{ "best_metric": 0.404656408603883, "best_model_checkpoint": "/cluster/home/torstefl/Master/saved_model/W2V/single/BB/30.05/checkpoint-13572", "epoch": 40.0, "eval_steps": 500, "global_step": 15080, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 1.0, "grad_norm": 6.709790229797363, "learning_rate": 2.5e-05, "loss": 2.4835, "step": 377 }, { "epoch": 1.0, "eval_rundkast_loss": 0.419185996055603, "eval_rundkast_runtime": 33.4408, "eval_rundkast_samples_per_second": 40.22, "eval_rundkast_steps_per_second": 1.286, "eval_rundkast_wer": 0.24005727012408526, "step": 377 }, { "epoch": 1.0, "eval_nb_samtale_loss": 0.5932812094688416, "eval_nb_samtale_runtime": 38.8625, "eval_nb_samtale_samples_per_second": 13.715, "eval_nb_samtale_steps_per_second": 0.437, "eval_nb_samtale_wer": 0.2626953125, "step": 377 }, { "epoch": 1.0, "eval_bigbrother_loss": 1.838633418083191, "eval_bigbrother_runtime": 41.4238, "eval_bigbrother_samples_per_second": 33.073, "eval_bigbrother_steps_per_second": 1.038, "eval_bigbrother_wer": 0.5936518166438411, "step": 377 }, { "epoch": 2.0, "grad_norm": 8.728084564208984, "learning_rate": 5.013333333333333e-05, "loss": 2.0585, "step": 754 }, { "epoch": 2.0, "eval_rundkast_loss": 0.40199896693229675, "eval_rundkast_runtime": 33.3039, "eval_rundkast_samples_per_second": 40.386, "eval_rundkast_steps_per_second": 1.291, "eval_rundkast_wer": 0.2357620108176901, "step": 754 }, { "epoch": 2.0, "eval_nb_samtale_loss": 0.5779798030853271, "eval_nb_samtale_runtime": 37.921, "eval_nb_samtale_samples_per_second": 14.056, "eval_nb_samtale_steps_per_second": 0.448, "eval_nb_samtale_wer": 0.253173828125, "step": 754 }, { "epoch": 2.0, "eval_bigbrother_loss": 1.5922973155975342, "eval_bigbrother_runtime": 41.1151, "eval_bigbrother_samples_per_second": 33.321, "eval_bigbrother_steps_per_second": 1.046, "eval_bigbrother_wer": 0.5574800612261339, "step": 754 }, { "epoch": 3.0, "grad_norm": 6.347009658813477, "learning_rate": 7.526666666666668e-05, "loss": 1.9142, "step": 1131 }, { "epoch": 3.0, "eval_rundkast_loss": 0.3938983082771301, "eval_rundkast_runtime": 33.5062, "eval_rundkast_samples_per_second": 40.142, "eval_rundkast_steps_per_second": 1.283, "eval_rundkast_wer": 0.24029589564110723, "step": 1131 }, { "epoch": 3.0, "eval_nb_samtale_loss": 0.5637267827987671, "eval_nb_samtale_runtime": 38.0506, "eval_nb_samtale_samples_per_second": 14.008, "eval_nb_samtale_steps_per_second": 0.447, "eval_nb_samtale_wer": 0.2652180989583333, "step": 1131 }, { "epoch": 3.0, "eval_bigbrother_loss": 1.4027199745178223, "eval_bigbrother_runtime": 41.4816, "eval_bigbrother_samples_per_second": 33.027, "eval_bigbrother_steps_per_second": 1.037, "eval_bigbrother_wer": 0.5358897929589946, "step": 1131 }, { "epoch": 4.0, "grad_norm": 13.580578804016113, "learning_rate": 9.99558173784978e-05, "loss": 1.816, "step": 1508 }, { "epoch": 4.0, "eval_rundkast_loss": 0.3921523690223694, "eval_rundkast_runtime": 33.1392, "eval_rundkast_samples_per_second": 40.586, "eval_rundkast_steps_per_second": 1.298, "eval_rundkast_wer": 0.24315940184537066, "step": 1508 }, { "epoch": 4.0, "eval_nb_samtale_loss": 0.5375664830207825, "eval_nb_samtale_runtime": 37.7821, "eval_nb_samtale_samples_per_second": 14.107, "eval_nb_samtale_steps_per_second": 0.45, "eval_nb_samtale_wer": 0.262939453125, "step": 1508 }, { "epoch": 4.0, "eval_bigbrother_loss": 1.2783823013305664, "eval_bigbrother_runtime": 41.1302, "eval_bigbrother_samples_per_second": 33.309, "eval_bigbrother_steps_per_second": 1.045, "eval_bigbrother_wer": 0.5196165310561508, "step": 1508 }, { "epoch": 5.0, "grad_norm": 5.716007709503174, "learning_rate": 9.717967599410898e-05, "loss": 1.7424, "step": 1885 }, { "epoch": 5.0, "eval_rundkast_loss": 0.40847668051719666, "eval_rundkast_runtime": 33.007, "eval_rundkast_samples_per_second": 40.749, "eval_rundkast_steps_per_second": 1.303, "eval_rundkast_wer": 0.23870505886096086, "step": 1885 }, { "epoch": 5.0, "eval_nb_samtale_loss": 0.5756209492683411, "eval_nb_samtale_runtime": 37.9326, "eval_nb_samtale_samples_per_second": 14.051, "eval_nb_samtale_steps_per_second": 0.448, "eval_nb_samtale_wer": 0.2607421875, "step": 1885 }, { "epoch": 5.0, "eval_bigbrother_loss": 1.3229293823242188, "eval_bigbrother_runtime": 40.9231, "eval_bigbrother_samples_per_second": 33.477, "eval_bigbrother_steps_per_second": 1.051, "eval_bigbrother_wer": 0.5068879400628373, "step": 1885 }, { "epoch": 6.0, "grad_norm": 4.037384510040283, "learning_rate": 9.440353460972018e-05, "loss": 1.6619, "step": 2262 }, { "epoch": 6.0, "eval_rundkast_loss": 0.4441112279891968, "eval_rundkast_runtime": 33.7921, "eval_rundkast_samples_per_second": 39.802, "eval_rundkast_steps_per_second": 1.272, "eval_rundkast_wer": 0.24315940184537066, "step": 2262 }, { "epoch": 6.0, "eval_nb_samtale_loss": 0.6132481694221497, "eval_nb_samtale_runtime": 38.4613, "eval_nb_samtale_samples_per_second": 13.858, "eval_nb_samtale_steps_per_second": 0.442, "eval_nb_samtale_wer": 0.2744140625, "step": 2262 }, { "epoch": 6.0, "eval_bigbrother_loss": 1.3384240865707397, "eval_bigbrother_runtime": 41.3107, "eval_bigbrother_samples_per_second": 33.163, "eval_bigbrother_steps_per_second": 1.041, "eval_bigbrother_wer": 0.5068879400628373, "step": 2262 }, { "epoch": 7.0, "grad_norm": 8.110424995422363, "learning_rate": 9.163475699558174e-05, "loss": 1.6139, "step": 2639 }, { "epoch": 7.0, "eval_rundkast_loss": 0.45466309785842896, "eval_rundkast_runtime": 33.255, "eval_rundkast_samples_per_second": 40.445, "eval_rundkast_steps_per_second": 1.293, "eval_rundkast_wer": 0.24713649379573654, "step": 2639 }, { "epoch": 7.0, "eval_nb_samtale_loss": 0.5841706991195679, "eval_nb_samtale_runtime": 38.3016, "eval_nb_samtale_samples_per_second": 13.916, "eval_nb_samtale_steps_per_second": 0.444, "eval_nb_samtale_wer": 0.2757975260416667, "step": 2639 }, { "epoch": 7.0, "eval_bigbrother_loss": 1.338714838027954, "eval_bigbrother_runtime": 41.6281, "eval_bigbrother_samples_per_second": 32.91, "eval_bigbrother_steps_per_second": 1.033, "eval_bigbrother_wer": 0.49625392733424634, "step": 2639 }, { "epoch": 8.0, "grad_norm": 19.2949275970459, "learning_rate": 8.885861561119293e-05, "loss": 1.5567, "step": 3016 }, { "epoch": 8.0, "eval_rundkast_loss": 0.45816686749458313, "eval_rundkast_runtime": 33.2799, "eval_rundkast_samples_per_second": 40.415, "eval_rundkast_steps_per_second": 1.292, "eval_rundkast_wer": 0.25095450206808784, "step": 3016 }, { "epoch": 8.0, "eval_nb_samtale_loss": 0.5909866690635681, "eval_nb_samtale_runtime": 37.7066, "eval_nb_samtale_samples_per_second": 14.135, "eval_nb_samtale_steps_per_second": 0.451, "eval_nb_samtale_wer": 0.2709147135416667, "step": 3016 }, { "epoch": 8.0, "eval_bigbrother_loss": 1.2419942617416382, "eval_bigbrother_runtime": 41.2436, "eval_bigbrother_samples_per_second": 33.217, "eval_bigbrother_steps_per_second": 1.043, "eval_bigbrother_wer": 0.49238701361475873, "step": 3016 }, { "epoch": 9.0, "grad_norm": 4.614249229431152, "learning_rate": 8.608247422680413e-05, "loss": 1.4942, "step": 3393 }, { "epoch": 9.0, "eval_rundkast_loss": 0.4503733217716217, "eval_rundkast_runtime": 33.2408, "eval_rundkast_samples_per_second": 40.462, "eval_rundkast_steps_per_second": 1.294, "eval_rundkast_wer": 0.2419662742602609, "step": 3393 }, { "epoch": 9.0, "eval_nb_samtale_loss": 0.5738528370857239, "eval_nb_samtale_runtime": 37.6697, "eval_nb_samtale_samples_per_second": 14.149, "eval_nb_samtale_steps_per_second": 0.451, "eval_nb_samtale_wer": 0.2655436197916667, "step": 3393 }, { "epoch": 9.0, "eval_bigbrother_loss": 1.207922101020813, "eval_bigbrother_runtime": 41.3161, "eval_bigbrother_samples_per_second": 33.159, "eval_bigbrother_steps_per_second": 1.041, "eval_bigbrother_wer": 0.4722468379924273, "step": 3393 }, { "epoch": 10.0, "grad_norm": 10.218366622924805, "learning_rate": 8.330633284241532e-05, "loss": 1.4412, "step": 3770 }, { "epoch": 10.0, "eval_rundkast_loss": 0.4589375853538513, "eval_rundkast_runtime": 33.4846, "eval_rundkast_samples_per_second": 40.168, "eval_rundkast_steps_per_second": 1.284, "eval_rundkast_wer": 0.2531816735602927, "step": 3770 }, { "epoch": 10.0, "eval_nb_samtale_loss": 0.5715627670288086, "eval_nb_samtale_runtime": 37.8642, "eval_nb_samtale_samples_per_second": 14.077, "eval_nb_samtale_steps_per_second": 0.449, "eval_nb_samtale_wer": 0.2689615885416667, "step": 3770 }, { "epoch": 10.0, "eval_bigbrother_loss": 1.1500192880630493, "eval_bigbrother_runtime": 41.3376, "eval_bigbrother_samples_per_second": 33.142, "eval_bigbrother_steps_per_second": 1.04, "eval_bigbrother_wer": 0.46588254249577055, "step": 3770 }, { "epoch": 11.0, "grad_norm": 3.9760067462921143, "learning_rate": 8.053019145802652e-05, "loss": 1.4267, "step": 4147 }, { "epoch": 11.0, "eval_rundkast_loss": 0.4483039081096649, "eval_rundkast_runtime": 33.8508, "eval_rundkast_samples_per_second": 39.733, "eval_rundkast_steps_per_second": 1.27, "eval_rundkast_wer": 0.25015908367801465, "step": 4147 }, { "epoch": 11.0, "eval_nb_samtale_loss": 0.5839167237281799, "eval_nb_samtale_runtime": 38.2514, "eval_nb_samtale_samples_per_second": 13.934, "eval_nb_samtale_steps_per_second": 0.444, "eval_nb_samtale_wer": 0.2657063802083333, "step": 4147 }, { "epoch": 11.0, "eval_bigbrother_loss": 1.246797800064087, "eval_bigbrother_runtime": 41.8091, "eval_bigbrother_samples_per_second": 32.768, "eval_bigbrother_steps_per_second": 1.028, "eval_bigbrother_wer": 0.45669862241198744, "step": 4147 }, { "epoch": 12.0, "grad_norm": 4.7413763999938965, "learning_rate": 7.77540500736377e-05, "loss": 1.3792, "step": 4524 }, { "epoch": 12.0, "eval_rundkast_loss": 0.4521505832672119, "eval_rundkast_runtime": 33.7711, "eval_rundkast_samples_per_second": 39.827, "eval_rundkast_steps_per_second": 1.273, "eval_rundkast_wer": 0.24872733057588292, "step": 4524 }, { "epoch": 12.0, "eval_nb_samtale_loss": 0.5718214511871338, "eval_nb_samtale_runtime": 37.7574, "eval_nb_samtale_samples_per_second": 14.116, "eval_nb_samtale_steps_per_second": 0.45, "eval_nb_samtale_wer": 0.2674967447916667, "step": 4524 }, { "epoch": 12.0, "eval_bigbrother_loss": 1.1487047672271729, "eval_bigbrother_runtime": 41.46, "eval_bigbrother_samples_per_second": 33.044, "eval_bigbrother_steps_per_second": 1.037, "eval_bigbrother_wer": 0.4530733907999678, "step": 4524 }, { "epoch": 13.0, "grad_norm": 5.937889099121094, "learning_rate": 7.49779086892489e-05, "loss": 1.3269, "step": 4901 }, { "epoch": 13.0, "eval_rundkast_loss": 0.483146995306015, "eval_rundkast_runtime": 33.3218, "eval_rundkast_samples_per_second": 40.364, "eval_rundkast_steps_per_second": 1.29, "eval_rundkast_wer": 0.26002227171492204, "step": 4901 }, { "epoch": 13.0, "eval_nb_samtale_loss": 0.6045836806297302, "eval_nb_samtale_runtime": 37.3201, "eval_nb_samtale_samples_per_second": 14.282, "eval_nb_samtale_steps_per_second": 0.456, "eval_nb_samtale_wer": 0.2681477864583333, "step": 4901 }, { "epoch": 13.0, "eval_bigbrother_loss": 1.1842252016067505, "eval_bigbrother_runtime": 41.1883, "eval_bigbrother_samples_per_second": 33.262, "eval_bigbrother_steps_per_second": 1.044, "eval_bigbrother_wer": 0.45057600902279865, "step": 4901 }, { "epoch": 14.0, "grad_norm": 6.761897087097168, "learning_rate": 7.22017673048601e-05, "loss": 1.2988, "step": 5278 }, { "epoch": 14.0, "eval_rundkast_loss": 0.5177704691886902, "eval_rundkast_runtime": 33.7773, "eval_rundkast_samples_per_second": 39.82, "eval_rundkast_steps_per_second": 1.273, "eval_rundkast_wer": 0.24443207126948774, "step": 5278 }, { "epoch": 14.0, "eval_nb_samtale_loss": 0.6508249044418335, "eval_nb_samtale_runtime": 37.9055, "eval_nb_samtale_samples_per_second": 14.061, "eval_nb_samtale_steps_per_second": 0.448, "eval_nb_samtale_wer": 0.2630208333333333, "step": 5278 }, { "epoch": 14.0, "eval_bigbrother_loss": 1.2824336290359497, "eval_bigbrother_runtime": 41.3751, "eval_bigbrother_samples_per_second": 33.112, "eval_bigbrother_steps_per_second": 1.039, "eval_bigbrother_wer": 0.4433255457987594, "step": 5278 }, { "epoch": 15.0, "grad_norm": 8.300821304321289, "learning_rate": 6.942562592047128e-05, "loss": 1.2819, "step": 5655 }, { "epoch": 15.0, "eval_rundkast_loss": 0.47504347562789917, "eval_rundkast_runtime": 33.2023, "eval_rundkast_samples_per_second": 40.509, "eval_rundkast_steps_per_second": 1.295, "eval_rundkast_wer": 0.24737511931275852, "step": 5655 }, { "epoch": 15.0, "eval_nb_samtale_loss": 0.6060231924057007, "eval_nb_samtale_runtime": 38.0225, "eval_nb_samtale_samples_per_second": 14.018, "eval_nb_samtale_steps_per_second": 0.447, "eval_nb_samtale_wer": 0.263427734375, "step": 5655 }, { "epoch": 15.0, "eval_bigbrother_loss": 1.193629503250122, "eval_bigbrother_runtime": 41.725, "eval_bigbrother_samples_per_second": 32.834, "eval_bigbrother_steps_per_second": 1.031, "eval_bigbrother_wer": 0.4493675984854588, "step": 5655 }, { "epoch": 16.0, "grad_norm": 6.88249397277832, "learning_rate": 6.665684830633285e-05, "loss": 1.2504, "step": 6032 }, { "epoch": 16.0, "eval_rundkast_loss": 0.5179128646850586, "eval_rundkast_runtime": 33.3288, "eval_rundkast_samples_per_second": 40.356, "eval_rundkast_steps_per_second": 1.29, "eval_rundkast_wer": 0.24856824689786827, "step": 6032 }, { "epoch": 16.0, "eval_nb_samtale_loss": 0.653479814529419, "eval_nb_samtale_runtime": 37.2804, "eval_nb_samtale_samples_per_second": 14.297, "eval_nb_samtale_steps_per_second": 0.456, "eval_nb_samtale_wer": 0.2644856770833333, "step": 6032 }, { "epoch": 16.0, "eval_bigbrother_loss": 1.2211058139801025, "eval_bigbrother_runtime": 42.0179, "eval_bigbrother_samples_per_second": 32.605, "eval_bigbrother_steps_per_second": 1.023, "eval_bigbrother_wer": 0.43792797873197453, "step": 6032 }, { "epoch": 17.0, "grad_norm": 3.6763479709625244, "learning_rate": 6.388070692194403e-05, "loss": 1.2295, "step": 6409 }, { "epoch": 17.0, "eval_rundkast_loss": 0.4937501847743988, "eval_rundkast_runtime": 37.7386, "eval_rundkast_samples_per_second": 35.64, "eval_rundkast_steps_per_second": 1.139, "eval_rundkast_wer": 0.2561247216035635, "step": 6409 }, { "epoch": 17.0, "eval_nb_samtale_loss": 0.6216253638267517, "eval_nb_samtale_runtime": 38.691, "eval_nb_samtale_samples_per_second": 13.776, "eval_nb_samtale_steps_per_second": 0.439, "eval_nb_samtale_wer": 0.2705078125, "step": 6409 }, { "epoch": 17.0, "eval_bigbrother_loss": 1.188571572303772, "eval_bigbrother_runtime": 41.7402, "eval_bigbrother_samples_per_second": 32.822, "eval_bigbrother_steps_per_second": 1.03, "eval_bigbrother_wer": 0.4334165793925723, "step": 6409 }, { "epoch": 18.0, "grad_norm": 5.536041259765625, "learning_rate": 6.110456553755524e-05, "loss": 1.2053, "step": 6786 }, { "epoch": 18.0, "eval_rundkast_loss": 0.4938836097717285, "eval_rundkast_runtime": 33.2744, "eval_rundkast_samples_per_second": 40.421, "eval_rundkast_steps_per_second": 1.292, "eval_rundkast_wer": 0.24856824689786827, "step": 6786 }, { "epoch": 18.0, "eval_nb_samtale_loss": 0.6238839030265808, "eval_nb_samtale_runtime": 37.8912, "eval_nb_samtale_samples_per_second": 14.067, "eval_nb_samtale_steps_per_second": 0.449, "eval_nb_samtale_wer": 0.263916015625, "step": 6786 }, { "epoch": 18.0, "eval_bigbrother_loss": 1.1581498384475708, "eval_bigbrother_runtime": 41.4842, "eval_bigbrother_samples_per_second": 33.025, "eval_bigbrother_steps_per_second": 1.037, "eval_bigbrother_wer": 0.4314831225328285, "step": 6786 }, { "epoch": 19.0, "grad_norm": 5.267818927764893, "learning_rate": 5.832842415316643e-05, "loss": 1.1815, "step": 7163 }, { "epoch": 19.0, "eval_rundkast_loss": 0.4980849623680115, "eval_rundkast_runtime": 33.6894, "eval_rundkast_samples_per_second": 39.923, "eval_rundkast_steps_per_second": 1.276, "eval_rundkast_wer": 0.24451161310849506, "step": 7163 }, { "epoch": 19.0, "eval_nb_samtale_loss": 0.6296201348304749, "eval_nb_samtale_runtime": 37.5618, "eval_nb_samtale_samples_per_second": 14.19, "eval_nb_samtale_steps_per_second": 0.453, "eval_nb_samtale_wer": 0.2608235677083333, "step": 7163 }, { "epoch": 19.0, "eval_bigbrother_loss": 1.1892309188842773, "eval_bigbrother_runtime": 41.4844, "eval_bigbrother_samples_per_second": 33.024, "eval_bigbrother_steps_per_second": 1.037, "eval_bigbrother_wer": 0.4253605091436397, "step": 7163 }, { "epoch": 20.0, "grad_norm": 41.6290283203125, "learning_rate": 5.5559646539027985e-05, "loss": 1.1703, "step": 7540 }, { "epoch": 20.0, "eval_rundkast_loss": 0.5075950026512146, "eval_rundkast_runtime": 33.2501, "eval_rundkast_samples_per_second": 40.451, "eval_rundkast_steps_per_second": 1.293, "eval_rundkast_wer": 0.2478523703468024, "step": 7540 }, { "epoch": 20.0, "eval_nb_samtale_loss": 0.6342408061027527, "eval_nb_samtale_runtime": 37.4134, "eval_nb_samtale_samples_per_second": 14.246, "eval_nb_samtale_steps_per_second": 0.454, "eval_nb_samtale_wer": 0.2586263020833333, "step": 7540 }, { "epoch": 20.0, "eval_bigbrother_loss": 1.206842303276062, "eval_bigbrother_runtime": 41.5395, "eval_bigbrother_samples_per_second": 32.981, "eval_bigbrother_steps_per_second": 1.035, "eval_bigbrother_wer": 0.42519938773866106, "step": 7540 }, { "epoch": 21.0, "grad_norm": 6.965384483337402, "learning_rate": 5.278350515463918e-05, "loss": 1.1446, "step": 7917 }, { "epoch": 21.0, "eval_rundkast_loss": 0.5136203169822693, "eval_rundkast_runtime": 33.5002, "eval_rundkast_samples_per_second": 40.149, "eval_rundkast_steps_per_second": 1.284, "eval_rundkast_wer": 0.24793191218580973, "step": 7917 }, { "epoch": 21.0, "eval_nb_samtale_loss": 0.6333425641059875, "eval_nb_samtale_runtime": 37.5282, "eval_nb_samtale_samples_per_second": 14.203, "eval_nb_samtale_steps_per_second": 0.453, "eval_nb_samtale_wer": 0.265625, "step": 7917 }, { "epoch": 21.0, "eval_bigbrother_loss": 1.186505913734436, "eval_bigbrother_runtime": 41.5636, "eval_bigbrother_samples_per_second": 32.962, "eval_bigbrother_steps_per_second": 1.035, "eval_bigbrother_wer": 0.42487714492870376, "step": 7917 }, { "epoch": 22.0, "grad_norm": 22.68753433227539, "learning_rate": 5.000736377025037e-05, "loss": 1.1384, "step": 8294 }, { "epoch": 22.0, "eval_rundkast_loss": 0.5014224052429199, "eval_rundkast_runtime": 33.6382, "eval_rundkast_samples_per_second": 39.984, "eval_rundkast_steps_per_second": 1.278, "eval_rundkast_wer": 0.24880687241489025, "step": 8294 }, { "epoch": 22.0, "eval_nb_samtale_loss": 0.6214331984519958, "eval_nb_samtale_runtime": 37.41, "eval_nb_samtale_samples_per_second": 14.248, "eval_nb_samtale_steps_per_second": 0.454, "eval_nb_samtale_wer": 0.2609049479166667, "step": 8294 }, { "epoch": 22.0, "eval_bigbrother_loss": 1.1910523176193237, "eval_bigbrother_runtime": 41.6527, "eval_bigbrother_samples_per_second": 32.891, "eval_bigbrother_steps_per_second": 1.032, "eval_bigbrother_wer": 0.42189639893659875, "step": 8294 }, { "epoch": 23.0, "grad_norm": 2.731438398361206, "learning_rate": 4.723122238586156e-05, "loss": 1.1324, "step": 8671 }, { "epoch": 23.0, "eval_rundkast_loss": 0.476345032453537, "eval_rundkast_runtime": 33.4012, "eval_rundkast_samples_per_second": 40.268, "eval_rundkast_steps_per_second": 1.287, "eval_rundkast_wer": 0.24968183264397073, "step": 8671 }, { "epoch": 23.0, "eval_nb_samtale_loss": 0.6042336821556091, "eval_nb_samtale_runtime": 37.9587, "eval_nb_samtale_samples_per_second": 14.042, "eval_nb_samtale_steps_per_second": 0.448, "eval_nb_samtale_wer": 0.2681477864583333, "step": 8671 }, { "epoch": 23.0, "eval_bigbrother_loss": 1.1591678857803345, "eval_bigbrother_runtime": 41.5785, "eval_bigbrother_samples_per_second": 32.95, "eval_bigbrother_steps_per_second": 1.034, "eval_bigbrother_wer": 0.4180294852171111, "step": 8671 }, { "epoch": 24.0, "grad_norm": 4.182296276092529, "learning_rate": 4.4455081001472755e-05, "loss": 1.0927, "step": 9048 }, { "epoch": 24.0, "eval_rundkast_loss": 0.49040549993515015, "eval_rundkast_runtime": 33.4922, "eval_rundkast_samples_per_second": 40.159, "eval_rundkast_steps_per_second": 1.284, "eval_rundkast_wer": 0.2406936048361438, "step": 9048 }, { "epoch": 24.0, "eval_nb_samtale_loss": 0.6173272132873535, "eval_nb_samtale_runtime": 37.5762, "eval_nb_samtale_samples_per_second": 14.184, "eval_nb_samtale_steps_per_second": 0.452, "eval_nb_samtale_wer": 0.26025390625, "step": 9048 }, { "epoch": 24.0, "eval_bigbrother_loss": 1.1952488422393799, "eval_bigbrother_runtime": 41.4629, "eval_bigbrother_samples_per_second": 33.042, "eval_bigbrother_steps_per_second": 1.037, "eval_bigbrother_wer": 0.41786836381213244, "step": 9048 }, { "epoch": 25.0, "grad_norm": 7.412995338439941, "learning_rate": 4.167893961708395e-05, "loss": 1.0897, "step": 9425 }, { "epoch": 25.0, "eval_rundkast_loss": 0.5106588006019592, "eval_rundkast_runtime": 33.3541, "eval_rundkast_samples_per_second": 40.325, "eval_rundkast_steps_per_second": 1.289, "eval_rundkast_wer": 0.25206808781419027, "step": 9425 }, { "epoch": 25.0, "eval_nb_samtale_loss": 0.6364408731460571, "eval_nb_samtale_runtime": 37.3682, "eval_nb_samtale_samples_per_second": 14.263, "eval_nb_samtale_steps_per_second": 0.455, "eval_nb_samtale_wer": 0.2630208333333333, "step": 9425 }, { "epoch": 25.0, "eval_bigbrother_loss": 1.2089372873306274, "eval_bigbrother_runtime": 41.0339, "eval_bigbrother_samples_per_second": 33.387, "eval_bigbrother_steps_per_second": 1.048, "eval_bigbrother_wer": 0.4167405139772819, "step": 9425 }, { "epoch": 26.0, "grad_norm": 20.474260330200195, "learning_rate": 3.890279823269514e-05, "loss": 1.0818, "step": 9802 }, { "epoch": 26.0, "eval_rundkast_loss": 0.49753764271736145, "eval_rundkast_runtime": 33.4596, "eval_rundkast_samples_per_second": 40.198, "eval_rundkast_steps_per_second": 1.285, "eval_rundkast_wer": 0.24562519885459752, "step": 9802 }, { "epoch": 26.0, "eval_nb_samtale_loss": 0.6342372298240662, "eval_nb_samtale_runtime": 37.5602, "eval_nb_samtale_samples_per_second": 14.191, "eval_nb_samtale_steps_per_second": 0.453, "eval_nb_samtale_wer": 0.260009765625, "step": 9802 }, { "epoch": 26.0, "eval_bigbrother_loss": 1.175524115562439, "eval_bigbrother_runtime": 41.4743, "eval_bigbrother_samples_per_second": 33.033, "eval_bigbrother_steps_per_second": 1.037, "eval_bigbrother_wer": 0.4159349069523886, "step": 9802 }, { "epoch": 27.0, "grad_norm": 3.3909595012664795, "learning_rate": 3.6126656848306336e-05, "loss": 1.0629, "step": 10179 }, { "epoch": 27.0, "eval_rundkast_loss": 0.5271292924880981, "eval_rundkast_runtime": 33.4554, "eval_rundkast_samples_per_second": 40.203, "eval_rundkast_steps_per_second": 1.285, "eval_rundkast_wer": 0.24347756920139993, "step": 10179 }, { "epoch": 27.0, "eval_nb_samtale_loss": 0.6458906531333923, "eval_nb_samtale_runtime": 37.3547, "eval_nb_samtale_samples_per_second": 14.269, "eval_nb_samtale_steps_per_second": 0.455, "eval_nb_samtale_wer": 0.2598470052083333, "step": 10179 }, { "epoch": 27.0, "eval_bigbrother_loss": 1.1968939304351807, "eval_bigbrother_runtime": 41.3751, "eval_bigbrother_samples_per_second": 33.112, "eval_bigbrother_steps_per_second": 1.039, "eval_bigbrother_wer": 0.4142431322001128, "step": 10179 }, { "epoch": 28.0, "grad_norm": 19.91695213317871, "learning_rate": 3.335787923416789e-05, "loss": 1.069, "step": 10556 }, { "epoch": 28.0, "eval_rundkast_loss": 0.5136268138885498, "eval_rundkast_runtime": 33.4249, "eval_rundkast_samples_per_second": 40.24, "eval_rundkast_steps_per_second": 1.286, "eval_rundkast_wer": 0.2433980273623926, "step": 10556 }, { "epoch": 28.0, "eval_nb_samtale_loss": 0.6378623247146606, "eval_nb_samtale_runtime": 37.4155, "eval_nb_samtale_samples_per_second": 14.245, "eval_nb_samtale_steps_per_second": 0.454, "eval_nb_samtale_wer": 0.25927734375, "step": 10556 }, { "epoch": 28.0, "eval_bigbrother_loss": 1.1700557470321655, "eval_bigbrother_runtime": 41.488, "eval_bigbrother_samples_per_second": 33.022, "eval_bigbrother_steps_per_second": 1.036, "eval_bigbrother_wer": 0.41102070410053976, "step": 10556 }, { "epoch": 29.0, "grad_norm": 4.6988959312438965, "learning_rate": 3.058173784977909e-05, "loss": 1.0561, "step": 10933 }, { "epoch": 29.0, "eval_rundkast_loss": 0.5006275773048401, "eval_rundkast_runtime": 33.4079, "eval_rundkast_samples_per_second": 40.26, "eval_rundkast_steps_per_second": 1.287, "eval_rundkast_wer": 0.24387527839643652, "step": 10933 }, { "epoch": 29.0, "eval_nb_samtale_loss": 0.6258318424224854, "eval_nb_samtale_runtime": 37.2465, "eval_nb_samtale_samples_per_second": 14.31, "eval_nb_samtale_steps_per_second": 0.456, "eval_nb_samtale_wer": 0.258056640625, "step": 10933 }, { "epoch": 29.0, "eval_bigbrother_loss": 1.1609516143798828, "eval_bigbrother_runtime": 41.9041, "eval_bigbrother_samples_per_second": 32.694, "eval_bigbrother_steps_per_second": 1.026, "eval_bigbrother_wer": 0.4105373398856038, "step": 10933 }, { "epoch": 30.0, "grad_norm": 6.915459156036377, "learning_rate": 2.7805596465390282e-05, "loss": 1.046, "step": 11310 }, { "epoch": 30.0, "eval_rundkast_loss": 0.5193740725517273, "eval_rundkast_runtime": 33.0243, "eval_rundkast_samples_per_second": 40.728, "eval_rundkast_steps_per_second": 1.302, "eval_rundkast_wer": 0.24530703149856825, "step": 11310 }, { "epoch": 30.0, "eval_nb_samtale_loss": 0.6532759666442871, "eval_nb_samtale_runtime": 37.2211, "eval_nb_samtale_samples_per_second": 14.32, "eval_nb_samtale_steps_per_second": 0.457, "eval_nb_samtale_wer": 0.260986328125, "step": 11310 }, { "epoch": 30.0, "eval_bigbrother_loss": 1.2057961225509644, "eval_bigbrother_runtime": 41.7437, "eval_bigbrother_samples_per_second": 32.819, "eval_bigbrother_steps_per_second": 1.03, "eval_bigbrother_wer": 0.4085233223233707, "step": 11310 }, { "epoch": 31.0, "grad_norm": 6.216676712036133, "learning_rate": 2.5029455081001475e-05, "loss": 1.0244, "step": 11687 }, { "epoch": 31.0, "eval_rundkast_loss": 0.5112914443016052, "eval_rundkast_runtime": 33.3775, "eval_rundkast_samples_per_second": 40.297, "eval_rundkast_steps_per_second": 1.288, "eval_rundkast_wer": 0.24347756920139993, "step": 11687 }, { "epoch": 31.0, "eval_nb_samtale_loss": 0.637014627456665, "eval_nb_samtale_runtime": 37.0028, "eval_nb_samtale_samples_per_second": 14.404, "eval_nb_samtale_steps_per_second": 0.459, "eval_nb_samtale_wer": 0.2584635416666667, "step": 11687 }, { "epoch": 31.0, "eval_bigbrother_loss": 1.196700096130371, "eval_bigbrother_runtime": 41.5693, "eval_bigbrother_samples_per_second": 32.957, "eval_bigbrother_steps_per_second": 1.034, "eval_bigbrother_wer": 0.40699266897607345, "step": 11687 }, { "epoch": 32.0, "grad_norm": 5.768988609313965, "learning_rate": 2.2253313696612664e-05, "loss": 1.0211, "step": 12064 }, { "epoch": 32.0, "eval_rundkast_loss": 0.5140534043312073, "eval_rundkast_runtime": 33.4214, "eval_rundkast_samples_per_second": 40.244, "eval_rundkast_steps_per_second": 1.287, "eval_rundkast_wer": 0.24307986000636334, "step": 12064 }, { "epoch": 32.0, "eval_nb_samtale_loss": 0.6492373943328857, "eval_nb_samtale_runtime": 37.5056, "eval_nb_samtale_samples_per_second": 14.211, "eval_nb_samtale_steps_per_second": 0.453, "eval_nb_samtale_wer": 0.2571614583333333, "step": 12064 }, { "epoch": 32.0, "eval_bigbrother_loss": 1.1793997287750244, "eval_bigbrother_runtime": 41.1907, "eval_bigbrother_samples_per_second": 33.26, "eval_bigbrother_steps_per_second": 1.044, "eval_bigbrother_wer": 0.40812051881092404, "step": 12064 }, { "epoch": 33.0, "grad_norm": 8.210163116455078, "learning_rate": 1.947717231222386e-05, "loss": 1.0221, "step": 12441 }, { "epoch": 33.0, "eval_rundkast_loss": 0.5195760726928711, "eval_rundkast_runtime": 33.4259, "eval_rundkast_samples_per_second": 40.238, "eval_rundkast_steps_per_second": 1.286, "eval_rundkast_wer": 0.24188673242125358, "step": 12441 }, { "epoch": 33.0, "eval_nb_samtale_loss": 0.6468714475631714, "eval_nb_samtale_runtime": 37.1982, "eval_nb_samtale_samples_per_second": 14.329, "eval_nb_samtale_steps_per_second": 0.457, "eval_nb_samtale_wer": 0.257568359375, "step": 12441 }, { "epoch": 33.0, "eval_bigbrother_loss": 1.1880689859390259, "eval_bigbrother_runtime": 41.3876, "eval_bigbrother_samples_per_second": 33.102, "eval_bigbrother_steps_per_second": 1.039, "eval_bigbrother_wer": 0.40747603319100945, "step": 12441 }, { "epoch": 34.0, "grad_norm": 4.438199043273926, "learning_rate": 1.670839469808542e-05, "loss": 1.0095, "step": 12818 }, { "epoch": 34.0, "eval_rundkast_loss": 0.5272213816642761, "eval_rundkast_runtime": 33.3742, "eval_rundkast_samples_per_second": 40.301, "eval_rundkast_steps_per_second": 1.288, "eval_rundkast_wer": 0.24212535793827553, "step": 12818 }, { "epoch": 34.0, "eval_nb_samtale_loss": 0.6708551645278931, "eval_nb_samtale_runtime": 37.2593, "eval_nb_samtale_samples_per_second": 14.305, "eval_nb_samtale_steps_per_second": 0.456, "eval_nb_samtale_wer": 0.2578125, "step": 12818 }, { "epoch": 34.0, "eval_bigbrother_loss": 1.2168066501617432, "eval_bigbrother_runtime": 41.3249, "eval_bigbrother_samples_per_second": 33.152, "eval_bigbrother_steps_per_second": 1.041, "eval_bigbrother_wer": 0.4064287440586482, "step": 12818 }, { "epoch": 35.0, "grad_norm": 18.065574645996094, "learning_rate": 1.3932253313696614e-05, "loss": 1.0011, "step": 13195 }, { "epoch": 35.0, "eval_rundkast_loss": 0.5081976652145386, "eval_rundkast_runtime": 33.3601, "eval_rundkast_samples_per_second": 40.318, "eval_rundkast_steps_per_second": 1.289, "eval_rundkast_wer": 0.2410117721921731, "step": 13195 }, { "epoch": 35.0, "eval_nb_samtale_loss": 0.6458988189697266, "eval_nb_samtale_runtime": 37.6087, "eval_nb_samtale_samples_per_second": 14.172, "eval_nb_samtale_steps_per_second": 0.452, "eval_nb_samtale_wer": 0.257080078125, "step": 13195 }, { "epoch": 35.0, "eval_bigbrother_loss": 1.1833205223083496, "eval_bigbrother_runtime": 41.2758, "eval_bigbrother_samples_per_second": 33.191, "eval_bigbrother_steps_per_second": 1.042, "eval_bigbrother_wer": 0.40667042616611615, "step": 13195 }, { "epoch": 36.0, "grad_norm": 14.747867584228516, "learning_rate": 1.1156111929307807e-05, "loss": 0.9849, "step": 13572 }, { "epoch": 36.0, "eval_rundkast_loss": 0.5170104503631592, "eval_rundkast_runtime": 33.5036, "eval_rundkast_samples_per_second": 40.145, "eval_rundkast_steps_per_second": 1.283, "eval_rundkast_wer": 0.24140948138720966, "step": 13572 }, { "epoch": 36.0, "eval_nb_samtale_loss": 0.6568956971168518, "eval_nb_samtale_runtime": 37.0353, "eval_nb_samtale_samples_per_second": 14.392, "eval_nb_samtale_steps_per_second": 0.459, "eval_nb_samtale_wer": 0.2569173177083333, "step": 13572 }, { "epoch": 36.0, "eval_bigbrother_loss": 1.2056487798690796, "eval_bigbrother_runtime": 41.6456, "eval_bigbrother_samples_per_second": 32.897, "eval_bigbrother_steps_per_second": 1.033, "eval_bigbrother_wer": 0.404656408603883, "step": 13572 }, { "epoch": 37.0, "grad_norm": 4.995520114898682, "learning_rate": 8.379970544918998e-06, "loss": 0.9785, "step": 13949 }, { "epoch": 37.0, "eval_rundkast_loss": 0.5311599969863892, "eval_rundkast_runtime": 33.4582, "eval_rundkast_samples_per_second": 40.199, "eval_rundkast_steps_per_second": 1.285, "eval_rundkast_wer": 0.2419662742602609, "step": 13949 }, { "epoch": 37.0, "eval_nb_samtale_loss": 0.6731538772583008, "eval_nb_samtale_runtime": 37.2532, "eval_nb_samtale_samples_per_second": 14.307, "eval_nb_samtale_steps_per_second": 0.456, "eval_nb_samtale_wer": 0.2586263020833333, "step": 13949 }, { "epoch": 37.0, "eval_bigbrother_loss": 1.2453778982162476, "eval_bigbrother_runtime": 41.2503, "eval_bigbrother_samples_per_second": 33.212, "eval_bigbrother_steps_per_second": 1.042, "eval_bigbrother_wer": 0.40667042616611615, "step": 13949 }, { "epoch": 38.0, "grad_norm": 4.370805740356445, "learning_rate": 5.603829160530192e-06, "loss": 0.9853, "step": 14326 }, { "epoch": 38.0, "eval_rundkast_loss": 0.514400064945221, "eval_rundkast_runtime": 33.9777, "eval_rundkast_samples_per_second": 39.585, "eval_rundkast_steps_per_second": 1.266, "eval_rundkast_wer": 0.2415685650652243, "step": 14326 }, { "epoch": 38.0, "eval_nb_samtale_loss": 0.6526333689689636, "eval_nb_samtale_runtime": 37.0789, "eval_nb_samtale_samples_per_second": 14.375, "eval_nb_samtale_steps_per_second": 0.458, "eval_nb_samtale_wer": 0.2571614583333333, "step": 14326 }, { "epoch": 38.0, "eval_bigbrother_loss": 1.2180824279785156, "eval_bigbrother_runtime": 41.4647, "eval_bigbrother_samples_per_second": 33.04, "eval_bigbrother_steps_per_second": 1.037, "eval_bigbrother_wer": 0.4073954724885201, "step": 14326 }, { "epoch": 39.0, "grad_norm": 7.822154998779297, "learning_rate": 2.8350515463917527e-06, "loss": 0.9714, "step": 14703 }, { "epoch": 39.0, "eval_rundkast_loss": 0.5187526345252991, "eval_rundkast_runtime": 33.0413, "eval_rundkast_samples_per_second": 40.707, "eval_rundkast_steps_per_second": 1.301, "eval_rundkast_wer": 0.24268215081132677, "step": 14703 }, { "epoch": 39.0, "eval_nb_samtale_loss": 0.6581916809082031, "eval_nb_samtale_runtime": 37.1203, "eval_nb_samtale_samples_per_second": 14.359, "eval_nb_samtale_steps_per_second": 0.458, "eval_nb_samtale_wer": 0.257568359375, "step": 14703 }, { "epoch": 39.0, "eval_bigbrother_loss": 1.2226907014846802, "eval_bigbrother_runtime": 41.2361, "eval_bigbrother_samples_per_second": 33.223, "eval_bigbrother_steps_per_second": 1.043, "eval_bigbrother_wer": 0.404898090711351, "step": 14703 }, { "epoch": 40.0, "grad_norm": 5.108997821807861, "learning_rate": 5.8910162002945506e-08, "loss": 0.9743, "step": 15080 }, { "epoch": 40.0, "eval_rundkast_loss": 0.5217476487159729, "eval_rundkast_runtime": 33.378, "eval_rundkast_samples_per_second": 40.296, "eval_rundkast_steps_per_second": 1.288, "eval_rundkast_wer": 0.2423639834552975, "step": 15080 }, { "epoch": 40.0, "eval_nb_samtale_loss": 0.6616787314414978, "eval_nb_samtale_runtime": 37.4121, "eval_nb_samtale_samples_per_second": 14.247, "eval_nb_samtale_steps_per_second": 0.454, "eval_nb_samtale_wer": 0.2579752604166667, "step": 15080 }, { "epoch": 40.0, "eval_bigbrother_loss": 1.2213834524154663, "eval_bigbrother_runtime": 41.9733, "eval_bigbrother_samples_per_second": 32.64, "eval_bigbrother_steps_per_second": 1.024, "eval_bigbrother_wer": 0.40538145492628697, "step": 15080 } ], "logging_steps": 500, "max_steps": 15080, "num_input_tokens_seen": 0, "num_train_epochs": 40, "save_steps": 500, "total_flos": 8.278376708995564e+19, "train_batch_size": 48, "trial_name": null, "trial_params": null }