Tflatval's picture
Upload folder using huggingface_hub
364ee58 verified
raw
history blame contribute delete
No virus
53.8 kB
{
"best_metric": 0.2430986748380507,
"best_model_checkpoint": "/cluster/home/torstefl/Master/saved_model/W2V/Combined/BB-NB-RUND/30.05/checkpoint-65754",
"epoch": 40.0,
"eval_steps": 500,
"global_step": 67440,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 1.0,
"grad_norm": 81.12491607666016,
"learning_rate": 9.972247497725206e-05,
"loss": 0.8557,
"step": 1686
},
{
"epoch": 1.0,
"eval_bigbrother_loss": 1.6390442848205566,
"eval_bigbrother_runtime": 54.9493,
"eval_bigbrother_samples_per_second": 24.932,
"eval_bigbrother_steps_per_second": 0.783,
"eval_bigbrother_wer": 0.6789656005800371,
"step": 1686
},
{
"epoch": 1.0,
"eval_BB_NB_RUND_loss": 0.960404634475708,
"eval_BB_NB_RUND_runtime": 215.4543,
"eval_BB_NB_RUND_samples_per_second": 15.047,
"eval_BB_NB_RUND_steps_per_second": 0.473,
"eval_BB_NB_RUND_wer": 0.42735800876273417,
"step": 1686
},
{
"epoch": 1.0,
"eval_rundkast_loss": 0.3885672688484192,
"eval_rundkast_runtime": 39.7428,
"eval_rundkast_samples_per_second": 33.692,
"eval_rundkast_steps_per_second": 1.057,
"eval_rundkast_wer": 0.26931690929451285,
"step": 1686
},
{
"epoch": 1.0,
"eval_nb_samtale_loss": 0.6525214910507202,
"eval_nb_samtale_runtime": 47.5144,
"eval_nb_samtale_samples_per_second": 11.218,
"eval_nb_samtale_steps_per_second": 0.358,
"eval_nb_samtale_wer": 0.3343098958333333,
"step": 1686
},
{
"epoch": 2.0,
"grad_norm": 55.65499496459961,
"learning_rate": 9.71656050955414e-05,
"loss": 0.8819,
"step": 3372
},
{
"epoch": 2.0,
"eval_bigbrother_loss": 1.5301967859268188,
"eval_bigbrother_runtime": 55.6705,
"eval_bigbrother_samples_per_second": 24.609,
"eval_bigbrother_steps_per_second": 0.772,
"eval_bigbrother_wer": 0.6032385402400708,
"step": 3372
},
{
"epoch": 2.0,
"eval_BB_NB_RUND_loss": 0.8895238637924194,
"eval_BB_NB_RUND_runtime": 216.0717,
"eval_BB_NB_RUND_samples_per_second": 15.004,
"eval_BB_NB_RUND_steps_per_second": 0.472,
"eval_BB_NB_RUND_wer": 0.3907211783995914,
"step": 3372
},
{
"epoch": 2.0,
"eval_rundkast_loss": 0.3464510142803192,
"eval_rundkast_runtime": 39.915,
"eval_rundkast_samples_per_second": 33.546,
"eval_rundkast_steps_per_second": 1.052,
"eval_rundkast_wer": 0.2510798272276436,
"step": 3372
},
{
"epoch": 2.0,
"eval_nb_samtale_loss": 0.6075140237808228,
"eval_nb_samtale_runtime": 47.9711,
"eval_nb_samtale_samples_per_second": 11.111,
"eval_nb_samtale_steps_per_second": 0.354,
"eval_nb_samtale_wer": 0.3182779947916667,
"step": 3372
},
{
"epoch": 3.0,
"grad_norm": 29.089189529418945,
"learning_rate": 9.461025174400972e-05,
"loss": 0.8064,
"step": 5058
},
{
"epoch": 3.0,
"eval_bigbrother_loss": 1.5702797174453735,
"eval_bigbrother_runtime": 55.4022,
"eval_bigbrother_samples_per_second": 24.728,
"eval_bigbrother_steps_per_second": 0.776,
"eval_bigbrother_wer": 0.6068637718520905,
"step": 5058
},
{
"epoch": 3.0,
"eval_BB_NB_RUND_loss": 0.8996490240097046,
"eval_BB_NB_RUND_runtime": 238.1867,
"eval_BB_NB_RUND_samples_per_second": 13.611,
"eval_BB_NB_RUND_steps_per_second": 0.428,
"eval_BB_NB_RUND_wer": 0.3820659624223853,
"step": 5058
},
{
"epoch": 3.0,
"eval_rundkast_loss": 0.3366275429725647,
"eval_rundkast_runtime": 39.8274,
"eval_rundkast_samples_per_second": 33.62,
"eval_rundkast_steps_per_second": 1.055,
"eval_rundkast_wer": 0.23484242521196608,
"step": 5058
},
{
"epoch": 3.0,
"eval_nb_samtale_loss": 0.59026038646698,
"eval_nb_samtale_runtime": 47.7348,
"eval_nb_samtale_samples_per_second": 11.166,
"eval_nb_samtale_steps_per_second": 0.356,
"eval_nb_samtale_wer": 0.3050944010416667,
"step": 5058
},
{
"epoch": 4.0,
"grad_norm": 12.38161563873291,
"learning_rate": 9.205338186229906e-05,
"loss": 0.7614,
"step": 6744
},
{
"epoch": 4.0,
"eval_bigbrother_loss": 1.5530073642730713,
"eval_bigbrother_runtime": 55.7815,
"eval_bigbrother_samples_per_second": 24.56,
"eval_bigbrother_steps_per_second": 0.771,
"eval_bigbrother_wer": 0.5924434061065013,
"step": 6744
},
{
"epoch": 4.0,
"eval_BB_NB_RUND_loss": 0.891120970249176,
"eval_BB_NB_RUND_runtime": 215.5766,
"eval_BB_NB_RUND_samples_per_second": 15.039,
"eval_BB_NB_RUND_steps_per_second": 0.473,
"eval_BB_NB_RUND_wer": 0.37961992312447923,
"step": 6744
},
{
"epoch": 4.0,
"eval_rundkast_loss": 0.3337271213531494,
"eval_rundkast_runtime": 39.7645,
"eval_rundkast_samples_per_second": 33.673,
"eval_rundkast_steps_per_second": 1.056,
"eval_rundkast_wer": 0.2345224764037754,
"step": 6744
},
{
"epoch": 4.0,
"eval_nb_samtale_loss": 0.5904266238212585,
"eval_nb_samtale_runtime": 47.79,
"eval_nb_samtale_samples_per_second": 11.153,
"eval_nb_samtale_steps_per_second": 0.356,
"eval_nb_samtale_wer": 0.3128255208333333,
"step": 6744
},
{
"epoch": 5.0,
"grad_norm": 9.62388801574707,
"learning_rate": 8.949802851076737e-05,
"loss": 0.716,
"step": 8430
},
{
"epoch": 5.0,
"eval_bigbrother_loss": 1.3175848722457886,
"eval_bigbrother_runtime": 56.2695,
"eval_bigbrother_samples_per_second": 24.347,
"eval_bigbrother_steps_per_second": 0.764,
"eval_bigbrother_wer": 0.5748006122613389,
"step": 8430
},
{
"epoch": 5.0,
"eval_BB_NB_RUND_loss": 0.7703477144241333,
"eval_BB_NB_RUND_runtime": 213.244,
"eval_BB_NB_RUND_samples_per_second": 15.203,
"eval_BB_NB_RUND_steps_per_second": 0.478,
"eval_BB_NB_RUND_wer": 0.3668521355804639,
"step": 8430
},
{
"epoch": 5.0,
"eval_rundkast_loss": 0.30976158380508423,
"eval_rundkast_runtime": 39.6119,
"eval_rundkast_samples_per_second": 33.803,
"eval_rundkast_steps_per_second": 1.06,
"eval_rundkast_wer": 0.2258038713805791,
"step": 8430
},
{
"epoch": 5.0,
"eval_nb_samtale_loss": 0.520818829536438,
"eval_nb_samtale_runtime": 47.528,
"eval_nb_samtale_samples_per_second": 11.214,
"eval_nb_samtale_steps_per_second": 0.358,
"eval_nb_samtale_wer": 0.3003743489583333,
"step": 8430
},
{
"epoch": 6.0,
"grad_norm": 24.378128051757812,
"learning_rate": 8.694267515923567e-05,
"loss": 0.6852,
"step": 10116
},
{
"epoch": 6.0,
"eval_bigbrother_loss": 1.3726727962493896,
"eval_bigbrother_runtime": 56.2024,
"eval_bigbrother_samples_per_second": 24.376,
"eval_bigbrother_steps_per_second": 0.765,
"eval_bigbrother_wer": 0.5383066140336744,
"step": 10116
},
{
"epoch": 6.0,
"eval_BB_NB_RUND_loss": 0.798953115940094,
"eval_BB_NB_RUND_runtime": 235.4413,
"eval_BB_NB_RUND_samples_per_second": 13.77,
"eval_BB_NB_RUND_steps_per_second": 0.433,
"eval_BB_NB_RUND_wer": 0.3504556084186759,
"step": 10116
},
{
"epoch": 6.0,
"eval_rundkast_loss": 0.30640244483947754,
"eval_rundkast_runtime": 39.638,
"eval_rundkast_samples_per_second": 33.781,
"eval_rundkast_steps_per_second": 1.06,
"eval_rundkast_wer": 0.22228443449048152,
"step": 10116
},
{
"epoch": 6.0,
"eval_nb_samtale_loss": 0.5616967082023621,
"eval_nb_samtale_runtime": 47.4044,
"eval_nb_samtale_samples_per_second": 11.244,
"eval_nb_samtale_steps_per_second": 0.359,
"eval_nb_samtale_wer": 0.2908528645833333,
"step": 10116
},
{
"epoch": 7.0,
"grad_norm": 17.646387100219727,
"learning_rate": 8.438580527752503e-05,
"loss": 0.65,
"step": 11802
},
{
"epoch": 7.0,
"eval_bigbrother_loss": 1.296665906906128,
"eval_bigbrother_runtime": 56.8276,
"eval_bigbrother_samples_per_second": 24.108,
"eval_bigbrother_steps_per_second": 0.757,
"eval_bigbrother_wer": 0.5387899782486103,
"step": 11802
},
{
"epoch": 7.0,
"eval_BB_NB_RUND_loss": 0.7510452270507812,
"eval_BB_NB_RUND_runtime": 219.1946,
"eval_BB_NB_RUND_samples_per_second": 14.791,
"eval_BB_NB_RUND_steps_per_second": 0.465,
"eval_BB_NB_RUND_wer": 0.3475526167244577,
"step": 11802
},
{
"epoch": 7.0,
"eval_rundkast_loss": 0.2876454293727875,
"eval_rundkast_runtime": 39.7396,
"eval_rundkast_samples_per_second": 33.694,
"eval_rundkast_steps_per_second": 1.057,
"eval_rundkast_wer": 0.2191649336106223,
"step": 11802
},
{
"epoch": 7.0,
"eval_nb_samtale_loss": 0.5133534073829651,
"eval_nb_samtale_runtime": 47.6414,
"eval_nb_samtale_samples_per_second": 11.188,
"eval_nb_samtale_steps_per_second": 0.357,
"eval_nb_samtale_wer": 0.2849934895833333,
"step": 11802
},
{
"epoch": 8.0,
"grad_norm": 9.183784484863281,
"learning_rate": 8.183045192599334e-05,
"loss": 0.6195,
"step": 13488
},
{
"epoch": 8.0,
"eval_bigbrother_loss": 1.3931747674942017,
"eval_bigbrother_runtime": 55.455,
"eval_bigbrother_samples_per_second": 24.705,
"eval_bigbrother_steps_per_second": 0.775,
"eval_bigbrother_wer": 0.5308950294046564,
"step": 13488
},
{
"epoch": 8.0,
"eval_BB_NB_RUND_loss": 0.8028880953788757,
"eval_BB_NB_RUND_runtime": 174.7359,
"eval_BB_NB_RUND_samples_per_second": 18.554,
"eval_BB_NB_RUND_steps_per_second": 0.584,
"eval_BB_NB_RUND_wer": 0.34357444292126976,
"step": 13488
},
{
"epoch": 8.0,
"eval_rundkast_loss": 0.31355729699134827,
"eval_rundkast_runtime": 39.8585,
"eval_rundkast_samples_per_second": 33.594,
"eval_rundkast_steps_per_second": 1.054,
"eval_rundkast_wer": 0.21404575267957127,
"step": 13488
},
{
"epoch": 8.0,
"eval_nb_samtale_loss": 0.5146628618240356,
"eval_nb_samtale_runtime": 47.9963,
"eval_nb_samtale_samples_per_second": 11.105,
"eval_nb_samtale_steps_per_second": 0.354,
"eval_nb_samtale_wer": 0.2861328125,
"step": 13488
},
{
"epoch": 9.0,
"grad_norm": 9.670160293579102,
"learning_rate": 7.92735820442827e-05,
"loss": 0.592,
"step": 15174
},
{
"epoch": 9.0,
"eval_bigbrother_loss": 1.3019070625305176,
"eval_bigbrother_runtime": 55.2104,
"eval_bigbrother_samples_per_second": 24.814,
"eval_bigbrother_steps_per_second": 0.779,
"eval_bigbrother_wer": 0.5215499879158946,
"step": 15174
},
{
"epoch": 9.0,
"eval_BB_NB_RUND_loss": 0.7444170117378235,
"eval_BB_NB_RUND_runtime": 237.2263,
"eval_BB_NB_RUND_samples_per_second": 13.666,
"eval_BB_NB_RUND_steps_per_second": 0.43,
"eval_BB_NB_RUND_wer": 0.3310485713517727,
"step": 15174
},
{
"epoch": 9.0,
"eval_rundkast_loss": 0.2764076590538025,
"eval_rundkast_runtime": 40.1719,
"eval_rundkast_samples_per_second": 33.332,
"eval_rundkast_steps_per_second": 1.046,
"eval_rundkast_wer": 0.19956806910894256,
"step": 15174
},
{
"epoch": 9.0,
"eval_nb_samtale_loss": 0.4871513545513153,
"eval_nb_samtale_runtime": 47.9822,
"eval_nb_samtale_samples_per_second": 11.108,
"eval_nb_samtale_steps_per_second": 0.354,
"eval_nb_samtale_wer": 0.2718098958333333,
"step": 15174
},
{
"epoch": 10.0,
"grad_norm": 6.870215892791748,
"learning_rate": 7.671671216257204e-05,
"loss": 0.5668,
"step": 16860
},
{
"epoch": 10.0,
"eval_bigbrother_loss": 1.418025255203247,
"eval_bigbrother_runtime": 56.4643,
"eval_bigbrother_samples_per_second": 24.263,
"eval_bigbrother_steps_per_second": 0.762,
"eval_bigbrother_wer": 0.5194554096511722,
"step": 16860
},
{
"epoch": 10.0,
"eval_BB_NB_RUND_loss": 0.814452052116394,
"eval_BB_NB_RUND_runtime": 187.2873,
"eval_BB_NB_RUND_samples_per_second": 17.31,
"eval_BB_NB_RUND_steps_per_second": 0.545,
"eval_BB_NB_RUND_wer": 0.32844125473752117,
"step": 16860
},
{
"epoch": 10.0,
"eval_rundkast_loss": 0.3035307824611664,
"eval_rundkast_runtime": 39.9861,
"eval_rundkast_samples_per_second": 33.487,
"eval_rundkast_steps_per_second": 1.05,
"eval_rundkast_wer": 0.19844824828027516,
"step": 16860
},
{
"epoch": 10.0,
"eval_nb_samtale_loss": 0.5466247200965881,
"eval_nb_samtale_runtime": 47.5084,
"eval_nb_samtale_samples_per_second": 11.219,
"eval_nb_samtale_steps_per_second": 0.358,
"eval_nb_samtale_wer": 0.267333984375,
"step": 16860
},
{
"epoch": 11.0,
"grad_norm": 5.365777015686035,
"learning_rate": 7.41628753412193e-05,
"loss": 0.5501,
"step": 18546
},
{
"epoch": 11.0,
"eval_bigbrother_loss": 1.413320779800415,
"eval_bigbrother_runtime": 56.7058,
"eval_bigbrother_samples_per_second": 24.16,
"eval_bigbrother_steps_per_second": 0.758,
"eval_bigbrother_wer": 0.5231612019656812,
"step": 18546
},
{
"epoch": 11.0,
"eval_BB_NB_RUND_loss": 0.8070511817932129,
"eval_BB_NB_RUND_runtime": 186.3465,
"eval_BB_NB_RUND_samples_per_second": 17.398,
"eval_BB_NB_RUND_steps_per_second": 0.547,
"eval_BB_NB_RUND_wer": 0.32887132758110904,
"step": 18546
},
{
"epoch": 11.0,
"eval_rundkast_loss": 0.30373072624206543,
"eval_rundkast_runtime": 40.0188,
"eval_rundkast_samples_per_second": 33.459,
"eval_rundkast_steps_per_second": 1.05,
"eval_rundkast_wer": 0.19588865781474965,
"step": 18546
},
{
"epoch": 11.0,
"eval_nb_samtale_loss": 0.5129746198654175,
"eval_nb_samtale_runtime": 47.3772,
"eval_nb_samtale_samples_per_second": 11.25,
"eval_nb_samtale_steps_per_second": 0.359,
"eval_nb_samtale_wer": 0.2676595052083333,
"step": 18546
},
{
"epoch": 12.0,
"grad_norm": 13.1134033203125,
"learning_rate": 7.160600545950865e-05,
"loss": 0.5307,
"step": 20232
},
{
"epoch": 12.0,
"eval_bigbrother_loss": 1.269589900970459,
"eval_bigbrother_runtime": 54.8931,
"eval_bigbrother_samples_per_second": 24.958,
"eval_bigbrother_steps_per_second": 0.783,
"eval_bigbrother_wer": 0.49617336663175704,
"step": 20232
},
{
"epoch": 12.0,
"eval_BB_NB_RUND_loss": 0.7405148148536682,
"eval_BB_NB_RUND_runtime": 189.7158,
"eval_BB_NB_RUND_samples_per_second": 17.089,
"eval_BB_NB_RUND_steps_per_second": 0.538,
"eval_BB_NB_RUND_wer": 0.31540467166626346,
"step": 20232
},
{
"epoch": 12.0,
"eval_rundkast_loss": 0.2901303768157959,
"eval_rundkast_runtime": 39.9118,
"eval_rundkast_samples_per_second": 33.549,
"eval_rundkast_steps_per_second": 1.052,
"eval_rundkast_wer": 0.19612861942089266,
"step": 20232
},
{
"epoch": 12.0,
"eval_nb_samtale_loss": 0.5120977759361267,
"eval_nb_samtale_runtime": 47.5553,
"eval_nb_samtale_samples_per_second": 11.208,
"eval_nb_samtale_steps_per_second": 0.357,
"eval_nb_samtale_wer": 0.2537434895833333,
"step": 20232
},
{
"epoch": 13.0,
"grad_norm": 5.697801113128662,
"learning_rate": 6.9049135577798e-05,
"loss": 0.5134,
"step": 21918
},
{
"epoch": 13.0,
"eval_bigbrother_loss": 1.347646951675415,
"eval_bigbrother_runtime": 55.3541,
"eval_bigbrother_samples_per_second": 24.75,
"eval_bigbrother_steps_per_second": 0.777,
"eval_bigbrother_wer": 0.4957705631193104,
"step": 21918
},
{
"epoch": 13.0,
"eval_BB_NB_RUND_loss": 0.7735500335693359,
"eval_BB_NB_RUND_runtime": 187.3035,
"eval_BB_NB_RUND_samples_per_second": 17.309,
"eval_BB_NB_RUND_steps_per_second": 0.545,
"eval_BB_NB_RUND_wer": 0.3093030131978604,
"step": 21918
},
{
"epoch": 13.0,
"eval_rundkast_loss": 0.29282376170158386,
"eval_rundkast_runtime": 39.9603,
"eval_rundkast_samples_per_second": 33.508,
"eval_rundkast_steps_per_second": 1.051,
"eval_rundkast_wer": 0.18285074388097905,
"step": 21918
},
{
"epoch": 13.0,
"eval_nb_samtale_loss": 0.5057567954063416,
"eval_nb_samtale_runtime": 47.9932,
"eval_nb_samtale_samples_per_second": 11.106,
"eval_nb_samtale_steps_per_second": 0.354,
"eval_nb_samtale_wer": 0.24934895833333334,
"step": 21918
},
{
"epoch": 14.0,
"grad_norm": 8.790960311889648,
"learning_rate": 6.64937822262663e-05,
"loss": 0.4883,
"step": 23604
},
{
"epoch": 14.0,
"eval_bigbrother_loss": 1.3468968868255615,
"eval_bigbrother_runtime": 54.8463,
"eval_bigbrother_samples_per_second": 24.979,
"eval_bigbrother_steps_per_second": 0.784,
"eval_bigbrother_wer": 0.48843953919278177,
"step": 23604
},
{
"epoch": 14.0,
"eval_BB_NB_RUND_loss": 0.7681761384010315,
"eval_BB_NB_RUND_runtime": 174.885,
"eval_BB_NB_RUND_samples_per_second": 18.538,
"eval_BB_NB_RUND_steps_per_second": 0.583,
"eval_BB_NB_RUND_wer": 0.3044646937074967,
"step": 23604
},
{
"epoch": 14.0,
"eval_rundkast_loss": 0.27578702569007874,
"eval_rundkast_runtime": 39.9056,
"eval_rundkast_samples_per_second": 33.554,
"eval_rundkast_steps_per_second": 1.052,
"eval_rundkast_wer": 0.17957126859702446,
"step": 23604
},
{
"epoch": 14.0,
"eval_nb_samtale_loss": 0.5177371501922607,
"eval_nb_samtale_runtime": 47.3502,
"eval_nb_samtale_samples_per_second": 11.257,
"eval_nb_samtale_steps_per_second": 0.359,
"eval_nb_samtale_wer": 0.2451171875,
"step": 23604
},
{
"epoch": 15.0,
"grad_norm": 10.714680671691895,
"learning_rate": 6.393842887473462e-05,
"loss": 0.4689,
"step": 25290
},
{
"epoch": 15.0,
"eval_bigbrother_loss": 1.3112008571624756,
"eval_bigbrother_runtime": 54.8505,
"eval_bigbrother_samples_per_second": 24.977,
"eval_bigbrother_steps_per_second": 0.784,
"eval_bigbrother_wer": 0.49093692096995084,
"step": 25290
},
{
"epoch": 15.0,
"eval_BB_NB_RUND_loss": 0.7539634704589844,
"eval_BB_NB_RUND_runtime": 188.0576,
"eval_BB_NB_RUND_samples_per_second": 17.239,
"eval_BB_NB_RUND_steps_per_second": 0.542,
"eval_BB_NB_RUND_wer": 0.30505604386743007,
"step": 25290
},
{
"epoch": 15.0,
"eval_rundkast_loss": 0.28489232063293457,
"eval_rundkast_runtime": 40.0722,
"eval_rundkast_samples_per_second": 33.415,
"eval_rundkast_steps_per_second": 1.048,
"eval_rundkast_wer": 0.17997120460726285,
"step": 25290
},
{
"epoch": 15.0,
"eval_nb_samtale_loss": 0.4998992383480072,
"eval_nb_samtale_runtime": 47.6004,
"eval_nb_samtale_samples_per_second": 11.197,
"eval_nb_samtale_steps_per_second": 0.357,
"eval_nb_samtale_wer": 0.24503580729166666,
"step": 25290
},
{
"epoch": 16.0,
"grad_norm": 5.148988246917725,
"learning_rate": 6.138459205338186e-05,
"loss": 0.4555,
"step": 26976
},
{
"epoch": 16.0,
"eval_bigbrother_loss": 1.3192588090896606,
"eval_bigbrother_runtime": 54.8083,
"eval_bigbrother_samples_per_second": 24.996,
"eval_bigbrother_steps_per_second": 0.785,
"eval_bigbrother_wer": 0.475388705389511,
"step": 26976
},
{
"epoch": 16.0,
"eval_BB_NB_RUND_loss": 0.7727116942405701,
"eval_BB_NB_RUND_runtime": 175.6567,
"eval_BB_NB_RUND_samples_per_second": 18.456,
"eval_BB_NB_RUND_steps_per_second": 0.581,
"eval_BB_NB_RUND_wer": 0.2980404806064027,
"step": 26976
},
{
"epoch": 16.0,
"eval_rundkast_loss": 0.30636081099510193,
"eval_rundkast_runtime": 40.0387,
"eval_rundkast_samples_per_second": 33.443,
"eval_rundkast_steps_per_second": 1.049,
"eval_rundkast_wer": 0.17501199808030715,
"step": 26976
},
{
"epoch": 16.0,
"eval_nb_samtale_loss": 0.5385606288909912,
"eval_nb_samtale_runtime": 47.4641,
"eval_nb_samtale_samples_per_second": 11.23,
"eval_nb_samtale_steps_per_second": 0.358,
"eval_nb_samtale_wer": 0.24462890625,
"step": 26976
},
{
"epoch": 17.0,
"grad_norm": 7.1762614250183105,
"learning_rate": 5.8827722171671216e-05,
"loss": 0.4414,
"step": 28662
},
{
"epoch": 17.0,
"eval_bigbrother_loss": 1.3271602392196655,
"eval_bigbrother_runtime": 55.1307,
"eval_bigbrother_samples_per_second": 24.85,
"eval_bigbrother_steps_per_second": 0.78,
"eval_bigbrother_wer": 0.4732135664222992,
"step": 28662
},
{
"epoch": 17.0,
"eval_BB_NB_RUND_loss": 0.7706267237663269,
"eval_BB_NB_RUND_runtime": 186.2927,
"eval_BB_NB_RUND_samples_per_second": 17.403,
"eval_BB_NB_RUND_steps_per_second": 0.548,
"eval_BB_NB_RUND_wer": 0.2966427438647421,
"step": 28662
},
{
"epoch": 17.0,
"eval_rundkast_loss": 0.29671627283096313,
"eval_rundkast_runtime": 39.937,
"eval_rundkast_samples_per_second": 33.528,
"eval_rundkast_steps_per_second": 1.052,
"eval_rundkast_wer": 0.17677171652535595,
"step": 28662
},
{
"epoch": 17.0,
"eval_nb_samtale_loss": 0.5305598378181458,
"eval_nb_samtale_runtime": 47.5425,
"eval_nb_samtale_samples_per_second": 11.211,
"eval_nb_samtale_steps_per_second": 0.358,
"eval_nb_samtale_wer": 0.24007161458333334,
"step": 28662
},
{
"epoch": 18.0,
"grad_norm": 7.369349956512451,
"learning_rate": 5.627085228996057e-05,
"loss": 0.4215,
"step": 30348
},
{
"epoch": 18.0,
"eval_bigbrother_loss": 1.396024465560913,
"eval_bigbrother_runtime": 54.7768,
"eval_bigbrother_samples_per_second": 25.011,
"eval_bigbrother_steps_per_second": 0.785,
"eval_bigbrother_wer": 0.47377749133972447,
"step": 30348
},
{
"epoch": 18.0,
"eval_BB_NB_RUND_loss": 0.8181838393211365,
"eval_BB_NB_RUND_runtime": 176.3618,
"eval_BB_NB_RUND_samples_per_second": 18.383,
"eval_BB_NB_RUND_steps_per_second": 0.578,
"eval_BB_NB_RUND_wer": 0.2952450071230815,
"step": 30348
},
{
"epoch": 18.0,
"eval_rundkast_loss": 0.3431045413017273,
"eval_rundkast_runtime": 40.0058,
"eval_rundkast_samples_per_second": 33.47,
"eval_rundkast_steps_per_second": 1.05,
"eval_rundkast_wer": 0.17389217725163975,
"step": 30348
},
{
"epoch": 18.0,
"eval_nb_samtale_loss": 0.5264986753463745,
"eval_nb_samtale_runtime": 47.5607,
"eval_nb_samtale_samples_per_second": 11.207,
"eval_nb_samtale_steps_per_second": 0.357,
"eval_nb_samtale_wer": 0.237548828125,
"step": 30348
},
{
"epoch": 19.0,
"grad_norm": 5.667070388793945,
"learning_rate": 5.371549893842888e-05,
"loss": 0.407,
"step": 32034
},
{
"epoch": 19.0,
"eval_bigbrother_loss": 1.402126669883728,
"eval_bigbrother_runtime": 55.4841,
"eval_bigbrother_samples_per_second": 24.692,
"eval_bigbrother_steps_per_second": 0.775,
"eval_bigbrother_wer": 0.4772416015467655,
"step": 32034
},
{
"epoch": 19.0,
"eval_BB_NB_RUND_loss": 0.8144161701202393,
"eval_BB_NB_RUND_runtime": 193.9029,
"eval_BB_NB_RUND_samples_per_second": 16.72,
"eval_BB_NB_RUND_steps_per_second": 0.526,
"eval_BB_NB_RUND_wer": 0.29502997070128756,
"step": 32034
},
{
"epoch": 19.0,
"eval_rundkast_loss": 0.3256075978279114,
"eval_rundkast_runtime": 40.5438,
"eval_rundkast_samples_per_second": 33.026,
"eval_rundkast_steps_per_second": 1.036,
"eval_rundkast_wer": 0.17381219004959206,
"step": 32034
},
{
"epoch": 19.0,
"eval_nb_samtale_loss": 0.5318475365638733,
"eval_nb_samtale_runtime": 47.61,
"eval_nb_samtale_samples_per_second": 11.195,
"eval_nb_samtale_steps_per_second": 0.357,
"eval_nb_samtale_wer": 0.234619140625,
"step": 32034
},
{
"epoch": 20.0,
"grad_norm": NaN,
"learning_rate": 5.116166211707614e-05,
"loss": 0.394,
"step": 33720
},
{
"epoch": 20.0,
"eval_bigbrother_loss": 1.3861628770828247,
"eval_bigbrother_runtime": 55.5803,
"eval_bigbrother_samples_per_second": 24.649,
"eval_bigbrother_steps_per_second": 0.774,
"eval_bigbrother_wer": 0.47168291307500204,
"step": 33720
},
{
"epoch": 20.0,
"eval_BB_NB_RUND_loss": 0.8024275302886963,
"eval_BB_NB_RUND_runtime": 184.6934,
"eval_BB_NB_RUND_samples_per_second": 17.553,
"eval_BB_NB_RUND_steps_per_second": 0.552,
"eval_BB_NB_RUND_wer": 0.28771873236029355,
"step": 33720
},
{
"epoch": 20.0,
"eval_rundkast_loss": 0.31396913528442383,
"eval_rundkast_runtime": 40.1483,
"eval_rundkast_samples_per_second": 33.351,
"eval_rundkast_steps_per_second": 1.046,
"eval_rundkast_wer": 0.16221404575267956,
"step": 33720
},
{
"epoch": 20.0,
"eval_nb_samtale_loss": 0.5292519927024841,
"eval_nb_samtale_runtime": 47.252,
"eval_nb_samtale_samples_per_second": 11.28,
"eval_nb_samtale_steps_per_second": 0.36,
"eval_nb_samtale_wer": 0.229248046875,
"step": 33720
},
{
"epoch": 21.0,
"grad_norm": 10.585386276245117,
"learning_rate": 4.860479223536549e-05,
"loss": 0.3812,
"step": 35406
},
{
"epoch": 21.0,
"eval_bigbrother_loss": 1.368707537651062,
"eval_bigbrother_runtime": 55.1721,
"eval_bigbrother_samples_per_second": 24.831,
"eval_bigbrother_steps_per_second": 0.779,
"eval_bigbrother_wer": 0.45702086522194474,
"step": 35406
},
{
"epoch": 21.0,
"eval_BB_NB_RUND_loss": 0.7922731041908264,
"eval_BB_NB_RUND_runtime": 186.0861,
"eval_BB_NB_RUND_samples_per_second": 17.422,
"eval_BB_NB_RUND_steps_per_second": 0.548,
"eval_BB_NB_RUND_wer": 0.2828804128699298,
"step": 35406
},
{
"epoch": 21.0,
"eval_rundkast_loss": 0.3038511872291565,
"eval_rundkast_runtime": 40.036,
"eval_rundkast_samples_per_second": 33.445,
"eval_rundkast_steps_per_second": 1.049,
"eval_rundkast_wer": 0.16437370020796674,
"step": 35406
},
{
"epoch": 21.0,
"eval_nb_samtale_loss": 0.5377490520477295,
"eval_nb_samtale_runtime": 47.8978,
"eval_nb_samtale_samples_per_second": 11.128,
"eval_nb_samtale_steps_per_second": 0.355,
"eval_nb_samtale_wer": 0.227294921875,
"step": 35406
},
{
"epoch": 22.0,
"grad_norm": 5.612251281738281,
"learning_rate": 4.604792235365484e-05,
"loss": 0.3656,
"step": 37092
},
{
"epoch": 22.0,
"eval_bigbrother_loss": 1.3479688167572021,
"eval_bigbrother_runtime": 55.3759,
"eval_bigbrother_samples_per_second": 24.74,
"eval_bigbrother_steps_per_second": 0.777,
"eval_bigbrother_wer": 0.46588254249577055,
"step": 37092
},
{
"epoch": 22.0,
"eval_BB_NB_RUND_loss": 0.7788259387016296,
"eval_BB_NB_RUND_runtime": 186.5504,
"eval_BB_NB_RUND_samples_per_second": 17.379,
"eval_BB_NB_RUND_steps_per_second": 0.547,
"eval_BB_NB_RUND_wer": 0.285890922775045,
"step": 37092
},
{
"epoch": 22.0,
"eval_rundkast_loss": 0.30141252279281616,
"eval_rundkast_runtime": 40.0476,
"eval_rundkast_samples_per_second": 33.435,
"eval_rundkast_steps_per_second": 1.049,
"eval_rundkast_wer": 0.16549352103663414,
"step": 37092
},
{
"epoch": 22.0,
"eval_nb_samtale_loss": 0.5157301425933838,
"eval_nb_samtale_runtime": 47.2633,
"eval_nb_samtale_samples_per_second": 11.277,
"eval_nb_samtale_steps_per_second": 0.36,
"eval_nb_samtale_wer": 0.22599283854166666,
"step": 37092
},
{
"epoch": 23.0,
"grad_norm": 21.909698486328125,
"learning_rate": 4.349560206248104e-05,
"loss": 0.3507,
"step": 38778
},
{
"epoch": 23.0,
"eval_bigbrother_loss": 1.436195731163025,
"eval_bigbrother_runtime": 55.1641,
"eval_bigbrother_samples_per_second": 24.835,
"eval_bigbrother_steps_per_second": 0.779,
"eval_bigbrother_wer": 0.45677918311447674,
"step": 38778
},
{
"epoch": 23.0,
"eval_BB_NB_RUND_loss": 0.8368074893951416,
"eval_BB_NB_RUND_runtime": 174.9632,
"eval_BB_NB_RUND_samples_per_second": 18.53,
"eval_BB_NB_RUND_steps_per_second": 0.583,
"eval_BB_NB_RUND_wer": 0.2791979141467086,
"step": 38778
},
{
"epoch": 23.0,
"eval_rundkast_loss": 0.3419454097747803,
"eval_rundkast_runtime": 40.1893,
"eval_rundkast_samples_per_second": 33.317,
"eval_rundkast_steps_per_second": 1.045,
"eval_rundkast_wer": 0.15853463445848665,
"step": 38778
},
{
"epoch": 23.0,
"eval_nb_samtale_loss": 0.5392637848854065,
"eval_nb_samtale_runtime": 47.8657,
"eval_nb_samtale_samples_per_second": 11.135,
"eval_nb_samtale_steps_per_second": 0.355,
"eval_nb_samtale_wer": 0.22233072916666666,
"step": 38778
},
{
"epoch": 24.0,
"grad_norm": 17.75894546508789,
"learning_rate": 4.09387321807704e-05,
"loss": 0.3407,
"step": 40464
},
{
"epoch": 24.0,
"eval_bigbrother_loss": 1.394923448562622,
"eval_bigbrother_runtime": 55.3073,
"eval_bigbrother_samples_per_second": 24.771,
"eval_bigbrother_steps_per_second": 0.777,
"eval_bigbrother_wer": 0.4556513332796262,
"step": 40464
},
{
"epoch": 24.0,
"eval_BB_NB_RUND_loss": 0.8072466850280762,
"eval_BB_NB_RUND_runtime": 184.4773,
"eval_BB_NB_RUND_samples_per_second": 17.574,
"eval_BB_NB_RUND_steps_per_second": 0.553,
"eval_BB_NB_RUND_wer": 0.279681746095745,
"step": 40464
},
{
"epoch": 24.0,
"eval_rundkast_loss": 0.32181429862976074,
"eval_rundkast_runtime": 40.2553,
"eval_rundkast_samples_per_second": 33.263,
"eval_rundkast_steps_per_second": 1.043,
"eval_rundkast_wer": 0.15709486482162854,
"step": 40464
},
{
"epoch": 24.0,
"eval_nb_samtale_loss": 0.5160051584243774,
"eval_nb_samtale_runtime": 48.428,
"eval_nb_samtale_samples_per_second": 11.006,
"eval_nb_samtale_steps_per_second": 0.351,
"eval_nb_samtale_wer": 0.22639973958333334,
"step": 40464
},
{
"epoch": 25.0,
"grad_norm": 7.524557590484619,
"learning_rate": 3.838186229905975e-05,
"loss": 0.3306,
"step": 42150
},
{
"epoch": 25.0,
"eval_bigbrother_loss": 1.4151748418807983,
"eval_bigbrother_runtime": 54.9262,
"eval_bigbrother_samples_per_second": 24.943,
"eval_bigbrother_steps_per_second": 0.783,
"eval_bigbrother_wer": 0.45146217675018124,
"step": 42150
},
{
"epoch": 25.0,
"eval_BB_NB_RUND_loss": 0.8287575840950012,
"eval_BB_NB_RUND_runtime": 184.6246,
"eval_BB_NB_RUND_samples_per_second": 17.56,
"eval_BB_NB_RUND_steps_per_second": 0.552,
"eval_BB_NB_RUND_wer": 0.2747896674999328,
"step": 42150
},
{
"epoch": 25.0,
"eval_rundkast_loss": 0.34578415751457214,
"eval_rundkast_runtime": 40.2638,
"eval_rundkast_samples_per_second": 33.256,
"eval_rundkast_steps_per_second": 1.043,
"eval_rundkast_wer": 0.15877459606462965,
"step": 42150
},
{
"epoch": 25.0,
"eval_nb_samtale_loss": 0.5345997214317322,
"eval_nb_samtale_runtime": 47.4285,
"eval_nb_samtale_samples_per_second": 11.238,
"eval_nb_samtale_steps_per_second": 0.358,
"eval_nb_samtale_wer": 0.214111328125,
"step": 42150
},
{
"epoch": 26.0,
"grad_norm": 45.58380126953125,
"learning_rate": 3.582650894752806e-05,
"loss": 0.3192,
"step": 43836
},
{
"epoch": 26.0,
"eval_bigbrother_loss": 1.5123320817947388,
"eval_bigbrother_runtime": 54.6916,
"eval_bigbrother_samples_per_second": 25.05,
"eval_bigbrother_steps_per_second": 0.786,
"eval_bigbrother_wer": 0.4467896560058004,
"step": 43836
},
{
"epoch": 26.0,
"eval_BB_NB_RUND_loss": 0.8734822869300842,
"eval_BB_NB_RUND_runtime": 185.5859,
"eval_BB_NB_RUND_samples_per_second": 17.469,
"eval_BB_NB_RUND_steps_per_second": 0.55,
"eval_BB_NB_RUND_wer": 0.27237050775475097,
"step": 43836
},
{
"epoch": 26.0,
"eval_rundkast_loss": 0.3513795733451843,
"eval_rundkast_runtime": 40.0168,
"eval_rundkast_samples_per_second": 33.461,
"eval_rundkast_steps_per_second": 1.05,
"eval_rundkast_wer": 0.1575747880339146,
"step": 43836
},
{
"epoch": 26.0,
"eval_nb_samtale_loss": 0.5426519513130188,
"eval_nb_samtale_runtime": 47.2324,
"eval_nb_samtale_samples_per_second": 11.285,
"eval_nb_samtale_steps_per_second": 0.36,
"eval_nb_samtale_wer": 0.212890625,
"step": 43836
},
{
"epoch": 27.0,
"grad_norm": 51.679561614990234,
"learning_rate": 3.326963906581741e-05,
"loss": 0.3092,
"step": 45522
},
{
"epoch": 27.0,
"eval_bigbrother_loss": 1.4591954946517944,
"eval_bigbrother_runtime": 54.7449,
"eval_bigbrother_samples_per_second": 25.025,
"eval_bigbrother_steps_per_second": 0.785,
"eval_bigbrother_wer": 0.44324498509627,
"step": 45522
},
{
"epoch": 27.0,
"eval_BB_NB_RUND_loss": 0.8474361896514893,
"eval_BB_NB_RUND_runtime": 185.4261,
"eval_BB_NB_RUND_samples_per_second": 17.484,
"eval_BB_NB_RUND_steps_per_second": 0.55,
"eval_BB_NB_RUND_wer": 0.2704351799586055,
"step": 45522
},
{
"epoch": 27.0,
"eval_rundkast_loss": 0.34611085057258606,
"eval_rundkast_runtime": 39.9599,
"eval_rundkast_samples_per_second": 33.509,
"eval_rundkast_steps_per_second": 1.051,
"eval_rundkast_wer": 0.15381538953767399,
"step": 45522
},
{
"epoch": 27.0,
"eval_nb_samtale_loss": 0.5338240265846252,
"eval_nb_samtale_runtime": 47.1834,
"eval_nb_samtale_samples_per_second": 11.296,
"eval_nb_samtale_steps_per_second": 0.36,
"eval_nb_samtale_wer": 0.21500651041666666,
"step": 45522
},
{
"epoch": 28.0,
"grad_norm": 6.035738468170166,
"learning_rate": 3.071276918410677e-05,
"loss": 0.2935,
"step": 47208
},
{
"epoch": 28.0,
"eval_bigbrother_loss": 1.4708665609359741,
"eval_bigbrother_runtime": 55.0307,
"eval_bigbrother_samples_per_second": 24.895,
"eval_bigbrother_steps_per_second": 0.781,
"eval_bigbrother_wer": 0.4352694755498268,
"step": 47208
},
{
"epoch": 28.0,
"eval_BB_NB_RUND_loss": 0.8631957173347473,
"eval_BB_NB_RUND_runtime": 250.4068,
"eval_BB_NB_RUND_samples_per_second": 12.947,
"eval_BB_NB_RUND_steps_per_second": 0.407,
"eval_BB_NB_RUND_wer": 0.26624196973362363,
"step": 47208
},
{
"epoch": 28.0,
"eval_rundkast_loss": 0.35267603397369385,
"eval_rundkast_runtime": 40.1826,
"eval_rundkast_samples_per_second": 33.323,
"eval_rundkast_steps_per_second": 1.045,
"eval_rundkast_wer": 0.1529355303151496,
"step": 47208
},
{
"epoch": 28.0,
"eval_nb_samtale_loss": 0.583747923374176,
"eval_nb_samtale_runtime": 47.2983,
"eval_nb_samtale_samples_per_second": 11.269,
"eval_nb_samtale_steps_per_second": 0.359,
"eval_nb_samtale_wer": 0.21053059895833334,
"step": 47208
},
{
"epoch": 29.0,
"grad_norm": 6.352161407470703,
"learning_rate": 2.8155899302396122e-05,
"loss": 0.2854,
"step": 48894
},
{
"epoch": 29.0,
"eval_bigbrother_loss": 1.5383168458938599,
"eval_bigbrother_runtime": 54.9702,
"eval_bigbrother_samples_per_second": 24.923,
"eval_bigbrother_steps_per_second": 0.782,
"eval_bigbrother_wer": 0.4352694755498268,
"step": 48894
},
{
"epoch": 29.0,
"eval_BB_NB_RUND_loss": 0.8866830468177795,
"eval_BB_NB_RUND_runtime": 184.2665,
"eval_BB_NB_RUND_samples_per_second": 17.594,
"eval_BB_NB_RUND_steps_per_second": 0.554,
"eval_BB_NB_RUND_wer": 0.2632583393812327,
"step": 48894
},
{
"epoch": 29.0,
"eval_rundkast_loss": 0.3528152108192444,
"eval_rundkast_runtime": 40.0458,
"eval_rundkast_samples_per_second": 33.437,
"eval_rundkast_steps_per_second": 1.049,
"eval_rundkast_wer": 0.1476563749800032,
"step": 48894
},
{
"epoch": 29.0,
"eval_nb_samtale_loss": 0.5526814460754395,
"eval_nb_samtale_runtime": 47.2216,
"eval_nb_samtale_samples_per_second": 11.287,
"eval_nb_samtale_steps_per_second": 0.36,
"eval_nb_samtale_wer": 0.20662434895833334,
"step": 48894
},
{
"epoch": 30.0,
"grad_norm": 16.592275619506836,
"learning_rate": 2.5600545950864424e-05,
"loss": 0.2743,
"step": 50580
},
{
"epoch": 30.0,
"eval_bigbrother_loss": 1.491037130355835,
"eval_bigbrother_runtime": 55.562,
"eval_bigbrother_samples_per_second": 24.657,
"eval_bigbrother_steps_per_second": 0.774,
"eval_bigbrother_wer": 0.4344638685249335,
"step": 50580
},
{
"epoch": 30.0,
"eval_BB_NB_RUND_loss": 0.8672559261322021,
"eval_BB_NB_RUND_runtime": 206.7209,
"eval_BB_NB_RUND_samples_per_second": 15.683,
"eval_BB_NB_RUND_steps_per_second": 0.493,
"eval_BB_NB_RUND_wer": 0.2624519527995054,
"step": 50580
},
{
"epoch": 30.0,
"eval_rundkast_loss": 0.352713406085968,
"eval_rundkast_runtime": 40.2922,
"eval_rundkast_samples_per_second": 33.232,
"eval_rundkast_steps_per_second": 1.042,
"eval_rundkast_wer": 0.1478963365861462,
"step": 50580
},
{
"epoch": 30.0,
"eval_nb_samtale_loss": 0.5577544569969177,
"eval_nb_samtale_runtime": 47.5456,
"eval_nb_samtale_samples_per_second": 11.21,
"eval_nb_samtale_steps_per_second": 0.358,
"eval_nb_samtale_wer": 0.20467122395833334,
"step": 50580
},
{
"epoch": 31.0,
"grad_norm": 45.161354064941406,
"learning_rate": 2.3045192599332726e-05,
"loss": 0.2642,
"step": 52266
},
{
"epoch": 31.0,
"eval_bigbrother_loss": 1.5767817497253418,
"eval_bigbrother_runtime": 55.8334,
"eval_bigbrother_samples_per_second": 24.537,
"eval_bigbrother_steps_per_second": 0.77,
"eval_bigbrother_wer": 0.43317489728510433,
"step": 52266
},
{
"epoch": 31.0,
"eval_BB_NB_RUND_loss": 0.9109604358673096,
"eval_BB_NB_RUND_runtime": 200.853,
"eval_BB_NB_RUND_samples_per_second": 16.141,
"eval_BB_NB_RUND_steps_per_second": 0.508,
"eval_BB_NB_RUND_wer": 0.2612692524796387,
"step": 52266
},
{
"epoch": 31.0,
"eval_rundkast_loss": 0.3539213240146637,
"eval_rundkast_runtime": 40.1832,
"eval_rundkast_samples_per_second": 33.322,
"eval_rundkast_steps_per_second": 1.045,
"eval_rundkast_wer": 0.14285714285714285,
"step": 52266
},
{
"epoch": 31.0,
"eval_nb_samtale_loss": 0.5987378358840942,
"eval_nb_samtale_runtime": 47.6989,
"eval_nb_samtale_samples_per_second": 11.174,
"eval_nb_samtale_steps_per_second": 0.356,
"eval_nb_samtale_wer": 0.20768229166666666,
"step": 52266
},
{
"epoch": 32.0,
"grad_norm": 11.585915565490723,
"learning_rate": 2.0489839247801032e-05,
"loss": 0.2547,
"step": 53952
},
{
"epoch": 32.0,
"eval_bigbrother_loss": 1.529497742652893,
"eval_bigbrother_runtime": 55.5603,
"eval_bigbrother_samples_per_second": 24.658,
"eval_bigbrother_steps_per_second": 0.774,
"eval_bigbrother_wer": 0.42954966567308467,
"step": 53952
},
{
"epoch": 32.0,
"eval_BB_NB_RUND_loss": 0.8869587779045105,
"eval_BB_NB_RUND_runtime": 185.554,
"eval_BB_NB_RUND_samples_per_second": 17.472,
"eval_BB_NB_RUND_steps_per_second": 0.55,
"eval_BB_NB_RUND_wer": 0.25793618794183265,
"step": 53952
},
{
"epoch": 32.0,
"eval_rundkast_loss": 0.3568912148475647,
"eval_rundkast_runtime": 40.0636,
"eval_rundkast_samples_per_second": 33.422,
"eval_rundkast_steps_per_second": 1.048,
"eval_rundkast_wer": 0.14397696368581028,
"step": 53952
},
{
"epoch": 32.0,
"eval_nb_samtale_loss": 0.566994309425354,
"eval_nb_samtale_runtime": 47.5483,
"eval_nb_samtale_samples_per_second": 11.21,
"eval_nb_samtale_steps_per_second": 0.358,
"eval_nb_samtale_wer": 0.20003255208333334,
"step": 53952
},
{
"epoch": 33.0,
"grad_norm": 6.523238182067871,
"learning_rate": 1.7932969366090385e-05,
"loss": 0.2471,
"step": 55638
},
{
"epoch": 33.0,
"eval_bigbrother_loss": 1.589526891708374,
"eval_bigbrother_runtime": 54.9785,
"eval_bigbrother_samples_per_second": 24.919,
"eval_bigbrother_steps_per_second": 0.782,
"eval_bigbrother_wer": 0.41980182067187627,
"step": 55638
},
{
"epoch": 33.0,
"eval_BB_NB_RUND_loss": 0.9184404611587524,
"eval_BB_NB_RUND_runtime": 176.6606,
"eval_BB_NB_RUND_samples_per_second": 18.352,
"eval_BB_NB_RUND_steps_per_second": 0.577,
"eval_BB_NB_RUND_wer": 0.25258715694970835,
"step": 55638
},
{
"epoch": 33.0,
"eval_rundkast_loss": 0.3644249141216278,
"eval_rundkast_runtime": 40.0915,
"eval_rundkast_samples_per_second": 33.399,
"eval_rundkast_steps_per_second": 1.048,
"eval_rundkast_wer": 0.14117741161414174,
"step": 55638
},
{
"epoch": 33.0,
"eval_nb_samtale_loss": 0.5854535102844238,
"eval_nb_samtale_runtime": 47.6556,
"eval_nb_samtale_samples_per_second": 11.184,
"eval_nb_samtale_steps_per_second": 0.357,
"eval_nb_samtale_wer": 0.19669596354166666,
"step": 55638
},
{
"epoch": 34.0,
"grad_norm": 6.739225387573242,
"learning_rate": 1.5376099484379742e-05,
"loss": 0.2391,
"step": 57324
},
{
"epoch": 34.0,
"eval_bigbrother_loss": 1.5508168935775757,
"eval_bigbrother_runtime": 55.2494,
"eval_bigbrother_samples_per_second": 24.797,
"eval_bigbrother_steps_per_second": 0.778,
"eval_bigbrother_wer": 0.4182711673245791,
"step": 57324
},
{
"epoch": 34.0,
"eval_BB_NB_RUND_loss": 0.9037603139877319,
"eval_BB_NB_RUND_runtime": 189.0492,
"eval_BB_NB_RUND_samples_per_second": 17.149,
"eval_BB_NB_RUND_steps_per_second": 0.54,
"eval_BB_NB_RUND_wer": 0.2511356611025993,
"step": 57324
},
{
"epoch": 34.0,
"eval_rundkast_loss": 0.37484127283096313,
"eval_rundkast_runtime": 40.5007,
"eval_rundkast_samples_per_second": 33.061,
"eval_rundkast_steps_per_second": 1.037,
"eval_rundkast_wer": 0.1410174372100464,
"step": 57324
},
{
"epoch": 34.0,
"eval_nb_samtale_loss": 0.5698674917221069,
"eval_nb_samtale_runtime": 47.7367,
"eval_nb_samtale_samples_per_second": 11.165,
"eval_nb_samtale_steps_per_second": 0.356,
"eval_nb_samtale_wer": 0.19417317708333334,
"step": 57324
},
{
"epoch": 35.0,
"grad_norm": 8.39053726196289,
"learning_rate": 1.2819229602669094e-05,
"loss": 0.2279,
"step": 59010
},
{
"epoch": 35.0,
"eval_bigbrother_loss": 1.5247896909713745,
"eval_bigbrother_runtime": 55.883,
"eval_bigbrother_samples_per_second": 24.516,
"eval_bigbrother_steps_per_second": 0.769,
"eval_bigbrother_wer": 0.41843228872955773,
"step": 59010
},
{
"epoch": 35.0,
"eval_BB_NB_RUND_loss": 0.8858203291893005,
"eval_BB_NB_RUND_runtime": 186.0172,
"eval_BB_NB_RUND_samples_per_second": 17.428,
"eval_BB_NB_RUND_steps_per_second": 0.548,
"eval_BB_NB_RUND_wer": 0.24989920167728408,
"step": 59010
},
{
"epoch": 35.0,
"eval_rundkast_loss": 0.3675382137298584,
"eval_rundkast_runtime": 39.7117,
"eval_rundkast_samples_per_second": 33.718,
"eval_rundkast_steps_per_second": 1.058,
"eval_rundkast_wer": 0.13845784674452088,
"step": 59010
},
{
"epoch": 35.0,
"eval_nb_samtale_loss": 0.5457141995429993,
"eval_nb_samtale_runtime": 47.1407,
"eval_nb_samtale_samples_per_second": 11.307,
"eval_nb_samtale_steps_per_second": 0.361,
"eval_nb_samtale_wer": 0.19327799479166666,
"step": 59010
},
{
"epoch": 36.0,
"grad_norm": 32.44700622558594,
"learning_rate": 1.0263876251137398e-05,
"loss": 0.2202,
"step": 60696
},
{
"epoch": 36.0,
"eval_bigbrother_loss": 1.6295210123062134,
"eval_bigbrother_runtime": 55.7536,
"eval_bigbrother_samples_per_second": 24.572,
"eval_bigbrother_steps_per_second": 0.771,
"eval_bigbrother_wer": 0.4164988318698139,
"step": 60696
},
{
"epoch": 36.0,
"eval_BB_NB_RUND_loss": 0.9496392607688904,
"eval_BB_NB_RUND_runtime": 187.239,
"eval_BB_NB_RUND_samples_per_second": 17.315,
"eval_BB_NB_RUND_steps_per_second": 0.545,
"eval_BB_NB_RUND_wer": 0.24809827164475984,
"step": 60696
},
{
"epoch": 36.0,
"eval_rundkast_loss": 0.41033124923706055,
"eval_rundkast_runtime": 39.7349,
"eval_rundkast_samples_per_second": 33.698,
"eval_rundkast_steps_per_second": 1.057,
"eval_rundkast_wer": 0.1346184610462326,
"step": 60696
},
{
"epoch": 36.0,
"eval_nb_samtale_loss": 0.5565491318702698,
"eval_nb_samtale_runtime": 47.2102,
"eval_nb_samtale_samples_per_second": 11.29,
"eval_nb_samtale_steps_per_second": 0.36,
"eval_nb_samtale_wer": 0.193115234375,
"step": 60696
},
{
"epoch": 37.0,
"grad_norm": 6.115579128265381,
"learning_rate": 7.710039429784654e-06,
"loss": 0.2134,
"step": 62382
},
{
"epoch": 37.0,
"eval_bigbrother_loss": 1.6113148927688599,
"eval_bigbrother_runtime": 55.3118,
"eval_bigbrother_samples_per_second": 24.769,
"eval_bigbrother_steps_per_second": 0.777,
"eval_bigbrother_wer": 0.42068798839925886,
"step": 62382
},
{
"epoch": 37.0,
"eval_BB_NB_RUND_loss": 0.9415259957313538,
"eval_BB_NB_RUND_runtime": 186.5349,
"eval_BB_NB_RUND_samples_per_second": 17.38,
"eval_BB_NB_RUND_steps_per_second": 0.547,
"eval_BB_NB_RUND_wer": 0.24963040615004167,
"step": 62382
},
{
"epoch": 37.0,
"eval_rundkast_loss": 0.40727370977401733,
"eval_rundkast_runtime": 40.0362,
"eval_rundkast_samples_per_second": 33.445,
"eval_rundkast_steps_per_second": 1.049,
"eval_rundkast_wer": 0.1369380899056151,
"step": 62382
},
{
"epoch": 37.0,
"eval_nb_samtale_loss": 0.5639351606369019,
"eval_nb_samtale_runtime": 46.907,
"eval_nb_samtale_samples_per_second": 11.363,
"eval_nb_samtale_steps_per_second": 0.362,
"eval_nb_samtale_wer": 0.19140625,
"step": 62382
},
{
"epoch": 38.0,
"grad_norm": 10.58083438873291,
"learning_rate": 5.153169548074007e-06,
"loss": 0.2094,
"step": 64068
},
{
"epoch": 38.0,
"eval_bigbrother_loss": 1.6277378797531128,
"eval_bigbrother_runtime": 54.6605,
"eval_bigbrother_samples_per_second": 25.064,
"eval_bigbrother_steps_per_second": 0.787,
"eval_bigbrother_wer": 0.41037621848062517,
"step": 64068
},
{
"epoch": 38.0,
"eval_BB_NB_RUND_loss": 0.9503330588340759,
"eval_BB_NB_RUND_runtime": 187.301,
"eval_BB_NB_RUND_samples_per_second": 17.309,
"eval_BB_NB_RUND_steps_per_second": 0.545,
"eval_BB_NB_RUND_wer": 0.24508776173964467,
"step": 64068
},
{
"epoch": 38.0,
"eval_rundkast_loss": 0.4076659679412842,
"eval_rundkast_runtime": 40.0836,
"eval_rundkast_samples_per_second": 33.405,
"eval_rundkast_steps_per_second": 1.048,
"eval_rundkast_wer": 0.1334986402175652,
"step": 64068
},
{
"epoch": 38.0,
"eval_nb_samtale_loss": 0.5725387334823608,
"eval_nb_samtale_runtime": 48.2674,
"eval_nb_samtale_samples_per_second": 11.043,
"eval_nb_samtale_steps_per_second": 0.352,
"eval_nb_samtale_wer": 0.19156901041666666,
"step": 64068
},
{
"epoch": 39.0,
"grad_norm": 9.250320434570312,
"learning_rate": 2.5978161965423114e-06,
"loss": 0.203,
"step": 65754
},
{
"epoch": 39.0,
"eval_bigbrother_loss": 1.6648077964782715,
"eval_bigbrother_runtime": 55.4612,
"eval_bigbrother_samples_per_second": 24.702,
"eval_bigbrother_steps_per_second": 0.775,
"eval_bigbrother_wer": 0.40812051881092404,
"step": 65754
},
{
"epoch": 39.0,
"eval_BB_NB_RUND_loss": 0.9731335639953613,
"eval_BB_NB_RUND_runtime": 187.078,
"eval_BB_NB_RUND_samples_per_second": 17.33,
"eval_BB_NB_RUND_steps_per_second": 0.545,
"eval_BB_NB_RUND_wer": 0.2430986748380507,
"step": 65754
},
{
"epoch": 39.0,
"eval_rundkast_loss": 0.41892868280410767,
"eval_rundkast_runtime": 39.6687,
"eval_rundkast_samples_per_second": 33.755,
"eval_rundkast_steps_per_second": 1.059,
"eval_rundkast_wer": 0.13181890897456408,
"step": 65754
},
{
"epoch": 39.0,
"eval_nb_samtale_loss": 0.5878060460090637,
"eval_nb_samtale_runtime": 46.9935,
"eval_nb_samtale_samples_per_second": 11.342,
"eval_nb_samtale_steps_per_second": 0.362,
"eval_nb_samtale_wer": 0.189697265625,
"step": 65754
},
{
"epoch": 40.0,
"grad_norm": 5.386321067810059,
"learning_rate": 4.094631483166515e-08,
"loss": 0.1995,
"step": 67440
},
{
"epoch": 40.0,
"eval_bigbrother_loss": 1.6618098020553589,
"eval_bigbrother_runtime": 55.1914,
"eval_bigbrother_samples_per_second": 24.823,
"eval_bigbrother_steps_per_second": 0.779,
"eval_bigbrother_wer": 0.4102956577781358,
"step": 67440
},
{
"epoch": 40.0,
"eval_BB_NB_RUND_loss": 0.9701318740844727,
"eval_BB_NB_RUND_runtime": 185.1129,
"eval_BB_NB_RUND_samples_per_second": 17.514,
"eval_BB_NB_RUND_steps_per_second": 0.551,
"eval_BB_NB_RUND_wer": 0.24457705023788404,
"step": 67440
},
{
"epoch": 40.0,
"eval_rundkast_loss": 0.4156753122806549,
"eval_rundkast_runtime": 39.6972,
"eval_rundkast_samples_per_second": 33.73,
"eval_rundkast_steps_per_second": 1.058,
"eval_rundkast_wer": 0.13397856342985123,
"step": 67440
},
{
"epoch": 40.0,
"eval_nb_samtale_loss": 0.5852100849151611,
"eval_nb_samtale_runtime": 47.3346,
"eval_nb_samtale_samples_per_second": 11.26,
"eval_nb_samtale_steps_per_second": 0.359,
"eval_nb_samtale_wer": 0.18896484375,
"step": 67440
}
],
"logging_steps": 500,
"max_steps": 67440,
"num_input_tokens_seen": 0,
"num_train_epochs": 40,
"save_steps": 500,
"total_flos": 1.1424949557265292e+21,
"train_batch_size": 48,
"trial_name": null,
"trial_params": null
}