| { | |
| "best_metric": 27.962619834044954, | |
| "best_model_checkpoint": "/cluster/home/torstefl/Master/saved_model/Whisper/BB-03.06/checkpoint-1410", | |
| "epoch": 3.740053050397878, | |
| "eval_steps": 94, | |
| "global_step": 1410, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.25, | |
| "eval_rundkast_loss": 1.1710554361343384, | |
| "eval_rundkast_runtime": 394.6232, | |
| "eval_rundkast_samples_per_second": 3.393, | |
| "eval_rundkast_steps_per_second": 0.071, | |
| "eval_rundkast_wer": 17.813149896016636, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "eval_nb_samtale_loss": 1.1702924966812134, | |
| "eval_nb_samtale_runtime": 186.1995, | |
| "eval_nb_samtale_samples_per_second": 2.863, | |
| "eval_nb_samtale_steps_per_second": 0.064, | |
| "eval_nb_samtale_wer": 17.0166015625, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "eval_bigbrother_loss": 1.5188441276550293, | |
| "eval_bigbrother_runtime": 400.8668, | |
| "eval_bigbrother_samples_per_second": 3.418, | |
| "eval_bigbrother_steps_per_second": 0.072, | |
| "eval_bigbrother_wer": 36.69539998388786, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "grad_norm": 32.34502410888672, | |
| "learning_rate": 1.0000000000000002e-06, | |
| "loss": 1.9553, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_rundkast_loss": 0.6190834045410156, | |
| "eval_rundkast_runtime": 386.9299, | |
| "eval_rundkast_samples_per_second": 3.461, | |
| "eval_rundkast_steps_per_second": 0.072, | |
| "eval_rundkast_wer": 17.60518317069269, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_nb_samtale_loss": 0.6846473217010498, | |
| "eval_nb_samtale_runtime": 194.999, | |
| "eval_nb_samtale_samples_per_second": 2.733, | |
| "eval_nb_samtale_steps_per_second": 0.062, | |
| "eval_nb_samtale_wer": 20.646158854166664, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_bigbrother_loss": 0.8479565382003784, | |
| "eval_bigbrother_runtime": 419.9132, | |
| "eval_bigbrother_samples_per_second": 3.263, | |
| "eval_bigbrother_steps_per_second": 0.069, | |
| "eval_bigbrother_wer": 41.819060662208976, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "grad_norm": 7.441174030303955, | |
| "learning_rate": 2.0000000000000003e-06, | |
| "loss": 1.1636, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "eval_rundkast_loss": 0.4178731143474579, | |
| "eval_rundkast_runtime": 388.088, | |
| "eval_rundkast_samples_per_second": 3.45, | |
| "eval_rundkast_steps_per_second": 0.072, | |
| "eval_rundkast_wer": 17.069268916973286, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "eval_nb_samtale_loss": 0.4929106533527374, | |
| "eval_nb_samtale_runtime": 190.8866, | |
| "eval_nb_samtale_samples_per_second": 2.792, | |
| "eval_nb_samtale_steps_per_second": 0.063, | |
| "eval_nb_samtale_wer": 16.9677734375, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "eval_bigbrother_loss": 0.6370145678520203, | |
| "eval_bigbrother_runtime": 400.3961, | |
| "eval_bigbrother_samples_per_second": 3.422, | |
| "eval_bigbrother_steps_per_second": 0.072, | |
| "eval_bigbrother_wer": 32.86876661564489, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "grad_norm": 8.241803169250488, | |
| "learning_rate": 3e-06, | |
| "loss": 0.7531, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_rundkast_loss": 0.37628263235092163, | |
| "eval_rundkast_runtime": 393.1584, | |
| "eval_rundkast_samples_per_second": 3.406, | |
| "eval_rundkast_steps_per_second": 0.071, | |
| "eval_rundkast_wer": 16.73332266837306, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_nb_samtale_loss": 0.4400510787963867, | |
| "eval_nb_samtale_runtime": 194.3558, | |
| "eval_nb_samtale_samples_per_second": 2.742, | |
| "eval_nb_samtale_steps_per_second": 0.062, | |
| "eval_nb_samtale_wer": 16.219075520833336, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_bigbrother_loss": 0.5805255174636841, | |
| "eval_bigbrother_runtime": 420.4239, | |
| "eval_bigbrother_samples_per_second": 3.259, | |
| "eval_bigbrother_steps_per_second": 0.069, | |
| "eval_bigbrother_wer": 31.69258035930073, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "grad_norm": 6.843526363372803, | |
| "learning_rate": 4.000000000000001e-06, | |
| "loss": 0.644, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "eval_rundkast_loss": 0.3539932370185852, | |
| "eval_rundkast_runtime": 389.9336, | |
| "eval_rundkast_samples_per_second": 3.434, | |
| "eval_rundkast_steps_per_second": 0.072, | |
| "eval_rundkast_wer": 16.685330347144458, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "eval_nb_samtale_loss": 0.4174058139324188, | |
| "eval_nb_samtale_runtime": 187.98, | |
| "eval_nb_samtale_samples_per_second": 2.835, | |
| "eval_nb_samtale_steps_per_second": 0.064, | |
| "eval_nb_samtale_wer": 15.519205729166666, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "eval_bigbrother_loss": 0.5535444021224976, | |
| "eval_bigbrother_runtime": 401.275, | |
| "eval_bigbrother_samples_per_second": 3.414, | |
| "eval_bigbrother_steps_per_second": 0.072, | |
| "eval_bigbrother_wer": 31.297832917103037, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "grad_norm": 6.64098596572876, | |
| "learning_rate": 5e-06, | |
| "loss": 0.5711, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "eval_rundkast_loss": 0.3458848297595978, | |
| "eval_rundkast_runtime": 390.2988, | |
| "eval_rundkast_samples_per_second": 3.431, | |
| "eval_rundkast_steps_per_second": 0.072, | |
| "eval_rundkast_wer": 16.357382818749, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "eval_nb_samtale_loss": 0.4064616560935974, | |
| "eval_nb_samtale_runtime": 190.1597, | |
| "eval_nb_samtale_samples_per_second": 2.803, | |
| "eval_nb_samtale_steps_per_second": 0.063, | |
| "eval_nb_samtale_wer": 16.1376953125, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "eval_bigbrother_loss": 0.534324586391449, | |
| "eval_bigbrother_runtime": 402.9114, | |
| "eval_bigbrother_samples_per_second": 3.4, | |
| "eval_bigbrother_steps_per_second": 0.072, | |
| "eval_bigbrother_wer": 30.500281962458715, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "grad_norm": 5.4988203048706055, | |
| "learning_rate": 6e-06, | |
| "loss": 0.5376, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "eval_rundkast_loss": 0.3418006896972656, | |
| "eval_rundkast_runtime": 393.619, | |
| "eval_rundkast_samples_per_second": 3.402, | |
| "eval_rundkast_steps_per_second": 0.071, | |
| "eval_rundkast_wer": 16.42137258038714, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "eval_nb_samtale_loss": 0.3977779150009155, | |
| "eval_nb_samtale_runtime": 187.361, | |
| "eval_nb_samtale_samples_per_second": 2.845, | |
| "eval_nb_samtale_steps_per_second": 0.064, | |
| "eval_nb_samtale_wer": 15.950520833333334, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "eval_bigbrother_loss": 0.5196393728256226, | |
| "eval_bigbrother_runtime": 400.8125, | |
| "eval_bigbrother_samples_per_second": 3.418, | |
| "eval_bigbrother_steps_per_second": 0.072, | |
| "eval_bigbrother_wer": 29.01796503665512, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "grad_norm": 6.2353644371032715, | |
| "learning_rate": 7e-06, | |
| "loss": 0.5275, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "eval_rundkast_loss": 0.3388798236846924, | |
| "eval_rundkast_runtime": 390.6781, | |
| "eval_rundkast_samples_per_second": 3.427, | |
| "eval_rundkast_steps_per_second": 0.072, | |
| "eval_rundkast_wer": 16.405375139977604, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "eval_nb_samtale_loss": 0.3939466178417206, | |
| "eval_nb_samtale_runtime": 188.2555, | |
| "eval_nb_samtale_samples_per_second": 2.831, | |
| "eval_nb_samtale_steps_per_second": 0.064, | |
| "eval_nb_samtale_wer": 16.023763020833336, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "eval_bigbrother_loss": 0.5035152435302734, | |
| "eval_bigbrother_runtime": 407.9919, | |
| "eval_bigbrother_samples_per_second": 3.358, | |
| "eval_bigbrother_steps_per_second": 0.071, | |
| "eval_bigbrother_wer": 33.505196165310565, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "grad_norm": 5.442716598510742, | |
| "learning_rate": 8.000000000000001e-06, | |
| "loss": 0.4749, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "eval_rundkast_loss": 0.33767446875572205, | |
| "eval_rundkast_runtime": 393.1862, | |
| "eval_rundkast_samples_per_second": 3.406, | |
| "eval_rundkast_steps_per_second": 0.071, | |
| "eval_rundkast_wer": 16.221404575267957, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "eval_nb_samtale_loss": 0.38744962215423584, | |
| "eval_nb_samtale_runtime": 190.2295, | |
| "eval_nb_samtale_samples_per_second": 2.802, | |
| "eval_nb_samtale_steps_per_second": 0.063, | |
| "eval_nb_samtale_wer": 15.519205729166666, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "eval_bigbrother_loss": 0.5067595839500427, | |
| "eval_bigbrother_runtime": 402.7715, | |
| "eval_bigbrother_samples_per_second": 3.401, | |
| "eval_bigbrother_steps_per_second": 0.072, | |
| "eval_bigbrother_wer": 28.405703697736246, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "grad_norm": 5.219305515289307, | |
| "learning_rate": 9e-06, | |
| "loss": 0.3914, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "eval_rundkast_loss": 0.3438616394996643, | |
| "eval_rundkast_runtime": 387.6196, | |
| "eval_rundkast_samples_per_second": 3.454, | |
| "eval_rundkast_steps_per_second": 0.072, | |
| "eval_rundkast_wer": 16.76531754919213, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "eval_nb_samtale_loss": 0.39085230231285095, | |
| "eval_nb_samtale_runtime": 187.9942, | |
| "eval_nb_samtale_samples_per_second": 2.835, | |
| "eval_nb_samtale_steps_per_second": 0.064, | |
| "eval_nb_samtale_wer": 15.462239583333334, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "eval_bigbrother_loss": 0.49723753333091736, | |
| "eval_bigbrother_runtime": 399.7333, | |
| "eval_bigbrother_samples_per_second": 3.427, | |
| "eval_bigbrother_steps_per_second": 0.073, | |
| "eval_bigbrother_wer": 29.557721743333605, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "grad_norm": 5.705211639404297, | |
| "learning_rate": 1e-05, | |
| "loss": 0.3954, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "eval_rundkast_loss": 0.3441552519798279, | |
| "eval_rundkast_runtime": 412.597, | |
| "eval_rundkast_samples_per_second": 3.245, | |
| "eval_rundkast_steps_per_second": 0.068, | |
| "eval_rundkast_wer": 26.235802271636537, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "eval_nb_samtale_loss": 0.390921950340271, | |
| "eval_nb_samtale_runtime": 189.49, | |
| "eval_nb_samtale_samples_per_second": 2.813, | |
| "eval_nb_samtale_steps_per_second": 0.063, | |
| "eval_nb_samtale_wer": 15.804036458333334, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "eval_bigbrother_loss": 0.4992116689682007, | |
| "eval_bigbrother_runtime": 409.5244, | |
| "eval_bigbrother_samples_per_second": 3.345, | |
| "eval_bigbrother_steps_per_second": 0.071, | |
| "eval_bigbrother_wer": 31.52340288407315, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "grad_norm": 6.673018932342529, | |
| "learning_rate": 8.863636363636365e-06, | |
| "loss": 0.3922, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "eval_rundkast_loss": 0.34099873900413513, | |
| "eval_rundkast_runtime": 393.4534, | |
| "eval_rundkast_samples_per_second": 3.403, | |
| "eval_rundkast_steps_per_second": 0.071, | |
| "eval_rundkast_wer": 18.549032154855226, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "eval_nb_samtale_loss": 0.3950643241405487, | |
| "eval_nb_samtale_runtime": 191.5246, | |
| "eval_nb_samtale_samples_per_second": 2.783, | |
| "eval_nb_samtale_steps_per_second": 0.063, | |
| "eval_nb_samtale_wer": 17.618815104166664, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "eval_bigbrother_loss": 0.4919377565383911, | |
| "eval_bigbrother_runtime": 408.3017, | |
| "eval_bigbrother_samples_per_second": 3.355, | |
| "eval_bigbrother_steps_per_second": 0.071, | |
| "eval_bigbrother_wer": 30.653347297188432, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "grad_norm": 4.939565658569336, | |
| "learning_rate": 7.727272727272727e-06, | |
| "loss": 0.2929, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "eval_rundkast_loss": 0.35260480642318726, | |
| "eval_rundkast_runtime": 405.1086, | |
| "eval_rundkast_samples_per_second": 3.305, | |
| "eval_rundkast_steps_per_second": 0.069, | |
| "eval_rundkast_wer": 22.75635898256279, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "eval_nb_samtale_loss": 0.4051592946052551, | |
| "eval_nb_samtale_runtime": 189.9189, | |
| "eval_nb_samtale_samples_per_second": 2.806, | |
| "eval_nb_samtale_steps_per_second": 0.063, | |
| "eval_nb_samtale_wer": 16.300455729166664, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "eval_bigbrother_loss": 0.5098506212234497, | |
| "eval_bigbrother_runtime": 405.0466, | |
| "eval_bigbrother_samples_per_second": 3.382, | |
| "eval_bigbrother_steps_per_second": 0.072, | |
| "eval_bigbrother_wer": 28.695722226697818, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "grad_norm": 4.733194828033447, | |
| "learning_rate": 6.590909090909091e-06, | |
| "loss": 0.257, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "eval_rundkast_loss": 0.3553010821342468, | |
| "eval_rundkast_runtime": 395.5247, | |
| "eval_rundkast_samples_per_second": 3.385, | |
| "eval_rundkast_steps_per_second": 0.071, | |
| "eval_rundkast_wer": 21.036634138537835, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "eval_nb_samtale_loss": 0.40574032068252563, | |
| "eval_nb_samtale_runtime": 189.1442, | |
| "eval_nb_samtale_samples_per_second": 2.818, | |
| "eval_nb_samtale_steps_per_second": 0.063, | |
| "eval_nb_samtale_wer": 16.365559895833336, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "eval_bigbrother_loss": 0.5063855051994324, | |
| "eval_bigbrother_runtime": 407.0043, | |
| "eval_bigbrother_samples_per_second": 3.366, | |
| "eval_bigbrother_steps_per_second": 0.071, | |
| "eval_bigbrother_wer": 28.824619350680734, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "grad_norm": 4.505491733551025, | |
| "learning_rate": 5.4545454545454545e-06, | |
| "loss": 0.2694, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "eval_rundkast_loss": 0.35006436705589294, | |
| "eval_rundkast_runtime": 405.6289, | |
| "eval_rundkast_samples_per_second": 3.301, | |
| "eval_rundkast_steps_per_second": 0.069, | |
| "eval_rundkast_wer": 21.732522796352583, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "eval_nb_samtale_loss": 0.3994791805744171, | |
| "eval_nb_samtale_runtime": 189.4524, | |
| "eval_nb_samtale_samples_per_second": 2.813, | |
| "eval_nb_samtale_steps_per_second": 0.063, | |
| "eval_nb_samtale_wer": 15.966796875, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "eval_bigbrother_loss": 0.5046056509017944, | |
| "eval_bigbrother_runtime": 405.5821, | |
| "eval_bigbrother_samples_per_second": 3.378, | |
| "eval_bigbrother_steps_per_second": 0.072, | |
| "eval_bigbrother_wer": 27.962619834044954, | |
| "step": 1410 | |
| } | |
| ], | |
| "logging_steps": 100, | |
| "max_steps": 1885, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 5, | |
| "save_steps": 94, | |
| "total_flos": 1.950288993902592e+19, | |
| "train_batch_size": 48, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |