| { | |
| "best_metric": 0.14368463395012068, | |
| "best_model_checkpoint": "/cluster/home/torstefl/Master/saved_model/W2V/Combined/NB-RUND/22.05/checkpoint-48581", | |
| "epoch": 37.0, | |
| "eval_steps": 500, | |
| "global_step": 48581, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 3.1965060234069824, | |
| "learning_rate": 8.740000000000001e-05, | |
| "loss": 0.5607, | |
| "step": 1313 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_bigbrother_loss": 1.985207200050354, | |
| "eval_bigbrother_runtime": 42.9617, | |
| "eval_bigbrother_samples_per_second": 31.889, | |
| "eval_bigbrother_steps_per_second": 1.001, | |
| "eval_bigbrother_wer": 0.5811649077579957, | |
| "step": 1313 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_NB_RUND_loss": 0.3991325795650482, | |
| "eval_NB_RUND_runtime": 118.1565, | |
| "eval_NB_RUND_samples_per_second": 15.894, | |
| "eval_NB_RUND_steps_per_second": 0.499, | |
| "eval_NB_RUND_wer": 0.21319388576025744, | |
| "step": 1313 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_rundkast_loss": 0.3412451446056366, | |
| "eval_rundkast_runtime": 33.4748, | |
| "eval_rundkast_samples_per_second": 40.18, | |
| "eval_rundkast_steps_per_second": 1.285, | |
| "eval_rundkast_wer": 0.1870824053452116, | |
| "step": 1313 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_nb_samtale_loss": 0.5455772876739502, | |
| "eval_nb_samtale_runtime": 38.7172, | |
| "eval_nb_samtale_samples_per_second": 13.766, | |
| "eval_nb_samtale_steps_per_second": 0.439, | |
| "eval_nb_samtale_wer": 0.23990885416666666, | |
| "step": 1313 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 4.949782371520996, | |
| "learning_rate": 9.779694237553901e-05, | |
| "loss": 0.4692, | |
| "step": 2626 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_bigbrother_loss": 1.9343265295028687, | |
| "eval_bigbrother_runtime": 41.3509, | |
| "eval_bigbrother_samples_per_second": 33.131, | |
| "eval_bigbrother_steps_per_second": 1.04, | |
| "eval_bigbrother_wer": 0.5653750100700878, | |
| "step": 2626 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_NB_RUND_loss": 0.3552016317844391, | |
| "eval_NB_RUND_runtime": 86.1673, | |
| "eval_NB_RUND_samples_per_second": 21.795, | |
| "eval_NB_RUND_steps_per_second": 0.685, | |
| "eval_NB_RUND_wer": 0.1984714400643604, | |
| "step": 2626 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_rundkast_loss": 0.29250362515449524, | |
| "eval_rundkast_runtime": 33.3182, | |
| "eval_rundkast_samples_per_second": 40.368, | |
| "eval_rundkast_steps_per_second": 1.291, | |
| "eval_rundkast_wer": 0.17181037225580656, | |
| "step": 2626 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_nb_samtale_loss": 0.5133547782897949, | |
| "eval_nb_samtale_runtime": 37.6446, | |
| "eval_nb_samtale_samples_per_second": 14.159, | |
| "eval_nb_samtale_steps_per_second": 0.452, | |
| "eval_nb_samtale_wer": 0.224609375, | |
| "step": 2626 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 2.300737142562866, | |
| "learning_rate": 9.522344178753431e-05, | |
| "loss": 0.4264, | |
| "step": 3939 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_bigbrother_loss": 1.831033706665039, | |
| "eval_bigbrother_runtime": 41.3456, | |
| "eval_bigbrother_samples_per_second": 33.135, | |
| "eval_bigbrother_steps_per_second": 1.04, | |
| "eval_bigbrother_wer": 0.5553049222589221, | |
| "step": 3939 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_NB_RUND_loss": 0.33576926589012146, | |
| "eval_NB_RUND_runtime": 85.9654, | |
| "eval_NB_RUND_samples_per_second": 21.846, | |
| "eval_NB_RUND_steps_per_second": 0.686, | |
| "eval_NB_RUND_wer": 0.19296057924376508, | |
| "step": 3939 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_rundkast_loss": 0.2803370952606201, | |
| "eval_rundkast_runtime": 33.2875, | |
| "eval_rundkast_samples_per_second": 40.406, | |
| "eval_rundkast_steps_per_second": 1.292, | |
| "eval_rundkast_wer": 0.16457206490614062, | |
| "step": 3939 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_nb_samtale_loss": 0.47577109932899475, | |
| "eval_nb_samtale_runtime": 37.7954, | |
| "eval_nb_samtale_samples_per_second": 14.102, | |
| "eval_nb_samtale_steps_per_second": 0.45, | |
| "eval_nb_samtale_wer": 0.22102864583333334, | |
| "step": 3939 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 4.673075199127197, | |
| "learning_rate": 9.26499411995296e-05, | |
| "loss": 0.397, | |
| "step": 5252 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_bigbrother_loss": 1.7913014888763428, | |
| "eval_bigbrother_runtime": 41.2837, | |
| "eval_bigbrother_samples_per_second": 33.185, | |
| "eval_bigbrother_steps_per_second": 1.042, | |
| "eval_bigbrother_wer": 0.5413679207282688, | |
| "step": 5252 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_NB_RUND_loss": 0.3278275728225708, | |
| "eval_NB_RUND_runtime": 85.8041, | |
| "eval_NB_RUND_samples_per_second": 21.887, | |
| "eval_NB_RUND_steps_per_second": 0.688, | |
| "eval_NB_RUND_wer": 0.1829847144006436, | |
| "step": 5252 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_rundkast_loss": 0.2707855999469757, | |
| "eval_rundkast_runtime": 33.1752, | |
| "eval_rundkast_samples_per_second": 40.542, | |
| "eval_rundkast_steps_per_second": 1.296, | |
| "eval_rundkast_wer": 0.15431116767419661, | |
| "step": 5252 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_nb_samtale_loss": 0.47150808572769165, | |
| "eval_nb_samtale_runtime": 37.7895, | |
| "eval_nb_samtale_samples_per_second": 14.104, | |
| "eval_nb_samtale_steps_per_second": 0.45, | |
| "eval_nb_samtale_wer": 0.21158854166666666, | |
| "step": 5252 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 2.968538999557495, | |
| "learning_rate": 9.00764406115249e-05, | |
| "loss": 0.3742, | |
| "step": 6565 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_bigbrother_loss": 1.9187724590301514, | |
| "eval_bigbrother_runtime": 41.3465, | |
| "eval_bigbrother_samples_per_second": 33.135, | |
| "eval_bigbrother_steps_per_second": 1.04, | |
| "eval_bigbrother_wer": 0.5373398856038024, | |
| "step": 6565 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_NB_RUND_loss": 0.3254011571407318, | |
| "eval_NB_RUND_runtime": 85.8831, | |
| "eval_NB_RUND_samples_per_second": 21.867, | |
| "eval_NB_RUND_steps_per_second": 0.687, | |
| "eval_NB_RUND_wer": 0.17666934835076428, | |
| "step": 6565 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_rundkast_loss": 0.2743120491504669, | |
| "eval_rundkast_runtime": 32.9681, | |
| "eval_rundkast_samples_per_second": 40.797, | |
| "eval_rundkast_steps_per_second": 1.304, | |
| "eval_rundkast_wer": 0.14937957365574292, | |
| "step": 6565 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_nb_samtale_loss": 0.453761488199234, | |
| "eval_nb_samtale_runtime": 38.2375, | |
| "eval_nb_samtale_samples_per_second": 13.939, | |
| "eval_nb_samtale_steps_per_second": 0.445, | |
| "eval_nb_samtale_wer": 0.20475260416666666, | |
| "step": 6565 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 2.727647304534912, | |
| "learning_rate": 8.750490003920032e-05, | |
| "loss": 0.3559, | |
| "step": 7878 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_bigbrother_loss": 1.8001114130020142, | |
| "eval_bigbrother_runtime": 41.331, | |
| "eval_bigbrother_samples_per_second": 33.147, | |
| "eval_bigbrother_steps_per_second": 1.04, | |
| "eval_bigbrother_wer": 0.5302505437847418, | |
| "step": 7878 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_NB_RUND_loss": 0.31470128893852234, | |
| "eval_NB_RUND_runtime": 85.2967, | |
| "eval_NB_RUND_samples_per_second": 22.017, | |
| "eval_NB_RUND_steps_per_second": 0.692, | |
| "eval_NB_RUND_wer": 0.1740949316170555, | |
| "step": 7878 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_rundkast_loss": 0.26562032103538513, | |
| "eval_rundkast_runtime": 33.1186, | |
| "eval_rundkast_samples_per_second": 40.612, | |
| "eval_rundkast_steps_per_second": 1.298, | |
| "eval_rundkast_wer": 0.14802736239261852, | |
| "step": 7878 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_nb_samtale_loss": 0.43801799416542053, | |
| "eval_nb_samtale_runtime": 37.7602, | |
| "eval_nb_samtale_samples_per_second": 14.115, | |
| "eval_nb_samtale_steps_per_second": 0.45, | |
| "eval_nb_samtale_wer": 0.20100911458333334, | |
| "step": 7878 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 4.560679912567139, | |
| "learning_rate": 8.493139945119561e-05, | |
| "loss": 0.3425, | |
| "step": 9191 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_bigbrother_loss": 1.8266634941101074, | |
| "eval_bigbrother_runtime": 41.178, | |
| "eval_bigbrother_samples_per_second": 33.27, | |
| "eval_bigbrother_steps_per_second": 1.044, | |
| "eval_bigbrother_wer": 0.5329090469668896, | |
| "step": 9191 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_NB_RUND_loss": 0.3031991124153137, | |
| "eval_NB_RUND_runtime": 85.2734, | |
| "eval_NB_RUND_samples_per_second": 22.023, | |
| "eval_NB_RUND_steps_per_second": 0.692, | |
| "eval_NB_RUND_wer": 0.168141592920354, | |
| "step": 9191 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_rundkast_loss": 0.2632894814014435, | |
| "eval_rundkast_runtime": 33.1093, | |
| "eval_rundkast_samples_per_second": 40.623, | |
| "eval_rundkast_steps_per_second": 1.299, | |
| "eval_rundkast_wer": 0.14007317849188675, | |
| "step": 9191 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_nb_samtale_loss": 0.4041052460670471, | |
| "eval_nb_samtale_runtime": 37.7437, | |
| "eval_nb_samtale_samples_per_second": 14.122, | |
| "eval_nb_samtale_steps_per_second": 0.45, | |
| "eval_nb_samtale_wer": 0.1962890625, | |
| "step": 9191 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 2.754389524459839, | |
| "learning_rate": 8.235985887887103e-05, | |
| "loss": 0.3264, | |
| "step": 10504 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_bigbrother_loss": 1.930435299873352, | |
| "eval_bigbrother_runtime": 41.8539, | |
| "eval_bigbrother_samples_per_second": 32.733, | |
| "eval_bigbrother_steps_per_second": 1.027, | |
| "eval_bigbrother_wer": 0.5231612019656812, | |
| "step": 10504 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_NB_RUND_loss": 0.3151450455188751, | |
| "eval_NB_RUND_runtime": 106.9286, | |
| "eval_NB_RUND_samples_per_second": 17.563, | |
| "eval_NB_RUND_steps_per_second": 0.552, | |
| "eval_NB_RUND_wer": 0.16790024135156878, | |
| "step": 10504 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_rundkast_loss": 0.2726036012172699, | |
| "eval_rundkast_runtime": 37.1374, | |
| "eval_rundkast_samples_per_second": 36.217, | |
| "eval_rundkast_steps_per_second": 1.158, | |
| "eval_rundkast_wer": 0.1422208081450843, | |
| "step": 10504 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_nb_samtale_loss": 0.422342449426651, | |
| "eval_nb_samtale_runtime": 37.6557, | |
| "eval_nb_samtale_samples_per_second": 14.155, | |
| "eval_nb_samtale_steps_per_second": 0.451, | |
| "eval_nb_samtale_wer": 0.19490559895833334, | |
| "step": 10504 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 1.9258495569229126, | |
| "learning_rate": 7.978635829086632e-05, | |
| "loss": 0.3126, | |
| "step": 11817 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_bigbrother_loss": 1.8571630716323853, | |
| "eval_bigbrother_runtime": 41.3864, | |
| "eval_bigbrother_samples_per_second": 33.103, | |
| "eval_bigbrother_steps_per_second": 1.039, | |
| "eval_bigbrother_wer": 0.524611294610489, | |
| "step": 11817 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_NB_RUND_loss": 0.3117910325527191, | |
| "eval_NB_RUND_runtime": 86.1461, | |
| "eval_NB_RUND_samples_per_second": 21.8, | |
| "eval_NB_RUND_steps_per_second": 0.685, | |
| "eval_NB_RUND_wer": 0.16287208366854383, | |
| "step": 11817 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_rundkast_loss": 0.27052268385887146, | |
| "eval_rundkast_runtime": 33.2454, | |
| "eval_rundkast_samples_per_second": 40.457, | |
| "eval_rundkast_steps_per_second": 1.293, | |
| "eval_rundkast_wer": 0.1399140948138721, | |
| "step": 11817 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_nb_samtale_loss": 0.41521695256233215, | |
| "eval_nb_samtale_runtime": 37.6049, | |
| "eval_nb_samtale_samples_per_second": 14.174, | |
| "eval_nb_samtale_steps_per_second": 0.452, | |
| "eval_nb_samtale_wer": 0.18636067708333334, | |
| "step": 11817 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "grad_norm": 4.801841735839844, | |
| "learning_rate": 7.721285770286163e-05, | |
| "loss": 0.3009, | |
| "step": 13130 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_bigbrother_loss": 1.9679045677185059, | |
| "eval_bigbrother_runtime": 41.3496, | |
| "eval_bigbrother_samples_per_second": 33.132, | |
| "eval_bigbrother_steps_per_second": 1.04, | |
| "eval_bigbrother_wer": 0.5192942882461935, | |
| "step": 13130 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_NB_RUND_loss": 0.31853994727134705, | |
| "eval_NB_RUND_runtime": 85.1737, | |
| "eval_NB_RUND_samples_per_second": 22.049, | |
| "eval_NB_RUND_steps_per_second": 0.693, | |
| "eval_NB_RUND_wer": 0.16182622687047465, | |
| "step": 13130 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_rundkast_loss": 0.2728247046470642, | |
| "eval_rundkast_runtime": 32.73, | |
| "eval_rundkast_samples_per_second": 41.094, | |
| "eval_rundkast_steps_per_second": 1.314, | |
| "eval_rundkast_wer": 0.13880050906776964, | |
| "step": 13130 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_nb_samtale_loss": 0.4334540069103241, | |
| "eval_nb_samtale_runtime": 37.2959, | |
| "eval_nb_samtale_samples_per_second": 14.291, | |
| "eval_nb_samtale_steps_per_second": 0.456, | |
| "eval_nb_samtale_wer": 0.185302734375, | |
| "step": 13130 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "grad_norm": 2.734304428100586, | |
| "learning_rate": 7.464131713053705e-05, | |
| "loss": 0.2911, | |
| "step": 14443 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_bigbrother_loss": 2.009200096130371, | |
| "eval_bigbrother_runtime": 41.4463, | |
| "eval_bigbrother_samples_per_second": 33.055, | |
| "eval_bigbrother_steps_per_second": 1.037, | |
| "eval_bigbrother_wer": 0.5190526061387255, | |
| "step": 14443 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_NB_RUND_loss": 0.3390868902206421, | |
| "eval_NB_RUND_runtime": 85.2442, | |
| "eval_NB_RUND_samples_per_second": 22.031, | |
| "eval_NB_RUND_steps_per_second": 0.692, | |
| "eval_NB_RUND_wer": 0.16061946902654867, | |
| "step": 14443 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_rundkast_loss": 0.3032366931438446, | |
| "eval_rundkast_runtime": 33.1436, | |
| "eval_rundkast_samples_per_second": 40.581, | |
| "eval_rundkast_steps_per_second": 1.297, | |
| "eval_rundkast_wer": 0.1363347120585428, | |
| "step": 14443 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_nb_samtale_loss": 0.4296092391014099, | |
| "eval_nb_samtale_runtime": 37.8422, | |
| "eval_nb_samtale_samples_per_second": 14.085, | |
| "eval_nb_samtale_steps_per_second": 0.449, | |
| "eval_nb_samtale_wer": 0.184814453125, | |
| "step": 14443 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "grad_norm": 3.582427740097046, | |
| "learning_rate": 7.206781654253235e-05, | |
| "loss": 0.2795, | |
| "step": 15756 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_bigbrother_loss": 2.00494122505188, | |
| "eval_bigbrother_runtime": 41.5443, | |
| "eval_bigbrother_samples_per_second": 32.977, | |
| "eval_bigbrother_steps_per_second": 1.035, | |
| "eval_bigbrother_wer": 0.5176025134939176, | |
| "step": 15756 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_NB_RUND_loss": 0.3187481760978699, | |
| "eval_NB_RUND_runtime": 85.8022, | |
| "eval_NB_RUND_samples_per_second": 21.888, | |
| "eval_NB_RUND_steps_per_second": 0.688, | |
| "eval_NB_RUND_wer": 0.15901045856798068, | |
| "step": 15756 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_rundkast_loss": 0.2752975821495056, | |
| "eval_rundkast_runtime": 33.305, | |
| "eval_rundkast_samples_per_second": 40.384, | |
| "eval_rundkast_steps_per_second": 1.291, | |
| "eval_rundkast_wer": 0.1350620426344257, | |
| "step": 15756 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_nb_samtale_loss": 0.42811375856399536, | |
| "eval_nb_samtale_runtime": 37.713, | |
| "eval_nb_samtale_samples_per_second": 14.133, | |
| "eval_nb_samtale_steps_per_second": 0.451, | |
| "eval_nb_samtale_wer": 0.18310546875, | |
| "step": 15756 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "grad_norm": 4.836462020874023, | |
| "learning_rate": 6.94982359858879e-05, | |
| "loss": 0.2712, | |
| "step": 17069 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_bigbrother_loss": 1.9352320432662964, | |
| "eval_bigbrother_runtime": 41.182, | |
| "eval_bigbrother_samples_per_second": 33.267, | |
| "eval_bigbrother_steps_per_second": 1.044, | |
| "eval_bigbrother_wer": 0.5195359703536615, | |
| "step": 17069 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_NB_RUND_loss": 0.305833101272583, | |
| "eval_NB_RUND_runtime": 84.7929, | |
| "eval_NB_RUND_samples_per_second": 22.148, | |
| "eval_NB_RUND_steps_per_second": 0.696, | |
| "eval_NB_RUND_wer": 0.1578037007240547, | |
| "step": 17069 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_rundkast_loss": 0.26513412594795227, | |
| "eval_rundkast_runtime": 32.9292, | |
| "eval_rundkast_samples_per_second": 40.845, | |
| "eval_rundkast_steps_per_second": 1.306, | |
| "eval_rundkast_wer": 0.13434616608335984, | |
| "step": 17069 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_nb_samtale_loss": 0.4084182679653168, | |
| "eval_nb_samtale_runtime": 37.2213, | |
| "eval_nb_samtale_samples_per_second": 14.32, | |
| "eval_nb_samtale_steps_per_second": 0.457, | |
| "eval_nb_samtale_wer": 0.18131510416666666, | |
| "step": 17069 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "grad_norm": 3.973949909210205, | |
| "learning_rate": 6.692473539788319e-05, | |
| "loss": 0.2655, | |
| "step": 18382 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_bigbrother_loss": 1.9412481784820557, | |
| "eval_bigbrother_runtime": 41.0966, | |
| "eval_bigbrother_samples_per_second": 33.336, | |
| "eval_bigbrother_steps_per_second": 1.046, | |
| "eval_bigbrother_wer": 0.5118021429146862, | |
| "step": 18382 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_NB_RUND_loss": 0.3095969557762146, | |
| "eval_NB_RUND_runtime": 84.8936, | |
| "eval_NB_RUND_samples_per_second": 22.122, | |
| "eval_NB_RUND_steps_per_second": 0.695, | |
| "eval_NB_RUND_wer": 0.15530973451327434, | |
| "step": 18382 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_rundkast_loss": 0.2749696373939514, | |
| "eval_rundkast_runtime": 33.0503, | |
| "eval_rundkast_samples_per_second": 40.696, | |
| "eval_rundkast_steps_per_second": 1.301, | |
| "eval_rundkast_wer": 0.1327553293032135, | |
| "step": 18382 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_nb_samtale_loss": 0.395680695772171, | |
| "eval_nb_samtale_runtime": 37.1829, | |
| "eval_nb_samtale_samples_per_second": 14.335, | |
| "eval_nb_samtale_steps_per_second": 0.457, | |
| "eval_nb_samtale_wer": 0.17789713541666666, | |
| "step": 18382 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "grad_norm": 3.9156057834625244, | |
| "learning_rate": 6.435123480987849e-05, | |
| "loss": 0.2557, | |
| "step": 19695 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_bigbrother_loss": 1.9280781745910645, | |
| "eval_bigbrother_runtime": 42.2614, | |
| "eval_bigbrother_samples_per_second": 32.417, | |
| "eval_bigbrother_steps_per_second": 1.017, | |
| "eval_bigbrother_wer": 0.5109159751873036, | |
| "step": 19695 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_NB_RUND_loss": 0.2862784266471863, | |
| "eval_NB_RUND_runtime": 85.5508, | |
| "eval_NB_RUND_samples_per_second": 21.952, | |
| "eval_NB_RUND_steps_per_second": 0.69, | |
| "eval_NB_RUND_wer": 0.1563958165728077, | |
| "step": 19695 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_rundkast_loss": 0.2504226267337799, | |
| "eval_rundkast_runtime": 33.4378, | |
| "eval_rundkast_samples_per_second": 40.224, | |
| "eval_rundkast_steps_per_second": 1.286, | |
| "eval_rundkast_wer": 0.13426662424435254, | |
| "step": 19695 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_nb_samtale_loss": 0.37682539224624634, | |
| "eval_nb_samtale_runtime": 37.3809, | |
| "eval_nb_samtale_samples_per_second": 14.259, | |
| "eval_nb_samtale_steps_per_second": 0.455, | |
| "eval_nb_samtale_wer": 0.17887369791666666, | |
| "step": 19695 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "grad_norm": 4.767539024353027, | |
| "learning_rate": 6.17796942375539e-05, | |
| "loss": 0.2497, | |
| "step": 21008 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_bigbrother_loss": 2.0177502632141113, | |
| "eval_bigbrother_runtime": 41.1529, | |
| "eval_bigbrother_samples_per_second": 33.291, | |
| "eval_bigbrother_steps_per_second": 1.045, | |
| "eval_bigbrother_wer": 0.5074518649802626, | |
| "step": 21008 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_NB_RUND_loss": 0.32014137506484985, | |
| "eval_NB_RUND_runtime": 85.0566, | |
| "eval_NB_RUND_samples_per_second": 22.079, | |
| "eval_NB_RUND_steps_per_second": 0.694, | |
| "eval_NB_RUND_wer": 0.15522928399034594, | |
| "step": 21008 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_rundkast_loss": 0.28833797574043274, | |
| "eval_rundkast_runtime": 33.0799, | |
| "eval_rundkast_samples_per_second": 40.659, | |
| "eval_rundkast_steps_per_second": 1.3, | |
| "eval_rundkast_wer": 0.13219853643016227, | |
| "step": 21008 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_nb_samtale_loss": 0.40036484599113464, | |
| "eval_nb_samtale_runtime": 37.391, | |
| "eval_nb_samtale_samples_per_second": 14.255, | |
| "eval_nb_samtale_steps_per_second": 0.455, | |
| "eval_nb_samtale_wer": 0.17838541666666666, | |
| "step": 21008 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "grad_norm": 2.875094413757324, | |
| "learning_rate": 5.9206193649549204e-05, | |
| "loss": 0.2424, | |
| "step": 22321 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_bigbrother_loss": 2.1849660873413086, | |
| "eval_bigbrother_runtime": 41.489, | |
| "eval_bigbrother_samples_per_second": 33.021, | |
| "eval_bigbrother_steps_per_second": 1.036, | |
| "eval_bigbrother_wer": 0.51494401031177, | |
| "step": 22321 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_NB_RUND_loss": 0.32200556993484497, | |
| "eval_NB_RUND_runtime": 85.425, | |
| "eval_NB_RUND_samples_per_second": 21.984, | |
| "eval_NB_RUND_steps_per_second": 0.691, | |
| "eval_NB_RUND_wer": 0.15289621882542237, | |
| "step": 22321 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_rundkast_loss": 0.28678199648857117, | |
| "eval_rundkast_runtime": 33.351, | |
| "eval_rundkast_samples_per_second": 40.329, | |
| "eval_rundkast_steps_per_second": 1.289, | |
| "eval_rundkast_wer": 0.13180082723512568, | |
| "step": 22321 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_nb_samtale_loss": 0.4106721878051758, | |
| "eval_nb_samtale_runtime": 37.5903, | |
| "eval_nb_samtale_samples_per_second": 14.179, | |
| "eval_nb_samtale_steps_per_second": 0.452, | |
| "eval_nb_samtale_wer": 0.17415364583333334, | |
| "step": 22321 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "grad_norm": 4.949888229370117, | |
| "learning_rate": 5.66326930615445e-05, | |
| "loss": 0.2374, | |
| "step": 23634 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_bigbrother_loss": 2.1056079864501953, | |
| "eval_bigbrother_runtime": 41.6333, | |
| "eval_bigbrother_samples_per_second": 32.906, | |
| "eval_bigbrother_steps_per_second": 1.033, | |
| "eval_bigbrother_wer": 0.5105937323773463, | |
| "step": 23634 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_NB_RUND_loss": 0.32035475969314575, | |
| "eval_NB_RUND_runtime": 87.7221, | |
| "eval_NB_RUND_samples_per_second": 21.409, | |
| "eval_NB_RUND_steps_per_second": 0.673, | |
| "eval_NB_RUND_wer": 0.15329847144006437, | |
| "step": 23634 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_rundkast_loss": 0.29013389348983765, | |
| "eval_rundkast_runtime": 33.2901, | |
| "eval_rundkast_samples_per_second": 40.402, | |
| "eval_rundkast_steps_per_second": 1.292, | |
| "eval_rundkast_wer": 0.13005090677696468, | |
| "step": 23634 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_nb_samtale_loss": 0.39669641852378845, | |
| "eval_nb_samtale_runtime": 38.5507, | |
| "eval_nb_samtale_samples_per_second": 13.826, | |
| "eval_nb_samtale_steps_per_second": 0.441, | |
| "eval_nb_samtale_wer": 0.17724609375, | |
| "step": 23634 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "grad_norm": 5.62293004989624, | |
| "learning_rate": 5.4063112504900036e-05, | |
| "loss": 0.2313, | |
| "step": 24947 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_bigbrother_loss": 2.0347869396209717, | |
| "eval_bigbrother_runtime": 41.8028, | |
| "eval_bigbrother_samples_per_second": 32.773, | |
| "eval_bigbrother_steps_per_second": 1.029, | |
| "eval_bigbrother_wer": 0.503584951260775, | |
| "step": 24947 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_NB_RUND_loss": 0.3127482533454895, | |
| "eval_NB_RUND_runtime": 85.6096, | |
| "eval_NB_RUND_samples_per_second": 21.937, | |
| "eval_NB_RUND_steps_per_second": 0.689, | |
| "eval_NB_RUND_wer": 0.15100563153660498, | |
| "step": 24947 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_rundkast_loss": 0.2715360224246979, | |
| "eval_rundkast_runtime": 33.3269, | |
| "eval_rundkast_samples_per_second": 40.358, | |
| "eval_rundkast_steps_per_second": 1.29, | |
| "eval_rundkast_wer": 0.1292554883868915, | |
| "step": 24947 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_nb_samtale_loss": 0.4169124960899353, | |
| "eval_nb_samtale_runtime": 37.555, | |
| "eval_nb_samtale_samples_per_second": 14.193, | |
| "eval_nb_samtale_steps_per_second": 0.453, | |
| "eval_nb_samtale_wer": 0.17350260416666666, | |
| "step": 24947 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "grad_norm": 2.86464262008667, | |
| "learning_rate": 5.148961191689534e-05, | |
| "loss": 0.2227, | |
| "step": 26260 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_bigbrother_loss": 2.1283321380615234, | |
| "eval_bigbrother_runtime": 41.6851, | |
| "eval_bigbrother_samples_per_second": 32.865, | |
| "eval_bigbrother_steps_per_second": 1.032, | |
| "eval_bigbrother_wer": 0.5016514944010312, | |
| "step": 26260 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_NB_RUND_loss": 0.3282929062843323, | |
| "eval_NB_RUND_runtime": 86.0672, | |
| "eval_NB_RUND_samples_per_second": 21.82, | |
| "eval_NB_RUND_steps_per_second": 0.686, | |
| "eval_NB_RUND_wer": 0.15160901045856798, | |
| "step": 26260 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_rundkast_loss": 0.2934817969799042, | |
| "eval_rundkast_runtime": 33.1568, | |
| "eval_rundkast_samples_per_second": 40.565, | |
| "eval_rundkast_steps_per_second": 1.297, | |
| "eval_rundkast_wer": 0.1306076996500159, | |
| "step": 26260 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_nb_samtale_loss": 0.41641756892204285, | |
| "eval_nb_samtale_runtime": 37.7201, | |
| "eval_nb_samtale_samples_per_second": 14.13, | |
| "eval_nb_samtale_steps_per_second": 0.451, | |
| "eval_nb_samtale_wer": 0.17301432291666666, | |
| "step": 26260 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "grad_norm": 3.0193545818328857, | |
| "learning_rate": 4.8916111328890636e-05, | |
| "loss": 0.2214, | |
| "step": 27573 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "eval_bigbrother_loss": 2.2207038402557373, | |
| "eval_bigbrother_runtime": 41.5705, | |
| "eval_bigbrother_samples_per_second": 32.956, | |
| "eval_bigbrother_steps_per_second": 1.034, | |
| "eval_bigbrother_wer": 0.50656569725288, | |
| "step": 27573 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "eval_NB_RUND_loss": 0.32080039381980896, | |
| "eval_NB_RUND_runtime": 85.8822, | |
| "eval_NB_RUND_samples_per_second": 21.867, | |
| "eval_NB_RUND_steps_per_second": 0.687, | |
| "eval_NB_RUND_wer": 0.1498793242156074, | |
| "step": 27573 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "eval_rundkast_loss": 0.28392452001571655, | |
| "eval_rundkast_runtime": 33.3668, | |
| "eval_rundkast_samples_per_second": 40.31, | |
| "eval_rundkast_steps_per_second": 1.289, | |
| "eval_rundkast_wer": 0.1287782373528476, | |
| "step": 27573 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "eval_nb_samtale_loss": 0.4135919511318207, | |
| "eval_nb_samtale_runtime": 37.8061, | |
| "eval_nb_samtale_samples_per_second": 14.098, | |
| "eval_nb_samtale_steps_per_second": 0.45, | |
| "eval_nb_samtale_wer": 0.17097981770833334, | |
| "step": 27573 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "grad_norm": 3.5664854049682617, | |
| "learning_rate": 4.6344570756566055e-05, | |
| "loss": 0.2162, | |
| "step": 28886 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_bigbrother_loss": 2.1843111515045166, | |
| "eval_bigbrother_runtime": 40.759, | |
| "eval_bigbrother_samples_per_second": 33.612, | |
| "eval_bigbrother_steps_per_second": 1.055, | |
| "eval_bigbrother_wer": 0.5060823330379441, | |
| "step": 28886 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_NB_RUND_loss": 0.33185243606567383, | |
| "eval_NB_RUND_runtime": 84.3565, | |
| "eval_NB_RUND_samples_per_second": 22.263, | |
| "eval_NB_RUND_steps_per_second": 0.699, | |
| "eval_NB_RUND_wer": 0.1508447304907482, | |
| "step": 28886 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_rundkast_loss": 0.2995389699935913, | |
| "eval_rundkast_runtime": 32.6931, | |
| "eval_rundkast_samples_per_second": 41.14, | |
| "eval_rundkast_steps_per_second": 1.315, | |
| "eval_rundkast_wer": 0.12869869551384028, | |
| "step": 28886 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_nb_samtale_loss": 0.4128870964050293, | |
| "eval_nb_samtale_runtime": 36.8088, | |
| "eval_nb_samtale_samples_per_second": 14.48, | |
| "eval_nb_samtale_steps_per_second": 0.462, | |
| "eval_nb_samtale_wer": 0.173828125, | |
| "step": 28886 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "grad_norm": 4.249322414398193, | |
| "learning_rate": 4.37749901999216e-05, | |
| "loss": 0.2125, | |
| "step": 30199 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "eval_bigbrother_loss": 2.460817575454712, | |
| "eval_bigbrother_runtime": 40.6315, | |
| "eval_bigbrother_samples_per_second": 33.718, | |
| "eval_bigbrother_steps_per_second": 1.058, | |
| "eval_bigbrother_wer": 0.5047128010956256, | |
| "step": 30199 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "eval_NB_RUND_loss": 0.34763312339782715, | |
| "eval_NB_RUND_runtime": 84.2518, | |
| "eval_NB_RUND_samples_per_second": 22.29, | |
| "eval_NB_RUND_steps_per_second": 0.7, | |
| "eval_NB_RUND_wer": 0.14774738535800483, | |
| "step": 30199 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "eval_rundkast_loss": 0.3159354031085968, | |
| "eval_rundkast_runtime": 32.702, | |
| "eval_rundkast_samples_per_second": 41.129, | |
| "eval_rundkast_steps_per_second": 1.315, | |
| "eval_rundkast_wer": 0.12758510976773782, | |
| "step": 30199 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "eval_nb_samtale_loss": 0.42722073197364807, | |
| "eval_nb_samtale_runtime": 36.8013, | |
| "eval_nb_samtale_samples_per_second": 14.483, | |
| "eval_nb_samtale_steps_per_second": 0.462, | |
| "eval_nb_samtale_wer": 0.16796875, | |
| "step": 30199 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "grad_norm": 3.8950681686401367, | |
| "learning_rate": 4.12014896119169e-05, | |
| "loss": 0.2054, | |
| "step": 31512 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_bigbrother_loss": 2.2196638584136963, | |
| "eval_bigbrother_runtime": 41.3959, | |
| "eval_bigbrother_samples_per_second": 33.095, | |
| "eval_bigbrother_steps_per_second": 1.039, | |
| "eval_bigbrother_wer": 0.4956900024168211, | |
| "step": 31512 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_NB_RUND_loss": 0.31967854499816895, | |
| "eval_NB_RUND_runtime": 85.4915, | |
| "eval_NB_RUND_samples_per_second": 21.967, | |
| "eval_NB_RUND_steps_per_second": 0.69, | |
| "eval_NB_RUND_wer": 0.14903459372485922, | |
| "step": 31512 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_rundkast_loss": 0.28724637627601624, | |
| "eval_rundkast_runtime": 32.6772, | |
| "eval_rundkast_samples_per_second": 41.16, | |
| "eval_rundkast_steps_per_second": 1.316, | |
| "eval_rundkast_wer": 0.12726694241170855, | |
| "step": 31512 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_nb_samtale_loss": 0.401165634393692, | |
| "eval_nb_samtale_runtime": 37.0545, | |
| "eval_nb_samtale_samples_per_second": 14.384, | |
| "eval_nb_samtale_steps_per_second": 0.459, | |
| "eval_nb_samtale_wer": 0.17106119791666666, | |
| "step": 31512 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "grad_norm": 3.483825922012329, | |
| "learning_rate": 3.8627989023912194e-05, | |
| "loss": 0.2016, | |
| "step": 32825 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "eval_bigbrother_loss": 2.3513925075531006, | |
| "eval_bigbrother_runtime": 40.9522, | |
| "eval_bigbrother_samples_per_second": 33.454, | |
| "eval_bigbrother_steps_per_second": 1.05, | |
| "eval_bigbrother_wer": 0.5014098122935632, | |
| "step": 32825 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "eval_NB_RUND_loss": 0.3408574163913727, | |
| "eval_NB_RUND_runtime": 84.79, | |
| "eval_NB_RUND_samples_per_second": 22.149, | |
| "eval_NB_RUND_steps_per_second": 0.696, | |
| "eval_NB_RUND_wer": 0.14714400643604184, | |
| "step": 32825 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "eval_rundkast_loss": 0.3113822042942047, | |
| "eval_rundkast_runtime": 33.2613, | |
| "eval_rundkast_samples_per_second": 40.437, | |
| "eval_rundkast_steps_per_second": 1.293, | |
| "eval_rundkast_wer": 0.12599427298759147, | |
| "step": 32825 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "eval_nb_samtale_loss": 0.41482433676719666, | |
| "eval_nb_samtale_runtime": 36.9383, | |
| "eval_nb_samtale_samples_per_second": 14.429, | |
| "eval_nb_samtale_steps_per_second": 0.46, | |
| "eval_nb_samtale_wer": 0.16861979166666666, | |
| "step": 32825 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "grad_norm": 5.882791042327881, | |
| "learning_rate": 3.6054488435907494e-05, | |
| "loss": 0.1986, | |
| "step": 34138 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "eval_bigbrother_loss": 2.2447853088378906, | |
| "eval_bigbrother_runtime": 41.1993, | |
| "eval_bigbrother_samples_per_second": 33.253, | |
| "eval_bigbrother_steps_per_second": 1.044, | |
| "eval_bigbrother_wer": 0.4957705631193104, | |
| "step": 34138 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "eval_NB_RUND_loss": 0.3391417860984802, | |
| "eval_NB_RUND_runtime": 86.9591, | |
| "eval_NB_RUND_samples_per_second": 21.596, | |
| "eval_NB_RUND_steps_per_second": 0.678, | |
| "eval_NB_RUND_wer": 0.14734513274336283, | |
| "step": 34138 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "eval_rundkast_loss": 0.3099968433380127, | |
| "eval_rundkast_runtime": 32.6851, | |
| "eval_rundkast_samples_per_second": 41.15, | |
| "eval_rundkast_steps_per_second": 1.316, | |
| "eval_rundkast_wer": 0.12575564747056953, | |
| "step": 34138 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "eval_nb_samtale_loss": 0.41232603788375854, | |
| "eval_nb_samtale_runtime": 37.0492, | |
| "eval_nb_samtale_samples_per_second": 14.386, | |
| "eval_nb_samtale_steps_per_second": 0.459, | |
| "eval_nb_samtale_wer": 0.16951497395833334, | |
| "step": 34138 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "grad_norm": 2.837421178817749, | |
| "learning_rate": 3.348098784790279e-05, | |
| "loss": 0.1953, | |
| "step": 35451 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "eval_bigbrother_loss": 2.240429401397705, | |
| "eval_bigbrother_runtime": 41.126, | |
| "eval_bigbrother_samples_per_second": 33.312, | |
| "eval_bigbrother_steps_per_second": 1.046, | |
| "eval_bigbrother_wer": 0.49552888101184245, | |
| "step": 35451 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "eval_NB_RUND_loss": 0.3420700132846832, | |
| "eval_NB_RUND_runtime": 84.1024, | |
| "eval_NB_RUND_samples_per_second": 22.33, | |
| "eval_NB_RUND_steps_per_second": 0.702, | |
| "eval_NB_RUND_wer": 0.14738535800482702, | |
| "step": 35451 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "eval_rundkast_loss": 0.3145124018192291, | |
| "eval_rundkast_runtime": 32.9045, | |
| "eval_rundkast_samples_per_second": 40.876, | |
| "eval_rundkast_steps_per_second": 1.307, | |
| "eval_rundkast_wer": 0.12591473114858415, | |
| "step": 35451 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "eval_nb_samtale_loss": 0.4117932617664337, | |
| "eval_nb_samtale_runtime": 36.8227, | |
| "eval_nb_samtale_samples_per_second": 14.475, | |
| "eval_nb_samtale_steps_per_second": 0.462, | |
| "eval_nb_samtale_wer": 0.16943359375, | |
| "step": 35451 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "grad_norm": 3.3868250846862793, | |
| "learning_rate": 3.0909447275578206e-05, | |
| "loss": 0.1939, | |
| "step": 36764 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "eval_bigbrother_loss": 2.409648895263672, | |
| "eval_bigbrother_runtime": 41.1082, | |
| "eval_bigbrother_samples_per_second": 33.327, | |
| "eval_bigbrother_steps_per_second": 1.046, | |
| "eval_bigbrother_wer": 0.49536775960686374, | |
| "step": 36764 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "eval_NB_RUND_loss": 0.34609299898147583, | |
| "eval_NB_RUND_runtime": 86.3212, | |
| "eval_NB_RUND_samples_per_second": 21.756, | |
| "eval_NB_RUND_steps_per_second": 0.683, | |
| "eval_NB_RUND_wer": 0.14569589702333066, | |
| "step": 36764 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "eval_rundkast_loss": 0.31334388256073, | |
| "eval_rundkast_runtime": 32.8372, | |
| "eval_rundkast_samples_per_second": 40.96, | |
| "eval_rundkast_steps_per_second": 1.309, | |
| "eval_rundkast_wer": 0.12368755965637926, | |
| "step": 36764 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "eval_nb_samtale_loss": 0.42863455414772034, | |
| "eval_nb_samtale_runtime": 37.033, | |
| "eval_nb_samtale_samples_per_second": 14.393, | |
| "eval_nb_samtale_steps_per_second": 0.459, | |
| "eval_nb_samtale_wer": 0.16829427083333334, | |
| "step": 36764 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "grad_norm": 3.9038360118865967, | |
| "learning_rate": 2.8337906703253626e-05, | |
| "loss": 0.1883, | |
| "step": 38077 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "eval_bigbrother_loss": 2.32140851020813, | |
| "eval_bigbrother_runtime": 40.9774, | |
| "eval_bigbrother_samples_per_second": 33.433, | |
| "eval_bigbrother_steps_per_second": 1.049, | |
| "eval_bigbrother_wer": 0.4960122452267784, | |
| "step": 38077 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "eval_NB_RUND_loss": 0.35102906823158264, | |
| "eval_NB_RUND_runtime": 84.5748, | |
| "eval_NB_RUND_samples_per_second": 22.205, | |
| "eval_NB_RUND_steps_per_second": 0.698, | |
| "eval_NB_RUND_wer": 0.14646017699115044, | |
| "step": 38077 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "eval_rundkast_loss": 0.31922492384910583, | |
| "eval_rundkast_runtime": 32.7767, | |
| "eval_rundkast_samples_per_second": 41.035, | |
| "eval_rundkast_steps_per_second": 1.312, | |
| "eval_rundkast_wer": 0.12496022908049634, | |
| "step": 38077 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "eval_nb_samtale_loss": 0.4308791756629944, | |
| "eval_nb_samtale_runtime": 36.5599, | |
| "eval_nb_samtale_samples_per_second": 14.579, | |
| "eval_nb_samtale_steps_per_second": 0.465, | |
| "eval_nb_samtale_wer": 0.167724609375, | |
| "step": 38077 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "grad_norm": 2.5224320888519287, | |
| "learning_rate": 2.5764406115248922e-05, | |
| "loss": 0.1857, | |
| "step": 39390 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "eval_bigbrother_loss": 2.3643717765808105, | |
| "eval_bigbrother_runtime": 41.8962, | |
| "eval_bigbrother_samples_per_second": 32.7, | |
| "eval_bigbrother_steps_per_second": 1.026, | |
| "eval_bigbrother_wer": 0.496415048739225, | |
| "step": 39390 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "eval_NB_RUND_loss": 0.3557446002960205, | |
| "eval_NB_RUND_runtime": 143.0815, | |
| "eval_NB_RUND_samples_per_second": 13.125, | |
| "eval_NB_RUND_steps_per_second": 0.412, | |
| "eval_NB_RUND_wer": 0.14617860016090103, | |
| "step": 39390 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "eval_rundkast_loss": 0.3290172815322876, | |
| "eval_rundkast_runtime": 32.9137, | |
| "eval_rundkast_samples_per_second": 40.864, | |
| "eval_rundkast_steps_per_second": 1.306, | |
| "eval_rundkast_wer": 0.12623289850461342, | |
| "step": 39390 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "eval_nb_samtale_loss": 0.42328643798828125, | |
| "eval_nb_samtale_runtime": 36.9569, | |
| "eval_nb_samtale_samples_per_second": 14.422, | |
| "eval_nb_samtale_steps_per_second": 0.46, | |
| "eval_nb_samtale_wer": 0.16715494791666666, | |
| "step": 39390 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "grad_norm": 8.252937316894531, | |
| "learning_rate": 2.319090552724422e-05, | |
| "loss": 0.184, | |
| "step": 40703 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "eval_bigbrother_loss": 2.3923449516296387, | |
| "eval_bigbrother_runtime": 41.1367, | |
| "eval_bigbrother_samples_per_second": 33.304, | |
| "eval_bigbrother_steps_per_second": 1.045, | |
| "eval_bigbrother_wer": 0.4919036493998228, | |
| "step": 40703 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "eval_NB_RUND_loss": 0.35123324394226074, | |
| "eval_NB_RUND_runtime": 85.5059, | |
| "eval_NB_RUND_samples_per_second": 21.963, | |
| "eval_NB_RUND_steps_per_second": 0.69, | |
| "eval_NB_RUND_wer": 0.14473049074818986, | |
| "step": 40703 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "eval_rundkast_loss": 0.31798994541168213, | |
| "eval_rundkast_runtime": 33.3879, | |
| "eval_rundkast_samples_per_second": 40.284, | |
| "eval_rundkast_steps_per_second": 1.288, | |
| "eval_rundkast_wer": 0.1243238943684378, | |
| "step": 40703 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "eval_nb_samtale_loss": 0.4355938732624054, | |
| "eval_nb_samtale_runtime": 37.0847, | |
| "eval_nb_samtale_samples_per_second": 14.372, | |
| "eval_nb_samtale_steps_per_second": 0.458, | |
| "eval_nb_samtale_wer": 0.16536458333333334, | |
| "step": 40703 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "grad_norm": 4.856067657470703, | |
| "learning_rate": 2.061936495491964e-05, | |
| "loss": 0.1795, | |
| "step": 42016 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "eval_bigbrother_loss": 2.3352065086364746, | |
| "eval_bigbrother_runtime": 41.1475, | |
| "eval_bigbrother_samples_per_second": 33.295, | |
| "eval_bigbrother_steps_per_second": 1.045, | |
| "eval_bigbrother_wer": 0.4930314992346733, | |
| "step": 42016 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "eval_NB_RUND_loss": 0.3572410047054291, | |
| "eval_NB_RUND_runtime": 105.2885, | |
| "eval_NB_RUND_samples_per_second": 17.837, | |
| "eval_NB_RUND_steps_per_second": 0.56, | |
| "eval_NB_RUND_wer": 0.14477071600965405, | |
| "step": 42016 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "eval_rundkast_loss": 0.32713791728019714, | |
| "eval_rundkast_runtime": 33.0242, | |
| "eval_rundkast_samples_per_second": 40.728, | |
| "eval_rundkast_steps_per_second": 1.302, | |
| "eval_rundkast_wer": 0.1243238943684378, | |
| "step": 42016 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "eval_nb_samtale_loss": 0.43330347537994385, | |
| "eval_nb_samtale_runtime": 36.808, | |
| "eval_nb_samtale_samples_per_second": 14.481, | |
| "eval_nb_samtale_steps_per_second": 0.462, | |
| "eval_nb_samtale_wer": 0.16560872395833334, | |
| "step": 42016 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "grad_norm": 5.330729961395264, | |
| "learning_rate": 1.8045864366914934e-05, | |
| "loss": 0.1797, | |
| "step": 43329 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "eval_bigbrother_loss": 2.4726736545562744, | |
| "eval_bigbrother_runtime": 41.9313, | |
| "eval_bigbrother_samples_per_second": 32.672, | |
| "eval_bigbrother_steps_per_second": 1.025, | |
| "eval_bigbrother_wer": 0.4939982276645452, | |
| "step": 43329 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "eval_NB_RUND_loss": 0.36341410875320435, | |
| "eval_NB_RUND_runtime": 84.523, | |
| "eval_NB_RUND_samples_per_second": 22.219, | |
| "eval_NB_RUND_steps_per_second": 0.698, | |
| "eval_NB_RUND_wer": 0.14710378117457765, | |
| "step": 43329 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "eval_rundkast_loss": 0.3344702422618866, | |
| "eval_rundkast_runtime": 32.8493, | |
| "eval_rundkast_samples_per_second": 40.945, | |
| "eval_rundkast_steps_per_second": 1.309, | |
| "eval_rundkast_wer": 0.12750556792873052, | |
| "step": 43329 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "eval_nb_samtale_loss": 0.43638789653778076, | |
| "eval_nb_samtale_runtime": 36.859, | |
| "eval_nb_samtale_samples_per_second": 14.46, | |
| "eval_nb_samtale_steps_per_second": 0.461, | |
| "eval_nb_samtale_wer": 0.16756184895833334, | |
| "step": 43329 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "grad_norm": 2.472703218460083, | |
| "learning_rate": 1.5474323794590357e-05, | |
| "loss": 0.1792, | |
| "step": 44642 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "eval_bigbrother_loss": 2.369781255722046, | |
| "eval_bigbrother_runtime": 41.5232, | |
| "eval_bigbrother_samples_per_second": 32.994, | |
| "eval_bigbrother_steps_per_second": 1.036, | |
| "eval_bigbrother_wer": 0.4910980423749295, | |
| "step": 44642 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "eval_NB_RUND_loss": 0.3522418141365051, | |
| "eval_NB_RUND_runtime": 86.176, | |
| "eval_NB_RUND_samples_per_second": 21.793, | |
| "eval_NB_RUND_steps_per_second": 0.685, | |
| "eval_NB_RUND_wer": 0.14509251810136767, | |
| "step": 44642 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "eval_rundkast_loss": 0.3241034746170044, | |
| "eval_rundkast_runtime": 32.7879, | |
| "eval_rundkast_samples_per_second": 41.021, | |
| "eval_rundkast_steps_per_second": 1.311, | |
| "eval_rundkast_wer": 0.12360801781737193, | |
| "step": 44642 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "eval_nb_samtale_loss": 0.42330336570739746, | |
| "eval_nb_samtale_runtime": 37.3979, | |
| "eval_nb_samtale_samples_per_second": 14.252, | |
| "eval_nb_samtale_steps_per_second": 0.455, | |
| "eval_nb_samtale_wer": 0.16707356770833334, | |
| "step": 44642 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "grad_norm": 5.900428295135498, | |
| "learning_rate": 1.2900823206585652e-05, | |
| "loss": 0.175, | |
| "step": 45955 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "eval_bigbrother_loss": 2.3766987323760986, | |
| "eval_bigbrother_runtime": 42.6165, | |
| "eval_bigbrother_samples_per_second": 32.147, | |
| "eval_bigbrother_steps_per_second": 1.009, | |
| "eval_bigbrother_wer": 0.48900346411020706, | |
| "step": 45955 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "eval_NB_RUND_loss": 0.359698623418808, | |
| "eval_NB_RUND_runtime": 84.4668, | |
| "eval_NB_RUND_samples_per_second": 22.234, | |
| "eval_NB_RUND_steps_per_second": 0.698, | |
| "eval_NB_RUND_wer": 0.14400643604183427, | |
| "step": 45955 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "eval_rundkast_loss": 0.32820314168930054, | |
| "eval_rundkast_runtime": 33.145, | |
| "eval_rundkast_samples_per_second": 40.579, | |
| "eval_rundkast_steps_per_second": 1.297, | |
| "eval_rundkast_wer": 0.12313076678332803, | |
| "step": 45955 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "eval_nb_samtale_loss": 0.438490629196167, | |
| "eval_nb_samtale_runtime": 36.783, | |
| "eval_nb_samtale_samples_per_second": 14.49, | |
| "eval_nb_samtale_steps_per_second": 0.462, | |
| "eval_nb_samtale_wer": 0.16552734375, | |
| "step": 45955 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "grad_norm": 9.065438270568848, | |
| "learning_rate": 1.0329282634261075e-05, | |
| "loss": 0.1735, | |
| "step": 47268 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "eval_bigbrother_loss": 2.3984920978546143, | |
| "eval_bigbrother_runtime": 41.0024, | |
| "eval_bigbrother_samples_per_second": 33.413, | |
| "eval_bigbrother_steps_per_second": 1.049, | |
| "eval_bigbrother_wer": 0.4888423427052284, | |
| "step": 47268 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "eval_NB_RUND_loss": 0.3673810064792633, | |
| "eval_NB_RUND_runtime": 85.7426, | |
| "eval_NB_RUND_samples_per_second": 21.903, | |
| "eval_NB_RUND_steps_per_second": 0.688, | |
| "eval_NB_RUND_wer": 0.14384553499597746, | |
| "step": 47268 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "eval_rundkast_loss": 0.3365083932876587, | |
| "eval_rundkast_runtime": 32.9335, | |
| "eval_rundkast_samples_per_second": 40.84, | |
| "eval_rundkast_steps_per_second": 1.306, | |
| "eval_rundkast_wer": 0.12336939230034999, | |
| "step": 47268 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "eval_nb_samtale_loss": 0.44506001472473145, | |
| "eval_nb_samtale_runtime": 36.8094, | |
| "eval_nb_samtale_samples_per_second": 14.48, | |
| "eval_nb_samtale_steps_per_second": 0.462, | |
| "eval_nb_samtale_wer": 0.16463216145833334, | |
| "step": 47268 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "grad_norm": 18.192665100097656, | |
| "learning_rate": 7.757742061936496e-06, | |
| "loss": 0.1731, | |
| "step": 48581 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "eval_bigbrother_loss": 2.399851083755493, | |
| "eval_bigbrother_runtime": 41.7999, | |
| "eval_bigbrother_samples_per_second": 32.775, | |
| "eval_bigbrother_steps_per_second": 1.029, | |
| "eval_bigbrother_wer": 0.49117860307741884, | |
| "step": 48581 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "eval_NB_RUND_loss": 0.3626614809036255, | |
| "eval_NB_RUND_runtime": 84.556, | |
| "eval_NB_RUND_samples_per_second": 22.21, | |
| "eval_NB_RUND_steps_per_second": 0.698, | |
| "eval_NB_RUND_wer": 0.14368463395012068, | |
| "step": 48581 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "eval_rundkast_loss": 0.33268144726753235, | |
| "eval_rundkast_runtime": 33.4742, | |
| "eval_rundkast_samples_per_second": 40.18, | |
| "eval_rundkast_steps_per_second": 1.285, | |
| "eval_rundkast_wer": 0.12289214126630607, | |
| "step": 48581 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "eval_nb_samtale_loss": 0.4378024637699127, | |
| "eval_nb_samtale_runtime": 36.9248, | |
| "eval_nb_samtale_samples_per_second": 14.435, | |
| "eval_nb_samtale_steps_per_second": 0.46, | |
| "eval_nb_samtale_wer": 0.164794921875, | |
| "step": 48581 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 52520, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 40, | |
| "save_steps": 500, | |
| "total_flos": 2.707743528931453e+20, | |
| "train_batch_size": 48, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |