diff --git "a/v3.0/trainer_state.json" "b/v3.0/trainer_state.json" new file mode 100644--- /dev/null +++ "b/v3.0/trainer_state.json" @@ -0,0 +1,21541 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 25.0, + "global_step": 153725, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.02, + "learning_rate": 2e-05, + "loss": 13.0218, + "step": 100 + }, + { + "epoch": 0.02, + "eval_loss": 12.945786476135254, + "eval_runtime": 7.9731, + "eval_samples_per_second": 12.542, + "eval_wer": 1.0, + "step": 100 + }, + { + "epoch": 0.03, + "learning_rate": 4e-05, + "loss": 4.4749, + "step": 200 + }, + { + "epoch": 0.03, + "eval_loss": 3.8371927738189697, + "eval_runtime": 7.0147, + "eval_samples_per_second": 14.256, + "eval_wer": 1.0, + "step": 200 + }, + { + "epoch": 0.05, + "learning_rate": 6e-05, + "loss": 3.1121, + "step": 300 + }, + { + "epoch": 0.05, + "eval_loss": 3.498441457748413, + "eval_runtime": 8.8314, + "eval_samples_per_second": 11.323, + "eval_wer": 1.0, + "step": 300 + }, + { + "epoch": 0.07, + "learning_rate": 8e-05, + "loss": 3.0216, + "step": 400 + }, + { + "epoch": 0.07, + "eval_loss": 3.28515625, + "eval_runtime": 7.5144, + "eval_samples_per_second": 13.308, + "eval_wer": 1.0, + "step": 400 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001, + "loss": 2.9786, + "step": 500 + }, + { + "epoch": 0.08, + "eval_loss": 3.120776891708374, + "eval_runtime": 7.7151, + "eval_samples_per_second": 12.962, + "eval_wer": 1.0, + "step": 500 + }, + { + "epoch": 0.1, + "learning_rate": 9.993473649861316e-05, + "loss": 2.9376, + "step": 600 + }, + { + "epoch": 0.1, + "eval_loss": 2.971864700317383, + "eval_runtime": 8.4632, + "eval_samples_per_second": 11.816, + "eval_wer": 1.0, + "step": 600 + }, + { + "epoch": 0.11, + "learning_rate": 9.986947299722631e-05, + "loss": 2.0802, + "step": 700 + }, + { + "epoch": 0.11, + "eval_loss": 1.857584834098816, + "eval_runtime": 9.723, + "eval_samples_per_second": 10.285, + "eval_wer": 0.8855054811205847, + "step": 700 + }, + { + "epoch": 0.13, + "learning_rate": 9.980420949583945e-05, + "loss": 1.0262, + "step": 800 + }, + { + "epoch": 0.13, + "eval_loss": 1.1796859502792358, + "eval_runtime": 7.4195, + "eval_samples_per_second": 13.478, + "eval_wer": 0.6942752740560292, + "step": 800 + }, + { + "epoch": 0.15, + "learning_rate": 9.97389459944526e-05, + "loss": 0.7875, + "step": 900 + }, + { + "epoch": 0.15, + "eval_loss": 1.219588041305542, + "eval_runtime": 9.0106, + "eval_samples_per_second": 11.098, + "eval_wer": 0.682095006090134, + "step": 900 + }, + { + "epoch": 0.16, + "learning_rate": 9.967368249306576e-05, + "loss": 0.688, + "step": 1000 + }, + { + "epoch": 0.16, + "eval_loss": 0.8769925832748413, + "eval_runtime": 8.365, + "eval_samples_per_second": 11.955, + "eval_wer": 0.6065773447015834, + "step": 1000 + }, + { + "epoch": 0.18, + "learning_rate": 9.96084189916789e-05, + "loss": 0.6377, + "step": 1100 + }, + { + "epoch": 0.18, + "eval_loss": 0.7903584241867065, + "eval_runtime": 7.3855, + "eval_samples_per_second": 13.54, + "eval_wer": 0.5700365408038977, + "step": 1100 + }, + { + "epoch": 0.2, + "learning_rate": 9.954315549029206e-05, + "loss": 0.5704, + "step": 1200 + }, + { + "epoch": 0.2, + "eval_loss": 0.8232672214508057, + "eval_runtime": 7.6284, + "eval_samples_per_second": 13.109, + "eval_wer": 0.5651644336175395, + "step": 1200 + }, + { + "epoch": 0.21, + "learning_rate": 9.947789198890522e-05, + "loss": 0.5187, + "step": 1300 + }, + { + "epoch": 0.21, + "eval_loss": 0.7398794293403625, + "eval_runtime": 8.3859, + "eval_samples_per_second": 11.925, + "eval_wer": 0.5237515225334958, + "step": 1300 + }, + { + "epoch": 0.23, + "learning_rate": 9.941262848751835e-05, + "loss": Infinity, + "step": 1400 + }, + { + "epoch": 0.23, + "eval_loss": 0.6848248839378357, + "eval_runtime": 8.3309, + "eval_samples_per_second": 12.004, + "eval_wer": 0.5261875761266748, + "step": 1400 + }, + { + "epoch": 0.24, + "learning_rate": 9.934736498613151e-05, + "loss": NaN, + "step": 1500 + }, + { + "epoch": 0.24, + "eval_loss": 0.712200403213501, + "eval_runtime": 8.7637, + "eval_samples_per_second": 11.411, + "eval_wer": 0.5408038976857491, + "step": 1500 + }, + { + "epoch": 0.26, + "learning_rate": 9.928210148474466e-05, + "loss": NaN, + "step": 1600 + }, + { + "epoch": 0.26, + "eval_loss": 0.7104419469833374, + "eval_runtime": 8.8754, + "eval_samples_per_second": 11.267, + "eval_wer": 0.5371498172959805, + "step": 1600 + }, + { + "epoch": 0.28, + "learning_rate": 9.921683798335781e-05, + "loss": NaN, + "step": 1700 + }, + { + "epoch": 0.28, + "eval_loss": 0.6172800660133362, + "eval_runtime": 7.0857, + "eval_samples_per_second": 14.113, + "eval_wer": 0.5018270401948843, + "step": 1700 + }, + { + "epoch": 0.29, + "learning_rate": 9.915157448197096e-05, + "loss": NaN, + "step": 1800 + }, + { + "epoch": 0.29, + "eval_loss": 0.6506848335266113, + "eval_runtime": 7.7525, + "eval_samples_per_second": 12.899, + "eval_wer": 0.5006090133982948, + "step": 1800 + }, + { + "epoch": 0.31, + "learning_rate": 9.908631098058412e-05, + "loss": NaN, + "step": 1900 + }, + { + "epoch": 0.31, + "eval_loss": 0.6602973937988281, + "eval_runtime": 7.4177, + "eval_samples_per_second": 13.481, + "eval_wer": 0.5042630937880633, + "step": 1900 + }, + { + "epoch": 0.33, + "learning_rate": 9.902104747919726e-05, + "loss": NaN, + "step": 2000 + }, + { + "epoch": 0.33, + "eval_loss": 0.7591428160667419, + "eval_runtime": 8.4029, + "eval_samples_per_second": 11.901, + "eval_wer": 0.4981729598051157, + "step": 2000 + }, + { + "epoch": 0.34, + "learning_rate": 9.895578397781041e-05, + "loss": NaN, + "step": 2100 + }, + { + "epoch": 0.34, + "eval_loss": 0.645115077495575, + "eval_runtime": 8.9522, + "eval_samples_per_second": 11.17, + "eval_wer": 0.47990255785627284, + "step": 2100 + }, + { + "epoch": 0.36, + "learning_rate": 9.889052047642357e-05, + "loss": NaN, + "step": 2200 + }, + { + "epoch": 0.36, + "eval_loss": 0.7210723757743835, + "eval_runtime": 8.4131, + "eval_samples_per_second": 11.886, + "eval_wer": 0.5152253349573691, + "step": 2200 + }, + { + "epoch": 0.37, + "learning_rate": 9.882525697503671e-05, + "loss": NaN, + "step": 2300 + }, + { + "epoch": 0.37, + "eval_loss": 0.6183672547340393, + "eval_runtime": 8.3896, + "eval_samples_per_second": 11.92, + "eval_wer": 0.4725943970767357, + "step": 2300 + }, + { + "epoch": 0.39, + "learning_rate": 9.875999347364987e-05, + "loss": NaN, + "step": 2400 + }, + { + "epoch": 0.39, + "eval_loss": 0.6048241257667542, + "eval_runtime": 10.2676, + "eval_samples_per_second": 9.739, + "eval_wer": 0.464068209500609, + "step": 2400 + }, + { + "epoch": 0.41, + "learning_rate": 9.869472997226302e-05, + "loss": NaN, + "step": 2500 + }, + { + "epoch": 0.41, + "eval_loss": 0.6044276356697083, + "eval_runtime": 8.3824, + "eval_samples_per_second": 11.93, + "eval_wer": 0.49573690621193667, + "step": 2500 + }, + { + "epoch": 0.42, + "learning_rate": 9.862946647087616e-05, + "loss": NaN, + "step": 2600 + }, + { + "epoch": 0.42, + "eval_loss": 0.6002959609031677, + "eval_runtime": 9.9471, + "eval_samples_per_second": 10.053, + "eval_wer": 0.4835566382460414, + "step": 2600 + }, + { + "epoch": 0.44, + "learning_rate": 9.856420296948932e-05, + "loss": NaN, + "step": 2700 + }, + { + "epoch": 0.44, + "eval_loss": 0.5933263897895813, + "eval_runtime": 9.0583, + "eval_samples_per_second": 11.04, + "eval_wer": 0.4835566382460414, + "step": 2700 + }, + { + "epoch": 0.46, + "learning_rate": 9.849893946810246e-05, + "loss": NaN, + "step": 2800 + }, + { + "epoch": 0.46, + "eval_loss": 0.6014515161514282, + "eval_runtime": 8.0695, + "eval_samples_per_second": 12.392, + "eval_wer": 0.46650426309378806, + "step": 2800 + }, + { + "epoch": 0.47, + "learning_rate": 9.843367596671562e-05, + "loss": NaN, + "step": 2900 + }, + { + "epoch": 0.47, + "eval_loss": 0.5901432037353516, + "eval_runtime": 6.798, + "eval_samples_per_second": 14.71, + "eval_wer": 0.4762484774665043, + "step": 2900 + }, + { + "epoch": 0.49, + "learning_rate": 9.836841246532877e-05, + "loss": NaN, + "step": 3000 + }, + { + "epoch": 0.49, + "eval_loss": 0.626334011554718, + "eval_runtime": 7.8972, + "eval_samples_per_second": 12.663, + "eval_wer": 0.5066991473812423, + "step": 3000 + }, + { + "epoch": 0.5, + "learning_rate": 9.830314896394191e-05, + "loss": NaN, + "step": 3100 + }, + { + "epoch": 0.5, + "eval_loss": 0.5547911524772644, + "eval_runtime": 9.1187, + "eval_samples_per_second": 10.966, + "eval_wer": 0.464068209500609, + "step": 3100 + }, + { + "epoch": 0.52, + "learning_rate": 9.823788546255507e-05, + "loss": NaN, + "step": 3200 + }, + { + "epoch": 0.52, + "eval_loss": 0.5864753127098083, + "eval_runtime": 8.1423, + "eval_samples_per_second": 12.282, + "eval_wer": 0.48112058465286234, + "step": 3200 + }, + { + "epoch": 0.54, + "learning_rate": 9.817262196116822e-05, + "loss": NaN, + "step": 3300 + }, + { + "epoch": 0.54, + "eval_loss": 0.573021650314331, + "eval_runtime": 7.0496, + "eval_samples_per_second": 14.185, + "eval_wer": 0.45554202192448234, + "step": 3300 + }, + { + "epoch": 0.55, + "learning_rate": 9.810735845978138e-05, + "loss": NaN, + "step": 3400 + }, + { + "epoch": 0.55, + "eval_loss": 0.5912113785743713, + "eval_runtime": 7.7808, + "eval_samples_per_second": 12.852, + "eval_wer": 0.4543239951278928, + "step": 3400 + }, + { + "epoch": 0.57, + "learning_rate": 9.804209495839452e-05, + "loss": NaN, + "step": 3500 + }, + { + "epoch": 0.57, + "eval_loss": 0.5591552257537842, + "eval_runtime": 8.6018, + "eval_samples_per_second": 11.625, + "eval_wer": 0.4835566382460414, + "step": 3500 + }, + { + "epoch": 0.59, + "learning_rate": 9.797683145700768e-05, + "loss": NaN, + "step": 3600 + }, + { + "epoch": 0.59, + "eval_loss": 0.5811930298805237, + "eval_runtime": 11.1558, + "eval_samples_per_second": 8.964, + "eval_wer": 0.47990255785627284, + "step": 3600 + }, + { + "epoch": 0.6, + "learning_rate": 9.791156795562083e-05, + "loss": NaN, + "step": 3700 + }, + { + "epoch": 0.6, + "eval_loss": 0.634550154209137, + "eval_runtime": 6.9892, + "eval_samples_per_second": 14.308, + "eval_wer": 0.464068209500609, + "step": 3700 + }, + { + "epoch": 0.62, + "learning_rate": 9.784630445423397e-05, + "loss": NaN, + "step": 3800 + }, + { + "epoch": 0.62, + "eval_loss": 0.6496349573135376, + "eval_runtime": 8.2428, + "eval_samples_per_second": 12.132, + "eval_wer": 0.4993909866017052, + "step": 3800 + }, + { + "epoch": 0.63, + "learning_rate": 9.778104095284713e-05, + "loss": NaN, + "step": 3900 + }, + { + "epoch": 0.63, + "eval_loss": 0.6067304611206055, + "eval_runtime": 7.9623, + "eval_samples_per_second": 12.559, + "eval_wer": 0.464068209500609, + "step": 3900 + }, + { + "epoch": 0.65, + "learning_rate": 9.771577745146027e-05, + "loss": NaN, + "step": 4000 + }, + { + "epoch": 0.65, + "eval_loss": 0.6350998878479004, + "eval_runtime": 7.586, + "eval_samples_per_second": 13.182, + "eval_wer": 0.47381242387332523, + "step": 4000 + }, + { + "epoch": 0.67, + "learning_rate": 9.765051395007343e-05, + "loss": NaN, + "step": 4100 + }, + { + "epoch": 0.67, + "eval_loss": 0.54295814037323, + "eval_runtime": 8.9327, + "eval_samples_per_second": 11.195, + "eval_wer": 0.45554202192448234, + "step": 4100 + }, + { + "epoch": 0.68, + "learning_rate": 9.758525044868658e-05, + "loss": NaN, + "step": 4200 + }, + { + "epoch": 0.68, + "eval_loss": 0.6605328321456909, + "eval_runtime": 9.5705, + "eval_samples_per_second": 10.449, + "eval_wer": 0.46528623629719856, + "step": 4200 + }, + { + "epoch": 0.7, + "learning_rate": 9.751998694729972e-05, + "loss": NaN, + "step": 4300 + }, + { + "epoch": 0.7, + "eval_loss": 0.580009937286377, + "eval_runtime": 10.5327, + "eval_samples_per_second": 9.494, + "eval_wer": 0.464068209500609, + "step": 4300 + }, + { + "epoch": 0.72, + "learning_rate": 9.745472344591288e-05, + "loss": NaN, + "step": 4400 + }, + { + "epoch": 0.72, + "eval_loss": 0.579453706741333, + "eval_runtime": 10.8819, + "eval_samples_per_second": 9.19, + "eval_wer": 0.46163215590742995, + "step": 4400 + }, + { + "epoch": 0.73, + "learning_rate": 9.738945994452603e-05, + "loss": NaN, + "step": 4500 + }, + { + "epoch": 0.73, + "eval_loss": 0.655251145362854, + "eval_runtime": 9.1533, + "eval_samples_per_second": 10.925, + "eval_wer": 0.4591961023142509, + "step": 4500 + }, + { + "epoch": 0.75, + "learning_rate": 9.732419644313919e-05, + "loss": NaN, + "step": 4600 + }, + { + "epoch": 0.75, + "eval_loss": 0.5785291790962219, + "eval_runtime": 6.6729, + "eval_samples_per_second": 14.986, + "eval_wer": 0.44336175395858707, + "step": 4600 + }, + { + "epoch": 0.76, + "learning_rate": 9.725893294175233e-05, + "loss": NaN, + "step": 4700 + }, + { + "epoch": 0.76, + "eval_loss": 0.5562949776649475, + "eval_runtime": 12.0553, + "eval_samples_per_second": 8.295, + "eval_wer": 0.4762484774665043, + "step": 4700 + }, + { + "epoch": 0.78, + "learning_rate": 9.719366944036548e-05, + "loss": NaN, + "step": 4800 + }, + { + "epoch": 0.78, + "eval_loss": 0.6431012749671936, + "eval_runtime": 10.2668, + "eval_samples_per_second": 9.74, + "eval_wer": 0.46041412911084045, + "step": 4800 + }, + { + "epoch": 0.8, + "learning_rate": 9.712840593897863e-05, + "loss": NaN, + "step": 4900 + }, + { + "epoch": 0.8, + "eval_loss": 0.5410518050193787, + "eval_runtime": 7.7698, + "eval_samples_per_second": 12.87, + "eval_wer": 0.44336175395858707, + "step": 4900 + }, + { + "epoch": 0.81, + "learning_rate": 9.706314243759178e-05, + "loss": NaN, + "step": 5000 + }, + { + "epoch": 0.81, + "eval_loss": 0.5471655130386353, + "eval_runtime": 9.2576, + "eval_samples_per_second": 10.802, + "eval_wer": 0.4591961023142509, + "step": 5000 + }, + { + "epoch": 0.83, + "learning_rate": 9.699787893620494e-05, + "loss": NaN, + "step": 5100 + }, + { + "epoch": 0.83, + "eval_loss": 0.5607233643531799, + "eval_runtime": 9.7403, + "eval_samples_per_second": 10.267, + "eval_wer": 0.4457978075517661, + "step": 5100 + }, + { + "epoch": 0.85, + "learning_rate": 9.693261543481808e-05, + "loss": NaN, + "step": 5200 + }, + { + "epoch": 0.85, + "eval_loss": 0.5090478658676147, + "eval_runtime": 7.6231, + "eval_samples_per_second": 13.118, + "eval_wer": 0.42509135200974424, + "step": 5200 + }, + { + "epoch": 0.86, + "learning_rate": 9.686735193343124e-05, + "loss": NaN, + "step": 5300 + }, + { + "epoch": 0.86, + "eval_loss": 0.5385191440582275, + "eval_runtime": 8.5521, + "eval_samples_per_second": 11.693, + "eval_wer": 0.44336175395858707, + "step": 5300 + }, + { + "epoch": 0.88, + "learning_rate": 9.680208843204437e-05, + "loss": NaN, + "step": 5400 + }, + { + "epoch": 0.88, + "eval_loss": 0.5557727217674255, + "eval_runtime": 7.4167, + "eval_samples_per_second": 13.483, + "eval_wer": 0.4591961023142509, + "step": 5400 + }, + { + "epoch": 0.89, + "learning_rate": 9.673682493065753e-05, + "loss": NaN, + "step": 5500 + }, + { + "epoch": 0.89, + "eval_loss": 0.5240246653556824, + "eval_runtime": 8.9322, + "eval_samples_per_second": 11.195, + "eval_wer": 0.42630937880633374, + "step": 5500 + }, + { + "epoch": 0.91, + "learning_rate": 9.667156142927069e-05, + "loss": NaN, + "step": 5600 + }, + { + "epoch": 0.91, + "eval_loss": 0.5305324792861938, + "eval_runtime": 6.8749, + "eval_samples_per_second": 14.546, + "eval_wer": 0.440925700365408, + "step": 5600 + }, + { + "epoch": 0.93, + "learning_rate": 9.660629792788384e-05, + "loss": NaN, + "step": 5700 + }, + { + "epoch": 0.93, + "eval_loss": 0.4999062716960907, + "eval_runtime": 8.9284, + "eval_samples_per_second": 11.2, + "eval_wer": 0.43727161997563946, + "step": 5700 + }, + { + "epoch": 0.94, + "learning_rate": 9.6541034426497e-05, + "loss": NaN, + "step": 5800 + }, + { + "epoch": 0.94, + "eval_loss": 0.5356720089912415, + "eval_runtime": 7.3019, + "eval_samples_per_second": 13.695, + "eval_wer": 0.4299634591961023, + "step": 5800 + }, + { + "epoch": 0.96, + "learning_rate": 9.647577092511013e-05, + "loss": NaN, + "step": 5900 + }, + { + "epoch": 0.96, + "eval_loss": 0.5162432193756104, + "eval_runtime": 7.7242, + "eval_samples_per_second": 12.946, + "eval_wer": 0.4177831912302071, + "step": 5900 + }, + { + "epoch": 0.98, + "learning_rate": 9.641050742372329e-05, + "loss": NaN, + "step": 6000 + }, + { + "epoch": 0.98, + "eval_loss": 0.541946291923523, + "eval_runtime": 8.7592, + "eval_samples_per_second": 11.417, + "eval_wer": 0.440925700365408, + "step": 6000 + }, + { + "epoch": 0.99, + "learning_rate": 9.634524392233643e-05, + "loss": NaN, + "step": 6100 + }, + { + "epoch": 0.99, + "eval_loss": 0.5452945828437805, + "eval_runtime": 8.1867, + "eval_samples_per_second": 12.215, + "eval_wer": 0.44336175395858707, + "step": 6100 + }, + { + "epoch": 1.01, + "learning_rate": 9.627998042094959e-05, + "loss": NaN, + "step": 6200 + }, + { + "epoch": 1.01, + "eval_loss": 0.6135980486869812, + "eval_runtime": 8.1992, + "eval_samples_per_second": 12.196, + "eval_wer": 0.44336175395858707, + "step": 6200 + }, + { + "epoch": 1.02, + "learning_rate": 9.621471691956275e-05, + "loss": NaN, + "step": 6300 + }, + { + "epoch": 1.02, + "eval_loss": 0.5763392448425293, + "eval_runtime": 8.7786, + "eval_samples_per_second": 11.391, + "eval_wer": 0.4457978075517661, + "step": 6300 + }, + { + "epoch": 1.04, + "learning_rate": 9.61494534181759e-05, + "loss": NaN, + "step": 6400 + }, + { + "epoch": 1.04, + "eval_loss": 0.5169751048088074, + "eval_runtime": 8.8344, + "eval_samples_per_second": 11.319, + "eval_wer": 0.44945188794153473, + "step": 6400 + }, + { + "epoch": 1.06, + "learning_rate": 9.608418991678904e-05, + "loss": NaN, + "step": 6500 + }, + { + "epoch": 1.06, + "eval_loss": 0.5410892367362976, + "eval_runtime": 8.9659, + "eval_samples_per_second": 11.153, + "eval_wer": 0.41291108404384896, + "step": 6500 + }, + { + "epoch": 1.07, + "learning_rate": 9.601892641540218e-05, + "loss": NaN, + "step": 6600 + }, + { + "epoch": 1.07, + "eval_loss": 0.6024890542030334, + "eval_runtime": 9.0589, + "eval_samples_per_second": 11.039, + "eval_wer": 0.46650426309378806, + "step": 6600 + }, + { + "epoch": 1.09, + "learning_rate": 9.595366291401534e-05, + "loss": NaN, + "step": 6700 + }, + { + "epoch": 1.09, + "eval_loss": 0.5081002116203308, + "eval_runtime": 6.9031, + "eval_samples_per_second": 14.486, + "eval_wer": 0.4445797807551766, + "step": 6700 + }, + { + "epoch": 1.11, + "learning_rate": 9.588839941262849e-05, + "loss": NaN, + "step": 6800 + }, + { + "epoch": 1.11, + "eval_loss": 0.537754237651825, + "eval_runtime": 6.9434, + "eval_samples_per_second": 14.402, + "eval_wer": 0.44336175395858707, + "step": 6800 + }, + { + "epoch": 1.12, + "learning_rate": 9.582313591124165e-05, + "loss": NaN, + "step": 6900 + }, + { + "epoch": 1.12, + "eval_loss": 0.5683107972145081, + "eval_runtime": 7.8602, + "eval_samples_per_second": 12.722, + "eval_wer": 0.4336175395858709, + "step": 6900 + }, + { + "epoch": 1.14, + "learning_rate": 9.57578724098548e-05, + "loss": NaN, + "step": 7000 + }, + { + "epoch": 1.14, + "eval_loss": 0.6018743515014648, + "eval_runtime": 7.5695, + "eval_samples_per_second": 13.211, + "eval_wer": 0.4348355663824604, + "step": 7000 + }, + { + "epoch": 1.15, + "learning_rate": 9.569260890846794e-05, + "loss": NaN, + "step": 7100 + }, + { + "epoch": 1.15, + "eval_loss": 0.5463069677352905, + "eval_runtime": 8.1373, + "eval_samples_per_second": 12.289, + "eval_wer": 0.42752740560292324, + "step": 7100 + }, + { + "epoch": 1.17, + "learning_rate": 9.56273454070811e-05, + "loss": NaN, + "step": 7200 + }, + { + "epoch": 1.17, + "eval_loss": 0.5245855450630188, + "eval_runtime": 7.3379, + "eval_samples_per_second": 13.628, + "eval_wer": 0.42752740560292324, + "step": 7200 + }, + { + "epoch": 1.19, + "learning_rate": 9.556208190569424e-05, + "loss": NaN, + "step": 7300 + }, + { + "epoch": 1.19, + "eval_loss": 0.48042479157447815, + "eval_runtime": 6.6664, + "eval_samples_per_second": 15.001, + "eval_wer": 0.4336175395858709, + "step": 7300 + }, + { + "epoch": 1.2, + "learning_rate": 9.54968184043074e-05, + "loss": NaN, + "step": 7400 + }, + { + "epoch": 1.2, + "eval_loss": 0.4897633492946625, + "eval_runtime": 6.6761, + "eval_samples_per_second": 14.979, + "eval_wer": 0.4445797807551766, + "step": 7400 + }, + { + "epoch": 1.22, + "learning_rate": 9.543155490292054e-05, + "loss": NaN, + "step": 7500 + }, + { + "epoch": 1.22, + "eval_loss": 0.5400981903076172, + "eval_runtime": 7.744, + "eval_samples_per_second": 12.913, + "eval_wer": 0.4238733252131547, + "step": 7500 + }, + { + "epoch": 1.24, + "learning_rate": 9.536629140153369e-05, + "loss": NaN, + "step": 7600 + }, + { + "epoch": 1.24, + "eval_loss": 0.5461037158966064, + "eval_runtime": 7.6292, + "eval_samples_per_second": 13.108, + "eval_wer": 0.43118148599269185, + "step": 7600 + }, + { + "epoch": 1.25, + "learning_rate": 9.530102790014685e-05, + "loss": NaN, + "step": 7700 + }, + { + "epoch": 1.25, + "eval_loss": 0.4961685538291931, + "eval_runtime": 7.2783, + "eval_samples_per_second": 13.739, + "eval_wer": 0.4226552984165652, + "step": 7700 + }, + { + "epoch": 1.27, + "learning_rate": 9.523576439875999e-05, + "loss": NaN, + "step": 7800 + }, + { + "epoch": 1.27, + "eval_loss": 0.6107471585273743, + "eval_runtime": 8.6517, + "eval_samples_per_second": 11.558, + "eval_wer": 0.438489646772229, + "step": 7800 + }, + { + "epoch": 1.28, + "learning_rate": 9.517050089737315e-05, + "loss": NaN, + "step": 7900 + }, + { + "epoch": 1.28, + "eval_loss": 0.5769979953765869, + "eval_runtime": 7.4406, + "eval_samples_per_second": 13.44, + "eval_wer": 0.43605359317904996, + "step": 7900 + }, + { + "epoch": 1.3, + "learning_rate": 9.51052373959863e-05, + "loss": NaN, + "step": 8000 + }, + { + "epoch": 1.3, + "eval_loss": 0.5004762411117554, + "eval_runtime": 8.9614, + "eval_samples_per_second": 11.159, + "eval_wer": 0.43239951278928135, + "step": 8000 + }, + { + "epoch": 1.32, + "learning_rate": 9.503997389459946e-05, + "loss": NaN, + "step": 8100 + }, + { + "epoch": 1.32, + "eval_loss": 0.46339115500450134, + "eval_runtime": 7.2515, + "eval_samples_per_second": 13.79, + "eval_wer": 0.4104750304506699, + "step": 8100 + }, + { + "epoch": 1.33, + "learning_rate": 9.49747103932126e-05, + "loss": NaN, + "step": 8200 + }, + { + "epoch": 1.33, + "eval_loss": 0.5336301922798157, + "eval_runtime": 8.7749, + "eval_samples_per_second": 11.396, + "eval_wer": 0.44336175395858707, + "step": 8200 + }, + { + "epoch": 1.35, + "learning_rate": 9.490944689182575e-05, + "loss": NaN, + "step": 8300 + }, + { + "epoch": 1.35, + "eval_loss": 0.5296890139579773, + "eval_runtime": 8.6952, + "eval_samples_per_second": 11.501, + "eval_wer": 0.4238733252131547, + "step": 8300 + }, + { + "epoch": 1.37, + "learning_rate": 9.48441833904389e-05, + "loss": NaN, + "step": 8400 + }, + { + "epoch": 1.37, + "eval_loss": 0.5209415555000305, + "eval_runtime": 8.7658, + "eval_samples_per_second": 11.408, + "eval_wer": 0.4482338611449452, + "step": 8400 + }, + { + "epoch": 1.38, + "learning_rate": 9.477891988905205e-05, + "loss": NaN, + "step": 8500 + }, + { + "epoch": 1.38, + "eval_loss": 0.5313872694969177, + "eval_runtime": 8.6447, + "eval_samples_per_second": 11.568, + "eval_wer": 0.4287454323995128, + "step": 8500 + }, + { + "epoch": 1.4, + "learning_rate": 9.471365638766521e-05, + "loss": NaN, + "step": 8600 + }, + { + "epoch": 1.4, + "eval_loss": 0.5957339406013489, + "eval_runtime": 9.3851, + "eval_samples_per_second": 10.655, + "eval_wer": 0.4445797807551766, + "step": 8600 + }, + { + "epoch": 1.41, + "learning_rate": 9.464839288627835e-05, + "loss": NaN, + "step": 8700 + }, + { + "epoch": 1.41, + "eval_loss": 0.4665583074092865, + "eval_runtime": 7.7597, + "eval_samples_per_second": 12.887, + "eval_wer": 0.4336175395858709, + "step": 8700 + }, + { + "epoch": 1.43, + "learning_rate": 9.45831293848915e-05, + "loss": NaN, + "step": 8800 + }, + { + "epoch": 1.43, + "eval_loss": 0.5392510890960693, + "eval_runtime": 8.1949, + "eval_samples_per_second": 12.203, + "eval_wer": 0.4336175395858709, + "step": 8800 + }, + { + "epoch": 1.45, + "learning_rate": 9.451786588350466e-05, + "loss": NaN, + "step": 8900 + }, + { + "epoch": 1.45, + "eval_loss": 0.5618553161621094, + "eval_runtime": 7.3195, + "eval_samples_per_second": 13.662, + "eval_wer": 0.4482338611449452, + "step": 8900 + }, + { + "epoch": 1.46, + "learning_rate": 9.44526023821178e-05, + "loss": NaN, + "step": 9000 + }, + { + "epoch": 1.46, + "eval_loss": 0.6219630837440491, + "eval_runtime": 7.7967, + "eval_samples_per_second": 12.826, + "eval_wer": 0.4713763702801462, + "step": 9000 + }, + { + "epoch": 1.48, + "learning_rate": 9.438733888073096e-05, + "loss": NaN, + "step": 9100 + }, + { + "epoch": 1.48, + "eval_loss": 0.6084015369415283, + "eval_runtime": 8.535, + "eval_samples_per_second": 11.717, + "eval_wer": 0.4470158343483557, + "step": 9100 + }, + { + "epoch": 1.5, + "learning_rate": 9.43220753793441e-05, + "loss": NaN, + "step": 9200 + }, + { + "epoch": 1.5, + "eval_loss": 0.5832752585411072, + "eval_runtime": 8.3754, + "eval_samples_per_second": 11.94, + "eval_wer": 0.43727161997563946, + "step": 9200 + }, + { + "epoch": 1.51, + "learning_rate": 9.425681187795725e-05, + "loss": NaN, + "step": 9300 + }, + { + "epoch": 1.51, + "eval_loss": 0.5465306043624878, + "eval_runtime": 6.9903, + "eval_samples_per_second": 14.306, + "eval_wer": 0.43118148599269185, + "step": 9300 + }, + { + "epoch": 1.53, + "learning_rate": 9.419154837657041e-05, + "loss": NaN, + "step": 9400 + }, + { + "epoch": 1.53, + "eval_loss": 0.527911901473999, + "eval_runtime": 7.4382, + "eval_samples_per_second": 13.444, + "eval_wer": 0.4518879415347138, + "step": 9400 + }, + { + "epoch": 1.54, + "learning_rate": 9.412628487518356e-05, + "loss": NaN, + "step": 9500 + }, + { + "epoch": 1.54, + "eval_loss": 0.5745791792869568, + "eval_runtime": 9.6658, + "eval_samples_per_second": 10.346, + "eval_wer": 0.4238733252131547, + "step": 9500 + }, + { + "epoch": 1.56, + "learning_rate": 9.406102137379671e-05, + "loss": NaN, + "step": 9600 + }, + { + "epoch": 1.56, + "eval_loss": 0.6281310319900513, + "eval_runtime": 10.5392, + "eval_samples_per_second": 9.488, + "eval_wer": 0.4531059683313033, + "step": 9600 + }, + { + "epoch": 1.58, + "learning_rate": 9.399575787240986e-05, + "loss": NaN, + "step": 9700 + }, + { + "epoch": 1.58, + "eval_loss": 0.5570322871208191, + "eval_runtime": 7.2179, + "eval_samples_per_second": 13.854, + "eval_wer": 0.44336175395858707, + "step": 9700 + }, + { + "epoch": 1.59, + "learning_rate": 9.393049437102302e-05, + "loss": NaN, + "step": 9800 + }, + { + "epoch": 1.59, + "eval_loss": 0.5097599625587463, + "eval_runtime": 8.5291, + "eval_samples_per_second": 11.725, + "eval_wer": 0.43605359317904996, + "step": 9800 + }, + { + "epoch": 1.61, + "learning_rate": 9.386523086963615e-05, + "loss": NaN, + "step": 9900 + }, + { + "epoch": 1.61, + "eval_loss": 0.5071770548820496, + "eval_runtime": 8.0326, + "eval_samples_per_second": 12.449, + "eval_wer": 0.4190012180267966, + "step": 9900 + }, + { + "epoch": 1.63, + "learning_rate": 9.379996736824931e-05, + "loss": NaN, + "step": 10000 + }, + { + "epoch": 1.63, + "eval_loss": 0.5283228158950806, + "eval_runtime": 7.0447, + "eval_samples_per_second": 14.195, + "eval_wer": 0.42752740560292324, + "step": 10000 + }, + { + "epoch": 1.64, + "learning_rate": 9.373470386686247e-05, + "loss": NaN, + "step": 10100 + }, + { + "epoch": 1.64, + "eval_loss": 0.6241239309310913, + "eval_runtime": 7.6906, + "eval_samples_per_second": 13.003, + "eval_wer": 0.45676004872107184, + "step": 10100 + }, + { + "epoch": 1.66, + "learning_rate": 9.366944036547561e-05, + "loss": NaN, + "step": 10200 + }, + { + "epoch": 1.66, + "eval_loss": 0.5286669731140137, + "eval_runtime": 7.6387, + "eval_samples_per_second": 13.091, + "eval_wer": 0.4445797807551766, + "step": 10200 + }, + { + "epoch": 1.68, + "learning_rate": 9.360417686408877e-05, + "loss": NaN, + "step": 10300 + }, + { + "epoch": 1.68, + "eval_loss": 0.560374915599823, + "eval_runtime": 7.7761, + "eval_samples_per_second": 12.86, + "eval_wer": 0.4518879415347138, + "step": 10300 + }, + { + "epoch": 1.69, + "learning_rate": 9.353891336270192e-05, + "loss": NaN, + "step": 10400 + }, + { + "epoch": 1.69, + "eval_loss": 0.5197404623031616, + "eval_runtime": 7.8548, + "eval_samples_per_second": 12.731, + "eval_wer": 0.4445797807551766, + "step": 10400 + }, + { + "epoch": 1.71, + "learning_rate": 9.347364986131506e-05, + "loss": NaN, + "step": 10500 + }, + { + "epoch": 1.71, + "eval_loss": 0.5106395483016968, + "eval_runtime": 8.6218, + "eval_samples_per_second": 11.598, + "eval_wer": 0.44214372716199757, + "step": 10500 + }, + { + "epoch": 1.72, + "learning_rate": 9.34083863599282e-05, + "loss": NaN, + "step": 10600 + }, + { + "epoch": 1.72, + "eval_loss": 0.5029647350311279, + "eval_runtime": 9.875, + "eval_samples_per_second": 10.127, + "eval_wer": 0.4445797807551766, + "step": 10600 + }, + { + "epoch": 1.74, + "learning_rate": 9.334312285854136e-05, + "loss": NaN, + "step": 10700 + }, + { + "epoch": 1.74, + "eval_loss": 0.5408429503440857, + "eval_runtime": 8.0516, + "eval_samples_per_second": 12.42, + "eval_wer": 0.4397076735688185, + "step": 10700 + }, + { + "epoch": 1.76, + "learning_rate": 9.327785935715452e-05, + "loss": NaN, + "step": 10800 + }, + { + "epoch": 1.76, + "eval_loss": 0.599218487739563, + "eval_runtime": 9.8313, + "eval_samples_per_second": 10.172, + "eval_wer": 0.43118148599269185, + "step": 10800 + }, + { + "epoch": 1.77, + "learning_rate": 9.321259585576767e-05, + "loss": NaN, + "step": 10900 + }, + { + "epoch": 1.77, + "eval_loss": 0.5169634819030762, + "eval_runtime": 7.002, + "eval_samples_per_second": 14.282, + "eval_wer": 0.4092570036540804, + "step": 10900 + }, + { + "epoch": 1.79, + "learning_rate": 9.314733235438081e-05, + "loss": NaN, + "step": 11000 + }, + { + "epoch": 1.79, + "eval_loss": 0.5207093954086304, + "eval_runtime": 7.4268, + "eval_samples_per_second": 13.465, + "eval_wer": 0.42752740560292324, + "step": 11000 + }, + { + "epoch": 1.81, + "learning_rate": 9.308206885299396e-05, + "loss": NaN, + "step": 11100 + }, + { + "epoch": 1.81, + "eval_loss": 0.5467461347579956, + "eval_runtime": 7.8754, + "eval_samples_per_second": 12.698, + "eval_wer": 0.4226552984165652, + "step": 11100 + }, + { + "epoch": 1.82, + "learning_rate": 9.301680535160712e-05, + "loss": NaN, + "step": 11200 + }, + { + "epoch": 1.82, + "eval_loss": 0.49166274070739746, + "eval_runtime": 7.5148, + "eval_samples_per_second": 13.307, + "eval_wer": 0.4104750304506699, + "step": 11200 + }, + { + "epoch": 1.84, + "learning_rate": 9.295154185022028e-05, + "loss": NaN, + "step": 11300 + }, + { + "epoch": 1.84, + "eval_loss": 0.49259471893310547, + "eval_runtime": 8.1781, + "eval_samples_per_second": 12.228, + "eval_wer": 0.4336175395858709, + "step": 11300 + }, + { + "epoch": 1.85, + "learning_rate": 9.288627834883342e-05, + "loss": NaN, + "step": 11400 + }, + { + "epoch": 1.85, + "eval_loss": 0.666587233543396, + "eval_runtime": 7.8835, + "eval_samples_per_second": 12.685, + "eval_wer": 0.43605359317904996, + "step": 11400 + }, + { + "epoch": 1.87, + "learning_rate": 9.282101484744658e-05, + "loss": NaN, + "step": 11500 + }, + { + "epoch": 1.87, + "eval_loss": 0.5402622222900391, + "eval_runtime": 7.727, + "eval_samples_per_second": 12.942, + "eval_wer": 0.42630937880633374, + "step": 11500 + }, + { + "epoch": 1.89, + "learning_rate": 9.275575134605971e-05, + "loss": NaN, + "step": 11600 + }, + { + "epoch": 1.89, + "eval_loss": 0.5431535840034485, + "eval_runtime": 8.2898, + "eval_samples_per_second": 12.063, + "eval_wer": 0.4397076735688185, + "step": 11600 + }, + { + "epoch": 1.9, + "learning_rate": 9.269048784467287e-05, + "loss": NaN, + "step": 11700 + }, + { + "epoch": 1.9, + "eval_loss": 0.5058331489562988, + "eval_runtime": 7.6428, + "eval_samples_per_second": 13.084, + "eval_wer": 0.4190012180267966, + "step": 11700 + }, + { + "epoch": 1.92, + "learning_rate": 9.262522434328602e-05, + "loss": NaN, + "step": 11800 + }, + { + "epoch": 1.92, + "eval_loss": 0.5385850667953491, + "eval_runtime": 9.2261, + "eval_samples_per_second": 10.839, + "eval_wer": 0.4579780755176614, + "step": 11800 + }, + { + "epoch": 1.94, + "learning_rate": 9.255996084189917e-05, + "loss": NaN, + "step": 11900 + }, + { + "epoch": 1.94, + "eval_loss": 0.5211225748062134, + "eval_runtime": 8.2929, + "eval_samples_per_second": 12.058, + "eval_wer": 0.4470158343483557, + "step": 11900 + }, + { + "epoch": 1.95, + "learning_rate": 9.249469734051233e-05, + "loss": NaN, + "step": 12000 + }, + { + "epoch": 1.95, + "eval_loss": 0.44943609833717346, + "eval_runtime": 7.5, + "eval_samples_per_second": 13.333, + "eval_wer": 0.4287454323995128, + "step": 12000 + }, + { + "epoch": 1.97, + "learning_rate": 9.242943383912548e-05, + "loss": NaN, + "step": 12100 + }, + { + "epoch": 1.97, + "eval_loss": 0.44961708784103394, + "eval_runtime": 8.559, + "eval_samples_per_second": 11.684, + "eval_wer": 0.4007308160779537, + "step": 12100 + }, + { + "epoch": 1.98, + "learning_rate": 9.236417033773862e-05, + "loss": NaN, + "step": 12200 + }, + { + "epoch": 1.98, + "eval_loss": 0.5775151252746582, + "eval_runtime": 8.2678, + "eval_samples_per_second": 12.095, + "eval_wer": 0.42630937880633374, + "step": 12200 + }, + { + "epoch": 2.0, + "learning_rate": 9.229890683635177e-05, + "loss": NaN, + "step": 12300 + }, + { + "epoch": 2.0, + "eval_loss": 0.5101304054260254, + "eval_runtime": 9.5041, + "eval_samples_per_second": 10.522, + "eval_wer": 0.4470158343483557, + "step": 12300 + }, + { + "epoch": 2.02, + "learning_rate": 9.223364333496493e-05, + "loss": NaN, + "step": 12400 + }, + { + "epoch": 2.02, + "eval_loss": 0.5802700519561768, + "eval_runtime": 8.978, + "eval_samples_per_second": 11.138, + "eval_wer": 0.42021924482338613, + "step": 12400 + }, + { + "epoch": 2.03, + "learning_rate": 9.216837983357807e-05, + "loss": NaN, + "step": 12500 + }, + { + "epoch": 2.03, + "eval_loss": 0.4817795157432556, + "eval_runtime": 8.6515, + "eval_samples_per_second": 11.559, + "eval_wer": 0.4177831912302071, + "step": 12500 + }, + { + "epoch": 2.05, + "learning_rate": 9.210311633219123e-05, + "loss": NaN, + "step": 12600 + }, + { + "epoch": 2.05, + "eval_loss": 0.5166882276535034, + "eval_runtime": 9.9407, + "eval_samples_per_second": 10.06, + "eval_wer": 0.44214372716199757, + "step": 12600 + }, + { + "epoch": 2.07, + "learning_rate": 9.203785283080438e-05, + "loss": NaN, + "step": 12700 + }, + { + "epoch": 2.07, + "eval_loss": 0.49081742763519287, + "eval_runtime": 7.4879, + "eval_samples_per_second": 13.355, + "eval_wer": 0.42752740560292324, + "step": 12700 + }, + { + "epoch": 2.08, + "learning_rate": 9.197258932941752e-05, + "loss": NaN, + "step": 12800 + }, + { + "epoch": 2.08, + "eval_loss": 0.5209938287734985, + "eval_runtime": 9.1719, + "eval_samples_per_second": 10.903, + "eval_wer": 0.4457978075517661, + "step": 12800 + }, + { + "epoch": 2.1, + "learning_rate": 9.190732582803068e-05, + "loss": NaN, + "step": 12900 + }, + { + "epoch": 2.1, + "eval_loss": 0.4667002856731415, + "eval_runtime": 8.2262, + "eval_samples_per_second": 12.156, + "eval_wer": 0.4190012180267966, + "step": 12900 + }, + { + "epoch": 2.11, + "learning_rate": 9.184206232664382e-05, + "loss": NaN, + "step": 13000 + }, + { + "epoch": 2.11, + "eval_loss": 0.44741660356521606, + "eval_runtime": 6.9715, + "eval_samples_per_second": 14.344, + "eval_wer": 0.42630937880633374, + "step": 13000 + }, + { + "epoch": 2.13, + "learning_rate": 9.177679882525698e-05, + "loss": NaN, + "step": 13100 + }, + { + "epoch": 2.13, + "eval_loss": 0.45468372106552124, + "eval_runtime": 8.1778, + "eval_samples_per_second": 12.228, + "eval_wer": 0.4177831912302071, + "step": 13100 + }, + { + "epoch": 2.15, + "learning_rate": 9.171153532387013e-05, + "loss": NaN, + "step": 13200 + }, + { + "epoch": 2.15, + "eval_loss": 0.47137129306793213, + "eval_runtime": 9.2926, + "eval_samples_per_second": 10.761, + "eval_wer": 0.43605359317904996, + "step": 13200 + }, + { + "epoch": 2.16, + "learning_rate": 9.164627182248327e-05, + "loss": NaN, + "step": 13300 + }, + { + "epoch": 2.16, + "eval_loss": 0.505387544631958, + "eval_runtime": 8.3638, + "eval_samples_per_second": 11.956, + "eval_wer": 0.415347137637028, + "step": 13300 + }, + { + "epoch": 2.18, + "learning_rate": 9.158100832109643e-05, + "loss": NaN, + "step": 13400 + }, + { + "epoch": 2.18, + "eval_loss": 0.4471493661403656, + "eval_runtime": 7.3344, + "eval_samples_per_second": 13.634, + "eval_wer": 0.4226552984165652, + "step": 13400 + }, + { + "epoch": 2.2, + "learning_rate": 9.151574481970958e-05, + "loss": NaN, + "step": 13500 + }, + { + "epoch": 2.2, + "eval_loss": 0.4644159972667694, + "eval_runtime": 9.2128, + "eval_samples_per_second": 10.854, + "eval_wer": 0.3995127892813642, + "step": 13500 + }, + { + "epoch": 2.21, + "learning_rate": 9.145048131832274e-05, + "loss": NaN, + "step": 13600 + }, + { + "epoch": 2.21, + "eval_loss": 0.4804657995700836, + "eval_runtime": 7.5792, + "eval_samples_per_second": 13.194, + "eval_wer": 0.415347137637028, + "step": 13600 + }, + { + "epoch": 2.23, + "learning_rate": 9.138521781693588e-05, + "loss": NaN, + "step": 13700 + }, + { + "epoch": 2.23, + "eval_loss": 0.5253716111183167, + "eval_runtime": 9.797, + "eval_samples_per_second": 10.207, + "eval_wer": 0.42752740560292324, + "step": 13700 + }, + { + "epoch": 2.24, + "learning_rate": 9.131995431554904e-05, + "loss": NaN, + "step": 13800 + }, + { + "epoch": 2.24, + "eval_loss": 0.4461153745651245, + "eval_runtime": 8.0888, + "eval_samples_per_second": 12.363, + "eval_wer": 0.4177831912302071, + "step": 13800 + }, + { + "epoch": 2.26, + "learning_rate": 9.125469081416219e-05, + "loss": NaN, + "step": 13900 + }, + { + "epoch": 2.26, + "eval_loss": 0.5731090307235718, + "eval_runtime": 8.4905, + "eval_samples_per_second": 11.778, + "eval_wer": 0.43118148599269185, + "step": 13900 + }, + { + "epoch": 2.28, + "learning_rate": 9.118942731277533e-05, + "loss": NaN, + "step": 14000 + }, + { + "epoch": 2.28, + "eval_loss": 0.5009161829948425, + "eval_runtime": 8.2825, + "eval_samples_per_second": 12.074, + "eval_wer": 0.42021924482338613, + "step": 14000 + }, + { + "epoch": 2.29, + "learning_rate": 9.112416381138849e-05, + "loss": NaN, + "step": 14100 + }, + { + "epoch": 2.29, + "eval_loss": 0.4805867075920105, + "eval_runtime": 8.9811, + "eval_samples_per_second": 11.134, + "eval_wer": 0.39707673568818513, + "step": 14100 + }, + { + "epoch": 2.31, + "learning_rate": 9.105890031000163e-05, + "loss": NaN, + "step": 14200 + }, + { + "epoch": 2.31, + "eval_loss": 0.5124616026878357, + "eval_runtime": 7.2004, + "eval_samples_per_second": 13.888, + "eval_wer": 0.3982947624847747, + "step": 14200 + }, + { + "epoch": 2.33, + "learning_rate": 9.099363680861479e-05, + "loss": NaN, + "step": 14300 + }, + { + "epoch": 2.33, + "eval_loss": 0.45692193508148193, + "eval_runtime": 7.2681, + "eval_samples_per_second": 13.759, + "eval_wer": 0.4177831912302071, + "step": 14300 + }, + { + "epoch": 2.34, + "learning_rate": 9.092837330722794e-05, + "loss": NaN, + "step": 14400 + }, + { + "epoch": 2.34, + "eval_loss": 0.44331681728363037, + "eval_runtime": 8.5939, + "eval_samples_per_second": 11.636, + "eval_wer": 0.4177831912302071, + "step": 14400 + }, + { + "epoch": 2.36, + "learning_rate": 9.086310980584108e-05, + "loss": NaN, + "step": 14500 + }, + { + "epoch": 2.36, + "eval_loss": 0.494161456823349, + "eval_runtime": 7.6191, + "eval_samples_per_second": 13.125, + "eval_wer": 0.4299634591961023, + "step": 14500 + }, + { + "epoch": 2.37, + "learning_rate": 9.079784630445424e-05, + "loss": NaN, + "step": 14600 + }, + { + "epoch": 2.37, + "eval_loss": 0.5291484594345093, + "eval_runtime": 7.9069, + "eval_samples_per_second": 12.647, + "eval_wer": 0.4177831912302071, + "step": 14600 + }, + { + "epoch": 2.39, + "learning_rate": 9.073258280306739e-05, + "loss": NaN, + "step": 14700 + }, + { + "epoch": 2.39, + "eval_loss": 0.45012572407722473, + "eval_runtime": 9.4293, + "eval_samples_per_second": 10.605, + "eval_wer": 0.4056029232643118, + "step": 14700 + }, + { + "epoch": 2.41, + "learning_rate": 9.066731930168055e-05, + "loss": NaN, + "step": 14800 + }, + { + "epoch": 2.41, + "eval_loss": 0.44526755809783936, + "eval_runtime": 7.9253, + "eval_samples_per_second": 12.618, + "eval_wer": 0.42021924482338613, + "step": 14800 + }, + { + "epoch": 2.42, + "learning_rate": 9.060205580029369e-05, + "loss": NaN, + "step": 14900 + }, + { + "epoch": 2.42, + "eval_loss": 0.4633622467517853, + "eval_runtime": 7.3284, + "eval_samples_per_second": 13.646, + "eval_wer": 0.42752740560292324, + "step": 14900 + }, + { + "epoch": 2.44, + "learning_rate": 9.053679229890684e-05, + "loss": NaN, + "step": 15000 + }, + { + "epoch": 2.44, + "eval_loss": 0.45304596424102783, + "eval_runtime": 8.2189, + "eval_samples_per_second": 12.167, + "eval_wer": 0.4190012180267966, + "step": 15000 + }, + { + "epoch": 2.46, + "learning_rate": 9.047152879752e-05, + "loss": NaN, + "step": 15100 + }, + { + "epoch": 2.46, + "eval_loss": 0.563227653503418, + "eval_runtime": 7.3261, + "eval_samples_per_second": 13.65, + "eval_wer": 0.4238733252131547, + "step": 15100 + }, + { + "epoch": 2.47, + "learning_rate": 9.040626529613314e-05, + "loss": NaN, + "step": 15200 + }, + { + "epoch": 2.47, + "eval_loss": 0.4682408273220062, + "eval_runtime": 9.085, + "eval_samples_per_second": 11.007, + "eval_wer": 0.43118148599269185, + "step": 15200 + }, + { + "epoch": 2.49, + "learning_rate": 9.03410017947463e-05, + "loss": NaN, + "step": 15300 + }, + { + "epoch": 2.49, + "eval_loss": 0.4495187997817993, + "eval_runtime": 8.9075, + "eval_samples_per_second": 11.227, + "eval_wer": 0.4007308160779537, + "step": 15300 + }, + { + "epoch": 2.5, + "learning_rate": 9.027573829335944e-05, + "loss": NaN, + "step": 15400 + }, + { + "epoch": 2.5, + "eval_loss": 0.4402164816856384, + "eval_runtime": 10.3596, + "eval_samples_per_second": 9.653, + "eval_wer": 0.4092570036540804, + "step": 15400 + }, + { + "epoch": 2.52, + "learning_rate": 9.02104747919726e-05, + "loss": NaN, + "step": 15500 + }, + { + "epoch": 2.52, + "eval_loss": 0.45120611786842346, + "eval_runtime": 8.2246, + "eval_samples_per_second": 12.159, + "eval_wer": 0.40803897685749085, + "step": 15500 + }, + { + "epoch": 2.54, + "learning_rate": 9.014521129058573e-05, + "loss": NaN, + "step": 15600 + }, + { + "epoch": 2.54, + "eval_loss": 0.44330108165740967, + "eval_runtime": 7.0996, + "eval_samples_per_second": 14.085, + "eval_wer": 0.4007308160779537, + "step": 15600 + }, + { + "epoch": 2.55, + "learning_rate": 9.007994778919889e-05, + "loss": NaN, + "step": 15700 + }, + { + "epoch": 2.55, + "eval_loss": 0.42252305150032043, + "eval_runtime": 7.1285, + "eval_samples_per_second": 14.028, + "eval_wer": 0.3934226552984166, + "step": 15700 + }, + { + "epoch": 2.57, + "learning_rate": 9.001468428781205e-05, + "loss": NaN, + "step": 15800 + }, + { + "epoch": 2.57, + "eval_loss": 0.4437943696975708, + "eval_runtime": 8.6965, + "eval_samples_per_second": 11.499, + "eval_wer": 0.41169305724725946, + "step": 15800 + }, + { + "epoch": 2.59, + "learning_rate": 8.99494207864252e-05, + "loss": NaN, + "step": 15900 + }, + { + "epoch": 2.59, + "eval_loss": 0.47090038657188416, + "eval_runtime": 7.2716, + "eval_samples_per_second": 13.752, + "eval_wer": 0.4056029232643118, + "step": 15900 + }, + { + "epoch": 2.6, + "learning_rate": 8.988415728503836e-05, + "loss": NaN, + "step": 16000 + }, + { + "epoch": 2.6, + "eval_loss": 0.45720386505126953, + "eval_runtime": 8.165, + "eval_samples_per_second": 12.247, + "eval_wer": 0.43118148599269185, + "step": 16000 + }, + { + "epoch": 2.62, + "learning_rate": 8.981889378365149e-05, + "loss": NaN, + "step": 16100 + }, + { + "epoch": 2.62, + "eval_loss": 0.4764085114002228, + "eval_runtime": 8.2382, + "eval_samples_per_second": 12.139, + "eval_wer": 0.4043848964677223, + "step": 16100 + }, + { + "epoch": 2.63, + "learning_rate": 8.975363028226465e-05, + "loss": NaN, + "step": 16200 + }, + { + "epoch": 2.63, + "eval_loss": 0.4751797020435333, + "eval_runtime": 9.3113, + "eval_samples_per_second": 10.74, + "eval_wer": 0.41412911084043846, + "step": 16200 + }, + { + "epoch": 2.65, + "learning_rate": 8.968836678087779e-05, + "loss": NaN, + "step": 16300 + }, + { + "epoch": 2.65, + "eval_loss": 0.44270026683807373, + "eval_runtime": 7.4624, + "eval_samples_per_second": 13.401, + "eval_wer": 0.40194884287454324, + "step": 16300 + }, + { + "epoch": 2.67, + "learning_rate": 8.962310327949095e-05, + "loss": NaN, + "step": 16400 + }, + { + "epoch": 2.67, + "eval_loss": 0.4818960428237915, + "eval_runtime": 7.8783, + "eval_samples_per_second": 12.693, + "eval_wer": 0.43118148599269185, + "step": 16400 + }, + { + "epoch": 2.68, + "learning_rate": 8.955783977810411e-05, + "loss": NaN, + "step": 16500 + }, + { + "epoch": 2.68, + "eval_loss": 0.4510990083217621, + "eval_runtime": 7.04, + "eval_samples_per_second": 14.205, + "eval_wer": 0.4043848964677223, + "step": 16500 + }, + { + "epoch": 2.7, + "learning_rate": 8.949257627671725e-05, + "loss": NaN, + "step": 16600 + }, + { + "epoch": 2.7, + "eval_loss": 0.4735144078731537, + "eval_runtime": 10.0014, + "eval_samples_per_second": 9.999, + "eval_wer": 0.3995127892813642, + "step": 16600 + }, + { + "epoch": 2.72, + "learning_rate": 8.94273127753304e-05, + "loss": NaN, + "step": 16700 + }, + { + "epoch": 2.72, + "eval_loss": 0.4041927456855774, + "eval_runtime": 6.9626, + "eval_samples_per_second": 14.362, + "eval_wer": 0.40316686967113274, + "step": 16700 + }, + { + "epoch": 2.73, + "learning_rate": 8.936204927394354e-05, + "loss": NaN, + "step": 16800 + }, + { + "epoch": 2.73, + "eval_loss": 0.4579096734523773, + "eval_runtime": 8.5814, + "eval_samples_per_second": 11.653, + "eval_wer": 0.42509135200974424, + "step": 16800 + }, + { + "epoch": 2.75, + "learning_rate": 8.92967857725567e-05, + "loss": NaN, + "step": 16900 + }, + { + "epoch": 2.75, + "eval_loss": 0.46610623598098755, + "eval_runtime": 7.2885, + "eval_samples_per_second": 13.72, + "eval_wer": 0.42143727161997563, + "step": 16900 + }, + { + "epoch": 2.76, + "learning_rate": 8.923152227116985e-05, + "loss": NaN, + "step": 17000 + }, + { + "epoch": 2.76, + "eval_loss": 0.46158257126808167, + "eval_runtime": 9.1346, + "eval_samples_per_second": 10.947, + "eval_wer": 0.4177831912302071, + "step": 17000 + }, + { + "epoch": 2.78, + "learning_rate": 8.9166258769783e-05, + "loss": NaN, + "step": 17100 + }, + { + "epoch": 2.78, + "eval_loss": 0.46586254239082336, + "eval_runtime": 8.0894, + "eval_samples_per_second": 12.362, + "eval_wer": 0.42509135200974424, + "step": 17100 + }, + { + "epoch": 2.8, + "learning_rate": 8.910099526839616e-05, + "loss": NaN, + "step": 17200 + }, + { + "epoch": 2.8, + "eval_loss": 0.42319783568382263, + "eval_runtime": 8.6611, + "eval_samples_per_second": 11.546, + "eval_wer": 0.4104750304506699, + "step": 17200 + }, + { + "epoch": 2.81, + "learning_rate": 8.90357317670093e-05, + "loss": NaN, + "step": 17300 + }, + { + "epoch": 2.81, + "eval_loss": 0.45863696932792664, + "eval_runtime": 8.0216, + "eval_samples_per_second": 12.466, + "eval_wer": 0.4092570036540804, + "step": 17300 + }, + { + "epoch": 2.83, + "learning_rate": 8.897046826562245e-05, + "loss": NaN, + "step": 17400 + }, + { + "epoch": 2.83, + "eval_loss": 0.52719646692276, + "eval_runtime": 7.6618, + "eval_samples_per_second": 13.052, + "eval_wer": 0.45066991473812423, + "step": 17400 + }, + { + "epoch": 2.85, + "learning_rate": 8.89052047642356e-05, + "loss": NaN, + "step": 17500 + }, + { + "epoch": 2.85, + "eval_loss": 0.4623071551322937, + "eval_runtime": 8.3826, + "eval_samples_per_second": 11.929, + "eval_wer": 0.4104750304506699, + "step": 17500 + }, + { + "epoch": 2.86, + "learning_rate": 8.883994126284876e-05, + "loss": NaN, + "step": 17600 + }, + { + "epoch": 2.86, + "eval_loss": 0.47391974925994873, + "eval_runtime": 8.8856, + "eval_samples_per_second": 11.254, + "eval_wer": 0.4092570036540804, + "step": 17600 + }, + { + "epoch": 2.88, + "learning_rate": 8.877467776146192e-05, + "loss": NaN, + "step": 17700 + }, + { + "epoch": 2.88, + "eval_loss": 0.4068062901496887, + "eval_runtime": 7.4887, + "eval_samples_per_second": 13.353, + "eval_wer": 0.392204628501827, + "step": 17700 + }, + { + "epoch": 2.89, + "learning_rate": 8.870941426007505e-05, + "loss": NaN, + "step": 17800 + }, + { + "epoch": 2.89, + "eval_loss": 0.45875948667526245, + "eval_runtime": 8.0538, + "eval_samples_per_second": 12.416, + "eval_wer": 0.42509135200974424, + "step": 17800 + }, + { + "epoch": 2.91, + "learning_rate": 8.864415075868821e-05, + "loss": NaN, + "step": 17900 + }, + { + "epoch": 2.91, + "eval_loss": 0.5207622051239014, + "eval_runtime": 8.8428, + "eval_samples_per_second": 11.309, + "eval_wer": 0.3995127892813642, + "step": 17900 + }, + { + "epoch": 2.93, + "learning_rate": 8.857888725730135e-05, + "loss": NaN, + "step": 18000 + }, + { + "epoch": 2.93, + "eval_loss": 0.4633566737174988, + "eval_runtime": 7.2767, + "eval_samples_per_second": 13.743, + "eval_wer": 0.40803897685749085, + "step": 18000 + }, + { + "epoch": 2.94, + "learning_rate": 8.851362375591451e-05, + "loss": NaN, + "step": 18100 + }, + { + "epoch": 2.94, + "eval_loss": 0.4258524179458618, + "eval_runtime": 7.062, + "eval_samples_per_second": 14.16, + "eval_wer": 0.3848964677222899, + "step": 18100 + }, + { + "epoch": 2.96, + "learning_rate": 8.844836025452766e-05, + "loss": NaN, + "step": 18200 + }, + { + "epoch": 2.96, + "eval_loss": 0.42992159724235535, + "eval_runtime": 7.4964, + "eval_samples_per_second": 13.34, + "eval_wer": 0.4043848964677223, + "step": 18200 + }, + { + "epoch": 2.98, + "learning_rate": 8.838309675314082e-05, + "loss": NaN, + "step": 18300 + }, + { + "epoch": 2.98, + "eval_loss": 0.4557535946369171, + "eval_runtime": 8.3341, + "eval_samples_per_second": 11.999, + "eval_wer": 0.4177831912302071, + "step": 18300 + }, + { + "epoch": 2.99, + "learning_rate": 8.831783325175396e-05, + "loss": NaN, + "step": 18400 + }, + { + "epoch": 2.99, + "eval_loss": 0.4576800465583801, + "eval_runtime": 7.8065, + "eval_samples_per_second": 12.81, + "eval_wer": 0.39707673568818513, + "step": 18400 + }, + { + "epoch": 3.01, + "learning_rate": 8.82525697503671e-05, + "loss": NaN, + "step": 18500 + }, + { + "epoch": 3.01, + "eval_loss": 0.4770311713218689, + "eval_runtime": 8.6894, + "eval_samples_per_second": 11.508, + "eval_wer": 0.4007308160779537, + "step": 18500 + }, + { + "epoch": 3.02, + "learning_rate": 8.818730624898026e-05, + "loss": NaN, + "step": 18600 + }, + { + "epoch": 3.02, + "eval_loss": 0.44065818190574646, + "eval_runtime": 9.5635, + "eval_samples_per_second": 10.456, + "eval_wer": 0.41169305724725946, + "step": 18600 + }, + { + "epoch": 3.04, + "learning_rate": 8.812204274759341e-05, + "loss": NaN, + "step": 18700 + }, + { + "epoch": 3.04, + "eval_loss": 0.4866321086883545, + "eval_runtime": 7.252, + "eval_samples_per_second": 13.789, + "eval_wer": 0.42143727161997563, + "step": 18700 + }, + { + "epoch": 3.06, + "learning_rate": 8.805677924620657e-05, + "loss": NaN, + "step": 18800 + }, + { + "epoch": 3.06, + "eval_loss": 0.4869626760482788, + "eval_runtime": 9.3184, + "eval_samples_per_second": 10.731, + "eval_wer": 0.415347137637028, + "step": 18800 + }, + { + "epoch": 3.07, + "learning_rate": 8.799151574481971e-05, + "loss": NaN, + "step": 18900 + }, + { + "epoch": 3.07, + "eval_loss": 0.580832839012146, + "eval_runtime": 8.0833, + "eval_samples_per_second": 12.371, + "eval_wer": 0.40803897685749085, + "step": 18900 + }, + { + "epoch": 3.09, + "learning_rate": 8.792625224343286e-05, + "loss": NaN, + "step": 19000 + }, + { + "epoch": 3.09, + "eval_loss": 0.5339367389678955, + "eval_runtime": 8.6454, + "eval_samples_per_second": 11.567, + "eval_wer": 0.39707673568818513, + "step": 19000 + }, + { + "epoch": 3.11, + "learning_rate": 8.786098874204602e-05, + "loss": NaN, + "step": 19100 + }, + { + "epoch": 3.11, + "eval_loss": 0.47444137930870056, + "eval_runtime": 8.4893, + "eval_samples_per_second": 11.78, + "eval_wer": 0.40682095006090135, + "step": 19100 + }, + { + "epoch": 3.12, + "learning_rate": 8.779572524065916e-05, + "loss": NaN, + "step": 19200 + }, + { + "epoch": 3.12, + "eval_loss": 0.4808363616466522, + "eval_runtime": 8.229, + "eval_samples_per_second": 12.152, + "eval_wer": 0.42143727161997563, + "step": 19200 + }, + { + "epoch": 3.14, + "learning_rate": 8.773046173927232e-05, + "loss": NaN, + "step": 19300 + }, + { + "epoch": 3.14, + "eval_loss": 0.5487423539161682, + "eval_runtime": 8.6936, + "eval_samples_per_second": 11.503, + "eval_wer": 0.4287454323995128, + "step": 19300 + }, + { + "epoch": 3.15, + "learning_rate": 8.766519823788547e-05, + "loss": NaN, + "step": 19400 + }, + { + "epoch": 3.15, + "eval_loss": 0.5379164218902588, + "eval_runtime": 7.8502, + "eval_samples_per_second": 12.739, + "eval_wer": 0.4238733252131547, + "step": 19400 + }, + { + "epoch": 3.17, + "learning_rate": 8.759993473649861e-05, + "loss": NaN, + "step": 19500 + }, + { + "epoch": 3.17, + "eval_loss": 0.4747236669063568, + "eval_runtime": 6.9979, + "eval_samples_per_second": 14.29, + "eval_wer": 0.40316686967113274, + "step": 19500 + }, + { + "epoch": 3.19, + "learning_rate": 8.753467123511177e-05, + "loss": NaN, + "step": 19600 + }, + { + "epoch": 3.19, + "eval_loss": 0.4517196714878082, + "eval_runtime": 6.8378, + "eval_samples_per_second": 14.624, + "eval_wer": 0.40194884287454324, + "step": 19600 + }, + { + "epoch": 3.2, + "learning_rate": 8.746940773372492e-05, + "loss": NaN, + "step": 19700 + }, + { + "epoch": 3.2, + "eval_loss": 0.473175972700119, + "eval_runtime": 7.9557, + "eval_samples_per_second": 12.57, + "eval_wer": 0.3861144945188794, + "step": 19700 + }, + { + "epoch": 3.22, + "learning_rate": 8.740414423233807e-05, + "loss": NaN, + "step": 19800 + }, + { + "epoch": 3.22, + "eval_loss": 0.4839608371257782, + "eval_runtime": 7.511, + "eval_samples_per_second": 13.314, + "eval_wer": 0.40194884287454324, + "step": 19800 + }, + { + "epoch": 3.24, + "learning_rate": 8.733888073095122e-05, + "loss": NaN, + "step": 19900 + }, + { + "epoch": 3.24, + "eval_loss": 0.45369553565979004, + "eval_runtime": 7.3681, + "eval_samples_per_second": 13.572, + "eval_wer": 0.37758830694275275, + "step": 19900 + }, + { + "epoch": 3.25, + "learning_rate": 8.727361722956438e-05, + "loss": NaN, + "step": 20000 + }, + { + "epoch": 3.25, + "eval_loss": 0.44059714674949646, + "eval_runtime": 8.867, + "eval_samples_per_second": 11.278, + "eval_wer": 0.3861144945188794, + "step": 20000 + }, + { + "epoch": 3.27, + "learning_rate": 8.720835372817752e-05, + "loss": NaN, + "step": 20100 + }, + { + "epoch": 3.27, + "eval_loss": 0.4481971859931946, + "eval_runtime": 7.089, + "eval_samples_per_second": 14.106, + "eval_wer": 0.40194884287454324, + "step": 20100 + }, + { + "epoch": 3.29, + "learning_rate": 8.714309022679067e-05, + "loss": NaN, + "step": 20200 + }, + { + "epoch": 3.29, + "eval_loss": 0.49499496817588806, + "eval_runtime": 9.068, + "eval_samples_per_second": 11.028, + "eval_wer": 0.40194884287454324, + "step": 20200 + }, + { + "epoch": 3.3, + "learning_rate": 8.707782672540383e-05, + "loss": NaN, + "step": 20300 + }, + { + "epoch": 3.3, + "eval_loss": 0.4684819281101227, + "eval_runtime": 8.8297, + "eval_samples_per_second": 11.325, + "eval_wer": 0.3946406820950061, + "step": 20300 + }, + { + "epoch": 3.32, + "learning_rate": 8.701256322401697e-05, + "loss": NaN, + "step": 20400 + }, + { + "epoch": 3.32, + "eval_loss": 0.5416879057884216, + "eval_runtime": 8.4935, + "eval_samples_per_second": 11.774, + "eval_wer": 0.3982947624847747, + "step": 20400 + }, + { + "epoch": 3.33, + "learning_rate": 8.694729972263013e-05, + "loss": NaN, + "step": 20500 + }, + { + "epoch": 3.33, + "eval_loss": 0.49328354001045227, + "eval_runtime": 7.5212, + "eval_samples_per_second": 13.296, + "eval_wer": 0.39585870889159563, + "step": 20500 + }, + { + "epoch": 3.35, + "learning_rate": 8.688203622124328e-05, + "loss": NaN, + "step": 20600 + }, + { + "epoch": 3.35, + "eval_loss": 0.48222553730010986, + "eval_runtime": 9.8089, + "eval_samples_per_second": 10.195, + "eval_wer": 0.4165651644336175, + "step": 20600 + }, + { + "epoch": 3.37, + "learning_rate": 8.681677271985642e-05, + "loss": NaN, + "step": 20700 + }, + { + "epoch": 3.37, + "eval_loss": 0.4871315658092499, + "eval_runtime": 8.0731, + "eval_samples_per_second": 12.387, + "eval_wer": 0.4104750304506699, + "step": 20700 + }, + { + "epoch": 3.38, + "learning_rate": 8.675150921846958e-05, + "loss": NaN, + "step": 20800 + }, + { + "epoch": 3.38, + "eval_loss": 0.4995521605014801, + "eval_runtime": 8.1255, + "eval_samples_per_second": 12.307, + "eval_wer": 0.42021924482338613, + "step": 20800 + }, + { + "epoch": 3.4, + "learning_rate": 8.668624571708272e-05, + "loss": NaN, + "step": 20900 + }, + { + "epoch": 3.4, + "eval_loss": 0.46075180172920227, + "eval_runtime": 7.3922, + "eval_samples_per_second": 13.528, + "eval_wer": 0.41412911084043846, + "step": 20900 + }, + { + "epoch": 3.42, + "learning_rate": 8.662098221569588e-05, + "loss": NaN, + "step": 21000 + }, + { + "epoch": 3.42, + "eval_loss": 0.4704657793045044, + "eval_runtime": 8.87, + "eval_samples_per_second": 11.274, + "eval_wer": 0.40316686967113274, + "step": 21000 + }, + { + "epoch": 3.43, + "learning_rate": 8.655571871430903e-05, + "loss": NaN, + "step": 21100 + }, + { + "epoch": 3.43, + "eval_loss": 0.44366922974586487, + "eval_runtime": 8.0362, + "eval_samples_per_second": 12.444, + "eval_wer": 0.3982947624847747, + "step": 21100 + }, + { + "epoch": 3.45, + "learning_rate": 8.649045521292217e-05, + "loss": NaN, + "step": 21200 + }, + { + "epoch": 3.45, + "eval_loss": 0.5748487710952759, + "eval_runtime": 8.4229, + "eval_samples_per_second": 11.872, + "eval_wer": 0.4226552984165652, + "step": 21200 + }, + { + "epoch": 3.46, + "learning_rate": 8.642519171153532e-05, + "loss": NaN, + "step": 21300 + }, + { + "epoch": 3.46, + "eval_loss": 0.4763486385345459, + "eval_runtime": 7.3956, + "eval_samples_per_second": 13.522, + "eval_wer": 0.3800243605359318, + "step": 21300 + }, + { + "epoch": 3.48, + "learning_rate": 8.635992821014848e-05, + "loss": NaN, + "step": 21400 + }, + { + "epoch": 3.48, + "eval_loss": 0.45793360471725464, + "eval_runtime": 7.217, + "eval_samples_per_second": 13.856, + "eval_wer": 0.4104750304506699, + "step": 21400 + }, + { + "epoch": 3.5, + "learning_rate": 8.629466470876164e-05, + "loss": NaN, + "step": 21500 + }, + { + "epoch": 3.5, + "eval_loss": 0.4132789373397827, + "eval_runtime": 9.7025, + "eval_samples_per_second": 10.307, + "eval_wer": 0.4056029232643118, + "step": 21500 + }, + { + "epoch": 3.51, + "learning_rate": 8.622940120737478e-05, + "loss": NaN, + "step": 21600 + }, + { + "epoch": 3.51, + "eval_loss": 0.4666365385055542, + "eval_runtime": 9.4541, + "eval_samples_per_second": 10.577, + "eval_wer": 0.40682095006090135, + "step": 21600 + }, + { + "epoch": 3.53, + "learning_rate": 8.616413770598794e-05, + "loss": NaN, + "step": 21700 + }, + { + "epoch": 3.53, + "eval_loss": 0.43131592869758606, + "eval_runtime": 7.5373, + "eval_samples_per_second": 13.267, + "eval_wer": 0.37758830694275275, + "step": 21700 + }, + { + "epoch": 3.55, + "learning_rate": 8.609887420460107e-05, + "loss": NaN, + "step": 21800 + }, + { + "epoch": 3.55, + "eval_loss": 0.4624130129814148, + "eval_runtime": 8.6512, + "eval_samples_per_second": 11.559, + "eval_wer": 0.3934226552984166, + "step": 21800 + }, + { + "epoch": 3.56, + "learning_rate": 8.603361070321423e-05, + "loss": NaN, + "step": 21900 + }, + { + "epoch": 3.56, + "eval_loss": 0.47029244899749756, + "eval_runtime": 7.0891, + "eval_samples_per_second": 14.106, + "eval_wer": 0.39707673568818513, + "step": 21900 + }, + { + "epoch": 3.58, + "learning_rate": 8.596834720182738e-05, + "loss": NaN, + "step": 22000 + }, + { + "epoch": 3.58, + "eval_loss": 0.4115927219390869, + "eval_runtime": 7.9478, + "eval_samples_per_second": 12.582, + "eval_wer": 0.38855054811205847, + "step": 22000 + }, + { + "epoch": 3.59, + "learning_rate": 8.590308370044053e-05, + "loss": NaN, + "step": 22100 + }, + { + "epoch": 3.59, + "eval_loss": 0.4755612313747406, + "eval_runtime": 8.2831, + "eval_samples_per_second": 12.073, + "eval_wer": 0.3848964677222899, + "step": 22100 + }, + { + "epoch": 3.61, + "learning_rate": 8.583782019905369e-05, + "loss": NaN, + "step": 22200 + }, + { + "epoch": 3.61, + "eval_loss": 0.42175978422164917, + "eval_runtime": 10.3113, + "eval_samples_per_second": 9.698, + "eval_wer": 0.38855054811205847, + "step": 22200 + }, + { + "epoch": 3.63, + "learning_rate": 8.577255669766684e-05, + "loss": NaN, + "step": 22300 + }, + { + "epoch": 3.63, + "eval_loss": 0.4504680633544922, + "eval_runtime": 8.4957, + "eval_samples_per_second": 11.771, + "eval_wer": 0.40316686967113274, + "step": 22300 + }, + { + "epoch": 3.64, + "learning_rate": 8.570729319627998e-05, + "loss": NaN, + "step": 22400 + }, + { + "epoch": 3.64, + "eval_loss": 0.4168296754360199, + "eval_runtime": 7.4037, + "eval_samples_per_second": 13.507, + "eval_wer": 0.40194884287454324, + "step": 22400 + }, + { + "epoch": 3.66, + "learning_rate": 8.564202969489313e-05, + "loss": NaN, + "step": 22500 + }, + { + "epoch": 3.66, + "eval_loss": 0.3921906650066376, + "eval_runtime": 7.4162, + "eval_samples_per_second": 13.484, + "eval_wer": 0.37271619975639464, + "step": 22500 + }, + { + "epoch": 3.68, + "learning_rate": 8.557676619350629e-05, + "loss": NaN, + "step": 22600 + }, + { + "epoch": 3.68, + "eval_loss": 0.4271007478237152, + "eval_runtime": 8.8145, + "eval_samples_per_second": 11.345, + "eval_wer": 0.38855054811205847, + "step": 22600 + }, + { + "epoch": 3.69, + "learning_rate": 8.551150269211943e-05, + "loss": NaN, + "step": 22700 + }, + { + "epoch": 3.69, + "eval_loss": 0.416711688041687, + "eval_runtime": 8.4178, + "eval_samples_per_second": 11.88, + "eval_wer": 0.40682095006090135, + "step": 22700 + }, + { + "epoch": 3.71, + "learning_rate": 8.544623919073259e-05, + "loss": NaN, + "step": 22800 + }, + { + "epoch": 3.71, + "eval_loss": 0.42215126752853394, + "eval_runtime": 7.8605, + "eval_samples_per_second": 12.722, + "eval_wer": 0.392204628501827, + "step": 22800 + }, + { + "epoch": 3.72, + "learning_rate": 8.538097568934574e-05, + "loss": NaN, + "step": 22900 + }, + { + "epoch": 3.72, + "eval_loss": 0.4496414065361023, + "eval_runtime": 7.9977, + "eval_samples_per_second": 12.504, + "eval_wer": 0.4177831912302071, + "step": 22900 + }, + { + "epoch": 3.74, + "learning_rate": 8.531571218795888e-05, + "loss": NaN, + "step": 23000 + }, + { + "epoch": 3.74, + "eval_loss": 0.4615870714187622, + "eval_runtime": 8.1022, + "eval_samples_per_second": 12.342, + "eval_wer": 0.38855054811205847, + "step": 23000 + }, + { + "epoch": 3.76, + "learning_rate": 8.525044868657204e-05, + "loss": NaN, + "step": 23100 + }, + { + "epoch": 3.76, + "eval_loss": 0.4659249186515808, + "eval_runtime": 8.1138, + "eval_samples_per_second": 12.325, + "eval_wer": 0.4007308160779537, + "step": 23100 + }, + { + "epoch": 3.77, + "learning_rate": 8.518518518518518e-05, + "loss": NaN, + "step": 23200 + }, + { + "epoch": 3.77, + "eval_loss": 0.45942869782447815, + "eval_runtime": 7.7263, + "eval_samples_per_second": 12.943, + "eval_wer": 0.36784409257003653, + "step": 23200 + }, + { + "epoch": 3.79, + "learning_rate": 8.511992168379834e-05, + "loss": NaN, + "step": 23300 + }, + { + "epoch": 3.79, + "eval_loss": 0.4240889847278595, + "eval_runtime": 7.2849, + "eval_samples_per_second": 13.727, + "eval_wer": 0.37880633373934225, + "step": 23300 + }, + { + "epoch": 3.81, + "learning_rate": 8.505465818241149e-05, + "loss": NaN, + "step": 23400 + }, + { + "epoch": 3.81, + "eval_loss": 0.4439373314380646, + "eval_runtime": 7.5713, + "eval_samples_per_second": 13.208, + "eval_wer": 0.39707673568818513, + "step": 23400 + }, + { + "epoch": 3.82, + "learning_rate": 8.498939468102463e-05, + "loss": NaN, + "step": 23500 + }, + { + "epoch": 3.82, + "eval_loss": 0.4574301838874817, + "eval_runtime": 7.723, + "eval_samples_per_second": 12.948, + "eval_wer": 0.39707673568818513, + "step": 23500 + }, + { + "epoch": 3.84, + "learning_rate": 8.492413117963779e-05, + "loss": NaN, + "step": 23600 + }, + { + "epoch": 3.84, + "eval_loss": 0.41990235447883606, + "eval_runtime": 7.2988, + "eval_samples_per_second": 13.701, + "eval_wer": 0.38246041412911086, + "step": 23600 + }, + { + "epoch": 3.85, + "learning_rate": 8.485886767825094e-05, + "loss": NaN, + "step": 23700 + }, + { + "epoch": 3.85, + "eval_loss": 0.4570240080356598, + "eval_runtime": 9.4247, + "eval_samples_per_second": 10.61, + "eval_wer": 0.3934226552984166, + "step": 23700 + }, + { + "epoch": 3.87, + "learning_rate": 8.47936041768641e-05, + "loss": NaN, + "step": 23800 + }, + { + "epoch": 3.87, + "eval_loss": 0.42279544472694397, + "eval_runtime": 7.429, + "eval_samples_per_second": 13.461, + "eval_wer": 0.37880633373934225, + "step": 23800 + }, + { + "epoch": 3.89, + "learning_rate": 8.472834067547724e-05, + "loss": NaN, + "step": 23900 + }, + { + "epoch": 3.89, + "eval_loss": 0.4250124394893646, + "eval_runtime": 8.7078, + "eval_samples_per_second": 11.484, + "eval_wer": 0.39585870889159563, + "step": 23900 + }, + { + "epoch": 3.9, + "learning_rate": 8.46630771740904e-05, + "loss": NaN, + "step": 24000 + }, + { + "epoch": 3.9, + "eval_loss": 0.48117172718048096, + "eval_runtime": 7.6756, + "eval_samples_per_second": 13.028, + "eval_wer": 0.40803897685749085, + "step": 24000 + }, + { + "epoch": 3.92, + "learning_rate": 8.459781367270355e-05, + "loss": NaN, + "step": 24100 + }, + { + "epoch": 3.92, + "eval_loss": 0.4824243485927582, + "eval_runtime": 9.2626, + "eval_samples_per_second": 10.796, + "eval_wer": 0.3812423873325213, + "step": 24100 + }, + { + "epoch": 3.94, + "learning_rate": 8.453255017131669e-05, + "loss": NaN, + "step": 24200 + }, + { + "epoch": 3.94, + "eval_loss": 0.45317116379737854, + "eval_runtime": 7.8742, + "eval_samples_per_second": 12.7, + "eval_wer": 0.39585870889159563, + "step": 24200 + }, + { + "epoch": 3.95, + "learning_rate": 8.446728666992985e-05, + "loss": NaN, + "step": 24300 + }, + { + "epoch": 3.95, + "eval_loss": 0.5323064923286438, + "eval_runtime": 7.1451, + "eval_samples_per_second": 13.996, + "eval_wer": 0.42509135200974424, + "step": 24300 + }, + { + "epoch": 3.97, + "learning_rate": 8.4402023168543e-05, + "loss": NaN, + "step": 24400 + }, + { + "epoch": 3.97, + "eval_loss": 0.5017148852348328, + "eval_runtime": 8.2004, + "eval_samples_per_second": 12.194, + "eval_wer": 0.40803897685749085, + "step": 24400 + }, + { + "epoch": 3.98, + "learning_rate": 8.433675966715615e-05, + "loss": NaN, + "step": 24500 + }, + { + "epoch": 3.98, + "eval_loss": 0.4483763873577118, + "eval_runtime": 7.3109, + "eval_samples_per_second": 13.678, + "eval_wer": 0.38246041412911086, + "step": 24500 + }, + { + "epoch": 4.0, + "learning_rate": 8.42714961657693e-05, + "loss": NaN, + "step": 24600 + }, + { + "epoch": 4.0, + "eval_loss": 0.4591113030910492, + "eval_runtime": 7.7451, + "eval_samples_per_second": 12.911, + "eval_wer": 0.3800243605359318, + "step": 24600 + }, + { + "epoch": 4.02, + "learning_rate": 8.420623266438244e-05, + "loss": NaN, + "step": 24700 + }, + { + "epoch": 4.02, + "eval_loss": 0.45395427942276, + "eval_runtime": 7.718, + "eval_samples_per_second": 12.957, + "eval_wer": 0.3763702801461632, + "step": 24700 + }, + { + "epoch": 4.03, + "learning_rate": 8.41409691629956e-05, + "loss": NaN, + "step": 24800 + }, + { + "epoch": 4.03, + "eval_loss": 0.5075406432151794, + "eval_runtime": 8.7905, + "eval_samples_per_second": 11.376, + "eval_wer": 0.40803897685749085, + "step": 24800 + }, + { + "epoch": 4.05, + "learning_rate": 8.407570566160875e-05, + "loss": NaN, + "step": 24900 + }, + { + "epoch": 4.05, + "eval_loss": 0.5291510224342346, + "eval_runtime": 8.3479, + "eval_samples_per_second": 11.979, + "eval_wer": 0.4104750304506699, + "step": 24900 + }, + { + "epoch": 4.07, + "learning_rate": 8.40104421602219e-05, + "loss": NaN, + "step": 25000 + }, + { + "epoch": 4.07, + "eval_loss": 0.4835575819015503, + "eval_runtime": 9.1852, + "eval_samples_per_second": 10.887, + "eval_wer": 0.3934226552984166, + "step": 25000 + }, + { + "epoch": 4.08, + "learning_rate": 8.394517865883505e-05, + "loss": NaN, + "step": 25100 + }, + { + "epoch": 4.08, + "eval_loss": 0.4627467691898346, + "eval_runtime": 8.3255, + "eval_samples_per_second": 12.011, + "eval_wer": 0.40803897685749085, + "step": 25100 + }, + { + "epoch": 4.1, + "learning_rate": 8.38799151574482e-05, + "loss": NaN, + "step": 25200 + }, + { + "epoch": 4.1, + "eval_loss": 0.45444557070732117, + "eval_runtime": 7.7259, + "eval_samples_per_second": 12.943, + "eval_wer": 0.40316686967113274, + "step": 25200 + }, + { + "epoch": 4.11, + "learning_rate": 8.381465165606135e-05, + "loss": NaN, + "step": 25300 + }, + { + "epoch": 4.11, + "eval_loss": 0.4336519241333008, + "eval_runtime": 7.905, + "eval_samples_per_second": 12.65, + "eval_wer": 0.3873325213154689, + "step": 25300 + }, + { + "epoch": 4.13, + "learning_rate": 8.37493881546745e-05, + "loss": NaN, + "step": 25400 + }, + { + "epoch": 4.13, + "eval_loss": 0.44815555214881897, + "eval_runtime": 8.3801, + "eval_samples_per_second": 11.933, + "eval_wer": 0.4104750304506699, + "step": 25400 + }, + { + "epoch": 4.15, + "learning_rate": 8.368412465328766e-05, + "loss": NaN, + "step": 25500 + }, + { + "epoch": 4.15, + "eval_loss": 0.45955729484558105, + "eval_runtime": 7.7863, + "eval_samples_per_second": 12.843, + "eval_wer": 0.38855054811205847, + "step": 25500 + }, + { + "epoch": 4.16, + "learning_rate": 8.36188611519008e-05, + "loss": NaN, + "step": 25600 + }, + { + "epoch": 4.16, + "eval_loss": 0.47062352299690247, + "eval_runtime": 9.5871, + "eval_samples_per_second": 10.431, + "eval_wer": 0.39707673568818513, + "step": 25600 + }, + { + "epoch": 4.18, + "learning_rate": 8.355359765051396e-05, + "loss": NaN, + "step": 25700 + }, + { + "epoch": 4.18, + "eval_loss": 0.44303008913993835, + "eval_runtime": 8.2157, + "eval_samples_per_second": 12.172, + "eval_wer": 0.3995127892813642, + "step": 25700 + }, + { + "epoch": 4.2, + "learning_rate": 8.348833414912711e-05, + "loss": NaN, + "step": 25800 + }, + { + "epoch": 4.2, + "eval_loss": 0.527855396270752, + "eval_runtime": 8.507, + "eval_samples_per_second": 11.755, + "eval_wer": 0.4104750304506699, + "step": 25800 + }, + { + "epoch": 4.21, + "learning_rate": 8.342307064774025e-05, + "loss": NaN, + "step": 25900 + }, + { + "epoch": 4.21, + "eval_loss": 0.4706363379955292, + "eval_runtime": 6.3997, + "eval_samples_per_second": 15.626, + "eval_wer": 0.4007308160779537, + "step": 25900 + }, + { + "epoch": 4.23, + "learning_rate": 8.335780714635341e-05, + "loss": NaN, + "step": 26000 + }, + { + "epoch": 4.23, + "eval_loss": 0.46070727705955505, + "eval_runtime": 6.6071, + "eval_samples_per_second": 15.135, + "eval_wer": 0.37880633373934225, + "step": 26000 + }, + { + "epoch": 4.24, + "learning_rate": 8.329254364496656e-05, + "loss": NaN, + "step": 26100 + }, + { + "epoch": 4.24, + "eval_loss": 0.47955620288848877, + "eval_runtime": 8.053, + "eval_samples_per_second": 12.418, + "eval_wer": 0.3812423873325213, + "step": 26100 + }, + { + "epoch": 4.26, + "learning_rate": 8.322728014357972e-05, + "loss": NaN, + "step": 26200 + }, + { + "epoch": 4.26, + "eval_loss": 0.4731149971485138, + "eval_runtime": 8.8267, + "eval_samples_per_second": 11.329, + "eval_wer": 0.40682095006090135, + "step": 26200 + }, + { + "epoch": 4.28, + "learning_rate": 8.316201664219285e-05, + "loss": NaN, + "step": 26300 + }, + { + "epoch": 4.28, + "eval_loss": 0.4844764471054077, + "eval_runtime": 7.374, + "eval_samples_per_second": 13.561, + "eval_wer": 0.39585870889159563, + "step": 26300 + }, + { + "epoch": 4.29, + "learning_rate": 8.3096753140806e-05, + "loss": NaN, + "step": 26400 + }, + { + "epoch": 4.29, + "eval_loss": 0.4396247863769531, + "eval_runtime": 9.6514, + "eval_samples_per_second": 10.361, + "eval_wer": 0.3751522533495737, + "step": 26400 + }, + { + "epoch": 4.31, + "learning_rate": 8.303148963941916e-05, + "loss": NaN, + "step": 26500 + }, + { + "epoch": 4.31, + "eval_loss": 0.4746050238609314, + "eval_runtime": 8.1321, + "eval_samples_per_second": 12.297, + "eval_wer": 0.38246041412911086, + "step": 26500 + }, + { + "epoch": 4.33, + "learning_rate": 8.296622613803231e-05, + "loss": NaN, + "step": 26600 + }, + { + "epoch": 4.33, + "eval_loss": 0.4225209951400757, + "eval_runtime": 7.7046, + "eval_samples_per_second": 12.979, + "eval_wer": 0.3702801461632156, + "step": 26600 + }, + { + "epoch": 4.34, + "learning_rate": 8.290096263664547e-05, + "loss": NaN, + "step": 26700 + }, + { + "epoch": 4.34, + "eval_loss": 0.46304523944854736, + "eval_runtime": 8.8314, + "eval_samples_per_second": 11.323, + "eval_wer": 0.40194884287454324, + "step": 26700 + }, + { + "epoch": 4.36, + "learning_rate": 8.283569913525861e-05, + "loss": NaN, + "step": 26800 + }, + { + "epoch": 4.36, + "eval_loss": 0.45539891719818115, + "eval_runtime": 8.3892, + "eval_samples_per_second": 11.92, + "eval_wer": 0.3848964677222899, + "step": 26800 + }, + { + "epoch": 4.37, + "learning_rate": 8.277043563387176e-05, + "loss": NaN, + "step": 26900 + }, + { + "epoch": 4.37, + "eval_loss": 0.4747115671634674, + "eval_runtime": 8.0532, + "eval_samples_per_second": 12.417, + "eval_wer": 0.3861144945188794, + "step": 26900 + }, + { + "epoch": 4.39, + "learning_rate": 8.27051721324849e-05, + "loss": NaN, + "step": 27000 + }, + { + "epoch": 4.39, + "eval_loss": 0.4937517046928406, + "eval_runtime": 8.6063, + "eval_samples_per_second": 11.619, + "eval_wer": 0.3982947624847747, + "step": 27000 + }, + { + "epoch": 4.41, + "learning_rate": 8.263990863109806e-05, + "loss": NaN, + "step": 27100 + }, + { + "epoch": 4.41, + "eval_loss": 0.5522226095199585, + "eval_runtime": 7.3096, + "eval_samples_per_second": 13.681, + "eval_wer": 0.38855054811205847, + "step": 27100 + }, + { + "epoch": 4.42, + "learning_rate": 8.257464512971122e-05, + "loss": NaN, + "step": 27200 + }, + { + "epoch": 4.42, + "eval_loss": 0.48202434182167053, + "eval_runtime": 7.8854, + "eval_samples_per_second": 12.682, + "eval_wer": 0.4056029232643118, + "step": 27200 + }, + { + "epoch": 4.44, + "learning_rate": 8.250938162832437e-05, + "loss": NaN, + "step": 27300 + }, + { + "epoch": 4.44, + "eval_loss": 0.4578789472579956, + "eval_runtime": 8.752, + "eval_samples_per_second": 11.426, + "eval_wer": 0.37393422655298414, + "step": 27300 + }, + { + "epoch": 4.46, + "learning_rate": 8.244411812693752e-05, + "loss": NaN, + "step": 27400 + }, + { + "epoch": 4.46, + "eval_loss": 0.44214507937431335, + "eval_runtime": 10.4981, + "eval_samples_per_second": 9.526, + "eval_wer": 0.3946406820950061, + "step": 27400 + }, + { + "epoch": 4.47, + "learning_rate": 8.237885462555066e-05, + "loss": NaN, + "step": 27500 + }, + { + "epoch": 4.47, + "eval_loss": 0.43562430143356323, + "eval_runtime": 7.8819, + "eval_samples_per_second": 12.687, + "eval_wer": 0.3848964677222899, + "step": 27500 + }, + { + "epoch": 4.49, + "learning_rate": 8.231359112416381e-05, + "loss": NaN, + "step": 27600 + }, + { + "epoch": 4.49, + "eval_loss": 0.5293126106262207, + "eval_runtime": 9.3755, + "eval_samples_per_second": 10.666, + "eval_wer": 0.3982947624847747, + "step": 27600 + }, + { + "epoch": 4.5, + "learning_rate": 8.224832762277696e-05, + "loss": NaN, + "step": 27700 + }, + { + "epoch": 4.5, + "eval_loss": 0.4340215027332306, + "eval_runtime": 8.0204, + "eval_samples_per_second": 12.468, + "eval_wer": 0.38855054811205847, + "step": 27700 + }, + { + "epoch": 4.52, + "learning_rate": 8.218306412139012e-05, + "loss": NaN, + "step": 27800 + }, + { + "epoch": 4.52, + "eval_loss": 0.4252225160598755, + "eval_runtime": 8.021, + "eval_samples_per_second": 12.467, + "eval_wer": 0.3763702801461632, + "step": 27800 + }, + { + "epoch": 4.54, + "learning_rate": 8.211780062000328e-05, + "loss": NaN, + "step": 27900 + }, + { + "epoch": 4.54, + "eval_loss": 0.43396663665771484, + "eval_runtime": 7.1016, + "eval_samples_per_second": 14.081, + "eval_wer": 0.38855054811205847, + "step": 27900 + }, + { + "epoch": 4.55, + "learning_rate": 8.205253711861641e-05, + "loss": NaN, + "step": 28000 + }, + { + "epoch": 4.55, + "eval_loss": 0.4182353615760803, + "eval_runtime": 9.2514, + "eval_samples_per_second": 10.809, + "eval_wer": 0.3800243605359318, + "step": 28000 + }, + { + "epoch": 4.57, + "learning_rate": 8.198727361722957e-05, + "loss": NaN, + "step": 28100 + }, + { + "epoch": 4.57, + "eval_loss": 0.5017396211624146, + "eval_runtime": 8.9953, + "eval_samples_per_second": 11.117, + "eval_wer": 0.40316686967113274, + "step": 28100 + }, + { + "epoch": 4.59, + "learning_rate": 8.192201011584271e-05, + "loss": NaN, + "step": 28200 + }, + { + "epoch": 4.59, + "eval_loss": 0.4595078229904175, + "eval_runtime": 9.3616, + "eval_samples_per_second": 10.682, + "eval_wer": 0.3909866017052375, + "step": 28200 + }, + { + "epoch": 4.6, + "learning_rate": 8.185674661445587e-05, + "loss": NaN, + "step": 28300 + }, + { + "epoch": 4.6, + "eval_loss": 0.44491440057754517, + "eval_runtime": 10.1045, + "eval_samples_per_second": 9.897, + "eval_wer": 0.3946406820950061, + "step": 28300 + }, + { + "epoch": 4.62, + "learning_rate": 8.179148311306902e-05, + "loss": NaN, + "step": 28400 + }, + { + "epoch": 4.62, + "eval_loss": 0.4460849463939667, + "eval_runtime": 9.8214, + "eval_samples_per_second": 10.182, + "eval_wer": 0.3909866017052375, + "step": 28400 + }, + { + "epoch": 4.63, + "learning_rate": 8.172621961168218e-05, + "loss": NaN, + "step": 28500 + }, + { + "epoch": 4.63, + "eval_loss": 0.49524936079978943, + "eval_runtime": 7.8304, + "eval_samples_per_second": 12.771, + "eval_wer": 0.40194884287454324, + "step": 28500 + }, + { + "epoch": 4.65, + "learning_rate": 8.166095611029532e-05, + "loss": NaN, + "step": 28600 + }, + { + "epoch": 4.65, + "eval_loss": 0.481218159198761, + "eval_runtime": 8.0581, + "eval_samples_per_second": 12.41, + "eval_wer": 0.4043848964677223, + "step": 28600 + }, + { + "epoch": 4.67, + "learning_rate": 8.159569260890847e-05, + "loss": NaN, + "step": 28700 + }, + { + "epoch": 4.67, + "eval_loss": 0.4491060674190521, + "eval_runtime": 7.4142, + "eval_samples_per_second": 13.488, + "eval_wer": 0.3909866017052375, + "step": 28700 + }, + { + "epoch": 4.68, + "learning_rate": 8.153042910752162e-05, + "loss": NaN, + "step": 28800 + }, + { + "epoch": 4.68, + "eval_loss": 0.5148893594741821, + "eval_runtime": 6.7875, + "eval_samples_per_second": 14.733, + "eval_wer": 0.3946406820950061, + "step": 28800 + }, + { + "epoch": 4.7, + "learning_rate": 8.146516560613477e-05, + "loss": NaN, + "step": 28900 + }, + { + "epoch": 4.7, + "eval_loss": 0.46138709783554077, + "eval_runtime": 7.3893, + "eval_samples_per_second": 13.533, + "eval_wer": 0.4007308160779537, + "step": 28900 + }, + { + "epoch": 4.72, + "learning_rate": 8.139990210474793e-05, + "loss": NaN, + "step": 29000 + }, + { + "epoch": 4.72, + "eval_loss": 0.44223907589912415, + "eval_runtime": 8.6011, + "eval_samples_per_second": 11.626, + "eval_wer": 0.3873325213154689, + "step": 29000 + }, + { + "epoch": 4.73, + "learning_rate": 8.133463860336107e-05, + "loss": NaN, + "step": 29100 + }, + { + "epoch": 4.73, + "eval_loss": 0.471282958984375, + "eval_runtime": 10.4884, + "eval_samples_per_second": 9.534, + "eval_wer": 0.40803897685749085, + "step": 29100 + }, + { + "epoch": 4.75, + "learning_rate": 8.126937510197422e-05, + "loss": NaN, + "step": 29200 + }, + { + "epoch": 4.75, + "eval_loss": 0.43840956687927246, + "eval_runtime": 8.0732, + "eval_samples_per_second": 12.387, + "eval_wer": 0.41412911084043846, + "step": 29200 + }, + { + "epoch": 4.77, + "learning_rate": 8.120411160058738e-05, + "loss": NaN, + "step": 29300 + }, + { + "epoch": 4.77, + "eval_loss": 0.4980277717113495, + "eval_runtime": 8.0117, + "eval_samples_per_second": 12.482, + "eval_wer": 0.4007308160779537, + "step": 29300 + }, + { + "epoch": 4.78, + "learning_rate": 8.113884809920052e-05, + "loss": NaN, + "step": 29400 + }, + { + "epoch": 4.78, + "eval_loss": 0.46275556087493896, + "eval_runtime": 8.4323, + "eval_samples_per_second": 11.859, + "eval_wer": 0.41169305724725946, + "step": 29400 + }, + { + "epoch": 4.8, + "learning_rate": 8.107358459781368e-05, + "loss": NaN, + "step": 29500 + }, + { + "epoch": 4.8, + "eval_loss": 0.5011305809020996, + "eval_runtime": 9.4028, + "eval_samples_per_second": 10.635, + "eval_wer": 0.39707673568818513, + "step": 29500 + }, + { + "epoch": 4.81, + "learning_rate": 8.100832109642683e-05, + "loss": NaN, + "step": 29600 + }, + { + "epoch": 4.81, + "eval_loss": 0.4675903022289276, + "eval_runtime": 8.4503, + "eval_samples_per_second": 11.834, + "eval_wer": 0.3861144945188794, + "step": 29600 + }, + { + "epoch": 4.83, + "learning_rate": 8.094305759503997e-05, + "loss": NaN, + "step": 29700 + }, + { + "epoch": 4.83, + "eval_loss": 0.4845035970211029, + "eval_runtime": 10.1458, + "eval_samples_per_second": 9.856, + "eval_wer": 0.40316686967113274, + "step": 29700 + }, + { + "epoch": 4.85, + "learning_rate": 8.087779409365313e-05, + "loss": NaN, + "step": 29800 + }, + { + "epoch": 4.85, + "eval_loss": 0.47924065589904785, + "eval_runtime": 9.6324, + "eval_samples_per_second": 10.382, + "eval_wer": 0.41169305724725946, + "step": 29800 + }, + { + "epoch": 4.86, + "learning_rate": 8.081253059226628e-05, + "loss": NaN, + "step": 29900 + }, + { + "epoch": 4.86, + "eval_loss": 0.4541124701499939, + "eval_runtime": 7.4256, + "eval_samples_per_second": 13.467, + "eval_wer": 0.4007308160779537, + "step": 29900 + }, + { + "epoch": 4.88, + "learning_rate": 8.074726709087943e-05, + "loss": NaN, + "step": 30000 + }, + { + "epoch": 4.88, + "eval_loss": 0.4285271167755127, + "eval_runtime": 6.7649, + "eval_samples_per_second": 14.782, + "eval_wer": 0.4056029232643118, + "step": 30000 + }, + { + "epoch": 4.9, + "learning_rate": 8.068200358949258e-05, + "loss": NaN, + "step": 30100 + }, + { + "epoch": 4.9, + "eval_loss": 0.4844328761100769, + "eval_runtime": 9.5693, + "eval_samples_per_second": 10.45, + "eval_wer": 0.3934226552984166, + "step": 30100 + }, + { + "epoch": 4.91, + "learning_rate": 8.061674008810574e-05, + "loss": NaN, + "step": 30200 + }, + { + "epoch": 4.91, + "eval_loss": 0.4602464735507965, + "eval_runtime": 10.2251, + "eval_samples_per_second": 9.78, + "eval_wer": 0.3909866017052375, + "step": 30200 + }, + { + "epoch": 4.93, + "learning_rate": 8.055147658671888e-05, + "loss": NaN, + "step": 30300 + }, + { + "epoch": 4.93, + "eval_loss": 0.431302934885025, + "eval_runtime": 8.621, + "eval_samples_per_second": 11.6, + "eval_wer": 0.3982947624847747, + "step": 30300 + }, + { + "epoch": 4.94, + "learning_rate": 8.048621308533203e-05, + "loss": NaN, + "step": 30400 + }, + { + "epoch": 4.94, + "eval_loss": 0.4665343463420868, + "eval_runtime": 8.6252, + "eval_samples_per_second": 11.594, + "eval_wer": 0.4092570036540804, + "step": 30400 + }, + { + "epoch": 4.96, + "learning_rate": 8.042094958394519e-05, + "loss": NaN, + "step": 30500 + }, + { + "epoch": 4.96, + "eval_loss": 0.4470652639865875, + "eval_runtime": 8.3879, + "eval_samples_per_second": 11.922, + "eval_wer": 0.3946406820950061, + "step": 30500 + }, + { + "epoch": 4.98, + "learning_rate": 8.035568608255833e-05, + "loss": NaN, + "step": 30600 + }, + { + "epoch": 4.98, + "eval_loss": 0.4649478495121002, + "eval_runtime": 8.0029, + "eval_samples_per_second": 12.495, + "eval_wer": 0.3934226552984166, + "step": 30600 + }, + { + "epoch": 4.99, + "learning_rate": 8.029042258117149e-05, + "loss": NaN, + "step": 30700 + }, + { + "epoch": 4.99, + "eval_loss": 0.42199084162712097, + "eval_runtime": 7.8648, + "eval_samples_per_second": 12.715, + "eval_wer": 0.38367844092570036, + "step": 30700 + }, + { + "epoch": 5.01, + "learning_rate": 8.022515907978464e-05, + "loss": NaN, + "step": 30800 + }, + { + "epoch": 5.01, + "eval_loss": 0.43867814540863037, + "eval_runtime": 10.0022, + "eval_samples_per_second": 9.998, + "eval_wer": 0.4092570036540804, + "step": 30800 + }, + { + "epoch": 5.03, + "learning_rate": 8.015989557839778e-05, + "loss": NaN, + "step": 30900 + }, + { + "epoch": 5.03, + "eval_loss": 0.4615475535392761, + "eval_runtime": 7.5843, + "eval_samples_per_second": 13.185, + "eval_wer": 0.4104750304506699, + "step": 30900 + }, + { + "epoch": 5.04, + "learning_rate": 8.009463207701094e-05, + "loss": NaN, + "step": 31000 + }, + { + "epoch": 5.04, + "eval_loss": 0.4595627188682556, + "eval_runtime": 8.6951, + "eval_samples_per_second": 11.501, + "eval_wer": 0.41169305724725946, + "step": 31000 + }, + { + "epoch": 5.06, + "learning_rate": 8.002936857562408e-05, + "loss": NaN, + "step": 31100 + }, + { + "epoch": 5.06, + "eval_loss": 0.48309993743896484, + "eval_runtime": 7.7623, + "eval_samples_per_second": 12.883, + "eval_wer": 0.41169305724725946, + "step": 31100 + }, + { + "epoch": 5.07, + "learning_rate": 7.996410507423724e-05, + "loss": NaN, + "step": 31200 + }, + { + "epoch": 5.07, + "eval_loss": 0.4405969977378845, + "eval_runtime": 8.6906, + "eval_samples_per_second": 11.507, + "eval_wer": 0.392204628501827, + "step": 31200 + }, + { + "epoch": 5.09, + "learning_rate": 7.989884157285039e-05, + "loss": NaN, + "step": 31300 + }, + { + "epoch": 5.09, + "eval_loss": 0.46763721108436584, + "eval_runtime": 8.1484, + "eval_samples_per_second": 12.272, + "eval_wer": 0.3909866017052375, + "step": 31300 + }, + { + "epoch": 5.11, + "learning_rate": 7.983357807146353e-05, + "loss": NaN, + "step": 31400 + }, + { + "epoch": 5.11, + "eval_loss": 0.4569970369338989, + "eval_runtime": 8.3373, + "eval_samples_per_second": 11.994, + "eval_wer": 0.41169305724725946, + "step": 31400 + }, + { + "epoch": 5.12, + "learning_rate": 7.976831457007669e-05, + "loss": NaN, + "step": 31500 + }, + { + "epoch": 5.12, + "eval_loss": 0.4363997280597687, + "eval_runtime": 8.1102, + "eval_samples_per_second": 12.33, + "eval_wer": 0.3848964677222899, + "step": 31500 + }, + { + "epoch": 5.14, + "learning_rate": 7.970305106868984e-05, + "loss": NaN, + "step": 31600 + }, + { + "epoch": 5.14, + "eval_loss": 0.4483918249607086, + "eval_runtime": 8.4558, + "eval_samples_per_second": 11.826, + "eval_wer": 0.38246041412911086, + "step": 31600 + }, + { + "epoch": 5.16, + "learning_rate": 7.9637787567303e-05, + "loss": NaN, + "step": 31700 + }, + { + "epoch": 5.16, + "eval_loss": 0.40626582503318787, + "eval_runtime": 7.4323, + "eval_samples_per_second": 13.455, + "eval_wer": 0.3751522533495737, + "step": 31700 + }, + { + "epoch": 5.17, + "learning_rate": 7.957252406591614e-05, + "loss": NaN, + "step": 31800 + }, + { + "epoch": 5.17, + "eval_loss": 0.4829193949699402, + "eval_runtime": 7.8066, + "eval_samples_per_second": 12.81, + "eval_wer": 0.40194884287454324, + "step": 31800 + }, + { + "epoch": 5.19, + "learning_rate": 7.95072605645293e-05, + "loss": NaN, + "step": 31900 + }, + { + "epoch": 5.19, + "eval_loss": 0.49380388855934143, + "eval_runtime": 8.2852, + "eval_samples_per_second": 12.07, + "eval_wer": 0.38246041412911086, + "step": 31900 + }, + { + "epoch": 5.2, + "learning_rate": 7.944199706314243e-05, + "loss": NaN, + "step": 32000 + }, + { + "epoch": 5.2, + "eval_loss": 0.4846552014350891, + "eval_runtime": 8.2309, + "eval_samples_per_second": 12.149, + "eval_wer": 0.3812423873325213, + "step": 32000 + }, + { + "epoch": 5.22, + "learning_rate": 7.937673356175559e-05, + "loss": NaN, + "step": 32100 + }, + { + "epoch": 5.22, + "eval_loss": 0.4324251413345337, + "eval_runtime": 7.5408, + "eval_samples_per_second": 13.261, + "eval_wer": 0.38976857490864797, + "step": 32100 + }, + { + "epoch": 5.24, + "learning_rate": 7.931147006036875e-05, + "loss": NaN, + "step": 32200 + }, + { + "epoch": 5.24, + "eval_loss": 0.4314603805541992, + "eval_runtime": 8.6415, + "eval_samples_per_second": 11.572, + "eval_wer": 0.3909866017052375, + "step": 32200 + }, + { + "epoch": 5.25, + "learning_rate": 7.92462065589819e-05, + "loss": NaN, + "step": 32300 + }, + { + "epoch": 5.25, + "eval_loss": 0.4612935185432434, + "eval_runtime": 8.2981, + "eval_samples_per_second": 12.051, + "eval_wer": 0.4092570036540804, + "step": 32300 + }, + { + "epoch": 5.27, + "learning_rate": 7.918094305759505e-05, + "loss": NaN, + "step": 32400 + }, + { + "epoch": 5.27, + "eval_loss": 0.4439612329006195, + "eval_runtime": 6.8526, + "eval_samples_per_second": 14.593, + "eval_wer": 0.3946406820950061, + "step": 32400 + }, + { + "epoch": 5.29, + "learning_rate": 7.91156795562082e-05, + "loss": NaN, + "step": 32500 + }, + { + "epoch": 5.29, + "eval_loss": 0.5218726396560669, + "eval_runtime": 10.5449, + "eval_samples_per_second": 9.483, + "eval_wer": 0.42143727161997563, + "step": 32500 + }, + { + "epoch": 5.3, + "learning_rate": 7.905041605482134e-05, + "loss": NaN, + "step": 32600 + }, + { + "epoch": 5.3, + "eval_loss": 0.45300713181495667, + "eval_runtime": 7.9081, + "eval_samples_per_second": 12.645, + "eval_wer": 0.3812423873325213, + "step": 32600 + }, + { + "epoch": 5.32, + "learning_rate": 7.898515255343449e-05, + "loss": NaN, + "step": 32700 + }, + { + "epoch": 5.32, + "eval_loss": 0.46696189045906067, + "eval_runtime": 8.375, + "eval_samples_per_second": 11.94, + "eval_wer": 0.36662606577344703, + "step": 32700 + }, + { + "epoch": 5.33, + "learning_rate": 7.891988905204765e-05, + "loss": NaN, + "step": 32800 + }, + { + "epoch": 5.33, + "eval_loss": 0.4551876485347748, + "eval_runtime": 8.9235, + "eval_samples_per_second": 11.206, + "eval_wer": 0.39707673568818513, + "step": 32800 + }, + { + "epoch": 5.35, + "learning_rate": 7.88546255506608e-05, + "loss": NaN, + "step": 32900 + }, + { + "epoch": 5.35, + "eval_loss": 0.49849167466163635, + "eval_runtime": 7.4164, + "eval_samples_per_second": 13.484, + "eval_wer": 0.3763702801461632, + "step": 32900 + }, + { + "epoch": 5.37, + "learning_rate": 7.878936204927395e-05, + "loss": NaN, + "step": 33000 + }, + { + "epoch": 5.37, + "eval_loss": 0.4134560525417328, + "eval_runtime": 8.9386, + "eval_samples_per_second": 11.187, + "eval_wer": 0.3800243605359318, + "step": 33000 + }, + { + "epoch": 5.38, + "learning_rate": 7.87240985478871e-05, + "loss": NaN, + "step": 33100 + }, + { + "epoch": 5.38, + "eval_loss": 0.43791449069976807, + "eval_runtime": 7.9006, + "eval_samples_per_second": 12.657, + "eval_wer": 0.37149817295980514, + "step": 33100 + }, + { + "epoch": 5.4, + "learning_rate": 7.865883504650024e-05, + "loss": NaN, + "step": 33200 + }, + { + "epoch": 5.4, + "eval_loss": 0.3854129910469055, + "eval_runtime": 8.888, + "eval_samples_per_second": 11.251, + "eval_wer": 0.3702801461632156, + "step": 33200 + }, + { + "epoch": 5.42, + "learning_rate": 7.85935715451134e-05, + "loss": NaN, + "step": 33300 + }, + { + "epoch": 5.42, + "eval_loss": 0.45465293526649475, + "eval_runtime": 7.5638, + "eval_samples_per_second": 13.221, + "eval_wer": 0.392204628501827, + "step": 33300 + }, + { + "epoch": 5.43, + "learning_rate": 7.852830804372654e-05, + "loss": NaN, + "step": 33400 + }, + { + "epoch": 5.43, + "eval_loss": 0.4339330196380615, + "eval_runtime": 7.2942, + "eval_samples_per_second": 13.709, + "eval_wer": 0.37880633373934225, + "step": 33400 + }, + { + "epoch": 5.45, + "learning_rate": 7.84630445423397e-05, + "loss": NaN, + "step": 33500 + }, + { + "epoch": 5.45, + "eval_loss": 0.47783201932907104, + "eval_runtime": 10.4013, + "eval_samples_per_second": 9.614, + "eval_wer": 0.38976857490864797, + "step": 33500 + }, + { + "epoch": 5.46, + "learning_rate": 7.839778104095286e-05, + "loss": NaN, + "step": 33600 + }, + { + "epoch": 5.46, + "eval_loss": 0.451149582862854, + "eval_runtime": 10.0363, + "eval_samples_per_second": 9.964, + "eval_wer": 0.3934226552984166, + "step": 33600 + }, + { + "epoch": 5.48, + "learning_rate": 7.8332517539566e-05, + "loss": NaN, + "step": 33700 + }, + { + "epoch": 5.48, + "eval_loss": 0.4253043830394745, + "eval_runtime": 7.7756, + "eval_samples_per_second": 12.861, + "eval_wer": 0.3751522533495737, + "step": 33700 + }, + { + "epoch": 5.5, + "learning_rate": 7.826725403817915e-05, + "loss": NaN, + "step": 33800 + }, + { + "epoch": 5.5, + "eval_loss": 0.4464089274406433, + "eval_runtime": 6.8445, + "eval_samples_per_second": 14.61, + "eval_wer": 0.38976857490864797, + "step": 33800 + }, + { + "epoch": 5.51, + "learning_rate": 7.82019905367923e-05, + "loss": NaN, + "step": 33900 + }, + { + "epoch": 5.51, + "eval_loss": 0.4914042353630066, + "eval_runtime": 8.3493, + "eval_samples_per_second": 11.977, + "eval_wer": 0.39585870889159563, + "step": 33900 + }, + { + "epoch": 5.53, + "learning_rate": 7.813672703540546e-05, + "loss": NaN, + "step": 34000 + }, + { + "epoch": 5.53, + "eval_loss": 0.4946918487548828, + "eval_runtime": 7.3641, + "eval_samples_per_second": 13.579, + "eval_wer": 0.3995127892813642, + "step": 34000 + }, + { + "epoch": 5.55, + "learning_rate": 7.80714635340186e-05, + "loss": NaN, + "step": 34100 + }, + { + "epoch": 5.55, + "eval_loss": 0.5067921280860901, + "eval_runtime": 8.8184, + "eval_samples_per_second": 11.34, + "eval_wer": 0.4092570036540804, + "step": 34100 + }, + { + "epoch": 5.56, + "learning_rate": 7.800620003263176e-05, + "loss": NaN, + "step": 34200 + }, + { + "epoch": 5.56, + "eval_loss": 0.517976701259613, + "eval_runtime": 8.7478, + "eval_samples_per_second": 11.431, + "eval_wer": 0.40682095006090135, + "step": 34200 + }, + { + "epoch": 5.58, + "learning_rate": 7.79409365312449e-05, + "loss": NaN, + "step": 34300 + }, + { + "epoch": 5.58, + "eval_loss": 0.5033119916915894, + "eval_runtime": 9.201, + "eval_samples_per_second": 10.868, + "eval_wer": 0.392204628501827, + "step": 34300 + }, + { + "epoch": 5.59, + "learning_rate": 7.787567302985805e-05, + "loss": NaN, + "step": 34400 + }, + { + "epoch": 5.59, + "eval_loss": 0.49291473627090454, + "eval_runtime": 11.455, + "eval_samples_per_second": 8.73, + "eval_wer": 0.40316686967113274, + "step": 34400 + }, + { + "epoch": 5.61, + "learning_rate": 7.781040952847121e-05, + "loss": NaN, + "step": 34500 + }, + { + "epoch": 5.61, + "eval_loss": 0.4840911626815796, + "eval_runtime": 9.864, + "eval_samples_per_second": 10.138, + "eval_wer": 0.39707673568818513, + "step": 34500 + }, + { + "epoch": 5.63, + "learning_rate": 7.774514602708435e-05, + "loss": NaN, + "step": 34600 + }, + { + "epoch": 5.63, + "eval_loss": 0.4436954855918884, + "eval_runtime": 7.5792, + "eval_samples_per_second": 13.194, + "eval_wer": 0.37393422655298414, + "step": 34600 + }, + { + "epoch": 5.64, + "learning_rate": 7.767988252569751e-05, + "loss": NaN, + "step": 34700 + }, + { + "epoch": 5.64, + "eval_loss": 0.42483794689178467, + "eval_runtime": 9.6007, + "eval_samples_per_second": 10.416, + "eval_wer": 0.36784409257003653, + "step": 34700 + }, + { + "epoch": 5.66, + "learning_rate": 7.761461902431066e-05, + "loss": NaN, + "step": 34800 + }, + { + "epoch": 5.66, + "eval_loss": 0.4381932020187378, + "eval_runtime": 8.7213, + "eval_samples_per_second": 11.466, + "eval_wer": 0.3909866017052375, + "step": 34800 + }, + { + "epoch": 5.68, + "learning_rate": 7.75493555229238e-05, + "loss": NaN, + "step": 34900 + }, + { + "epoch": 5.68, + "eval_loss": 0.40743523836135864, + "eval_runtime": 9.6342, + "eval_samples_per_second": 10.38, + "eval_wer": 0.38367844092570036, + "step": 34900 + }, + { + "epoch": 5.69, + "learning_rate": 7.748409202153696e-05, + "loss": NaN, + "step": 35000 + }, + { + "epoch": 5.69, + "eval_loss": 0.4256589412689209, + "eval_runtime": 7.6861, + "eval_samples_per_second": 13.01, + "eval_wer": 0.392204628501827, + "step": 35000 + }, + { + "epoch": 5.71, + "learning_rate": 7.741882852015011e-05, + "loss": NaN, + "step": 35100 + }, + { + "epoch": 5.71, + "eval_loss": 0.41759517788887024, + "eval_runtime": 7.5282, + "eval_samples_per_second": 13.283, + "eval_wer": 0.3946406820950061, + "step": 35100 + }, + { + "epoch": 5.72, + "learning_rate": 7.735356501876327e-05, + "loss": NaN, + "step": 35200 + }, + { + "epoch": 5.72, + "eval_loss": 0.4593577980995178, + "eval_runtime": 8.4088, + "eval_samples_per_second": 11.892, + "eval_wer": 0.4104750304506699, + "step": 35200 + }, + { + "epoch": 5.74, + "learning_rate": 7.728830151737641e-05, + "loss": NaN, + "step": 35300 + }, + { + "epoch": 5.74, + "eval_loss": 0.4901994466781616, + "eval_runtime": 7.524, + "eval_samples_per_second": 13.291, + "eval_wer": 0.4056029232643118, + "step": 35300 + }, + { + "epoch": 5.76, + "learning_rate": 7.722303801598956e-05, + "loss": NaN, + "step": 35400 + }, + { + "epoch": 5.76, + "eval_loss": 0.47598525881767273, + "eval_runtime": 8.5548, + "eval_samples_per_second": 11.689, + "eval_wer": 0.38246041412911086, + "step": 35400 + }, + { + "epoch": 5.77, + "learning_rate": 7.715777451460271e-05, + "loss": NaN, + "step": 35500 + }, + { + "epoch": 5.77, + "eval_loss": 0.4955834150314331, + "eval_runtime": 8.2229, + "eval_samples_per_second": 12.161, + "eval_wer": 0.3995127892813642, + "step": 35500 + }, + { + "epoch": 5.79, + "learning_rate": 7.709251101321586e-05, + "loss": NaN, + "step": 35600 + }, + { + "epoch": 5.79, + "eval_loss": 0.5013775825500488, + "eval_runtime": 7.3332, + "eval_samples_per_second": 13.637, + "eval_wer": 0.37758830694275275, + "step": 35600 + }, + { + "epoch": 5.81, + "learning_rate": 7.702724751182902e-05, + "loss": NaN, + "step": 35700 + }, + { + "epoch": 5.81, + "eval_loss": 0.4535202085971832, + "eval_runtime": 6.7442, + "eval_samples_per_second": 14.828, + "eval_wer": 0.3617539585870889, + "step": 35700 + }, + { + "epoch": 5.82, + "learning_rate": 7.696198401044216e-05, + "loss": NaN, + "step": 35800 + }, + { + "epoch": 5.82, + "eval_loss": 0.43038153648376465, + "eval_runtime": 7.9798, + "eval_samples_per_second": 12.532, + "eval_wer": 0.38367844092570036, + "step": 35800 + }, + { + "epoch": 5.84, + "learning_rate": 7.689672050905532e-05, + "loss": NaN, + "step": 35900 + }, + { + "epoch": 5.84, + "eval_loss": 0.4333917200565338, + "eval_runtime": 8.5583, + "eval_samples_per_second": 11.685, + "eval_wer": 0.3800243605359318, + "step": 35900 + }, + { + "epoch": 5.85, + "learning_rate": 7.683145700766847e-05, + "loss": NaN, + "step": 36000 + }, + { + "epoch": 5.85, + "eval_loss": 0.49249768257141113, + "eval_runtime": 9.3291, + "eval_samples_per_second": 10.719, + "eval_wer": 0.40316686967113274, + "step": 36000 + }, + { + "epoch": 5.87, + "learning_rate": 7.676619350628161e-05, + "loss": NaN, + "step": 36100 + }, + { + "epoch": 5.87, + "eval_loss": 0.47578758001327515, + "eval_runtime": 7.9207, + "eval_samples_per_second": 12.625, + "eval_wer": 0.38976857490864797, + "step": 36100 + }, + { + "epoch": 5.89, + "learning_rate": 7.670093000489477e-05, + "loss": NaN, + "step": 36200 + }, + { + "epoch": 5.89, + "eval_loss": 0.44561660289764404, + "eval_runtime": 6.6969, + "eval_samples_per_second": 14.932, + "eval_wer": 0.3873325213154689, + "step": 36200 + }, + { + "epoch": 5.9, + "learning_rate": 7.663566650350792e-05, + "loss": NaN, + "step": 36300 + }, + { + "epoch": 5.9, + "eval_loss": 0.46735045313835144, + "eval_runtime": 7.2262, + "eval_samples_per_second": 13.839, + "eval_wer": 0.39585870889159563, + "step": 36300 + }, + { + "epoch": 5.92, + "learning_rate": 7.657040300212108e-05, + "loss": NaN, + "step": 36400 + }, + { + "epoch": 5.92, + "eval_loss": 0.42316126823425293, + "eval_runtime": 8.7463, + "eval_samples_per_second": 11.433, + "eval_wer": 0.37758830694275275, + "step": 36400 + }, + { + "epoch": 5.94, + "learning_rate": 7.65051395007342e-05, + "loss": NaN, + "step": 36500 + }, + { + "epoch": 5.94, + "eval_loss": 0.46479174494743347, + "eval_runtime": 9.1318, + "eval_samples_per_second": 10.951, + "eval_wer": 0.39585870889159563, + "step": 36500 + }, + { + "epoch": 5.95, + "learning_rate": 7.643987599934737e-05, + "loss": NaN, + "step": 36600 + }, + { + "epoch": 5.95, + "eval_loss": 0.42728888988494873, + "eval_runtime": 8.4942, + "eval_samples_per_second": 11.773, + "eval_wer": 0.37758830694275275, + "step": 36600 + }, + { + "epoch": 5.97, + "learning_rate": 7.637461249796052e-05, + "loss": NaN, + "step": 36700 + }, + { + "epoch": 5.97, + "eval_loss": 0.4792998433113098, + "eval_runtime": 8.1451, + "eval_samples_per_second": 12.277, + "eval_wer": 0.4007308160779537, + "step": 36700 + }, + { + "epoch": 5.98, + "learning_rate": 7.630934899657367e-05, + "loss": NaN, + "step": 36800 + }, + { + "epoch": 5.98, + "eval_loss": 0.4131828546524048, + "eval_runtime": 6.536, + "eval_samples_per_second": 15.3, + "eval_wer": 0.3629719853836784, + "step": 36800 + }, + { + "epoch": 6.0, + "learning_rate": 7.624408549518683e-05, + "loss": NaN, + "step": 36900 + }, + { + "epoch": 6.0, + "eval_loss": 0.5964876413345337, + "eval_runtime": 8.3503, + "eval_samples_per_second": 11.976, + "eval_wer": 0.392204628501827, + "step": 36900 + }, + { + "epoch": 6.02, + "learning_rate": 7.617882199379997e-05, + "loss": NaN, + "step": 37000 + }, + { + "epoch": 6.02, + "eval_loss": 0.5113080739974976, + "eval_runtime": 7.8786, + "eval_samples_per_second": 12.693, + "eval_wer": 0.40803897685749085, + "step": 37000 + }, + { + "epoch": 6.03, + "learning_rate": 7.611355849241312e-05, + "loss": NaN, + "step": 37100 + }, + { + "epoch": 6.03, + "eval_loss": 0.44438639283180237, + "eval_runtime": 7.8581, + "eval_samples_per_second": 12.726, + "eval_wer": 0.3812423873325213, + "step": 37100 + }, + { + "epoch": 6.05, + "learning_rate": 7.604829499102626e-05, + "loss": NaN, + "step": 37200 + }, + { + "epoch": 6.05, + "eval_loss": 0.44501668214797974, + "eval_runtime": 6.8166, + "eval_samples_per_second": 14.67, + "eval_wer": 0.3861144945188794, + "step": 37200 + }, + { + "epoch": 6.07, + "learning_rate": 7.598303148963942e-05, + "loss": NaN, + "step": 37300 + }, + { + "epoch": 6.07, + "eval_loss": 0.46016010642051697, + "eval_runtime": 6.8605, + "eval_samples_per_second": 14.576, + "eval_wer": 0.37758830694275275, + "step": 37300 + }, + { + "epoch": 6.08, + "learning_rate": 7.591776798825258e-05, + "loss": NaN, + "step": 37400 + }, + { + "epoch": 6.08, + "eval_loss": 0.4487974941730499, + "eval_runtime": 8.698, + "eval_samples_per_second": 11.497, + "eval_wer": 0.392204628501827, + "step": 37400 + }, + { + "epoch": 6.1, + "learning_rate": 7.585250448686573e-05, + "loss": NaN, + "step": 37500 + }, + { + "epoch": 6.1, + "eval_loss": 0.46871453523635864, + "eval_runtime": 7.2092, + "eval_samples_per_second": 13.871, + "eval_wer": 0.3909866017052375, + "step": 37500 + }, + { + "epoch": 6.11, + "learning_rate": 7.578724098547888e-05, + "loss": NaN, + "step": 37600 + }, + { + "epoch": 6.11, + "eval_loss": 0.5187182426452637, + "eval_runtime": 8.235, + "eval_samples_per_second": 12.143, + "eval_wer": 0.3848964677222899, + "step": 37600 + }, + { + "epoch": 6.13, + "learning_rate": 7.572197748409202e-05, + "loss": NaN, + "step": 37700 + }, + { + "epoch": 6.13, + "eval_loss": 0.4755428433418274, + "eval_runtime": 8.6418, + "eval_samples_per_second": 11.572, + "eval_wer": 0.3909866017052375, + "step": 37700 + }, + { + "epoch": 6.15, + "learning_rate": 7.565671398270517e-05, + "loss": NaN, + "step": 37800 + }, + { + "epoch": 6.15, + "eval_loss": 0.4716055393218994, + "eval_runtime": 8.6568, + "eval_samples_per_second": 11.552, + "eval_wer": 0.39585870889159563, + "step": 37800 + }, + { + "epoch": 6.16, + "learning_rate": 7.559145048131832e-05, + "loss": NaN, + "step": 37900 + }, + { + "epoch": 6.16, + "eval_loss": 0.43883880972862244, + "eval_runtime": 8.0179, + "eval_samples_per_second": 12.472, + "eval_wer": 0.3861144945188794, + "step": 37900 + }, + { + "epoch": 6.18, + "learning_rate": 7.552618697993148e-05, + "loss": NaN, + "step": 38000 + }, + { + "epoch": 6.18, + "eval_loss": 0.447729229927063, + "eval_runtime": 7.3355, + "eval_samples_per_second": 13.632, + "eval_wer": 0.38855054811205847, + "step": 38000 + }, + { + "epoch": 6.2, + "learning_rate": 7.546092347854464e-05, + "loss": NaN, + "step": 38100 + }, + { + "epoch": 6.2, + "eval_loss": 0.5071135759353638, + "eval_runtime": 9.0115, + "eval_samples_per_second": 11.097, + "eval_wer": 0.4092570036540804, + "step": 38100 + }, + { + "epoch": 6.21, + "learning_rate": 7.539565997715777e-05, + "loss": NaN, + "step": 38200 + }, + { + "epoch": 6.21, + "eval_loss": 0.45388340950012207, + "eval_runtime": 7.6236, + "eval_samples_per_second": 13.117, + "eval_wer": 0.3873325213154689, + "step": 38200 + }, + { + "epoch": 6.23, + "learning_rate": 7.533039647577093e-05, + "loss": NaN, + "step": 38300 + }, + { + "epoch": 6.23, + "eval_loss": 0.4727196991443634, + "eval_runtime": 9.3783, + "eval_samples_per_second": 10.663, + "eval_wer": 0.392204628501827, + "step": 38300 + }, + { + "epoch": 6.24, + "learning_rate": 7.526513297438407e-05, + "loss": NaN, + "step": 38400 + }, + { + "epoch": 6.24, + "eval_loss": 0.45815300941467285, + "eval_runtime": 8.7937, + "eval_samples_per_second": 11.372, + "eval_wer": 0.3995127892813642, + "step": 38400 + }, + { + "epoch": 6.26, + "learning_rate": 7.519986947299723e-05, + "loss": NaN, + "step": 38500 + }, + { + "epoch": 6.26, + "eval_loss": 0.45697593688964844, + "eval_runtime": 7.849, + "eval_samples_per_second": 12.74, + "eval_wer": 0.38367844092570036, + "step": 38500 + }, + { + "epoch": 6.28, + "learning_rate": 7.513460597161039e-05, + "loss": NaN, + "step": 38600 + }, + { + "epoch": 6.28, + "eval_loss": 0.4512711465358734, + "eval_runtime": 8.0969, + "eval_samples_per_second": 12.35, + "eval_wer": 0.4056029232643118, + "step": 38600 + }, + { + "epoch": 6.29, + "learning_rate": 7.506934247022354e-05, + "loss": NaN, + "step": 38700 + }, + { + "epoch": 6.29, + "eval_loss": 0.47323668003082275, + "eval_runtime": 9.4309, + "eval_samples_per_second": 10.603, + "eval_wer": 0.4007308160779537, + "step": 38700 + }, + { + "epoch": 6.31, + "learning_rate": 7.500407896883668e-05, + "loss": NaN, + "step": 38800 + }, + { + "epoch": 6.31, + "eval_loss": 0.46344485878944397, + "eval_runtime": 7.4323, + "eval_samples_per_second": 13.455, + "eval_wer": 0.3909866017052375, + "step": 38800 + }, + { + "epoch": 6.33, + "learning_rate": 7.493881546744983e-05, + "loss": NaN, + "step": 38900 + }, + { + "epoch": 6.33, + "eval_loss": 0.46373656392097473, + "eval_runtime": 8.8968, + "eval_samples_per_second": 11.24, + "eval_wer": 0.4043848964677223, + "step": 38900 + }, + { + "epoch": 6.34, + "learning_rate": 7.487355196606298e-05, + "loss": NaN, + "step": 39000 + }, + { + "epoch": 6.34, + "eval_loss": 0.5335507392883301, + "eval_runtime": 7.8243, + "eval_samples_per_second": 12.781, + "eval_wer": 0.3982947624847747, + "step": 39000 + }, + { + "epoch": 6.36, + "learning_rate": 7.480828846467613e-05, + "loss": NaN, + "step": 39100 + }, + { + "epoch": 6.36, + "eval_loss": 0.4426437318325043, + "eval_runtime": 9.3026, + "eval_samples_per_second": 10.75, + "eval_wer": 0.38246041412911086, + "step": 39100 + }, + { + "epoch": 6.38, + "learning_rate": 7.474302496328929e-05, + "loss": NaN, + "step": 39200 + }, + { + "epoch": 6.38, + "eval_loss": 0.4251578450202942, + "eval_runtime": 7.5958, + "eval_samples_per_second": 13.165, + "eval_wer": 0.38976857490864797, + "step": 39200 + }, + { + "epoch": 6.39, + "learning_rate": 7.467776146190245e-05, + "loss": NaN, + "step": 39300 + }, + { + "epoch": 6.39, + "eval_loss": 0.45868903398513794, + "eval_runtime": 7.9208, + "eval_samples_per_second": 12.625, + "eval_wer": 0.37271619975639464, + "step": 39300 + }, + { + "epoch": 6.41, + "learning_rate": 7.461249796051558e-05, + "loss": NaN, + "step": 39400 + }, + { + "epoch": 6.41, + "eval_loss": 0.4029034376144409, + "eval_runtime": 7.9815, + "eval_samples_per_second": 12.529, + "eval_wer": 0.37271619975639464, + "step": 39400 + }, + { + "epoch": 6.42, + "learning_rate": 7.454723445912874e-05, + "loss": NaN, + "step": 39500 + }, + { + "epoch": 6.42, + "eval_loss": 0.43351614475250244, + "eval_runtime": 8.7741, + "eval_samples_per_second": 11.397, + "eval_wer": 0.3848964677222899, + "step": 39500 + }, + { + "epoch": 6.44, + "learning_rate": 7.448197095774188e-05, + "loss": NaN, + "step": 39600 + }, + { + "epoch": 6.44, + "eval_loss": 0.4573790729045868, + "eval_runtime": 7.9568, + "eval_samples_per_second": 12.568, + "eval_wer": 0.38367844092570036, + "step": 39600 + }, + { + "epoch": 6.46, + "learning_rate": 7.441670745635504e-05, + "loss": NaN, + "step": 39700 + }, + { + "epoch": 6.46, + "eval_loss": 0.47754156589508057, + "eval_runtime": 8.0778, + "eval_samples_per_second": 12.38, + "eval_wer": 0.3848964677222899, + "step": 39700 + }, + { + "epoch": 6.47, + "learning_rate": 7.435144395496819e-05, + "loss": NaN, + "step": 39800 + }, + { + "epoch": 6.47, + "eval_loss": 0.5262435674667358, + "eval_runtime": 9.0367, + "eval_samples_per_second": 11.066, + "eval_wer": 0.3934226552984166, + "step": 39800 + }, + { + "epoch": 6.49, + "learning_rate": 7.428618045358133e-05, + "loss": NaN, + "step": 39900 + }, + { + "epoch": 6.49, + "eval_loss": 0.45043036341667175, + "eval_runtime": 8.4329, + "eval_samples_per_second": 11.858, + "eval_wer": 0.38855054811205847, + "step": 39900 + }, + { + "epoch": 6.51, + "learning_rate": 7.422091695219449e-05, + "loss": NaN, + "step": 40000 + }, + { + "epoch": 6.51, + "eval_loss": 0.5360258221626282, + "eval_runtime": 8.3169, + "eval_samples_per_second": 12.024, + "eval_wer": 0.38976857490864797, + "step": 40000 + }, + { + "epoch": 6.52, + "learning_rate": 7.415565345080763e-05, + "loss": NaN, + "step": 40100 + }, + { + "epoch": 6.52, + "eval_loss": 0.43279510736465454, + "eval_runtime": 8.3411, + "eval_samples_per_second": 11.989, + "eval_wer": 0.37880633373934225, + "step": 40100 + }, + { + "epoch": 6.54, + "learning_rate": 7.40903899494208e-05, + "loss": NaN, + "step": 40200 + }, + { + "epoch": 6.54, + "eval_loss": 0.436869353055954, + "eval_runtime": 7.5681, + "eval_samples_per_second": 13.213, + "eval_wer": 0.37758830694275275, + "step": 40200 + }, + { + "epoch": 6.55, + "learning_rate": 7.402512644803394e-05, + "loss": NaN, + "step": 40300 + }, + { + "epoch": 6.55, + "eval_loss": 0.48069870471954346, + "eval_runtime": 7.7372, + "eval_samples_per_second": 12.924, + "eval_wer": 0.4177831912302071, + "step": 40300 + }, + { + "epoch": 6.57, + "learning_rate": 7.39598629466471e-05, + "loss": NaN, + "step": 40400 + }, + { + "epoch": 6.57, + "eval_loss": 0.43905335664749146, + "eval_runtime": 7.4851, + "eval_samples_per_second": 13.36, + "eval_wer": 0.3848964677222899, + "step": 40400 + }, + { + "epoch": 6.59, + "learning_rate": 7.389459944526024e-05, + "loss": NaN, + "step": 40500 + }, + { + "epoch": 6.59, + "eval_loss": 0.45289239287376404, + "eval_runtime": 8.056, + "eval_samples_per_second": 12.413, + "eval_wer": 0.3873325213154689, + "step": 40500 + }, + { + "epoch": 6.6, + "learning_rate": 7.382933594387339e-05, + "loss": NaN, + "step": 40600 + }, + { + "epoch": 6.6, + "eval_loss": 0.4513566195964813, + "eval_runtime": 8.3785, + "eval_samples_per_second": 11.935, + "eval_wer": 0.3848964677222899, + "step": 40600 + }, + { + "epoch": 6.62, + "learning_rate": 7.376407244248655e-05, + "loss": NaN, + "step": 40700 + }, + { + "epoch": 6.62, + "eval_loss": 0.45488113164901733, + "eval_runtime": 9.389, + "eval_samples_per_second": 10.651, + "eval_wer": 0.37149817295980514, + "step": 40700 + }, + { + "epoch": 6.64, + "learning_rate": 7.369880894109969e-05, + "loss": NaN, + "step": 40800 + }, + { + "epoch": 6.64, + "eval_loss": 0.4812730550765991, + "eval_runtime": 9.6857, + "eval_samples_per_second": 10.324, + "eval_wer": 0.37149817295980514, + "step": 40800 + }, + { + "epoch": 6.65, + "learning_rate": 7.363354543971285e-05, + "loss": NaN, + "step": 40900 + }, + { + "epoch": 6.65, + "eval_loss": 0.4662674367427826, + "eval_runtime": 9.1141, + "eval_samples_per_second": 10.972, + "eval_wer": 0.37271619975639464, + "step": 40900 + }, + { + "epoch": 6.67, + "learning_rate": 7.3568281938326e-05, + "loss": NaN, + "step": 41000 + }, + { + "epoch": 6.67, + "eval_loss": 0.43189141154289246, + "eval_runtime": 7.7905, + "eval_samples_per_second": 12.836, + "eval_wer": 0.38855054811205847, + "step": 41000 + }, + { + "epoch": 6.68, + "learning_rate": 7.350301843693914e-05, + "loss": NaN, + "step": 41100 + }, + { + "epoch": 6.68, + "eval_loss": 0.4479624032974243, + "eval_runtime": 6.7026, + "eval_samples_per_second": 14.919, + "eval_wer": 0.392204628501827, + "step": 41100 + }, + { + "epoch": 6.7, + "learning_rate": 7.34377549355523e-05, + "loss": NaN, + "step": 41200 + }, + { + "epoch": 6.7, + "eval_loss": 0.5534685254096985, + "eval_runtime": 7.9313, + "eval_samples_per_second": 12.608, + "eval_wer": 0.392204628501827, + "step": 41200 + }, + { + "epoch": 6.72, + "learning_rate": 7.337249143416544e-05, + "loss": NaN, + "step": 41300 + }, + { + "epoch": 6.72, + "eval_loss": 0.42362719774246216, + "eval_runtime": 7.9752, + "eval_samples_per_second": 12.539, + "eval_wer": 0.3763702801461632, + "step": 41300 + }, + { + "epoch": 6.73, + "learning_rate": 7.33072279327786e-05, + "loss": NaN, + "step": 41400 + }, + { + "epoch": 6.73, + "eval_loss": 0.4316897690296173, + "eval_runtime": 9.0445, + "eval_samples_per_second": 11.056, + "eval_wer": 0.3800243605359318, + "step": 41400 + }, + { + "epoch": 6.75, + "learning_rate": 7.324196443139175e-05, + "loss": NaN, + "step": 41500 + }, + { + "epoch": 6.75, + "eval_loss": 0.4458038806915283, + "eval_runtime": 8.5585, + "eval_samples_per_second": 11.684, + "eval_wer": 0.38246041412911086, + "step": 41500 + }, + { + "epoch": 6.77, + "learning_rate": 7.31767009300049e-05, + "loss": NaN, + "step": 41600 + }, + { + "epoch": 6.77, + "eval_loss": 0.4252176582813263, + "eval_runtime": 6.9682, + "eval_samples_per_second": 14.351, + "eval_wer": 0.38855054811205847, + "step": 41600 + }, + { + "epoch": 6.78, + "learning_rate": 7.311143742861805e-05, + "loss": NaN, + "step": 41700 + }, + { + "epoch": 6.78, + "eval_loss": 0.42161300778388977, + "eval_runtime": 8.2804, + "eval_samples_per_second": 12.077, + "eval_wer": 0.36053593179049936, + "step": 41700 + }, + { + "epoch": 6.8, + "learning_rate": 7.30461739272312e-05, + "loss": NaN, + "step": 41800 + }, + { + "epoch": 6.8, + "eval_loss": 0.43236517906188965, + "eval_runtime": 6.8741, + "eval_samples_per_second": 14.547, + "eval_wer": 0.3763702801461632, + "step": 41800 + }, + { + "epoch": 6.81, + "learning_rate": 7.298091042584436e-05, + "loss": NaN, + "step": 41900 + }, + { + "epoch": 6.81, + "eval_loss": 0.42260417342185974, + "eval_runtime": 8.7571, + "eval_samples_per_second": 11.419, + "eval_wer": 0.3763702801461632, + "step": 41900 + }, + { + "epoch": 6.83, + "learning_rate": 7.29156469244575e-05, + "loss": NaN, + "step": 42000 + }, + { + "epoch": 6.83, + "eval_loss": 0.4526902735233307, + "eval_runtime": 9.9377, + "eval_samples_per_second": 10.063, + "eval_wer": 0.364190012180268, + "step": 42000 + }, + { + "epoch": 6.85, + "learning_rate": 7.285038342307066e-05, + "loss": NaN, + "step": 42100 + }, + { + "epoch": 6.85, + "eval_loss": 0.4244474768638611, + "eval_runtime": 7.3805, + "eval_samples_per_second": 13.549, + "eval_wer": 0.3861144945188794, + "step": 42100 + }, + { + "epoch": 6.86, + "learning_rate": 7.278511992168379e-05, + "loss": NaN, + "step": 42200 + }, + { + "epoch": 6.86, + "eval_loss": 0.3987332284450531, + "eval_runtime": 7.9724, + "eval_samples_per_second": 12.543, + "eval_wer": 0.3690621193666261, + "step": 42200 + }, + { + "epoch": 6.88, + "learning_rate": 7.271985642029695e-05, + "loss": NaN, + "step": 42300 + }, + { + "epoch": 6.88, + "eval_loss": 0.4404261112213135, + "eval_runtime": 8.8576, + "eval_samples_per_second": 11.29, + "eval_wer": 0.38976857490864797, + "step": 42300 + }, + { + "epoch": 6.9, + "learning_rate": 7.265459291891011e-05, + "loss": NaN, + "step": 42400 + }, + { + "epoch": 6.9, + "eval_loss": 0.4243963658809662, + "eval_runtime": 9.9726, + "eval_samples_per_second": 10.027, + "eval_wer": 0.37758830694275275, + "step": 42400 + }, + { + "epoch": 6.91, + "learning_rate": 7.258932941752325e-05, + "loss": NaN, + "step": 42500 + }, + { + "epoch": 6.91, + "eval_loss": 0.44823479652404785, + "eval_runtime": 9.933, + "eval_samples_per_second": 10.067, + "eval_wer": 0.3995127892813642, + "step": 42500 + }, + { + "epoch": 6.93, + "learning_rate": 7.252406591613641e-05, + "loss": NaN, + "step": 42600 + }, + { + "epoch": 6.93, + "eval_loss": 0.39962050318717957, + "eval_runtime": 6.5522, + "eval_samples_per_second": 15.262, + "eval_wer": 0.3848964677222899, + "step": 42600 + }, + { + "epoch": 6.94, + "learning_rate": 7.245880241474956e-05, + "loss": NaN, + "step": 42700 + }, + { + "epoch": 6.94, + "eval_loss": 0.43537867069244385, + "eval_runtime": 7.7406, + "eval_samples_per_second": 12.919, + "eval_wer": 0.3812423873325213, + "step": 42700 + }, + { + "epoch": 6.96, + "learning_rate": 7.23935389133627e-05, + "loss": NaN, + "step": 42800 + }, + { + "epoch": 6.96, + "eval_loss": 0.43869927525520325, + "eval_runtime": 8.2951, + "eval_samples_per_second": 12.055, + "eval_wer": 0.392204628501827, + "step": 42800 + }, + { + "epoch": 6.98, + "learning_rate": 7.232827541197585e-05, + "loss": NaN, + "step": 42900 + }, + { + "epoch": 6.98, + "eval_loss": 0.4649328589439392, + "eval_runtime": 8.5972, + "eval_samples_per_second": 11.632, + "eval_wer": 0.4092570036540804, + "step": 42900 + }, + { + "epoch": 6.99, + "learning_rate": 7.2263011910589e-05, + "loss": NaN, + "step": 43000 + }, + { + "epoch": 6.99, + "eval_loss": 0.3986141085624695, + "eval_runtime": 10.2183, + "eval_samples_per_second": 9.786, + "eval_wer": 0.37149817295980514, + "step": 43000 + }, + { + "epoch": 7.01, + "learning_rate": 7.219774840920217e-05, + "loss": NaN, + "step": 43100 + }, + { + "epoch": 7.01, + "eval_loss": 0.41271498799324036, + "eval_runtime": 8.2539, + "eval_samples_per_second": 12.116, + "eval_wer": 0.39585870889159563, + "step": 43100 + }, + { + "epoch": 7.03, + "learning_rate": 7.213248490781531e-05, + "loss": NaN, + "step": 43200 + }, + { + "epoch": 7.03, + "eval_loss": 0.42347651720046997, + "eval_runtime": 6.7075, + "eval_samples_per_second": 14.909, + "eval_wer": 0.3848964677222899, + "step": 43200 + }, + { + "epoch": 7.04, + "learning_rate": 7.206722140642846e-05, + "loss": NaN, + "step": 43300 + }, + { + "epoch": 7.04, + "eval_loss": 0.4063185453414917, + "eval_runtime": 7.6729, + "eval_samples_per_second": 13.033, + "eval_wer": 0.36662606577344703, + "step": 43300 + }, + { + "epoch": 7.06, + "learning_rate": 7.20019579050416e-05, + "loss": NaN, + "step": 43400 + }, + { + "epoch": 7.06, + "eval_loss": 0.49583899974823, + "eval_runtime": 6.9774, + "eval_samples_per_second": 14.332, + "eval_wer": 0.38855054811205847, + "step": 43400 + }, + { + "epoch": 7.07, + "learning_rate": 7.193669440365476e-05, + "loss": NaN, + "step": 43500 + }, + { + "epoch": 7.07, + "eval_loss": 0.4709340035915375, + "eval_runtime": 7.0993, + "eval_samples_per_second": 14.086, + "eval_wer": 0.4056029232643118, + "step": 43500 + }, + { + "epoch": 7.09, + "learning_rate": 7.18714309022679e-05, + "loss": NaN, + "step": 43600 + }, + { + "epoch": 7.09, + "eval_loss": 0.45330917835235596, + "eval_runtime": 6.7893, + "eval_samples_per_second": 14.729, + "eval_wer": 0.3702801461632156, + "step": 43600 + }, + { + "epoch": 7.11, + "learning_rate": 7.180616740088106e-05, + "loss": NaN, + "step": 43700 + }, + { + "epoch": 7.11, + "eval_loss": 0.444570392370224, + "eval_runtime": 7.3413, + "eval_samples_per_second": 13.622, + "eval_wer": 0.4043848964677223, + "step": 43700 + }, + { + "epoch": 7.12, + "learning_rate": 7.174090389949422e-05, + "loss": NaN, + "step": 43800 + }, + { + "epoch": 7.12, + "eval_loss": 0.42336639761924744, + "eval_runtime": 7.0892, + "eval_samples_per_second": 14.106, + "eval_wer": 0.3690621193666261, + "step": 43800 + }, + { + "epoch": 7.14, + "learning_rate": 7.167564039810735e-05, + "loss": NaN, + "step": 43900 + }, + { + "epoch": 7.14, + "eval_loss": 0.4557478427886963, + "eval_runtime": 9.6857, + "eval_samples_per_second": 10.325, + "eval_wer": 0.38246041412911086, + "step": 43900 + }, + { + "epoch": 7.16, + "learning_rate": 7.161037689672051e-05, + "loss": NaN, + "step": 44000 + }, + { + "epoch": 7.16, + "eval_loss": 0.4820619225502014, + "eval_runtime": 8.8382, + "eval_samples_per_second": 11.314, + "eval_wer": 0.37393422655298414, + "step": 44000 + }, + { + "epoch": 7.17, + "learning_rate": 7.154511339533366e-05, + "loss": NaN, + "step": 44100 + }, + { + "epoch": 7.17, + "eval_loss": 0.4444971978664398, + "eval_runtime": 9.4662, + "eval_samples_per_second": 10.564, + "eval_wer": 0.38976857490864797, + "step": 44100 + }, + { + "epoch": 7.19, + "learning_rate": 7.147984989394682e-05, + "loss": NaN, + "step": 44200 + }, + { + "epoch": 7.19, + "eval_loss": 0.4306584596633911, + "eval_runtime": 9.2167, + "eval_samples_per_second": 10.85, + "eval_wer": 0.37758830694275275, + "step": 44200 + }, + { + "epoch": 7.2, + "learning_rate": 7.141458639255996e-05, + "loss": NaN, + "step": 44300 + }, + { + "epoch": 7.2, + "eval_loss": 0.43316277861595154, + "eval_runtime": 6.9757, + "eval_samples_per_second": 14.336, + "eval_wer": 0.364190012180268, + "step": 44300 + }, + { + "epoch": 7.22, + "learning_rate": 7.13493228911731e-05, + "loss": NaN, + "step": 44400 + }, + { + "epoch": 7.22, + "eval_loss": 0.4009106755256653, + "eval_runtime": 7.8376, + "eval_samples_per_second": 12.759, + "eval_wer": 0.3629719853836784, + "step": 44400 + }, + { + "epoch": 7.24, + "learning_rate": 7.128405938978626e-05, + "loss": NaN, + "step": 44500 + }, + { + "epoch": 7.24, + "eval_loss": 0.3961714804172516, + "eval_runtime": 7.8289, + "eval_samples_per_second": 12.773, + "eval_wer": 0.3629719853836784, + "step": 44500 + }, + { + "epoch": 7.25, + "learning_rate": 7.121879588839941e-05, + "loss": NaN, + "step": 44600 + }, + { + "epoch": 7.25, + "eval_loss": 0.4771730303764343, + "eval_runtime": 10.2069, + "eval_samples_per_second": 9.797, + "eval_wer": 0.3861144945188794, + "step": 44600 + }, + { + "epoch": 7.27, + "learning_rate": 7.115353238701257e-05, + "loss": NaN, + "step": 44700 + }, + { + "epoch": 7.27, + "eval_loss": 0.41656962037086487, + "eval_runtime": 8.6363, + "eval_samples_per_second": 11.579, + "eval_wer": 0.3751522533495737, + "step": 44700 + }, + { + "epoch": 7.29, + "learning_rate": 7.108826888562571e-05, + "loss": NaN, + "step": 44800 + }, + { + "epoch": 7.29, + "eval_loss": 0.4536881744861603, + "eval_runtime": 6.9673, + "eval_samples_per_second": 14.353, + "eval_wer": 0.3909866017052375, + "step": 44800 + }, + { + "epoch": 7.3, + "learning_rate": 7.102300538423887e-05, + "loss": NaN, + "step": 44900 + }, + { + "epoch": 7.3, + "eval_loss": 0.4048987627029419, + "eval_runtime": 7.9455, + "eval_samples_per_second": 12.586, + "eval_wer": 0.37758830694275275, + "step": 44900 + }, + { + "epoch": 7.32, + "learning_rate": 7.095774188285202e-05, + "loss": NaN, + "step": 45000 + }, + { + "epoch": 7.32, + "eval_loss": 0.4246349334716797, + "eval_runtime": 7.4556, + "eval_samples_per_second": 13.413, + "eval_wer": 0.3861144945188794, + "step": 45000 + }, + { + "epoch": 7.33, + "learning_rate": 7.089247838146516e-05, + "loss": NaN, + "step": 45100 + }, + { + "epoch": 7.33, + "eval_loss": 0.49522098898887634, + "eval_runtime": 6.8596, + "eval_samples_per_second": 14.578, + "eval_wer": 0.38246041412911086, + "step": 45100 + }, + { + "epoch": 7.35, + "learning_rate": 7.082721488007832e-05, + "loss": NaN, + "step": 45200 + }, + { + "epoch": 7.35, + "eval_loss": 0.47712311148643494, + "eval_runtime": 9.0655, + "eval_samples_per_second": 11.031, + "eval_wer": 0.40194884287454324, + "step": 45200 + }, + { + "epoch": 7.37, + "learning_rate": 7.076195137869147e-05, + "loss": NaN, + "step": 45300 + }, + { + "epoch": 7.37, + "eval_loss": 0.407779723405838, + "eval_runtime": 8.3949, + "eval_samples_per_second": 11.912, + "eval_wer": 0.3861144945188794, + "step": 45300 + }, + { + "epoch": 7.38, + "learning_rate": 7.069668787730463e-05, + "loss": NaN, + "step": 45400 + }, + { + "epoch": 7.38, + "eval_loss": 0.42167598009109497, + "eval_runtime": 7.6909, + "eval_samples_per_second": 13.002, + "eval_wer": 0.38976857490864797, + "step": 45400 + }, + { + "epoch": 7.4, + "learning_rate": 7.063142437591777e-05, + "loss": NaN, + "step": 45500 + }, + { + "epoch": 7.4, + "eval_loss": 0.46571946144104004, + "eval_runtime": 8.1474, + "eval_samples_per_second": 12.274, + "eval_wer": 0.3934226552984166, + "step": 45500 + }, + { + "epoch": 7.42, + "learning_rate": 7.056616087453092e-05, + "loss": NaN, + "step": 45600 + }, + { + "epoch": 7.42, + "eval_loss": 0.40125495195388794, + "eval_runtime": 7.1588, + "eval_samples_per_second": 13.969, + "eval_wer": 0.4007308160779537, + "step": 45600 + }, + { + "epoch": 7.43, + "learning_rate": 7.050089737314407e-05, + "loss": NaN, + "step": 45700 + }, + { + "epoch": 7.43, + "eval_loss": 0.41013047099113464, + "eval_runtime": 8.2392, + "eval_samples_per_second": 12.137, + "eval_wer": 0.39707673568818513, + "step": 45700 + }, + { + "epoch": 7.45, + "learning_rate": 7.043563387175722e-05, + "loss": NaN, + "step": 45800 + }, + { + "epoch": 7.45, + "eval_loss": 0.3986717164516449, + "eval_runtime": 6.5584, + "eval_samples_per_second": 15.248, + "eval_wer": 0.37393422655298414, + "step": 45800 + }, + { + "epoch": 7.46, + "learning_rate": 7.037037037037038e-05, + "loss": NaN, + "step": 45900 + }, + { + "epoch": 7.46, + "eval_loss": 0.3988329768180847, + "eval_runtime": 8.4951, + "eval_samples_per_second": 11.771, + "eval_wer": 0.3763702801461632, + "step": 45900 + }, + { + "epoch": 7.48, + "learning_rate": 7.030510686898352e-05, + "loss": NaN, + "step": 46000 + }, + { + "epoch": 7.48, + "eval_loss": 0.45179134607315063, + "eval_runtime": 9.3929, + "eval_samples_per_second": 10.646, + "eval_wer": 0.3800243605359318, + "step": 46000 + }, + { + "epoch": 7.5, + "learning_rate": 7.023984336759667e-05, + "loss": NaN, + "step": 46100 + }, + { + "epoch": 7.5, + "eval_loss": 0.4907733201980591, + "eval_runtime": 8.1588, + "eval_samples_per_second": 12.257, + "eval_wer": 0.41169305724725946, + "step": 46100 + }, + { + "epoch": 7.51, + "learning_rate": 7.017457986620983e-05, + "loss": NaN, + "step": 46200 + }, + { + "epoch": 7.51, + "eval_loss": 0.4354599416255951, + "eval_runtime": 7.4922, + "eval_samples_per_second": 13.347, + "eval_wer": 0.3909866017052375, + "step": 46200 + }, + { + "epoch": 7.53, + "learning_rate": 7.010931636482297e-05, + "loss": NaN, + "step": 46300 + }, + { + "epoch": 7.53, + "eval_loss": 0.4466521441936493, + "eval_runtime": 10.9543, + "eval_samples_per_second": 9.129, + "eval_wer": 0.3909866017052375, + "step": 46300 + }, + { + "epoch": 7.55, + "learning_rate": 7.004405286343613e-05, + "loss": NaN, + "step": 46400 + }, + { + "epoch": 7.55, + "eval_loss": 0.4440977871417999, + "eval_runtime": 9.6264, + "eval_samples_per_second": 10.388, + "eval_wer": 0.40316686967113274, + "step": 46400 + }, + { + "epoch": 7.56, + "learning_rate": 6.997878936204928e-05, + "loss": NaN, + "step": 46500 + }, + { + "epoch": 7.56, + "eval_loss": 0.40619930624961853, + "eval_runtime": 8.2096, + "eval_samples_per_second": 12.181, + "eval_wer": 0.37149817295980514, + "step": 46500 + }, + { + "epoch": 7.58, + "learning_rate": 6.991352586066243e-05, + "loss": NaN, + "step": 46600 + }, + { + "epoch": 7.58, + "eval_loss": 0.4126388430595398, + "eval_runtime": 8.8109, + "eval_samples_per_second": 11.35, + "eval_wer": 0.3909866017052375, + "step": 46600 + }, + { + "epoch": 7.59, + "learning_rate": 6.984826235927558e-05, + "loss": NaN, + "step": 46700 + }, + { + "epoch": 7.59, + "eval_loss": 0.41685402393341064, + "eval_runtime": 8.6911, + "eval_samples_per_second": 11.506, + "eval_wer": 0.3800243605359318, + "step": 46700 + }, + { + "epoch": 7.61, + "learning_rate": 6.978299885788873e-05, + "loss": NaN, + "step": 46800 + }, + { + "epoch": 7.61, + "eval_loss": 0.42471861839294434, + "eval_runtime": 6.7824, + "eval_samples_per_second": 14.744, + "eval_wer": 0.39585870889159563, + "step": 46800 + }, + { + "epoch": 7.63, + "learning_rate": 6.971773535650188e-05, + "loss": NaN, + "step": 46900 + }, + { + "epoch": 7.63, + "eval_loss": 0.4128079116344452, + "eval_runtime": 7.183, + "eval_samples_per_second": 13.922, + "eval_wer": 0.3946406820950061, + "step": 46900 + }, + { + "epoch": 7.64, + "learning_rate": 6.965247185511503e-05, + "loss": NaN, + "step": 47000 + }, + { + "epoch": 7.64, + "eval_loss": 0.3996775448322296, + "eval_runtime": 8.3343, + "eval_samples_per_second": 11.999, + "eval_wer": 0.3654080389768575, + "step": 47000 + }, + { + "epoch": 7.66, + "learning_rate": 6.958720835372819e-05, + "loss": NaN, + "step": 47100 + }, + { + "epoch": 7.66, + "eval_loss": 0.3690239191055298, + "eval_runtime": 6.5963, + "eval_samples_per_second": 15.16, + "eval_wer": 0.37880633373934225, + "step": 47100 + }, + { + "epoch": 7.68, + "learning_rate": 6.952194485234133e-05, + "loss": NaN, + "step": 47200 + }, + { + "epoch": 7.68, + "eval_loss": 0.42326632142066956, + "eval_runtime": 8.3908, + "eval_samples_per_second": 11.918, + "eval_wer": 0.3909866017052375, + "step": 47200 + }, + { + "epoch": 7.69, + "learning_rate": 6.945668135095448e-05, + "loss": NaN, + "step": 47300 + }, + { + "epoch": 7.69, + "eval_loss": 0.39741602540016174, + "eval_runtime": 8.5855, + "eval_samples_per_second": 11.648, + "eval_wer": 0.36784409257003653, + "step": 47300 + }, + { + "epoch": 7.71, + "learning_rate": 6.939141784956764e-05, + "loss": NaN, + "step": 47400 + }, + { + "epoch": 7.71, + "eval_loss": 0.42626139521598816, + "eval_runtime": 8.2558, + "eval_samples_per_second": 12.113, + "eval_wer": 0.3982947624847747, + "step": 47400 + }, + { + "epoch": 7.72, + "learning_rate": 6.932615434818078e-05, + "loss": NaN, + "step": 47500 + }, + { + "epoch": 7.72, + "eval_loss": 0.41164088249206543, + "eval_runtime": 7.9604, + "eval_samples_per_second": 12.562, + "eval_wer": 0.392204628501827, + "step": 47500 + }, + { + "epoch": 7.74, + "learning_rate": 6.926089084679394e-05, + "loss": NaN, + "step": 47600 + }, + { + "epoch": 7.74, + "eval_loss": 0.41196590662002563, + "eval_runtime": 7.4856, + "eval_samples_per_second": 13.359, + "eval_wer": 0.3654080389768575, + "step": 47600 + }, + { + "epoch": 7.76, + "learning_rate": 6.919562734540709e-05, + "loss": NaN, + "step": 47700 + }, + { + "epoch": 7.76, + "eval_loss": 0.4575519263744354, + "eval_runtime": 7.5753, + "eval_samples_per_second": 13.201, + "eval_wer": 0.392204628501827, + "step": 47700 + }, + { + "epoch": 7.77, + "learning_rate": 6.913036384402023e-05, + "loss": NaN, + "step": 47800 + }, + { + "epoch": 7.77, + "eval_loss": 0.47335246205329895, + "eval_runtime": 8.6796, + "eval_samples_per_second": 11.521, + "eval_wer": 0.40194884287454324, + "step": 47800 + }, + { + "epoch": 7.79, + "learning_rate": 6.906510034263338e-05, + "loss": NaN, + "step": 47900 + }, + { + "epoch": 7.79, + "eval_loss": 0.46512919664382935, + "eval_runtime": 7.9927, + "eval_samples_per_second": 12.511, + "eval_wer": 0.3873325213154689, + "step": 47900 + }, + { + "epoch": 7.81, + "learning_rate": 6.899983684124653e-05, + "loss": NaN, + "step": 48000 + }, + { + "epoch": 7.81, + "eval_loss": 0.45921266078948975, + "eval_runtime": 8.1681, + "eval_samples_per_second": 12.243, + "eval_wer": 0.3946406820950061, + "step": 48000 + }, + { + "epoch": 7.82, + "learning_rate": 6.89345733398597e-05, + "loss": NaN, + "step": 48100 + }, + { + "epoch": 7.82, + "eval_loss": 0.4549075961112976, + "eval_runtime": 9.0376, + "eval_samples_per_second": 11.065, + "eval_wer": 0.38246041412911086, + "step": 48100 + }, + { + "epoch": 7.84, + "learning_rate": 6.886930983847284e-05, + "loss": NaN, + "step": 48200 + }, + { + "epoch": 7.84, + "eval_loss": 0.45289474725723267, + "eval_runtime": 7.9376, + "eval_samples_per_second": 12.598, + "eval_wer": 0.3873325213154689, + "step": 48200 + }, + { + "epoch": 7.85, + "learning_rate": 6.8804046337086e-05, + "loss": NaN, + "step": 48300 + }, + { + "epoch": 7.85, + "eval_loss": 0.4604376256465912, + "eval_runtime": 8.0346, + "eval_samples_per_second": 12.446, + "eval_wer": 0.3861144945188794, + "step": 48300 + }, + { + "epoch": 7.87, + "learning_rate": 6.873878283569913e-05, + "loss": NaN, + "step": 48400 + }, + { + "epoch": 7.87, + "eval_loss": 0.42993414402008057, + "eval_runtime": 9.3655, + "eval_samples_per_second": 10.677, + "eval_wer": 0.38367844092570036, + "step": 48400 + }, + { + "epoch": 7.89, + "learning_rate": 6.867351933431229e-05, + "loss": NaN, + "step": 48500 + }, + { + "epoch": 7.89, + "eval_loss": 0.5003374218940735, + "eval_runtime": 9.3394, + "eval_samples_per_second": 10.707, + "eval_wer": 0.3946406820950061, + "step": 48500 + }, + { + "epoch": 7.9, + "learning_rate": 6.860825583292543e-05, + "loss": NaN, + "step": 48600 + }, + { + "epoch": 7.9, + "eval_loss": 0.4476844072341919, + "eval_runtime": 8.2996, + "eval_samples_per_second": 12.049, + "eval_wer": 0.3982947624847747, + "step": 48600 + }, + { + "epoch": 7.92, + "learning_rate": 6.854299233153859e-05, + "loss": NaN, + "step": 48700 + }, + { + "epoch": 7.92, + "eval_loss": 0.4162767827510834, + "eval_runtime": 7.2206, + "eval_samples_per_second": 13.849, + "eval_wer": 0.3861144945188794, + "step": 48700 + }, + { + "epoch": 7.94, + "learning_rate": 6.847772883015175e-05, + "loss": NaN, + "step": 48800 + }, + { + "epoch": 7.94, + "eval_loss": 0.5044607520103455, + "eval_runtime": 8.7024, + "eval_samples_per_second": 11.491, + "eval_wer": 0.3909866017052375, + "step": 48800 + }, + { + "epoch": 7.95, + "learning_rate": 6.84124653287649e-05, + "loss": NaN, + "step": 48900 + }, + { + "epoch": 7.95, + "eval_loss": 0.42168164253234863, + "eval_runtime": 6.5122, + "eval_samples_per_second": 15.356, + "eval_wer": 0.3800243605359318, + "step": 48900 + }, + { + "epoch": 7.97, + "learning_rate": 6.834720182737804e-05, + "loss": NaN, + "step": 49000 + }, + { + "epoch": 7.97, + "eval_loss": 0.4774813950061798, + "eval_runtime": 8.4021, + "eval_samples_per_second": 11.902, + "eval_wer": 0.3873325213154689, + "step": 49000 + }, + { + "epoch": 7.99, + "learning_rate": 6.828193832599119e-05, + "loss": NaN, + "step": 49100 + }, + { + "epoch": 7.99, + "eval_loss": 0.47412794828414917, + "eval_runtime": 7.8667, + "eval_samples_per_second": 12.712, + "eval_wer": 0.3946406820950061, + "step": 49100 + }, + { + "epoch": 8.0, + "learning_rate": 6.821667482460434e-05, + "loss": NaN, + "step": 49200 + }, + { + "epoch": 8.0, + "eval_loss": 0.4439965784549713, + "eval_runtime": 8.4464, + "eval_samples_per_second": 11.839, + "eval_wer": 0.3873325213154689, + "step": 49200 + }, + { + "epoch": 8.02, + "learning_rate": 6.815141132321749e-05, + "loss": NaN, + "step": 49300 + }, + { + "epoch": 8.02, + "eval_loss": 0.47220298647880554, + "eval_runtime": 8.4551, + "eval_samples_per_second": 11.827, + "eval_wer": 0.38367844092570036, + "step": 49300 + }, + { + "epoch": 8.03, + "learning_rate": 6.808614782183065e-05, + "loss": NaN, + "step": 49400 + }, + { + "epoch": 8.03, + "eval_loss": 0.41892436146736145, + "eval_runtime": 7.9787, + "eval_samples_per_second": 12.533, + "eval_wer": 0.37880633373934225, + "step": 49400 + }, + { + "epoch": 8.05, + "learning_rate": 6.802088432044379e-05, + "loss": NaN, + "step": 49500 + }, + { + "epoch": 8.05, + "eval_loss": 0.4536736309528351, + "eval_runtime": 7.9322, + "eval_samples_per_second": 12.607, + "eval_wer": 0.39585870889159563, + "step": 49500 + }, + { + "epoch": 8.07, + "learning_rate": 6.795562081905694e-05, + "loss": NaN, + "step": 49600 + }, + { + "epoch": 8.07, + "eval_loss": 0.45775213837623596, + "eval_runtime": 8.207, + "eval_samples_per_second": 12.185, + "eval_wer": 0.3873325213154689, + "step": 49600 + }, + { + "epoch": 8.08, + "learning_rate": 6.78903573176701e-05, + "loss": NaN, + "step": 49700 + }, + { + "epoch": 8.08, + "eval_loss": 0.44402602314949036, + "eval_runtime": 8.5892, + "eval_samples_per_second": 11.643, + "eval_wer": 0.38976857490864797, + "step": 49700 + }, + { + "epoch": 8.1, + "learning_rate": 6.782509381628324e-05, + "loss": NaN, + "step": 49800 + }, + { + "epoch": 8.1, + "eval_loss": 0.46547335386276245, + "eval_runtime": 9.3683, + "eval_samples_per_second": 10.674, + "eval_wer": 0.3934226552984166, + "step": 49800 + }, + { + "epoch": 8.12, + "learning_rate": 6.77598303148964e-05, + "loss": NaN, + "step": 49900 + }, + { + "epoch": 8.12, + "eval_loss": 0.46032240986824036, + "eval_runtime": 8.8425, + "eval_samples_per_second": 11.309, + "eval_wer": 0.3848964677222899, + "step": 49900 + }, + { + "epoch": 8.13, + "learning_rate": 6.769456681350955e-05, + "loss": NaN, + "step": 50000 + }, + { + "epoch": 8.13, + "eval_loss": 0.4585595726966858, + "eval_runtime": 8.5708, + "eval_samples_per_second": 11.667, + "eval_wer": 0.3848964677222899, + "step": 50000 + }, + { + "epoch": 8.15, + "learning_rate": 6.762930331212269e-05, + "loss": NaN, + "step": 50100 + }, + { + "epoch": 8.15, + "eval_loss": 0.41307058930397034, + "eval_runtime": 7.0952, + "eval_samples_per_second": 14.094, + "eval_wer": 0.37271619975639464, + "step": 50100 + }, + { + "epoch": 8.16, + "learning_rate": 6.756403981073585e-05, + "loss": NaN, + "step": 50200 + }, + { + "epoch": 8.16, + "eval_loss": 0.41041550040245056, + "eval_runtime": 7.1426, + "eval_samples_per_second": 14.001, + "eval_wer": 0.36784409257003653, + "step": 50200 + }, + { + "epoch": 8.18, + "learning_rate": 6.7498776309349e-05, + "loss": NaN, + "step": 50300 + }, + { + "epoch": 8.18, + "eval_loss": 0.44717052578926086, + "eval_runtime": 7.0047, + "eval_samples_per_second": 14.276, + "eval_wer": 0.39707673568818513, + "step": 50300 + }, + { + "epoch": 8.2, + "learning_rate": 6.743351280796215e-05, + "loss": NaN, + "step": 50400 + }, + { + "epoch": 8.2, + "eval_loss": 0.5247046947479248, + "eval_runtime": 7.6638, + "eval_samples_per_second": 13.048, + "eval_wer": 0.3982947624847747, + "step": 50400 + }, + { + "epoch": 8.21, + "learning_rate": 6.73682493065753e-05, + "loss": NaN, + "step": 50500 + }, + { + "epoch": 8.21, + "eval_loss": 0.5030866861343384, + "eval_runtime": 7.6449, + "eval_samples_per_second": 13.081, + "eval_wer": 0.38246041412911086, + "step": 50500 + }, + { + "epoch": 8.23, + "learning_rate": 6.730298580518846e-05, + "loss": NaN, + "step": 50600 + }, + { + "epoch": 8.23, + "eval_loss": 0.4644200801849365, + "eval_runtime": 7.0234, + "eval_samples_per_second": 14.238, + "eval_wer": 0.364190012180268, + "step": 50600 + }, + { + "epoch": 8.25, + "learning_rate": 6.72377223038016e-05, + "loss": NaN, + "step": 50700 + }, + { + "epoch": 8.25, + "eval_loss": 0.4306560158729553, + "eval_runtime": 8.0867, + "eval_samples_per_second": 12.366, + "eval_wer": 0.3800243605359318, + "step": 50700 + }, + { + "epoch": 8.26, + "learning_rate": 6.717245880241475e-05, + "loss": NaN, + "step": 50800 + }, + { + "epoch": 8.26, + "eval_loss": 0.45848017930984497, + "eval_runtime": 8.7036, + "eval_samples_per_second": 11.489, + "eval_wer": 0.39707673568818513, + "step": 50800 + }, + { + "epoch": 8.28, + "learning_rate": 6.71071953010279e-05, + "loss": NaN, + "step": 50900 + }, + { + "epoch": 8.28, + "eval_loss": 0.5098088979721069, + "eval_runtime": 7.2954, + "eval_samples_per_second": 13.707, + "eval_wer": 0.3861144945188794, + "step": 50900 + }, + { + "epoch": 8.29, + "learning_rate": 6.704193179964105e-05, + "loss": NaN, + "step": 51000 + }, + { + "epoch": 8.29, + "eval_loss": 0.5031391382217407, + "eval_runtime": 7.9446, + "eval_samples_per_second": 12.587, + "eval_wer": 0.3934226552984166, + "step": 51000 + }, + { + "epoch": 8.31, + "learning_rate": 6.697666829825421e-05, + "loss": NaN, + "step": 51100 + }, + { + "epoch": 8.31, + "eval_loss": 0.455885648727417, + "eval_runtime": 8.4028, + "eval_samples_per_second": 11.901, + "eval_wer": 0.392204628501827, + "step": 51100 + }, + { + "epoch": 8.33, + "learning_rate": 6.691140479686736e-05, + "loss": NaN, + "step": 51200 + }, + { + "epoch": 8.33, + "eval_loss": 0.4358394145965576, + "eval_runtime": 9.3301, + "eval_samples_per_second": 10.718, + "eval_wer": 0.3702801461632156, + "step": 51200 + }, + { + "epoch": 8.34, + "learning_rate": 6.68461412954805e-05, + "loss": NaN, + "step": 51300 + }, + { + "epoch": 8.34, + "eval_loss": 0.4508705139160156, + "eval_runtime": 7.2511, + "eval_samples_per_second": 13.791, + "eval_wer": 0.37880633373934225, + "step": 51300 + }, + { + "epoch": 8.36, + "learning_rate": 6.678087779409366e-05, + "loss": NaN, + "step": 51400 + }, + { + "epoch": 8.36, + "eval_loss": 0.5324064493179321, + "eval_runtime": 7.3965, + "eval_samples_per_second": 13.52, + "eval_wer": 0.37880633373934225, + "step": 51400 + }, + { + "epoch": 8.38, + "learning_rate": 6.67156142927068e-05, + "loss": NaN, + "step": 51500 + }, + { + "epoch": 8.38, + "eval_loss": 0.4332933723926544, + "eval_runtime": 7.9068, + "eval_samples_per_second": 12.647, + "eval_wer": 0.3751522533495737, + "step": 51500 + }, + { + "epoch": 8.39, + "learning_rate": 6.665035079131996e-05, + "loss": NaN, + "step": 51600 + }, + { + "epoch": 8.39, + "eval_loss": 0.5386966466903687, + "eval_runtime": 6.9317, + "eval_samples_per_second": 14.426, + "eval_wer": 0.364190012180268, + "step": 51600 + }, + { + "epoch": 8.41, + "learning_rate": 6.658508728993311e-05, + "loss": NaN, + "step": 51700 + }, + { + "epoch": 8.41, + "eval_loss": 0.47883322834968567, + "eval_runtime": 8.0805, + "eval_samples_per_second": 12.376, + "eval_wer": 0.3861144945188794, + "step": 51700 + }, + { + "epoch": 8.42, + "learning_rate": 6.651982378854625e-05, + "loss": NaN, + "step": 51800 + }, + { + "epoch": 8.42, + "eval_loss": 0.49707546830177307, + "eval_runtime": 6.9503, + "eval_samples_per_second": 14.388, + "eval_wer": 0.3763702801461632, + "step": 51800 + }, + { + "epoch": 8.44, + "learning_rate": 6.645456028715941e-05, + "loss": NaN, + "step": 51900 + }, + { + "epoch": 8.44, + "eval_loss": 0.4724676012992859, + "eval_runtime": 8.5223, + "eval_samples_per_second": 11.734, + "eval_wer": 0.37758830694275275, + "step": 51900 + }, + { + "epoch": 8.46, + "learning_rate": 6.638929678577256e-05, + "loss": NaN, + "step": 52000 + }, + { + "epoch": 8.46, + "eval_loss": 0.49637308716773987, + "eval_runtime": 6.8621, + "eval_samples_per_second": 14.573, + "eval_wer": 0.37393422655298414, + "step": 52000 + }, + { + "epoch": 8.47, + "learning_rate": 6.632403328438572e-05, + "loss": NaN, + "step": 52100 + }, + { + "epoch": 8.47, + "eval_loss": 0.46207770705223083, + "eval_runtime": 9.6685, + "eval_samples_per_second": 10.343, + "eval_wer": 0.3812423873325213, + "step": 52100 + }, + { + "epoch": 8.49, + "learning_rate": 6.625876978299886e-05, + "loss": NaN, + "step": 52200 + }, + { + "epoch": 8.49, + "eval_loss": 0.47856566309928894, + "eval_runtime": 9.3206, + "eval_samples_per_second": 10.729, + "eval_wer": 0.3751522533495737, + "step": 52200 + }, + { + "epoch": 8.51, + "learning_rate": 6.619350628161202e-05, + "loss": NaN, + "step": 52300 + }, + { + "epoch": 8.51, + "eval_loss": 0.49372223019599915, + "eval_runtime": 8.558, + "eval_samples_per_second": 11.685, + "eval_wer": 0.3873325213154689, + "step": 52300 + }, + { + "epoch": 8.52, + "learning_rate": 6.612824278022516e-05, + "loss": NaN, + "step": 52400 + }, + { + "epoch": 8.52, + "eval_loss": 0.4097326695919037, + "eval_runtime": 7.8765, + "eval_samples_per_second": 12.696, + "eval_wer": 0.3568818514007308, + "step": 52400 + }, + { + "epoch": 8.54, + "learning_rate": 6.606297927883831e-05, + "loss": NaN, + "step": 52500 + }, + { + "epoch": 8.54, + "eval_loss": 0.510248601436615, + "eval_runtime": 8.7686, + "eval_samples_per_second": 11.404, + "eval_wer": 0.3751522533495737, + "step": 52500 + }, + { + "epoch": 8.55, + "learning_rate": 6.599771577745147e-05, + "loss": NaN, + "step": 52600 + }, + { + "epoch": 8.55, + "eval_loss": 0.4330621063709259, + "eval_runtime": 7.4599, + "eval_samples_per_second": 13.405, + "eval_wer": 0.36662606577344703, + "step": 52600 + }, + { + "epoch": 8.57, + "learning_rate": 6.593245227606461e-05, + "loss": NaN, + "step": 52700 + }, + { + "epoch": 8.57, + "eval_loss": 0.4965771436691284, + "eval_runtime": 9.6722, + "eval_samples_per_second": 10.339, + "eval_wer": 0.392204628501827, + "step": 52700 + }, + { + "epoch": 8.59, + "learning_rate": 6.586718877467777e-05, + "loss": NaN, + "step": 52800 + }, + { + "epoch": 8.59, + "eval_loss": 0.5258552432060242, + "eval_runtime": 8.3154, + "eval_samples_per_second": 12.026, + "eval_wer": 0.3934226552984166, + "step": 52800 + }, + { + "epoch": 8.6, + "learning_rate": 6.58019252732909e-05, + "loss": NaN, + "step": 52900 + }, + { + "epoch": 8.6, + "eval_loss": 0.47305095195770264, + "eval_runtime": 8.8487, + "eval_samples_per_second": 11.301, + "eval_wer": 0.3873325213154689, + "step": 52900 + }, + { + "epoch": 8.62, + "learning_rate": 6.573666177190406e-05, + "loss": NaN, + "step": 53000 + }, + { + "epoch": 8.62, + "eval_loss": 0.5063592791557312, + "eval_runtime": 8.5975, + "eval_samples_per_second": 11.631, + "eval_wer": 0.39585870889159563, + "step": 53000 + }, + { + "epoch": 8.64, + "learning_rate": 6.567139827051722e-05, + "loss": NaN, + "step": 53100 + }, + { + "epoch": 8.64, + "eval_loss": 0.5428659319877625, + "eval_runtime": 8.4203, + "eval_samples_per_second": 11.876, + "eval_wer": 0.3800243605359318, + "step": 53100 + }, + { + "epoch": 8.65, + "learning_rate": 6.560613476913037e-05, + "loss": NaN, + "step": 53200 + }, + { + "epoch": 8.65, + "eval_loss": 0.44813072681427, + "eval_runtime": 8.2801, + "eval_samples_per_second": 12.077, + "eval_wer": 0.37880633373934225, + "step": 53200 + }, + { + "epoch": 8.67, + "learning_rate": 6.554087126774353e-05, + "loss": NaN, + "step": 53300 + }, + { + "epoch": 8.67, + "eval_loss": 0.4521804451942444, + "eval_runtime": 8.2143, + "eval_samples_per_second": 12.174, + "eval_wer": 0.37393422655298414, + "step": 53300 + }, + { + "epoch": 8.68, + "learning_rate": 6.547560776635667e-05, + "loss": NaN, + "step": 53400 + }, + { + "epoch": 8.68, + "eval_loss": 0.45459070801734924, + "eval_runtime": 8.9144, + "eval_samples_per_second": 11.218, + "eval_wer": 0.37149817295980514, + "step": 53400 + }, + { + "epoch": 8.7, + "learning_rate": 6.541034426496982e-05, + "loss": NaN, + "step": 53500 + }, + { + "epoch": 8.7, + "eval_loss": 0.5054107904434204, + "eval_runtime": 6.4804, + "eval_samples_per_second": 15.431, + "eval_wer": 0.3763702801461632, + "step": 53500 + }, + { + "epoch": 8.72, + "learning_rate": 6.534508076358296e-05, + "loss": NaN, + "step": 53600 + }, + { + "epoch": 8.72, + "eval_loss": 0.4579046666622162, + "eval_runtime": 6.6279, + "eval_samples_per_second": 15.088, + "eval_wer": 0.3702801461632156, + "step": 53600 + }, + { + "epoch": 8.73, + "learning_rate": 6.527981726219612e-05, + "loss": NaN, + "step": 53700 + }, + { + "epoch": 8.73, + "eval_loss": 0.4322802722454071, + "eval_runtime": 9.2316, + "eval_samples_per_second": 10.832, + "eval_wer": 0.3763702801461632, + "step": 53700 + }, + { + "epoch": 8.75, + "learning_rate": 6.521455376080928e-05, + "loss": NaN, + "step": 53800 + }, + { + "epoch": 8.75, + "eval_loss": 0.42796722054481506, + "eval_runtime": 7.68, + "eval_samples_per_second": 13.021, + "eval_wer": 0.37271619975639464, + "step": 53800 + }, + { + "epoch": 8.77, + "learning_rate": 6.514929025942242e-05, + "loss": NaN, + "step": 53900 + }, + { + "epoch": 8.77, + "eval_loss": 0.44873932003974915, + "eval_runtime": 7.9481, + "eval_samples_per_second": 12.582, + "eval_wer": 0.38855054811205847, + "step": 53900 + }, + { + "epoch": 8.78, + "learning_rate": 6.508402675803558e-05, + "loss": NaN, + "step": 54000 + }, + { + "epoch": 8.78, + "eval_loss": 0.4511549472808838, + "eval_runtime": 9.1786, + "eval_samples_per_second": 10.895, + "eval_wer": 0.3654080389768575, + "step": 54000 + }, + { + "epoch": 8.8, + "learning_rate": 6.501876325664871e-05, + "loss": NaN, + "step": 54100 + }, + { + "epoch": 8.8, + "eval_loss": 0.42401379346847534, + "eval_runtime": 7.9905, + "eval_samples_per_second": 12.515, + "eval_wer": 0.3629719853836784, + "step": 54100 + }, + { + "epoch": 8.81, + "learning_rate": 6.495349975526187e-05, + "loss": NaN, + "step": 54200 + }, + { + "epoch": 8.81, + "eval_loss": 0.4265538156032562, + "eval_runtime": 9.2408, + "eval_samples_per_second": 10.822, + "eval_wer": 0.3751522533495737, + "step": 54200 + }, + { + "epoch": 8.83, + "learning_rate": 6.488823625387502e-05, + "loss": NaN, + "step": 54300 + }, + { + "epoch": 8.83, + "eval_loss": 0.505720853805542, + "eval_runtime": 8.1012, + "eval_samples_per_second": 12.344, + "eval_wer": 0.37393422655298414, + "step": 54300 + }, + { + "epoch": 8.85, + "learning_rate": 6.482297275248818e-05, + "loss": NaN, + "step": 54400 + }, + { + "epoch": 8.85, + "eval_loss": 0.4096541702747345, + "eval_runtime": 7.7623, + "eval_samples_per_second": 12.883, + "eval_wer": 0.37149817295980514, + "step": 54400 + }, + { + "epoch": 8.86, + "learning_rate": 6.475770925110133e-05, + "loss": NaN, + "step": 54500 + }, + { + "epoch": 8.86, + "eval_loss": 0.4086014926433563, + "eval_runtime": 8.886, + "eval_samples_per_second": 11.254, + "eval_wer": 0.3763702801461632, + "step": 54500 + }, + { + "epoch": 8.88, + "learning_rate": 6.469244574971447e-05, + "loss": NaN, + "step": 54600 + }, + { + "epoch": 8.88, + "eval_loss": 0.43217840790748596, + "eval_runtime": 7.9198, + "eval_samples_per_second": 12.627, + "eval_wer": 0.36662606577344703, + "step": 54600 + }, + { + "epoch": 8.9, + "learning_rate": 6.462718224832762e-05, + "loss": NaN, + "step": 54700 + }, + { + "epoch": 8.9, + "eval_loss": 0.428193062543869, + "eval_runtime": 8.0067, + "eval_samples_per_second": 12.49, + "eval_wer": 0.3873325213154689, + "step": 54700 + }, + { + "epoch": 8.91, + "learning_rate": 6.456191874694077e-05, + "loss": NaN, + "step": 54800 + }, + { + "epoch": 8.91, + "eval_loss": 0.43840643763542175, + "eval_runtime": 8.3593, + "eval_samples_per_second": 11.963, + "eval_wer": 0.37393422655298414, + "step": 54800 + }, + { + "epoch": 8.93, + "learning_rate": 6.449665524555393e-05, + "loss": NaN, + "step": 54900 + }, + { + "epoch": 8.93, + "eval_loss": 0.41468119621276855, + "eval_runtime": 7.1084, + "eval_samples_per_second": 14.068, + "eval_wer": 0.36784409257003653, + "step": 54900 + }, + { + "epoch": 8.94, + "learning_rate": 6.443139174416707e-05, + "loss": NaN, + "step": 55000 + }, + { + "epoch": 8.94, + "eval_loss": 0.4655587673187256, + "eval_runtime": 7.4051, + "eval_samples_per_second": 13.504, + "eval_wer": 0.364190012180268, + "step": 55000 + }, + { + "epoch": 8.96, + "learning_rate": 6.436612824278023e-05, + "loss": NaN, + "step": 55100 + }, + { + "epoch": 8.96, + "eval_loss": 0.4038917124271393, + "eval_runtime": 8.3584, + "eval_samples_per_second": 11.964, + "eval_wer": 0.36784409257003653, + "step": 55100 + }, + { + "epoch": 8.98, + "learning_rate": 6.430086474139338e-05, + "loss": NaN, + "step": 55200 + }, + { + "epoch": 8.98, + "eval_loss": 0.39275607466697693, + "eval_runtime": 7.452, + "eval_samples_per_second": 13.419, + "eval_wer": 0.36784409257003653, + "step": 55200 + }, + { + "epoch": 8.99, + "learning_rate": 6.423560124000652e-05, + "loss": NaN, + "step": 55300 + }, + { + "epoch": 8.99, + "eval_loss": 0.4186645448207855, + "eval_runtime": 8.5472, + "eval_samples_per_second": 11.7, + "eval_wer": 0.364190012180268, + "step": 55300 + }, + { + "epoch": 9.01, + "learning_rate": 6.417033773861968e-05, + "loss": NaN, + "step": 55400 + }, + { + "epoch": 9.01, + "eval_loss": 0.44433704018592834, + "eval_runtime": 7.986, + "eval_samples_per_second": 12.522, + "eval_wer": 0.36662606577344703, + "step": 55400 + }, + { + "epoch": 9.03, + "learning_rate": 6.410507423723283e-05, + "loss": NaN, + "step": 55500 + }, + { + "epoch": 9.03, + "eval_loss": 0.43365246057510376, + "eval_runtime": 7.3226, + "eval_samples_per_second": 13.656, + "eval_wer": 0.35931790499390986, + "step": 55500 + }, + { + "epoch": 9.04, + "learning_rate": 6.403981073584599e-05, + "loss": NaN, + "step": 55600 + }, + { + "epoch": 9.04, + "eval_loss": 0.4543744623661041, + "eval_runtime": 8.7477, + "eval_samples_per_second": 11.432, + "eval_wer": 0.37271619975639464, + "step": 55600 + }, + { + "epoch": 9.06, + "learning_rate": 6.397454723445913e-05, + "loss": NaN, + "step": 55700 + }, + { + "epoch": 9.06, + "eval_loss": 0.43953046202659607, + "eval_runtime": 7.711, + "eval_samples_per_second": 12.968, + "eval_wer": 0.3934226552984166, + "step": 55700 + }, + { + "epoch": 9.07, + "learning_rate": 6.390928373307228e-05, + "loss": NaN, + "step": 55800 + }, + { + "epoch": 9.07, + "eval_loss": 0.43022480607032776, + "eval_runtime": 7.9745, + "eval_samples_per_second": 12.54, + "eval_wer": 0.36053593179049936, + "step": 55800 + }, + { + "epoch": 9.09, + "learning_rate": 6.384402023168543e-05, + "loss": NaN, + "step": 55900 + }, + { + "epoch": 9.09, + "eval_loss": 0.4893076419830322, + "eval_runtime": 8.1627, + "eval_samples_per_second": 12.251, + "eval_wer": 0.3507917174177832, + "step": 55900 + }, + { + "epoch": 9.11, + "learning_rate": 6.377875673029858e-05, + "loss": NaN, + "step": 56000 + }, + { + "epoch": 9.11, + "eval_loss": 0.4135149419307709, + "eval_runtime": 7.9225, + "eval_samples_per_second": 12.622, + "eval_wer": 0.3763702801461632, + "step": 56000 + }, + { + "epoch": 9.12, + "learning_rate": 6.371349322891174e-05, + "loss": NaN, + "step": 56100 + }, + { + "epoch": 9.12, + "eval_loss": 0.4469653069972992, + "eval_runtime": 8.096, + "eval_samples_per_second": 12.352, + "eval_wer": 0.3800243605359318, + "step": 56100 + }, + { + "epoch": 9.14, + "learning_rate": 6.364822972752488e-05, + "loss": NaN, + "step": 56200 + }, + { + "epoch": 9.14, + "eval_loss": 0.42649006843566895, + "eval_runtime": 8.89, + "eval_samples_per_second": 11.249, + "eval_wer": 0.3690621193666261, + "step": 56200 + }, + { + "epoch": 9.16, + "learning_rate": 6.358296622613803e-05, + "loss": NaN, + "step": 56300 + }, + { + "epoch": 9.16, + "eval_loss": 0.46478691697120667, + "eval_runtime": 7.1008, + "eval_samples_per_second": 14.083, + "eval_wer": 0.36784409257003653, + "step": 56300 + }, + { + "epoch": 9.17, + "learning_rate": 6.351770272475119e-05, + "loss": NaN, + "step": 56400 + }, + { + "epoch": 9.17, + "eval_loss": 0.45123374462127686, + "eval_runtime": 8.3492, + "eval_samples_per_second": 11.977, + "eval_wer": 0.37393422655298414, + "step": 56400 + }, + { + "epoch": 9.19, + "learning_rate": 6.345243922336433e-05, + "loss": NaN, + "step": 56500 + }, + { + "epoch": 9.19, + "eval_loss": 0.42654842138290405, + "eval_runtime": 6.8554, + "eval_samples_per_second": 14.587, + "eval_wer": 0.3812423873325213, + "step": 56500 + }, + { + "epoch": 9.2, + "learning_rate": 6.338717572197749e-05, + "loss": NaN, + "step": 56600 + }, + { + "epoch": 9.2, + "eval_loss": 0.4205102026462555, + "eval_runtime": 8.9024, + "eval_samples_per_second": 11.233, + "eval_wer": 0.37271619975639464, + "step": 56600 + }, + { + "epoch": 9.22, + "learning_rate": 6.332191222059064e-05, + "loss": NaN, + "step": 56700 + }, + { + "epoch": 9.22, + "eval_loss": 0.4153713583946228, + "eval_runtime": 8.0624, + "eval_samples_per_second": 12.403, + "eval_wer": 0.35931790499390986, + "step": 56700 + }, + { + "epoch": 9.24, + "learning_rate": 6.32566487192038e-05, + "loss": NaN, + "step": 56800 + }, + { + "epoch": 9.24, + "eval_loss": 0.4380166232585907, + "eval_runtime": 7.3533, + "eval_samples_per_second": 13.599, + "eval_wer": 0.3471376370280146, + "step": 56800 + }, + { + "epoch": 9.25, + "learning_rate": 6.319138521781694e-05, + "loss": NaN, + "step": 56900 + }, + { + "epoch": 9.25, + "eval_loss": 0.3888612389564514, + "eval_runtime": 9.3301, + "eval_samples_per_second": 10.718, + "eval_wer": 0.3520097442143727, + "step": 56900 + }, + { + "epoch": 9.27, + "learning_rate": 6.312612171643009e-05, + "loss": NaN, + "step": 57000 + }, + { + "epoch": 9.27, + "eval_loss": 0.39410483837127686, + "eval_runtime": 8.4812, + "eval_samples_per_second": 11.791, + "eval_wer": 0.3568818514007308, + "step": 57000 + }, + { + "epoch": 9.29, + "learning_rate": 6.306085821504324e-05, + "loss": NaN, + "step": 57100 + }, + { + "epoch": 9.29, + "eval_loss": 0.4467742443084717, + "eval_runtime": 7.367, + "eval_samples_per_second": 13.574, + "eval_wer": 0.35931790499390986, + "step": 57100 + }, + { + "epoch": 9.3, + "learning_rate": 6.299559471365639e-05, + "loss": NaN, + "step": 57200 + }, + { + "epoch": 9.3, + "eval_loss": 0.47365981340408325, + "eval_runtime": 8.3932, + "eval_samples_per_second": 11.914, + "eval_wer": 0.3800243605359318, + "step": 57200 + }, + { + "epoch": 9.32, + "learning_rate": 6.293033121226955e-05, + "loss": NaN, + "step": 57300 + }, + { + "epoch": 9.32, + "eval_loss": 0.4668024182319641, + "eval_runtime": 8.2093, + "eval_samples_per_second": 12.181, + "eval_wer": 0.3690621193666261, + "step": 57300 + }, + { + "epoch": 9.33, + "learning_rate": 6.286506771088269e-05, + "loss": NaN, + "step": 57400 + }, + { + "epoch": 9.33, + "eval_loss": 0.4525511860847473, + "eval_runtime": 9.8628, + "eval_samples_per_second": 10.139, + "eval_wer": 0.37758830694275275, + "step": 57400 + }, + { + "epoch": 9.35, + "learning_rate": 6.279980420949584e-05, + "loss": NaN, + "step": 57500 + }, + { + "epoch": 9.35, + "eval_loss": 0.45667481422424316, + "eval_runtime": 7.42, + "eval_samples_per_second": 13.477, + "eval_wer": 0.38367844092570036, + "step": 57500 + }, + { + "epoch": 9.37, + "learning_rate": 6.2734540708109e-05, + "loss": NaN, + "step": 57600 + }, + { + "epoch": 9.37, + "eval_loss": 0.4748976230621338, + "eval_runtime": 7.9502, + "eval_samples_per_second": 12.578, + "eval_wer": 0.37271619975639464, + "step": 57600 + }, + { + "epoch": 9.38, + "learning_rate": 6.266927720672214e-05, + "loss": NaN, + "step": 57700 + }, + { + "epoch": 9.38, + "eval_loss": 0.41124123334884644, + "eval_runtime": 8.7336, + "eval_samples_per_second": 11.45, + "eval_wer": 0.3629719853836784, + "step": 57700 + }, + { + "epoch": 9.4, + "learning_rate": 6.26040137053353e-05, + "loss": NaN, + "step": 57800 + }, + { + "epoch": 9.4, + "eval_loss": 0.4126984477043152, + "eval_runtime": 7.3596, + "eval_samples_per_second": 13.588, + "eval_wer": 0.3654080389768575, + "step": 57800 + }, + { + "epoch": 9.42, + "learning_rate": 6.253875020394845e-05, + "loss": NaN, + "step": 57900 + }, + { + "epoch": 9.42, + "eval_loss": 0.41617000102996826, + "eval_runtime": 8.9616, + "eval_samples_per_second": 11.159, + "eval_wer": 0.3617539585870889, + "step": 57900 + }, + { + "epoch": 9.43, + "learning_rate": 6.247348670256159e-05, + "loss": NaN, + "step": 58000 + }, + { + "epoch": 9.43, + "eval_loss": 0.4667081832885742, + "eval_runtime": 7.9176, + "eval_samples_per_second": 12.63, + "eval_wer": 0.37271619975639464, + "step": 58000 + }, + { + "epoch": 9.45, + "learning_rate": 6.240822320117474e-05, + "loss": NaN, + "step": 58100 + }, + { + "epoch": 9.45, + "eval_loss": 0.4130760431289673, + "eval_runtime": 9.1619, + "eval_samples_per_second": 10.915, + "eval_wer": 0.3520097442143727, + "step": 58100 + }, + { + "epoch": 9.46, + "learning_rate": 6.23429596997879e-05, + "loss": NaN, + "step": 58200 + }, + { + "epoch": 9.46, + "eval_loss": 0.3896174728870392, + "eval_runtime": 6.9796, + "eval_samples_per_second": 14.328, + "eval_wer": 0.3507917174177832, + "step": 58200 + }, + { + "epoch": 9.48, + "learning_rate": 6.227769619840105e-05, + "loss": NaN, + "step": 58300 + }, + { + "epoch": 9.48, + "eval_loss": 0.44675758481025696, + "eval_runtime": 10.5739, + "eval_samples_per_second": 9.457, + "eval_wer": 0.37393422655298414, + "step": 58300 + }, + { + "epoch": 9.5, + "learning_rate": 6.22124326970142e-05, + "loss": NaN, + "step": 58400 + }, + { + "epoch": 9.5, + "eval_loss": 0.45445725321769714, + "eval_runtime": 8.0137, + "eval_samples_per_second": 12.479, + "eval_wer": 0.35322777101096225, + "step": 58400 + }, + { + "epoch": 9.51, + "learning_rate": 6.214716919562736e-05, + "loss": NaN, + "step": 58500 + }, + { + "epoch": 9.51, + "eval_loss": 0.47014904022216797, + "eval_runtime": 8.4495, + "eval_samples_per_second": 11.835, + "eval_wer": 0.36784409257003653, + "step": 58500 + }, + { + "epoch": 9.53, + "learning_rate": 6.208190569424049e-05, + "loss": NaN, + "step": 58600 + }, + { + "epoch": 9.53, + "eval_loss": 0.42686671018600464, + "eval_runtime": 7.4504, + "eval_samples_per_second": 13.422, + "eval_wer": 0.36053593179049936, + "step": 58600 + }, + { + "epoch": 9.55, + "learning_rate": 6.201664219285365e-05, + "loss": NaN, + "step": 58700 + }, + { + "epoch": 9.55, + "eval_loss": 0.45379403233528137, + "eval_runtime": 7.6774, + "eval_samples_per_second": 13.025, + "eval_wer": 0.3617539585870889, + "step": 58700 + }, + { + "epoch": 9.56, + "learning_rate": 6.19513786914668e-05, + "loss": NaN, + "step": 58800 + }, + { + "epoch": 9.56, + "eval_loss": 0.4386783838272095, + "eval_runtime": 9.7583, + "eval_samples_per_second": 10.248, + "eval_wer": 0.36662606577344703, + "step": 58800 + }, + { + "epoch": 9.58, + "learning_rate": 6.188611519007995e-05, + "loss": NaN, + "step": 58900 + }, + { + "epoch": 9.58, + "eval_loss": 0.40192508697509766, + "eval_runtime": 9.6879, + "eval_samples_per_second": 10.322, + "eval_wer": 0.35809987819732036, + "step": 58900 + }, + { + "epoch": 9.6, + "learning_rate": 6.182085168869311e-05, + "loss": NaN, + "step": 59000 + }, + { + "epoch": 9.6, + "eval_loss": 0.46459415555000305, + "eval_runtime": 8.441, + "eval_samples_per_second": 11.847, + "eval_wer": 0.35931790499390986, + "step": 59000 + }, + { + "epoch": 9.61, + "learning_rate": 6.175558818730625e-05, + "loss": NaN, + "step": 59100 + }, + { + "epoch": 9.61, + "eval_loss": 0.41784995794296265, + "eval_runtime": 7.9289, + "eval_samples_per_second": 12.612, + "eval_wer": 0.38246041412911086, + "step": 59100 + }, + { + "epoch": 9.63, + "learning_rate": 6.16903246859194e-05, + "loss": NaN, + "step": 59200 + }, + { + "epoch": 9.63, + "eval_loss": 0.44782283902168274, + "eval_runtime": 6.8119, + "eval_samples_per_second": 14.68, + "eval_wer": 0.37758830694275275, + "step": 59200 + }, + { + "epoch": 9.64, + "learning_rate": 6.162506118453255e-05, + "loss": NaN, + "step": 59300 + }, + { + "epoch": 9.64, + "eval_loss": 0.40649664402008057, + "eval_runtime": 7.3722, + "eval_samples_per_second": 13.564, + "eval_wer": 0.3520097442143727, + "step": 59300 + }, + { + "epoch": 9.66, + "learning_rate": 6.15597976831457e-05, + "loss": NaN, + "step": 59400 + }, + { + "epoch": 9.66, + "eval_loss": 0.405249685049057, + "eval_runtime": 7.7895, + "eval_samples_per_second": 12.838, + "eval_wer": 0.3690621193666261, + "step": 59400 + }, + { + "epoch": 9.68, + "learning_rate": 6.149453418175886e-05, + "loss": NaN, + "step": 59500 + }, + { + "epoch": 9.68, + "eval_loss": 0.37005943059921265, + "eval_runtime": 7.3848, + "eval_samples_per_second": 13.541, + "eval_wer": 0.3507917174177832, + "step": 59500 + }, + { + "epoch": 9.69, + "learning_rate": 6.142927068037201e-05, + "loss": NaN, + "step": 59600 + }, + { + "epoch": 9.69, + "eval_loss": 0.4289380609989166, + "eval_runtime": 8.9702, + "eval_samples_per_second": 11.148, + "eval_wer": 0.36053593179049936, + "step": 59600 + }, + { + "epoch": 9.71, + "learning_rate": 6.136400717898515e-05, + "loss": NaN, + "step": 59700 + }, + { + "epoch": 9.71, + "eval_loss": 0.39418941736221313, + "eval_runtime": 7.4148, + "eval_samples_per_second": 13.487, + "eval_wer": 0.3654080389768575, + "step": 59700 + }, + { + "epoch": 9.73, + "learning_rate": 6.12987436775983e-05, + "loss": NaN, + "step": 59800 + }, + { + "epoch": 9.73, + "eval_loss": 0.44876864552497864, + "eval_runtime": 10.228, + "eval_samples_per_second": 9.777, + "eval_wer": 0.3556638246041413, + "step": 59800 + }, + { + "epoch": 9.74, + "learning_rate": 6.123348017621146e-05, + "loss": NaN, + "step": 59900 + }, + { + "epoch": 9.74, + "eval_loss": 0.49101337790489197, + "eval_runtime": 8.1552, + "eval_samples_per_second": 12.262, + "eval_wer": 0.3848964677222899, + "step": 59900 + }, + { + "epoch": 9.76, + "learning_rate": 6.11682166748246e-05, + "loss": NaN, + "step": 60000 + }, + { + "epoch": 9.76, + "eval_loss": 0.3982294797897339, + "eval_runtime": 8.4742, + "eval_samples_per_second": 11.8, + "eval_wer": 0.37149817295980514, + "step": 60000 + }, + { + "epoch": 9.77, + "learning_rate": 6.110295317343776e-05, + "loss": NaN, + "step": 60100 + }, + { + "epoch": 9.77, + "eval_loss": 0.3967539668083191, + "eval_runtime": 6.4434, + "eval_samples_per_second": 15.52, + "eval_wer": 0.36053593179049936, + "step": 60100 + }, + { + "epoch": 9.79, + "learning_rate": 6.103768967205092e-05, + "loss": NaN, + "step": 60200 + }, + { + "epoch": 9.79, + "eval_loss": 0.4385029673576355, + "eval_runtime": 7.5002, + "eval_samples_per_second": 13.333, + "eval_wer": 0.3617539585870889, + "step": 60200 + }, + { + "epoch": 9.81, + "learning_rate": 6.097242617066406e-05, + "loss": NaN, + "step": 60300 + }, + { + "epoch": 9.81, + "eval_loss": 0.4130534827709198, + "eval_runtime": 8.817, + "eval_samples_per_second": 11.342, + "eval_wer": 0.3751522533495737, + "step": 60300 + }, + { + "epoch": 9.82, + "learning_rate": 6.090716266927721e-05, + "loss": NaN, + "step": 60400 + }, + { + "epoch": 9.82, + "eval_loss": 0.5191701650619507, + "eval_runtime": 7.9727, + "eval_samples_per_second": 12.543, + "eval_wer": 0.37880633373934225, + "step": 60400 + }, + { + "epoch": 9.84, + "learning_rate": 6.084189916789036e-05, + "loss": NaN, + "step": 60500 + }, + { + "epoch": 9.84, + "eval_loss": 0.4418995678424835, + "eval_runtime": 9.1327, + "eval_samples_per_second": 10.95, + "eval_wer": 0.37880633373934225, + "step": 60500 + }, + { + "epoch": 9.86, + "learning_rate": 6.077663566650351e-05, + "loss": NaN, + "step": 60600 + }, + { + "epoch": 9.86, + "eval_loss": 0.4954804480075836, + "eval_runtime": 9.5074, + "eval_samples_per_second": 10.518, + "eval_wer": 0.4056029232643118, + "step": 60600 + }, + { + "epoch": 9.87, + "learning_rate": 6.0711372165116665e-05, + "loss": NaN, + "step": 60700 + }, + { + "epoch": 9.87, + "eval_loss": 0.45987486839294434, + "eval_runtime": 6.9269, + "eval_samples_per_second": 14.436, + "eval_wer": 0.36784409257003653, + "step": 60700 + }, + { + "epoch": 9.89, + "learning_rate": 6.064610866372982e-05, + "loss": NaN, + "step": 60800 + }, + { + "epoch": 9.89, + "eval_loss": 0.38861092925071716, + "eval_runtime": 9.0697, + "eval_samples_per_second": 11.026, + "eval_wer": 0.35931790499390986, + "step": 60800 + }, + { + "epoch": 9.9, + "learning_rate": 6.0580845162342956e-05, + "loss": NaN, + "step": 60900 + }, + { + "epoch": 9.9, + "eval_loss": 0.38833603262901306, + "eval_runtime": 7.6516, + "eval_samples_per_second": 13.069, + "eval_wer": 0.364190012180268, + "step": 60900 + }, + { + "epoch": 9.92, + "learning_rate": 6.0515581660956114e-05, + "loss": NaN, + "step": 61000 + }, + { + "epoch": 9.92, + "eval_loss": 0.44373536109924316, + "eval_runtime": 7.1189, + "eval_samples_per_second": 14.047, + "eval_wer": 0.35809987819732036, + "step": 61000 + }, + { + "epoch": 9.94, + "learning_rate": 6.0450318159569266e-05, + "loss": NaN, + "step": 61100 + }, + { + "epoch": 9.94, + "eval_loss": 0.41120070219039917, + "eval_runtime": 7.3599, + "eval_samples_per_second": 13.587, + "eval_wer": 0.3459196102314251, + "step": 61100 + }, + { + "epoch": 9.95, + "learning_rate": 6.038505465818242e-05, + "loss": NaN, + "step": 61200 + }, + { + "epoch": 9.95, + "eval_loss": 0.41615867614746094, + "eval_runtime": 9.4386, + "eval_samples_per_second": 10.595, + "eval_wer": 0.3568818514007308, + "step": 61200 + }, + { + "epoch": 9.97, + "learning_rate": 6.031979115679557e-05, + "loss": NaN, + "step": 61300 + }, + { + "epoch": 9.97, + "eval_loss": 0.4973611831665039, + "eval_runtime": 8.1559, + "eval_samples_per_second": 12.261, + "eval_wer": 0.3702801461632156, + "step": 61300 + }, + { + "epoch": 9.99, + "learning_rate": 6.025452765540871e-05, + "loss": NaN, + "step": 61400 + }, + { + "epoch": 9.99, + "eval_loss": 0.448307603597641, + "eval_runtime": 9.8048, + "eval_samples_per_second": 10.199, + "eval_wer": 0.3690621193666261, + "step": 61400 + }, + { + "epoch": 10.0, + "learning_rate": 6.018926415402186e-05, + "loss": NaN, + "step": 61500 + }, + { + "epoch": 10.0, + "eval_loss": 0.5115909576416016, + "eval_runtime": 7.945, + "eval_samples_per_second": 12.587, + "eval_wer": 0.37393422655298414, + "step": 61500 + }, + { + "epoch": 10.02, + "learning_rate": 6.012400065263501e-05, + "loss": NaN, + "step": 61600 + }, + { + "epoch": 10.02, + "eval_loss": 0.42501771450042725, + "eval_runtime": 9.4293, + "eval_samples_per_second": 10.605, + "eval_wer": 0.3690621193666261, + "step": 61600 + }, + { + "epoch": 10.03, + "learning_rate": 6.005873715124817e-05, + "loss": NaN, + "step": 61700 + }, + { + "epoch": 10.03, + "eval_loss": 0.4137573540210724, + "eval_runtime": 8.5889, + "eval_samples_per_second": 11.643, + "eval_wer": 0.3520097442143727, + "step": 61700 + }, + { + "epoch": 10.05, + "learning_rate": 5.999347364986132e-05, + "loss": NaN, + "step": 61800 + }, + { + "epoch": 10.05, + "eval_loss": 0.4213745892047882, + "eval_runtime": 9.5865, + "eval_samples_per_second": 10.431, + "eval_wer": 0.36053593179049936, + "step": 61800 + }, + { + "epoch": 10.07, + "learning_rate": 5.9928210148474475e-05, + "loss": NaN, + "step": 61900 + }, + { + "epoch": 10.07, + "eval_loss": 0.4096081256866455, + "eval_runtime": 8.1463, + "eval_samples_per_second": 12.275, + "eval_wer": 0.3690621193666261, + "step": 61900 + }, + { + "epoch": 10.08, + "learning_rate": 5.986294664708761e-05, + "loss": NaN, + "step": 62000 + }, + { + "epoch": 10.08, + "eval_loss": 0.41392749547958374, + "eval_runtime": 9.4692, + "eval_samples_per_second": 10.561, + "eval_wer": 0.36662606577344703, + "step": 62000 + }, + { + "epoch": 10.1, + "learning_rate": 5.9797683145700765e-05, + "loss": NaN, + "step": 62100 + }, + { + "epoch": 10.1, + "eval_loss": 0.41633856296539307, + "eval_runtime": 8.7223, + "eval_samples_per_second": 11.465, + "eval_wer": 0.3690621193666261, + "step": 62100 + }, + { + "epoch": 10.12, + "learning_rate": 5.973241964431392e-05, + "loss": NaN, + "step": 62200 + }, + { + "epoch": 10.12, + "eval_loss": 0.45410457253456116, + "eval_runtime": 7.382, + "eval_samples_per_second": 13.546, + "eval_wer": 0.3520097442143727, + "step": 62200 + }, + { + "epoch": 10.13, + "learning_rate": 5.966715614292707e-05, + "loss": NaN, + "step": 62300 + }, + { + "epoch": 10.13, + "eval_loss": 0.456071674823761, + "eval_runtime": 8.3133, + "eval_samples_per_second": 12.029, + "eval_wer": 0.3690621193666261, + "step": 62300 + }, + { + "epoch": 10.15, + "learning_rate": 5.960189264154023e-05, + "loss": NaN, + "step": 62400 + }, + { + "epoch": 10.15, + "eval_loss": 0.4523433446884155, + "eval_runtime": 7.4187, + "eval_samples_per_second": 13.479, + "eval_wer": 0.35931790499390986, + "step": 62400 + }, + { + "epoch": 10.16, + "learning_rate": 5.953662914015338e-05, + "loss": NaN, + "step": 62500 + }, + { + "epoch": 10.16, + "eval_loss": 0.4045345187187195, + "eval_runtime": 9.0224, + "eval_samples_per_second": 11.083, + "eval_wer": 0.3556638246041413, + "step": 62500 + }, + { + "epoch": 10.18, + "learning_rate": 5.947136563876652e-05, + "loss": NaN, + "step": 62600 + }, + { + "epoch": 10.18, + "eval_loss": 0.4067111313343048, + "eval_runtime": 7.5666, + "eval_samples_per_second": 13.216, + "eval_wer": 0.364190012180268, + "step": 62600 + }, + { + "epoch": 10.2, + "learning_rate": 5.940610213737967e-05, + "loss": NaN, + "step": 62700 + }, + { + "epoch": 10.2, + "eval_loss": 0.4985601007938385, + "eval_runtime": 10.6701, + "eval_samples_per_second": 9.372, + "eval_wer": 0.3812423873325213, + "step": 62700 + }, + { + "epoch": 10.21, + "learning_rate": 5.934083863599282e-05, + "loss": NaN, + "step": 62800 + }, + { + "epoch": 10.21, + "eval_loss": 0.4605454206466675, + "eval_runtime": 7.6443, + "eval_samples_per_second": 13.082, + "eval_wer": 0.37880633373934225, + "step": 62800 + }, + { + "epoch": 10.23, + "learning_rate": 5.9275575134605974e-05, + "loss": NaN, + "step": 62900 + }, + { + "epoch": 10.23, + "eval_loss": 0.41414105892181396, + "eval_runtime": 9.3718, + "eval_samples_per_second": 10.67, + "eval_wer": 0.35931790499390986, + "step": 62900 + }, + { + "epoch": 10.25, + "learning_rate": 5.9210311633219125e-05, + "loss": NaN, + "step": 63000 + }, + { + "epoch": 10.25, + "eval_loss": 0.4487077593803406, + "eval_runtime": 8.7749, + "eval_samples_per_second": 11.396, + "eval_wer": 0.37149817295980514, + "step": 63000 + }, + { + "epoch": 10.26, + "learning_rate": 5.914504813183227e-05, + "loss": NaN, + "step": 63100 + }, + { + "epoch": 10.26, + "eval_loss": 0.5015296339988708, + "eval_runtime": 7.5183, + "eval_samples_per_second": 13.301, + "eval_wer": 0.3861144945188794, + "step": 63100 + }, + { + "epoch": 10.28, + "learning_rate": 5.907978463044542e-05, + "loss": NaN, + "step": 63200 + }, + { + "epoch": 10.28, + "eval_loss": 0.49254322052001953, + "eval_runtime": 7.8387, + "eval_samples_per_second": 12.757, + "eval_wer": 0.3617539585870889, + "step": 63200 + }, + { + "epoch": 10.29, + "learning_rate": 5.9014521129058574e-05, + "loss": NaN, + "step": 63300 + }, + { + "epoch": 10.29, + "eval_loss": 0.4693681001663208, + "eval_runtime": 9.0717, + "eval_samples_per_second": 11.023, + "eval_wer": 0.3812423873325213, + "step": 63300 + }, + { + "epoch": 10.31, + "learning_rate": 5.8949257627671726e-05, + "loss": NaN, + "step": 63400 + }, + { + "epoch": 10.31, + "eval_loss": 0.4481465220451355, + "eval_runtime": 8.4682, + "eval_samples_per_second": 11.809, + "eval_wer": 0.3702801461632156, + "step": 63400 + }, + { + "epoch": 10.33, + "learning_rate": 5.888399412628488e-05, + "loss": NaN, + "step": 63500 + }, + { + "epoch": 10.33, + "eval_loss": 0.43376660346984863, + "eval_runtime": 9.0956, + "eval_samples_per_second": 10.994, + "eval_wer": 0.3702801461632156, + "step": 63500 + }, + { + "epoch": 10.34, + "learning_rate": 5.881873062489803e-05, + "loss": NaN, + "step": 63600 + }, + { + "epoch": 10.34, + "eval_loss": 0.4156494438648224, + "eval_runtime": 8.4037, + "eval_samples_per_second": 11.899, + "eval_wer": 0.35444579780755175, + "step": 63600 + }, + { + "epoch": 10.36, + "learning_rate": 5.8753467123511175e-05, + "loss": NaN, + "step": 63700 + }, + { + "epoch": 10.36, + "eval_loss": 0.4458553194999695, + "eval_runtime": 8.7165, + "eval_samples_per_second": 11.473, + "eval_wer": 0.37393422655298414, + "step": 63700 + }, + { + "epoch": 10.38, + "learning_rate": 5.868820362212433e-05, + "loss": NaN, + "step": 63800 + }, + { + "epoch": 10.38, + "eval_loss": 0.466782808303833, + "eval_runtime": 8.1618, + "eval_samples_per_second": 12.252, + "eval_wer": 0.3873325213154689, + "step": 63800 + }, + { + "epoch": 10.39, + "learning_rate": 5.862294012073748e-05, + "loss": NaN, + "step": 63900 + }, + { + "epoch": 10.39, + "eval_loss": 0.5930745601654053, + "eval_runtime": 9.5258, + "eval_samples_per_second": 10.498, + "eval_wer": 0.39707673568818513, + "step": 63900 + }, + { + "epoch": 10.41, + "learning_rate": 5.855767661935063e-05, + "loss": NaN, + "step": 64000 + }, + { + "epoch": 10.41, + "eval_loss": 0.45110756158828735, + "eval_runtime": 9.4694, + "eval_samples_per_second": 10.56, + "eval_wer": 0.3617539585870889, + "step": 64000 + }, + { + "epoch": 10.42, + "learning_rate": 5.849241311796378e-05, + "loss": NaN, + "step": 64100 + }, + { + "epoch": 10.42, + "eval_loss": 0.4962243139743805, + "eval_runtime": 8.4633, + "eval_samples_per_second": 11.816, + "eval_wer": 0.3800243605359318, + "step": 64100 + }, + { + "epoch": 10.44, + "learning_rate": 5.8427149616576935e-05, + "loss": NaN, + "step": 64200 + }, + { + "epoch": 10.44, + "eval_loss": 0.43860459327697754, + "eval_runtime": 8.0809, + "eval_samples_per_second": 12.375, + "eval_wer": 0.3568818514007308, + "step": 64200 + }, + { + "epoch": 10.46, + "learning_rate": 5.836188611519008e-05, + "loss": NaN, + "step": 64300 + }, + { + "epoch": 10.46, + "eval_loss": 0.4754991829395294, + "eval_runtime": 10.3787, + "eval_samples_per_second": 9.635, + "eval_wer": 0.3800243605359318, + "step": 64300 + }, + { + "epoch": 10.47, + "learning_rate": 5.829662261380323e-05, + "loss": NaN, + "step": 64400 + }, + { + "epoch": 10.47, + "eval_loss": 0.44975659251213074, + "eval_runtime": 8.6363, + "eval_samples_per_second": 11.579, + "eval_wer": 0.37393422655298414, + "step": 64400 + }, + { + "epoch": 10.49, + "learning_rate": 5.8231359112416384e-05, + "loss": NaN, + "step": 64500 + }, + { + "epoch": 10.49, + "eval_loss": 0.4558520019054413, + "eval_runtime": 8.2941, + "eval_samples_per_second": 12.057, + "eval_wer": 0.37271619975639464, + "step": 64500 + }, + { + "epoch": 10.51, + "learning_rate": 5.8166095611029536e-05, + "loss": NaN, + "step": 64600 + }, + { + "epoch": 10.51, + "eval_loss": 0.40871143341064453, + "eval_runtime": 8.7219, + "eval_samples_per_second": 11.465, + "eval_wer": 0.3556638246041413, + "step": 64600 + }, + { + "epoch": 10.52, + "learning_rate": 5.810083210964269e-05, + "loss": NaN, + "step": 64700 + }, + { + "epoch": 10.52, + "eval_loss": 0.5071795582771301, + "eval_runtime": 8.6766, + "eval_samples_per_second": 11.525, + "eval_wer": 0.3848964677222899, + "step": 64700 + }, + { + "epoch": 10.54, + "learning_rate": 5.803556860825583e-05, + "loss": NaN, + "step": 64800 + }, + { + "epoch": 10.54, + "eval_loss": 0.4381246864795685, + "eval_runtime": 7.7019, + "eval_samples_per_second": 12.984, + "eval_wer": 0.364190012180268, + "step": 64800 + }, + { + "epoch": 10.55, + "learning_rate": 5.7970305106868985e-05, + "loss": NaN, + "step": 64900 + }, + { + "epoch": 10.55, + "eval_loss": 0.42241111397743225, + "eval_runtime": 6.9119, + "eval_samples_per_second": 14.468, + "eval_wer": 0.3520097442143727, + "step": 64900 + }, + { + "epoch": 10.57, + "learning_rate": 5.790504160548214e-05, + "loss": NaN, + "step": 65000 + }, + { + "epoch": 10.57, + "eval_loss": 0.44368839263916016, + "eval_runtime": 8.9514, + "eval_samples_per_second": 11.171, + "eval_wer": 0.37393422655298414, + "step": 65000 + }, + { + "epoch": 10.59, + "learning_rate": 5.783977810409529e-05, + "loss": NaN, + "step": 65100 + }, + { + "epoch": 10.59, + "eval_loss": 0.4559805393218994, + "eval_runtime": 9.0066, + "eval_samples_per_second": 11.103, + "eval_wer": 0.3702801461632156, + "step": 65100 + }, + { + "epoch": 10.6, + "learning_rate": 5.777451460270844e-05, + "loss": NaN, + "step": 65200 + }, + { + "epoch": 10.6, + "eval_loss": 0.3956843912601471, + "eval_runtime": 8.016, + "eval_samples_per_second": 12.475, + "eval_wer": 0.3800243605359318, + "step": 65200 + }, + { + "epoch": 10.62, + "learning_rate": 5.770925110132159e-05, + "loss": NaN, + "step": 65300 + }, + { + "epoch": 10.62, + "eval_loss": 0.3806883990764618, + "eval_runtime": 6.54, + "eval_samples_per_second": 15.291, + "eval_wer": 0.3568818514007308, + "step": 65300 + }, + { + "epoch": 10.64, + "learning_rate": 5.764398759993474e-05, + "loss": NaN, + "step": 65400 + }, + { + "epoch": 10.64, + "eval_loss": 0.4815838634967804, + "eval_runtime": 8.6182, + "eval_samples_per_second": 11.603, + "eval_wer": 0.37758830694275275, + "step": 65400 + }, + { + "epoch": 10.65, + "learning_rate": 5.757872409854789e-05, + "loss": NaN, + "step": 65500 + }, + { + "epoch": 10.65, + "eval_loss": 0.42734912037849426, + "eval_runtime": 7.7612, + "eval_samples_per_second": 12.885, + "eval_wer": 0.3702801461632156, + "step": 65500 + }, + { + "epoch": 10.67, + "learning_rate": 5.751346059716104e-05, + "loss": NaN, + "step": 65600 + }, + { + "epoch": 10.67, + "eval_loss": 0.453123539686203, + "eval_runtime": 8.3059, + "eval_samples_per_second": 12.04, + "eval_wer": 0.3861144945188794, + "step": 65600 + }, + { + "epoch": 10.68, + "learning_rate": 5.744819709577419e-05, + "loss": NaN, + "step": 65700 + }, + { + "epoch": 10.68, + "eval_loss": 0.47986891865730286, + "eval_runtime": 8.2303, + "eval_samples_per_second": 12.15, + "eval_wer": 0.3763702801461632, + "step": 65700 + }, + { + "epoch": 10.7, + "learning_rate": 5.7382933594387345e-05, + "loss": NaN, + "step": 65800 + }, + { + "epoch": 10.7, + "eval_loss": 0.3873835802078247, + "eval_runtime": 7.8326, + "eval_samples_per_second": 12.767, + "eval_wer": 0.3568818514007308, + "step": 65800 + }, + { + "epoch": 10.72, + "learning_rate": 5.73176700930005e-05, + "loss": NaN, + "step": 65900 + }, + { + "epoch": 10.72, + "eval_loss": 0.41347476840019226, + "eval_runtime": 10.5146, + "eval_samples_per_second": 9.511, + "eval_wer": 0.36784409257003653, + "step": 65900 + }, + { + "epoch": 10.73, + "learning_rate": 5.7252406591613636e-05, + "loss": NaN, + "step": 66000 + }, + { + "epoch": 10.73, + "eval_loss": 0.39395031332969666, + "eval_runtime": 7.1626, + "eval_samples_per_second": 13.961, + "eval_wer": 0.3629719853836784, + "step": 66000 + }, + { + "epoch": 10.75, + "learning_rate": 5.7187143090226794e-05, + "loss": NaN, + "step": 66100 + }, + { + "epoch": 10.75, + "eval_loss": 0.38374292850494385, + "eval_runtime": 8.2206, + "eval_samples_per_second": 12.165, + "eval_wer": 0.3690621193666261, + "step": 66100 + }, + { + "epoch": 10.77, + "learning_rate": 5.7121879588839946e-05, + "loss": NaN, + "step": 66200 + }, + { + "epoch": 10.77, + "eval_loss": 0.4342725872993469, + "eval_runtime": 8.1692, + "eval_samples_per_second": 12.241, + "eval_wer": 0.37149817295980514, + "step": 66200 + }, + { + "epoch": 10.78, + "learning_rate": 5.70566160874531e-05, + "loss": NaN, + "step": 66300 + }, + { + "epoch": 10.78, + "eval_loss": 0.4155646562576294, + "eval_runtime": 6.8934, + "eval_samples_per_second": 14.507, + "eval_wer": 0.3690621193666261, + "step": 66300 + }, + { + "epoch": 10.8, + "learning_rate": 5.699135258606625e-05, + "loss": NaN, + "step": 66400 + }, + { + "epoch": 10.8, + "eval_loss": 0.410979300737381, + "eval_runtime": 7.7957, + "eval_samples_per_second": 12.828, + "eval_wer": 0.37149817295980514, + "step": 66400 + }, + { + "epoch": 10.81, + "learning_rate": 5.692608908467939e-05, + "loss": NaN, + "step": 66500 + }, + { + "epoch": 10.81, + "eval_loss": 0.3924597203731537, + "eval_runtime": 7.8199, + "eval_samples_per_second": 12.788, + "eval_wer": 0.3702801461632156, + "step": 66500 + }, + { + "epoch": 10.83, + "learning_rate": 5.686082558329254e-05, + "loss": NaN, + "step": 66600 + }, + { + "epoch": 10.83, + "eval_loss": 0.4514992833137512, + "eval_runtime": 8.7106, + "eval_samples_per_second": 11.48, + "eval_wer": 0.3763702801461632, + "step": 66600 + }, + { + "epoch": 10.85, + "learning_rate": 5.679556208190569e-05, + "loss": NaN, + "step": 66700 + }, + { + "epoch": 10.85, + "eval_loss": 0.46335548162460327, + "eval_runtime": 6.7296, + "eval_samples_per_second": 14.86, + "eval_wer": 0.37271619975639464, + "step": 66700 + }, + { + "epoch": 10.86, + "learning_rate": 5.673029858051885e-05, + "loss": NaN, + "step": 66800 + }, + { + "epoch": 10.86, + "eval_loss": 0.48735862970352173, + "eval_runtime": 8.529, + "eval_samples_per_second": 11.725, + "eval_wer": 0.38367844092570036, + "step": 66800 + }, + { + "epoch": 10.88, + "learning_rate": 5.6665035079132e-05, + "loss": NaN, + "step": 66900 + }, + { + "epoch": 10.88, + "eval_loss": 0.4417389929294586, + "eval_runtime": 8.1548, + "eval_samples_per_second": 12.263, + "eval_wer": 0.364190012180268, + "step": 66900 + }, + { + "epoch": 10.9, + "learning_rate": 5.6599771577745155e-05, + "loss": NaN, + "step": 67000 + }, + { + "epoch": 10.9, + "eval_loss": 0.38281384110450745, + "eval_runtime": 7.2321, + "eval_samples_per_second": 13.827, + "eval_wer": 0.34957369062119364, + "step": 67000 + }, + { + "epoch": 10.91, + "learning_rate": 5.653450807635829e-05, + "loss": NaN, + "step": 67100 + }, + { + "epoch": 10.91, + "eval_loss": 0.4453480839729309, + "eval_runtime": 9.8708, + "eval_samples_per_second": 10.131, + "eval_wer": 0.37149817295980514, + "step": 67100 + }, + { + "epoch": 10.93, + "learning_rate": 5.6469244574971445e-05, + "loss": NaN, + "step": 67200 + }, + { + "epoch": 10.93, + "eval_loss": 0.43574249744415283, + "eval_runtime": 6.7934, + "eval_samples_per_second": 14.72, + "eval_wer": 0.3654080389768575, + "step": 67200 + }, + { + "epoch": 10.94, + "learning_rate": 5.64039810735846e-05, + "loss": NaN, + "step": 67300 + }, + { + "epoch": 10.94, + "eval_loss": 0.4049379825592041, + "eval_runtime": 7.1607, + "eval_samples_per_second": 13.965, + "eval_wer": 0.35931790499390986, + "step": 67300 + }, + { + "epoch": 10.96, + "learning_rate": 5.6338717572197756e-05, + "loss": NaN, + "step": 67400 + }, + { + "epoch": 10.96, + "eval_loss": 0.40730178356170654, + "eval_runtime": 8.7286, + "eval_samples_per_second": 11.457, + "eval_wer": 0.36784409257003653, + "step": 67400 + }, + { + "epoch": 10.98, + "learning_rate": 5.627345407081091e-05, + "loss": NaN, + "step": 67500 + }, + { + "epoch": 10.98, + "eval_loss": 0.42172133922576904, + "eval_runtime": 7.5099, + "eval_samples_per_second": 13.316, + "eval_wer": 0.3617539585870889, + "step": 67500 + }, + { + "epoch": 10.99, + "learning_rate": 5.620819056942406e-05, + "loss": NaN, + "step": 67600 + }, + { + "epoch": 10.99, + "eval_loss": 0.4412268400192261, + "eval_runtime": 8.7502, + "eval_samples_per_second": 11.428, + "eval_wer": 0.3848964677222899, + "step": 67600 + }, + { + "epoch": 11.01, + "learning_rate": 5.61429270680372e-05, + "loss": NaN, + "step": 67700 + }, + { + "epoch": 11.01, + "eval_loss": 0.4201769232749939, + "eval_runtime": 9.2704, + "eval_samples_per_second": 10.787, + "eval_wer": 0.364190012180268, + "step": 67700 + }, + { + "epoch": 11.03, + "learning_rate": 5.607766356665035e-05, + "loss": NaN, + "step": 67800 + }, + { + "epoch": 11.03, + "eval_loss": 0.418453574180603, + "eval_runtime": 8.1205, + "eval_samples_per_second": 12.315, + "eval_wer": 0.37758830694275275, + "step": 67800 + }, + { + "epoch": 11.04, + "learning_rate": 5.60124000652635e-05, + "loss": NaN, + "step": 67900 + }, + { + "epoch": 11.04, + "eval_loss": 0.4066586196422577, + "eval_runtime": 7.045, + "eval_samples_per_second": 14.194, + "eval_wer": 0.36053593179049936, + "step": 67900 + }, + { + "epoch": 11.06, + "learning_rate": 5.5947136563876653e-05, + "loss": NaN, + "step": 68000 + }, + { + "epoch": 11.06, + "eval_loss": 0.43720924854278564, + "eval_runtime": 8.1791, + "eval_samples_per_second": 12.226, + "eval_wer": 0.3800243605359318, + "step": 68000 + }, + { + "epoch": 11.07, + "learning_rate": 5.588187306248981e-05, + "loss": NaN, + "step": 68100 + }, + { + "epoch": 11.07, + "eval_loss": 0.40105950832366943, + "eval_runtime": 9.6195, + "eval_samples_per_second": 10.396, + "eval_wer": 0.37880633373934225, + "step": 68100 + }, + { + "epoch": 11.09, + "learning_rate": 5.581660956110295e-05, + "loss": NaN, + "step": 68200 + }, + { + "epoch": 11.09, + "eval_loss": 0.365195631980896, + "eval_runtime": 6.5706, + "eval_samples_per_second": 15.219, + "eval_wer": 0.3447015834348356, + "step": 68200 + }, + { + "epoch": 11.11, + "learning_rate": 5.57513460597161e-05, + "loss": NaN, + "step": 68300 + }, + { + "epoch": 11.11, + "eval_loss": 0.4046388268470764, + "eval_runtime": 8.1044, + "eval_samples_per_second": 12.339, + "eval_wer": 0.3617539585870889, + "step": 68300 + }, + { + "epoch": 11.12, + "learning_rate": 5.5686082558329254e-05, + "loss": NaN, + "step": 68400 + }, + { + "epoch": 11.12, + "eval_loss": 0.39111071825027466, + "eval_runtime": 6.8887, + "eval_samples_per_second": 14.517, + "eval_wer": 0.35931790499390986, + "step": 68400 + }, + { + "epoch": 11.14, + "learning_rate": 5.5620819056942406e-05, + "loss": NaN, + "step": 68500 + }, + { + "epoch": 11.14, + "eval_loss": 0.4509762227535248, + "eval_runtime": 8.7816, + "eval_samples_per_second": 11.388, + "eval_wer": 0.36053593179049936, + "step": 68500 + }, + { + "epoch": 11.16, + "learning_rate": 5.555555555555556e-05, + "loss": NaN, + "step": 68600 + }, + { + "epoch": 11.16, + "eval_loss": 0.43242478370666504, + "eval_runtime": 9.8683, + "eval_samples_per_second": 10.133, + "eval_wer": 0.37271619975639464, + "step": 68600 + }, + { + "epoch": 11.17, + "learning_rate": 5.549029205416871e-05, + "loss": NaN, + "step": 68700 + }, + { + "epoch": 11.17, + "eval_loss": 0.3907853364944458, + "eval_runtime": 7.1506, + "eval_samples_per_second": 13.985, + "eval_wer": 0.3617539585870889, + "step": 68700 + }, + { + "epoch": 11.19, + "learning_rate": 5.5425028552781855e-05, + "loss": NaN, + "step": 68800 + }, + { + "epoch": 11.19, + "eval_loss": 0.5252321362495422, + "eval_runtime": 7.5772, + "eval_samples_per_second": 13.197, + "eval_wer": 0.3751522533495737, + "step": 68800 + }, + { + "epoch": 11.21, + "learning_rate": 5.535976505139501e-05, + "loss": NaN, + "step": 68900 + }, + { + "epoch": 11.21, + "eval_loss": 0.3976127505302429, + "eval_runtime": 8.8027, + "eval_samples_per_second": 11.36, + "eval_wer": 0.35809987819732036, + "step": 68900 + }, + { + "epoch": 11.22, + "learning_rate": 5.529450155000816e-05, + "loss": NaN, + "step": 69000 + }, + { + "epoch": 11.22, + "eval_loss": 0.3901364505290985, + "eval_runtime": 7.7268, + "eval_samples_per_second": 12.942, + "eval_wer": 0.364190012180268, + "step": 69000 + }, + { + "epoch": 11.24, + "learning_rate": 5.522923804862131e-05, + "loss": NaN, + "step": 69100 + }, + { + "epoch": 11.24, + "eval_loss": 0.42438361048698425, + "eval_runtime": 7.1569, + "eval_samples_per_second": 13.972, + "eval_wer": 0.3751522533495737, + "step": 69100 + }, + { + "epoch": 11.25, + "learning_rate": 5.516397454723446e-05, + "loss": NaN, + "step": 69200 + }, + { + "epoch": 11.25, + "eval_loss": 0.4562244415283203, + "eval_runtime": 8.5825, + "eval_samples_per_second": 11.652, + "eval_wer": 0.3800243605359318, + "step": 69200 + }, + { + "epoch": 11.27, + "learning_rate": 5.5098711045847615e-05, + "loss": NaN, + "step": 69300 + }, + { + "epoch": 11.27, + "eval_loss": 0.4827350974082947, + "eval_runtime": 7.0866, + "eval_samples_per_second": 14.111, + "eval_wer": 0.37149817295980514, + "step": 69300 + }, + { + "epoch": 11.29, + "learning_rate": 5.503344754446076e-05, + "loss": NaN, + "step": 69400 + }, + { + "epoch": 11.29, + "eval_loss": 0.43090012669563293, + "eval_runtime": 7.5916, + "eval_samples_per_second": 13.172, + "eval_wer": 0.36784409257003653, + "step": 69400 + }, + { + "epoch": 11.3, + "learning_rate": 5.496818404307391e-05, + "loss": NaN, + "step": 69500 + }, + { + "epoch": 11.3, + "eval_loss": 0.42443668842315674, + "eval_runtime": 8.0483, + "eval_samples_per_second": 12.425, + "eval_wer": 0.36053593179049936, + "step": 69500 + }, + { + "epoch": 11.32, + "learning_rate": 5.4902920541687064e-05, + "loss": NaN, + "step": 69600 + }, + { + "epoch": 11.32, + "eval_loss": 0.42483994364738464, + "eval_runtime": 8.6393, + "eval_samples_per_second": 11.575, + "eval_wer": 0.37880633373934225, + "step": 69600 + }, + { + "epoch": 11.34, + "learning_rate": 5.4837657040300216e-05, + "loss": NaN, + "step": 69700 + }, + { + "epoch": 11.34, + "eval_loss": 0.47487872838974, + "eval_runtime": 7.5637, + "eval_samples_per_second": 13.221, + "eval_wer": 0.37149817295980514, + "step": 69700 + }, + { + "epoch": 11.35, + "learning_rate": 5.477239353891337e-05, + "loss": NaN, + "step": 69800 + }, + { + "epoch": 11.35, + "eval_loss": 0.4218917191028595, + "eval_runtime": 9.9857, + "eval_samples_per_second": 10.014, + "eval_wer": 0.3995127892813642, + "step": 69800 + }, + { + "epoch": 11.37, + "learning_rate": 5.470713003752651e-05, + "loss": NaN, + "step": 69900 + }, + { + "epoch": 11.37, + "eval_loss": 0.4314954876899719, + "eval_runtime": 9.0835, + "eval_samples_per_second": 11.009, + "eval_wer": 0.3751522533495737, + "step": 69900 + }, + { + "epoch": 11.38, + "learning_rate": 5.4641866536139665e-05, + "loss": NaN, + "step": 70000 + }, + { + "epoch": 11.38, + "eval_loss": 0.38264960050582886, + "eval_runtime": 8.6353, + "eval_samples_per_second": 11.58, + "eval_wer": 0.3751522533495737, + "step": 70000 + }, + { + "epoch": 11.4, + "learning_rate": 5.457660303475282e-05, + "loss": NaN, + "step": 70100 + }, + { + "epoch": 11.4, + "eval_loss": 0.3891557455062866, + "eval_runtime": 8.2828, + "eval_samples_per_second": 12.073, + "eval_wer": 0.3812423873325213, + "step": 70100 + }, + { + "epoch": 11.42, + "learning_rate": 5.451133953336597e-05, + "loss": NaN, + "step": 70200 + }, + { + "epoch": 11.42, + "eval_loss": 0.4659789800643921, + "eval_runtime": 7.9385, + "eval_samples_per_second": 12.597, + "eval_wer": 0.37758830694275275, + "step": 70200 + }, + { + "epoch": 11.43, + "learning_rate": 5.444607603197912e-05, + "loss": NaN, + "step": 70300 + }, + { + "epoch": 11.43, + "eval_loss": 0.4575248062610626, + "eval_runtime": 7.6225, + "eval_samples_per_second": 13.119, + "eval_wer": 0.3702801461632156, + "step": 70300 + }, + { + "epoch": 11.45, + "learning_rate": 5.438081253059227e-05, + "loss": NaN, + "step": 70400 + }, + { + "epoch": 11.45, + "eval_loss": 0.4173835515975952, + "eval_runtime": 7.9279, + "eval_samples_per_second": 12.614, + "eval_wer": 0.36053593179049936, + "step": 70400 + }, + { + "epoch": 11.47, + "learning_rate": 5.431554902920542e-05, + "loss": NaN, + "step": 70500 + }, + { + "epoch": 11.47, + "eval_loss": 0.4074372947216034, + "eval_runtime": 8.2601, + "eval_samples_per_second": 12.106, + "eval_wer": 0.3556638246041413, + "step": 70500 + }, + { + "epoch": 11.48, + "learning_rate": 5.425028552781857e-05, + "loss": NaN, + "step": 70600 + }, + { + "epoch": 11.48, + "eval_loss": 0.4152701497077942, + "eval_runtime": 8.2546, + "eval_samples_per_second": 12.114, + "eval_wer": 0.3617539585870889, + "step": 70600 + }, + { + "epoch": 11.5, + "learning_rate": 5.418502202643172e-05, + "loss": NaN, + "step": 70700 + }, + { + "epoch": 11.5, + "eval_loss": 0.422438383102417, + "eval_runtime": 7.4806, + "eval_samples_per_second": 13.368, + "eval_wer": 0.3556638246041413, + "step": 70700 + }, + { + "epoch": 11.51, + "learning_rate": 5.411975852504487e-05, + "loss": NaN, + "step": 70800 + }, + { + "epoch": 11.51, + "eval_loss": 0.40204957127571106, + "eval_runtime": 8.0201, + "eval_samples_per_second": 12.469, + "eval_wer": 0.3654080389768575, + "step": 70800 + }, + { + "epoch": 11.53, + "learning_rate": 5.4054495023658025e-05, + "loss": NaN, + "step": 70900 + }, + { + "epoch": 11.53, + "eval_loss": 0.4267421066761017, + "eval_runtime": 8.9633, + "eval_samples_per_second": 11.157, + "eval_wer": 0.36053593179049936, + "step": 70900 + }, + { + "epoch": 11.55, + "learning_rate": 5.398923152227118e-05, + "loss": NaN, + "step": 71000 + }, + { + "epoch": 11.55, + "eval_loss": 0.4199873208999634, + "eval_runtime": 8.0756, + "eval_samples_per_second": 12.383, + "eval_wer": 0.3617539585870889, + "step": 71000 + }, + { + "epoch": 11.56, + "learning_rate": 5.392396802088432e-05, + "loss": NaN, + "step": 71100 + }, + { + "epoch": 11.56, + "eval_loss": 0.4814501702785492, + "eval_runtime": 8.2177, + "eval_samples_per_second": 12.169, + "eval_wer": 0.3690621193666261, + "step": 71100 + }, + { + "epoch": 11.58, + "learning_rate": 5.3858704519497474e-05, + "loss": NaN, + "step": 71200 + }, + { + "epoch": 11.58, + "eval_loss": 0.4721526801586151, + "eval_runtime": 9.4886, + "eval_samples_per_second": 10.539, + "eval_wer": 0.36784409257003653, + "step": 71200 + }, + { + "epoch": 11.6, + "learning_rate": 5.3793441018110626e-05, + "loss": NaN, + "step": 71300 + }, + { + "epoch": 11.6, + "eval_loss": 0.41694724559783936, + "eval_runtime": 8.7547, + "eval_samples_per_second": 11.422, + "eval_wer": 0.3556638246041413, + "step": 71300 + }, + { + "epoch": 11.61, + "learning_rate": 5.372817751672378e-05, + "loss": NaN, + "step": 71400 + }, + { + "epoch": 11.61, + "eval_loss": 0.43105384707450867, + "eval_runtime": 8.4226, + "eval_samples_per_second": 11.873, + "eval_wer": 0.35322777101096225, + "step": 71400 + }, + { + "epoch": 11.63, + "learning_rate": 5.366291401533693e-05, + "loss": NaN, + "step": 71500 + }, + { + "epoch": 11.63, + "eval_loss": 0.43795666098594666, + "eval_runtime": 7.2699, + "eval_samples_per_second": 13.755, + "eval_wer": 0.36784409257003653, + "step": 71500 + }, + { + "epoch": 11.64, + "learning_rate": 5.359765051395007e-05, + "loss": NaN, + "step": 71600 + }, + { + "epoch": 11.64, + "eval_loss": 0.4947461783885956, + "eval_runtime": 7.7101, + "eval_samples_per_second": 12.97, + "eval_wer": 0.36662606577344703, + "step": 71600 + }, + { + "epoch": 11.66, + "learning_rate": 5.353238701256322e-05, + "loss": NaN, + "step": 71700 + }, + { + "epoch": 11.66, + "eval_loss": 0.4540432095527649, + "eval_runtime": 6.9467, + "eval_samples_per_second": 14.395, + "eval_wer": 0.35931790499390986, + "step": 71700 + }, + { + "epoch": 11.68, + "learning_rate": 5.346712351117638e-05, + "loss": NaN, + "step": 71800 + }, + { + "epoch": 11.68, + "eval_loss": 0.3812675476074219, + "eval_runtime": 7.2252, + "eval_samples_per_second": 13.84, + "eval_wer": 0.3471376370280146, + "step": 71800 + }, + { + "epoch": 11.69, + "learning_rate": 5.340186000978953e-05, + "loss": NaN, + "step": 71900 + }, + { + "epoch": 11.69, + "eval_loss": 0.3847273290157318, + "eval_runtime": 7.6934, + "eval_samples_per_second": 12.998, + "eval_wer": 0.3568818514007308, + "step": 71900 + }, + { + "epoch": 11.71, + "learning_rate": 5.333659650840268e-05, + "loss": NaN, + "step": 72000 + }, + { + "epoch": 11.71, + "eval_loss": 0.4283711910247803, + "eval_runtime": 8.7201, + "eval_samples_per_second": 11.468, + "eval_wer": 0.3751522533495737, + "step": 72000 + }, + { + "epoch": 11.73, + "learning_rate": 5.3271333007015835e-05, + "loss": NaN, + "step": 72100 + }, + { + "epoch": 11.73, + "eval_loss": 0.4184917211532593, + "eval_runtime": 7.324, + "eval_samples_per_second": 13.654, + "eval_wer": 0.3568818514007308, + "step": 72100 + }, + { + "epoch": 11.74, + "learning_rate": 5.320606950562897e-05, + "loss": NaN, + "step": 72200 + }, + { + "epoch": 11.74, + "eval_loss": 0.4423336088657379, + "eval_runtime": 8.1944, + "eval_samples_per_second": 12.203, + "eval_wer": 0.37758830694275275, + "step": 72200 + }, + { + "epoch": 11.76, + "learning_rate": 5.3140806004242125e-05, + "loss": NaN, + "step": 72300 + }, + { + "epoch": 11.76, + "eval_loss": 0.3999084234237671, + "eval_runtime": 8.6145, + "eval_samples_per_second": 11.608, + "eval_wer": 0.36053593179049936, + "step": 72300 + }, + { + "epoch": 11.77, + "learning_rate": 5.307554250285528e-05, + "loss": NaN, + "step": 72400 + }, + { + "epoch": 11.77, + "eval_loss": 0.428960919380188, + "eval_runtime": 7.2399, + "eval_samples_per_second": 13.812, + "eval_wer": 0.36053593179049936, + "step": 72400 + }, + { + "epoch": 11.79, + "learning_rate": 5.3010279001468436e-05, + "loss": NaN, + "step": 72500 + }, + { + "epoch": 11.79, + "eval_loss": 0.3995371162891388, + "eval_runtime": 7.3237, + "eval_samples_per_second": 13.654, + "eval_wer": 0.35322777101096225, + "step": 72500 + }, + { + "epoch": 11.81, + "learning_rate": 5.294501550008159e-05, + "loss": NaN, + "step": 72600 + }, + { + "epoch": 11.81, + "eval_loss": 0.3626769185066223, + "eval_runtime": 8.6716, + "eval_samples_per_second": 11.532, + "eval_wer": 0.35322777101096225, + "step": 72600 + }, + { + "epoch": 11.82, + "learning_rate": 5.287975199869474e-05, + "loss": NaN, + "step": 72700 + }, + { + "epoch": 11.82, + "eval_loss": 0.4408608675003052, + "eval_runtime": 8.3313, + "eval_samples_per_second": 12.003, + "eval_wer": 0.38976857490864797, + "step": 72700 + }, + { + "epoch": 11.84, + "learning_rate": 5.281448849730788e-05, + "loss": NaN, + "step": 72800 + }, + { + "epoch": 11.84, + "eval_loss": 0.39294520020484924, + "eval_runtime": 9.9612, + "eval_samples_per_second": 10.039, + "eval_wer": 0.36784409257003653, + "step": 72800 + }, + { + "epoch": 11.86, + "learning_rate": 5.274922499592103e-05, + "loss": NaN, + "step": 72900 + }, + { + "epoch": 11.86, + "eval_loss": 0.4848403036594391, + "eval_runtime": 7.6218, + "eval_samples_per_second": 13.12, + "eval_wer": 0.37149817295980514, + "step": 72900 + }, + { + "epoch": 11.87, + "learning_rate": 5.268396149453418e-05, + "loss": NaN, + "step": 73000 + }, + { + "epoch": 11.87, + "eval_loss": 0.41433045268058777, + "eval_runtime": 8.3506, + "eval_samples_per_second": 11.975, + "eval_wer": 0.37271619975639464, + "step": 73000 + }, + { + "epoch": 11.89, + "learning_rate": 5.2618697993147333e-05, + "loss": NaN, + "step": 73100 + }, + { + "epoch": 11.89, + "eval_loss": 0.38409173488616943, + "eval_runtime": 7.2038, + "eval_samples_per_second": 13.881, + "eval_wer": 0.35444579780755175, + "step": 73100 + }, + { + "epoch": 11.9, + "learning_rate": 5.255343449176049e-05, + "loss": NaN, + "step": 73200 + }, + { + "epoch": 11.9, + "eval_loss": 0.4268048405647278, + "eval_runtime": 9.8281, + "eval_samples_per_second": 10.175, + "eval_wer": 0.38976857490864797, + "step": 73200 + }, + { + "epoch": 11.92, + "learning_rate": 5.248817099037363e-05, + "loss": NaN, + "step": 73300 + }, + { + "epoch": 11.92, + "eval_loss": 0.43448886275291443, + "eval_runtime": 7.4958, + "eval_samples_per_second": 13.341, + "eval_wer": 0.3556638246041413, + "step": 73300 + }, + { + "epoch": 11.94, + "learning_rate": 5.242290748898678e-05, + "loss": NaN, + "step": 73400 + }, + { + "epoch": 11.94, + "eval_loss": 0.43121451139450073, + "eval_runtime": 9.4109, + "eval_samples_per_second": 10.626, + "eval_wer": 0.37880633373934225, + "step": 73400 + }, + { + "epoch": 11.95, + "learning_rate": 5.2357643987599934e-05, + "loss": NaN, + "step": 73500 + }, + { + "epoch": 11.95, + "eval_loss": 0.3783913552761078, + "eval_runtime": 7.8278, + "eval_samples_per_second": 12.775, + "eval_wer": 0.364190012180268, + "step": 73500 + }, + { + "epoch": 11.97, + "learning_rate": 5.2292380486213086e-05, + "loss": NaN, + "step": 73600 + }, + { + "epoch": 11.97, + "eval_loss": 0.4278438687324524, + "eval_runtime": 9.8748, + "eval_samples_per_second": 10.127, + "eval_wer": 0.32399512789281365, + "step": 73600 + }, + { + "epoch": 11.99, + "learning_rate": 5.222711698482624e-05, + "loss": NaN, + "step": 73700 + }, + { + "epoch": 11.99, + "eval_loss": 0.3995451331138611, + "eval_runtime": 6.8971, + "eval_samples_per_second": 14.499, + "eval_wer": 0.3629719853836784, + "step": 73700 + }, + { + "epoch": 12.0, + "learning_rate": 5.21618534834394e-05, + "loss": NaN, + "step": 73800 + }, + { + "epoch": 12.0, + "eval_loss": 0.43313267827033997, + "eval_runtime": 6.688, + "eval_samples_per_second": 14.952, + "eval_wer": 0.36053593179049936, + "step": 73800 + }, + { + "epoch": 12.02, + "learning_rate": 5.2096589982052535e-05, + "loss": NaN, + "step": 73900 + }, + { + "epoch": 12.02, + "eval_loss": 0.3490147292613983, + "eval_runtime": 8.1685, + "eval_samples_per_second": 12.242, + "eval_wer": 0.3447015834348356, + "step": 73900 + }, + { + "epoch": 12.03, + "learning_rate": 5.203132648066569e-05, + "loss": NaN, + "step": 74000 + }, + { + "epoch": 12.03, + "eval_loss": 0.37458693981170654, + "eval_runtime": 8.6478, + "eval_samples_per_second": 11.564, + "eval_wer": 0.3520097442143727, + "step": 74000 + }, + { + "epoch": 12.05, + "learning_rate": 5.196606297927884e-05, + "loss": NaN, + "step": 74100 + }, + { + "epoch": 12.05, + "eval_loss": 0.3997301459312439, + "eval_runtime": 9.0658, + "eval_samples_per_second": 11.031, + "eval_wer": 0.3617539585870889, + "step": 74100 + }, + { + "epoch": 12.07, + "learning_rate": 5.190079947789199e-05, + "loss": NaN, + "step": 74200 + }, + { + "epoch": 12.07, + "eval_loss": 0.4225488305091858, + "eval_runtime": 8.5571, + "eval_samples_per_second": 11.686, + "eval_wer": 0.364190012180268, + "step": 74200 + }, + { + "epoch": 12.08, + "learning_rate": 5.183553597650514e-05, + "loss": NaN, + "step": 74300 + }, + { + "epoch": 12.08, + "eval_loss": 0.4431927800178528, + "eval_runtime": 7.6665, + "eval_samples_per_second": 13.044, + "eval_wer": 0.34957369062119364, + "step": 74300 + }, + { + "epoch": 12.1, + "learning_rate": 5.1770272475118295e-05, + "loss": NaN, + "step": 74400 + }, + { + "epoch": 12.1, + "eval_loss": 0.39965328574180603, + "eval_runtime": 6.9617, + "eval_samples_per_second": 14.364, + "eval_wer": 0.36053593179049936, + "step": 74400 + }, + { + "epoch": 12.12, + "learning_rate": 5.170500897373144e-05, + "loss": NaN, + "step": 74500 + }, + { + "epoch": 12.12, + "eval_loss": 0.4133952260017395, + "eval_runtime": 10.1253, + "eval_samples_per_second": 9.876, + "eval_wer": 0.36053593179049936, + "step": 74500 + }, + { + "epoch": 12.13, + "learning_rate": 5.163974547234459e-05, + "loss": NaN, + "step": 74600 + }, + { + "epoch": 12.13, + "eval_loss": 0.436084508895874, + "eval_runtime": 9.4077, + "eval_samples_per_second": 10.63, + "eval_wer": 0.35931790499390986, + "step": 74600 + }, + { + "epoch": 12.15, + "learning_rate": 5.1574481970957744e-05, + "loss": NaN, + "step": 74700 + }, + { + "epoch": 12.15, + "eval_loss": 0.42807960510253906, + "eval_runtime": 8.0262, + "eval_samples_per_second": 12.459, + "eval_wer": 0.3520097442143727, + "step": 74700 + }, + { + "epoch": 12.16, + "learning_rate": 5.1509218469570896e-05, + "loss": NaN, + "step": 74800 + }, + { + "epoch": 12.16, + "eval_loss": 0.6679572463035583, + "eval_runtime": 9.4255, + "eval_samples_per_second": 10.61, + "eval_wer": 0.35444579780755175, + "step": 74800 + }, + { + "epoch": 12.18, + "learning_rate": 5.144395496818405e-05, + "loss": NaN, + "step": 74900 + }, + { + "epoch": 12.18, + "eval_loss": 0.3985752463340759, + "eval_runtime": 7.9339, + "eval_samples_per_second": 12.604, + "eval_wer": 0.35809987819732036, + "step": 74900 + }, + { + "epoch": 12.2, + "learning_rate": 5.137869146679719e-05, + "loss": NaN, + "step": 75000 + }, + { + "epoch": 12.2, + "eval_loss": 0.3972986936569214, + "eval_runtime": 9.0029, + "eval_samples_per_second": 11.108, + "eval_wer": 0.35322777101096225, + "step": 75000 + }, + { + "epoch": 12.21, + "learning_rate": 5.1313427965410345e-05, + "loss": NaN, + "step": 75100 + }, + { + "epoch": 12.21, + "eval_loss": 0.41875389218330383, + "eval_runtime": 7.5122, + "eval_samples_per_second": 13.312, + "eval_wer": 0.3556638246041413, + "step": 75100 + }, + { + "epoch": 12.23, + "learning_rate": 5.1248164464023497e-05, + "loss": NaN, + "step": 75200 + }, + { + "epoch": 12.23, + "eval_loss": 0.4159642159938812, + "eval_runtime": 7.1412, + "eval_samples_per_second": 14.003, + "eval_wer": 0.33617539585870887, + "step": 75200 + }, + { + "epoch": 12.25, + "learning_rate": 5.118290096263665e-05, + "loss": NaN, + "step": 75300 + }, + { + "epoch": 12.25, + "eval_loss": 0.44466495513916016, + "eval_runtime": 9.1084, + "eval_samples_per_second": 10.979, + "eval_wer": 0.34957369062119364, + "step": 75300 + }, + { + "epoch": 12.26, + "learning_rate": 5.11176374612498e-05, + "loss": NaN, + "step": 75400 + }, + { + "epoch": 12.26, + "eval_loss": 0.45291373133659363, + "eval_runtime": 10.9268, + "eval_samples_per_second": 9.152, + "eval_wer": 0.3654080389768575, + "step": 75400 + }, + { + "epoch": 12.28, + "learning_rate": 5.105237395986295e-05, + "loss": NaN, + "step": 75500 + }, + { + "epoch": 12.28, + "eval_loss": 0.429513156414032, + "eval_runtime": 7.1347, + "eval_samples_per_second": 14.016, + "eval_wer": 0.3507917174177832, + "step": 75500 + }, + { + "epoch": 12.29, + "learning_rate": 5.09871104584761e-05, + "loss": NaN, + "step": 75600 + }, + { + "epoch": 12.29, + "eval_loss": 0.43607571721076965, + "eval_runtime": 8.106, + "eval_samples_per_second": 12.336, + "eval_wer": 0.36662606577344703, + "step": 75600 + }, + { + "epoch": 12.31, + "learning_rate": 5.092184695708925e-05, + "loss": NaN, + "step": 75700 + }, + { + "epoch": 12.31, + "eval_loss": 0.505099356174469, + "eval_runtime": 9.0741, + "eval_samples_per_second": 11.02, + "eval_wer": 0.36053593179049936, + "step": 75700 + }, + { + "epoch": 12.33, + "learning_rate": 5.08565834557024e-05, + "loss": NaN, + "step": 75800 + }, + { + "epoch": 12.33, + "eval_loss": 0.5338563919067383, + "eval_runtime": 9.1994, + "eval_samples_per_second": 10.87, + "eval_wer": 0.35322777101096225, + "step": 75800 + }, + { + "epoch": 12.34, + "learning_rate": 5.079131995431555e-05, + "loss": NaN, + "step": 75900 + }, + { + "epoch": 12.34, + "eval_loss": 0.43960484862327576, + "eval_runtime": 10.6506, + "eval_samples_per_second": 9.389, + "eval_wer": 0.364190012180268, + "step": 75900 + }, + { + "epoch": 12.36, + "learning_rate": 5.0726056452928705e-05, + "loss": NaN, + "step": 76000 + }, + { + "epoch": 12.36, + "eval_loss": 0.4200018644332886, + "eval_runtime": 8.4143, + "eval_samples_per_second": 11.885, + "eval_wer": 0.3459196102314251, + "step": 76000 + }, + { + "epoch": 12.38, + "learning_rate": 5.066079295154186e-05, + "loss": NaN, + "step": 76100 + }, + { + "epoch": 12.38, + "eval_loss": 0.44790372252464294, + "eval_runtime": 9.2667, + "eval_samples_per_second": 10.791, + "eval_wer": 0.36053593179049936, + "step": 76100 + }, + { + "epoch": 12.39, + "learning_rate": 5.0595529450155e-05, + "loss": NaN, + "step": 76200 + }, + { + "epoch": 12.39, + "eval_loss": 0.4088483154773712, + "eval_runtime": 9.4703, + "eval_samples_per_second": 10.559, + "eval_wer": 0.35809987819732036, + "step": 76200 + }, + { + "epoch": 12.41, + "learning_rate": 5.0530265948768154e-05, + "loss": NaN, + "step": 76300 + }, + { + "epoch": 12.41, + "eval_loss": 0.4267137944698334, + "eval_runtime": 8.6069, + "eval_samples_per_second": 11.619, + "eval_wer": 0.3763702801461632, + "step": 76300 + }, + { + "epoch": 12.42, + "learning_rate": 5.0465002447381306e-05, + "loss": NaN, + "step": 76400 + }, + { + "epoch": 12.42, + "eval_loss": 0.4056796133518219, + "eval_runtime": 7.7712, + "eval_samples_per_second": 12.868, + "eval_wer": 0.37149817295980514, + "step": 76400 + }, + { + "epoch": 12.44, + "learning_rate": 5.039973894599446e-05, + "loss": NaN, + "step": 76500 + }, + { + "epoch": 12.44, + "eval_loss": 0.3797760307788849, + "eval_runtime": 8.4756, + "eval_samples_per_second": 11.799, + "eval_wer": 0.341047503045067, + "step": 76500 + }, + { + "epoch": 12.46, + "learning_rate": 5.033447544460761e-05, + "loss": NaN, + "step": 76600 + }, + { + "epoch": 12.46, + "eval_loss": 0.3841753900051117, + "eval_runtime": 9.2622, + "eval_samples_per_second": 10.797, + "eval_wer": 0.3617539585870889, + "step": 76600 + }, + { + "epoch": 12.47, + "learning_rate": 5.026921194322075e-05, + "loss": NaN, + "step": 76700 + }, + { + "epoch": 12.47, + "eval_loss": 0.3942345380783081, + "eval_runtime": 7.1118, + "eval_samples_per_second": 14.061, + "eval_wer": 0.3459196102314251, + "step": 76700 + }, + { + "epoch": 12.49, + "learning_rate": 5.02039484418339e-05, + "loss": NaN, + "step": 76800 + }, + { + "epoch": 12.49, + "eval_loss": 0.4500068426132202, + "eval_runtime": 8.7652, + "eval_samples_per_second": 11.409, + "eval_wer": 0.36784409257003653, + "step": 76800 + }, + { + "epoch": 12.51, + "learning_rate": 5.013868494044706e-05, + "loss": NaN, + "step": 76900 + }, + { + "epoch": 12.51, + "eval_loss": 0.36947688460350037, + "eval_runtime": 7.5264, + "eval_samples_per_second": 13.286, + "eval_wer": 0.35322777101096225, + "step": 76900 + }, + { + "epoch": 12.52, + "learning_rate": 5.007342143906021e-05, + "loss": NaN, + "step": 77000 + }, + { + "epoch": 12.52, + "eval_loss": 0.4559879004955292, + "eval_runtime": 7.0948, + "eval_samples_per_second": 14.095, + "eval_wer": 0.36053593179049936, + "step": 77000 + }, + { + "epoch": 12.54, + "learning_rate": 5.000815793767336e-05, + "loss": NaN, + "step": 77100 + }, + { + "epoch": 12.54, + "eval_loss": 0.4029790759086609, + "eval_runtime": 8.456, + "eval_samples_per_second": 11.826, + "eval_wer": 0.34835566382460414, + "step": 77100 + }, + { + "epoch": 12.55, + "learning_rate": 4.994289443628651e-05, + "loss": NaN, + "step": 77200 + }, + { + "epoch": 12.55, + "eval_loss": 0.3982163369655609, + "eval_runtime": 7.6143, + "eval_samples_per_second": 13.133, + "eval_wer": 0.34226552984165654, + "step": 77200 + }, + { + "epoch": 12.57, + "learning_rate": 4.987763093489966e-05, + "loss": NaN, + "step": 77300 + }, + { + "epoch": 12.57, + "eval_loss": 0.5170900225639343, + "eval_runtime": 8.925, + "eval_samples_per_second": 11.204, + "eval_wer": 0.36784409257003653, + "step": 77300 + }, + { + "epoch": 12.59, + "learning_rate": 4.9812367433512805e-05, + "loss": NaN, + "step": 77400 + }, + { + "epoch": 12.59, + "eval_loss": 0.42326900362968445, + "eval_runtime": 8.9522, + "eval_samples_per_second": 11.17, + "eval_wer": 0.35931790499390986, + "step": 77400 + }, + { + "epoch": 12.6, + "learning_rate": 4.9747103932125964e-05, + "loss": NaN, + "step": 77500 + }, + { + "epoch": 12.6, + "eval_loss": 0.4721744656562805, + "eval_runtime": 9.3092, + "eval_samples_per_second": 10.742, + "eval_wer": 0.3763702801461632, + "step": 77500 + }, + { + "epoch": 12.62, + "learning_rate": 4.9681840430739115e-05, + "loss": NaN, + "step": 77600 + }, + { + "epoch": 12.62, + "eval_loss": 0.40708309412002563, + "eval_runtime": 8.9646, + "eval_samples_per_second": 11.155, + "eval_wer": 0.36053593179049936, + "step": 77600 + }, + { + "epoch": 12.64, + "learning_rate": 4.961657692935226e-05, + "loss": NaN, + "step": 77700 + }, + { + "epoch": 12.64, + "eval_loss": 0.3885576128959656, + "eval_runtime": 6.9779, + "eval_samples_per_second": 14.331, + "eval_wer": 0.3471376370280146, + "step": 77700 + }, + { + "epoch": 12.65, + "learning_rate": 4.955131342796541e-05, + "loss": NaN, + "step": 77800 + }, + { + "epoch": 12.65, + "eval_loss": 0.39132988452911377, + "eval_runtime": 9.1207, + "eval_samples_per_second": 10.964, + "eval_wer": 0.33617539585870887, + "step": 77800 + }, + { + "epoch": 12.67, + "learning_rate": 4.9486049926578564e-05, + "loss": NaN, + "step": 77900 + }, + { + "epoch": 12.67, + "eval_loss": 0.40842190384864807, + "eval_runtime": 7.0913, + "eval_samples_per_second": 14.102, + "eval_wer": 0.35809987819732036, + "step": 77900 + }, + { + "epoch": 12.68, + "learning_rate": 4.942078642519171e-05, + "loss": NaN, + "step": 78000 + }, + { + "epoch": 12.68, + "eval_loss": 0.3725109398365021, + "eval_runtime": 7.7538, + "eval_samples_per_second": 12.897, + "eval_wer": 0.35809987819732036, + "step": 78000 + }, + { + "epoch": 12.7, + "learning_rate": 4.935552292380486e-05, + "loss": NaN, + "step": 78100 + }, + { + "epoch": 12.7, + "eval_loss": 0.5190932750701904, + "eval_runtime": 8.1388, + "eval_samples_per_second": 12.287, + "eval_wer": 0.34835566382460414, + "step": 78100 + }, + { + "epoch": 12.72, + "learning_rate": 4.929025942241802e-05, + "loss": NaN, + "step": 78200 + }, + { + "epoch": 12.72, + "eval_loss": 0.37530940771102905, + "eval_runtime": 7.1222, + "eval_samples_per_second": 14.041, + "eval_wer": 0.35322777101096225, + "step": 78200 + }, + { + "epoch": 12.73, + "learning_rate": 4.9224995921031165e-05, + "loss": NaN, + "step": 78300 + }, + { + "epoch": 12.73, + "eval_loss": 0.4114912748336792, + "eval_runtime": 7.8419, + "eval_samples_per_second": 12.752, + "eval_wer": 0.36053593179049936, + "step": 78300 + }, + { + "epoch": 12.75, + "learning_rate": 4.915973241964432e-05, + "loss": NaN, + "step": 78400 + }, + { + "epoch": 12.75, + "eval_loss": 0.4053696095943451, + "eval_runtime": 6.6142, + "eval_samples_per_second": 15.119, + "eval_wer": 0.3507917174177832, + "step": 78400 + }, + { + "epoch": 12.77, + "learning_rate": 4.909446891825747e-05, + "loss": NaN, + "step": 78500 + }, + { + "epoch": 12.77, + "eval_loss": 0.39146101474761963, + "eval_runtime": 9.407, + "eval_samples_per_second": 10.63, + "eval_wer": 0.3556638246041413, + "step": 78500 + }, + { + "epoch": 12.78, + "learning_rate": 4.9029205416870614e-05, + "loss": NaN, + "step": 78600 + }, + { + "epoch": 12.78, + "eval_loss": 0.40449854731559753, + "eval_runtime": 7.3375, + "eval_samples_per_second": 13.629, + "eval_wer": 0.3568818514007308, + "step": 78600 + }, + { + "epoch": 12.8, + "learning_rate": 4.8963941915483766e-05, + "loss": NaN, + "step": 78700 + }, + { + "epoch": 12.8, + "eval_loss": 0.4020620584487915, + "eval_runtime": 10.3327, + "eval_samples_per_second": 9.678, + "eval_wer": 0.35809987819732036, + "step": 78700 + }, + { + "epoch": 12.82, + "learning_rate": 4.889867841409692e-05, + "loss": NaN, + "step": 78800 + }, + { + "epoch": 12.82, + "eval_loss": 0.39416953921318054, + "eval_runtime": 7.3981, + "eval_samples_per_second": 13.517, + "eval_wer": 0.3507917174177832, + "step": 78800 + }, + { + "epoch": 12.83, + "learning_rate": 4.883341491271007e-05, + "loss": NaN, + "step": 78900 + }, + { + "epoch": 12.83, + "eval_loss": 0.3987289369106293, + "eval_runtime": 8.9882, + "eval_samples_per_second": 11.126, + "eval_wer": 0.3507917174177832, + "step": 78900 + }, + { + "epoch": 12.85, + "learning_rate": 4.876815141132322e-05, + "loss": NaN, + "step": 79000 + }, + { + "epoch": 12.85, + "eval_loss": 0.3875060975551605, + "eval_runtime": 8.7165, + "eval_samples_per_second": 11.472, + "eval_wer": 0.35809987819732036, + "step": 79000 + }, + { + "epoch": 12.86, + "learning_rate": 4.870288790993637e-05, + "loss": NaN, + "step": 79100 + }, + { + "epoch": 12.86, + "eval_loss": 0.4113094210624695, + "eval_runtime": 9.2773, + "eval_samples_per_second": 10.779, + "eval_wer": 0.3702801461632156, + "step": 79100 + }, + { + "epoch": 12.88, + "learning_rate": 4.863762440854952e-05, + "loss": NaN, + "step": 79200 + }, + { + "epoch": 12.88, + "eval_loss": 0.4063952565193176, + "eval_runtime": 8.6562, + "eval_samples_per_second": 11.552, + "eval_wer": 0.3751522533495737, + "step": 79200 + }, + { + "epoch": 12.9, + "learning_rate": 4.857236090716267e-05, + "loss": NaN, + "step": 79300 + }, + { + "epoch": 12.9, + "eval_loss": 0.3837190270423889, + "eval_runtime": 7.4933, + "eval_samples_per_second": 13.345, + "eval_wer": 0.3568818514007308, + "step": 79300 + }, + { + "epoch": 12.91, + "learning_rate": 4.850709740577582e-05, + "loss": NaN, + "step": 79400 + }, + { + "epoch": 12.91, + "eval_loss": 0.40857169032096863, + "eval_runtime": 8.4689, + "eval_samples_per_second": 11.808, + "eval_wer": 0.3568818514007308, + "step": 79400 + }, + { + "epoch": 12.93, + "learning_rate": 4.8441833904388975e-05, + "loss": NaN, + "step": 79500 + }, + { + "epoch": 12.93, + "eval_loss": 0.3926541209220886, + "eval_runtime": 6.7162, + "eval_samples_per_second": 14.889, + "eval_wer": 0.3459196102314251, + "step": 79500 + }, + { + "epoch": 12.95, + "learning_rate": 4.837657040300213e-05, + "loss": NaN, + "step": 79600 + }, + { + "epoch": 12.95, + "eval_loss": 0.4522092044353485, + "eval_runtime": 8.4503, + "eval_samples_per_second": 11.834, + "eval_wer": 0.35322777101096225, + "step": 79600 + }, + { + "epoch": 12.96, + "learning_rate": 4.831130690161527e-05, + "loss": NaN, + "step": 79700 + }, + { + "epoch": 12.96, + "eval_loss": 0.4004019498825073, + "eval_runtime": 7.2333, + "eval_samples_per_second": 13.825, + "eval_wer": 0.35809987819732036, + "step": 79700 + }, + { + "epoch": 12.98, + "learning_rate": 4.8246043400228424e-05, + "loss": NaN, + "step": 79800 + }, + { + "epoch": 12.98, + "eval_loss": 0.36814793944358826, + "eval_runtime": 9.1464, + "eval_samples_per_second": 10.933, + "eval_wer": 0.34348355663824603, + "step": 79800 + }, + { + "epoch": 12.99, + "learning_rate": 4.8180779898841576e-05, + "loss": NaN, + "step": 79900 + }, + { + "epoch": 12.99, + "eval_loss": 0.3880935609340668, + "eval_runtime": 8.0402, + "eval_samples_per_second": 12.438, + "eval_wer": 0.364190012180268, + "step": 79900 + }, + { + "epoch": 13.01, + "learning_rate": 4.811551639745472e-05, + "loss": NaN, + "step": 80000 + }, + { + "epoch": 13.01, + "eval_loss": 0.39711371064186096, + "eval_runtime": 8.0714, + "eval_samples_per_second": 12.389, + "eval_wer": 0.3520097442143727, + "step": 80000 + }, + { + "epoch": 13.03, + "learning_rate": 4.805025289606788e-05, + "loss": NaN, + "step": 80100 + }, + { + "epoch": 13.03, + "eval_loss": 0.4093989133834839, + "eval_runtime": 8.8057, + "eval_samples_per_second": 11.356, + "eval_wer": 0.3568818514007308, + "step": 80100 + }, + { + "epoch": 13.04, + "learning_rate": 4.798498939468103e-05, + "loss": NaN, + "step": 80200 + }, + { + "epoch": 13.04, + "eval_loss": 0.39422935247421265, + "eval_runtime": 8.3368, + "eval_samples_per_second": 11.995, + "eval_wer": 0.36784409257003653, + "step": 80200 + }, + { + "epoch": 13.06, + "learning_rate": 4.7919725893294177e-05, + "loss": NaN, + "step": 80300 + }, + { + "epoch": 13.06, + "eval_loss": 0.40601789951324463, + "eval_runtime": 7.678, + "eval_samples_per_second": 13.024, + "eval_wer": 0.36053593179049936, + "step": 80300 + }, + { + "epoch": 13.08, + "learning_rate": 4.785446239190733e-05, + "loss": NaN, + "step": 80400 + }, + { + "epoch": 13.08, + "eval_loss": 0.42190930247306824, + "eval_runtime": 7.1522, + "eval_samples_per_second": 13.982, + "eval_wer": 0.3568818514007308, + "step": 80400 + }, + { + "epoch": 13.09, + "learning_rate": 4.778919889052048e-05, + "loss": NaN, + "step": 80500 + }, + { + "epoch": 13.09, + "eval_loss": 0.45500144362449646, + "eval_runtime": 7.293, + "eval_samples_per_second": 13.712, + "eval_wer": 0.3568818514007308, + "step": 80500 + }, + { + "epoch": 13.11, + "learning_rate": 4.7723935389133625e-05, + "loss": NaN, + "step": 80600 + }, + { + "epoch": 13.11, + "eval_loss": 0.4279215335845947, + "eval_runtime": 8.0851, + "eval_samples_per_second": 12.368, + "eval_wer": 0.364190012180268, + "step": 80600 + }, + { + "epoch": 13.12, + "learning_rate": 4.7658671887746784e-05, + "loss": NaN, + "step": 80700 + }, + { + "epoch": 13.12, + "eval_loss": 0.379155695438385, + "eval_runtime": 8.3158, + "eval_samples_per_second": 12.025, + "eval_wer": 0.3447015834348356, + "step": 80700 + }, + { + "epoch": 13.14, + "learning_rate": 4.759340838635993e-05, + "loss": NaN, + "step": 80800 + }, + { + "epoch": 13.14, + "eval_loss": 0.41694533824920654, + "eval_runtime": 9.1011, + "eval_samples_per_second": 10.988, + "eval_wer": 0.3702801461632156, + "step": 80800 + }, + { + "epoch": 13.16, + "learning_rate": 4.752814488497308e-05, + "loss": NaN, + "step": 80900 + }, + { + "epoch": 13.16, + "eval_loss": 0.43820008635520935, + "eval_runtime": 8.0662, + "eval_samples_per_second": 12.397, + "eval_wer": 0.35931790499390986, + "step": 80900 + }, + { + "epoch": 13.17, + "learning_rate": 4.746288138358623e-05, + "loss": NaN, + "step": 81000 + }, + { + "epoch": 13.17, + "eval_loss": 0.38130292296409607, + "eval_runtime": 6.6507, + "eval_samples_per_second": 15.036, + "eval_wer": 0.3471376370280146, + "step": 81000 + }, + { + "epoch": 13.19, + "learning_rate": 4.739761788219938e-05, + "loss": NaN, + "step": 81100 + }, + { + "epoch": 13.19, + "eval_loss": 0.4210968315601349, + "eval_runtime": 8.56, + "eval_samples_per_second": 11.682, + "eval_wer": 0.35322777101096225, + "step": 81100 + }, + { + "epoch": 13.21, + "learning_rate": 4.733235438081253e-05, + "loss": NaN, + "step": 81200 + }, + { + "epoch": 13.21, + "eval_loss": 0.379304438829422, + "eval_runtime": 7.6806, + "eval_samples_per_second": 13.02, + "eval_wer": 0.3373934226552984, + "step": 81200 + }, + { + "epoch": 13.22, + "learning_rate": 4.726709087942568e-05, + "loss": NaN, + "step": 81300 + }, + { + "epoch": 13.22, + "eval_loss": 0.4642069935798645, + "eval_runtime": 7.3598, + "eval_samples_per_second": 13.587, + "eval_wer": 0.3629719853836784, + "step": 81300 + }, + { + "epoch": 13.24, + "learning_rate": 4.7201827378038834e-05, + "loss": NaN, + "step": 81400 + }, + { + "epoch": 13.24, + "eval_loss": 0.39892637729644775, + "eval_runtime": 8.1349, + "eval_samples_per_second": 12.293, + "eval_wer": 0.35444579780755175, + "step": 81400 + }, + { + "epoch": 13.25, + "learning_rate": 4.7136563876651986e-05, + "loss": NaN, + "step": 81500 + }, + { + "epoch": 13.25, + "eval_loss": 0.3699035942554474, + "eval_runtime": 7.0381, + "eval_samples_per_second": 14.208, + "eval_wer": 0.3386114494518879, + "step": 81500 + }, + { + "epoch": 13.27, + "learning_rate": 4.707130037526514e-05, + "loss": NaN, + "step": 81600 + }, + { + "epoch": 13.27, + "eval_loss": 0.4748342037200928, + "eval_runtime": 8.5553, + "eval_samples_per_second": 11.689, + "eval_wer": 0.3654080389768575, + "step": 81600 + }, + { + "epoch": 13.29, + "learning_rate": 4.700603687387828e-05, + "loss": NaN, + "step": 81700 + }, + { + "epoch": 13.29, + "eval_loss": 0.4026525616645813, + "eval_runtime": 6.9498, + "eval_samples_per_second": 14.389, + "eval_wer": 0.36053593179049936, + "step": 81700 + }, + { + "epoch": 13.3, + "learning_rate": 4.6940773372491435e-05, + "loss": NaN, + "step": 81800 + }, + { + "epoch": 13.3, + "eval_loss": 0.4426953196525574, + "eval_runtime": 9.5632, + "eval_samples_per_second": 10.457, + "eval_wer": 0.36053593179049936, + "step": 81800 + }, + { + "epoch": 13.32, + "learning_rate": 4.687550987110459e-05, + "loss": NaN, + "step": 81900 + }, + { + "epoch": 13.32, + "eval_loss": 0.4907471835613251, + "eval_runtime": 9.1012, + "eval_samples_per_second": 10.988, + "eval_wer": 0.38246041412911086, + "step": 81900 + }, + { + "epoch": 13.34, + "learning_rate": 4.681024636971774e-05, + "loss": NaN, + "step": 82000 + }, + { + "epoch": 13.34, + "eval_loss": 0.39570945501327515, + "eval_runtime": 9.1482, + "eval_samples_per_second": 10.931, + "eval_wer": 0.35931790499390986, + "step": 82000 + }, + { + "epoch": 13.35, + "learning_rate": 4.674498286833089e-05, + "loss": NaN, + "step": 82100 + }, + { + "epoch": 13.35, + "eval_loss": 0.39692896604537964, + "eval_runtime": 8.067, + "eval_samples_per_second": 12.396, + "eval_wer": 0.3520097442143727, + "step": 82100 + }, + { + "epoch": 13.37, + "learning_rate": 4.667971936694404e-05, + "loss": NaN, + "step": 82200 + }, + { + "epoch": 13.37, + "eval_loss": 0.41192084550857544, + "eval_runtime": 8.128, + "eval_samples_per_second": 12.303, + "eval_wer": 0.3617539585870889, + "step": 82200 + }, + { + "epoch": 13.38, + "learning_rate": 4.661445586555719e-05, + "loss": NaN, + "step": 82300 + }, + { + "epoch": 13.38, + "eval_loss": 0.38953062891960144, + "eval_runtime": 9.918, + "eval_samples_per_second": 10.083, + "eval_wer": 0.36053593179049936, + "step": 82300 + }, + { + "epoch": 13.4, + "learning_rate": 4.654919236417034e-05, + "loss": NaN, + "step": 82400 + }, + { + "epoch": 13.4, + "eval_loss": 0.37186986207962036, + "eval_runtime": 8.1376, + "eval_samples_per_second": 12.289, + "eval_wer": 0.34835566382460414, + "step": 82400 + }, + { + "epoch": 13.42, + "learning_rate": 4.6483928862783485e-05, + "loss": NaN, + "step": 82500 + }, + { + "epoch": 13.42, + "eval_loss": 0.40984976291656494, + "eval_runtime": 8.9721, + "eval_samples_per_second": 11.146, + "eval_wer": 0.34226552984165654, + "step": 82500 + }, + { + "epoch": 13.43, + "learning_rate": 4.6418665361396643e-05, + "loss": NaN, + "step": 82600 + }, + { + "epoch": 13.43, + "eval_loss": 0.4037730097770691, + "eval_runtime": 6.8614, + "eval_samples_per_second": 14.574, + "eval_wer": 0.34226552984165654, + "step": 82600 + }, + { + "epoch": 13.45, + "learning_rate": 4.6353401860009795e-05, + "loss": NaN, + "step": 82700 + }, + { + "epoch": 13.45, + "eval_loss": 0.38672852516174316, + "eval_runtime": 7.8242, + "eval_samples_per_second": 12.781, + "eval_wer": 0.3556638246041413, + "step": 82700 + }, + { + "epoch": 13.47, + "learning_rate": 4.628813835862294e-05, + "loss": NaN, + "step": 82800 + }, + { + "epoch": 13.47, + "eval_loss": 0.41303497552871704, + "eval_runtime": 7.7371, + "eval_samples_per_second": 12.925, + "eval_wer": 0.34957369062119364, + "step": 82800 + }, + { + "epoch": 13.48, + "learning_rate": 4.622287485723609e-05, + "loss": NaN, + "step": 82900 + }, + { + "epoch": 13.48, + "eval_loss": 0.3807326853275299, + "eval_runtime": 8.6603, + "eval_samples_per_second": 11.547, + "eval_wer": 0.35444579780755175, + "step": 82900 + }, + { + "epoch": 13.5, + "learning_rate": 4.6157611355849244e-05, + "loss": NaN, + "step": 83000 + }, + { + "epoch": 13.5, + "eval_loss": 0.3885740637779236, + "eval_runtime": 8.1938, + "eval_samples_per_second": 12.204, + "eval_wer": 0.34348355663824603, + "step": 83000 + }, + { + "epoch": 13.51, + "learning_rate": 4.609234785446239e-05, + "loss": NaN, + "step": 83100 + }, + { + "epoch": 13.51, + "eval_loss": 0.40991002321243286, + "eval_runtime": 8.0699, + "eval_samples_per_second": 12.392, + "eval_wer": 0.34957369062119364, + "step": 83100 + }, + { + "epoch": 13.53, + "learning_rate": 4.602708435307554e-05, + "loss": NaN, + "step": 83200 + }, + { + "epoch": 13.53, + "eval_loss": 0.45967453718185425, + "eval_runtime": 9.6326, + "eval_samples_per_second": 10.381, + "eval_wer": 0.3654080389768575, + "step": 83200 + }, + { + "epoch": 13.55, + "learning_rate": 4.59618208516887e-05, + "loss": NaN, + "step": 83300 + }, + { + "epoch": 13.55, + "eval_loss": 0.41274598240852356, + "eval_runtime": 7.4414, + "eval_samples_per_second": 13.438, + "eval_wer": 0.3520097442143727, + "step": 83300 + }, + { + "epoch": 13.56, + "learning_rate": 4.5896557350301845e-05, + "loss": NaN, + "step": 83400 + }, + { + "epoch": 13.56, + "eval_loss": 0.39359050989151, + "eval_runtime": 7.6776, + "eval_samples_per_second": 13.025, + "eval_wer": 0.3520097442143727, + "step": 83400 + }, + { + "epoch": 13.58, + "learning_rate": 4.5831293848915e-05, + "loss": NaN, + "step": 83500 + }, + { + "epoch": 13.58, + "eval_loss": 0.3984470069408417, + "eval_runtime": 8.3263, + "eval_samples_per_second": 12.01, + "eval_wer": 0.3556638246041413, + "step": 83500 + }, + { + "epoch": 13.6, + "learning_rate": 4.576603034752815e-05, + "loss": NaN, + "step": 83600 + }, + { + "epoch": 13.6, + "eval_loss": 0.39405977725982666, + "eval_runtime": 7.0798, + "eval_samples_per_second": 14.125, + "eval_wer": 0.35809987819732036, + "step": 83600 + }, + { + "epoch": 13.61, + "learning_rate": 4.5700766846141294e-05, + "loss": NaN, + "step": 83700 + }, + { + "epoch": 13.61, + "eval_loss": 0.45130589604377747, + "eval_runtime": 7.6376, + "eval_samples_per_second": 13.093, + "eval_wer": 0.3629719853836784, + "step": 83700 + }, + { + "epoch": 13.63, + "learning_rate": 4.5635503344754446e-05, + "loss": NaN, + "step": 83800 + }, + { + "epoch": 13.63, + "eval_loss": 0.4328234791755676, + "eval_runtime": 6.7166, + "eval_samples_per_second": 14.889, + "eval_wer": 0.3520097442143727, + "step": 83800 + }, + { + "epoch": 13.64, + "learning_rate": 4.55702398433676e-05, + "loss": NaN, + "step": 83900 + }, + { + "epoch": 13.64, + "eval_loss": 0.40169548988342285, + "eval_runtime": 8.601, + "eval_samples_per_second": 11.626, + "eval_wer": 0.35444579780755175, + "step": 83900 + }, + { + "epoch": 13.66, + "learning_rate": 4.550497634198075e-05, + "loss": NaN, + "step": 84000 + }, + { + "epoch": 13.66, + "eval_loss": 0.40670937299728394, + "eval_runtime": 8.6181, + "eval_samples_per_second": 11.603, + "eval_wer": 0.36662606577344703, + "step": 84000 + }, + { + "epoch": 13.68, + "learning_rate": 4.54397128405939e-05, + "loss": NaN, + "step": 84100 + }, + { + "epoch": 13.68, + "eval_loss": 0.4055994153022766, + "eval_runtime": 8.9478, + "eval_samples_per_second": 11.176, + "eval_wer": 0.3556638246041413, + "step": 84100 + }, + { + "epoch": 13.69, + "learning_rate": 4.537444933920705e-05, + "loss": NaN, + "step": 84200 + }, + { + "epoch": 13.69, + "eval_loss": 0.37059950828552246, + "eval_runtime": 8.4585, + "eval_samples_per_second": 11.822, + "eval_wer": 0.3520097442143727, + "step": 84200 + }, + { + "epoch": 13.71, + "learning_rate": 4.53091858378202e-05, + "loss": NaN, + "step": 84300 + }, + { + "epoch": 13.71, + "eval_loss": 0.3316919207572937, + "eval_runtime": 7.6508, + "eval_samples_per_second": 13.071, + "eval_wer": 0.32521315468940315, + "step": 84300 + }, + { + "epoch": 13.73, + "learning_rate": 4.524392233643335e-05, + "loss": NaN, + "step": 84400 + }, + { + "epoch": 13.73, + "eval_loss": 0.38926997780799866, + "eval_runtime": 7.2992, + "eval_samples_per_second": 13.7, + "eval_wer": 0.3459196102314251, + "step": 84400 + }, + { + "epoch": 13.74, + "learning_rate": 4.51786588350465e-05, + "loss": NaN, + "step": 84500 + }, + { + "epoch": 13.74, + "eval_loss": 0.3776205778121948, + "eval_runtime": 7.205, + "eval_samples_per_second": 13.879, + "eval_wer": 0.3398294762484775, + "step": 84500 + }, + { + "epoch": 13.76, + "learning_rate": 4.5113395333659655e-05, + "loss": NaN, + "step": 84600 + }, + { + "epoch": 13.76, + "eval_loss": 0.4353153705596924, + "eval_runtime": 8.1838, + "eval_samples_per_second": 12.219, + "eval_wer": 0.34957369062119364, + "step": 84600 + }, + { + "epoch": 13.77, + "learning_rate": 4.5048131832272807e-05, + "loss": NaN, + "step": 84700 + }, + { + "epoch": 13.77, + "eval_loss": 0.3977775573730469, + "eval_runtime": 8.5838, + "eval_samples_per_second": 11.65, + "eval_wer": 0.34226552984165654, + "step": 84700 + }, + { + "epoch": 13.79, + "learning_rate": 4.498286833088595e-05, + "loss": NaN, + "step": 84800 + }, + { + "epoch": 13.79, + "eval_loss": 0.3765636384487152, + "eval_runtime": 7.8414, + "eval_samples_per_second": 12.753, + "eval_wer": 0.34226552984165654, + "step": 84800 + }, + { + "epoch": 13.81, + "learning_rate": 4.4917604829499104e-05, + "loss": NaN, + "step": 84900 + }, + { + "epoch": 13.81, + "eval_loss": 0.3906586766242981, + "eval_runtime": 9.8562, + "eval_samples_per_second": 10.146, + "eval_wer": 0.36053593179049936, + "step": 84900 + }, + { + "epoch": 13.82, + "learning_rate": 4.4852341328112256e-05, + "loss": NaN, + "step": 85000 + }, + { + "epoch": 13.82, + "eval_loss": 0.39076992869377136, + "eval_runtime": 7.9184, + "eval_samples_per_second": 12.629, + "eval_wer": 0.34348355663824603, + "step": 85000 + }, + { + "epoch": 13.84, + "learning_rate": 4.478707782672541e-05, + "loss": NaN, + "step": 85100 + }, + { + "epoch": 13.84, + "eval_loss": 0.38430699706077576, + "eval_runtime": 7.852, + "eval_samples_per_second": 12.736, + "eval_wer": 0.3471376370280146, + "step": 85100 + }, + { + "epoch": 13.86, + "learning_rate": 4.472181432533856e-05, + "loss": NaN, + "step": 85200 + }, + { + "epoch": 13.86, + "eval_loss": 0.36845266819000244, + "eval_runtime": 7.4373, + "eval_samples_per_second": 13.446, + "eval_wer": 0.3459196102314251, + "step": 85200 + }, + { + "epoch": 13.87, + "learning_rate": 4.465655082395171e-05, + "loss": NaN, + "step": 85300 + }, + { + "epoch": 13.87, + "eval_loss": 0.39804258942604065, + "eval_runtime": 9.8439, + "eval_samples_per_second": 10.159, + "eval_wer": 0.3459196102314251, + "step": 85300 + }, + { + "epoch": 13.89, + "learning_rate": 4.4591287322564856e-05, + "loss": NaN, + "step": 85400 + }, + { + "epoch": 13.89, + "eval_loss": 0.3661288321018219, + "eval_runtime": 8.5748, + "eval_samples_per_second": 11.662, + "eval_wer": 0.3398294762484775, + "step": 85400 + }, + { + "epoch": 13.9, + "learning_rate": 4.452602382117801e-05, + "loss": NaN, + "step": 85500 + }, + { + "epoch": 13.9, + "eval_loss": 0.3940080404281616, + "eval_runtime": 8.4536, + "eval_samples_per_second": 11.829, + "eval_wer": 0.35444579780755175, + "step": 85500 + }, + { + "epoch": 13.92, + "learning_rate": 4.446076031979116e-05, + "loss": NaN, + "step": 85600 + }, + { + "epoch": 13.92, + "eval_loss": 0.38597309589385986, + "eval_runtime": 7.5301, + "eval_samples_per_second": 13.28, + "eval_wer": 0.3471376370280146, + "step": 85600 + }, + { + "epoch": 13.94, + "learning_rate": 4.4395496818404305e-05, + "loss": NaN, + "step": 85700 + }, + { + "epoch": 13.94, + "eval_loss": 0.3475070297718048, + "eval_runtime": 7.4895, + "eval_samples_per_second": 13.352, + "eval_wer": 0.3447015834348356, + "step": 85700 + }, + { + "epoch": 13.95, + "learning_rate": 4.4330233317017464e-05, + "loss": NaN, + "step": 85800 + }, + { + "epoch": 13.95, + "eval_loss": 0.3472925126552582, + "eval_runtime": 7.1585, + "eval_samples_per_second": 13.969, + "eval_wer": 0.3459196102314251, + "step": 85800 + }, + { + "epoch": 13.97, + "learning_rate": 4.426496981563061e-05, + "loss": NaN, + "step": 85900 + }, + { + "epoch": 13.97, + "eval_loss": 0.4307803213596344, + "eval_runtime": 8.4562, + "eval_samples_per_second": 11.826, + "eval_wer": 0.35444579780755175, + "step": 85900 + }, + { + "epoch": 13.99, + "learning_rate": 4.419970631424376e-05, + "loss": NaN, + "step": 86000 + }, + { + "epoch": 13.99, + "eval_loss": 0.3824266791343689, + "eval_runtime": 7.6614, + "eval_samples_per_second": 13.052, + "eval_wer": 0.35444579780755175, + "step": 86000 + }, + { + "epoch": 14.0, + "learning_rate": 4.413444281285691e-05, + "loss": NaN, + "step": 86100 + }, + { + "epoch": 14.0, + "eval_loss": 0.4147208333015442, + "eval_runtime": 8.0603, + "eval_samples_per_second": 12.407, + "eval_wer": 0.34348355663824603, + "step": 86100 + }, + { + "epoch": 14.02, + "learning_rate": 4.406917931147006e-05, + "loss": NaN, + "step": 86200 + }, + { + "epoch": 14.02, + "eval_loss": 0.37703704833984375, + "eval_runtime": 8.6635, + "eval_samples_per_second": 11.543, + "eval_wer": 0.35322777101096225, + "step": 86200 + }, + { + "epoch": 14.03, + "learning_rate": 4.400391581008321e-05, + "loss": NaN, + "step": 86300 + }, + { + "epoch": 14.03, + "eval_loss": 0.38355034589767456, + "eval_runtime": 8.3757, + "eval_samples_per_second": 11.939, + "eval_wer": 0.3507917174177832, + "step": 86300 + }, + { + "epoch": 14.05, + "learning_rate": 4.393865230869636e-05, + "loss": NaN, + "step": 86400 + }, + { + "epoch": 14.05, + "eval_loss": 0.4052790403366089, + "eval_runtime": 9.7614, + "eval_samples_per_second": 10.244, + "eval_wer": 0.3398294762484775, + "step": 86400 + }, + { + "epoch": 14.07, + "learning_rate": 4.3873388807309514e-05, + "loss": NaN, + "step": 86500 + }, + { + "epoch": 14.07, + "eval_loss": 0.3577238917350769, + "eval_runtime": 7.8702, + "eval_samples_per_second": 12.706, + "eval_wer": 0.34348355663824603, + "step": 86500 + }, + { + "epoch": 14.08, + "learning_rate": 4.3808125305922666e-05, + "loss": NaN, + "step": 86600 + }, + { + "epoch": 14.08, + "eval_loss": 0.3567332327365875, + "eval_runtime": 8.207, + "eval_samples_per_second": 12.185, + "eval_wer": 0.3373934226552984, + "step": 86600 + }, + { + "epoch": 14.1, + "learning_rate": 4.374286180453582e-05, + "loss": NaN, + "step": 86700 + }, + { + "epoch": 14.1, + "eval_loss": 0.46958306431770325, + "eval_runtime": 7.2452, + "eval_samples_per_second": 13.802, + "eval_wer": 0.35444579780755175, + "step": 86700 + }, + { + "epoch": 14.12, + "learning_rate": 4.367759830314896e-05, + "loss": NaN, + "step": 86800 + }, + { + "epoch": 14.12, + "eval_loss": 0.37138259410858154, + "eval_runtime": 8.0019, + "eval_samples_per_second": 12.497, + "eval_wer": 0.3325213154689403, + "step": 86800 + }, + { + "epoch": 14.13, + "learning_rate": 4.3612334801762115e-05, + "loss": 0.1053, + "step": 86900 + }, + { + "epoch": 14.13, + "eval_loss": 0.3933388590812683, + "eval_runtime": 7.7892, + "eval_samples_per_second": 12.838, + "eval_wer": 0.35444579780755175, + "step": 86900 + }, + { + "epoch": 14.15, + "learning_rate": 4.354707130037527e-05, + "loss": 0.1085, + "step": 87000 + }, + { + "epoch": 14.15, + "eval_loss": 0.4454494118690491, + "eval_runtime": 9.0708, + "eval_samples_per_second": 11.024, + "eval_wer": 0.34835566382460414, + "step": 87000 + }, + { + "epoch": 14.16, + "learning_rate": 4.348180779898842e-05, + "loss": 0.1044, + "step": 87100 + }, + { + "epoch": 14.16, + "eval_loss": 0.35896095633506775, + "eval_runtime": 7.3353, + "eval_samples_per_second": 13.633, + "eval_wer": 0.341047503045067, + "step": 87100 + }, + { + "epoch": 14.18, + "learning_rate": 4.341654429760157e-05, + "loss": 0.1131, + "step": 87200 + }, + { + "epoch": 14.18, + "eval_loss": 0.4209243655204773, + "eval_runtime": 7.4199, + "eval_samples_per_second": 13.477, + "eval_wer": 0.3459196102314251, + "step": 87200 + }, + { + "epoch": 14.2, + "learning_rate": 4.335128079621472e-05, + "loss": 0.1069, + "step": 87300 + }, + { + "epoch": 14.2, + "eval_loss": 0.4241653084754944, + "eval_runtime": 7.8927, + "eval_samples_per_second": 12.67, + "eval_wer": 0.341047503045067, + "step": 87300 + }, + { + "epoch": 14.21, + "learning_rate": 4.328601729482787e-05, + "loss": 0.1042, + "step": 87400 + }, + { + "epoch": 14.21, + "eval_loss": 0.3520914912223816, + "eval_runtime": 7.2723, + "eval_samples_per_second": 13.751, + "eval_wer": 0.3459196102314251, + "step": 87400 + }, + { + "epoch": 14.23, + "learning_rate": 4.322075379344102e-05, + "loss": 0.1017, + "step": 87500 + }, + { + "epoch": 14.23, + "eval_loss": 0.3735422194004059, + "eval_runtime": 7.8531, + "eval_samples_per_second": 12.734, + "eval_wer": 0.3337393422655298, + "step": 87500 + }, + { + "epoch": 14.25, + "learning_rate": 4.315549029205417e-05, + "loss": 0.099, + "step": 87600 + }, + { + "epoch": 14.25, + "eval_loss": 0.4072035253047943, + "eval_runtime": 6.9486, + "eval_samples_per_second": 14.391, + "eval_wer": 0.3520097442143727, + "step": 87600 + }, + { + "epoch": 14.26, + "learning_rate": 4.3090226790667323e-05, + "loss": 0.1119, + "step": 87700 + }, + { + "epoch": 14.26, + "eval_loss": 0.4090457856655121, + "eval_runtime": 9.1301, + "eval_samples_per_second": 10.953, + "eval_wer": 0.33008526187576126, + "step": 87700 + }, + { + "epoch": 14.28, + "learning_rate": 4.3024963289280475e-05, + "loss": 0.1072, + "step": 87800 + }, + { + "epoch": 14.28, + "eval_loss": 0.3945271372795105, + "eval_runtime": 8.9475, + "eval_samples_per_second": 11.176, + "eval_wer": 0.3556638246041413, + "step": 87800 + }, + { + "epoch": 14.3, + "learning_rate": 4.295969978789362e-05, + "loss": 0.1132, + "step": 87900 + }, + { + "epoch": 14.3, + "eval_loss": 0.3662361800670624, + "eval_runtime": 7.1294, + "eval_samples_per_second": 14.026, + "eval_wer": 0.33617539585870887, + "step": 87900 + }, + { + "epoch": 14.31, + "learning_rate": 4.289443628650677e-05, + "loss": 0.1037, + "step": 88000 + }, + { + "epoch": 14.31, + "eval_loss": 0.3741585910320282, + "eval_runtime": 7.5993, + "eval_samples_per_second": 13.159, + "eval_wer": 0.3398294762484775, + "step": 88000 + }, + { + "epoch": 14.33, + "learning_rate": 4.2829172785119924e-05, + "loss": 0.1045, + "step": 88100 + }, + { + "epoch": 14.33, + "eval_loss": 0.41879794001579285, + "eval_runtime": 7.0732, + "eval_samples_per_second": 14.138, + "eval_wer": 0.3568818514007308, + "step": 88100 + }, + { + "epoch": 14.34, + "learning_rate": 4.276390928373307e-05, + "loss": 0.0999, + "step": 88200 + }, + { + "epoch": 14.34, + "eval_loss": 0.39614391326904297, + "eval_runtime": 9.6972, + "eval_samples_per_second": 10.312, + "eval_wer": 0.35444579780755175, + "step": 88200 + }, + { + "epoch": 14.36, + "learning_rate": 4.269864578234623e-05, + "loss": 0.1047, + "step": 88300 + }, + { + "epoch": 14.36, + "eval_loss": 0.3773258328437805, + "eval_runtime": 8.0348, + "eval_samples_per_second": 12.446, + "eval_wer": 0.34835566382460414, + "step": 88300 + }, + { + "epoch": 14.38, + "learning_rate": 4.263338228095938e-05, + "loss": 0.114, + "step": 88400 + }, + { + "epoch": 14.38, + "eval_loss": 0.37774306535720825, + "eval_runtime": 7.4926, + "eval_samples_per_second": 13.346, + "eval_wer": 0.34835566382460414, + "step": 88400 + }, + { + "epoch": 14.39, + "learning_rate": 4.2568118779572525e-05, + "loss": 0.1038, + "step": 88500 + }, + { + "epoch": 14.39, + "eval_loss": 0.3713986575603485, + "eval_runtime": 7.4849, + "eval_samples_per_second": 13.36, + "eval_wer": 0.34957369062119364, + "step": 88500 + }, + { + "epoch": 14.41, + "learning_rate": 4.250285527818568e-05, + "loss": 0.0978, + "step": 88600 + }, + { + "epoch": 14.41, + "eval_loss": 0.3625016510486603, + "eval_runtime": 7.1558, + "eval_samples_per_second": 13.975, + "eval_wer": 0.3398294762484775, + "step": 88600 + }, + { + "epoch": 14.43, + "learning_rate": 4.243759177679883e-05, + "loss": 0.0954, + "step": 88700 + }, + { + "epoch": 14.43, + "eval_loss": 0.40445342659950256, + "eval_runtime": 7.6543, + "eval_samples_per_second": 13.064, + "eval_wer": 0.34835566382460414, + "step": 88700 + }, + { + "epoch": 14.44, + "learning_rate": 4.2372328275411974e-05, + "loss": 0.1027, + "step": 88800 + }, + { + "epoch": 14.44, + "eval_loss": 0.4006795585155487, + "eval_runtime": 7.2382, + "eval_samples_per_second": 13.816, + "eval_wer": 0.34226552984165654, + "step": 88800 + }, + { + "epoch": 14.46, + "learning_rate": 4.2307064774025126e-05, + "loss": Infinity, + "step": 88900 + }, + { + "epoch": 14.46, + "eval_loss": 0.35563650727272034, + "eval_runtime": 8.396, + "eval_samples_per_second": 11.91, + "eval_wer": 0.34957369062119364, + "step": 88900 + }, + { + "epoch": 14.47, + "learning_rate": 4.2241801272638285e-05, + "loss": NaN, + "step": 89000 + }, + { + "epoch": 14.47, + "eval_loss": 0.3528667688369751, + "eval_runtime": 7.9985, + "eval_samples_per_second": 12.502, + "eval_wer": 0.34835566382460414, + "step": 89000 + }, + { + "epoch": 14.49, + "learning_rate": 4.217653777125143e-05, + "loss": NaN, + "step": 89100 + }, + { + "epoch": 14.49, + "eval_loss": 0.416147381067276, + "eval_runtime": 7.9316, + "eval_samples_per_second": 12.608, + "eval_wer": 0.35444579780755175, + "step": 89100 + }, + { + "epoch": 14.51, + "learning_rate": 4.211127426986458e-05, + "loss": NaN, + "step": 89200 + }, + { + "epoch": 14.51, + "eval_loss": 0.4058997631072998, + "eval_runtime": 8.2486, + "eval_samples_per_second": 12.123, + "eval_wer": 0.3520097442143727, + "step": 89200 + }, + { + "epoch": 14.52, + "learning_rate": 4.204601076847773e-05, + "loss": NaN, + "step": 89300 + }, + { + "epoch": 14.52, + "eval_loss": 0.38170912861824036, + "eval_runtime": 9.2084, + "eval_samples_per_second": 10.86, + "eval_wer": 0.3398294762484775, + "step": 89300 + }, + { + "epoch": 14.54, + "learning_rate": 4.198074726709088e-05, + "loss": NaN, + "step": 89400 + }, + { + "epoch": 14.54, + "eval_loss": 0.37428218126296997, + "eval_runtime": 8.1719, + "eval_samples_per_second": 12.237, + "eval_wer": 0.341047503045067, + "step": 89400 + }, + { + "epoch": 14.56, + "learning_rate": 4.191548376570403e-05, + "loss": NaN, + "step": 89500 + }, + { + "epoch": 14.56, + "eval_loss": 0.3986841142177582, + "eval_runtime": 10.8754, + "eval_samples_per_second": 9.195, + "eval_wer": 0.34835566382460414, + "step": 89500 + }, + { + "epoch": 14.57, + "learning_rate": 4.185022026431718e-05, + "loss": NaN, + "step": 89600 + }, + { + "epoch": 14.57, + "eval_loss": 0.38741618394851685, + "eval_runtime": 9.3064, + "eval_samples_per_second": 10.745, + "eval_wer": 0.3520097442143727, + "step": 89600 + }, + { + "epoch": 14.59, + "learning_rate": 4.1784956762930335e-05, + "loss": NaN, + "step": 89700 + }, + { + "epoch": 14.59, + "eval_loss": 0.4405732750892639, + "eval_runtime": 7.6728, + "eval_samples_per_second": 13.033, + "eval_wer": 0.3471376370280146, + "step": 89700 + }, + { + "epoch": 14.6, + "learning_rate": 4.1719693261543487e-05, + "loss": NaN, + "step": 89800 + }, + { + "epoch": 14.6, + "eval_loss": 0.3700186014175415, + "eval_runtime": 7.4389, + "eval_samples_per_second": 13.443, + "eval_wer": 0.341047503045067, + "step": 89800 + }, + { + "epoch": 14.62, + "learning_rate": 4.165442976015663e-05, + "loss": NaN, + "step": 89900 + }, + { + "epoch": 14.62, + "eval_loss": 0.4012463092803955, + "eval_runtime": 8.227, + "eval_samples_per_second": 12.155, + "eval_wer": 0.3386114494518879, + "step": 89900 + }, + { + "epoch": 14.64, + "learning_rate": 4.1589166258769784e-05, + "loss": NaN, + "step": 90000 + }, + { + "epoch": 14.64, + "eval_loss": 0.4593057930469513, + "eval_runtime": 7.5267, + "eval_samples_per_second": 13.286, + "eval_wer": 0.3471376370280146, + "step": 90000 + }, + { + "epoch": 14.65, + "learning_rate": 4.1523902757382936e-05, + "loss": NaN, + "step": 90100 + }, + { + "epoch": 14.65, + "eval_loss": 0.3738018870353699, + "eval_runtime": 8.684, + "eval_samples_per_second": 11.515, + "eval_wer": 0.3325213154689403, + "step": 90100 + }, + { + "epoch": 14.67, + "learning_rate": 4.145863925599609e-05, + "loss": NaN, + "step": 90200 + }, + { + "epoch": 14.67, + "eval_loss": 0.4285886287689209, + "eval_runtime": 8.4528, + "eval_samples_per_second": 11.83, + "eval_wer": 0.35322777101096225, + "step": 90200 + }, + { + "epoch": 14.69, + "learning_rate": 4.139337575460924e-05, + "loss": NaN, + "step": 90300 + }, + { + "epoch": 14.69, + "eval_loss": 0.3823952078819275, + "eval_runtime": 8.2181, + "eval_samples_per_second": 12.168, + "eval_wer": 0.33008526187576126, + "step": 90300 + }, + { + "epoch": 14.7, + "learning_rate": 4.132811225322239e-05, + "loss": NaN, + "step": 90400 + }, + { + "epoch": 14.7, + "eval_loss": 0.4831365644931793, + "eval_runtime": 9.718, + "eval_samples_per_second": 10.29, + "eval_wer": 0.3373934226552984, + "step": 90400 + }, + { + "epoch": 14.72, + "learning_rate": 4.1262848751835536e-05, + "loss": NaN, + "step": 90500 + }, + { + "epoch": 14.72, + "eval_loss": 0.40132513642311096, + "eval_runtime": 8.1178, + "eval_samples_per_second": 12.319, + "eval_wer": 0.34835566382460414, + "step": 90500 + }, + { + "epoch": 14.73, + "learning_rate": 4.119758525044869e-05, + "loss": NaN, + "step": 90600 + }, + { + "epoch": 14.73, + "eval_loss": 0.40187588334083557, + "eval_runtime": 9.5226, + "eval_samples_per_second": 10.501, + "eval_wer": 0.35931790499390986, + "step": 90600 + }, + { + "epoch": 14.75, + "learning_rate": 4.113232174906184e-05, + "loss": NaN, + "step": 90700 + }, + { + "epoch": 14.75, + "eval_loss": 0.46043136715888977, + "eval_runtime": 10.1706, + "eval_samples_per_second": 9.832, + "eval_wer": 0.364190012180268, + "step": 90700 + }, + { + "epoch": 14.77, + "learning_rate": 4.106705824767499e-05, + "loss": NaN, + "step": 90800 + }, + { + "epoch": 14.77, + "eval_loss": 0.37815314531326294, + "eval_runtime": 8.7882, + "eval_samples_per_second": 11.379, + "eval_wer": 0.3459196102314251, + "step": 90800 + }, + { + "epoch": 14.78, + "learning_rate": 4.1001794746288144e-05, + "loss": NaN, + "step": 90900 + }, + { + "epoch": 14.78, + "eval_loss": 0.4232279658317566, + "eval_runtime": 9.4671, + "eval_samples_per_second": 10.563, + "eval_wer": 0.3556638246041413, + "step": 90900 + }, + { + "epoch": 14.8, + "learning_rate": 4.093653124490129e-05, + "loss": NaN, + "step": 91000 + }, + { + "epoch": 14.8, + "eval_loss": 0.42844250798225403, + "eval_runtime": 8.6741, + "eval_samples_per_second": 11.529, + "eval_wer": 0.3654080389768575, + "step": 91000 + }, + { + "epoch": 14.82, + "learning_rate": 4.087126774351444e-05, + "loss": NaN, + "step": 91100 + }, + { + "epoch": 14.82, + "eval_loss": 0.3949730396270752, + "eval_runtime": 7.3535, + "eval_samples_per_second": 13.599, + "eval_wer": 0.34348355663824603, + "step": 91100 + }, + { + "epoch": 14.83, + "learning_rate": 4.080600424212759e-05, + "loss": NaN, + "step": 91200 + }, + { + "epoch": 14.83, + "eval_loss": 0.4596826136112213, + "eval_runtime": 7.979, + "eval_samples_per_second": 12.533, + "eval_wer": 0.34226552984165654, + "step": 91200 + }, + { + "epoch": 14.85, + "learning_rate": 4.074074074074074e-05, + "loss": NaN, + "step": 91300 + }, + { + "epoch": 14.85, + "eval_loss": 0.3969116508960724, + "eval_runtime": 7.3471, + "eval_samples_per_second": 13.611, + "eval_wer": 0.34348355663824603, + "step": 91300 + }, + { + "epoch": 14.86, + "learning_rate": 4.067547723935389e-05, + "loss": NaN, + "step": 91400 + }, + { + "epoch": 14.86, + "eval_loss": 0.42892399430274963, + "eval_runtime": 7.7448, + "eval_samples_per_second": 12.912, + "eval_wer": 0.33617539585870887, + "step": 91400 + }, + { + "epoch": 14.88, + "learning_rate": 4.061021373796705e-05, + "loss": NaN, + "step": 91500 + }, + { + "epoch": 14.88, + "eval_loss": 0.3805890679359436, + "eval_runtime": 7.337, + "eval_samples_per_second": 13.629, + "eval_wer": 0.3398294762484775, + "step": 91500 + }, + { + "epoch": 14.9, + "learning_rate": 4.0544950236580194e-05, + "loss": NaN, + "step": 91600 + }, + { + "epoch": 14.9, + "eval_loss": 0.34959712624549866, + "eval_runtime": 9.7552, + "eval_samples_per_second": 10.251, + "eval_wer": 0.30572472594397077, + "step": 91600 + }, + { + "epoch": 14.91, + "learning_rate": 4.0479686735193346e-05, + "loss": NaN, + "step": 91700 + }, + { + "epoch": 14.91, + "eval_loss": 0.3602063059806824, + "eval_runtime": 6.6405, + "eval_samples_per_second": 15.059, + "eval_wer": 0.3373934226552984, + "step": 91700 + }, + { + "epoch": 14.93, + "learning_rate": 4.04144232338065e-05, + "loss": NaN, + "step": 91800 + }, + { + "epoch": 14.93, + "eval_loss": 0.442261666059494, + "eval_runtime": 10.1904, + "eval_samples_per_second": 9.813, + "eval_wer": 0.34835566382460414, + "step": 91800 + }, + { + "epoch": 14.95, + "learning_rate": 4.034915973241964e-05, + "loss": NaN, + "step": 91900 + }, + { + "epoch": 14.95, + "eval_loss": 0.3955381512641907, + "eval_runtime": 9.4798, + "eval_samples_per_second": 10.549, + "eval_wer": 0.3398294762484775, + "step": 91900 + }, + { + "epoch": 14.96, + "learning_rate": 4.0283896231032795e-05, + "loss": NaN, + "step": 92000 + }, + { + "epoch": 14.96, + "eval_loss": 0.3891383409500122, + "eval_runtime": 8.2087, + "eval_samples_per_second": 12.182, + "eval_wer": 0.3215590742996346, + "step": 92000 + }, + { + "epoch": 14.98, + "learning_rate": 4.021863272964595e-05, + "loss": NaN, + "step": 92100 + }, + { + "epoch": 14.98, + "eval_loss": 0.38750743865966797, + "eval_runtime": 7.6982, + "eval_samples_per_second": 12.99, + "eval_wer": 0.3325213154689403, + "step": 92100 + }, + { + "epoch": 14.99, + "learning_rate": 4.01533692282591e-05, + "loss": NaN, + "step": 92200 + }, + { + "epoch": 14.99, + "eval_loss": 0.45837047696113586, + "eval_runtime": 7.4055, + "eval_samples_per_second": 13.504, + "eval_wer": 0.3507917174177832, + "step": 92200 + }, + { + "epoch": 15.01, + "learning_rate": 4.008810572687225e-05, + "loss": NaN, + "step": 92300 + }, + { + "epoch": 15.01, + "eval_loss": 0.4188990890979767, + "eval_runtime": 8.4196, + "eval_samples_per_second": 11.877, + "eval_wer": 0.37271619975639464, + "step": 92300 + }, + { + "epoch": 15.03, + "learning_rate": 4.00228422254854e-05, + "loss": NaN, + "step": 92400 + }, + { + "epoch": 15.03, + "eval_loss": 0.4422924816608429, + "eval_runtime": 8.3856, + "eval_samples_per_second": 11.925, + "eval_wer": 0.35322777101096225, + "step": 92400 + }, + { + "epoch": 15.04, + "learning_rate": 3.995757872409855e-05, + "loss": NaN, + "step": 92500 + }, + { + "epoch": 15.04, + "eval_loss": 0.414529412984848, + "eval_runtime": 8.9456, + "eval_samples_per_second": 11.179, + "eval_wer": 0.35931790499390986, + "step": 92500 + }, + { + "epoch": 15.06, + "learning_rate": 3.98923152227117e-05, + "loss": NaN, + "step": 92600 + }, + { + "epoch": 15.06, + "eval_loss": 0.3688591718673706, + "eval_runtime": 8.5338, + "eval_samples_per_second": 11.718, + "eval_wer": 0.33495736906211937, + "step": 92600 + }, + { + "epoch": 15.08, + "learning_rate": 3.982705172132485e-05, + "loss": NaN, + "step": 92700 + }, + { + "epoch": 15.08, + "eval_loss": 0.396541565656662, + "eval_runtime": 8.586, + "eval_samples_per_second": 11.647, + "eval_wer": 0.3337393422655298, + "step": 92700 + }, + { + "epoch": 15.09, + "learning_rate": 3.9761788219938e-05, + "loss": NaN, + "step": 92800 + }, + { + "epoch": 15.09, + "eval_loss": 0.4621961712837219, + "eval_runtime": 9.2469, + "eval_samples_per_second": 10.814, + "eval_wer": 0.3507917174177832, + "step": 92800 + }, + { + "epoch": 15.11, + "learning_rate": 3.9696524718551155e-05, + "loss": NaN, + "step": 92900 + }, + { + "epoch": 15.11, + "eval_loss": 0.39792075753211975, + "eval_runtime": 7.6153, + "eval_samples_per_second": 13.131, + "eval_wer": 0.3447015834348356, + "step": 92900 + }, + { + "epoch": 15.12, + "learning_rate": 3.96312612171643e-05, + "loss": NaN, + "step": 93000 + }, + { + "epoch": 15.12, + "eval_loss": 0.4314855635166168, + "eval_runtime": 8.7472, + "eval_samples_per_second": 11.432, + "eval_wer": 0.34835566382460414, + "step": 93000 + }, + { + "epoch": 15.14, + "learning_rate": 3.956599771577745e-05, + "loss": NaN, + "step": 93100 + }, + { + "epoch": 15.14, + "eval_loss": 0.4228772222995758, + "eval_runtime": 6.9716, + "eval_samples_per_second": 14.344, + "eval_wer": 0.34835566382460414, + "step": 93100 + }, + { + "epoch": 15.16, + "learning_rate": 3.9500734214390604e-05, + "loss": NaN, + "step": 93200 + }, + { + "epoch": 15.16, + "eval_loss": 0.4500848054885864, + "eval_runtime": 8.5782, + "eval_samples_per_second": 11.657, + "eval_wer": 0.3629719853836784, + "step": 93200 + }, + { + "epoch": 15.17, + "learning_rate": 3.943547071300375e-05, + "loss": NaN, + "step": 93300 + }, + { + "epoch": 15.17, + "eval_loss": 0.4819047451019287, + "eval_runtime": 7.8844, + "eval_samples_per_second": 12.683, + "eval_wer": 0.3313032886723508, + "step": 93300 + }, + { + "epoch": 15.19, + "learning_rate": 3.937020721161691e-05, + "loss": NaN, + "step": 93400 + }, + { + "epoch": 15.19, + "eval_loss": 0.40653935074806213, + "eval_runtime": 8.8297, + "eval_samples_per_second": 11.325, + "eval_wer": 0.3568818514007308, + "step": 93400 + }, + { + "epoch": 15.21, + "learning_rate": 3.930494371023006e-05, + "loss": NaN, + "step": 93500 + }, + { + "epoch": 15.21, + "eval_loss": 0.4287218451499939, + "eval_runtime": 8.8101, + "eval_samples_per_second": 11.351, + "eval_wer": 0.3568818514007308, + "step": 93500 + }, + { + "epoch": 15.22, + "learning_rate": 3.9239680208843205e-05, + "loss": NaN, + "step": 93600 + }, + { + "epoch": 15.22, + "eval_loss": 0.3936750888824463, + "eval_runtime": 8.9968, + "eval_samples_per_second": 11.115, + "eval_wer": 0.33495736906211937, + "step": 93600 + }, + { + "epoch": 15.24, + "learning_rate": 3.917441670745636e-05, + "loss": NaN, + "step": 93700 + }, + { + "epoch": 15.24, + "eval_loss": 0.4512450098991394, + "eval_runtime": 9.9939, + "eval_samples_per_second": 10.006, + "eval_wer": 0.35444579780755175, + "step": 93700 + }, + { + "epoch": 15.25, + "learning_rate": 3.910915320606951e-05, + "loss": NaN, + "step": 93800 + }, + { + "epoch": 15.25, + "eval_loss": 0.38099244236946106, + "eval_runtime": 8.1003, + "eval_samples_per_second": 12.345, + "eval_wer": 0.33008526187576126, + "step": 93800 + }, + { + "epoch": 15.27, + "learning_rate": 3.9043889704682654e-05, + "loss": NaN, + "step": 93900 + }, + { + "epoch": 15.27, + "eval_loss": 0.4231901466846466, + "eval_runtime": 8.4197, + "eval_samples_per_second": 11.877, + "eval_wer": 0.35809987819732036, + "step": 93900 + }, + { + "epoch": 15.29, + "learning_rate": 3.8978626203295806e-05, + "loss": NaN, + "step": 94000 + }, + { + "epoch": 15.29, + "eval_loss": 0.41978344321250916, + "eval_runtime": 9.9333, + "eval_samples_per_second": 10.067, + "eval_wer": 0.34348355663824603, + "step": 94000 + }, + { + "epoch": 15.3, + "learning_rate": 3.8913362701908965e-05, + "loss": NaN, + "step": 94100 + }, + { + "epoch": 15.3, + "eval_loss": 0.35810530185699463, + "eval_runtime": 8.2325, + "eval_samples_per_second": 12.147, + "eval_wer": 0.32886723507917176, + "step": 94100 + }, + { + "epoch": 15.32, + "learning_rate": 3.884809920052211e-05, + "loss": NaN, + "step": 94200 + }, + { + "epoch": 15.32, + "eval_loss": 0.40189048647880554, + "eval_runtime": 7.9965, + "eval_samples_per_second": 12.505, + "eval_wer": 0.3325213154689403, + "step": 94200 + }, + { + "epoch": 15.34, + "learning_rate": 3.878283569913526e-05, + "loss": NaN, + "step": 94300 + }, + { + "epoch": 15.34, + "eval_loss": 0.39003273844718933, + "eval_runtime": 9.1613, + "eval_samples_per_second": 10.916, + "eval_wer": 0.34835566382460414, + "step": 94300 + }, + { + "epoch": 15.35, + "learning_rate": 3.871757219774841e-05, + "loss": NaN, + "step": 94400 + }, + { + "epoch": 15.35, + "eval_loss": 0.37178894877433777, + "eval_runtime": 8.3496, + "eval_samples_per_second": 11.977, + "eval_wer": 0.3373934226552984, + "step": 94400 + }, + { + "epoch": 15.37, + "learning_rate": 3.865230869636156e-05, + "loss": NaN, + "step": 94500 + }, + { + "epoch": 15.37, + "eval_loss": 0.43365535140037537, + "eval_runtime": 7.7206, + "eval_samples_per_second": 12.952, + "eval_wer": 0.3337393422655298, + "step": 94500 + }, + { + "epoch": 15.38, + "learning_rate": 3.858704519497471e-05, + "loss": NaN, + "step": 94600 + }, + { + "epoch": 15.38, + "eval_loss": 0.4589609205722809, + "eval_runtime": 9.1951, + "eval_samples_per_second": 10.875, + "eval_wer": 0.3507917174177832, + "step": 94600 + }, + { + "epoch": 15.4, + "learning_rate": 3.852178169358786e-05, + "loss": NaN, + "step": 94700 + }, + { + "epoch": 15.4, + "eval_loss": 0.424941748380661, + "eval_runtime": 7.5287, + "eval_samples_per_second": 13.283, + "eval_wer": 0.34226552984165654, + "step": 94700 + }, + { + "epoch": 15.42, + "learning_rate": 3.8456518192201015e-05, + "loss": NaN, + "step": 94800 + }, + { + "epoch": 15.42, + "eval_loss": 0.3637240529060364, + "eval_runtime": 7.2763, + "eval_samples_per_second": 13.743, + "eval_wer": 0.33617539585870887, + "step": 94800 + }, + { + "epoch": 15.43, + "learning_rate": 3.8391254690814166e-05, + "loss": NaN, + "step": 94900 + }, + { + "epoch": 15.43, + "eval_loss": 0.4193127453327179, + "eval_runtime": 8.6177, + "eval_samples_per_second": 11.604, + "eval_wer": 0.34957369062119364, + "step": 94900 + }, + { + "epoch": 15.45, + "learning_rate": 3.832599118942731e-05, + "loss": NaN, + "step": 95000 + }, + { + "epoch": 15.45, + "eval_loss": 0.4117446839809418, + "eval_runtime": 8.8573, + "eval_samples_per_second": 11.29, + "eval_wer": 0.35322777101096225, + "step": 95000 + }, + { + "epoch": 15.47, + "learning_rate": 3.8260727688040464e-05, + "loss": NaN, + "step": 95100 + }, + { + "epoch": 15.47, + "eval_loss": 0.4040590226650238, + "eval_runtime": 8.3094, + "eval_samples_per_second": 12.035, + "eval_wer": 0.3325213154689403, + "step": 95100 + }, + { + "epoch": 15.48, + "learning_rate": 3.8195464186653615e-05, + "loss": NaN, + "step": 95200 + }, + { + "epoch": 15.48, + "eval_loss": 0.4011731445789337, + "eval_runtime": 8.1566, + "eval_samples_per_second": 12.26, + "eval_wer": 0.3447015834348356, + "step": 95200 + }, + { + "epoch": 15.5, + "learning_rate": 3.813020068526677e-05, + "loss": NaN, + "step": 95300 + }, + { + "epoch": 15.5, + "eval_loss": 0.4071398675441742, + "eval_runtime": 8.9786, + "eval_samples_per_second": 11.138, + "eval_wer": 0.3690621193666261, + "step": 95300 + }, + { + "epoch": 15.51, + "learning_rate": 3.806493718387992e-05, + "loss": NaN, + "step": 95400 + }, + { + "epoch": 15.51, + "eval_loss": 0.4124356806278229, + "eval_runtime": 7.7821, + "eval_samples_per_second": 12.85, + "eval_wer": 0.36053593179049936, + "step": 95400 + }, + { + "epoch": 15.53, + "learning_rate": 3.799967368249307e-05, + "loss": NaN, + "step": 95500 + }, + { + "epoch": 15.53, + "eval_loss": 0.43873709440231323, + "eval_runtime": 8.5817, + "eval_samples_per_second": 11.653, + "eval_wer": 0.35322777101096225, + "step": 95500 + }, + { + "epoch": 15.55, + "learning_rate": 3.7934410181106216e-05, + "loss": NaN, + "step": 95600 + }, + { + "epoch": 15.55, + "eval_loss": 0.43614262342453003, + "eval_runtime": 7.3882, + "eval_samples_per_second": 13.535, + "eval_wer": 0.35322777101096225, + "step": 95600 + }, + { + "epoch": 15.56, + "learning_rate": 3.786914667971937e-05, + "loss": NaN, + "step": 95700 + }, + { + "epoch": 15.56, + "eval_loss": 0.4307575225830078, + "eval_runtime": 7.3712, + "eval_samples_per_second": 13.566, + "eval_wer": 0.3617539585870889, + "step": 95700 + }, + { + "epoch": 15.58, + "learning_rate": 3.780388317833252e-05, + "loss": NaN, + "step": 95800 + }, + { + "epoch": 15.58, + "eval_loss": 0.39022716879844666, + "eval_runtime": 8.1214, + "eval_samples_per_second": 12.313, + "eval_wer": 0.34348355663824603, + "step": 95800 + }, + { + "epoch": 15.6, + "learning_rate": 3.773861967694567e-05, + "loss": NaN, + "step": 95900 + }, + { + "epoch": 15.6, + "eval_loss": 0.5400065779685974, + "eval_runtime": 9.4699, + "eval_samples_per_second": 10.56, + "eval_wer": 0.34226552984165654, + "step": 95900 + }, + { + "epoch": 15.61, + "learning_rate": 3.7673356175558824e-05, + "loss": NaN, + "step": 96000 + }, + { + "epoch": 15.61, + "eval_loss": 0.4195316433906555, + "eval_runtime": 6.7762, + "eval_samples_per_second": 14.758, + "eval_wer": 0.3471376370280146, + "step": 96000 + }, + { + "epoch": 15.63, + "learning_rate": 3.760809267417197e-05, + "loss": NaN, + "step": 96100 + }, + { + "epoch": 15.63, + "eval_loss": 0.5240188837051392, + "eval_runtime": 7.6344, + "eval_samples_per_second": 13.099, + "eval_wer": 0.35931790499390986, + "step": 96100 + }, + { + "epoch": 15.64, + "learning_rate": 3.754282917278512e-05, + "loss": NaN, + "step": 96200 + }, + { + "epoch": 15.64, + "eval_loss": 0.4596652686595917, + "eval_runtime": 8.7291, + "eval_samples_per_second": 11.456, + "eval_wer": 0.3471376370280146, + "step": 96200 + }, + { + "epoch": 15.66, + "learning_rate": 3.747756567139827e-05, + "loss": NaN, + "step": 96300 + }, + { + "epoch": 15.66, + "eval_loss": 0.5023432374000549, + "eval_runtime": 10.5167, + "eval_samples_per_second": 9.509, + "eval_wer": 0.3459196102314251, + "step": 96300 + }, + { + "epoch": 15.68, + "learning_rate": 3.741230217001142e-05, + "loss": NaN, + "step": 96400 + }, + { + "epoch": 15.68, + "eval_loss": 0.42950576543807983, + "eval_runtime": 9.5929, + "eval_samples_per_second": 10.424, + "eval_wer": 0.3617539585870889, + "step": 96400 + }, + { + "epoch": 15.69, + "learning_rate": 3.734703866862457e-05, + "loss": NaN, + "step": 96500 + }, + { + "epoch": 15.69, + "eval_loss": 0.6256393194198608, + "eval_runtime": 8.8475, + "eval_samples_per_second": 11.303, + "eval_wer": 0.3617539585870889, + "step": 96500 + }, + { + "epoch": 15.71, + "learning_rate": 3.728177516723773e-05, + "loss": NaN, + "step": 96600 + }, + { + "epoch": 15.71, + "eval_loss": 0.40215057134628296, + "eval_runtime": 7.4719, + "eval_samples_per_second": 13.383, + "eval_wer": 0.34835566382460414, + "step": 96600 + }, + { + "epoch": 15.73, + "learning_rate": 3.7216511665850874e-05, + "loss": NaN, + "step": 96700 + }, + { + "epoch": 15.73, + "eval_loss": 0.5013496279716492, + "eval_runtime": 8.1012, + "eval_samples_per_second": 12.344, + "eval_wer": 0.3520097442143727, + "step": 96700 + }, + { + "epoch": 15.74, + "learning_rate": 3.7151248164464026e-05, + "loss": NaN, + "step": 96800 + }, + { + "epoch": 15.74, + "eval_loss": 0.4012104868888855, + "eval_runtime": 6.7403, + "eval_samples_per_second": 14.836, + "eval_wer": 0.3459196102314251, + "step": 96800 + }, + { + "epoch": 15.76, + "learning_rate": 3.708598466307718e-05, + "loss": NaN, + "step": 96900 + }, + { + "epoch": 15.76, + "eval_loss": 0.39129239320755005, + "eval_runtime": 7.4628, + "eval_samples_per_second": 13.4, + "eval_wer": 0.35322777101096225, + "step": 96900 + }, + { + "epoch": 15.77, + "learning_rate": 3.702072116169032e-05, + "loss": NaN, + "step": 97000 + }, + { + "epoch": 15.77, + "eval_loss": 0.4360794126987457, + "eval_runtime": 7.5057, + "eval_samples_per_second": 13.323, + "eval_wer": 0.35444579780755175, + "step": 97000 + }, + { + "epoch": 15.79, + "learning_rate": 3.6955457660303475e-05, + "loss": NaN, + "step": 97100 + }, + { + "epoch": 15.79, + "eval_loss": 0.42406460642814636, + "eval_runtime": 7.6982, + "eval_samples_per_second": 12.99, + "eval_wer": 0.3398294762484775, + "step": 97100 + }, + { + "epoch": 15.81, + "learning_rate": 3.689019415891663e-05, + "loss": NaN, + "step": 97200 + }, + { + "epoch": 15.81, + "eval_loss": 0.4149155020713806, + "eval_runtime": 6.9699, + "eval_samples_per_second": 14.347, + "eval_wer": 0.3556638246041413, + "step": 97200 + }, + { + "epoch": 15.82, + "learning_rate": 3.682493065752978e-05, + "loss": NaN, + "step": 97300 + }, + { + "epoch": 15.82, + "eval_loss": 0.4104059934616089, + "eval_runtime": 7.1554, + "eval_samples_per_second": 13.976, + "eval_wer": 0.34226552984165654, + "step": 97300 + }, + { + "epoch": 15.84, + "learning_rate": 3.675966715614293e-05, + "loss": NaN, + "step": 97400 + }, + { + "epoch": 15.84, + "eval_loss": 0.3772718906402588, + "eval_runtime": 7.8874, + "eval_samples_per_second": 12.678, + "eval_wer": 0.3447015834348356, + "step": 97400 + }, + { + "epoch": 15.86, + "learning_rate": 3.669440365475608e-05, + "loss": NaN, + "step": 97500 + }, + { + "epoch": 15.86, + "eval_loss": 0.37940847873687744, + "eval_runtime": 6.9895, + "eval_samples_per_second": 14.307, + "eval_wer": 0.3373934226552984, + "step": 97500 + }, + { + "epoch": 15.87, + "learning_rate": 3.662914015336923e-05, + "loss": NaN, + "step": 97600 + }, + { + "epoch": 15.87, + "eval_loss": 0.41447925567626953, + "eval_runtime": 7.7777, + "eval_samples_per_second": 12.857, + "eval_wer": 0.36053593179049936, + "step": 97600 + }, + { + "epoch": 15.89, + "learning_rate": 3.656387665198238e-05, + "loss": NaN, + "step": 97700 + }, + { + "epoch": 15.89, + "eval_loss": 0.41132599115371704, + "eval_runtime": 8.7663, + "eval_samples_per_second": 11.407, + "eval_wer": 0.37149817295980514, + "step": 97700 + }, + { + "epoch": 15.91, + "learning_rate": 3.649861315059553e-05, + "loss": NaN, + "step": 97800 + }, + { + "epoch": 15.91, + "eval_loss": 0.5543003678321838, + "eval_runtime": 9.1096, + "eval_samples_per_second": 10.977, + "eval_wer": 0.35444579780755175, + "step": 97800 + }, + { + "epoch": 15.92, + "learning_rate": 3.643334964920868e-05, + "loss": NaN, + "step": 97900 + }, + { + "epoch": 15.92, + "eval_loss": 0.4039764106273651, + "eval_runtime": 8.5411, + "eval_samples_per_second": 11.708, + "eval_wer": 0.34226552984165654, + "step": 97900 + }, + { + "epoch": 15.94, + "learning_rate": 3.6368086147821835e-05, + "loss": NaN, + "step": 98000 + }, + { + "epoch": 15.94, + "eval_loss": 0.39489075541496277, + "eval_runtime": 7.4333, + "eval_samples_per_second": 13.453, + "eval_wer": 0.3459196102314251, + "step": 98000 + }, + { + "epoch": 15.95, + "learning_rate": 3.630282264643498e-05, + "loss": NaN, + "step": 98100 + }, + { + "epoch": 15.95, + "eval_loss": 0.44759106636047363, + "eval_runtime": 7.8336, + "eval_samples_per_second": 12.766, + "eval_wer": 0.35931790499390986, + "step": 98100 + }, + { + "epoch": 15.97, + "learning_rate": 3.623755914504813e-05, + "loss": NaN, + "step": 98200 + }, + { + "epoch": 15.97, + "eval_loss": 0.42032337188720703, + "eval_runtime": 9.2764, + "eval_samples_per_second": 10.78, + "eval_wer": 0.36053593179049936, + "step": 98200 + }, + { + "epoch": 15.99, + "learning_rate": 3.6172295643661284e-05, + "loss": NaN, + "step": 98300 + }, + { + "epoch": 15.99, + "eval_loss": 0.3860644996166229, + "eval_runtime": 7.7163, + "eval_samples_per_second": 12.96, + "eval_wer": 0.34226552984165654, + "step": 98300 + }, + { + "epoch": 16.0, + "learning_rate": 3.6107032142274436e-05, + "loss": NaN, + "step": 98400 + }, + { + "epoch": 16.0, + "eval_loss": 0.5123631954193115, + "eval_runtime": 8.6602, + "eval_samples_per_second": 11.547, + "eval_wer": 0.3459196102314251, + "step": 98400 + }, + { + "epoch": 16.02, + "learning_rate": 3.604176864088759e-05, + "loss": NaN, + "step": 98500 + }, + { + "epoch": 16.02, + "eval_loss": 0.39611780643463135, + "eval_runtime": 7.1716, + "eval_samples_per_second": 13.944, + "eval_wer": 0.3459196102314251, + "step": 98500 + }, + { + "epoch": 16.04, + "learning_rate": 3.597650513950074e-05, + "loss": NaN, + "step": 98600 + }, + { + "epoch": 16.04, + "eval_loss": 0.47687721252441406, + "eval_runtime": 7.4875, + "eval_samples_per_second": 13.356, + "eval_wer": 0.3556638246041413, + "step": 98600 + }, + { + "epoch": 16.05, + "learning_rate": 3.5911241638113885e-05, + "loss": NaN, + "step": 98700 + }, + { + "epoch": 16.05, + "eval_loss": 0.5570668578147888, + "eval_runtime": 9.3365, + "eval_samples_per_second": 10.711, + "eval_wer": 0.35931790499390986, + "step": 98700 + }, + { + "epoch": 16.07, + "learning_rate": 3.584597813672704e-05, + "loss": NaN, + "step": 98800 + }, + { + "epoch": 16.07, + "eval_loss": 0.4510762691497803, + "eval_runtime": 8.9657, + "eval_samples_per_second": 11.154, + "eval_wer": 0.3629719853836784, + "step": 98800 + }, + { + "epoch": 16.08, + "learning_rate": 3.578071463534019e-05, + "loss": NaN, + "step": 98900 + }, + { + "epoch": 16.08, + "eval_loss": 0.41739127039909363, + "eval_runtime": 8.5583, + "eval_samples_per_second": 11.685, + "eval_wer": 0.3459196102314251, + "step": 98900 + }, + { + "epoch": 16.1, + "learning_rate": 3.5715451133953334e-05, + "loss": NaN, + "step": 99000 + }, + { + "epoch": 16.1, + "eval_loss": 0.36998283863067627, + "eval_runtime": 9.3054, + "eval_samples_per_second": 10.747, + "eval_wer": 0.33008526187576126, + "step": 99000 + }, + { + "epoch": 16.12, + "learning_rate": 3.565018763256649e-05, + "loss": NaN, + "step": 99100 + }, + { + "epoch": 16.12, + "eval_loss": 0.4438319504261017, + "eval_runtime": 7.1139, + "eval_samples_per_second": 14.057, + "eval_wer": 0.35931790499390986, + "step": 99100 + }, + { + "epoch": 16.13, + "learning_rate": 3.5584924131179645e-05, + "loss": NaN, + "step": 99200 + }, + { + "epoch": 16.13, + "eval_loss": 0.42432862520217896, + "eval_runtime": 7.9048, + "eval_samples_per_second": 12.651, + "eval_wer": 0.30450669914738127, + "step": 99200 + }, + { + "epoch": 16.15, + "learning_rate": 3.551966062979279e-05, + "loss": NaN, + "step": 99300 + }, + { + "epoch": 16.15, + "eval_loss": 0.39224445819854736, + "eval_runtime": 7.1957, + "eval_samples_per_second": 13.897, + "eval_wer": 0.3447015834348356, + "step": 99300 + }, + { + "epoch": 16.17, + "learning_rate": 3.545439712840594e-05, + "loss": NaN, + "step": 99400 + }, + { + "epoch": 16.17, + "eval_loss": 0.44131773710250854, + "eval_runtime": 8.1898, + "eval_samples_per_second": 12.21, + "eval_wer": 0.3507917174177832, + "step": 99400 + }, + { + "epoch": 16.18, + "learning_rate": 3.538913362701909e-05, + "loss": NaN, + "step": 99500 + }, + { + "epoch": 16.18, + "eval_loss": 0.38120517134666443, + "eval_runtime": 6.6009, + "eval_samples_per_second": 15.149, + "eval_wer": 0.3398294762484775, + "step": 99500 + }, + { + "epoch": 16.2, + "learning_rate": 3.532387012563224e-05, + "loss": NaN, + "step": 99600 + }, + { + "epoch": 16.2, + "eval_loss": 0.4235430955886841, + "eval_runtime": 9.0979, + "eval_samples_per_second": 10.991, + "eval_wer": 0.3447015834348356, + "step": 99600 + }, + { + "epoch": 16.21, + "learning_rate": 3.525860662424539e-05, + "loss": NaN, + "step": 99700 + }, + { + "epoch": 16.21, + "eval_loss": 0.39221322536468506, + "eval_runtime": 7.4156, + "eval_samples_per_second": 13.485, + "eval_wer": 0.3507917174177832, + "step": 99700 + }, + { + "epoch": 16.23, + "learning_rate": 3.519334312285854e-05, + "loss": NaN, + "step": 99800 + }, + { + "epoch": 16.23, + "eval_loss": 0.49404698610305786, + "eval_runtime": 7.4446, + "eval_samples_per_second": 13.433, + "eval_wer": 0.3556638246041413, + "step": 99800 + }, + { + "epoch": 16.25, + "learning_rate": 3.5128079621471695e-05, + "loss": NaN, + "step": 99900 + }, + { + "epoch": 16.25, + "eval_loss": 0.4537053406238556, + "eval_runtime": 9.1442, + "eval_samples_per_second": 10.936, + "eval_wer": 0.3471376370280146, + "step": 99900 + }, + { + "epoch": 16.26, + "learning_rate": 3.5062816120084846e-05, + "loss": NaN, + "step": 100000 + }, + { + "epoch": 16.26, + "eval_loss": 0.42373108863830566, + "eval_runtime": 8.1063, + "eval_samples_per_second": 12.336, + "eval_wer": 0.33617539585870887, + "step": 100000 + }, + { + "epoch": 16.28, + "learning_rate": 3.499755261869799e-05, + "loss": NaN, + "step": 100100 + }, + { + "epoch": 16.28, + "eval_loss": 0.5141391754150391, + "eval_runtime": 9.0075, + "eval_samples_per_second": 11.102, + "eval_wer": 0.3459196102314251, + "step": 100100 + }, + { + "epoch": 16.3, + "learning_rate": 3.4932289117311143e-05, + "loss": NaN, + "step": 100200 + }, + { + "epoch": 16.3, + "eval_loss": 0.4039132297039032, + "eval_runtime": 9.3683, + "eval_samples_per_second": 10.674, + "eval_wer": 0.3459196102314251, + "step": 100200 + }, + { + "epoch": 16.31, + "learning_rate": 3.4867025615924295e-05, + "loss": NaN, + "step": 100300 + }, + { + "epoch": 16.31, + "eval_loss": 0.46978646516799927, + "eval_runtime": 6.7436, + "eval_samples_per_second": 14.829, + "eval_wer": 0.3471376370280146, + "step": 100300 + }, + { + "epoch": 16.33, + "learning_rate": 3.480176211453745e-05, + "loss": NaN, + "step": 100400 + }, + { + "epoch": 16.33, + "eval_loss": 0.41287657618522644, + "eval_runtime": 8.0877, + "eval_samples_per_second": 12.364, + "eval_wer": 0.3471376370280146, + "step": 100400 + }, + { + "epoch": 16.34, + "learning_rate": 3.47364986131506e-05, + "loss": NaN, + "step": 100500 + }, + { + "epoch": 16.34, + "eval_loss": 0.4684438407421112, + "eval_runtime": 8.5044, + "eval_samples_per_second": 11.759, + "eval_wer": 0.3520097442143727, + "step": 100500 + }, + { + "epoch": 16.36, + "learning_rate": 3.467123511176375e-05, + "loss": NaN, + "step": 100600 + }, + { + "epoch": 16.36, + "eval_loss": 0.4466744363307953, + "eval_runtime": 8.8741, + "eval_samples_per_second": 11.269, + "eval_wer": 0.35809987819732036, + "step": 100600 + }, + { + "epoch": 16.38, + "learning_rate": 3.4605971610376896e-05, + "loss": NaN, + "step": 100700 + }, + { + "epoch": 16.38, + "eval_loss": 0.38788944482803345, + "eval_runtime": 7.8104, + "eval_samples_per_second": 12.803, + "eval_wer": 0.33008526187576126, + "step": 100700 + }, + { + "epoch": 16.39, + "learning_rate": 3.454070810899005e-05, + "loss": NaN, + "step": 100800 + }, + { + "epoch": 16.39, + "eval_loss": 0.40568241477012634, + "eval_runtime": 9.8132, + "eval_samples_per_second": 10.19, + "eval_wer": 0.341047503045067, + "step": 100800 + }, + { + "epoch": 16.41, + "learning_rate": 3.44754446076032e-05, + "loss": NaN, + "step": 100900 + }, + { + "epoch": 16.41, + "eval_loss": 0.43558773398399353, + "eval_runtime": 7.63, + "eval_samples_per_second": 13.106, + "eval_wer": 0.3568818514007308, + "step": 100900 + }, + { + "epoch": 16.43, + "learning_rate": 3.441018110621635e-05, + "loss": NaN, + "step": 101000 + }, + { + "epoch": 16.43, + "eval_loss": 0.4001496136188507, + "eval_runtime": 6.8773, + "eval_samples_per_second": 14.541, + "eval_wer": 0.3471376370280146, + "step": 101000 + }, + { + "epoch": 16.44, + "learning_rate": 3.4344917604829504e-05, + "loss": NaN, + "step": 101100 + }, + { + "epoch": 16.44, + "eval_loss": 0.4105696976184845, + "eval_runtime": 7.7809, + "eval_samples_per_second": 12.852, + "eval_wer": 0.3325213154689403, + "step": 101100 + }, + { + "epoch": 16.46, + "learning_rate": 3.427965410344265e-05, + "loss": NaN, + "step": 101200 + }, + { + "epoch": 16.46, + "eval_loss": 0.4008920192718506, + "eval_runtime": 7.3769, + "eval_samples_per_second": 13.556, + "eval_wer": 0.34348355663824603, + "step": 101200 + }, + { + "epoch": 16.47, + "learning_rate": 3.42143906020558e-05, + "loss": NaN, + "step": 101300 + }, + { + "epoch": 16.47, + "eval_loss": 0.4045610725879669, + "eval_runtime": 8.9725, + "eval_samples_per_second": 11.145, + "eval_wer": 0.341047503045067, + "step": 101300 + }, + { + "epoch": 16.49, + "learning_rate": 3.414912710066895e-05, + "loss": NaN, + "step": 101400 + }, + { + "epoch": 16.49, + "eval_loss": 0.43567919731140137, + "eval_runtime": 8.8256, + "eval_samples_per_second": 11.331, + "eval_wer": 0.36053593179049936, + "step": 101400 + }, + { + "epoch": 16.51, + "learning_rate": 3.40838635992821e-05, + "loss": NaN, + "step": 101500 + }, + { + "epoch": 16.51, + "eval_loss": 0.41516247391700745, + "eval_runtime": 7.2706, + "eval_samples_per_second": 13.754, + "eval_wer": 0.35444579780755175, + "step": 101500 + }, + { + "epoch": 16.52, + "learning_rate": 3.401860009789526e-05, + "loss": NaN, + "step": 101600 + }, + { + "epoch": 16.52, + "eval_loss": 0.3579968512058258, + "eval_runtime": 7.7267, + "eval_samples_per_second": 12.942, + "eval_wer": 0.34226552984165654, + "step": 101600 + }, + { + "epoch": 16.54, + "learning_rate": 3.395333659650841e-05, + "loss": NaN, + "step": 101700 + }, + { + "epoch": 16.54, + "eval_loss": 0.3978956639766693, + "eval_runtime": 7.5352, + "eval_samples_per_second": 13.271, + "eval_wer": 0.3373934226552984, + "step": 101700 + }, + { + "epoch": 16.56, + "learning_rate": 3.3888073095121554e-05, + "loss": NaN, + "step": 101800 + }, + { + "epoch": 16.56, + "eval_loss": 0.4536603093147278, + "eval_runtime": 7.7523, + "eval_samples_per_second": 12.899, + "eval_wer": 0.3105968331303289, + "step": 101800 + }, + { + "epoch": 16.57, + "learning_rate": 3.3822809593734706e-05, + "loss": NaN, + "step": 101900 + }, + { + "epoch": 16.57, + "eval_loss": 0.44562357664108276, + "eval_runtime": 7.9874, + "eval_samples_per_second": 12.52, + "eval_wer": 0.3447015834348356, + "step": 101900 + }, + { + "epoch": 16.59, + "learning_rate": 3.375754609234786e-05, + "loss": NaN, + "step": 102000 + }, + { + "epoch": 16.59, + "eval_loss": 0.4070366322994232, + "eval_runtime": 7.5818, + "eval_samples_per_second": 13.189, + "eval_wer": 0.33495736906211937, + "step": 102000 + }, + { + "epoch": 16.6, + "learning_rate": 3.3692282590961e-05, + "loss": NaN, + "step": 102100 + }, + { + "epoch": 16.6, + "eval_loss": 0.4052729904651642, + "eval_runtime": 9.8038, + "eval_samples_per_second": 10.2, + "eval_wer": 0.3398294762484775, + "step": 102100 + }, + { + "epoch": 16.62, + "learning_rate": 3.3627019089574155e-05, + "loss": NaN, + "step": 102200 + }, + { + "epoch": 16.62, + "eval_loss": 0.43835657835006714, + "eval_runtime": 9.1381, + "eval_samples_per_second": 10.943, + "eval_wer": 0.3276492082825822, + "step": 102200 + }, + { + "epoch": 16.64, + "learning_rate": 3.3561755588187313e-05, + "loss": NaN, + "step": 102300 + }, + { + "epoch": 16.64, + "eval_loss": 0.3671765923500061, + "eval_runtime": 8.8073, + "eval_samples_per_second": 11.354, + "eval_wer": 0.33008526187576126, + "step": 102300 + }, + { + "epoch": 16.65, + "learning_rate": 3.349649208680046e-05, + "loss": NaN, + "step": 102400 + }, + { + "epoch": 16.65, + "eval_loss": 0.3550013303756714, + "eval_runtime": 8.81, + "eval_samples_per_second": 11.351, + "eval_wer": 0.3276492082825822, + "step": 102400 + }, + { + "epoch": 16.67, + "learning_rate": 3.343122858541361e-05, + "loss": NaN, + "step": 102500 + }, + { + "epoch": 16.67, + "eval_loss": 0.37561336159706116, + "eval_runtime": 8.2237, + "eval_samples_per_second": 12.16, + "eval_wer": 0.341047503045067, + "step": 102500 + }, + { + "epoch": 16.69, + "learning_rate": 3.336596508402676e-05, + "loss": NaN, + "step": 102600 + }, + { + "epoch": 16.69, + "eval_loss": 0.42471256852149963, + "eval_runtime": 10.0, + "eval_samples_per_second": 10.0, + "eval_wer": 0.341047503045067, + "step": 102600 + }, + { + "epoch": 16.7, + "learning_rate": 3.330070158263991e-05, + "loss": NaN, + "step": 102700 + }, + { + "epoch": 16.7, + "eval_loss": 0.43302589654922485, + "eval_runtime": 8.4874, + "eval_samples_per_second": 11.782, + "eval_wer": 0.3337393422655298, + "step": 102700 + }, + { + "epoch": 16.72, + "learning_rate": 3.323543808125306e-05, + "loss": NaN, + "step": 102800 + }, + { + "epoch": 16.72, + "eval_loss": 0.3747265636920929, + "eval_runtime": 8.2183, + "eval_samples_per_second": 12.168, + "eval_wer": 0.3398294762484775, + "step": 102800 + }, + { + "epoch": 16.73, + "learning_rate": 3.317017457986621e-05, + "loss": NaN, + "step": 102900 + }, + { + "epoch": 16.73, + "eval_loss": 0.44977033138275146, + "eval_runtime": 9.9224, + "eval_samples_per_second": 10.078, + "eval_wer": 0.33617539585870887, + "step": 102900 + }, + { + "epoch": 16.75, + "learning_rate": 3.310491107847936e-05, + "loss": NaN, + "step": 103000 + }, + { + "epoch": 16.75, + "eval_loss": 0.3952389061450958, + "eval_runtime": 7.7792, + "eval_samples_per_second": 12.855, + "eval_wer": 0.35322777101096225, + "step": 103000 + }, + { + "epoch": 16.77, + "learning_rate": 3.3039647577092515e-05, + "loss": NaN, + "step": 103100 + }, + { + "epoch": 16.77, + "eval_loss": 0.40458202362060547, + "eval_runtime": 8.518, + "eval_samples_per_second": 11.74, + "eval_wer": 0.3386114494518879, + "step": 103100 + }, + { + "epoch": 16.78, + "learning_rate": 3.297438407570566e-05, + "loss": NaN, + "step": 103200 + }, + { + "epoch": 16.78, + "eval_loss": 0.4162432849407196, + "eval_runtime": 9.1287, + "eval_samples_per_second": 10.954, + "eval_wer": 0.33617539585870887, + "step": 103200 + }, + { + "epoch": 16.8, + "learning_rate": 3.290912057431881e-05, + "loss": NaN, + "step": 103300 + }, + { + "epoch": 16.8, + "eval_loss": 0.4272727966308594, + "eval_runtime": 10.4704, + "eval_samples_per_second": 9.551, + "eval_wer": 0.3471376370280146, + "step": 103300 + }, + { + "epoch": 16.82, + "learning_rate": 3.2843857072931964e-05, + "loss": NaN, + "step": 103400 + }, + { + "epoch": 16.82, + "eval_loss": 0.40407755970954895, + "eval_runtime": 9.8048, + "eval_samples_per_second": 10.199, + "eval_wer": 0.3471376370280146, + "step": 103400 + }, + { + "epoch": 16.83, + "learning_rate": 3.2778593571545116e-05, + "loss": NaN, + "step": 103500 + }, + { + "epoch": 16.83, + "eval_loss": 0.36328965425491333, + "eval_runtime": 7.0675, + "eval_samples_per_second": 14.149, + "eval_wer": 0.3337393422655298, + "step": 103500 + }, + { + "epoch": 16.85, + "learning_rate": 3.271333007015827e-05, + "loss": NaN, + "step": 103600 + }, + { + "epoch": 16.85, + "eval_loss": 0.3919774293899536, + "eval_runtime": 8.2233, + "eval_samples_per_second": 12.161, + "eval_wer": 0.3227771010962241, + "step": 103600 + }, + { + "epoch": 16.86, + "learning_rate": 3.264806656877142e-05, + "loss": NaN, + "step": 103700 + }, + { + "epoch": 16.86, + "eval_loss": 0.41714799404144287, + "eval_runtime": 7.8888, + "eval_samples_per_second": 12.676, + "eval_wer": 0.3459196102314251, + "step": 103700 + }, + { + "epoch": 16.88, + "learning_rate": 3.2582803067384565e-05, + "loss": NaN, + "step": 103800 + }, + { + "epoch": 16.88, + "eval_loss": 0.3970955014228821, + "eval_runtime": 8.7637, + "eval_samples_per_second": 11.411, + "eval_wer": 0.34348355663824603, + "step": 103800 + }, + { + "epoch": 16.9, + "learning_rate": 3.251753956599772e-05, + "loss": NaN, + "step": 103900 + }, + { + "epoch": 16.9, + "eval_loss": 0.39458972215652466, + "eval_runtime": 7.5247, + "eval_samples_per_second": 13.29, + "eval_wer": 0.3313032886723508, + "step": 103900 + }, + { + "epoch": 16.91, + "learning_rate": 3.245227606461087e-05, + "loss": NaN, + "step": 104000 + }, + { + "epoch": 16.91, + "eval_loss": 0.3992546498775482, + "eval_runtime": 7.666, + "eval_samples_per_second": 13.045, + "eval_wer": 0.34348355663824603, + "step": 104000 + }, + { + "epoch": 16.93, + "learning_rate": 3.2387012563224014e-05, + "loss": NaN, + "step": 104100 + }, + { + "epoch": 16.93, + "eval_loss": 0.3565240502357483, + "eval_runtime": 6.8455, + "eval_samples_per_second": 14.608, + "eval_wer": 0.3313032886723508, + "step": 104100 + }, + { + "epoch": 16.95, + "learning_rate": 3.232174906183717e-05, + "loss": NaN, + "step": 104200 + }, + { + "epoch": 16.95, + "eval_loss": 0.39272624254226685, + "eval_runtime": 8.7645, + "eval_samples_per_second": 11.41, + "eval_wer": 0.3398294762484775, + "step": 104200 + }, + { + "epoch": 16.96, + "learning_rate": 3.2256485560450325e-05, + "loss": NaN, + "step": 104300 + }, + { + "epoch": 16.96, + "eval_loss": 0.377750962972641, + "eval_runtime": 7.4931, + "eval_samples_per_second": 13.346, + "eval_wer": 0.32886723507917176, + "step": 104300 + }, + { + "epoch": 16.98, + "learning_rate": 3.219122205906347e-05, + "loss": NaN, + "step": 104400 + }, + { + "epoch": 16.98, + "eval_loss": 0.3803618550300598, + "eval_runtime": 7.7196, + "eval_samples_per_second": 12.954, + "eval_wer": 0.32521315468940315, + "step": 104400 + }, + { + "epoch": 16.99, + "learning_rate": 3.212595855767662e-05, + "loss": NaN, + "step": 104500 + }, + { + "epoch": 16.99, + "eval_loss": 0.4160638153553009, + "eval_runtime": 7.8399, + "eval_samples_per_second": 12.755, + "eval_wer": 0.35931790499390986, + "step": 104500 + }, + { + "epoch": 17.01, + "learning_rate": 3.206069505628977e-05, + "loss": NaN, + "step": 104600 + }, + { + "epoch": 17.01, + "eval_loss": 0.3985629677772522, + "eval_runtime": 7.8601, + "eval_samples_per_second": 12.723, + "eval_wer": 0.36662606577344703, + "step": 104600 + }, + { + "epoch": 17.03, + "learning_rate": 3.199543155490292e-05, + "loss": NaN, + "step": 104700 + }, + { + "epoch": 17.03, + "eval_loss": 0.362797349691391, + "eval_runtime": 7.5799, + "eval_samples_per_second": 13.193, + "eval_wer": 0.33495736906211937, + "step": 104700 + }, + { + "epoch": 17.04, + "learning_rate": 3.193016805351608e-05, + "loss": NaN, + "step": 104800 + }, + { + "epoch": 17.04, + "eval_loss": 0.46020495891571045, + "eval_runtime": 10.2517, + "eval_samples_per_second": 9.754, + "eval_wer": 0.35322777101096225, + "step": 104800 + }, + { + "epoch": 17.06, + "learning_rate": 3.186490455212922e-05, + "loss": NaN, + "step": 104900 + }, + { + "epoch": 17.06, + "eval_loss": 0.4332337975502014, + "eval_runtime": 8.5335, + "eval_samples_per_second": 11.719, + "eval_wer": 0.33617539585870887, + "step": 104900 + }, + { + "epoch": 17.08, + "learning_rate": 3.1799641050742374e-05, + "loss": NaN, + "step": 105000 + }, + { + "epoch": 17.08, + "eval_loss": 0.39707785844802856, + "eval_runtime": 9.4106, + "eval_samples_per_second": 10.626, + "eval_wer": 0.34348355663824603, + "step": 105000 + }, + { + "epoch": 17.09, + "learning_rate": 3.1734377549355526e-05, + "loss": NaN, + "step": 105100 + }, + { + "epoch": 17.09, + "eval_loss": 0.4423673152923584, + "eval_runtime": 7.0369, + "eval_samples_per_second": 14.211, + "eval_wer": 0.3373934226552984, + "step": 105100 + }, + { + "epoch": 17.11, + "learning_rate": 3.166911404796867e-05, + "loss": NaN, + "step": 105200 + }, + { + "epoch": 17.11, + "eval_loss": 0.4622967541217804, + "eval_runtime": 8.2948, + "eval_samples_per_second": 12.056, + "eval_wer": 0.3654080389768575, + "step": 105200 + }, + { + "epoch": 17.12, + "learning_rate": 3.1603850546581823e-05, + "loss": NaN, + "step": 105300 + }, + { + "epoch": 17.12, + "eval_loss": 0.47677239775657654, + "eval_runtime": 7.7449, + "eval_samples_per_second": 12.912, + "eval_wer": 0.34348355663824603, + "step": 105300 + }, + { + "epoch": 17.14, + "learning_rate": 3.1538587045194975e-05, + "loss": NaN, + "step": 105400 + }, + { + "epoch": 17.14, + "eval_loss": 0.3996973931789398, + "eval_runtime": 8.3747, + "eval_samples_per_second": 11.941, + "eval_wer": 0.3471376370280146, + "step": 105400 + }, + { + "epoch": 17.16, + "learning_rate": 3.147332354380813e-05, + "loss": NaN, + "step": 105500 + }, + { + "epoch": 17.16, + "eval_loss": 0.45852556824684143, + "eval_runtime": 7.9151, + "eval_samples_per_second": 12.634, + "eval_wer": 0.34957369062119364, + "step": 105500 + }, + { + "epoch": 17.17, + "learning_rate": 3.140806004242128e-05, + "loss": NaN, + "step": 105600 + }, + { + "epoch": 17.17, + "eval_loss": 0.3811168670654297, + "eval_runtime": 7.7269, + "eval_samples_per_second": 12.942, + "eval_wer": 0.32886723507917176, + "step": 105600 + }, + { + "epoch": 17.19, + "learning_rate": 3.134279654103443e-05, + "loss": NaN, + "step": 105700 + }, + { + "epoch": 17.19, + "eval_loss": 0.3853236436843872, + "eval_runtime": 8.0078, + "eval_samples_per_second": 12.488, + "eval_wer": 0.32521315468940315, + "step": 105700 + }, + { + "epoch": 17.21, + "learning_rate": 3.1277533039647576e-05, + "loss": NaN, + "step": 105800 + }, + { + "epoch": 17.21, + "eval_loss": 0.43565693497657776, + "eval_runtime": 8.6327, + "eval_samples_per_second": 11.584, + "eval_wer": 0.3690621193666261, + "step": 105800 + }, + { + "epoch": 17.22, + "learning_rate": 3.121226953826073e-05, + "loss": NaN, + "step": 105900 + }, + { + "epoch": 17.22, + "eval_loss": 0.40283602476119995, + "eval_runtime": 7.4114, + "eval_samples_per_second": 13.493, + "eval_wer": 0.3264311814859927, + "step": 105900 + }, + { + "epoch": 17.24, + "learning_rate": 3.114700603687388e-05, + "loss": NaN, + "step": 106000 + }, + { + "epoch": 17.24, + "eval_loss": 0.4006456434726715, + "eval_runtime": 7.4096, + "eval_samples_per_second": 13.496, + "eval_wer": 0.3325213154689403, + "step": 106000 + }, + { + "epoch": 17.25, + "learning_rate": 3.108174253548703e-05, + "loss": NaN, + "step": 106100 + }, + { + "epoch": 17.25, + "eval_loss": 0.43152371048927307, + "eval_runtime": 7.2864, + "eval_samples_per_second": 13.724, + "eval_wer": 0.3386114494518879, + "step": 106100 + }, + { + "epoch": 17.27, + "learning_rate": 3.1016479034100184e-05, + "loss": NaN, + "step": 106200 + }, + { + "epoch": 17.27, + "eval_loss": 0.3727237582206726, + "eval_runtime": 9.6507, + "eval_samples_per_second": 10.362, + "eval_wer": 0.3227771010962241, + "step": 106200 + }, + { + "epoch": 17.29, + "learning_rate": 3.095121553271333e-05, + "loss": NaN, + "step": 106300 + }, + { + "epoch": 17.29, + "eval_loss": 0.4210117757320404, + "eval_runtime": 10.0021, + "eval_samples_per_second": 9.998, + "eval_wer": 0.3264311814859927, + "step": 106300 + }, + { + "epoch": 17.3, + "learning_rate": 3.088595203132648e-05, + "loss": NaN, + "step": 106400 + }, + { + "epoch": 17.3, + "eval_loss": 0.40674594044685364, + "eval_runtime": 9.4605, + "eval_samples_per_second": 10.57, + "eval_wer": 0.34226552984165654, + "step": 106400 + }, + { + "epoch": 17.32, + "learning_rate": 3.082068852993963e-05, + "loss": NaN, + "step": 106500 + }, + { + "epoch": 17.32, + "eval_loss": 0.42881014943122864, + "eval_runtime": 8.646, + "eval_samples_per_second": 11.566, + "eval_wer": 0.3325213154689403, + "step": 106500 + }, + { + "epoch": 17.34, + "learning_rate": 3.075542502855278e-05, + "loss": NaN, + "step": 106600 + }, + { + "epoch": 17.34, + "eval_loss": 0.3960503339767456, + "eval_runtime": 7.5435, + "eval_samples_per_second": 13.256, + "eval_wer": 0.3313032886723508, + "step": 106600 + }, + { + "epoch": 17.35, + "learning_rate": 3.069016152716594e-05, + "loss": NaN, + "step": 106700 + }, + { + "epoch": 17.35, + "eval_loss": 0.5093197226524353, + "eval_runtime": 8.4275, + "eval_samples_per_second": 11.866, + "eval_wer": 0.34226552984165654, + "step": 106700 + }, + { + "epoch": 17.37, + "learning_rate": 3.062489802577909e-05, + "loss": NaN, + "step": 106800 + }, + { + "epoch": 17.37, + "eval_loss": 0.4141642451286316, + "eval_runtime": 8.7304, + "eval_samples_per_second": 11.454, + "eval_wer": 0.3459196102314251, + "step": 106800 + }, + { + "epoch": 17.38, + "learning_rate": 3.0559634524392234e-05, + "loss": NaN, + "step": 106900 + }, + { + "epoch": 17.38, + "eval_loss": 0.3708423376083374, + "eval_runtime": 9.1322, + "eval_samples_per_second": 10.95, + "eval_wer": 0.3276492082825822, + "step": 106900 + }, + { + "epoch": 17.4, + "learning_rate": 3.0494371023005386e-05, + "loss": NaN, + "step": 107000 + }, + { + "epoch": 17.4, + "eval_loss": 0.43578729033470154, + "eval_runtime": 8.1187, + "eval_samples_per_second": 12.317, + "eval_wer": 0.3398294762484775, + "step": 107000 + }, + { + "epoch": 17.42, + "learning_rate": 3.0429107521618538e-05, + "loss": NaN, + "step": 107100 + }, + { + "epoch": 17.42, + "eval_loss": 0.38338860869407654, + "eval_runtime": 8.1539, + "eval_samples_per_second": 12.264, + "eval_wer": 0.32521315468940315, + "step": 107100 + }, + { + "epoch": 17.43, + "learning_rate": 3.0363844020231686e-05, + "loss": NaN, + "step": 107200 + }, + { + "epoch": 17.43, + "eval_loss": 0.3750736713409424, + "eval_runtime": 8.3902, + "eval_samples_per_second": 11.919, + "eval_wer": 0.34348355663824603, + "step": 107200 + }, + { + "epoch": 17.45, + "learning_rate": 3.0298580518844838e-05, + "loss": NaN, + "step": 107300 + }, + { + "epoch": 17.45, + "eval_loss": 0.39097708463668823, + "eval_runtime": 9.385, + "eval_samples_per_second": 10.655, + "eval_wer": 0.34348355663824603, + "step": 107300 + }, + { + "epoch": 17.47, + "learning_rate": 3.023331701745799e-05, + "loss": NaN, + "step": 107400 + }, + { + "epoch": 17.47, + "eval_loss": 0.4204353094100952, + "eval_runtime": 8.9165, + "eval_samples_per_second": 11.215, + "eval_wer": 0.35444579780755175, + "step": 107400 + }, + { + "epoch": 17.48, + "learning_rate": 3.016805351607114e-05, + "loss": NaN, + "step": 107500 + }, + { + "epoch": 17.48, + "eval_loss": 0.40133535861968994, + "eval_runtime": 8.5063, + "eval_samples_per_second": 11.756, + "eval_wer": 0.34226552984165654, + "step": 107500 + }, + { + "epoch": 17.5, + "learning_rate": 3.010279001468429e-05, + "loss": NaN, + "step": 107600 + }, + { + "epoch": 17.5, + "eval_loss": 0.3767768144607544, + "eval_runtime": 7.8527, + "eval_samples_per_second": 12.735, + "eval_wer": 0.32521315468940315, + "step": 107600 + }, + { + "epoch": 17.52, + "learning_rate": 3.0037526513297442e-05, + "loss": NaN, + "step": 107700 + }, + { + "epoch": 17.52, + "eval_loss": 0.3812255859375, + "eval_runtime": 8.1107, + "eval_samples_per_second": 12.329, + "eval_wer": 0.3398294762484775, + "step": 107700 + }, + { + "epoch": 17.53, + "learning_rate": 2.997226301191059e-05, + "loss": NaN, + "step": 107800 + }, + { + "epoch": 17.53, + "eval_loss": 0.40557724237442017, + "eval_runtime": 8.1245, + "eval_samples_per_second": 12.308, + "eval_wer": 0.33008526187576126, + "step": 107800 + }, + { + "epoch": 17.55, + "learning_rate": 2.9906999510523743e-05, + "loss": NaN, + "step": 107900 + }, + { + "epoch": 17.55, + "eval_loss": 0.42703160643577576, + "eval_runtime": 8.9403, + "eval_samples_per_second": 11.185, + "eval_wer": 0.3471376370280146, + "step": 107900 + }, + { + "epoch": 17.56, + "learning_rate": 2.9841736009136888e-05, + "loss": NaN, + "step": 108000 + }, + { + "epoch": 17.56, + "eval_loss": 0.38233357667922974, + "eval_runtime": 7.8203, + "eval_samples_per_second": 12.787, + "eval_wer": 0.34835566382460414, + "step": 108000 + }, + { + "epoch": 17.58, + "learning_rate": 2.9776472507750043e-05, + "loss": NaN, + "step": 108100 + }, + { + "epoch": 17.58, + "eval_loss": 0.4634896218776703, + "eval_runtime": 8.3672, + "eval_samples_per_second": 11.951, + "eval_wer": 0.3520097442143727, + "step": 108100 + }, + { + "epoch": 17.6, + "learning_rate": 2.9711209006363195e-05, + "loss": NaN, + "step": 108200 + }, + { + "epoch": 17.6, + "eval_loss": 0.3950602412223816, + "eval_runtime": 8.3188, + "eval_samples_per_second": 12.021, + "eval_wer": 0.33617539585870887, + "step": 108200 + }, + { + "epoch": 17.61, + "learning_rate": 2.964594550497634e-05, + "loss": NaN, + "step": 108300 + }, + { + "epoch": 17.61, + "eval_loss": 0.3925233781337738, + "eval_runtime": 6.759, + "eval_samples_per_second": 14.795, + "eval_wer": 0.3373934226552984, + "step": 108300 + }, + { + "epoch": 17.63, + "learning_rate": 2.9580682003589492e-05, + "loss": NaN, + "step": 108400 + }, + { + "epoch": 17.63, + "eval_loss": 0.49706822633743286, + "eval_runtime": 9.3441, + "eval_samples_per_second": 10.702, + "eval_wer": 0.35809987819732036, + "step": 108400 + }, + { + "epoch": 17.65, + "learning_rate": 2.9515418502202647e-05, + "loss": NaN, + "step": 108500 + }, + { + "epoch": 17.65, + "eval_loss": 0.4032347798347473, + "eval_runtime": 8.8075, + "eval_samples_per_second": 11.354, + "eval_wer": 0.3398294762484775, + "step": 108500 + }, + { + "epoch": 17.66, + "learning_rate": 2.9450155000815793e-05, + "loss": NaN, + "step": 108600 + }, + { + "epoch": 17.66, + "eval_loss": 0.39194339513778687, + "eval_runtime": 10.3759, + "eval_samples_per_second": 9.638, + "eval_wer": 0.32521315468940315, + "step": 108600 + }, + { + "epoch": 17.68, + "learning_rate": 2.9384891499428945e-05, + "loss": NaN, + "step": 108700 + }, + { + "epoch": 17.68, + "eval_loss": 0.36986175179481506, + "eval_runtime": 7.8636, + "eval_samples_per_second": 12.717, + "eval_wer": 0.32886723507917176, + "step": 108700 + }, + { + "epoch": 17.69, + "learning_rate": 2.93196279980421e-05, + "loss": NaN, + "step": 108800 + }, + { + "epoch": 17.69, + "eval_loss": 0.3867470920085907, + "eval_runtime": 8.137, + "eval_samples_per_second": 12.289, + "eval_wer": 0.34348355663824603, + "step": 108800 + }, + { + "epoch": 17.71, + "learning_rate": 2.9254364496655245e-05, + "loss": NaN, + "step": 108900 + }, + { + "epoch": 17.71, + "eval_loss": 0.3782542943954468, + "eval_runtime": 8.4236, + "eval_samples_per_second": 11.871, + "eval_wer": 0.3325213154689403, + "step": 108900 + }, + { + "epoch": 17.73, + "learning_rate": 2.9189100995268397e-05, + "loss": NaN, + "step": 109000 + }, + { + "epoch": 17.73, + "eval_loss": 0.4252021312713623, + "eval_runtime": 8.512, + "eval_samples_per_second": 11.748, + "eval_wer": 0.3325213154689403, + "step": 109000 + }, + { + "epoch": 17.74, + "learning_rate": 2.9123837493881552e-05, + "loss": NaN, + "step": 109100 + }, + { + "epoch": 17.74, + "eval_loss": 0.3819904327392578, + "eval_runtime": 6.8511, + "eval_samples_per_second": 14.596, + "eval_wer": 0.3373934226552984, + "step": 109100 + }, + { + "epoch": 17.76, + "learning_rate": 2.9058573992494697e-05, + "loss": NaN, + "step": 109200 + }, + { + "epoch": 17.76, + "eval_loss": 0.3850736618041992, + "eval_runtime": 8.4679, + "eval_samples_per_second": 11.809, + "eval_wer": 0.33008526187576126, + "step": 109200 + }, + { + "epoch": 17.78, + "learning_rate": 2.899331049110785e-05, + "loss": NaN, + "step": 109300 + }, + { + "epoch": 17.78, + "eval_loss": 0.4139541685581207, + "eval_runtime": 7.6909, + "eval_samples_per_second": 13.002, + "eval_wer": 0.3386114494518879, + "step": 109300 + }, + { + "epoch": 17.79, + "learning_rate": 2.8928046989721e-05, + "loss": NaN, + "step": 109400 + }, + { + "epoch": 17.79, + "eval_loss": 0.4256812334060669, + "eval_runtime": 7.0399, + "eval_samples_per_second": 14.205, + "eval_wer": 0.3507917174177832, + "step": 109400 + }, + { + "epoch": 17.81, + "learning_rate": 2.886278348833415e-05, + "loss": NaN, + "step": 109500 + }, + { + "epoch": 17.81, + "eval_loss": 0.4498865008354187, + "eval_runtime": 9.6258, + "eval_samples_per_second": 10.389, + "eval_wer": 0.3471376370280146, + "step": 109500 + }, + { + "epoch": 17.82, + "learning_rate": 2.87975199869473e-05, + "loss": NaN, + "step": 109600 + }, + { + "epoch": 17.82, + "eval_loss": 0.4236893057823181, + "eval_runtime": 9.1871, + "eval_samples_per_second": 10.885, + "eval_wer": 0.3617539585870889, + "step": 109600 + }, + { + "epoch": 17.84, + "learning_rate": 2.873225648556045e-05, + "loss": NaN, + "step": 109700 + }, + { + "epoch": 17.84, + "eval_loss": 0.46277669072151184, + "eval_runtime": 9.6292, + "eval_samples_per_second": 10.385, + "eval_wer": 0.35809987819732036, + "step": 109700 + }, + { + "epoch": 17.86, + "learning_rate": 2.8666992984173602e-05, + "loss": NaN, + "step": 109800 + }, + { + "epoch": 17.86, + "eval_loss": 0.46583592891693115, + "eval_runtime": 7.8086, + "eval_samples_per_second": 12.806, + "eval_wer": 0.35322777101096225, + "step": 109800 + }, + { + "epoch": 17.87, + "learning_rate": 2.8601729482786754e-05, + "loss": NaN, + "step": 109900 + }, + { + "epoch": 17.87, + "eval_loss": 0.3932786285877228, + "eval_runtime": 6.8409, + "eval_samples_per_second": 14.618, + "eval_wer": 0.3373934226552984, + "step": 109900 + }, + { + "epoch": 17.89, + "learning_rate": 2.8536465981399902e-05, + "loss": NaN, + "step": 110000 + }, + { + "epoch": 17.89, + "eval_loss": 0.40608084201812744, + "eval_runtime": 7.3141, + "eval_samples_per_second": 13.672, + "eval_wer": 0.3373934226552984, + "step": 110000 + }, + { + "epoch": 17.91, + "learning_rate": 2.8471202480013054e-05, + "loss": NaN, + "step": 110100 + }, + { + "epoch": 17.91, + "eval_loss": 0.4084034860134125, + "eval_runtime": 7.1928, + "eval_samples_per_second": 13.903, + "eval_wer": 0.3373934226552984, + "step": 110100 + }, + { + "epoch": 17.92, + "learning_rate": 2.8405938978626206e-05, + "loss": NaN, + "step": 110200 + }, + { + "epoch": 17.92, + "eval_loss": 0.40555787086486816, + "eval_runtime": 8.5682, + "eval_samples_per_second": 11.671, + "eval_wer": 0.3459196102314251, + "step": 110200 + }, + { + "epoch": 17.94, + "learning_rate": 2.8340675477239355e-05, + "loss": NaN, + "step": 110300 + }, + { + "epoch": 17.94, + "eval_loss": 0.4647526443004608, + "eval_runtime": 8.7905, + "eval_samples_per_second": 11.376, + "eval_wer": 0.341047503045067, + "step": 110300 + }, + { + "epoch": 17.95, + "learning_rate": 2.8275411975852507e-05, + "loss": NaN, + "step": 110400 + }, + { + "epoch": 17.95, + "eval_loss": 0.3823550343513489, + "eval_runtime": 8.631, + "eval_samples_per_second": 11.586, + "eval_wer": 0.341047503045067, + "step": 110400 + }, + { + "epoch": 17.97, + "learning_rate": 2.821014847446566e-05, + "loss": NaN, + "step": 110500 + }, + { + "epoch": 17.97, + "eval_loss": 0.4006991982460022, + "eval_runtime": 7.1345, + "eval_samples_per_second": 14.016, + "eval_wer": 0.3471376370280146, + "step": 110500 + }, + { + "epoch": 17.99, + "learning_rate": 2.8144884973078807e-05, + "loss": NaN, + "step": 110600 + }, + { + "epoch": 17.99, + "eval_loss": 0.3922583758831024, + "eval_runtime": 6.9349, + "eval_samples_per_second": 14.42, + "eval_wer": 0.3507917174177832, + "step": 110600 + }, + { + "epoch": 18.0, + "learning_rate": 2.807962147169196e-05, + "loss": NaN, + "step": 110700 + }, + { + "epoch": 18.0, + "eval_loss": 0.3946372866630554, + "eval_runtime": 6.8519, + "eval_samples_per_second": 14.594, + "eval_wer": 0.341047503045067, + "step": 110700 + }, + { + "epoch": 18.02, + "learning_rate": 2.801435797030511e-05, + "loss": NaN, + "step": 110800 + }, + { + "epoch": 18.02, + "eval_loss": 0.4183802306652069, + "eval_runtime": 7.1866, + "eval_samples_per_second": 13.915, + "eval_wer": 0.341047503045067, + "step": 110800 + }, + { + "epoch": 18.04, + "learning_rate": 2.7949094468918256e-05, + "loss": NaN, + "step": 110900 + }, + { + "epoch": 18.04, + "eval_loss": 0.37478265166282654, + "eval_runtime": 8.3068, + "eval_samples_per_second": 12.038, + "eval_wer": 0.3471376370280146, + "step": 110900 + }, + { + "epoch": 18.05, + "learning_rate": 2.788383096753141e-05, + "loss": NaN, + "step": 111000 + }, + { + "epoch": 18.05, + "eval_loss": 0.3887958526611328, + "eval_runtime": 7.7706, + "eval_samples_per_second": 12.869, + "eval_wer": 0.34957369062119364, + "step": 111000 + }, + { + "epoch": 18.07, + "learning_rate": 2.7818567466144557e-05, + "loss": NaN, + "step": 111100 + }, + { + "epoch": 18.07, + "eval_loss": 0.3766718804836273, + "eval_runtime": 7.7942, + "eval_samples_per_second": 12.83, + "eval_wer": 0.3447015834348356, + "step": 111100 + }, + { + "epoch": 18.08, + "learning_rate": 2.775330396475771e-05, + "loss": NaN, + "step": 111200 + }, + { + "epoch": 18.08, + "eval_loss": 0.39117416739463806, + "eval_runtime": 8.1811, + "eval_samples_per_second": 12.223, + "eval_wer": 0.341047503045067, + "step": 111200 + }, + { + "epoch": 18.1, + "learning_rate": 2.7688040463370864e-05, + "loss": NaN, + "step": 111300 + }, + { + "epoch": 18.1, + "eval_loss": 0.41156256198883057, + "eval_runtime": 8.6071, + "eval_samples_per_second": 11.618, + "eval_wer": 0.341047503045067, + "step": 111300 + }, + { + "epoch": 18.12, + "learning_rate": 2.762277696198401e-05, + "loss": NaN, + "step": 111400 + }, + { + "epoch": 18.12, + "eval_loss": 0.39620694518089294, + "eval_runtime": 8.9372, + "eval_samples_per_second": 11.189, + "eval_wer": 0.35322777101096225, + "step": 111400 + }, + { + "epoch": 18.13, + "learning_rate": 2.755751346059716e-05, + "loss": NaN, + "step": 111500 + }, + { + "epoch": 18.13, + "eval_loss": 0.40653663873672485, + "eval_runtime": 7.1536, + "eval_samples_per_second": 13.979, + "eval_wer": 0.3313032886723508, + "step": 111500 + }, + { + "epoch": 18.15, + "learning_rate": 2.7492249959210313e-05, + "loss": NaN, + "step": 111600 + }, + { + "epoch": 18.15, + "eval_loss": 0.43940940499305725, + "eval_runtime": 7.214, + "eval_samples_per_second": 13.862, + "eval_wer": 0.34835566382460414, + "step": 111600 + }, + { + "epoch": 18.17, + "learning_rate": 2.742698645782346e-05, + "loss": NaN, + "step": 111700 + }, + { + "epoch": 18.17, + "eval_loss": 0.43023011088371277, + "eval_runtime": 9.4948, + "eval_samples_per_second": 10.532, + "eval_wer": 0.35931790499390986, + "step": 111700 + }, + { + "epoch": 18.18, + "learning_rate": 2.7361722956436613e-05, + "loss": NaN, + "step": 111800 + }, + { + "epoch": 18.18, + "eval_loss": 0.47506821155548096, + "eval_runtime": 8.7209, + "eval_samples_per_second": 11.467, + "eval_wer": 0.3507917174177832, + "step": 111800 + }, + { + "epoch": 18.2, + "learning_rate": 2.7296459455049765e-05, + "loss": NaN, + "step": 111900 + }, + { + "epoch": 18.2, + "eval_loss": 0.4082556962966919, + "eval_runtime": 9.4452, + "eval_samples_per_second": 10.587, + "eval_wer": 0.364190012180268, + "step": 111900 + }, + { + "epoch": 18.21, + "learning_rate": 2.7231195953662914e-05, + "loss": NaN, + "step": 112000 + }, + { + "epoch": 18.21, + "eval_loss": 0.44821274280548096, + "eval_runtime": 7.7256, + "eval_samples_per_second": 12.944, + "eval_wer": 0.3313032886723508, + "step": 112000 + }, + { + "epoch": 18.23, + "learning_rate": 2.7165932452276066e-05, + "loss": NaN, + "step": 112100 + }, + { + "epoch": 18.23, + "eval_loss": 0.4030204117298126, + "eval_runtime": 8.0067, + "eval_samples_per_second": 12.49, + "eval_wer": 0.3507917174177832, + "step": 112100 + }, + { + "epoch": 18.25, + "learning_rate": 2.7100668950889218e-05, + "loss": NaN, + "step": 112200 + }, + { + "epoch": 18.25, + "eval_loss": 0.3854163885116577, + "eval_runtime": 8.2701, + "eval_samples_per_second": 12.092, + "eval_wer": 0.3398294762484775, + "step": 112200 + }, + { + "epoch": 18.26, + "learning_rate": 2.7035405449502366e-05, + "loss": NaN, + "step": 112300 + }, + { + "epoch": 18.26, + "eval_loss": 0.4053157866001129, + "eval_runtime": 7.0908, + "eval_samples_per_second": 14.103, + "eval_wer": 0.3313032886723508, + "step": 112300 + }, + { + "epoch": 18.28, + "learning_rate": 2.6970141948115518e-05, + "loss": NaN, + "step": 112400 + }, + { + "epoch": 18.28, + "eval_loss": 0.3890162706375122, + "eval_runtime": 7.2231, + "eval_samples_per_second": 13.844, + "eval_wer": 0.33008526187576126, + "step": 112400 + }, + { + "epoch": 18.3, + "learning_rate": 2.690487844672867e-05, + "loss": NaN, + "step": 112500 + }, + { + "epoch": 18.3, + "eval_loss": 0.3915853500366211, + "eval_runtime": 8.7452, + "eval_samples_per_second": 11.435, + "eval_wer": 0.3447015834348356, + "step": 112500 + }, + { + "epoch": 18.31, + "learning_rate": 2.683961494534182e-05, + "loss": NaN, + "step": 112600 + }, + { + "epoch": 18.31, + "eval_loss": 0.3701234459877014, + "eval_runtime": 7.3669, + "eval_samples_per_second": 13.574, + "eval_wer": 0.3313032886723508, + "step": 112600 + }, + { + "epoch": 18.33, + "learning_rate": 2.677435144395497e-05, + "loss": NaN, + "step": 112700 + }, + { + "epoch": 18.33, + "eval_loss": 0.399958074092865, + "eval_runtime": 8.536, + "eval_samples_per_second": 11.715, + "eval_wer": 0.34226552984165654, + "step": 112700 + }, + { + "epoch": 18.34, + "learning_rate": 2.670908794256812e-05, + "loss": NaN, + "step": 112800 + }, + { + "epoch": 18.34, + "eval_loss": 0.3959391117095947, + "eval_runtime": 7.6581, + "eval_samples_per_second": 13.058, + "eval_wer": 0.341047503045067, + "step": 112800 + }, + { + "epoch": 18.36, + "learning_rate": 2.664382444118127e-05, + "loss": NaN, + "step": 112900 + }, + { + "epoch": 18.36, + "eval_loss": 0.4151996672153473, + "eval_runtime": 8.5593, + "eval_samples_per_second": 11.683, + "eval_wer": 0.3398294762484775, + "step": 112900 + }, + { + "epoch": 18.38, + "learning_rate": 2.6578560939794423e-05, + "loss": NaN, + "step": 113000 + }, + { + "epoch": 18.38, + "eval_loss": 0.388653039932251, + "eval_runtime": 7.3021, + "eval_samples_per_second": 13.695, + "eval_wer": 0.341047503045067, + "step": 113000 + }, + { + "epoch": 18.39, + "learning_rate": 2.6513297438407568e-05, + "loss": NaN, + "step": 113100 + }, + { + "epoch": 18.39, + "eval_loss": 0.3932548761367798, + "eval_runtime": 7.3475, + "eval_samples_per_second": 13.61, + "eval_wer": 0.3447015834348356, + "step": 113100 + }, + { + "epoch": 18.41, + "learning_rate": 2.6448033937020723e-05, + "loss": NaN, + "step": 113200 + }, + { + "epoch": 18.41, + "eval_loss": 0.5715663433074951, + "eval_runtime": 6.4752, + "eval_samples_per_second": 15.444, + "eval_wer": 0.3459196102314251, + "step": 113200 + }, + { + "epoch": 18.43, + "learning_rate": 2.6382770435633875e-05, + "loss": NaN, + "step": 113300 + }, + { + "epoch": 18.43, + "eval_loss": 0.5505608320236206, + "eval_runtime": 8.5504, + "eval_samples_per_second": 11.695, + "eval_wer": 0.3459196102314251, + "step": 113300 + }, + { + "epoch": 18.44, + "learning_rate": 2.631750693424702e-05, + "loss": NaN, + "step": 113400 + }, + { + "epoch": 18.44, + "eval_loss": 0.4306776523590088, + "eval_runtime": 8.0297, + "eval_samples_per_second": 12.454, + "eval_wer": 0.33008526187576126, + "step": 113400 + }, + { + "epoch": 18.46, + "learning_rate": 2.6252243432860175e-05, + "loss": NaN, + "step": 113500 + }, + { + "epoch": 18.46, + "eval_loss": 0.3789026737213135, + "eval_runtime": 8.1239, + "eval_samples_per_second": 12.309, + "eval_wer": 0.3325213154689403, + "step": 113500 + }, + { + "epoch": 18.47, + "learning_rate": 2.6186979931473327e-05, + "loss": NaN, + "step": 113600 + }, + { + "epoch": 18.47, + "eval_loss": 0.42162245512008667, + "eval_runtime": 8.6954, + "eval_samples_per_second": 11.5, + "eval_wer": 0.35322777101096225, + "step": 113600 + }, + { + "epoch": 18.49, + "learning_rate": 2.6121716430086473e-05, + "loss": NaN, + "step": 113700 + }, + { + "epoch": 18.49, + "eval_loss": 0.5023617744445801, + "eval_runtime": 9.6846, + "eval_samples_per_second": 10.326, + "eval_wer": 0.3507917174177832, + "step": 113700 + }, + { + "epoch": 18.51, + "learning_rate": 2.6056452928699628e-05, + "loss": NaN, + "step": 113800 + }, + { + "epoch": 18.51, + "eval_loss": 0.395231693983078, + "eval_runtime": 9.953, + "eval_samples_per_second": 10.047, + "eval_wer": 0.3471376370280146, + "step": 113800 + }, + { + "epoch": 18.52, + "learning_rate": 2.599118942731278e-05, + "loss": NaN, + "step": 113900 + }, + { + "epoch": 18.52, + "eval_loss": 0.39416807889938354, + "eval_runtime": 7.819, + "eval_samples_per_second": 12.789, + "eval_wer": 0.33617539585870887, + "step": 113900 + }, + { + "epoch": 18.54, + "learning_rate": 2.5925925925925925e-05, + "loss": NaN, + "step": 114000 + }, + { + "epoch": 18.54, + "eval_loss": 0.5159714221954346, + "eval_runtime": 7.203, + "eval_samples_per_second": 13.883, + "eval_wer": 0.3337393422655298, + "step": 114000 + }, + { + "epoch": 18.56, + "learning_rate": 2.5860662424539077e-05, + "loss": NaN, + "step": 114100 + }, + { + "epoch": 18.56, + "eval_loss": 0.39711394906044006, + "eval_runtime": 8.1764, + "eval_samples_per_second": 12.23, + "eval_wer": 0.341047503045067, + "step": 114100 + }, + { + "epoch": 18.57, + "learning_rate": 2.5795398923152232e-05, + "loss": NaN, + "step": 114200 + }, + { + "epoch": 18.57, + "eval_loss": 0.3980695605278015, + "eval_runtime": 7.8237, + "eval_samples_per_second": 12.782, + "eval_wer": 0.341047503045067, + "step": 114200 + }, + { + "epoch": 18.59, + "learning_rate": 2.5730135421765377e-05, + "loss": NaN, + "step": 114300 + }, + { + "epoch": 18.59, + "eval_loss": 0.39846745133399963, + "eval_runtime": 9.4234, + "eval_samples_per_second": 10.612, + "eval_wer": 0.3398294762484775, + "step": 114300 + }, + { + "epoch": 18.6, + "learning_rate": 2.566487192037853e-05, + "loss": NaN, + "step": 114400 + }, + { + "epoch": 18.6, + "eval_loss": 0.4071855843067169, + "eval_runtime": 7.858, + "eval_samples_per_second": 12.726, + "eval_wer": 0.341047503045067, + "step": 114400 + }, + { + "epoch": 18.62, + "learning_rate": 2.5599608418991678e-05, + "loss": NaN, + "step": 114500 + }, + { + "epoch": 18.62, + "eval_loss": 0.41952818632125854, + "eval_runtime": 8.1596, + "eval_samples_per_second": 12.256, + "eval_wer": 0.3447015834348356, + "step": 114500 + }, + { + "epoch": 18.64, + "learning_rate": 2.553434491760483e-05, + "loss": NaN, + "step": 114600 + }, + { + "epoch": 18.64, + "eval_loss": 0.44777679443359375, + "eval_runtime": 8.1577, + "eval_samples_per_second": 12.258, + "eval_wer": 0.3325213154689403, + "step": 114600 + }, + { + "epoch": 18.65, + "learning_rate": 2.546908141621798e-05, + "loss": NaN, + "step": 114700 + }, + { + "epoch": 18.65, + "eval_loss": 0.39139124751091003, + "eval_runtime": 7.223, + "eval_samples_per_second": 13.845, + "eval_wer": 0.32399512789281365, + "step": 114700 + }, + { + "epoch": 18.67, + "learning_rate": 2.540381791483113e-05, + "loss": NaN, + "step": 114800 + }, + { + "epoch": 18.67, + "eval_loss": 0.40552714467048645, + "eval_runtime": 8.0589, + "eval_samples_per_second": 12.409, + "eval_wer": 0.3313032886723508, + "step": 114800 + }, + { + "epoch": 18.69, + "learning_rate": 2.5338554413444282e-05, + "loss": NaN, + "step": 114900 + }, + { + "epoch": 18.69, + "eval_loss": 0.4118463397026062, + "eval_runtime": 8.077, + "eval_samples_per_second": 12.381, + "eval_wer": 0.3276492082825822, + "step": 114900 + }, + { + "epoch": 18.7, + "learning_rate": 2.5273290912057434e-05, + "loss": NaN, + "step": 115000 + }, + { + "epoch": 18.7, + "eval_loss": 0.4166834354400635, + "eval_runtime": 8.1784, + "eval_samples_per_second": 12.227, + "eval_wer": 0.34348355663824603, + "step": 115000 + }, + { + "epoch": 18.72, + "learning_rate": 2.5208027410670582e-05, + "loss": NaN, + "step": 115100 + }, + { + "epoch": 18.72, + "eval_loss": 0.4071851074695587, + "eval_runtime": 10.5991, + "eval_samples_per_second": 9.435, + "eval_wer": 0.3276492082825822, + "step": 115100 + }, + { + "epoch": 18.73, + "learning_rate": 2.5142763909283734e-05, + "loss": NaN, + "step": 115200 + }, + { + "epoch": 18.73, + "eval_loss": 0.44569680094718933, + "eval_runtime": 7.985, + "eval_samples_per_second": 12.523, + "eval_wer": 0.3398294762484775, + "step": 115200 + }, + { + "epoch": 18.75, + "learning_rate": 2.5077500407896886e-05, + "loss": NaN, + "step": 115300 + }, + { + "epoch": 18.75, + "eval_loss": 0.4303815960884094, + "eval_runtime": 7.2375, + "eval_samples_per_second": 13.817, + "eval_wer": 0.3337393422655298, + "step": 115300 + }, + { + "epoch": 18.77, + "learning_rate": 2.5012236906510035e-05, + "loss": NaN, + "step": 115400 + }, + { + "epoch": 18.77, + "eval_loss": 0.4490242898464203, + "eval_runtime": 8.2404, + "eval_samples_per_second": 12.135, + "eval_wer": 0.34226552984165654, + "step": 115400 + }, + { + "epoch": 18.78, + "learning_rate": 2.4946973405123187e-05, + "loss": NaN, + "step": 115500 + }, + { + "epoch": 18.78, + "eval_loss": 0.49515098333358765, + "eval_runtime": 7.3002, + "eval_samples_per_second": 13.698, + "eval_wer": 0.3373934226552984, + "step": 115500 + }, + { + "epoch": 18.8, + "learning_rate": 2.4881709903736335e-05, + "loss": NaN, + "step": 115600 + }, + { + "epoch": 18.8, + "eval_loss": 0.4241500496864319, + "eval_runtime": 8.6244, + "eval_samples_per_second": 11.595, + "eval_wer": 0.3276492082825822, + "step": 115600 + }, + { + "epoch": 18.82, + "learning_rate": 2.4816446402349487e-05, + "loss": NaN, + "step": 115700 + }, + { + "epoch": 18.82, + "eval_loss": 0.4969961941242218, + "eval_runtime": 8.4269, + "eval_samples_per_second": 11.867, + "eval_wer": 0.34835566382460414, + "step": 115700 + }, + { + "epoch": 18.83, + "learning_rate": 2.475118290096264e-05, + "loss": NaN, + "step": 115800 + }, + { + "epoch": 18.83, + "eval_loss": 0.5238353610038757, + "eval_runtime": 8.3935, + "eval_samples_per_second": 11.914, + "eval_wer": 0.34226552984165654, + "step": 115800 + }, + { + "epoch": 18.85, + "learning_rate": 2.4685919399575788e-05, + "loss": NaN, + "step": 115900 + }, + { + "epoch": 18.85, + "eval_loss": 0.43089020252227783, + "eval_runtime": 7.219, + "eval_samples_per_second": 13.852, + "eval_wer": 0.32034104750304504, + "step": 115900 + }, + { + "epoch": 18.86, + "learning_rate": 2.462065589818894e-05, + "loss": NaN, + "step": 116000 + }, + { + "epoch": 18.86, + "eval_loss": 0.41796696186065674, + "eval_runtime": 7.5558, + "eval_samples_per_second": 13.235, + "eval_wer": 0.3313032886723508, + "step": 116000 + }, + { + "epoch": 18.88, + "learning_rate": 2.455539239680209e-05, + "loss": NaN, + "step": 116100 + }, + { + "epoch": 18.88, + "eval_loss": 0.4422737956047058, + "eval_runtime": 7.9398, + "eval_samples_per_second": 12.595, + "eval_wer": 0.32886723507917176, + "step": 116100 + }, + { + "epoch": 18.9, + "learning_rate": 2.449012889541524e-05, + "loss": NaN, + "step": 116200 + }, + { + "epoch": 18.9, + "eval_loss": 0.43862995505332947, + "eval_runtime": 8.9085, + "eval_samples_per_second": 11.225, + "eval_wer": 0.3337393422655298, + "step": 116200 + }, + { + "epoch": 18.91, + "learning_rate": 2.442486539402839e-05, + "loss": NaN, + "step": 116300 + }, + { + "epoch": 18.91, + "eval_loss": 0.4497430920600891, + "eval_runtime": 9.2945, + "eval_samples_per_second": 10.759, + "eval_wer": 0.3447015834348356, + "step": 116300 + }, + { + "epoch": 18.93, + "learning_rate": 2.4359601892641544e-05, + "loss": NaN, + "step": 116400 + }, + { + "epoch": 18.93, + "eval_loss": 0.4064001739025116, + "eval_runtime": 9.3152, + "eval_samples_per_second": 10.735, + "eval_wer": 0.3276492082825822, + "step": 116400 + }, + { + "epoch": 18.95, + "learning_rate": 2.4294338391254692e-05, + "loss": NaN, + "step": 116500 + }, + { + "epoch": 18.95, + "eval_loss": 0.5357220768928528, + "eval_runtime": 10.268, + "eval_samples_per_second": 9.739, + "eval_wer": 0.35444579780755175, + "step": 116500 + }, + { + "epoch": 18.96, + "learning_rate": 2.422907488986784e-05, + "loss": NaN, + "step": 116600 + }, + { + "epoch": 18.96, + "eval_loss": 0.4020291864871979, + "eval_runtime": 6.8978, + "eval_samples_per_second": 14.497, + "eval_wer": 0.341047503045067, + "step": 116600 + }, + { + "epoch": 18.98, + "learning_rate": 2.4163811388480993e-05, + "loss": NaN, + "step": 116700 + }, + { + "epoch": 18.98, + "eval_loss": 0.5173760056495667, + "eval_runtime": 8.2411, + "eval_samples_per_second": 12.134, + "eval_wer": 0.3447015834348356, + "step": 116700 + }, + { + "epoch": 18.99, + "learning_rate": 2.4098547887094145e-05, + "loss": NaN, + "step": 116800 + }, + { + "epoch": 18.99, + "eval_loss": 0.4323920011520386, + "eval_runtime": 8.3858, + "eval_samples_per_second": 11.925, + "eval_wer": 0.3556638246041413, + "step": 116800 + }, + { + "epoch": 19.01, + "learning_rate": 2.4033284385707293e-05, + "loss": NaN, + "step": 116900 + }, + { + "epoch": 19.01, + "eval_loss": 0.4196859300136566, + "eval_runtime": 8.9105, + "eval_samples_per_second": 11.223, + "eval_wer": 0.33617539585870887, + "step": 116900 + }, + { + "epoch": 19.03, + "learning_rate": 2.3968020884320445e-05, + "loss": NaN, + "step": 117000 + }, + { + "epoch": 19.03, + "eval_loss": 0.4224309027194977, + "eval_runtime": 7.9389, + "eval_samples_per_second": 12.596, + "eval_wer": 0.3325213154689403, + "step": 117000 + }, + { + "epoch": 19.04, + "learning_rate": 2.3902757382933597e-05, + "loss": NaN, + "step": 117100 + }, + { + "epoch": 19.04, + "eval_loss": 0.44540169835090637, + "eval_runtime": 8.0457, + "eval_samples_per_second": 12.429, + "eval_wer": 0.33008526187576126, + "step": 117100 + }, + { + "epoch": 19.06, + "learning_rate": 2.3837493881546746e-05, + "loss": NaN, + "step": 117200 + }, + { + "epoch": 19.06, + "eval_loss": 0.5561581254005432, + "eval_runtime": 8.6307, + "eval_samples_per_second": 11.587, + "eval_wer": 0.32521315468940315, + "step": 117200 + }, + { + "epoch": 19.08, + "learning_rate": 2.3772230380159897e-05, + "loss": NaN, + "step": 117300 + }, + { + "epoch": 19.08, + "eval_loss": 0.49928075075149536, + "eval_runtime": 7.705, + "eval_samples_per_second": 12.979, + "eval_wer": 0.3386114494518879, + "step": 117300 + }, + { + "epoch": 19.09, + "learning_rate": 2.370696687877305e-05, + "loss": NaN, + "step": 117400 + }, + { + "epoch": 19.09, + "eval_loss": 0.36428502202033997, + "eval_runtime": 8.6228, + "eval_samples_per_second": 11.597, + "eval_wer": 0.32886723507917176, + "step": 117400 + }, + { + "epoch": 19.11, + "learning_rate": 2.3641703377386198e-05, + "loss": NaN, + "step": 117500 + }, + { + "epoch": 19.11, + "eval_loss": 0.4284897744655609, + "eval_runtime": 9.3991, + "eval_samples_per_second": 10.639, + "eval_wer": 0.3373934226552984, + "step": 117500 + }, + { + "epoch": 19.13, + "learning_rate": 2.357643987599935e-05, + "loss": NaN, + "step": 117600 + }, + { + "epoch": 19.13, + "eval_loss": 0.4161425828933716, + "eval_runtime": 7.4568, + "eval_samples_per_second": 13.411, + "eval_wer": 0.3313032886723508, + "step": 117600 + }, + { + "epoch": 19.14, + "learning_rate": 2.35111763746125e-05, + "loss": NaN, + "step": 117700 + }, + { + "epoch": 19.14, + "eval_loss": 0.4765840470790863, + "eval_runtime": 9.8648, + "eval_samples_per_second": 10.137, + "eval_wer": 0.34226552984165654, + "step": 117700 + }, + { + "epoch": 19.16, + "learning_rate": 2.344591287322565e-05, + "loss": NaN, + "step": 117800 + }, + { + "epoch": 19.16, + "eval_loss": 0.41543564200401306, + "eval_runtime": 7.7288, + "eval_samples_per_second": 12.939, + "eval_wer": 0.32886723507917176, + "step": 117800 + }, + { + "epoch": 19.17, + "learning_rate": 2.33806493718388e-05, + "loss": NaN, + "step": 117900 + }, + { + "epoch": 19.17, + "eval_loss": 0.4077683985233307, + "eval_runtime": 6.5049, + "eval_samples_per_second": 15.373, + "eval_wer": 0.32886723507917176, + "step": 117900 + }, + { + "epoch": 19.19, + "learning_rate": 2.331538587045195e-05, + "loss": NaN, + "step": 118000 + }, + { + "epoch": 19.19, + "eval_loss": 0.4288666546344757, + "eval_runtime": 7.3525, + "eval_samples_per_second": 13.601, + "eval_wer": 0.3325213154689403, + "step": 118000 + }, + { + "epoch": 19.21, + "learning_rate": 2.3250122369065103e-05, + "loss": NaN, + "step": 118100 + }, + { + "epoch": 19.21, + "eval_loss": 0.3833557963371277, + "eval_runtime": 8.6001, + "eval_samples_per_second": 11.628, + "eval_wer": 0.32521315468940315, + "step": 118100 + }, + { + "epoch": 19.22, + "learning_rate": 2.318485886767825e-05, + "loss": NaN, + "step": 118200 + }, + { + "epoch": 19.22, + "eval_loss": 0.4251144528388977, + "eval_runtime": 8.0377, + "eval_samples_per_second": 12.441, + "eval_wer": 0.315468940316687, + "step": 118200 + }, + { + "epoch": 19.24, + "learning_rate": 2.3119595366291403e-05, + "loss": NaN, + "step": 118300 + }, + { + "epoch": 19.24, + "eval_loss": 0.48235252499580383, + "eval_runtime": 9.9197, + "eval_samples_per_second": 10.081, + "eval_wer": 0.3373934226552984, + "step": 118300 + }, + { + "epoch": 19.26, + "learning_rate": 2.305433186490455e-05, + "loss": NaN, + "step": 118400 + }, + { + "epoch": 19.26, + "eval_loss": 0.45876410603523254, + "eval_runtime": 8.0238, + "eval_samples_per_second": 12.463, + "eval_wer": 0.3398294762484775, + "step": 118400 + }, + { + "epoch": 19.27, + "learning_rate": 2.2989068363517704e-05, + "loss": NaN, + "step": 118500 + }, + { + "epoch": 19.27, + "eval_loss": 0.42962944507598877, + "eval_runtime": 7.6922, + "eval_samples_per_second": 13.0, + "eval_wer": 0.32886723507917176, + "step": 118500 + }, + { + "epoch": 19.29, + "learning_rate": 2.2923804862130855e-05, + "loss": NaN, + "step": 118600 + }, + { + "epoch": 19.29, + "eval_loss": 0.43103399872779846, + "eval_runtime": 8.5247, + "eval_samples_per_second": 11.731, + "eval_wer": 0.3313032886723508, + "step": 118600 + }, + { + "epoch": 19.3, + "learning_rate": 2.2858541360744004e-05, + "loss": NaN, + "step": 118700 + }, + { + "epoch": 19.3, + "eval_loss": 0.41755372285842896, + "eval_runtime": 7.9327, + "eval_samples_per_second": 12.606, + "eval_wer": 0.3264311814859927, + "step": 118700 + }, + { + "epoch": 19.32, + "learning_rate": 2.2793277859357156e-05, + "loss": NaN, + "step": 118800 + }, + { + "epoch": 19.32, + "eval_loss": 0.40416061878204346, + "eval_runtime": 7.7154, + "eval_samples_per_second": 12.961, + "eval_wer": 0.3215590742996346, + "step": 118800 + }, + { + "epoch": 19.34, + "learning_rate": 2.2728014357970308e-05, + "loss": NaN, + "step": 118900 + }, + { + "epoch": 19.34, + "eval_loss": 0.42045220732688904, + "eval_runtime": 7.9703, + "eval_samples_per_second": 12.547, + "eval_wer": 0.3325213154689403, + "step": 118900 + }, + { + "epoch": 19.35, + "learning_rate": 2.2662750856583456e-05, + "loss": NaN, + "step": 119000 + }, + { + "epoch": 19.35, + "eval_loss": 0.4497748613357544, + "eval_runtime": 8.152, + "eval_samples_per_second": 12.267, + "eval_wer": 0.33495736906211937, + "step": 119000 + }, + { + "epoch": 19.37, + "learning_rate": 2.2597487355196608e-05, + "loss": NaN, + "step": 119100 + }, + { + "epoch": 19.37, + "eval_loss": 0.4549383819103241, + "eval_runtime": 8.0924, + "eval_samples_per_second": 12.357, + "eval_wer": 0.3447015834348356, + "step": 119100 + }, + { + "epoch": 19.39, + "learning_rate": 2.253222385380976e-05, + "loss": NaN, + "step": 119200 + }, + { + "epoch": 19.39, + "eval_loss": 0.4458554983139038, + "eval_runtime": 8.5049, + "eval_samples_per_second": 11.758, + "eval_wer": 0.33495736906211937, + "step": 119200 + }, + { + "epoch": 19.4, + "learning_rate": 2.246696035242291e-05, + "loss": NaN, + "step": 119300 + }, + { + "epoch": 19.4, + "eval_loss": 0.43120619654655457, + "eval_runtime": 7.0431, + "eval_samples_per_second": 14.198, + "eval_wer": 0.3264311814859927, + "step": 119300 + }, + { + "epoch": 19.42, + "learning_rate": 2.2401696851036057e-05, + "loss": NaN, + "step": 119400 + }, + { + "epoch": 19.42, + "eval_loss": 0.4251594841480255, + "eval_runtime": 7.4385, + "eval_samples_per_second": 13.444, + "eval_wer": 0.3386114494518879, + "step": 119400 + }, + { + "epoch": 19.43, + "learning_rate": 2.233643334964921e-05, + "loss": NaN, + "step": 119500 + }, + { + "epoch": 19.43, + "eval_loss": 0.42931661009788513, + "eval_runtime": 9.2934, + "eval_samples_per_second": 10.76, + "eval_wer": 0.3507917174177832, + "step": 119500 + }, + { + "epoch": 19.45, + "learning_rate": 2.227116984826236e-05, + "loss": NaN, + "step": 119600 + }, + { + "epoch": 19.45, + "eval_loss": 0.42289912700653076, + "eval_runtime": 9.665, + "eval_samples_per_second": 10.347, + "eval_wer": 0.3398294762484775, + "step": 119600 + }, + { + "epoch": 19.47, + "learning_rate": 2.220590634687551e-05, + "loss": NaN, + "step": 119700 + }, + { + "epoch": 19.47, + "eval_loss": 0.487996369600296, + "eval_runtime": 8.5886, + "eval_samples_per_second": 11.643, + "eval_wer": 0.34348355663824603, + "step": 119700 + }, + { + "epoch": 19.48, + "learning_rate": 2.214064284548866e-05, + "loss": NaN, + "step": 119800 + }, + { + "epoch": 19.48, + "eval_loss": 0.550142228603363, + "eval_runtime": 9.4924, + "eval_samples_per_second": 10.535, + "eval_wer": 0.3556638246041413, + "step": 119800 + }, + { + "epoch": 19.5, + "learning_rate": 2.2075379344101813e-05, + "loss": NaN, + "step": 119900 + }, + { + "epoch": 19.5, + "eval_loss": 0.43226540088653564, + "eval_runtime": 7.492, + "eval_samples_per_second": 13.348, + "eval_wer": 0.32886723507917176, + "step": 119900 + }, + { + "epoch": 19.52, + "learning_rate": 2.2010115842714962e-05, + "loss": NaN, + "step": 120000 + }, + { + "epoch": 19.52, + "eval_loss": 0.3998276889324188, + "eval_runtime": 7.4602, + "eval_samples_per_second": 13.404, + "eval_wer": 0.3325213154689403, + "step": 120000 + }, + { + "epoch": 19.53, + "learning_rate": 2.194485234132811e-05, + "loss": NaN, + "step": 120100 + }, + { + "epoch": 19.53, + "eval_loss": 0.4583325982093811, + "eval_runtime": 7.7177, + "eval_samples_per_second": 12.957, + "eval_wer": 0.34348355663824603, + "step": 120100 + }, + { + "epoch": 19.55, + "learning_rate": 2.1879588839941266e-05, + "loss": NaN, + "step": 120200 + }, + { + "epoch": 19.55, + "eval_loss": 0.44406771659851074, + "eval_runtime": 8.2416, + "eval_samples_per_second": 12.134, + "eval_wer": 0.341047503045067, + "step": 120200 + }, + { + "epoch": 19.56, + "learning_rate": 2.1814325338554414e-05, + "loss": NaN, + "step": 120300 + }, + { + "epoch": 19.56, + "eval_loss": 0.41883566975593567, + "eval_runtime": 7.3107, + "eval_samples_per_second": 13.679, + "eval_wer": 0.34348355663824603, + "step": 120300 + }, + { + "epoch": 19.58, + "learning_rate": 2.1749061837167563e-05, + "loss": NaN, + "step": 120400 + }, + { + "epoch": 19.58, + "eval_loss": 0.4323764741420746, + "eval_runtime": 8.8296, + "eval_samples_per_second": 11.325, + "eval_wer": 0.341047503045067, + "step": 120400 + }, + { + "epoch": 19.6, + "learning_rate": 2.1683798335780718e-05, + "loss": NaN, + "step": 120500 + }, + { + "epoch": 19.6, + "eval_loss": 0.41844385862350464, + "eval_runtime": 8.8708, + "eval_samples_per_second": 11.273, + "eval_wer": 0.34957369062119364, + "step": 120500 + }, + { + "epoch": 19.61, + "learning_rate": 2.1618534834393867e-05, + "loss": NaN, + "step": 120600 + }, + { + "epoch": 19.61, + "eval_loss": 0.39420318603515625, + "eval_runtime": 9.8796, + "eval_samples_per_second": 10.122, + "eval_wer": 0.3325213154689403, + "step": 120600 + }, + { + "epoch": 19.63, + "learning_rate": 2.1553271333007015e-05, + "loss": NaN, + "step": 120700 + }, + { + "epoch": 19.63, + "eval_loss": 0.475872278213501, + "eval_runtime": 8.1761, + "eval_samples_per_second": 12.231, + "eval_wer": 0.3471376370280146, + "step": 120700 + }, + { + "epoch": 19.65, + "learning_rate": 2.148800783162017e-05, + "loss": NaN, + "step": 120800 + }, + { + "epoch": 19.65, + "eval_loss": 0.4005354642868042, + "eval_runtime": 8.1599, + "eval_samples_per_second": 12.255, + "eval_wer": 0.34226552984165654, + "step": 120800 + }, + { + "epoch": 19.66, + "learning_rate": 2.142274433023332e-05, + "loss": NaN, + "step": 120900 + }, + { + "epoch": 19.66, + "eval_loss": 0.3836045563220978, + "eval_runtime": 8.6109, + "eval_samples_per_second": 11.613, + "eval_wer": 0.3459196102314251, + "step": 120900 + }, + { + "epoch": 19.68, + "learning_rate": 2.1357480828846468e-05, + "loss": NaN, + "step": 121000 + }, + { + "epoch": 19.68, + "eval_loss": 0.36968758702278137, + "eval_runtime": 8.2911, + "eval_samples_per_second": 12.061, + "eval_wer": 0.33495736906211937, + "step": 121000 + }, + { + "epoch": 19.69, + "learning_rate": 2.129221732745962e-05, + "loss": NaN, + "step": 121100 + }, + { + "epoch": 19.69, + "eval_loss": 0.4322676360607147, + "eval_runtime": 8.1328, + "eval_samples_per_second": 12.296, + "eval_wer": 0.34348355663824603, + "step": 121100 + }, + { + "epoch": 19.71, + "learning_rate": 2.122695382607277e-05, + "loss": NaN, + "step": 121200 + }, + { + "epoch": 19.71, + "eval_loss": 0.42114943265914917, + "eval_runtime": 9.3436, + "eval_samples_per_second": 10.703, + "eval_wer": 0.341047503045067, + "step": 121200 + }, + { + "epoch": 19.73, + "learning_rate": 2.116169032468592e-05, + "loss": NaN, + "step": 121300 + }, + { + "epoch": 19.73, + "eval_loss": 0.40339621901512146, + "eval_runtime": 7.8643, + "eval_samples_per_second": 12.716, + "eval_wer": 0.3337393422655298, + "step": 121300 + }, + { + "epoch": 19.74, + "learning_rate": 2.1096426823299072e-05, + "loss": NaN, + "step": 121400 + }, + { + "epoch": 19.74, + "eval_loss": 0.3980773985385895, + "eval_runtime": 7.5335, + "eval_samples_per_second": 13.274, + "eval_wer": 0.3373934226552984, + "step": 121400 + }, + { + "epoch": 19.76, + "learning_rate": 2.1031163321912224e-05, + "loss": NaN, + "step": 121500 + }, + { + "epoch": 19.76, + "eval_loss": 0.5062117576599121, + "eval_runtime": 8.2053, + "eval_samples_per_second": 12.187, + "eval_wer": 0.3471376370280146, + "step": 121500 + }, + { + "epoch": 19.78, + "learning_rate": 2.0965899820525372e-05, + "loss": NaN, + "step": 121600 + }, + { + "epoch": 19.78, + "eval_loss": 0.4986831247806549, + "eval_runtime": 8.3443, + "eval_samples_per_second": 11.984, + "eval_wer": 0.3447015834348356, + "step": 121600 + }, + { + "epoch": 19.79, + "learning_rate": 2.090063631913852e-05, + "loss": NaN, + "step": 121700 + }, + { + "epoch": 19.79, + "eval_loss": 0.5382255911827087, + "eval_runtime": 9.0263, + "eval_samples_per_second": 11.079, + "eval_wer": 0.3568818514007308, + "step": 121700 + }, + { + "epoch": 19.81, + "learning_rate": 2.0835372817751673e-05, + "loss": NaN, + "step": 121800 + }, + { + "epoch": 19.81, + "eval_loss": 0.5431269407272339, + "eval_runtime": 10.0339, + "eval_samples_per_second": 9.966, + "eval_wer": 0.3471376370280146, + "step": 121800 + }, + { + "epoch": 19.82, + "learning_rate": 2.0770109316364825e-05, + "loss": NaN, + "step": 121900 + }, + { + "epoch": 19.82, + "eval_loss": 0.39853399991989136, + "eval_runtime": 8.1763, + "eval_samples_per_second": 12.231, + "eval_wer": 0.3373934226552984, + "step": 121900 + }, + { + "epoch": 19.84, + "learning_rate": 2.0704845814977973e-05, + "loss": NaN, + "step": 122000 + }, + { + "epoch": 19.84, + "eval_loss": 0.563309907913208, + "eval_runtime": 7.9985, + "eval_samples_per_second": 12.502, + "eval_wer": 0.3447015834348356, + "step": 122000 + }, + { + "epoch": 19.86, + "learning_rate": 2.0639582313591125e-05, + "loss": NaN, + "step": 122100 + }, + { + "epoch": 19.86, + "eval_loss": 0.38848280906677246, + "eval_runtime": 7.9406, + "eval_samples_per_second": 12.593, + "eval_wer": 0.33495736906211937, + "step": 122100 + }, + { + "epoch": 19.87, + "learning_rate": 2.0574318812204277e-05, + "loss": NaN, + "step": 122200 + }, + { + "epoch": 19.87, + "eval_loss": 0.3947732448577881, + "eval_runtime": 8.1644, + "eval_samples_per_second": 12.248, + "eval_wer": 0.35444579780755175, + "step": 122200 + }, + { + "epoch": 19.89, + "learning_rate": 2.0509055310817426e-05, + "loss": NaN, + "step": 122300 + }, + { + "epoch": 19.89, + "eval_loss": 0.4206204116344452, + "eval_runtime": 8.6753, + "eval_samples_per_second": 11.527, + "eval_wer": 0.3471376370280146, + "step": 122300 + }, + { + "epoch": 19.91, + "learning_rate": 2.0443791809430577e-05, + "loss": NaN, + "step": 122400 + }, + { + "epoch": 19.91, + "eval_loss": 0.41722992062568665, + "eval_runtime": 7.7422, + "eval_samples_per_second": 12.916, + "eval_wer": 0.341047503045067, + "step": 122400 + }, + { + "epoch": 19.92, + "learning_rate": 2.0378528308043726e-05, + "loss": NaN, + "step": 122500 + }, + { + "epoch": 19.92, + "eval_loss": 0.4375925362110138, + "eval_runtime": 6.9864, + "eval_samples_per_second": 14.313, + "eval_wer": 0.34348355663824603, + "step": 122500 + }, + { + "epoch": 19.94, + "learning_rate": 2.0313264806656878e-05, + "loss": NaN, + "step": 122600 + }, + { + "epoch": 19.94, + "eval_loss": 0.40097764134407043, + "eval_runtime": 6.79, + "eval_samples_per_second": 14.728, + "eval_wer": 0.33617539585870887, + "step": 122600 + }, + { + "epoch": 19.95, + "learning_rate": 2.024800130527003e-05, + "loss": NaN, + "step": 122700 + }, + { + "epoch": 19.95, + "eval_loss": 0.4545166790485382, + "eval_runtime": 7.8384, + "eval_samples_per_second": 12.758, + "eval_wer": 0.3398294762484775, + "step": 122700 + }, + { + "epoch": 19.97, + "learning_rate": 2.0182737803883178e-05, + "loss": NaN, + "step": 122800 + }, + { + "epoch": 19.97, + "eval_loss": 0.3846965730190277, + "eval_runtime": 7.3303, + "eval_samples_per_second": 13.642, + "eval_wer": 0.3264311814859927, + "step": 122800 + }, + { + "epoch": 19.99, + "learning_rate": 2.011747430249633e-05, + "loss": NaN, + "step": 122900 + }, + { + "epoch": 19.99, + "eval_loss": 0.38395482301712036, + "eval_runtime": 7.3923, + "eval_samples_per_second": 13.528, + "eval_wer": 0.3373934226552984, + "step": 122900 + }, + { + "epoch": 20.0, + "learning_rate": 2.0052210801109482e-05, + "loss": NaN, + "step": 123000 + }, + { + "epoch": 20.0, + "eval_loss": 0.43827033042907715, + "eval_runtime": 8.7235, + "eval_samples_per_second": 11.463, + "eval_wer": 0.3507917174177832, + "step": 123000 + }, + { + "epoch": 20.02, + "learning_rate": 1.998694729972263e-05, + "loss": NaN, + "step": 123100 + }, + { + "epoch": 20.02, + "eval_loss": 0.43740010261535645, + "eval_runtime": 8.0253, + "eval_samples_per_second": 12.461, + "eval_wer": 0.34348355663824603, + "step": 123100 + }, + { + "epoch": 20.04, + "learning_rate": 1.9921683798335783e-05, + "loss": NaN, + "step": 123200 + }, + { + "epoch": 20.04, + "eval_loss": 0.3877253830432892, + "eval_runtime": 7.9016, + "eval_samples_per_second": 12.656, + "eval_wer": 0.33008526187576126, + "step": 123200 + }, + { + "epoch": 20.05, + "learning_rate": 1.985642029694893e-05, + "loss": NaN, + "step": 123300 + }, + { + "epoch": 20.05, + "eval_loss": 0.46851950883865356, + "eval_runtime": 8.2945, + "eval_samples_per_second": 12.056, + "eval_wer": 0.3690621193666261, + "step": 123300 + }, + { + "epoch": 20.07, + "learning_rate": 1.9791156795562083e-05, + "loss": NaN, + "step": 123400 + }, + { + "epoch": 20.07, + "eval_loss": 0.5597333312034607, + "eval_runtime": 8.7953, + "eval_samples_per_second": 11.37, + "eval_wer": 0.34835566382460414, + "step": 123400 + }, + { + "epoch": 20.08, + "learning_rate": 1.972589329417523e-05, + "loss": NaN, + "step": 123500 + }, + { + "epoch": 20.08, + "eval_loss": 0.4044766128063202, + "eval_runtime": 8.1152, + "eval_samples_per_second": 12.323, + "eval_wer": 0.33495736906211937, + "step": 123500 + }, + { + "epoch": 20.1, + "learning_rate": 1.9660629792788383e-05, + "loss": NaN, + "step": 123600 + }, + { + "epoch": 20.1, + "eval_loss": 0.39406195282936096, + "eval_runtime": 7.6836, + "eval_samples_per_second": 13.015, + "eval_wer": 0.3325213154689403, + "step": 123600 + }, + { + "epoch": 20.12, + "learning_rate": 1.9595366291401535e-05, + "loss": NaN, + "step": 123700 + }, + { + "epoch": 20.12, + "eval_loss": 0.4256425201892853, + "eval_runtime": 8.9263, + "eval_samples_per_second": 11.203, + "eval_wer": 0.33495736906211937, + "step": 123700 + }, + { + "epoch": 20.13, + "learning_rate": 1.9530102790014684e-05, + "loss": NaN, + "step": 123800 + }, + { + "epoch": 20.13, + "eval_loss": 0.39307454228401184, + "eval_runtime": 7.7688, + "eval_samples_per_second": 12.872, + "eval_wer": 0.3398294762484775, + "step": 123800 + }, + { + "epoch": 20.15, + "learning_rate": 1.9464839288627836e-05, + "loss": NaN, + "step": 123900 + }, + { + "epoch": 20.15, + "eval_loss": 0.4074282944202423, + "eval_runtime": 7.4446, + "eval_samples_per_second": 13.433, + "eval_wer": 0.33617539585870887, + "step": 123900 + }, + { + "epoch": 20.17, + "learning_rate": 1.9399575787240988e-05, + "loss": NaN, + "step": 124000 + }, + { + "epoch": 20.17, + "eval_loss": 0.38621366024017334, + "eval_runtime": 8.5491, + "eval_samples_per_second": 11.697, + "eval_wer": 0.341047503045067, + "step": 124000 + }, + { + "epoch": 20.18, + "learning_rate": 1.9334312285854136e-05, + "loss": NaN, + "step": 124100 + }, + { + "epoch": 20.18, + "eval_loss": 0.3924228549003601, + "eval_runtime": 8.8405, + "eval_samples_per_second": 11.312, + "eval_wer": 0.34348355663824603, + "step": 124100 + }, + { + "epoch": 20.2, + "learning_rate": 1.9269048784467285e-05, + "loss": NaN, + "step": 124200 + }, + { + "epoch": 20.2, + "eval_loss": 0.3702751398086548, + "eval_runtime": 6.8704, + "eval_samples_per_second": 14.555, + "eval_wer": 0.3386114494518879, + "step": 124200 + }, + { + "epoch": 20.21, + "learning_rate": 1.920378528308044e-05, + "loss": NaN, + "step": 124300 + }, + { + "epoch": 20.21, + "eval_loss": 0.4109724760055542, + "eval_runtime": 8.6105, + "eval_samples_per_second": 11.614, + "eval_wer": 0.3568818514007308, + "step": 124300 + }, + { + "epoch": 20.23, + "learning_rate": 1.913852178169359e-05, + "loss": NaN, + "step": 124400 + }, + { + "epoch": 20.23, + "eval_loss": 0.40841957926750183, + "eval_runtime": 7.8666, + "eval_samples_per_second": 12.712, + "eval_wer": 0.3264311814859927, + "step": 124400 + }, + { + "epoch": 20.25, + "learning_rate": 1.9073258280306737e-05, + "loss": NaN, + "step": 124500 + }, + { + "epoch": 20.25, + "eval_loss": 0.44555091857910156, + "eval_runtime": 8.1778, + "eval_samples_per_second": 12.228, + "eval_wer": 0.341047503045067, + "step": 124500 + }, + { + "epoch": 20.26, + "learning_rate": 1.9007994778919892e-05, + "loss": NaN, + "step": 124600 + }, + { + "epoch": 20.26, + "eval_loss": 0.5727394819259644, + "eval_runtime": 8.2574, + "eval_samples_per_second": 12.11, + "eval_wer": 0.3386114494518879, + "step": 124600 + }, + { + "epoch": 20.28, + "learning_rate": 1.894273127753304e-05, + "loss": NaN, + "step": 124700 + }, + { + "epoch": 20.28, + "eval_loss": 0.43890708684921265, + "eval_runtime": 7.5362, + "eval_samples_per_second": 13.269, + "eval_wer": 0.3507917174177832, + "step": 124700 + }, + { + "epoch": 20.3, + "learning_rate": 1.887746777614619e-05, + "loss": NaN, + "step": 124800 + }, + { + "epoch": 20.3, + "eval_loss": 0.4767882525920868, + "eval_runtime": 9.0744, + "eval_samples_per_second": 11.02, + "eval_wer": 0.34226552984165654, + "step": 124800 + }, + { + "epoch": 20.31, + "learning_rate": 1.881220427475934e-05, + "loss": NaN, + "step": 124900 + }, + { + "epoch": 20.31, + "eval_loss": 0.3942064046859741, + "eval_runtime": 9.3212, + "eval_samples_per_second": 10.728, + "eval_wer": 0.3373934226552984, + "step": 124900 + }, + { + "epoch": 20.33, + "learning_rate": 1.8746940773372493e-05, + "loss": NaN, + "step": 125000 + }, + { + "epoch": 20.33, + "eval_loss": 0.4494655132293701, + "eval_runtime": 9.6714, + "eval_samples_per_second": 10.34, + "eval_wer": 0.3447015834348356, + "step": 125000 + }, + { + "epoch": 20.34, + "learning_rate": 1.8681677271985642e-05, + "loss": NaN, + "step": 125100 + }, + { + "epoch": 20.34, + "eval_loss": 0.39226436614990234, + "eval_runtime": 8.7962, + "eval_samples_per_second": 11.369, + "eval_wer": 0.32886723507917176, + "step": 125100 + }, + { + "epoch": 20.36, + "learning_rate": 1.8616413770598794e-05, + "loss": NaN, + "step": 125200 + }, + { + "epoch": 20.36, + "eval_loss": 0.41099902987480164, + "eval_runtime": 8.4883, + "eval_samples_per_second": 11.781, + "eval_wer": 0.3373934226552984, + "step": 125200 + }, + { + "epoch": 20.38, + "learning_rate": 1.8551150269211946e-05, + "loss": NaN, + "step": 125300 + }, + { + "epoch": 20.38, + "eval_loss": 0.4286768436431885, + "eval_runtime": 7.7395, + "eval_samples_per_second": 12.921, + "eval_wer": 0.341047503045067, + "step": 125300 + }, + { + "epoch": 20.39, + "learning_rate": 1.8485886767825094e-05, + "loss": NaN, + "step": 125400 + }, + { + "epoch": 20.39, + "eval_loss": 0.4009086489677429, + "eval_runtime": 7.1964, + "eval_samples_per_second": 13.896, + "eval_wer": 0.33495736906211937, + "step": 125400 + }, + { + "epoch": 20.41, + "learning_rate": 1.8420623266438243e-05, + "loss": NaN, + "step": 125500 + }, + { + "epoch": 20.41, + "eval_loss": 0.40821143984794617, + "eval_runtime": 8.3655, + "eval_samples_per_second": 11.954, + "eval_wer": 0.34226552984165654, + "step": 125500 + }, + { + "epoch": 20.43, + "learning_rate": 1.8355359765051398e-05, + "loss": NaN, + "step": 125600 + }, + { + "epoch": 20.43, + "eval_loss": 0.4229655861854553, + "eval_runtime": 8.8125, + "eval_samples_per_second": 11.348, + "eval_wer": 0.3568818514007308, + "step": 125600 + }, + { + "epoch": 20.44, + "learning_rate": 1.8290096263664547e-05, + "loss": NaN, + "step": 125700 + }, + { + "epoch": 20.44, + "eval_loss": 0.45079144835472107, + "eval_runtime": 9.8702, + "eval_samples_per_second": 10.131, + "eval_wer": 0.34348355663824603, + "step": 125700 + }, + { + "epoch": 20.46, + "learning_rate": 1.8224832762277695e-05, + "loss": NaN, + "step": 125800 + }, + { + "epoch": 20.46, + "eval_loss": 0.42513442039489746, + "eval_runtime": 7.4321, + "eval_samples_per_second": 13.455, + "eval_wer": 0.3313032886723508, + "step": 125800 + }, + { + "epoch": 20.47, + "learning_rate": 1.8159569260890847e-05, + "loss": NaN, + "step": 125900 + }, + { + "epoch": 20.47, + "eval_loss": 0.4858975112438202, + "eval_runtime": 7.8101, + "eval_samples_per_second": 12.804, + "eval_wer": 0.3568818514007308, + "step": 125900 + }, + { + "epoch": 20.49, + "learning_rate": 1.8094305759504e-05, + "loss": NaN, + "step": 126000 + }, + { + "epoch": 20.49, + "eval_loss": 0.4102497398853302, + "eval_runtime": 9.5076, + "eval_samples_per_second": 10.518, + "eval_wer": 0.35444579780755175, + "step": 126000 + }, + { + "epoch": 20.51, + "learning_rate": 1.8029042258117147e-05, + "loss": NaN, + "step": 126100 + }, + { + "epoch": 20.51, + "eval_loss": 0.4684319794178009, + "eval_runtime": 7.5644, + "eval_samples_per_second": 13.22, + "eval_wer": 0.3373934226552984, + "step": 126100 + }, + { + "epoch": 20.52, + "learning_rate": 1.79637787567303e-05, + "loss": NaN, + "step": 126200 + }, + { + "epoch": 20.52, + "eval_loss": 0.6590189933776855, + "eval_runtime": 8.2168, + "eval_samples_per_second": 12.17, + "eval_wer": 0.35931790499390986, + "step": 126200 + }, + { + "epoch": 20.54, + "learning_rate": 1.789851525534345e-05, + "loss": NaN, + "step": 126300 + }, + { + "epoch": 20.54, + "eval_loss": 0.4641081988811493, + "eval_runtime": 7.7606, + "eval_samples_per_second": 12.886, + "eval_wer": 0.315468940316687, + "step": 126300 + }, + { + "epoch": 20.56, + "learning_rate": 1.78332517539566e-05, + "loss": NaN, + "step": 126400 + }, + { + "epoch": 20.56, + "eval_loss": 0.5528498888015747, + "eval_runtime": 9.1648, + "eval_samples_per_second": 10.911, + "eval_wer": 0.3629719853836784, + "step": 126400 + }, + { + "epoch": 20.57, + "learning_rate": 1.7767988252569752e-05, + "loss": NaN, + "step": 126500 + }, + { + "epoch": 20.57, + "eval_loss": 0.433257132768631, + "eval_runtime": 7.6174, + "eval_samples_per_second": 13.128, + "eval_wer": 0.34348355663824603, + "step": 126500 + }, + { + "epoch": 20.59, + "learning_rate": 1.7702724751182904e-05, + "loss": NaN, + "step": 126600 + }, + { + "epoch": 20.59, + "eval_loss": 0.3999103307723999, + "eval_runtime": 7.888, + "eval_samples_per_second": 12.677, + "eval_wer": 0.3373934226552984, + "step": 126600 + }, + { + "epoch": 20.6, + "learning_rate": 1.7637461249796052e-05, + "loss": NaN, + "step": 126700 + }, + { + "epoch": 20.6, + "eval_loss": 0.4195931553840637, + "eval_runtime": 7.5791, + "eval_samples_per_second": 13.194, + "eval_wer": 0.3447015834348356, + "step": 126700 + }, + { + "epoch": 20.62, + "learning_rate": 1.7572197748409204e-05, + "loss": NaN, + "step": 126800 + }, + { + "epoch": 20.62, + "eval_loss": 0.4389355778694153, + "eval_runtime": 6.9812, + "eval_samples_per_second": 14.324, + "eval_wer": 0.34226552984165654, + "step": 126800 + }, + { + "epoch": 20.64, + "learning_rate": 1.7506934247022353e-05, + "loss": NaN, + "step": 126900 + }, + { + "epoch": 20.64, + "eval_loss": 0.4806211590766907, + "eval_runtime": 9.6568, + "eval_samples_per_second": 10.355, + "eval_wer": 0.34957369062119364, + "step": 126900 + }, + { + "epoch": 20.65, + "learning_rate": 1.7441670745635505e-05, + "loss": NaN, + "step": 127000 + }, + { + "epoch": 20.65, + "eval_loss": 0.48015937209129333, + "eval_runtime": 9.1133, + "eval_samples_per_second": 10.973, + "eval_wer": 0.35322777101096225, + "step": 127000 + }, + { + "epoch": 20.67, + "learning_rate": 1.7376407244248653e-05, + "loss": NaN, + "step": 127100 + }, + { + "epoch": 20.67, + "eval_loss": 0.4087853729724884, + "eval_runtime": 7.6535, + "eval_samples_per_second": 13.066, + "eval_wer": 0.3337393422655298, + "step": 127100 + }, + { + "epoch": 20.69, + "learning_rate": 1.7311143742861805e-05, + "loss": NaN, + "step": 127200 + }, + { + "epoch": 20.69, + "eval_loss": 0.43684786558151245, + "eval_runtime": 8.3464, + "eval_samples_per_second": 11.981, + "eval_wer": 0.33495736906211937, + "step": 127200 + }, + { + "epoch": 20.7, + "learning_rate": 1.7245880241474957e-05, + "loss": NaN, + "step": 127300 + }, + { + "epoch": 20.7, + "eval_loss": 0.46926018595695496, + "eval_runtime": 9.6189, + "eval_samples_per_second": 10.396, + "eval_wer": 0.3447015834348356, + "step": 127300 + }, + { + "epoch": 20.72, + "learning_rate": 1.7180616740088105e-05, + "loss": NaN, + "step": 127400 + }, + { + "epoch": 20.72, + "eval_loss": 0.45676276087760925, + "eval_runtime": 9.0225, + "eval_samples_per_second": 11.083, + "eval_wer": 0.3459196102314251, + "step": 127400 + }, + { + "epoch": 20.74, + "learning_rate": 1.7115353238701257e-05, + "loss": NaN, + "step": 127500 + }, + { + "epoch": 20.74, + "eval_loss": 0.4165736734867096, + "eval_runtime": 8.8912, + "eval_samples_per_second": 11.247, + "eval_wer": 0.3459196102314251, + "step": 127500 + }, + { + "epoch": 20.75, + "learning_rate": 1.7050089737314406e-05, + "loss": NaN, + "step": 127600 + }, + { + "epoch": 20.75, + "eval_loss": 0.4191476106643677, + "eval_runtime": 8.1249, + "eval_samples_per_second": 12.308, + "eval_wer": 0.3398294762484775, + "step": 127600 + }, + { + "epoch": 20.77, + "learning_rate": 1.6984826235927558e-05, + "loss": NaN, + "step": 127700 + }, + { + "epoch": 20.77, + "eval_loss": 0.4200890362262726, + "eval_runtime": 7.7999, + "eval_samples_per_second": 12.821, + "eval_wer": 0.3313032886723508, + "step": 127700 + }, + { + "epoch": 20.78, + "learning_rate": 1.691956273454071e-05, + "loss": NaN, + "step": 127800 + }, + { + "epoch": 20.78, + "eval_loss": 0.41486749053001404, + "eval_runtime": 9.2941, + "eval_samples_per_second": 10.76, + "eval_wer": 0.34348355663824603, + "step": 127800 + }, + { + "epoch": 20.8, + "learning_rate": 1.6854299233153858e-05, + "loss": NaN, + "step": 127900 + }, + { + "epoch": 20.8, + "eval_loss": 0.4273871183395386, + "eval_runtime": 9.7275, + "eval_samples_per_second": 10.28, + "eval_wer": 0.3386114494518879, + "step": 127900 + }, + { + "epoch": 20.82, + "learning_rate": 1.678903573176701e-05, + "loss": NaN, + "step": 128000 + }, + { + "epoch": 20.82, + "eval_loss": 0.44087138772010803, + "eval_runtime": 8.9574, + "eval_samples_per_second": 11.164, + "eval_wer": 0.3447015834348356, + "step": 128000 + }, + { + "epoch": 20.83, + "learning_rate": 1.6723772230380162e-05, + "loss": NaN, + "step": 128100 + }, + { + "epoch": 20.83, + "eval_loss": 0.3781212270259857, + "eval_runtime": 7.5604, + "eval_samples_per_second": 13.227, + "eval_wer": 0.3325213154689403, + "step": 128100 + }, + { + "epoch": 20.85, + "learning_rate": 1.665850872899331e-05, + "loss": NaN, + "step": 128200 + }, + { + "epoch": 20.85, + "eval_loss": 0.4753413498401642, + "eval_runtime": 8.6899, + "eval_samples_per_second": 11.508, + "eval_wer": 0.3337393422655298, + "step": 128200 + }, + { + "epoch": 20.87, + "learning_rate": 1.6593245227606463e-05, + "loss": NaN, + "step": 128300 + }, + { + "epoch": 20.87, + "eval_loss": 0.40558329224586487, + "eval_runtime": 7.5404, + "eval_samples_per_second": 13.262, + "eval_wer": 0.3313032886723508, + "step": 128300 + }, + { + "epoch": 20.88, + "learning_rate": 1.6527981726219614e-05, + "loss": NaN, + "step": 128400 + }, + { + "epoch": 20.88, + "eval_loss": 0.4515651762485504, + "eval_runtime": 8.2888, + "eval_samples_per_second": 12.065, + "eval_wer": 0.3398294762484775, + "step": 128400 + }, + { + "epoch": 20.9, + "learning_rate": 1.6462718224832763e-05, + "loss": NaN, + "step": 128500 + }, + { + "epoch": 20.9, + "eval_loss": 0.43877291679382324, + "eval_runtime": 7.2241, + "eval_samples_per_second": 13.843, + "eval_wer": 0.3507917174177832, + "step": 128500 + }, + { + "epoch": 20.91, + "learning_rate": 1.639745472344591e-05, + "loss": NaN, + "step": 128600 + }, + { + "epoch": 20.91, + "eval_loss": 0.457927405834198, + "eval_runtime": 8.7365, + "eval_samples_per_second": 11.446, + "eval_wer": 0.34348355663824603, + "step": 128600 + }, + { + "epoch": 20.93, + "learning_rate": 1.6332191222059063e-05, + "loss": NaN, + "step": 128700 + }, + { + "epoch": 20.93, + "eval_loss": 0.40025395154953003, + "eval_runtime": 8.0467, + "eval_samples_per_second": 12.428, + "eval_wer": 0.34348355663824603, + "step": 128700 + }, + { + "epoch": 20.95, + "learning_rate": 1.6266927720672215e-05, + "loss": NaN, + "step": 128800 + }, + { + "epoch": 20.95, + "eval_loss": 0.5182440280914307, + "eval_runtime": 7.1852, + "eval_samples_per_second": 13.918, + "eval_wer": 0.35444579780755175, + "step": 128800 + }, + { + "epoch": 20.96, + "learning_rate": 1.6201664219285364e-05, + "loss": NaN, + "step": 128900 + }, + { + "epoch": 20.96, + "eval_loss": 0.4123252034187317, + "eval_runtime": 8.1302, + "eval_samples_per_second": 12.3, + "eval_wer": 0.3386114494518879, + "step": 128900 + }, + { + "epoch": 20.98, + "learning_rate": 1.6136400717898516e-05, + "loss": NaN, + "step": 129000 + }, + { + "epoch": 20.98, + "eval_loss": 0.4326474070549011, + "eval_runtime": 7.6986, + "eval_samples_per_second": 12.989, + "eval_wer": 0.3398294762484775, + "step": 129000 + }, + { + "epoch": 21.0, + "learning_rate": 1.6071137216511668e-05, + "loss": NaN, + "step": 129100 + }, + { + "epoch": 21.0, + "eval_loss": 0.4750351011753082, + "eval_runtime": 7.5824, + "eval_samples_per_second": 13.188, + "eval_wer": 0.3373934226552984, + "step": 129100 + }, + { + "epoch": 21.01, + "learning_rate": 1.6005873715124816e-05, + "loss": NaN, + "step": 129200 + }, + { + "epoch": 21.01, + "eval_loss": 0.4125843942165375, + "eval_runtime": 8.0836, + "eval_samples_per_second": 12.371, + "eval_wer": 0.3386114494518879, + "step": 129200 + }, + { + "epoch": 21.03, + "learning_rate": 1.5940610213737968e-05, + "loss": NaN, + "step": 129300 + }, + { + "epoch": 21.03, + "eval_loss": 0.4247395396232605, + "eval_runtime": 8.0802, + "eval_samples_per_second": 12.376, + "eval_wer": 0.3373934226552984, + "step": 129300 + }, + { + "epoch": 21.04, + "learning_rate": 1.587534671235112e-05, + "loss": NaN, + "step": 129400 + }, + { + "epoch": 21.04, + "eval_loss": 0.43337181210517883, + "eval_runtime": 7.6212, + "eval_samples_per_second": 13.121, + "eval_wer": 0.32886723507917176, + "step": 129400 + }, + { + "epoch": 21.06, + "learning_rate": 1.581008321096427e-05, + "loss": NaN, + "step": 129500 + }, + { + "epoch": 21.06, + "eval_loss": 0.35774993896484375, + "eval_runtime": 8.103, + "eval_samples_per_second": 12.341, + "eval_wer": 0.315468940316687, + "step": 129500 + }, + { + "epoch": 21.08, + "learning_rate": 1.5744819709577417e-05, + "loss": NaN, + "step": 129600 + }, + { + "epoch": 21.08, + "eval_loss": 0.4800620973110199, + "eval_runtime": 8.5222, + "eval_samples_per_second": 11.734, + "eval_wer": 0.32521315468940315, + "step": 129600 + }, + { + "epoch": 21.09, + "learning_rate": 1.5679556208190572e-05, + "loss": NaN, + "step": 129700 + }, + { + "epoch": 21.09, + "eval_loss": 0.4503170847892761, + "eval_runtime": 8.4578, + "eval_samples_per_second": 11.823, + "eval_wer": 0.34226552984165654, + "step": 129700 + }, + { + "epoch": 21.11, + "learning_rate": 1.561429270680372e-05, + "loss": NaN, + "step": 129800 + }, + { + "epoch": 21.11, + "eval_loss": 0.4079561233520508, + "eval_runtime": 7.3872, + "eval_samples_per_second": 13.537, + "eval_wer": 0.3325213154689403, + "step": 129800 + }, + { + "epoch": 21.13, + "learning_rate": 1.554902920541687e-05, + "loss": NaN, + "step": 129900 + }, + { + "epoch": 21.13, + "eval_loss": 0.44905129075050354, + "eval_runtime": 7.1484, + "eval_samples_per_second": 13.989, + "eval_wer": 0.3398294762484775, + "step": 129900 + }, + { + "epoch": 21.14, + "learning_rate": 1.5483765704030025e-05, + "loss": NaN, + "step": 130000 + }, + { + "epoch": 21.14, + "eval_loss": 0.4295308291912079, + "eval_runtime": 8.0526, + "eval_samples_per_second": 12.418, + "eval_wer": 0.33617539585870887, + "step": 130000 + }, + { + "epoch": 21.16, + "learning_rate": 1.5418502202643173e-05, + "loss": NaN, + "step": 130100 + }, + { + "epoch": 21.16, + "eval_loss": 0.3925670385360718, + "eval_runtime": 8.4253, + "eval_samples_per_second": 11.869, + "eval_wer": 0.3325213154689403, + "step": 130100 + }, + { + "epoch": 21.17, + "learning_rate": 1.5353238701256322e-05, + "loss": NaN, + "step": 130200 + }, + { + "epoch": 21.17, + "eval_loss": 0.46943891048431396, + "eval_runtime": 6.9244, + "eval_samples_per_second": 14.442, + "eval_wer": 0.34957369062119364, + "step": 130200 + }, + { + "epoch": 21.19, + "learning_rate": 1.5287975199869474e-05, + "loss": NaN, + "step": 130300 + }, + { + "epoch": 21.19, + "eval_loss": 0.3773401975631714, + "eval_runtime": 9.0744, + "eval_samples_per_second": 11.02, + "eval_wer": 0.3471376370280146, + "step": 130300 + }, + { + "epoch": 21.21, + "learning_rate": 1.5222711698482626e-05, + "loss": NaN, + "step": 130400 + }, + { + "epoch": 21.21, + "eval_loss": 0.4215690493583679, + "eval_runtime": 8.1736, + "eval_samples_per_second": 12.235, + "eval_wer": 0.341047503045067, + "step": 130400 + }, + { + "epoch": 21.22, + "learning_rate": 1.5157448197095774e-05, + "loss": NaN, + "step": 130500 + }, + { + "epoch": 21.22, + "eval_loss": 0.4928508400917053, + "eval_runtime": 8.1815, + "eval_samples_per_second": 12.223, + "eval_wer": 0.3459196102314251, + "step": 130500 + }, + { + "epoch": 21.24, + "learning_rate": 1.5092184695708924e-05, + "loss": NaN, + "step": 130600 + }, + { + "epoch": 21.24, + "eval_loss": 0.4258655309677124, + "eval_runtime": 8.8623, + "eval_samples_per_second": 11.284, + "eval_wer": 0.33617539585870887, + "step": 130600 + }, + { + "epoch": 21.26, + "learning_rate": 1.5026921194322078e-05, + "loss": NaN, + "step": 130700 + }, + { + "epoch": 21.26, + "eval_loss": 0.40194839239120483, + "eval_runtime": 8.8628, + "eval_samples_per_second": 11.283, + "eval_wer": 0.3276492082825822, + "step": 130700 + }, + { + "epoch": 21.27, + "learning_rate": 1.4961657692935227e-05, + "loss": NaN, + "step": 130800 + }, + { + "epoch": 21.27, + "eval_loss": 0.4380399286746979, + "eval_runtime": 8.0782, + "eval_samples_per_second": 12.379, + "eval_wer": 0.34835566382460414, + "step": 130800 + }, + { + "epoch": 21.29, + "learning_rate": 1.4896394191548377e-05, + "loss": NaN, + "step": 130900 + }, + { + "epoch": 21.29, + "eval_loss": 0.40339186787605286, + "eval_runtime": 9.2102, + "eval_samples_per_second": 10.858, + "eval_wer": 0.34226552984165654, + "step": 130900 + }, + { + "epoch": 21.3, + "learning_rate": 1.4831130690161527e-05, + "loss": NaN, + "step": 131000 + }, + { + "epoch": 21.3, + "eval_loss": 0.4713993966579437, + "eval_runtime": 8.5421, + "eval_samples_per_second": 11.707, + "eval_wer": 0.3507917174177832, + "step": 131000 + }, + { + "epoch": 21.32, + "learning_rate": 1.4765867188774679e-05, + "loss": NaN, + "step": 131100 + }, + { + "epoch": 21.32, + "eval_loss": 0.5234803557395935, + "eval_runtime": 7.9161, + "eval_samples_per_second": 12.632, + "eval_wer": 0.3568818514007308, + "step": 131100 + }, + { + "epoch": 21.34, + "learning_rate": 1.4700603687387829e-05, + "loss": NaN, + "step": 131200 + }, + { + "epoch": 21.34, + "eval_loss": 0.46438494324684143, + "eval_runtime": 9.1223, + "eval_samples_per_second": 10.962, + "eval_wer": 0.3520097442143727, + "step": 131200 + }, + { + "epoch": 21.35, + "learning_rate": 1.463534018600098e-05, + "loss": NaN, + "step": 131300 + }, + { + "epoch": 21.35, + "eval_loss": 0.44316136837005615, + "eval_runtime": 9.5471, + "eval_samples_per_second": 10.474, + "eval_wer": 0.3373934226552984, + "step": 131300 + }, + { + "epoch": 21.37, + "learning_rate": 1.4570076684614131e-05, + "loss": NaN, + "step": 131400 + }, + { + "epoch": 21.37, + "eval_loss": 0.442475825548172, + "eval_runtime": 7.1952, + "eval_samples_per_second": 13.898, + "eval_wer": 0.33617539585870887, + "step": 131400 + }, + { + "epoch": 21.39, + "learning_rate": 1.4504813183227281e-05, + "loss": NaN, + "step": 131500 + }, + { + "epoch": 21.39, + "eval_loss": 0.5654551982879639, + "eval_runtime": 8.1407, + "eval_samples_per_second": 12.284, + "eval_wer": 0.35322777101096225, + "step": 131500 + }, + { + "epoch": 21.4, + "learning_rate": 1.443954968184043e-05, + "loss": NaN, + "step": 131600 + }, + { + "epoch": 21.4, + "eval_loss": 0.4513855278491974, + "eval_runtime": 8.8967, + "eval_samples_per_second": 11.24, + "eval_wer": 0.34835566382460414, + "step": 131600 + }, + { + "epoch": 21.42, + "learning_rate": 1.4374286180453584e-05, + "loss": NaN, + "step": 131700 + }, + { + "epoch": 21.42, + "eval_loss": 0.4416217803955078, + "eval_runtime": 9.1103, + "eval_samples_per_second": 10.977, + "eval_wer": 0.34835566382460414, + "step": 131700 + }, + { + "epoch": 21.43, + "learning_rate": 1.4309022679066734e-05, + "loss": NaN, + "step": 131800 + }, + { + "epoch": 21.43, + "eval_loss": 0.391195148229599, + "eval_runtime": 8.6493, + "eval_samples_per_second": 11.562, + "eval_wer": 0.3386114494518879, + "step": 131800 + }, + { + "epoch": 21.45, + "learning_rate": 1.4243759177679882e-05, + "loss": NaN, + "step": 131900 + }, + { + "epoch": 21.45, + "eval_loss": 0.4384614825248718, + "eval_runtime": 8.481, + "eval_samples_per_second": 11.791, + "eval_wer": 0.3447015834348356, + "step": 131900 + }, + { + "epoch": 21.47, + "learning_rate": 1.4178495676293033e-05, + "loss": NaN, + "step": 132000 + }, + { + "epoch": 21.47, + "eval_loss": 0.42818814516067505, + "eval_runtime": 8.6608, + "eval_samples_per_second": 11.546, + "eval_wer": 0.3386114494518879, + "step": 132000 + }, + { + "epoch": 21.48, + "learning_rate": 1.4113232174906185e-05, + "loss": NaN, + "step": 132100 + }, + { + "epoch": 21.48, + "eval_loss": 0.4142271876335144, + "eval_runtime": 7.784, + "eval_samples_per_second": 12.847, + "eval_wer": 0.32521315468940315, + "step": 132100 + }, + { + "epoch": 21.5, + "learning_rate": 1.4047968673519335e-05, + "loss": NaN, + "step": 132200 + }, + { + "epoch": 21.5, + "eval_loss": 0.42976322770118713, + "eval_runtime": 11.0885, + "eval_samples_per_second": 9.018, + "eval_wer": 0.35809987819732036, + "step": 132200 + }, + { + "epoch": 21.52, + "learning_rate": 1.3982705172132485e-05, + "loss": NaN, + "step": 132300 + }, + { + "epoch": 21.52, + "eval_loss": 0.44189947843551636, + "eval_runtime": 8.1376, + "eval_samples_per_second": 12.289, + "eval_wer": 0.3447015834348356, + "step": 132300 + }, + { + "epoch": 21.53, + "learning_rate": 1.3917441670745637e-05, + "loss": NaN, + "step": 132400 + }, + { + "epoch": 21.53, + "eval_loss": 0.4482198655605316, + "eval_runtime": 8.3756, + "eval_samples_per_second": 11.939, + "eval_wer": 0.3507917174177832, + "step": 132400 + }, + { + "epoch": 21.55, + "learning_rate": 1.3852178169358787e-05, + "loss": NaN, + "step": 132500 + }, + { + "epoch": 21.55, + "eval_loss": 0.37959039211273193, + "eval_runtime": 7.1877, + "eval_samples_per_second": 13.913, + "eval_wer": 0.3337393422655298, + "step": 132500 + }, + { + "epoch": 21.56, + "learning_rate": 1.3786914667971937e-05, + "loss": NaN, + "step": 132600 + }, + { + "epoch": 21.56, + "eval_loss": 0.4015367925167084, + "eval_runtime": 7.0749, + "eval_samples_per_second": 14.134, + "eval_wer": 0.3373934226552984, + "step": 132600 + }, + { + "epoch": 21.58, + "learning_rate": 1.3721651166585086e-05, + "loss": NaN, + "step": 132700 + }, + { + "epoch": 21.58, + "eval_loss": 0.4340609014034271, + "eval_runtime": 10.0027, + "eval_samples_per_second": 9.997, + "eval_wer": 0.34835566382460414, + "step": 132700 + }, + { + "epoch": 21.6, + "learning_rate": 1.365638766519824e-05, + "loss": NaN, + "step": 132800 + }, + { + "epoch": 21.6, + "eval_loss": 0.5372726917266846, + "eval_runtime": 8.4121, + "eval_samples_per_second": 11.888, + "eval_wer": 0.33617539585870887, + "step": 132800 + }, + { + "epoch": 21.61, + "learning_rate": 1.359112416381139e-05, + "loss": NaN, + "step": 132900 + }, + { + "epoch": 21.61, + "eval_loss": 0.39388373494148254, + "eval_runtime": 8.9479, + "eval_samples_per_second": 11.176, + "eval_wer": 0.3264311814859927, + "step": 132900 + }, + { + "epoch": 21.63, + "learning_rate": 1.3525860662424538e-05, + "loss": NaN, + "step": 133000 + }, + { + "epoch": 21.63, + "eval_loss": 0.44166165590286255, + "eval_runtime": 8.2657, + "eval_samples_per_second": 12.098, + "eval_wer": 0.3471376370280146, + "step": 133000 + }, + { + "epoch": 21.65, + "learning_rate": 1.3460597161037692e-05, + "loss": NaN, + "step": 133100 + }, + { + "epoch": 21.65, + "eval_loss": 0.41540205478668213, + "eval_runtime": 8.1998, + "eval_samples_per_second": 12.195, + "eval_wer": 0.34348355663824603, + "step": 133100 + }, + { + "epoch": 21.66, + "learning_rate": 1.339533365965084e-05, + "loss": NaN, + "step": 133200 + }, + { + "epoch": 21.66, + "eval_loss": 0.4051262140274048, + "eval_runtime": 7.0885, + "eval_samples_per_second": 14.107, + "eval_wer": 0.33617539585870887, + "step": 133200 + }, + { + "epoch": 21.68, + "learning_rate": 1.333007015826399e-05, + "loss": NaN, + "step": 133300 + }, + { + "epoch": 21.68, + "eval_loss": 0.41227516531944275, + "eval_runtime": 8.2062, + "eval_samples_per_second": 12.186, + "eval_wer": 0.3459196102314251, + "step": 133300 + }, + { + "epoch": 21.69, + "learning_rate": 1.3264806656877144e-05, + "loss": NaN, + "step": 133400 + }, + { + "epoch": 21.69, + "eval_loss": 0.42322754859924316, + "eval_runtime": 9.0314, + "eval_samples_per_second": 11.073, + "eval_wer": 0.3398294762484775, + "step": 133400 + }, + { + "epoch": 21.71, + "learning_rate": 1.3199543155490293e-05, + "loss": NaN, + "step": 133500 + }, + { + "epoch": 21.71, + "eval_loss": 0.3802301287651062, + "eval_runtime": 8.7387, + "eval_samples_per_second": 11.443, + "eval_wer": 0.3325213154689403, + "step": 133500 + }, + { + "epoch": 21.73, + "learning_rate": 1.3134279654103443e-05, + "loss": NaN, + "step": 133600 + }, + { + "epoch": 21.73, + "eval_loss": 0.40790998935699463, + "eval_runtime": 7.0036, + "eval_samples_per_second": 14.278, + "eval_wer": 0.3398294762484775, + "step": 133600 + }, + { + "epoch": 21.74, + "learning_rate": 1.3069016152716593e-05, + "loss": NaN, + "step": 133700 + }, + { + "epoch": 21.74, + "eval_loss": 0.41051357984542847, + "eval_runtime": 8.2179, + "eval_samples_per_second": 12.168, + "eval_wer": 0.341047503045067, + "step": 133700 + }, + { + "epoch": 21.76, + "learning_rate": 1.3003752651329745e-05, + "loss": NaN, + "step": 133800 + }, + { + "epoch": 21.76, + "eval_loss": 0.40564852952957153, + "eval_runtime": 8.5003, + "eval_samples_per_second": 11.764, + "eval_wer": 0.3325213154689403, + "step": 133800 + }, + { + "epoch": 21.78, + "learning_rate": 1.2938489149942895e-05, + "loss": NaN, + "step": 133900 + }, + { + "epoch": 21.78, + "eval_loss": 0.4392382800579071, + "eval_runtime": 7.6928, + "eval_samples_per_second": 12.999, + "eval_wer": 0.34835566382460414, + "step": 133900 + }, + { + "epoch": 21.79, + "learning_rate": 1.2873225648556045e-05, + "loss": NaN, + "step": 134000 + }, + { + "epoch": 21.79, + "eval_loss": 0.3991428017616272, + "eval_runtime": 9.3346, + "eval_samples_per_second": 10.713, + "eval_wer": 0.3386114494518879, + "step": 134000 + }, + { + "epoch": 21.81, + "learning_rate": 1.2807962147169197e-05, + "loss": NaN, + "step": 134100 + }, + { + "epoch": 21.81, + "eval_loss": 0.4022715389728546, + "eval_runtime": 7.981, + "eval_samples_per_second": 12.53, + "eval_wer": 0.34226552984165654, + "step": 134100 + }, + { + "epoch": 21.82, + "learning_rate": 1.2742698645782348e-05, + "loss": NaN, + "step": 134200 + }, + { + "epoch": 21.82, + "eval_loss": 0.4230504631996155, + "eval_runtime": 8.7343, + "eval_samples_per_second": 11.449, + "eval_wer": 0.341047503045067, + "step": 134200 + }, + { + "epoch": 21.84, + "learning_rate": 1.2677435144395496e-05, + "loss": NaN, + "step": 134300 + }, + { + "epoch": 21.84, + "eval_loss": 0.45141318440437317, + "eval_runtime": 9.2929, + "eval_samples_per_second": 10.761, + "eval_wer": 0.3447015834348356, + "step": 134300 + }, + { + "epoch": 21.86, + "learning_rate": 1.2612171643008646e-05, + "loss": NaN, + "step": 134400 + }, + { + "epoch": 21.86, + "eval_loss": 0.42951327562332153, + "eval_runtime": 9.4873, + "eval_samples_per_second": 10.54, + "eval_wer": 0.31303288672350793, + "step": 134400 + }, + { + "epoch": 21.87, + "learning_rate": 1.25469081416218e-05, + "loss": NaN, + "step": 134500 + }, + { + "epoch": 21.87, + "eval_loss": 0.46236321330070496, + "eval_runtime": 10.0535, + "eval_samples_per_second": 9.947, + "eval_wer": 0.3337393422655298, + "step": 134500 + }, + { + "epoch": 21.89, + "learning_rate": 1.2481644640234949e-05, + "loss": NaN, + "step": 134600 + }, + { + "epoch": 21.89, + "eval_loss": 0.49196091294288635, + "eval_runtime": 8.3873, + "eval_samples_per_second": 11.923, + "eval_wer": 0.33495736906211937, + "step": 134600 + }, + { + "epoch": 21.91, + "learning_rate": 1.24163811388481e-05, + "loss": NaN, + "step": 134700 + }, + { + "epoch": 21.91, + "eval_loss": 0.4672481417655945, + "eval_runtime": 9.8356, + "eval_samples_per_second": 10.167, + "eval_wer": 0.35444579780755175, + "step": 134700 + }, + { + "epoch": 21.92, + "learning_rate": 1.235111763746125e-05, + "loss": NaN, + "step": 134800 + }, + { + "epoch": 21.92, + "eval_loss": 0.4398852586746216, + "eval_runtime": 7.9762, + "eval_samples_per_second": 12.537, + "eval_wer": 0.35931790499390986, + "step": 134800 + }, + { + "epoch": 21.94, + "learning_rate": 1.2285854136074401e-05, + "loss": NaN, + "step": 134900 + }, + { + "epoch": 21.94, + "eval_loss": 0.40451163053512573, + "eval_runtime": 7.6202, + "eval_samples_per_second": 13.123, + "eval_wer": 0.3337393422655298, + "step": 134900 + }, + { + "epoch": 21.95, + "learning_rate": 1.2220590634687551e-05, + "loss": NaN, + "step": 135000 + }, + { + "epoch": 21.95, + "eval_loss": 0.41924548149108887, + "eval_runtime": 9.3656, + "eval_samples_per_second": 10.677, + "eval_wer": 0.34226552984165654, + "step": 135000 + }, + { + "epoch": 21.97, + "learning_rate": 1.2155327133300701e-05, + "loss": NaN, + "step": 135100 + }, + { + "epoch": 21.97, + "eval_loss": 0.42516928911209106, + "eval_runtime": 8.4901, + "eval_samples_per_second": 11.778, + "eval_wer": 0.3447015834348356, + "step": 135100 + }, + { + "epoch": 21.99, + "learning_rate": 1.2090063631913853e-05, + "loss": NaN, + "step": 135200 + }, + { + "epoch": 21.99, + "eval_loss": 0.40347903966903687, + "eval_runtime": 8.9415, + "eval_samples_per_second": 11.184, + "eval_wer": 0.33008526187576126, + "step": 135200 + }, + { + "epoch": 22.0, + "learning_rate": 1.2024800130527003e-05, + "loss": NaN, + "step": 135300 + }, + { + "epoch": 22.0, + "eval_loss": 0.5743076801300049, + "eval_runtime": 8.6933, + "eval_samples_per_second": 11.503, + "eval_wer": 0.3373934226552984, + "step": 135300 + }, + { + "epoch": 22.02, + "learning_rate": 1.1959536629140154e-05, + "loss": NaN, + "step": 135400 + }, + { + "epoch": 22.02, + "eval_loss": 0.40245911478996277, + "eval_runtime": 8.8702, + "eval_samples_per_second": 11.274, + "eval_wer": 0.32886723507917176, + "step": 135400 + }, + { + "epoch": 22.04, + "learning_rate": 1.1894273127753304e-05, + "loss": NaN, + "step": 135500 + }, + { + "epoch": 22.04, + "eval_loss": 0.4054236114025116, + "eval_runtime": 8.0844, + "eval_samples_per_second": 12.369, + "eval_wer": 0.33617539585870887, + "step": 135500 + }, + { + "epoch": 22.05, + "learning_rate": 1.1829009626366456e-05, + "loss": NaN, + "step": 135600 + }, + { + "epoch": 22.05, + "eval_loss": 0.40440499782562256, + "eval_runtime": 8.0393, + "eval_samples_per_second": 12.439, + "eval_wer": 0.33495736906211937, + "step": 135600 + }, + { + "epoch": 22.07, + "learning_rate": 1.1763746124979606e-05, + "loss": NaN, + "step": 135700 + }, + { + "epoch": 22.07, + "eval_loss": 0.41659045219421387, + "eval_runtime": 7.7624, + "eval_samples_per_second": 12.883, + "eval_wer": 0.35322777101096225, + "step": 135700 + }, + { + "epoch": 22.08, + "learning_rate": 1.1698482623592756e-05, + "loss": NaN, + "step": 135800 + }, + { + "epoch": 22.08, + "eval_loss": 0.47099122405052185, + "eval_runtime": 6.8327, + "eval_samples_per_second": 14.635, + "eval_wer": 0.341047503045067, + "step": 135800 + }, + { + "epoch": 22.1, + "learning_rate": 1.1633219122205906e-05, + "loss": NaN, + "step": 135900 + }, + { + "epoch": 22.1, + "eval_loss": 0.4006298780441284, + "eval_runtime": 6.5785, + "eval_samples_per_second": 15.201, + "eval_wer": 0.33008526187576126, + "step": 135900 + }, + { + "epoch": 22.12, + "learning_rate": 1.1567955620819057e-05, + "loss": NaN, + "step": 136000 + }, + { + "epoch": 22.12, + "eval_loss": 0.47020214796066284, + "eval_runtime": 8.3159, + "eval_samples_per_second": 12.025, + "eval_wer": 0.3471376370280146, + "step": 136000 + }, + { + "epoch": 22.13, + "learning_rate": 1.1502692119432209e-05, + "loss": NaN, + "step": 136100 + }, + { + "epoch": 22.13, + "eval_loss": 0.43791496753692627, + "eval_runtime": 8.3887, + "eval_samples_per_second": 11.921, + "eval_wer": 0.3471376370280146, + "step": 136100 + }, + { + "epoch": 22.15, + "learning_rate": 1.1437428618045359e-05, + "loss": NaN, + "step": 136200 + }, + { + "epoch": 22.15, + "eval_loss": 0.4248058795928955, + "eval_runtime": 10.0383, + "eval_samples_per_second": 9.962, + "eval_wer": 0.3447015834348356, + "step": 136200 + }, + { + "epoch": 22.17, + "learning_rate": 1.1372165116658509e-05, + "loss": NaN, + "step": 136300 + }, + { + "epoch": 22.17, + "eval_loss": 0.40560001134872437, + "eval_runtime": 7.8536, + "eval_samples_per_second": 12.733, + "eval_wer": 0.3325213154689403, + "step": 136300 + }, + { + "epoch": 22.18, + "learning_rate": 1.1306901615271661e-05, + "loss": NaN, + "step": 136400 + }, + { + "epoch": 22.18, + "eval_loss": 0.41141241788864136, + "eval_runtime": 7.712, + "eval_samples_per_second": 12.967, + "eval_wer": 0.33617539585870887, + "step": 136400 + }, + { + "epoch": 22.2, + "learning_rate": 1.124163811388481e-05, + "loss": NaN, + "step": 136500 + }, + { + "epoch": 22.2, + "eval_loss": 0.37809231877326965, + "eval_runtime": 6.7786, + "eval_samples_per_second": 14.752, + "eval_wer": 0.3398294762484775, + "step": 136500 + }, + { + "epoch": 22.21, + "learning_rate": 1.1176374612497961e-05, + "loss": NaN, + "step": 136600 + }, + { + "epoch": 22.21, + "eval_loss": 0.48994365334510803, + "eval_runtime": 8.9346, + "eval_samples_per_second": 11.192, + "eval_wer": 0.3471376370280146, + "step": 136600 + }, + { + "epoch": 22.23, + "learning_rate": 1.1111111111111112e-05, + "loss": NaN, + "step": 136700 + }, + { + "epoch": 22.23, + "eval_loss": 0.43884292244911194, + "eval_runtime": 9.9578, + "eval_samples_per_second": 10.042, + "eval_wer": 0.3520097442143727, + "step": 136700 + }, + { + "epoch": 22.25, + "learning_rate": 1.1045847609724262e-05, + "loss": NaN, + "step": 136800 + }, + { + "epoch": 22.25, + "eval_loss": 0.4258538782596588, + "eval_runtime": 7.3965, + "eval_samples_per_second": 13.52, + "eval_wer": 0.3459196102314251, + "step": 136800 + }, + { + "epoch": 22.26, + "learning_rate": 1.0980584108337414e-05, + "loss": NaN, + "step": 136900 + }, + { + "epoch": 22.26, + "eval_loss": 0.44286561012268066, + "eval_runtime": 8.7957, + "eval_samples_per_second": 11.369, + "eval_wer": 0.3520097442143727, + "step": 136900 + }, + { + "epoch": 22.28, + "learning_rate": 1.0915320606950562e-05, + "loss": NaN, + "step": 137000 + }, + { + "epoch": 22.28, + "eval_loss": 0.41971367597579956, + "eval_runtime": 9.2318, + "eval_samples_per_second": 10.832, + "eval_wer": 0.35809987819732036, + "step": 137000 + }, + { + "epoch": 22.3, + "learning_rate": 1.0850057105563714e-05, + "loss": NaN, + "step": 137100 + }, + { + "epoch": 22.3, + "eval_loss": 0.45038101077079773, + "eval_runtime": 7.8919, + "eval_samples_per_second": 12.671, + "eval_wer": 0.3447015834348356, + "step": 137100 + }, + { + "epoch": 22.31, + "learning_rate": 1.0784793604176864e-05, + "loss": NaN, + "step": 137200 + }, + { + "epoch": 22.31, + "eval_loss": 0.417258620262146, + "eval_runtime": 9.4776, + "eval_samples_per_second": 10.551, + "eval_wer": 0.3459196102314251, + "step": 137200 + }, + { + "epoch": 22.33, + "learning_rate": 1.0719530102790015e-05, + "loss": NaN, + "step": 137300 + }, + { + "epoch": 22.33, + "eval_loss": 0.40922966599464417, + "eval_runtime": 7.7162, + "eval_samples_per_second": 12.96, + "eval_wer": 0.33008526187576126, + "step": 137300 + }, + { + "epoch": 22.35, + "learning_rate": 1.0654266601403167e-05, + "loss": NaN, + "step": 137400 + }, + { + "epoch": 22.35, + "eval_loss": 0.42340758442878723, + "eval_runtime": 7.7407, + "eval_samples_per_second": 12.919, + "eval_wer": 0.3459196102314251, + "step": 137400 + }, + { + "epoch": 22.36, + "learning_rate": 1.0589003100016317e-05, + "loss": NaN, + "step": 137500 + }, + { + "epoch": 22.36, + "eval_loss": 0.43829768896102905, + "eval_runtime": 9.0473, + "eval_samples_per_second": 11.053, + "eval_wer": 0.3471376370280146, + "step": 137500 + }, + { + "epoch": 22.38, + "learning_rate": 1.0523739598629467e-05, + "loss": NaN, + "step": 137600 + }, + { + "epoch": 22.38, + "eval_loss": 0.4190714955329895, + "eval_runtime": 7.2713, + "eval_samples_per_second": 13.753, + "eval_wer": 0.3373934226552984, + "step": 137600 + }, + { + "epoch": 22.39, + "learning_rate": 1.0458476097242617e-05, + "loss": NaN, + "step": 137700 + }, + { + "epoch": 22.39, + "eval_loss": 0.38507279753685, + "eval_runtime": 8.2207, + "eval_samples_per_second": 12.164, + "eval_wer": 0.33617539585870887, + "step": 137700 + }, + { + "epoch": 22.41, + "learning_rate": 1.0393212595855767e-05, + "loss": NaN, + "step": 137800 + }, + { + "epoch": 22.41, + "eval_loss": 0.3674430847167969, + "eval_runtime": 9.7115, + "eval_samples_per_second": 10.297, + "eval_wer": 0.3313032886723508, + "step": 137800 + }, + { + "epoch": 22.43, + "learning_rate": 1.0327949094468918e-05, + "loss": NaN, + "step": 137900 + }, + { + "epoch": 22.43, + "eval_loss": 0.40191754698753357, + "eval_runtime": 7.543, + "eval_samples_per_second": 13.257, + "eval_wer": 0.33008526187576126, + "step": 137900 + }, + { + "epoch": 22.44, + "learning_rate": 1.026268559308207e-05, + "loss": NaN, + "step": 138000 + }, + { + "epoch": 22.44, + "eval_loss": 0.4558006525039673, + "eval_runtime": 8.3592, + "eval_samples_per_second": 11.963, + "eval_wer": 0.3507917174177832, + "step": 138000 + }, + { + "epoch": 22.46, + "learning_rate": 1.019742209169522e-05, + "loss": NaN, + "step": 138100 + }, + { + "epoch": 22.46, + "eval_loss": 0.5849547386169434, + "eval_runtime": 8.4615, + "eval_samples_per_second": 11.818, + "eval_wer": 0.33495736906211937, + "step": 138100 + }, + { + "epoch": 22.48, + "learning_rate": 1.013215859030837e-05, + "loss": NaN, + "step": 138200 + }, + { + "epoch": 22.48, + "eval_loss": 0.45433205366134644, + "eval_runtime": 8.487, + "eval_samples_per_second": 11.783, + "eval_wer": 0.33008526187576126, + "step": 138200 + }, + { + "epoch": 22.49, + "learning_rate": 1.0066895088921522e-05, + "loss": NaN, + "step": 138300 + }, + { + "epoch": 22.49, + "eval_loss": 0.420282781124115, + "eval_runtime": 9.2501, + "eval_samples_per_second": 10.811, + "eval_wer": 0.33617539585870887, + "step": 138300 + }, + { + "epoch": 22.51, + "learning_rate": 1.000163158753467e-05, + "loss": NaN, + "step": 138400 + }, + { + "epoch": 22.51, + "eval_loss": 0.4292939305305481, + "eval_runtime": 8.0963, + "eval_samples_per_second": 12.351, + "eval_wer": 0.3337393422655298, + "step": 138400 + }, + { + "epoch": 22.52, + "learning_rate": 9.936368086147822e-06, + "loss": NaN, + "step": 138500 + }, + { + "epoch": 22.52, + "eval_loss": 0.4376954138278961, + "eval_runtime": 8.5678, + "eval_samples_per_second": 11.672, + "eval_wer": 0.3093788063337393, + "step": 138500 + }, + { + "epoch": 22.54, + "learning_rate": 9.871104584760973e-06, + "loss": NaN, + "step": 138600 + }, + { + "epoch": 22.54, + "eval_loss": 0.4139672815799713, + "eval_runtime": 8.398, + "eval_samples_per_second": 11.908, + "eval_wer": 0.3313032886723508, + "step": 138600 + }, + { + "epoch": 22.56, + "learning_rate": 9.805841083374123e-06, + "loss": NaN, + "step": 138700 + }, + { + "epoch": 22.56, + "eval_loss": 0.4242633581161499, + "eval_runtime": 10.5113, + "eval_samples_per_second": 9.514, + "eval_wer": 0.34835566382460414, + "step": 138700 + }, + { + "epoch": 22.57, + "learning_rate": 9.740577581987275e-06, + "loss": NaN, + "step": 138800 + }, + { + "epoch": 22.57, + "eval_loss": 0.44767504930496216, + "eval_runtime": 7.6566, + "eval_samples_per_second": 13.061, + "eval_wer": 0.3386114494518879, + "step": 138800 + }, + { + "epoch": 22.59, + "learning_rate": 9.675314080600425e-06, + "loss": NaN, + "step": 138900 + }, + { + "epoch": 22.59, + "eval_loss": 0.43390411138534546, + "eval_runtime": 8.5202, + "eval_samples_per_second": 11.737, + "eval_wer": 0.341047503045067, + "step": 138900 + }, + { + "epoch": 22.61, + "learning_rate": 9.610050579213575e-06, + "loss": NaN, + "step": 139000 + }, + { + "epoch": 22.61, + "eval_loss": 0.42713436484336853, + "eval_runtime": 7.4028, + "eval_samples_per_second": 13.508, + "eval_wer": 0.32521315468940315, + "step": 139000 + }, + { + "epoch": 22.62, + "learning_rate": 9.544787077826727e-06, + "loss": NaN, + "step": 139100 + }, + { + "epoch": 22.62, + "eval_loss": 0.4049948751926422, + "eval_runtime": 9.4604, + "eval_samples_per_second": 10.57, + "eval_wer": 0.3276492082825822, + "step": 139100 + }, + { + "epoch": 22.64, + "learning_rate": 9.479523576439876e-06, + "loss": NaN, + "step": 139200 + }, + { + "epoch": 22.64, + "eval_loss": 0.4711925983428955, + "eval_runtime": 7.5699, + "eval_samples_per_second": 13.21, + "eval_wer": 0.341047503045067, + "step": 139200 + }, + { + "epoch": 22.65, + "learning_rate": 9.414260075053028e-06, + "loss": NaN, + "step": 139300 + }, + { + "epoch": 22.65, + "eval_loss": 0.40160369873046875, + "eval_runtime": 7.5911, + "eval_samples_per_second": 13.173, + "eval_wer": 0.32886723507917176, + "step": 139300 + }, + { + "epoch": 22.67, + "learning_rate": 9.348996573666178e-06, + "loss": NaN, + "step": 139400 + }, + { + "epoch": 22.67, + "eval_loss": 0.45362401008605957, + "eval_runtime": 6.8691, + "eval_samples_per_second": 14.558, + "eval_wer": 0.32886723507917176, + "step": 139400 + }, + { + "epoch": 22.69, + "learning_rate": 9.283733072279328e-06, + "loss": NaN, + "step": 139500 + }, + { + "epoch": 22.69, + "eval_loss": 0.4383563995361328, + "eval_runtime": 9.4025, + "eval_samples_per_second": 10.635, + "eval_wer": 0.3386114494518879, + "step": 139500 + }, + { + "epoch": 22.7, + "learning_rate": 9.218469570892478e-06, + "loss": NaN, + "step": 139600 + }, + { + "epoch": 22.7, + "eval_loss": 0.445679634809494, + "eval_runtime": 9.7736, + "eval_samples_per_second": 10.232, + "eval_wer": 0.3386114494518879, + "step": 139600 + }, + { + "epoch": 22.72, + "learning_rate": 9.15320606950563e-06, + "loss": NaN, + "step": 139700 + }, + { + "epoch": 22.72, + "eval_loss": 0.4067118763923645, + "eval_runtime": 7.3892, + "eval_samples_per_second": 13.533, + "eval_wer": 0.3264311814859927, + "step": 139700 + }, + { + "epoch": 22.74, + "learning_rate": 9.08794256811878e-06, + "loss": NaN, + "step": 139800 + }, + { + "epoch": 22.74, + "eval_loss": 0.3995942771434784, + "eval_runtime": 7.8855, + "eval_samples_per_second": 12.681, + "eval_wer": 0.3373934226552984, + "step": 139800 + }, + { + "epoch": 22.75, + "learning_rate": 9.02267906673193e-06, + "loss": NaN, + "step": 139900 + }, + { + "epoch": 22.75, + "eval_loss": 0.39302563667297363, + "eval_runtime": 8.6481, + "eval_samples_per_second": 11.563, + "eval_wer": 0.3398294762484775, + "step": 139900 + }, + { + "epoch": 22.77, + "learning_rate": 8.95741556534508e-06, + "loss": NaN, + "step": 140000 + }, + { + "epoch": 22.77, + "eval_loss": 0.5054168701171875, + "eval_runtime": 8.0714, + "eval_samples_per_second": 12.389, + "eval_wer": 0.3520097442143727, + "step": 140000 + }, + { + "epoch": 22.78, + "learning_rate": 8.892152063958231e-06, + "loss": NaN, + "step": 140100 + }, + { + "epoch": 22.78, + "eval_loss": 0.37638041377067566, + "eval_runtime": 7.885, + "eval_samples_per_second": 12.682, + "eval_wer": 0.3325213154689403, + "step": 140100 + }, + { + "epoch": 22.8, + "learning_rate": 8.826888562571383e-06, + "loss": NaN, + "step": 140200 + }, + { + "epoch": 22.8, + "eval_loss": 0.4191051423549652, + "eval_runtime": 8.5703, + "eval_samples_per_second": 11.668, + "eval_wer": 0.3459196102314251, + "step": 140200 + }, + { + "epoch": 22.82, + "learning_rate": 8.761625061184533e-06, + "loss": NaN, + "step": 140300 + }, + { + "epoch": 22.82, + "eval_loss": 0.47149497270584106, + "eval_runtime": 7.6448, + "eval_samples_per_second": 13.081, + "eval_wer": 0.3520097442143727, + "step": 140300 + }, + { + "epoch": 22.83, + "learning_rate": 8.696361559797683e-06, + "loss": NaN, + "step": 140400 + }, + { + "epoch": 22.83, + "eval_loss": 0.4407516419887543, + "eval_runtime": 7.1529, + "eval_samples_per_second": 13.98, + "eval_wer": 0.3386114494518879, + "step": 140400 + }, + { + "epoch": 22.85, + "learning_rate": 8.631098058410834e-06, + "loss": NaN, + "step": 140500 + }, + { + "epoch": 22.85, + "eval_loss": 0.40118682384490967, + "eval_runtime": 8.4283, + "eval_samples_per_second": 11.865, + "eval_wer": 0.341047503045067, + "step": 140500 + }, + { + "epoch": 22.87, + "learning_rate": 8.565834557023984e-06, + "loss": NaN, + "step": 140600 + }, + { + "epoch": 22.87, + "eval_loss": 0.4191232919692993, + "eval_runtime": 9.0572, + "eval_samples_per_second": 11.041, + "eval_wer": 0.3459196102314251, + "step": 140600 + }, + { + "epoch": 22.88, + "learning_rate": 8.500571055637136e-06, + "loss": NaN, + "step": 140700 + }, + { + "epoch": 22.88, + "eval_loss": 0.39035916328430176, + "eval_runtime": 8.5729, + "eval_samples_per_second": 11.665, + "eval_wer": 0.3520097442143727, + "step": 140700 + }, + { + "epoch": 22.9, + "learning_rate": 8.435307554250286e-06, + "loss": NaN, + "step": 140800 + }, + { + "epoch": 22.9, + "eval_loss": 0.3838728070259094, + "eval_runtime": 7.4544, + "eval_samples_per_second": 13.415, + "eval_wer": 0.32886723507917176, + "step": 140800 + }, + { + "epoch": 22.91, + "learning_rate": 8.370044052863436e-06, + "loss": NaN, + "step": 140900 + }, + { + "epoch": 22.91, + "eval_loss": 0.4352475702762604, + "eval_runtime": 8.9386, + "eval_samples_per_second": 11.187, + "eval_wer": 0.3520097442143727, + "step": 140900 + }, + { + "epoch": 22.93, + "learning_rate": 8.304780551476588e-06, + "loss": NaN, + "step": 141000 + }, + { + "epoch": 22.93, + "eval_loss": 0.37819182872772217, + "eval_runtime": 8.645, + "eval_samples_per_second": 11.567, + "eval_wer": 0.3373934226552984, + "step": 141000 + }, + { + "epoch": 22.95, + "learning_rate": 8.239517050089737e-06, + "loss": NaN, + "step": 141100 + }, + { + "epoch": 22.95, + "eval_loss": 0.3652733266353607, + "eval_runtime": 8.9186, + "eval_samples_per_second": 11.212, + "eval_wer": 0.3471376370280146, + "step": 141100 + }, + { + "epoch": 22.96, + "learning_rate": 8.174253548702889e-06, + "loss": NaN, + "step": 141200 + }, + { + "epoch": 22.96, + "eval_loss": 0.4218634068965912, + "eval_runtime": 7.892, + "eval_samples_per_second": 12.671, + "eval_wer": 0.34226552984165654, + "step": 141200 + }, + { + "epoch": 22.98, + "learning_rate": 8.108990047316039e-06, + "loss": NaN, + "step": 141300 + }, + { + "epoch": 22.98, + "eval_loss": 0.40318989753723145, + "eval_runtime": 8.686, + "eval_samples_per_second": 11.513, + "eval_wer": 0.34835566382460414, + "step": 141300 + }, + { + "epoch": 23.0, + "learning_rate": 8.043726545929189e-06, + "loss": NaN, + "step": 141400 + }, + { + "epoch": 23.0, + "eval_loss": 0.4164011478424072, + "eval_runtime": 8.8437, + "eval_samples_per_second": 11.307, + "eval_wer": 0.34835566382460414, + "step": 141400 + }, + { + "epoch": 23.01, + "learning_rate": 7.978463044542341e-06, + "loss": NaN, + "step": 141500 + }, + { + "epoch": 23.01, + "eval_loss": 0.4907635450363159, + "eval_runtime": 7.8965, + "eval_samples_per_second": 12.664, + "eval_wer": 0.33617539585870887, + "step": 141500 + }, + { + "epoch": 23.03, + "learning_rate": 7.913199543155491e-06, + "loss": NaN, + "step": 141600 + }, + { + "epoch": 23.03, + "eval_loss": 0.4023902118206024, + "eval_runtime": 8.1256, + "eval_samples_per_second": 12.307, + "eval_wer": 0.3398294762484775, + "step": 141600 + }, + { + "epoch": 23.04, + "learning_rate": 7.847936041768641e-06, + "loss": NaN, + "step": 141700 + }, + { + "epoch": 23.04, + "eval_loss": 0.4590035676956177, + "eval_runtime": 10.0026, + "eval_samples_per_second": 9.997, + "eval_wer": 0.3386114494518879, + "step": 141700 + }, + { + "epoch": 23.06, + "learning_rate": 7.782672540381792e-06, + "loss": NaN, + "step": 141800 + }, + { + "epoch": 23.06, + "eval_loss": 0.4697200655937195, + "eval_runtime": 8.4733, + "eval_samples_per_second": 11.802, + "eval_wer": 0.3276492082825822, + "step": 141800 + }, + { + "epoch": 23.08, + "learning_rate": 7.717409038994942e-06, + "loss": NaN, + "step": 141900 + }, + { + "epoch": 23.08, + "eval_loss": 0.4211934804916382, + "eval_runtime": 8.4528, + "eval_samples_per_second": 11.83, + "eval_wer": 0.33617539585870887, + "step": 141900 + }, + { + "epoch": 23.09, + "learning_rate": 7.652145537608094e-06, + "loss": NaN, + "step": 142000 + }, + { + "epoch": 23.09, + "eval_loss": 0.39060890674591064, + "eval_runtime": 7.5777, + "eval_samples_per_second": 13.197, + "eval_wer": 0.3373934226552984, + "step": 142000 + }, + { + "epoch": 23.11, + "learning_rate": 7.586882036221243e-06, + "loss": NaN, + "step": 142100 + }, + { + "epoch": 23.11, + "eval_loss": 0.37855765223503113, + "eval_runtime": 7.3666, + "eval_samples_per_second": 13.575, + "eval_wer": 0.29963459196102316, + "step": 142100 + }, + { + "epoch": 23.13, + "learning_rate": 7.521618534834394e-06, + "loss": NaN, + "step": 142200 + }, + { + "epoch": 23.13, + "eval_loss": 0.46709468960762024, + "eval_runtime": 7.4759, + "eval_samples_per_second": 13.376, + "eval_wer": 0.3337393422655298, + "step": 142200 + }, + { + "epoch": 23.14, + "learning_rate": 7.456355033447544e-06, + "loss": NaN, + "step": 142300 + }, + { + "epoch": 23.14, + "eval_loss": 0.3746023178100586, + "eval_runtime": 7.6965, + "eval_samples_per_second": 12.993, + "eval_wer": 0.3325213154689403, + "step": 142300 + }, + { + "epoch": 23.16, + "learning_rate": 7.3910915320606955e-06, + "loss": NaN, + "step": 142400 + }, + { + "epoch": 23.16, + "eval_loss": 0.4183931052684784, + "eval_runtime": 10.5082, + "eval_samples_per_second": 9.516, + "eval_wer": 0.33617539585870887, + "step": 142400 + }, + { + "epoch": 23.17, + "learning_rate": 7.3258280306738465e-06, + "loss": NaN, + "step": 142500 + }, + { + "epoch": 23.17, + "eval_loss": 0.387781023979187, + "eval_runtime": 8.2483, + "eval_samples_per_second": 12.124, + "eval_wer": 0.3337393422655298, + "step": 142500 + }, + { + "epoch": 23.19, + "learning_rate": 7.260564529286997e-06, + "loss": NaN, + "step": 142600 + }, + { + "epoch": 23.19, + "eval_loss": 0.4201037883758545, + "eval_runtime": 9.0845, + "eval_samples_per_second": 11.008, + "eval_wer": 0.33495736906211937, + "step": 142600 + }, + { + "epoch": 23.21, + "learning_rate": 7.195301027900148e-06, + "loss": NaN, + "step": 142700 + }, + { + "epoch": 23.21, + "eval_loss": 0.40640342235565186, + "eval_runtime": 8.611, + "eval_samples_per_second": 11.613, + "eval_wer": 0.33495736906211937, + "step": 142700 + }, + { + "epoch": 23.22, + "learning_rate": 7.130037526513297e-06, + "loss": NaN, + "step": 142800 + }, + { + "epoch": 23.22, + "eval_loss": 0.36859330534935, + "eval_runtime": 6.9585, + "eval_samples_per_second": 14.371, + "eval_wer": 0.33008526187576126, + "step": 142800 + }, + { + "epoch": 23.24, + "learning_rate": 7.064774025126448e-06, + "loss": NaN, + "step": 142900 + }, + { + "epoch": 23.24, + "eval_loss": 0.5641127824783325, + "eval_runtime": 9.3783, + "eval_samples_per_second": 10.663, + "eval_wer": 0.30572472594397077, + "step": 142900 + }, + { + "epoch": 23.26, + "learning_rate": 6.9995105237395985e-06, + "loss": NaN, + "step": 143000 + }, + { + "epoch": 23.26, + "eval_loss": 0.390995591878891, + "eval_runtime": 9.1316, + "eval_samples_per_second": 10.951, + "eval_wer": 0.32886723507917176, + "step": 143000 + }, + { + "epoch": 23.27, + "learning_rate": 6.9342470223527496e-06, + "loss": NaN, + "step": 143100 + }, + { + "epoch": 23.27, + "eval_loss": 0.3884507119655609, + "eval_runtime": 7.4684, + "eval_samples_per_second": 13.39, + "eval_wer": 0.33008526187576126, + "step": 143100 + }, + { + "epoch": 23.29, + "learning_rate": 6.868983520965901e-06, + "loss": NaN, + "step": 143200 + }, + { + "epoch": 23.29, + "eval_loss": 0.37767794728279114, + "eval_runtime": 7.8276, + "eval_samples_per_second": 12.775, + "eval_wer": 0.3337393422655298, + "step": 143200 + }, + { + "epoch": 23.3, + "learning_rate": 6.803720019579051e-06, + "loss": NaN, + "step": 143300 + }, + { + "epoch": 23.3, + "eval_loss": 0.39070072770118713, + "eval_runtime": 7.9301, + "eval_samples_per_second": 12.61, + "eval_wer": 0.3373934226552984, + "step": 143300 + }, + { + "epoch": 23.32, + "learning_rate": 6.738456518192202e-06, + "loss": NaN, + "step": 143400 + }, + { + "epoch": 23.32, + "eval_loss": 0.3968910276889801, + "eval_runtime": 8.388, + "eval_samples_per_second": 11.922, + "eval_wer": 0.33495736906211937, + "step": 143400 + }, + { + "epoch": 23.34, + "learning_rate": 6.673193016805351e-06, + "loss": NaN, + "step": 143500 + }, + { + "epoch": 23.34, + "eval_loss": 0.55922931432724, + "eval_runtime": 8.0437, + "eval_samples_per_second": 12.432, + "eval_wer": 0.3520097442143727, + "step": 143500 + }, + { + "epoch": 23.35, + "learning_rate": 6.607929515418502e-06, + "loss": NaN, + "step": 143600 + }, + { + "epoch": 23.35, + "eval_loss": 0.3922981321811676, + "eval_runtime": 7.1209, + "eval_samples_per_second": 14.043, + "eval_wer": 0.3386114494518879, + "step": 143600 + }, + { + "epoch": 23.37, + "learning_rate": 6.5426660140316534e-06, + "loss": NaN, + "step": 143700 + }, + { + "epoch": 23.37, + "eval_loss": 0.39172208309173584, + "eval_runtime": 7.6116, + "eval_samples_per_second": 13.138, + "eval_wer": 0.3373934226552984, + "step": 143700 + }, + { + "epoch": 23.39, + "learning_rate": 6.477402512644804e-06, + "loss": NaN, + "step": 143800 + }, + { + "epoch": 23.39, + "eval_loss": 0.4471363127231598, + "eval_runtime": 8.7619, + "eval_samples_per_second": 11.413, + "eval_wer": 0.3373934226552984, + "step": 143800 + }, + { + "epoch": 23.4, + "learning_rate": 6.412139011257955e-06, + "loss": NaN, + "step": 143900 + }, + { + "epoch": 23.4, + "eval_loss": 0.44637829065322876, + "eval_runtime": 7.6648, + "eval_samples_per_second": 13.047, + "eval_wer": 0.34226552984165654, + "step": 143900 + }, + { + "epoch": 23.42, + "learning_rate": 6.346875509871104e-06, + "loss": NaN, + "step": 144000 + }, + { + "epoch": 23.42, + "eval_loss": 0.3966550827026367, + "eval_runtime": 9.137, + "eval_samples_per_second": 10.944, + "eval_wer": 0.3373934226552984, + "step": 144000 + }, + { + "epoch": 23.43, + "learning_rate": 6.281612008484255e-06, + "loss": NaN, + "step": 144100 + }, + { + "epoch": 23.43, + "eval_loss": 0.3890346884727478, + "eval_runtime": 6.6729, + "eval_samples_per_second": 14.986, + "eval_wer": 0.33008526187576126, + "step": 144100 + }, + { + "epoch": 23.45, + "learning_rate": 6.216348507097406e-06, + "loss": NaN, + "step": 144200 + }, + { + "epoch": 23.45, + "eval_loss": 0.417468786239624, + "eval_runtime": 7.6605, + "eval_samples_per_second": 13.054, + "eval_wer": 0.3337393422655298, + "step": 144200 + }, + { + "epoch": 23.47, + "learning_rate": 6.1510850057105564e-06, + "loss": NaN, + "step": 144300 + }, + { + "epoch": 23.47, + "eval_loss": 0.47427332401275635, + "eval_runtime": 8.9388, + "eval_samples_per_second": 11.187, + "eval_wer": 0.3264311814859927, + "step": 144300 + }, + { + "epoch": 23.48, + "learning_rate": 6.085821504323707e-06, + "loss": NaN, + "step": 144400 + }, + { + "epoch": 23.48, + "eval_loss": 0.3865101933479309, + "eval_runtime": 7.3321, + "eval_samples_per_second": 13.639, + "eval_wer": 0.33008526187576126, + "step": 144400 + }, + { + "epoch": 23.5, + "learning_rate": 6.020558002936858e-06, + "loss": NaN, + "step": 144500 + }, + { + "epoch": 23.5, + "eval_loss": 0.3971826434135437, + "eval_runtime": 9.3902, + "eval_samples_per_second": 10.649, + "eval_wer": 0.3459196102314251, + "step": 144500 + }, + { + "epoch": 23.52, + "learning_rate": 5.955294501550009e-06, + "loss": NaN, + "step": 144600 + }, + { + "epoch": 23.52, + "eval_loss": 0.5336860418319702, + "eval_runtime": 7.7801, + "eval_samples_per_second": 12.853, + "eval_wer": 0.3276492082825822, + "step": 144600 + }, + { + "epoch": 23.53, + "learning_rate": 5.890031000163159e-06, + "loss": NaN, + "step": 144700 + }, + { + "epoch": 23.53, + "eval_loss": 0.4055691063404083, + "eval_runtime": 7.3304, + "eval_samples_per_second": 13.642, + "eval_wer": 0.3325213154689403, + "step": 144700 + }, + { + "epoch": 23.55, + "learning_rate": 5.824767498776309e-06, + "loss": NaN, + "step": 144800 + }, + { + "epoch": 23.55, + "eval_loss": 0.39876997470855713, + "eval_runtime": 8.2514, + "eval_samples_per_second": 12.119, + "eval_wer": 0.3398294762484775, + "step": 144800 + }, + { + "epoch": 23.56, + "learning_rate": 5.75950399738946e-06, + "loss": NaN, + "step": 144900 + }, + { + "epoch": 23.56, + "eval_loss": 0.46831241250038147, + "eval_runtime": 7.8267, + "eval_samples_per_second": 12.777, + "eval_wer": 0.33495736906211937, + "step": 144900 + }, + { + "epoch": 23.58, + "learning_rate": 5.6942404960026105e-06, + "loss": NaN, + "step": 145000 + }, + { + "epoch": 23.58, + "eval_loss": 0.402251273393631, + "eval_runtime": 8.6896, + "eval_samples_per_second": 11.508, + "eval_wer": 0.341047503045067, + "step": 145000 + }, + { + "epoch": 23.6, + "learning_rate": 5.628976994615762e-06, + "loss": NaN, + "step": 145100 + }, + { + "epoch": 23.6, + "eval_loss": 0.42078322172164917, + "eval_runtime": 7.0241, + "eval_samples_per_second": 14.237, + "eval_wer": 0.3386114494518879, + "step": 145100 + }, + { + "epoch": 23.61, + "learning_rate": 5.563713493228912e-06, + "loss": NaN, + "step": 145200 + }, + { + "epoch": 23.61, + "eval_loss": 0.37739893794059753, + "eval_runtime": 7.4681, + "eval_samples_per_second": 13.39, + "eval_wer": 0.3325213154689403, + "step": 145200 + }, + { + "epoch": 23.63, + "learning_rate": 5.498449991842063e-06, + "loss": NaN, + "step": 145300 + }, + { + "epoch": 23.63, + "eval_loss": 0.38974130153656006, + "eval_runtime": 8.1636, + "eval_samples_per_second": 12.249, + "eval_wer": 0.3325213154689403, + "step": 145300 + }, + { + "epoch": 23.65, + "learning_rate": 5.433186490455213e-06, + "loss": NaN, + "step": 145400 + }, + { + "epoch": 23.65, + "eval_loss": 0.42759647965431213, + "eval_runtime": 8.6656, + "eval_samples_per_second": 11.54, + "eval_wer": 0.3447015834348356, + "step": 145400 + }, + { + "epoch": 23.66, + "learning_rate": 5.367922989068363e-06, + "loss": NaN, + "step": 145500 + }, + { + "epoch": 23.66, + "eval_loss": 0.446329265832901, + "eval_runtime": 7.4281, + "eval_samples_per_second": 13.462, + "eval_wer": 0.3447015834348356, + "step": 145500 + }, + { + "epoch": 23.68, + "learning_rate": 5.302659487681514e-06, + "loss": NaN, + "step": 145600 + }, + { + "epoch": 23.68, + "eval_loss": 0.4557843804359436, + "eval_runtime": 8.6796, + "eval_samples_per_second": 11.521, + "eval_wer": 0.34957369062119364, + "step": 145600 + }, + { + "epoch": 23.69, + "learning_rate": 5.2373959862946655e-06, + "loss": NaN, + "step": 145700 + }, + { + "epoch": 23.69, + "eval_loss": 0.42350074648857117, + "eval_runtime": 6.9233, + "eval_samples_per_second": 14.444, + "eval_wer": 0.3313032886723508, + "step": 145700 + }, + { + "epoch": 23.71, + "learning_rate": 5.172132484907816e-06, + "loss": NaN, + "step": 145800 + }, + { + "epoch": 23.71, + "eval_loss": 0.4503900110721588, + "eval_runtime": 9.2067, + "eval_samples_per_second": 10.862, + "eval_wer": 0.34835566382460414, + "step": 145800 + }, + { + "epoch": 23.73, + "learning_rate": 5.106868983520966e-06, + "loss": NaN, + "step": 145900 + }, + { + "epoch": 23.73, + "eval_loss": 0.38946789503097534, + "eval_runtime": 6.7208, + "eval_samples_per_second": 14.879, + "eval_wer": 0.32521315468940315, + "step": 145900 + }, + { + "epoch": 23.74, + "learning_rate": 5.041605482134117e-06, + "loss": NaN, + "step": 146000 + }, + { + "epoch": 23.74, + "eval_loss": 0.42941275238990784, + "eval_runtime": 9.2546, + "eval_samples_per_second": 10.805, + "eval_wer": 0.3276492082825822, + "step": 146000 + }, + { + "epoch": 23.76, + "learning_rate": 4.976341980747267e-06, + "loss": NaN, + "step": 146100 + }, + { + "epoch": 23.76, + "eval_loss": 0.3988470435142517, + "eval_runtime": 8.8515, + "eval_samples_per_second": 11.298, + "eval_wer": 0.3337393422655298, + "step": 146100 + }, + { + "epoch": 23.78, + "learning_rate": 4.911078479360418e-06, + "loss": NaN, + "step": 146200 + }, + { + "epoch": 23.78, + "eval_loss": 0.37505969405174255, + "eval_runtime": 8.8108, + "eval_samples_per_second": 11.35, + "eval_wer": 0.3264311814859927, + "step": 146200 + }, + { + "epoch": 23.79, + "learning_rate": 4.8458149779735685e-06, + "loss": NaN, + "step": 146300 + }, + { + "epoch": 23.79, + "eval_loss": 0.4064733386039734, + "eval_runtime": 8.0955, + "eval_samples_per_second": 12.353, + "eval_wer": 0.33495736906211937, + "step": 146300 + }, + { + "epoch": 23.81, + "learning_rate": 4.7805514765867196e-06, + "loss": NaN, + "step": 146400 + }, + { + "epoch": 23.81, + "eval_loss": 0.38152438402175903, + "eval_runtime": 9.0081, + "eval_samples_per_second": 11.101, + "eval_wer": 0.3264311814859927, + "step": 146400 + }, + { + "epoch": 23.83, + "learning_rate": 4.71528797519987e-06, + "loss": NaN, + "step": 146500 + }, + { + "epoch": 23.83, + "eval_loss": 0.4286877512931824, + "eval_runtime": 34.8815, + "eval_samples_per_second": 2.867, + "eval_wer": 0.34957369062119364, + "step": 146500 + }, + { + "epoch": 23.84, + "learning_rate": 4.65002447381302e-06, + "loss": NaN, + "step": 146600 + }, + { + "epoch": 23.84, + "eval_loss": 0.3909842371940613, + "eval_runtime": 6.7922, + "eval_samples_per_second": 14.723, + "eval_wer": 0.3215590742996346, + "step": 146600 + }, + { + "epoch": 23.86, + "learning_rate": 4.58476097242617e-06, + "loss": NaN, + "step": 146700 + }, + { + "epoch": 23.86, + "eval_loss": 0.4343806505203247, + "eval_runtime": 8.1254, + "eval_samples_per_second": 12.307, + "eval_wer": 0.33008526187576126, + "step": 146700 + }, + { + "epoch": 23.87, + "learning_rate": 4.519497471039322e-06, + "loss": NaN, + "step": 146800 + }, + { + "epoch": 23.87, + "eval_loss": 0.38777297735214233, + "eval_runtime": 8.5033, + "eval_samples_per_second": 11.76, + "eval_wer": 0.3313032886723508, + "step": 146800 + }, + { + "epoch": 23.89, + "learning_rate": 4.454233969652472e-06, + "loss": NaN, + "step": 146900 + }, + { + "epoch": 23.89, + "eval_loss": 0.40899020433425903, + "eval_runtime": 7.6862, + "eval_samples_per_second": 13.01, + "eval_wer": 0.3227771010962241, + "step": 146900 + }, + { + "epoch": 23.91, + "learning_rate": 4.388970468265623e-06, + "loss": NaN, + "step": 147000 + }, + { + "epoch": 23.91, + "eval_loss": 0.39654624462127686, + "eval_runtime": 9.0568, + "eval_samples_per_second": 11.041, + "eval_wer": 0.3313032886723508, + "step": 147000 + }, + { + "epoch": 23.92, + "learning_rate": 4.323706966878773e-06, + "loss": NaN, + "step": 147100 + }, + { + "epoch": 23.92, + "eval_loss": 0.5828571915626526, + "eval_runtime": 9.0588, + "eval_samples_per_second": 11.039, + "eval_wer": 0.33617539585870887, + "step": 147100 + }, + { + "epoch": 23.94, + "learning_rate": 4.258443465491924e-06, + "loss": NaN, + "step": 147200 + }, + { + "epoch": 23.94, + "eval_loss": 0.4280412793159485, + "eval_runtime": 8.3581, + "eval_samples_per_second": 11.964, + "eval_wer": 0.33495736906211937, + "step": 147200 + }, + { + "epoch": 23.96, + "learning_rate": 4.193179964105074e-06, + "loss": NaN, + "step": 147300 + }, + { + "epoch": 23.96, + "eval_loss": 0.3880465030670166, + "eval_runtime": 8.3284, + "eval_samples_per_second": 12.007, + "eval_wer": 0.3215590742996346, + "step": 147300 + }, + { + "epoch": 23.97, + "learning_rate": 4.127916462718225e-06, + "loss": NaN, + "step": 147400 + }, + { + "epoch": 23.97, + "eval_loss": 0.4489409923553467, + "eval_runtime": 8.5866, + "eval_samples_per_second": 11.646, + "eval_wer": 0.33617539585870887, + "step": 147400 + }, + { + "epoch": 23.99, + "learning_rate": 4.062652961331375e-06, + "loss": NaN, + "step": 147500 + }, + { + "epoch": 23.99, + "eval_loss": 0.41607800126075745, + "eval_runtime": 8.1254, + "eval_samples_per_second": 12.307, + "eval_wer": 0.3471376370280146, + "step": 147500 + }, + { + "epoch": 24.0, + "learning_rate": 3.9973894599445265e-06, + "loss": NaN, + "step": 147600 + }, + { + "epoch": 24.0, + "eval_loss": 0.38375845551490784, + "eval_runtime": 8.373, + "eval_samples_per_second": 11.943, + "eval_wer": 0.3227771010962241, + "step": 147600 + }, + { + "epoch": 24.02, + "learning_rate": 3.932125958557677e-06, + "loss": NaN, + "step": 147700 + }, + { + "epoch": 24.02, + "eval_loss": 0.4201222360134125, + "eval_runtime": 8.7128, + "eval_samples_per_second": 11.477, + "eval_wer": 0.3447015834348356, + "step": 147700 + }, + { + "epoch": 24.04, + "learning_rate": 3.866862457170827e-06, + "loss": NaN, + "step": 147800 + }, + { + "epoch": 24.04, + "eval_loss": 0.3707936108112335, + "eval_runtime": 8.5162, + "eval_samples_per_second": 11.742, + "eval_wer": 0.3215590742996346, + "step": 147800 + }, + { + "epoch": 24.05, + "learning_rate": 3.8015989557839784e-06, + "loss": NaN, + "step": 147900 + }, + { + "epoch": 24.05, + "eval_loss": 0.4147247076034546, + "eval_runtime": 7.5178, + "eval_samples_per_second": 13.302, + "eval_wer": 0.3373934226552984, + "step": 147900 + }, + { + "epoch": 24.07, + "learning_rate": 3.7363354543971286e-06, + "loss": NaN, + "step": 148000 + }, + { + "epoch": 24.07, + "eval_loss": 0.3992792069911957, + "eval_runtime": 8.0681, + "eval_samples_per_second": 12.395, + "eval_wer": 0.3386114494518879, + "step": 148000 + }, + { + "epoch": 24.09, + "learning_rate": 3.6710719530102793e-06, + "loss": NaN, + "step": 148100 + }, + { + "epoch": 24.09, + "eval_loss": 0.4286440312862396, + "eval_runtime": 6.9843, + "eval_samples_per_second": 14.318, + "eval_wer": 0.34348355663824603, + "step": 148100 + }, + { + "epoch": 24.1, + "learning_rate": 3.60580845162343e-06, + "loss": NaN, + "step": 148200 + }, + { + "epoch": 24.1, + "eval_loss": 0.4791194200515747, + "eval_runtime": 7.4344, + "eval_samples_per_second": 13.451, + "eval_wer": 0.3447015834348356, + "step": 148200 + }, + { + "epoch": 24.12, + "learning_rate": 3.54054495023658e-06, + "loss": NaN, + "step": 148300 + }, + { + "epoch": 24.12, + "eval_loss": 0.3939591348171234, + "eval_runtime": 7.3768, + "eval_samples_per_second": 13.556, + "eval_wer": 0.34348355663824603, + "step": 148300 + }, + { + "epoch": 24.13, + "learning_rate": 3.4752814488497308e-06, + "loss": NaN, + "step": 148400 + }, + { + "epoch": 24.13, + "eval_loss": 0.4074297249317169, + "eval_runtime": 6.9657, + "eval_samples_per_second": 14.356, + "eval_wer": 0.3337393422655298, + "step": 148400 + }, + { + "epoch": 24.15, + "learning_rate": 3.410017947462882e-06, + "loss": NaN, + "step": 148500 + }, + { + "epoch": 24.15, + "eval_loss": 0.42010945081710815, + "eval_runtime": 7.4953, + "eval_samples_per_second": 13.342, + "eval_wer": 0.3373934226552984, + "step": 148500 + }, + { + "epoch": 24.17, + "learning_rate": 3.3447544460760325e-06, + "loss": NaN, + "step": 148600 + }, + { + "epoch": 24.17, + "eval_loss": 0.3944063186645508, + "eval_runtime": 8.1881, + "eval_samples_per_second": 12.213, + "eval_wer": 0.3313032886723508, + "step": 148600 + }, + { + "epoch": 24.18, + "learning_rate": 3.2794909446891827e-06, + "loss": NaN, + "step": 148700 + }, + { + "epoch": 24.18, + "eval_loss": 0.38928863406181335, + "eval_runtime": 9.1121, + "eval_samples_per_second": 10.974, + "eval_wer": 0.3325213154689403, + "step": 148700 + }, + { + "epoch": 24.2, + "learning_rate": 3.2142274433023334e-06, + "loss": NaN, + "step": 148800 + }, + { + "epoch": 24.2, + "eval_loss": 0.4040900766849518, + "eval_runtime": 7.8824, + "eval_samples_per_second": 12.687, + "eval_wer": 0.33495736906211937, + "step": 148800 + }, + { + "epoch": 24.22, + "learning_rate": 3.1489639419154836e-06, + "loss": NaN, + "step": 148900 + }, + { + "epoch": 24.22, + "eval_loss": 0.39032647013664246, + "eval_runtime": 6.7307, + "eval_samples_per_second": 14.857, + "eval_wer": 0.3215590742996346, + "step": 148900 + }, + { + "epoch": 24.23, + "learning_rate": 3.0837004405286347e-06, + "loss": NaN, + "step": 149000 + }, + { + "epoch": 24.23, + "eval_loss": 0.40411511063575745, + "eval_runtime": 8.2756, + "eval_samples_per_second": 12.084, + "eval_wer": 0.32886723507917176, + "step": 149000 + }, + { + "epoch": 24.25, + "learning_rate": 3.018436939141785e-06, + "loss": NaN, + "step": 149100 + }, + { + "epoch": 24.25, + "eval_loss": 0.43083474040031433, + "eval_runtime": 9.0012, + "eval_samples_per_second": 11.11, + "eval_wer": 0.3227771010962241, + "step": 149100 + }, + { + "epoch": 24.26, + "learning_rate": 2.953173437754936e-06, + "loss": NaN, + "step": 149200 + }, + { + "epoch": 24.26, + "eval_loss": 0.3937508463859558, + "eval_runtime": 7.7288, + "eval_samples_per_second": 12.939, + "eval_wer": 0.315468940316687, + "step": 149200 + }, + { + "epoch": 24.28, + "learning_rate": 2.887909936368086e-06, + "loss": NaN, + "step": 149300 + }, + { + "epoch": 24.28, + "eval_loss": 0.3944269120693207, + "eval_runtime": 6.9338, + "eval_samples_per_second": 14.422, + "eval_wer": 0.32399512789281365, + "step": 149300 + }, + { + "epoch": 24.3, + "learning_rate": 2.822646434981237e-06, + "loss": NaN, + "step": 149400 + }, + { + "epoch": 24.3, + "eval_loss": 0.5013560652732849, + "eval_runtime": 7.8923, + "eval_samples_per_second": 12.671, + "eval_wer": 0.34957369062119364, + "step": 149400 + }, + { + "epoch": 24.31, + "learning_rate": 2.7573829335943875e-06, + "loss": NaN, + "step": 149500 + }, + { + "epoch": 24.31, + "eval_loss": 0.39637017250061035, + "eval_runtime": 8.1021, + "eval_samples_per_second": 12.343, + "eval_wer": 0.3325213154689403, + "step": 149500 + }, + { + "epoch": 24.33, + "learning_rate": 2.692119432207538e-06, + "loss": NaN, + "step": 149600 + }, + { + "epoch": 24.33, + "eval_loss": 0.4175480306148529, + "eval_runtime": 8.4026, + "eval_samples_per_second": 11.901, + "eval_wer": 0.3325213154689403, + "step": 149600 + }, + { + "epoch": 24.35, + "learning_rate": 2.6268559308206883e-06, + "loss": NaN, + "step": 149700 + }, + { + "epoch": 24.35, + "eval_loss": 0.4383509159088135, + "eval_runtime": 8.2617, + "eval_samples_per_second": 12.104, + "eval_wer": 0.3276492082825822, + "step": 149700 + }, + { + "epoch": 24.36, + "learning_rate": 2.5615924294338394e-06, + "loss": NaN, + "step": 149800 + }, + { + "epoch": 24.36, + "eval_loss": 0.3883654475212097, + "eval_runtime": 7.6932, + "eval_samples_per_second": 12.999, + "eval_wer": 0.32886723507917176, + "step": 149800 + }, + { + "epoch": 24.38, + "learning_rate": 2.4963289280469896e-06, + "loss": NaN, + "step": 149900 + }, + { + "epoch": 24.38, + "eval_loss": 0.42128920555114746, + "eval_runtime": 7.3824, + "eval_samples_per_second": 13.546, + "eval_wer": 0.33008526187576126, + "step": 149900 + }, + { + "epoch": 24.39, + "learning_rate": 2.4310654266601403e-06, + "loss": NaN, + "step": 150000 + }, + { + "epoch": 24.39, + "eval_loss": 0.37016811966896057, + "eval_runtime": 7.3966, + "eval_samples_per_second": 13.52, + "eval_wer": 0.3264311814859927, + "step": 150000 + }, + { + "epoch": 24.41, + "learning_rate": 2.365801925273291e-06, + "loss": NaN, + "step": 150100 + }, + { + "epoch": 24.41, + "eval_loss": 0.4776330888271332, + "eval_runtime": 8.1524, + "eval_samples_per_second": 12.266, + "eval_wer": 0.3276492082825822, + "step": 150100 + }, + { + "epoch": 24.43, + "learning_rate": 2.3005384238864416e-06, + "loss": NaN, + "step": 150200 + }, + { + "epoch": 24.43, + "eval_loss": 0.40871936082839966, + "eval_runtime": 7.8643, + "eval_samples_per_second": 12.716, + "eval_wer": 0.3276492082825822, + "step": 150200 + }, + { + "epoch": 24.44, + "learning_rate": 2.235274922499592e-06, + "loss": NaN, + "step": 150300 + }, + { + "epoch": 24.44, + "eval_loss": 0.4232352375984192, + "eval_runtime": 7.5631, + "eval_samples_per_second": 13.222, + "eval_wer": 0.3325213154689403, + "step": 150300 + }, + { + "epoch": 24.46, + "learning_rate": 2.170011421112743e-06, + "loss": NaN, + "step": 150400 + }, + { + "epoch": 24.46, + "eval_loss": 0.3953474164009094, + "eval_runtime": 8.1896, + "eval_samples_per_second": 12.211, + "eval_wer": 0.32399512789281365, + "step": 150400 + }, + { + "epoch": 24.48, + "learning_rate": 2.1047479197258935e-06, + "loss": NaN, + "step": 150500 + }, + { + "epoch": 24.48, + "eval_loss": 0.40007010102272034, + "eval_runtime": 8.6682, + "eval_samples_per_second": 11.536, + "eval_wer": 0.32886723507917176, + "step": 150500 + }, + { + "epoch": 24.49, + "learning_rate": 2.039484418339044e-06, + "loss": NaN, + "step": 150600 + }, + { + "epoch": 24.49, + "eval_loss": 0.4761078357696533, + "eval_runtime": 9.2295, + "eval_samples_per_second": 10.835, + "eval_wer": 0.32886723507917176, + "step": 150600 + }, + { + "epoch": 24.51, + "learning_rate": 1.9742209169521948e-06, + "loss": NaN, + "step": 150700 + }, + { + "epoch": 24.51, + "eval_loss": 0.4079989194869995, + "eval_runtime": 8.3587, + "eval_samples_per_second": 11.964, + "eval_wer": 0.32886723507917176, + "step": 150700 + }, + { + "epoch": 24.52, + "learning_rate": 1.908957415565345e-06, + "loss": NaN, + "step": 150800 + }, + { + "epoch": 24.52, + "eval_loss": 0.3929882049560547, + "eval_runtime": 7.1125, + "eval_samples_per_second": 14.06, + "eval_wer": 0.33617539585870887, + "step": 150800 + }, + { + "epoch": 24.54, + "learning_rate": 1.8436939141784959e-06, + "loss": NaN, + "step": 150900 + }, + { + "epoch": 24.54, + "eval_loss": 0.44755759835243225, + "eval_runtime": 7.8989, + "eval_samples_per_second": 12.66, + "eval_wer": 0.3373934226552984, + "step": 150900 + }, + { + "epoch": 24.56, + "learning_rate": 1.7784304127916463e-06, + "loss": NaN, + "step": 151000 + }, + { + "epoch": 24.56, + "eval_loss": 0.3977559208869934, + "eval_runtime": 8.7185, + "eval_samples_per_second": 11.47, + "eval_wer": 0.32886723507917176, + "step": 151000 + }, + { + "epoch": 24.57, + "learning_rate": 1.713166911404797e-06, + "loss": NaN, + "step": 151100 + }, + { + "epoch": 24.57, + "eval_loss": 0.4052342176437378, + "eval_runtime": 7.8682, + "eval_samples_per_second": 12.709, + "eval_wer": 0.34226552984165654, + "step": 151100 + }, + { + "epoch": 24.59, + "learning_rate": 1.6479034100179476e-06, + "loss": NaN, + "step": 151200 + }, + { + "epoch": 24.59, + "eval_loss": 0.4712124168872833, + "eval_runtime": 8.3125, + "eval_samples_per_second": 12.03, + "eval_wer": 0.34226552984165654, + "step": 151200 + }, + { + "epoch": 24.61, + "learning_rate": 1.5826399086310982e-06, + "loss": NaN, + "step": 151300 + }, + { + "epoch": 24.61, + "eval_loss": 0.3818945288658142, + "eval_runtime": 7.968, + "eval_samples_per_second": 12.55, + "eval_wer": 0.33495736906211937, + "step": 151300 + }, + { + "epoch": 24.62, + "learning_rate": 1.5173764072442489e-06, + "loss": NaN, + "step": 151400 + }, + { + "epoch": 24.62, + "eval_loss": 0.44341355562210083, + "eval_runtime": 8.224, + "eval_samples_per_second": 12.16, + "eval_wer": 0.341047503045067, + "step": 151400 + }, + { + "epoch": 24.64, + "learning_rate": 1.4521129058573993e-06, + "loss": NaN, + "step": 151500 + }, + { + "epoch": 24.64, + "eval_loss": 0.4101918041706085, + "eval_runtime": 6.9383, + "eval_samples_per_second": 14.413, + "eval_wer": 0.3215590742996346, + "step": 151500 + }, + { + "epoch": 24.65, + "learning_rate": 1.38684940447055e-06, + "loss": NaN, + "step": 151600 + }, + { + "epoch": 24.65, + "eval_loss": 0.39534106850624084, + "eval_runtime": 8.9247, + "eval_samples_per_second": 11.205, + "eval_wer": 0.32886723507917176, + "step": 151600 + }, + { + "epoch": 24.67, + "learning_rate": 1.3215859030837006e-06, + "loss": NaN, + "step": 151700 + }, + { + "epoch": 24.67, + "eval_loss": 0.3738269805908203, + "eval_runtime": 7.6097, + "eval_samples_per_second": 13.141, + "eval_wer": 0.3264311814859927, + "step": 151700 + }, + { + "epoch": 24.69, + "learning_rate": 1.256322401696851e-06, + "loss": NaN, + "step": 151800 + }, + { + "epoch": 24.69, + "eval_loss": 0.5128779411315918, + "eval_runtime": 8.2012, + "eval_samples_per_second": 12.193, + "eval_wer": 0.3313032886723508, + "step": 151800 + }, + { + "epoch": 24.7, + "learning_rate": 1.1910589003100017e-06, + "loss": NaN, + "step": 151900 + }, + { + "epoch": 24.7, + "eval_loss": 0.41557371616363525, + "eval_runtime": 10.0221, + "eval_samples_per_second": 9.978, + "eval_wer": 0.3264311814859927, + "step": 151900 + }, + { + "epoch": 24.72, + "learning_rate": 1.1257953989231523e-06, + "loss": NaN, + "step": 152000 + }, + { + "epoch": 24.72, + "eval_loss": 0.3804296851158142, + "eval_runtime": 6.86, + "eval_samples_per_second": 14.577, + "eval_wer": 0.3166869671132765, + "step": 152000 + }, + { + "epoch": 24.74, + "learning_rate": 1.060531897536303e-06, + "loss": NaN, + "step": 152100 + }, + { + "epoch": 24.74, + "eval_loss": 0.37247443199157715, + "eval_runtime": 8.5022, + "eval_samples_per_second": 11.762, + "eval_wer": 0.31790499390986604, + "step": 152100 + }, + { + "epoch": 24.75, + "learning_rate": 9.952683961494534e-07, + "loss": NaN, + "step": 152200 + }, + { + "epoch": 24.75, + "eval_loss": 0.4154191315174103, + "eval_runtime": 7.3705, + "eval_samples_per_second": 13.568, + "eval_wer": 0.33495736906211937, + "step": 152200 + }, + { + "epoch": 24.77, + "learning_rate": 9.30004894762604e-07, + "loss": NaN, + "step": 152300 + }, + { + "epoch": 24.77, + "eval_loss": 0.46218574047088623, + "eval_runtime": 8.1433, + "eval_samples_per_second": 12.28, + "eval_wer": 0.34226552984165654, + "step": 152300 + }, + { + "epoch": 24.78, + "learning_rate": 8.647413933757547e-07, + "loss": NaN, + "step": 152400 + }, + { + "epoch": 24.78, + "eval_loss": 0.39375588297843933, + "eval_runtime": 6.9811, + "eval_samples_per_second": 14.324, + "eval_wer": 0.29963459196102316, + "step": 152400 + }, + { + "epoch": 24.8, + "learning_rate": 7.994778919889051e-07, + "loss": NaN, + "step": 152500 + }, + { + "epoch": 24.8, + "eval_loss": 0.4095069169998169, + "eval_runtime": 7.9503, + "eval_samples_per_second": 12.578, + "eval_wer": 0.3373934226552984, + "step": 152500 + }, + { + "epoch": 24.82, + "learning_rate": 7.342143906020558e-07, + "loss": NaN, + "step": 152600 + }, + { + "epoch": 24.82, + "eval_loss": 0.431011438369751, + "eval_runtime": 7.7649, + "eval_samples_per_second": 12.878, + "eval_wer": 0.3313032886723508, + "step": 152600 + }, + { + "epoch": 24.83, + "learning_rate": 6.689508892152064e-07, + "loss": NaN, + "step": 152700 + }, + { + "epoch": 24.83, + "eval_loss": 0.39817845821380615, + "eval_runtime": 6.8105, + "eval_samples_per_second": 14.683, + "eval_wer": 0.3264311814859927, + "step": 152700 + }, + { + "epoch": 24.85, + "learning_rate": 6.036873878283571e-07, + "loss": NaN, + "step": 152800 + }, + { + "epoch": 24.85, + "eval_loss": 0.3910774886608124, + "eval_runtime": 7.13, + "eval_samples_per_second": 14.025, + "eval_wer": 0.32034104750304504, + "step": 152800 + }, + { + "epoch": 24.87, + "learning_rate": 5.384238864415076e-07, + "loss": NaN, + "step": 152900 + }, + { + "epoch": 24.87, + "eval_loss": 0.40002116560935974, + "eval_runtime": 7.5393, + "eval_samples_per_second": 13.264, + "eval_wer": 0.31912302070645554, + "step": 152900 + }, + { + "epoch": 24.88, + "learning_rate": 4.7316038505465825e-07, + "loss": NaN, + "step": 153000 + }, + { + "epoch": 24.88, + "eval_loss": 0.40296486020088196, + "eval_runtime": 9.5006, + "eval_samples_per_second": 10.526, + "eval_wer": 0.33617539585870887, + "step": 153000 + }, + { + "epoch": 24.9, + "learning_rate": 4.078968836678088e-07, + "loss": NaN, + "step": 153100 + }, + { + "epoch": 24.9, + "eval_loss": 0.3949245810508728, + "eval_runtime": 9.1847, + "eval_samples_per_second": 10.888, + "eval_wer": 0.3313032886723508, + "step": 153100 + }, + { + "epoch": 24.91, + "learning_rate": 3.426333822809594e-07, + "loss": NaN, + "step": 153200 + }, + { + "epoch": 24.91, + "eval_loss": 0.43339037895202637, + "eval_runtime": 7.6951, + "eval_samples_per_second": 12.995, + "eval_wer": 0.34348355663824603, + "step": 153200 + }, + { + "epoch": 24.93, + "learning_rate": 2.7736988089410997e-07, + "loss": NaN, + "step": 153300 + }, + { + "epoch": 24.93, + "eval_loss": 0.439365416765213, + "eval_runtime": 6.9543, + "eval_samples_per_second": 14.38, + "eval_wer": 0.341047503045067, + "step": 153300 + }, + { + "epoch": 24.95, + "learning_rate": 2.121063795072606e-07, + "loss": NaN, + "step": 153400 + }, + { + "epoch": 24.95, + "eval_loss": 0.4255564510822296, + "eval_runtime": 8.4137, + "eval_samples_per_second": 11.885, + "eval_wer": 0.3373934226552984, + "step": 153400 + }, + { + "epoch": 24.96, + "learning_rate": 1.4684287812041115e-07, + "loss": NaN, + "step": 153500 + }, + { + "epoch": 24.96, + "eval_loss": 0.3864142894744873, + "eval_runtime": 9.5624, + "eval_samples_per_second": 10.458, + "eval_wer": 0.33008526187576126, + "step": 153500 + }, + { + "epoch": 24.98, + "learning_rate": 8.157937673356176e-08, + "loss": NaN, + "step": 153600 + }, + { + "epoch": 24.98, + "eval_loss": 0.4688411056995392, + "eval_runtime": 7.8079, + "eval_samples_per_second": 12.808, + "eval_wer": 0.3325213154689403, + "step": 153600 + }, + { + "epoch": 25.0, + "learning_rate": 1.6315875346712352e-08, + "loss": NaN, + "step": 153700 + }, + { + "epoch": 25.0, + "eval_loss": 0.40519940853118896, + "eval_runtime": 8.6405, + "eval_samples_per_second": 11.573, + "eval_wer": 0.32521315468940315, + "step": 153700 + }, + { + "epoch": 25.0, + "step": 153725, + "total_flos": 7.526023236355262e+20, + "train_runtime": 258038.6163, + "train_samples_per_second": 0.596 + } + ], + "max_steps": 153725, + "num_train_epochs": 25, + "total_flos": 7.526023236355262e+20, + "trial_name": null, + "trial_params": null +}