + "step": 29900 + }, + { + "epoch": 299.25, + "learning_rate": 9.575316455696202e-06, + "loss": 0.0, + "step": 29925 + }, + { + "epoch": 299.5, + "learning_rate": 9.551582278481012e-06, + "loss": 0.0, + "step": 29950 + }, + { + "epoch": 299.75, + "learning_rate": 9.527848101265821e-06, + "loss": 0.0, + "step": 29975 + }, + { + "epoch": 300.0, + "learning_rate": 9.504113924050633e-06, + "loss": 0.0, + "step": 30000 + }, + { + "epoch": 300.0, + "eval_loss": 0.5590522885322571, + "eval_runtime": 80.5922, + "eval_samples_per_second": 158.105, + "eval_steps_per_second": 1.241, + "eval_wer": 19.30592465218071, + "step": 30000 + }, + { + "epoch": 300.25, + "learning_rate": 9.480379746835442e-06, + "loss": 0.0, + "step": 30025 + }, + { + "epoch": 300.5, + "learning_rate": 9.456645569620252e-06, + "loss": 0.0, + "step": 30050 + }, + { + "epoch": 300.75, + "learning_rate": 9.432911392405061e-06, + "loss": 0.0, + "step": 30075 + }, + { + "epoch": 301.0, + "learning_rate": 9.409177215189872e-06, + "loss": 0.0, + "step": 30100 + }, + { + "epoch": 301.25, + "learning_rate": 9.385443037974682e-06, + "loss": 0.0, + "step": 30125 + }, + { + "epoch": 301.5, + "learning_rate": 9.361708860759493e-06, + "loss": 0.0, + "step": 30150 + }, + { + "epoch": 301.75, + "learning_rate": 9.337974683544302e-06, + "loss": 0.0, + "step": 30175 + }, + { + "epoch": 302.0, + "learning_rate": 9.314240506329114e-06, + "loss": 0.0, + "step": 30200 + }, + { + "epoch": 302.25, + "learning_rate": 9.290506329113923e-06, + "loss": 0.0, + "step": 30225 + }, + { + "epoch": 302.5, + "learning_rate": 9.266772151898733e-06, + "loss": 0.0, + "step": 30250 + }, + { + "epoch": 302.75, + "learning_rate": 9.243037974683544e-06, + "loss": 0.0, + "step": 30275 + }, + { + "epoch": 303.0, + "learning_rate": 9.219303797468353e-06, + "loss": 0.0, + "step": 30300 + }, + { + "epoch": 303.25, + "learning_rate": 9.195569620253165e-06, + "loss": 0.0, + "step": 30325 + }, + { + "epoch": 303.5, + "learning_rate": 9.171835443037974e-06, + "loss": 0.0, + "step": 30350 + }, + { + "epoch": 303.75, + "learning_rate": 9.148101265822784e-06, + "loss": 0.0, + "step": 30375 + }, + { + "epoch": 304.0, + "learning_rate": 9.124367088607593e-06, + "loss": 0.0, + "step": 30400 + }, + { + "epoch": 304.25, + "learning_rate": 9.100632911392404e-06, + "loss": 0.0, + "step": 30425 + }, + { + "epoch": 304.5, + "learning_rate": 9.076898734177214e-06, + "loss": 0.0, + "step": 30450 + }, + { + "epoch": 304.75, + "learning_rate": 9.053164556962025e-06, + "loss": 0.0, + "step": 30475 + }, + { + "epoch": 305.0, + "learning_rate": 9.029430379746835e-06, + "loss": 0.0, + "step": 30500 + }, + { + "epoch": 305.25, + "learning_rate": 9.005696202531646e-06, + "loss": 0.0, + "step": 30525 + }, + { + "epoch": 305.5, + "learning_rate": 8.981962025316455e-06, + "loss": 0.0, + "step": 30550 + }, + { + "epoch": 305.75, + "learning_rate": 8.958227848101265e-06, + "loss": 0.0, + "step": 30575 + }, + { + "epoch": 306.0, + "learning_rate": 8.934493670886074e-06, + "loss": 0.0, + "step": 30600 + }, + { + "epoch": 306.25, + "learning_rate": 8.910759493670885e-06, + "loss": 0.0, + "step": 30625 + }, + { + "epoch": 306.5, + "learning_rate": 8.887025316455695e-06, + "loss": 0.0, + "step": 30650 + }, + { + "epoch": 306.75, + "learning_rate": 8.863291139240504e-06, + "loss": 0.0, + "step": 30675 + }, + { + "epoch": 307.0, + "learning_rate": 8.839556962025316e-06, + "loss": 0.0, + "step": 30700 + }, + { + "epoch": 307.25, + "learning_rate": 8.815822784810127e-06, + "loss": 0.0, + "step": 30725 + }, + { + "epoch": 307.5, + "learning_rate": 8.792088607594936e-06, + "loss": 0.0, + "step": 30750 + }, + { + "epoch": 307.75, + "learning_rate": 8.768354430379746e-06, + "loss": 0.0, + "step": 30775 + }, + { + "epoch": 308.0, + "learning_rate": 8.744620253164555e-06, + "loss": 0.0, + "step": 30800 + }, + { + "epoch": 308.25, + "learning_rate": 8.720886075949367e-06, + "loss": 0.0, + "step": 30825 + }, + { + "epoch": 308.5, + "learning_rate": 8.697151898734176e-06, + "loss": 0.0, + "step": 30850 + }, + { + "epoch": 308.75, + "learning_rate": 8.673417721518986e-06, + "loss": 0.0, + "step": 30875 + }, + { + "epoch": 309.0, + "learning_rate": 8.649683544303797e-06, + "loss": 0.0, + "step": 30900 + }, + { + "epoch": 309.25, + "learning_rate": 8.625949367088606e-06, + "loss": 0.0, + "step": 30925 + }, + { + "epoch": 309.5, + "learning_rate": 8.602215189873418e-06, + "loss": 0.0, + "step": 30950 + }, + { + "epoch": 309.75, + "learning_rate": 8.578481012658227e-06, + "loss": 0.0, + "step": 30975 + }, + { + "epoch": 310.0, + "learning_rate": 8.554746835443037e-06, + "loss": 0.0, + "step": 31000 + }, + { + "epoch": 310.0, + "eval_loss": 0.566923975944519, + "eval_runtime": 80.5145, + "eval_samples_per_second": 158.257, + "eval_steps_per_second": 1.242, + "eval_wer": 19.313740816007506, + "step": 31000 + }, + { + "epoch": 310.25, + "learning_rate": 8.531012658227848e-06, + "loss": 0.0, + "step": 31025 + }, + { + "epoch": 310.5, + "learning_rate": 8.507278481012657e-06, + "loss": 0.0, + "step": 31050 + }, + { + "epoch": 310.75, + "learning_rate": 8.483544303797467e-06, + "loss": 0.0, + "step": 31075 + }, + { + "epoch": 311.0, + "learning_rate": 8.459810126582278e-06, + "loss": 0.0, + "step": 31100 + }, + { + "epoch": 311.25, + "learning_rate": 8.436075949367087e-06, + "loss": 0.0, + "step": 31125 + }, + { + "epoch": 311.5, + "learning_rate": 8.412341772151899e-06, + "loss": 0.0, + "step": 31150 + }, + { + "epoch": 311.75, + "learning_rate": 8.388607594936708e-06, + "loss": 0.0, + "step": 31175 + }, + { + "epoch": 312.0, + "learning_rate": 8.364873417721518e-06, + "loss": 0.0, + "step": 31200 + }, + { + "epoch": 312.25, + "learning_rate": 8.341139240506329e-06, + "loss": 0.0, + "step": 31225 + }, + { + "epoch": 312.5, + "learning_rate": 8.317405063291138e-06, + "loss": 0.0, + "step": 31250 + }, + { + "epoch": 312.75, + "learning_rate": 8.293670886075948e-06, + "loss": 0.0, + "step": 31275 + }, + { + "epoch": 313.0, + "learning_rate": 8.269936708860759e-06, + "loss": 0.0, + "step": 31300 + }, + { + "epoch": 313.25, + "learning_rate": 8.246202531645569e-06, + "loss": 0.0, + "step": 31325 + }, + { + "epoch": 313.5, + "learning_rate": 8.22246835443038e-06, + "loss": 0.0, + "step": 31350 + }, + { + "epoch": 313.75, + "learning_rate": 8.19873417721519e-06, + "loss": 0.0, + "step": 31375 + }, + { + "epoch": 314.0, + "learning_rate": 8.174999999999999e-06, + "loss": 0.0, + "step": 31400 + }, + { + "epoch": 314.25, + "learning_rate": 8.15126582278481e-06, + "loss": 0.0, + "step": 31425 + }, + { + "epoch": 314.5, + "learning_rate": 8.12753164556962e-06, + "loss": 0.0, + "step": 31450 + }, + { + "epoch": 314.75, + "learning_rate": 8.103797468354429e-06, + "loss": 0.0, + "step": 31475 + }, + { + "epoch": 315.0, + "learning_rate": 8.08006329113924e-06, + "loss": 0.0, + "step": 31500 + }, + { + "epoch": 315.25, + "learning_rate": 8.05632911392405e-06, + "loss": 0.0, + "step": 31525 + }, + { + "epoch": 315.5, + "learning_rate": 8.032594936708861e-06, + "loss": 0.0, + "step": 31550 + }, + { + "epoch": 315.75, + "learning_rate": 8.00886075949367e-06, + "loss": 0.0, + "step": 31575 + }, + { + "epoch": 316.0, + "learning_rate": 7.98512658227848e-06, + "loss": 0.0, + "step": 31600 + }, + { + "epoch": 316.25, + "learning_rate": 7.961392405063291e-06, + "loss": 0.0, + "step": 31625 + }, + { + "epoch": 316.5, + "learning_rate": 7.9376582278481e-06, + "loss": 0.0, + "step": 31650 + }, + { + "epoch": 316.75, + "learning_rate": 7.91392405063291e-06, + "loss": 0.0, + "step": 31675 + }, + { + "epoch": 317.0, + "learning_rate": 7.89018987341772e-06, + "loss": 0.0, + "step": 31700 + }, + { + "epoch": 317.25, + "learning_rate": 7.866455696202531e-06, + "loss": 0.0, + "step": 31725 + }, + { + "epoch": 317.5, + "learning_rate": 7.842721518987342e-06, + "loss": 0.0, + "step": 31750 + }, + { + "epoch": 317.75, + "learning_rate": 7.818987341772152e-06, + "loss": 0.0, + "step": 31775 + }, + { + "epoch": 318.0, + "learning_rate": 7.795253164556961e-06, + "loss": 0.0, + "step": 31800 + }, + { + "epoch": 318.25, + "learning_rate": 7.77151898734177e-06, + "loss": 0.0, + "step": 31825 + }, + { + "epoch": 318.5, + "learning_rate": 7.747784810126582e-06, + "loss": 0.0, + "step": 31850 + }, + { + "epoch": 318.75, + "learning_rate": 7.724050632911391e-06, + "loss": 0.0, + "step": 31875 + }, + { + "epoch": 319.0, + "learning_rate": 7.7003164556962e-06, + "loss": 0.0, + "step": 31900 + }, + { + "epoch": 319.25, + "learning_rate": 7.676582278481012e-06, + "loss": 0.0, + "step": 31925 + }, + { + "epoch": 319.5, + "learning_rate": 7.652848101265823e-06, + "loss": 0.0, + "step": 31950 + }, + { + "epoch": 319.75, + "learning_rate": 7.629113924050633e-06, + "loss": 0.0, + "step": 31975 + }, + { + "epoch": 320.0, + "learning_rate": 7.605379746835442e-06, + "loss": 0.0, + "step": 32000 + }, + { + "epoch": 320.0, + "eval_loss": 0.573691189289093, + "eval_runtime": 80.5487, + "eval_samples_per_second": 158.19, + "eval_steps_per_second": 1.241, + "eval_wer": 19.322534000312647, + "step": 32000 + }, + { + "epoch": 320.25, + "learning_rate": 7.5816455696202526e-06, + "loss": 0.0, + "step": 32025 + }, + { + "epoch": 320.5, + "learning_rate": 7.557911392405062e-06, + "loss": 0.0, + "step": 32050 + }, + { + "epoch": 320.75, + "learning_rate": 7.5341772151898724e-06, + "loss": 0.0, + "step": 32075 + }, + { + "epoch": 321.0, + "learning_rate": 7.510443037974683e-06, + "loss": 0.0, + "step": 32100 + }, + { + "epoch": 321.25, + "learning_rate": 7.486708860759493e-06, + "loss": 0.0, + "step": 32125 + }, + { + "epoch": 321.5, + "learning_rate": 7.4629746835443035e-06, + "loss": 0.0, + "step": 32150 + }, + { + "epoch": 321.75, + "learning_rate": 7.439240506329114e-06, + "loss": 0.0, + "step": 32175 + }, + { + "epoch": 322.0, + "learning_rate": 7.415506329113923e-06, + "loss": 0.0, + "step": 32200 + }, + { + "epoch": 322.25, + "learning_rate": 7.391772151898734e-06, + "loss": 0.0, + "step": 32225 + }, + { + "epoch": 322.5, + "learning_rate": 7.368037974683543e-06, + "loss": 0.0, + "step": 32250 + }, + { + "epoch": 322.75, + "learning_rate": 7.3443037974683536e-06, + "loss": 0.0, + "step": 32275 + }, + { + "epoch": 323.0, + "learning_rate": 7.320569620253164e-06, + "loss": 0.0, + "step": 32300 + }, + { + "epoch": 323.25, + "learning_rate": 7.296835443037974e-06, + "loss": 0.0, + "step": 32325 + }, + { + "epoch": 323.5, + "learning_rate": 7.273101265822785e-06, + "loss": 0.0, + "step": 32350 + }, + { + "epoch": 323.75, + "learning_rate": 7.249367088607594e-06, + "loss": 0.0, + "step": 32375 + }, + { + "epoch": 324.0, + "learning_rate": 7.2256329113924045e-06, + "loss": 0.0, + "step": 32400 + }, + { + "epoch": 324.25, + "learning_rate": 7.201898734177215e-06, + "loss": 0.0, + "step": 32425 + }, + { + "epoch": 324.5, + "learning_rate": 7.178164556962024e-06, + "loss": 0.0, + "step": 32450 + }, + { + "epoch": 324.75, + "learning_rate": 7.154430379746835e-06, + "loss": 0.0, + "step": 32475 + }, + { + "epoch": 325.0, + "learning_rate": 7.130696202531644e-06, + "loss": 0.0, + "step": 32500 + }, + { + "epoch": 325.25, + "learning_rate": 7.1069620253164546e-06, + "loss": 0.0, + "step": 32525 + }, + { + "epoch": 325.5, + "learning_rate": 7.083227848101266e-06, + "loss": 0.0, + "step": 32550 + }, + { + "epoch": 325.75, + "learning_rate": 7.059493670886075e-06, + "loss": 0.0, + "step": 32575 + }, + { + "epoch": 326.0, + "learning_rate": 7.035759493670886e-06, + "loss": 0.0, + "step": 32600 + }, + { + "epoch": 326.25, + "learning_rate": 7.012025316455696e-06, + "loss": 0.0, + "step": 32625 + }, + { + "epoch": 326.5, + "learning_rate": 6.9882911392405055e-06, + "loss": 0.0, + "step": 32650 + }, + { + "epoch": 326.75, + "learning_rate": 6.964556962025316e-06, + "loss": 0.0, + "step": 32675 + }, + { + "epoch": 327.0, + "learning_rate": 6.940822784810125e-06, + "loss": 0.0, + "step": 32700 + }, + { + "epoch": 327.25, + "learning_rate": 6.917088607594936e-06, + "loss": 0.0, + "step": 32725 + }, + { + "epoch": 327.5, + "learning_rate": 6.893354430379747e-06, + "loss": 0.0, + "step": 32750 + }, + { + "epoch": 327.75, + "learning_rate": 6.869620253164556e-06, + "loss": 0.0, + "step": 32775 + }, + { + "epoch": 328.0, + "learning_rate": 6.845886075949367e-06, + "loss": 0.0, + "step": 32800 + }, + { + "epoch": 328.25, + "learning_rate": 6.822151898734176e-06, + "loss": 0.0, + "step": 32825 + }, + { + "epoch": 328.5, + "learning_rate": 6.798417721518987e-06, + "loss": 0.0, + "step": 32850 + }, + { + "epoch": 328.75, + "learning_rate": 6.774683544303797e-06, + "loss": 0.0, + "step": 32875 + }, + { + "epoch": 329.0, + "learning_rate": 6.7509493670886065e-06, + "loss": 0.0, + "step": 32900 + }, + { + "epoch": 329.25, + "learning_rate": 6.727215189873417e-06, + "loss": 0.0, + "step": 32925 + }, + { + "epoch": 329.5, + "learning_rate": 6.703481012658228e-06, + "loss": 0.0, + "step": 32950 + }, + { + "epoch": 329.75, + "learning_rate": 6.6797468354430375e-06, + "loss": 0.0, + "step": 32975 + }, + { + "epoch": 330.0, + "learning_rate": 6.656012658227848e-06, + "loss": 0.0, + "step": 33000 + }, + { + "epoch": 330.0, + "eval_loss": 0.5798343420028687, + "eval_runtime": 80.7786, + "eval_samples_per_second": 157.74, + "eval_steps_per_second": 1.238, + "eval_wer": 19.288338283570425, + "step": 33000 + }, + { + "epoch": 330.25, + "learning_rate": 6.632278481012657e-06, + "loss": 0.0, + "step": 33025 + }, + { + "epoch": 330.5, + "learning_rate": 6.608544303797468e-06, + "loss": 0.0, + "step": 33050 + }, + { + "epoch": 330.75, + "learning_rate": 6.584810126582278e-06, + "loss": 0.0, + "step": 33075 + }, + { + "epoch": 331.0, + "learning_rate": 6.561075949367088e-06, + "loss": 0.0, + "step": 33100 + }, + { + "epoch": 331.25, + "learning_rate": 6.537341772151898e-06, + "loss": 0.0, + "step": 33125 + }, + { + "epoch": 331.5, + "learning_rate": 6.513607594936709e-06, + "loss": 0.0, + "step": 33150 + }, + { + "epoch": 331.75, + "learning_rate": 6.489873417721519e-06, + "loss": 0.0, + "step": 33175 + }, + { + "epoch": 332.0, + "learning_rate": 6.466139240506329e-06, + "loss": 0.0, + "step": 33200 + }, + { + "epoch": 332.25, + "learning_rate": 6.4424050632911385e-06, + "loss": 0.0, + "step": 33225 + }, + { + "epoch": 332.5, + "learning_rate": 6.418670886075949e-06, + "loss": 0.0, + "step": 33250 + }, + { + "epoch": 332.75, + "learning_rate": 6.394936708860758e-06, + "loss": 0.0, + "step": 33275 + }, + { + "epoch": 333.0, + "learning_rate": 6.371202531645569e-06, + "loss": 0.0, + "step": 33300 + }, + { + "epoch": 333.25, + "learning_rate": 6.347468354430379e-06, + "loss": 0.0, + "step": 33325 + }, + { + "epoch": 333.5, + "learning_rate": 6.3237341772151895e-06, + "loss": 0.0, + "step": 33350 + }, + { + "epoch": 333.75, + "learning_rate": 6.3e-06, + "loss": 0.0, + "step": 33375 + }, + { + "epoch": 334.0, + "learning_rate": 6.27626582278481e-06, + "loss": 0.0, + "step": 33400 + }, + { + "epoch": 334.25, + "learning_rate": 6.25253164556962e-06, + "loss": 0.0, + "step": 33425 + }, + { + "epoch": 334.5, + "learning_rate": 6.22879746835443e-06, + "loss": 0.0, + "step": 33450 + }, + { + "epoch": 334.75, + "learning_rate": 6.2050632911392395e-06, + "loss": 0.0, + "step": 33475 + }, + { + "epoch": 335.0, + "learning_rate": 6.18132911392405e-06, + "loss": 0.0, + "step": 33500 + }, + { + "epoch": 335.25, + "learning_rate": 6.15759493670886e-06, + "loss": 0.0, + "step": 33525 + }, + { + "epoch": 335.5, + "learning_rate": 6.13386075949367e-06, + "loss": 0.0, + "step": 33550 + }, + { + "epoch": 335.75, + "learning_rate": 6.110126582278481e-06, + "loss": 0.0, + "step": 33575 + }, + { + "epoch": 336.0, + "learning_rate": 6.086392405063291e-06, + "loss": 0.0, + "step": 33600 + }, + { + "epoch": 336.25, + "learning_rate": 6.062658227848101e-06, + "loss": 0.0, + "step": 33625 + }, + { + "epoch": 336.5, + "learning_rate": 6.038924050632911e-06, + "loss": 0.0, + "step": 33650 + }, + { + "epoch": 336.75, + "learning_rate": 6.015189873417721e-06, + "loss": 0.0, + "step": 33675 + }, + { + "epoch": 337.0, + "learning_rate": 5.991455696202531e-06, + "loss": 0.0, + "step": 33700 + }, + { + "epoch": 337.25, + "learning_rate": 5.9677215189873405e-06, + "loss": 0.0, + "step": 33725 + }, + { + "epoch": 337.5, + "learning_rate": 5.943987341772151e-06, + "loss": 0.0, + "step": 33750 + }, + { + "epoch": 337.75, + "learning_rate": 5.920253164556962e-06, + "loss": 0.0, + "step": 33775 + }, + { + "epoch": 338.0, + "learning_rate": 5.896518987341772e-06, + "loss": 0.0, + "step": 33800 + }, + { + "epoch": 338.25, + "learning_rate": 5.872784810126582e-06, + "loss": 0.0, + "step": 33825 + }, + { + "epoch": 338.5, + "learning_rate": 5.849050632911392e-06, + "loss": 0.0, + "step": 33850 + }, + { + "epoch": 338.75, + "learning_rate": 5.825316455696202e-06, + "loss": 0.0, + "step": 33875 + }, + { + "epoch": 339.0, + "learning_rate": 5.801582278481012e-06, + "loss": 0.0, + "step": 33900 + }, + { + "epoch": 339.25, + "learning_rate": 5.777848101265822e-06, + "loss": 0.0, + "step": 33925 + }, + { + "epoch": 339.5, + "learning_rate": 5.754113924050632e-06, + "loss": 0.0, + "step": 33950 + }, + { + "epoch": 339.75, + "learning_rate": 5.730379746835443e-06, + "loss": 0.0, + "step": 33975 + }, + { + "epoch": 340.0, + "learning_rate": 5.706645569620253e-06, + "loss": 0.0, + "step": 34000 + }, + { + "epoch": 340.0, + "eval_loss": 0.5856410264968872, + "eval_runtime": 80.4348, + "eval_samples_per_second": 158.414, + "eval_steps_per_second": 1.243, + "eval_wer": 19.26684383304674, + "step": 34000 + }, + { + "epoch": 340.25, + "learning_rate": 5.682911392405063e-06, + "loss": 0.0, + "step": 34025 + }, + { + "epoch": 340.5, + "learning_rate": 5.659177215189873e-06, + "loss": 0.0, + "step": 34050 + }, + { + "epoch": 340.75, + "learning_rate": 5.635443037974683e-06, + "loss": 0.0, + "step": 34075 + }, + { + "epoch": 341.0, + "learning_rate": 5.611708860759493e-06, + "loss": 0.0, + "step": 34100 + }, + { + "epoch": 341.25, + "learning_rate": 5.587974683544303e-06, + "loss": 0.0, + "step": 34125 + }, + { + "epoch": 341.5, + "learning_rate": 5.564240506329113e-06, + "loss": 0.0, + "step": 34150 + }, + { + "epoch": 341.75, + "learning_rate": 5.540506329113924e-06, + "loss": 0.0, + "step": 34175 + }, + { + "epoch": 342.0, + "learning_rate": 5.516772151898734e-06, + "loss": 0.0, + "step": 34200 + }, + { + "epoch": 342.25, + "learning_rate": 5.493037974683544e-06, + "loss": 0.0, + "step": 34225 + }, + { + "epoch": 342.5, + "learning_rate": 5.469303797468354e-06, + "loss": 0.0, + "step": 34250 + }, + { + "epoch": 342.75, + "learning_rate": 5.445569620253164e-06, + "loss": 0.0, + "step": 34275 + }, + { + "epoch": 343.0, + "learning_rate": 5.4218354430379744e-06, + "loss": 0.0, + "step": 34300 + }, + { + "epoch": 343.25, + "learning_rate": 5.398101265822784e-06, + "loss": 0.0, + "step": 34325 + }, + { + "epoch": 343.5, + "learning_rate": 5.374367088607594e-06, + "loss": 0.0, + "step": 34350 + }, + { + "epoch": 343.75, + "learning_rate": 5.350632911392404e-06, + "loss": 0.0, + "step": 34375 + }, + { + "epoch": 344.0, + "learning_rate": 5.326898734177215e-06, + "loss": 0.0, + "step": 34400 + }, + { + "epoch": 344.25, + "learning_rate": 5.303164556962025e-06, + "loss": 0.0, + "step": 34425 + }, + { + "epoch": 344.5, + "learning_rate": 5.279430379746835e-06, + "loss": 0.0, + "step": 34450 + }, + { + "epoch": 344.75, + "learning_rate": 5.255696202531645e-06, + "loss": 0.0, + "step": 34475 + }, + { + "epoch": 345.0, + "learning_rate": 5.231962025316455e-06, + "loss": 0.0, + "step": 34500 + }, + { + "epoch": 345.25, + "learning_rate": 5.208227848101265e-06, + "loss": 0.0, + "step": 34525 + }, + { + "epoch": 345.5, + "learning_rate": 5.1844936708860754e-06, + "loss": 0.0, + "step": 34550 + }, + { + "epoch": 345.75, + "learning_rate": 5.160759493670885e-06, + "loss": 0.0, + "step": 34575 + }, + { + "epoch": 346.0, + "learning_rate": 5.137025316455696e-06, + "loss": 0.0, + "step": 34600 + }, + { + "epoch": 346.25, + "learning_rate": 5.1132911392405065e-06, + "loss": 0.0, + "step": 34625 + }, + { + "epoch": 346.5, + "learning_rate": 5.089556962025316e-06, + "loss": 0.0, + "step": 34650 + }, + { + "epoch": 346.75, + "learning_rate": 5.065822784810126e-06, + "loss": 0.0, + "step": 34675 + }, + { + "epoch": 347.0, + "learning_rate": 5.042088607594936e-06, + "loss": 0.0, + "step": 34700 + }, + { + "epoch": 347.25, + "learning_rate": 5.018354430379746e-06, + "loss": 0.0, + "step": 34725 + }, + { + "epoch": 347.5, + "learning_rate": 4.9946202531645566e-06, + "loss": 0.0, + "step": 34750 + }, + { + "epoch": 347.75, + "learning_rate": 4.970886075949366e-06, + "loss": 0.0, + "step": 34775 + }, + { + "epoch": 348.0, + "learning_rate": 4.947151898734177e-06, + "loss": 0.0, + "step": 34800 + }, + { + "epoch": 348.25, + "learning_rate": 4.923417721518988e-06, + "loss": 0.0, + "step": 34825 + }, + { + "epoch": 348.5, + "learning_rate": 4.899683544303797e-06, + "loss": 0.0, + "step": 34850 + }, + { + "epoch": 348.75, + "learning_rate": 4.8759493670886075e-06, + "loss": 0.0, + "step": 34875 + }, + { + "epoch": 349.0, + "learning_rate": 4.852215189873417e-06, + "loss": 0.0, + "step": 34900 + }, + { + "epoch": 349.25, + "learning_rate": 4.828481012658227e-06, + "loss": 0.0, + "step": 34925 + }, + { + "epoch": 349.5, + "learning_rate": 4.804746835443037e-06, + "loss": 0.0, + "step": 34950 + }, + { + "epoch": 349.75, + "learning_rate": 4.781012658227847e-06, + "loss": 0.0, + "step": 34975 + }, + { + "epoch": 350.0, + "learning_rate": 4.757278481012658e-06, + "loss": 0.0, + "step": 35000 + }, + { + "epoch": 350.0, + "eval_loss": 0.5910990238189697, + "eval_runtime": 80.5464, + "eval_samples_per_second": 158.195, + "eval_steps_per_second": 1.242, + "eval_wer": 19.234602157261214, + "step": 35000 + }, + { + "epoch": 350.25, + "learning_rate": 4.733544303797468e-06, + "loss": 0.0, + "step": 35025 + }, + { + "epoch": 350.5, + "learning_rate": 4.709810126582278e-06, + "loss": 0.0, + "step": 35050 + }, + { + "epoch": 350.75, + "learning_rate": 4.686075949367089e-06, + "loss": 0.0, + "step": 35075 + }, + { + "epoch": 351.0, + "learning_rate": 4.662341772151898e-06, + "loss": 0.0, + "step": 35100 + }, + { + "epoch": 351.25, + "learning_rate": 4.6386075949367085e-06, + "loss": 0.0, + "step": 35125 + }, + { + "epoch": 351.5, + "learning_rate": 4.614873417721519e-06, + "loss": 0.0, + "step": 35150 + }, + { + "epoch": 351.75, + "learning_rate": 4.591139240506329e-06, + "loss": 0.0, + "step": 35175 + }, + { + "epoch": 352.0, + "learning_rate": 4.567405063291139e-06, + "loss": 0.0, + "step": 35200 + }, + { + "epoch": 352.25, + "learning_rate": 4.543670886075949e-06, + "loss": 0.0, + "step": 35225 + }, + { + "epoch": 352.5, + "learning_rate": 4.5199367088607586e-06, + "loss": 0.0, + "step": 35250 + }, + { + "epoch": 352.75, + "learning_rate": 4.49620253164557e-06, + "loss": 0.0, + "step": 35275 + }, + { + "epoch": 353.0, + "learning_rate": 4.472468354430379e-06, + "loss": 0.0, + "step": 35300 + }, + { + "epoch": 353.25, + "learning_rate": 4.44873417721519e-06, + "loss": 0.0, + "step": 35325 + }, + { + "epoch": 353.5, + "learning_rate": 4.424999999999999e-06, + "loss": 0.0, + "step": 35350 + }, + { + "epoch": 353.75, + "learning_rate": 4.4012658227848095e-06, + "loss": 0.0, + "step": 35375 + }, + { + "epoch": 354.0, + "learning_rate": 4.37753164556962e-06, + "loss": 0.0, + "step": 35400 + }, + { + "epoch": 354.25, + "learning_rate": 4.35379746835443e-06, + "loss": 0.0, + "step": 35425 + }, + { + "epoch": 354.5, + "learning_rate": 4.33006329113924e-06, + "loss": 0.0, + "step": 35450 + }, + { + "epoch": 354.75, + "learning_rate": 4.30632911392405e-06, + "loss": 0.0, + "step": 35475 + }, + { + "epoch": 355.0, + "learning_rate": 4.28259493670886e-06, + "loss": 0.0, + "step": 35500 + }, + { + "epoch": 355.25, + "learning_rate": 4.258860759493671e-06, + "loss": 0.0, + "step": 35525 + }, + { + "epoch": 355.5, + "learning_rate": 4.23512658227848e-06, + "loss": 0.0, + "step": 35550 + }, + { + "epoch": 355.75, + "learning_rate": 4.211392405063291e-06, + "loss": 0.0, + "step": 35575 + }, + { + "epoch": 356.0, + "learning_rate": 4.187658227848101e-06, + "loss": 0.0, + "step": 35600 + }, + { + "epoch": 356.25, + "learning_rate": 4.163924050632911e-06, + "loss": 0.0, + "step": 35625 + }, + { + "epoch": 356.5, + "learning_rate": 4.140189873417721e-06, + "loss": 0.0, + "step": 35650 + }, + { + "epoch": 356.75, + "learning_rate": 4.116455696202531e-06, + "loss": 0.0, + "step": 35675 + }, + { + "epoch": 357.0, + "learning_rate": 4.0927215189873416e-06, + "loss": 0.0, + "step": 35700 + }, + { + "epoch": 357.25, + "learning_rate": 4.068987341772152e-06, + "loss": 0.0, + "step": 35725 + }, + { + "epoch": 357.5, + "learning_rate": 4.045253164556961e-06, + "loss": 0.0, + "step": 35750 + }, + { + "epoch": 357.75, + "learning_rate": 4.021518987341772e-06, + "loss": 0.0, + "step": 35775 + }, + { + "epoch": 358.0, + "learning_rate": 3.997784810126582e-06, + "loss": 0.0, + "step": 35800 + }, + { + "epoch": 358.25, + "learning_rate": 3.974050632911392e-06, + "loss": 0.0, + "step": 35825 + }, + { + "epoch": 358.5, + "learning_rate": 3.950316455696202e-06, + "loss": 0.0, + "step": 35850 + }, + { + "epoch": 358.75, + "learning_rate": 3.926582278481012e-06, + "loss": 0.0, + "step": 35875 + }, + { + "epoch": 359.0, + "learning_rate": 3.902848101265823e-06, + "loss": 0.0, + "step": 35900 + }, + { + "epoch": 359.25, + "learning_rate": 3.879113924050632e-06, + "loss": 0.0, + "step": 35925 + }, + { + "epoch": 359.5, + "learning_rate": 3.8553797468354425e-06, + "loss": 0.0, + "step": 35950 + }, + { + "epoch": 359.75, + "learning_rate": 3.831645569620253e-06, + "loss": 0.0, + "step": 35975 + }, + { + "epoch": 360.0, + "learning_rate": 3.8088607594936704e-06, + "loss": 0.0, + "step": 36000 + }, + { + "epoch": 360.0, + "eval_loss": 0.5961853861808777, + "eval_runtime": 80.7236, + "eval_samples_per_second": 157.847, + "eval_steps_per_second": 1.239, + "eval_wer": 19.22874003439112, + "step": 36000 + }, + { + "epoch": 360.25, + "learning_rate": 3.7851265822784808e-06, + "loss": 0.0, + "step": 36025 + }, + { + "epoch": 360.5, + "learning_rate": 3.7613924050632907e-06, + "loss": 0.0, + "step": 36050 + }, + { + "epoch": 360.75, + "learning_rate": 3.737658227848101e-06, + "loss": 0.0, + "step": 36075 + }, + { + "epoch": 361.0, + "learning_rate": 3.713924050632911e-06, + "loss": 0.0, + "step": 36100 + }, + { + "epoch": 361.25, + "learning_rate": 3.690189873417721e-06, + "loss": 0.0, + "step": 36125 + }, + { + "epoch": 361.5, + "learning_rate": 3.6664556962025313e-06, + "loss": 0.0, + "step": 36150 + }, + { + "epoch": 361.75, + "learning_rate": 3.6427215189873416e-06, + "loss": 0.0, + "step": 36175 + }, + { + "epoch": 362.0, + "learning_rate": 3.6189873417721516e-06, + "loss": 0.0, + "step": 36200 + }, + { + "epoch": 362.25, + "learning_rate": 3.5952531645569615e-06, + "loss": 0.0, + "step": 36225 + }, + { + "epoch": 362.5, + "learning_rate": 3.571518987341772e-06, + "loss": 0.0, + "step": 36250 + }, + { + "epoch": 362.75, + "learning_rate": 3.547784810126582e-06, + "loss": 0.0, + "step": 36275 + }, + { + "epoch": 363.0, + "learning_rate": 3.524050632911392e-06, + "loss": 0.0, + "step": 36300 + }, + { + "epoch": 363.25, + "learning_rate": 3.500316455696202e-06, + "loss": 0.0, + "step": 36325 + }, + { + "epoch": 363.5, + "learning_rate": 3.476582278481012e-06, + "loss": 0.0, + "step": 36350 + }, + { + "epoch": 363.75, + "learning_rate": 3.4528481012658228e-06, + "loss": 0.0, + "step": 36375 + }, + { + "epoch": 364.0, + "learning_rate": 3.4291139240506327e-06, + "loss": 0.0, + "step": 36400 + }, + { + "epoch": 364.25, + "learning_rate": 3.4053797468354426e-06, + "loss": 0.0, + "step": 36425 + }, + { + "epoch": 364.5, + "learning_rate": 3.3816455696202526e-06, + "loss": 0.0, + "step": 36450 + }, + { + "epoch": 364.75, + "learning_rate": 3.3579113924050633e-06, + "loss": 0.0, + "step": 36475 + }, + { + "epoch": 365.0, + "learning_rate": 3.3341772151898733e-06, + "loss": 0.0, + "step": 36500 + }, + { + "epoch": 365.25, + "learning_rate": 3.310443037974683e-06, + "loss": 0.0, + "step": 36525 + }, + { + "epoch": 365.5, + "learning_rate": 3.286708860759493e-06, + "loss": 0.0, + "step": 36550 + }, + { + "epoch": 365.75, + "learning_rate": 3.262974683544304e-06, + "loss": 0.0, + "step": 36575 + }, + { + "epoch": 366.0, + "learning_rate": 3.239240506329114e-06, + "loss": 0.0, + "step": 36600 + }, + { + "epoch": 366.25, + "learning_rate": 3.2155063291139238e-06, + "loss": 0.0, + "step": 36625 + }, + { + "epoch": 366.5, + "learning_rate": 3.1917721518987337e-06, + "loss": 0.0, + "step": 36650 + }, + { + "epoch": 366.75, + "learning_rate": 3.1680379746835436e-06, + "loss": 0.0, + "step": 36675 + }, + { + "epoch": 367.0, + "learning_rate": 3.1443037974683544e-06, + "loss": 0.0, + "step": 36700 + }, + { + "epoch": 367.25, + "learning_rate": 3.1205696202531643e-06, + "loss": 0.0, + "step": 36725 + }, + { + "epoch": 367.5, + "learning_rate": 3.0968354430379743e-06, + "loss": 0.0, + "step": 36750 + }, + { + "epoch": 367.75, + "learning_rate": 3.073101265822784e-06, + "loss": 0.0, + "step": 36775 + }, + { + "epoch": 368.0, + "learning_rate": 3.049367088607595e-06, + "loss": 0.0, + "step": 36800 + }, + { + "epoch": 368.25, + "learning_rate": 3.025632911392405e-06, + "loss": 0.0, + "step": 36825 + }, + { + "epoch": 368.5, + "learning_rate": 3.001898734177215e-06, + "loss": 0.0, + "step": 36850 + }, + { + "epoch": 368.75, + "learning_rate": 2.9781645569620248e-06, + "loss": 0.0, + "step": 36875 + }, + { + "epoch": 369.0, + "learning_rate": 2.9544303797468355e-06, + "loss": 0.0, + "step": 36900 + }, + { + "epoch": 369.25, + "learning_rate": 2.9306962025316455e-06, + "loss": 0.0, + "step": 36925 + }, + { + "epoch": 369.5, + "learning_rate": 2.9069620253164554e-06, + "loss": 0.0, + "step": 36950 + }, + { + "epoch": 369.75, + "learning_rate": 2.8832278481012653e-06, + "loss": 0.0, + "step": 36975 + }, + { + "epoch": 370.0, + "learning_rate": 2.8594936708860757e-06, + "loss": 0.0, + "step": 37000 + }, + { + "epoch": 370.0, + "eval_loss": 0.6009894013404846, + "eval_runtime": 80.466, + "eval_samples_per_second": 158.353, + "eval_steps_per_second": 1.243, + "eval_wer": 19.232648116304517, + "step": 37000 + }, + { + "epoch": 370.25, + "learning_rate": 2.835759493670886e-06, + "loss": 0.0, + "step": 37025 + }, + { + "epoch": 370.5, + "learning_rate": 2.812025316455696e-06, + "loss": 0.0, + "step": 37050 + }, + { + "epoch": 370.75, + "learning_rate": 2.788291139240506e-06, + "loss": 0.0, + "step": 37075 + }, + { + "epoch": 371.0, + "learning_rate": 2.7645569620253162e-06, + "loss": 0.0, + "step": 37100 + }, + { + "epoch": 371.25, + "learning_rate": 2.7408227848101266e-06, + "loss": 0.0, + "step": 37125 + }, + { + "epoch": 371.5, + "learning_rate": 2.7170886075949365e-06, + "loss": 0.0, + "step": 37150 + }, + { + "epoch": 371.75, + "learning_rate": 2.6933544303797465e-06, + "loss": 0.0, + "step": 37175 + }, + { + "epoch": 372.0, + "learning_rate": 2.669620253164557e-06, + "loss": 0.0, + "step": 37200 + }, + { + "epoch": 372.25, + "learning_rate": 2.6458860759493667e-06, + "loss": 0.0, + "step": 37225 + }, + { + "epoch": 372.5, + "learning_rate": 2.622151898734177e-06, + "loss": 0.0, + "step": 37250 + }, + { + "epoch": 372.75, + "learning_rate": 2.598417721518987e-06, + "loss": 0.0, + "step": 37275 + }, + { + "epoch": 373.0, + "learning_rate": 2.5746835443037974e-06, + "loss": 0.0, + "step": 37300 + }, + { + "epoch": 373.25, + "learning_rate": 2.5509493670886073e-06, + "loss": 0.0, + "step": 37325 + }, + { + "epoch": 373.5, + "learning_rate": 2.5272151898734177e-06, + "loss": 0.0, + "step": 37350 + }, + { + "epoch": 373.75, + "learning_rate": 2.5034810126582276e-06, + "loss": 0.0, + "step": 37375 + }, + { + "epoch": 374.0, + "learning_rate": 2.479746835443038e-06, + "loss": 0.0, + "step": 37400 + }, + { + "epoch": 374.25, + "learning_rate": 2.456012658227848e-06, + "loss": 0.0, + "step": 37425 + }, + { + "epoch": 374.5, + "learning_rate": 2.432278481012658e-06, + "loss": 0.0, + "step": 37450 + }, + { + "epoch": 374.75, + "learning_rate": 2.408544303797468e-06, + "loss": 0.0, + "step": 37475 + }, + { + "epoch": 375.0, + "learning_rate": 2.384810126582278e-06, + "loss": 0.0, + "step": 37500 + }, + { + "epoch": 375.25, + "learning_rate": 2.3610759493670884e-06, + "loss": 0.0, + "step": 37525 + }, + { + "epoch": 375.5, + "learning_rate": 2.3373417721518984e-06, + "loss": 0.0, + "step": 37550 + }, + { + "epoch": 375.75, + "learning_rate": 2.3136075949367087e-06, + "loss": 0.0, + "step": 37575 + }, + { + "epoch": 376.0, + "learning_rate": 2.2898734177215187e-06, + "loss": 0.0, + "step": 37600 + }, + { + "epoch": 376.25, + "learning_rate": 2.2661392405063286e-06, + "loss": 0.0, + "step": 37625 + }, + { + "epoch": 376.5, + "learning_rate": 2.242405063291139e-06, + "loss": 0.0, + "step": 37650 + }, + { + "epoch": 376.75, + "learning_rate": 2.218670886075949e-06, + "loss": 0.0, + "step": 37675 + }, + { + "epoch": 377.0, + "learning_rate": 2.1949367088607592e-06, + "loss": 0.0, + "step": 37700 + }, + { + "epoch": 377.25, + "learning_rate": 2.171202531645569e-06, + "loss": 0.0, + "step": 37725 + }, + { + "epoch": 377.5, + "learning_rate": 2.1474683544303795e-06, + "loss": 0.0, + "step": 37750 + }, + { + "epoch": 377.75, + "learning_rate": 2.1237341772151894e-06, + "loss": 0.0, + "step": 37775 + }, + { + "epoch": 378.0, + "learning_rate": 2.1e-06, + "loss": 0.0, + "step": 37800 + }, + { + "epoch": 378.25, + "learning_rate": 2.0762658227848097e-06, + "loss": 0.0, + "step": 37825 + }, + { + "epoch": 378.5, + "learning_rate": 2.05253164556962e-06, + "loss": 0.0, + "step": 37850 + }, + { + "epoch": 378.75, + "learning_rate": 2.02879746835443e-06, + "loss": 0.0, + "step": 37875 + }, + { + "epoch": 379.0, + "learning_rate": 2.0050632911392404e-06, + "loss": 0.0, + "step": 37900 + }, + { + "epoch": 379.25, + "learning_rate": 1.9813291139240503e-06, + "loss": 0.0, + "step": 37925 + }, + { + "epoch": 379.5, + "learning_rate": 1.9575949367088607e-06, + "loss": 0.0, + "step": 37950 + }, + { + "epoch": 379.75, + "learning_rate": 1.9338607594936706e-06, + "loss": 0.0, + "step": 37975 + }, + { + "epoch": 380.0, + "learning_rate": 1.910126582278481e-06, + "loss": 0.0, + "step": 38000 + }, + { + "epoch": 380.0, + "eval_loss": 0.6050010919570923, + "eval_runtime": 80.5317, + "eval_samples_per_second": 158.223, + "eval_steps_per_second": 1.242, + "eval_wer": 19.22874003439112, + "step": 38000 + }, + { + "epoch": 380.25, + "learning_rate": 1.8863924050632909e-06, + "loss": 0.0, + "step": 38025 + }, + { + "epoch": 380.5, + "learning_rate": 1.8626582278481012e-06, + "loss": 0.0, + "step": 38050 + }, + { + "epoch": 380.75, + "learning_rate": 1.8389240506329112e-06, + "loss": 0.0, + "step": 38075 + }, + { + "epoch": 381.0, + "learning_rate": 1.8151898734177215e-06, + "loss": 0.0, + "step": 38100 + }, + { + "epoch": 381.25, + "learning_rate": 1.7914556962025314e-06, + "loss": 0.0, + "step": 38125 + }, + { + "epoch": 381.5, + "learning_rate": 1.7677215189873418e-06, + "loss": 0.0, + "step": 38150 + }, + { + "epoch": 381.75, + "learning_rate": 1.7439873417721517e-06, + "loss": 0.0, + "step": 38175 + }, + { + "epoch": 382.0, + "learning_rate": 1.720253164556962e-06, + "loss": 0.0, + "step": 38200 + }, + { + "epoch": 382.25, + "learning_rate": 1.696518987341772e-06, + "loss": 0.0, + "step": 38225 + }, + { + "epoch": 382.5, + "learning_rate": 1.6727848101265821e-06, + "loss": 0.0, + "step": 38250 + }, + { + "epoch": 382.75, + "learning_rate": 1.6490506329113923e-06, + "loss": 0.0, + "step": 38275 + }, + { + "epoch": 383.0, + "learning_rate": 1.6253164556962024e-06, + "loss": 0.0, + "step": 38300 + }, + { + "epoch": 383.25, + "learning_rate": 1.6015822784810126e-06, + "loss": 0.0, + "step": 38325 + }, + { + "epoch": 383.5, + "learning_rate": 1.5778481012658227e-06, + "loss": 0.0, + "step": 38350 + }, + { + "epoch": 383.75, + "learning_rate": 1.5541139240506329e-06, + "loss": 0.0, + "step": 38375 + }, + { + "epoch": 384.0, + "learning_rate": 1.5303797468354428e-06, + "loss": 0.0, + "step": 38400 + }, + { + "epoch": 384.25, + "learning_rate": 1.5066455696202531e-06, + "loss": 0.0, + "step": 38425 + }, + { + "epoch": 384.5, + "learning_rate": 1.482911392405063e-06, + "loss": 0.0, + "step": 38450 + }, + { + "epoch": 384.75, + "learning_rate": 1.4591772151898732e-06, + "loss": 0.0, + "step": 38475 + }, + { + "epoch": 385.0, + "learning_rate": 1.4354430379746834e-06, + "loss": 0.0, + "step": 38500 + }, + { + "epoch": 385.25, + "learning_rate": 1.4117088607594935e-06, + "loss": 0.0, + "step": 38525 + }, + { + "epoch": 385.5, + "learning_rate": 1.3879746835443036e-06, + "loss": 0.0, + "step": 38550 + }, + { + "epoch": 385.75, + "learning_rate": 1.3642405063291138e-06, + "loss": 0.0, + "step": 38575 + }, + { + "epoch": 386.0, + "learning_rate": 1.340506329113924e-06, + "loss": 0.0, + "step": 38600 + }, + { + "epoch": 386.25, + "learning_rate": 1.316772151898734e-06, + "loss": 0.0, + "step": 38625 + }, + { + "epoch": 386.5, + "learning_rate": 1.293037974683544e-06, + "loss": 0.0, + "step": 38650 + }, + { + "epoch": 386.75, + "learning_rate": 1.2693037974683543e-06, + "loss": 0.0, + "step": 38675 + }, + { + "epoch": 387.0, + "learning_rate": 1.2455696202531643e-06, + "loss": 0.0, + "step": 38700 + }, + { + "epoch": 387.25, + "learning_rate": 1.2218354430379746e-06, + "loss": 0.0, + "step": 38725 + }, + { + "epoch": 387.5, + "learning_rate": 1.1981012658227846e-06, + "loss": 0.0, + "step": 38750 + }, + { + "epoch": 387.75, + "learning_rate": 1.174367088607595e-06, + "loss": 0.0, + "step": 38775 + }, + { + "epoch": 388.0, + "learning_rate": 1.150632911392405e-06, + "loss": 0.0, + "step": 38800 + }, + { + "epoch": 388.25, + "learning_rate": 1.1268987341772152e-06, + "loss": 0.0, + "step": 38825 + }, + { + "epoch": 388.5, + "learning_rate": 1.1031645569620251e-06, + "loss": 0.0, + "step": 38850 + }, + { + "epoch": 388.75, + "learning_rate": 1.0794303797468353e-06, + "loss": 0.0, + "step": 38875 + }, + { + "epoch": 389.0, + "learning_rate": 1.0556962025316454e-06, + "loss": 0.0, + "step": 38900 + }, + { + "epoch": 389.25, + "learning_rate": 1.0319620253164556e-06, + "loss": 0.0, + "step": 38925 + }, + { + "epoch": 389.5, + "learning_rate": 1.0082278481012657e-06, + "loss": 0.0, + "step": 38950 + }, + { + "epoch": 389.75, + "learning_rate": 9.844936708860758e-07, + "loss": 0.0, + "step": 38975 + }, + { + "epoch": 390.0, + "learning_rate": 9.60759493670886e-07, + "loss": 0.0, + "step": 39000 + }, + { + "epoch": 390.0, + "eval_loss": 0.608101487159729, + "eval_runtime": 80.4312, + "eval_samples_per_second": 158.421, + "eval_steps_per_second": 1.243, + "eval_wer": 19.237533218696264, + "step": 39000 + }, + { + "epoch": 390.25, + "learning_rate": 9.370253164556961e-07, + "loss": 0.0, + "step": 39025 + }, + { + "epoch": 390.5, + "learning_rate": 9.132911392405062e-07, + "loss": 0.0, + "step": 39050 + }, + { + "epoch": 390.75, + "learning_rate": 8.895569620253163e-07, + "loss": 0.0, + "step": 39075 + }, + { + "epoch": 391.0, + "learning_rate": 8.658227848101264e-07, + "loss": 0.0, + "step": 39100 + }, + { + "epoch": 391.25, + "learning_rate": 8.420886075949366e-07, + "loss": 0.0, + "step": 39125 + }, + { + "epoch": 391.5, + "learning_rate": 8.183544303797467e-07, + "loss": 0.0, + "step": 39150 + }, + { + "epoch": 391.75, + "learning_rate": 7.946202531645569e-07, + "loss": 0.0, + "step": 39175 + }, + { + "epoch": 392.0, + "learning_rate": 7.70886075949367e-07, + "loss": 0.0, + "step": 39200 + }, + { + "epoch": 392.25, + "learning_rate": 7.471518987341772e-07, + "loss": 0.0, + "step": 39225 + }, + { + "epoch": 392.5, + "learning_rate": 7.234177215189873e-07, + "loss": 0.0, + "step": 39250 + }, + { + "epoch": 392.75, + "learning_rate": 6.996835443037974e-07, + "loss": 0.0, + "step": 39275 + }, + { + "epoch": 393.0, + "learning_rate": 6.759493670886076e-07, + "loss": 0.0, + "step": 39300 + }, + { + "epoch": 393.25, + "learning_rate": 6.522151898734176e-07, + "loss": 0.0, + "step": 39325 + }, + { + "epoch": 393.5, + "learning_rate": 6.284810126582278e-07, + "loss": 0.0, + "step": 39350 + }, + { + "epoch": 393.75, + "learning_rate": 6.047468354430379e-07, + "loss": 0.0, + "step": 39375 + }, + { + "epoch": 394.0, + "learning_rate": 5.81012658227848e-07, + "loss": 0.0, + "step": 39400 + }, + { + "epoch": 394.25, + "learning_rate": 5.572784810126582e-07, + "loss": 0.0, + "step": 39425 + }, + { + "epoch": 394.5, + "learning_rate": 5.335443037974683e-07, + "loss": 0.0, + "step": 39450 + }, + { + "epoch": 394.75, + "learning_rate": 5.098101265822785e-07, + "loss": 0.0, + "step": 39475 + }, + { + "epoch": 395.0, + "learning_rate": 4.860759493670886e-07, + "loss": 0.0, + "step": 39500 + }, + { + "epoch": 395.25, + "learning_rate": 4.6234177215189865e-07, + "loss": 0.0, + "step": 39525 + }, + { + "epoch": 395.5, + "learning_rate": 4.386075949367088e-07, + "loss": 0.0, + "step": 39550 + }, + { + "epoch": 395.75, + "learning_rate": 4.1487341772151893e-07, + "loss": 0.0, + "step": 39575 + }, + { + "epoch": 396.0, + "learning_rate": 3.911392405063291e-07, + "loss": 0.0, + "step": 39600 + }, + { + "epoch": 396.25, + "learning_rate": 3.674050632911392e-07, + "loss": 0.0, + "step": 39625 + }, + { + "epoch": 396.5, + "learning_rate": 3.436708860759493e-07, + "loss": 0.0, + "step": 39650 + }, + { + "epoch": 396.75, + "learning_rate": 3.1993670886075945e-07, + "loss": 0.0, + "step": 39675 + }, + { + "epoch": 397.0, + "learning_rate": 2.962025316455696e-07, + "loss": 0.0, + "step": 39700 + }, + { + "epoch": 397.25, + "learning_rate": 2.7246835443037973e-07, + "loss": 0.0, + "step": 39725 + }, + { + "epoch": 397.5, + "learning_rate": 2.487341772151899e-07, + "loss": 0.0, + "step": 39750 + }, + { + "epoch": 397.75, + "learning_rate": 2.25e-07, + "loss": 0.0, + "step": 39775 + }, + { + "epoch": 398.0, + "learning_rate": 2.012658227848101e-07, + "loss": 0.0, + "step": 39800 + }, + { + "epoch": 398.25, + "learning_rate": 1.7753164556962025e-07, + "loss": 0.0, + "step": 39825 + }, + { + "epoch": 398.5, + "learning_rate": 1.537974683544304e-07, + "loss": 0.0, + "step": 39850 + }, + { + "epoch": 398.75, + "learning_rate": 1.300632911392405e-07, + "loss": 0.0, + "step": 39875 + }, + { + "epoch": 399.0, + "learning_rate": 1.0632911392405063e-07, + "loss": 0.0, + "step": 39900 + }, + { + "epoch": 399.25, + "learning_rate": 8.259493670886075e-08, + "loss": 0.0, + "step": 39925 + }, + { + "epoch": 399.5, + "learning_rate": 5.886075949367088e-08, + "loss": 0.0, + "step": 39950 + }, + { + "epoch": 399.75, + "learning_rate": 3.5126582278481014e-08, + "loss": 0.0, + "step": 39975 + }, + { + "epoch": 400.0, + "learning_rate": 1.2341772151898733e-08, + "loss": 0.0, + "step": 40000 + }, + { + "epoch": 400.0, + "eval_loss": 0.6094558238983154, + "eval_runtime": 80.3325, + "eval_samples_per_second": 158.616, + "eval_steps_per_second": 1.245, + "eval_wer": 19.196498358605595, + "step": 40000 + }, + { + "epoch": 400.0, + "step": 40000, + "total_flos": 2.50955067359232e+20, + "train_loss": 0.010605198578223871, + "train_runtime": 40076.3871, + "train_samples_per_second": 255.512, + "train_steps_per_second": 0.998 + } + ], + "logging_steps": 25, + "max_steps": 40000, + "num_input_tokens_seen": 0, + "num_train_epochs": 400, + "save_steps": 1000, + "total_flos": 2.50955067359232e+20, + "train_batch_size": 256, + "trial_name": null, + "trial_params": null +}