{
  "best_global_step": 5500,
  "best_metric": 0.9997473023444335,
  "best_model_checkpoint": "/root/neobert-astraea-finetune/checkpoints/astraea-modernbert-full/checkpoint-5500",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 5799,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0,
      "loss_anomaly": 1.458340048789978,
      "loss_cls": 0.14811289310455322,
      "loss_mitre": 6.762622833251953,
      "loss_ner": 3.302584648132324,
      "step": 0
    },
    {
      "epoch": 0,
      "loss_anomaly": 1.8789876699447632,
      "loss_cls": 0.18254724144935608,
      "loss_mitre": 6.818868637084961,
      "loss_ner": 3.489518880844116,
      "step": 0
    },
    {
      "epoch": 0.0129366106080207,
      "grad_norm": 66.23992919921875,
      "learning_rate": 8.275862068965518e-07,
      "loss": 14.6671240234375,
      "step": 25
    },
    {
      "epoch": 0.0129366106080207,
      "loss_anomaly": 0.8223457932472229,
      "loss_cls": 0.1256418377161026,
      "loss_mitre": 6.621975421905518,
      "loss_ner": 3.215240955352783,
      "step": 25
    },
    {
      "epoch": 0.0129366106080207,
      "loss_anomaly": 1.0311975479125977,
      "loss_cls": 0.1335603892803192,
      "loss_mitre": 6.510723114013672,
      "loss_ner": 3.088139533996582,
      "step": 25
    },
    {
      "epoch": 0.0258732212160414,
      "grad_norm": 49.2056770324707,
      "learning_rate": 1.6896551724137933e-06,
      "loss": 12.811724853515624,
      "step": 50
    },
    {
      "epoch": 0.0258732212160414,
      "loss_anomaly": 0.32382556796073914,
      "loss_cls": 0.0800798311829567,
      "loss_mitre": 6.717785358428955,
      "loss_ner": 2.325465679168701,
      "step": 50
    },
    {
      "epoch": 0.0258732212160414,
      "loss_anomaly": 0.3541789650917053,
      "loss_cls": 0.07647467404603958,
      "loss_mitre": 7.045587062835693,
      "loss_ner": 2.242394208908081,
      "step": 50
    },
    {
      "epoch": 0.03880983182406209,
      "grad_norm": 19.00293731689453,
      "learning_rate": 2.5517241379310347e-06,
      "loss": 9.851192626953125,
      "step": 75
    },
    {
      "epoch": 0.03880983182406209,
      "loss_anomaly": 0.19930747151374817,
      "loss_cls": 0.04812509939074516,
      "loss_mitre": 5.816312313079834,
      "loss_ner": 0.9292773008346558,
      "step": 75
    },
    {
      "epoch": 0.03880983182406209,
      "loss_anomaly": 0.21306274831295013,
      "loss_cls": 0.0654156431555748,
      "loss_mitre": 6.500724792480469,
      "loss_ner": 0.9411810636520386,
      "step": 75
    },
    {
      "epoch": 0.0517464424320828,
      "grad_norm": 22.42531394958496,
      "learning_rate": 3.4137931034482764e-06,
      "loss": 7.84264892578125,
      "step": 100
    },
    {
      "epoch": 0.0517464424320828,
      "loss_anomaly": 0.234215646982193,
      "loss_cls": 0.027563823387026787,
      "loss_mitre": 5.189043045043945,
      "loss_ner": 0.8922914266586304,
      "step": 100
    },
    {
      "epoch": 0.0517464424320828,
      "loss_anomaly": 0.23188424110412598,
      "loss_cls": 0.022451482713222504,
      "loss_mitre": 5.847111225128174,
      "loss_ner": 0.8459073901176453,
      "step": 100
    },
    {
      "epoch": 0.0646830530401035,
      "grad_norm": 29.84002113342285,
      "learning_rate": 4.275862068965518e-06,
      "loss": 6.575426635742187,
      "step": 125
    },
    {
      "epoch": 0.0646830530401035,
      "loss_anomaly": 0.19701075553894043,
      "loss_cls": 0.030430851504206657,
      "loss_mitre": 4.0482940673828125,
      "loss_ner": 0.8195008635520935,
      "step": 125
    },
    {
      "epoch": 0.0646830530401035,
      "loss_anomaly": 0.2301376312971115,
      "loss_cls": 0.031123578548431396,
      "loss_mitre": 4.414292335510254,
      "loss_ner": 0.800862193107605,
      "step": 125
    },
    {
      "epoch": 0.07761966364812418,
      "grad_norm": 60.14553451538086,
      "learning_rate": 5.137931034482759e-06,
      "loss": 5.26430908203125,
      "step": 150
    },
    {
      "epoch": 0.07761966364812418,
      "loss_anomaly": 0.27928754687309265,
      "loss_cls": 0.014277160167694092,
      "loss_mitre": 2.4291622638702393,
      "loss_ner": 0.5783259868621826,
      "step": 150
    },
    {
      "epoch": 0.07761966364812418,
      "loss_anomaly": 0.39591261744499207,
      "loss_cls": 0.032565489411354065,
      "loss_mitre": 4.124593734741211,
      "loss_ner": 0.7729849815368652,
      "step": 150
    },
    {
      "epoch": 0.09055627425614489,
      "grad_norm": 32.38335418701172,
      "learning_rate": 6e-06,
      "loss": 4.355409851074219,
      "step": 175
    },
    {
      "epoch": 0.09055627425614489,
      "loss_anomaly": 0.18196144700050354,
      "loss_cls": 0.014510966837406158,
      "loss_mitre": 3.329522132873535,
      "loss_ner": 0.5992636680603027,
      "step": 175
    },
    {
      "epoch": 0.09055627425614489,
      "loss_anomaly": 0.18131959438323975,
      "loss_cls": 0.0041444613598287106,
      "loss_mitre": 2.4723010063171387,
      "loss_ner": 0.6262525916099548,
      "step": 175
    },
    {
      "epoch": 0.1034928848641656,
      "grad_norm": 396.8391418457031,
      "learning_rate": 6.862068965517242e-06,
      "loss": 3.3947799682617186,
      "step": 200
    },
    {
      "epoch": 0.1034928848641656,
      "loss_anomaly": 0.2402094602584839,
      "loss_cls": 0.010139903984963894,
      "loss_mitre": 1.4130696058273315,
      "loss_ner": 0.568060576915741,
      "step": 200
    },
    {
      "epoch": 0.1034928848641656,
      "loss_anomaly": 0.23881895840168,
      "loss_cls": 0.026661871001124382,
      "loss_mitre": 1.325908899307251,
      "loss_ner": 0.6285443305969238,
      "step": 200
    },
    {
      "epoch": 0.11642949547218628,
      "grad_norm": 169.10366821289062,
      "learning_rate": 7.724137931034483e-06,
      "loss": 2.2958746337890625,
      "step": 225
    },
    {
      "epoch": 0.11642949547218628,
      "loss_anomaly": 0.15743744373321533,
      "loss_cls": 0.004800485912710428,
      "loss_mitre": 0.31566348671913147,
      "loss_ner": 0.4057095944881439,
      "step": 225
    },
    {
      "epoch": 0.11642949547218628,
      "loss_anomaly": 0.20836271345615387,
      "loss_cls": 0.017825651913881302,
      "loss_mitre": 0.4337477684020996,
      "loss_ner": 0.48290449380874634,
      "step": 225
    },
    {
      "epoch": 0.129366106080207,
      "grad_norm": 27.531404495239258,
      "learning_rate": 8.586206896551726e-06,
      "loss": 2.0377734375,
      "step": 250
    },
    {
      "epoch": 0.129366106080207,
      "loss_anomaly": 0.19601386785507202,
      "loss_cls": 0.006822212599217892,
      "loss_mitre": 0.6947734355926514,
      "loss_ner": 0.4846152067184448,
      "step": 250
    },
    {
      "epoch": 0.129366106080207,
      "loss_anomaly": 0.1587674915790558,
      "loss_cls": 0.016658475622534752,
      "loss_mitre": 0.528306782245636,
      "loss_ner": 0.3810734450817108,
      "step": 250
    },
    {
      "epoch": 0.1423027166882277,
      "grad_norm": 17.436723709106445,
      "learning_rate": 9.448275862068967e-06,
      "loss": 1.4832650756835937,
      "step": 275
    },
    {
      "epoch": 0.1423027166882277,
      "loss_anomaly": 0.16191893815994263,
      "loss_cls": 0.008145017549395561,
      "loss_mitre": 0.2178678959608078,
      "loss_ner": 0.32947811484336853,
      "step": 275
    },
    {
      "epoch": 0.1423027166882277,
      "loss_anomaly": 0.22270818054676056,
      "loss_cls": 0.024416742846369743,
      "loss_mitre": 0.2518240511417389,
      "loss_ner": 0.2577015161514282,
      "step": 275
    },
    {
      "epoch": 0.15523932729624837,
      "grad_norm": 21.32984161376953,
      "learning_rate": 1.0310344827586208e-05,
      "loss": 1.3046697998046874,
      "step": 300
    },
    {
      "epoch": 0.15523932729624837,
      "loss_anomaly": 0.3059636950492859,
      "loss_cls": 0.011652686633169651,
      "loss_mitre": 0.29291433095932007,
      "loss_ner": 0.2802308201789856,
      "step": 300
    },
    {
      "epoch": 0.15523932729624837,
      "loss_anomaly": 0.20825400948524475,
      "loss_cls": 0.014717529527842999,
      "loss_mitre": 0.12829162180423737,
      "loss_ner": 0.2864997088909149,
      "step": 300
    },
    {
      "epoch": 0.16817593790426907,
      "grad_norm": 9.237492561340332,
      "learning_rate": 1.117241379310345e-05,
      "loss": 1.1254735565185547,
      "step": 325
    },
    {
      "epoch": 0.16817593790426907,
      "loss_anomaly": 0.2493964433670044,
      "loss_cls": 0.005455702543258667,
      "loss_mitre": 0.17017656564712524,
      "loss_ner": 0.25479021668434143,
      "step": 325
    },
    {
      "epoch": 0.16817593790426907,
      "loss_anomaly": 0.1609182059764862,
      "loss_cls": 0.0127034280449152,
      "loss_mitre": 0.11223185062408447,
      "loss_ner": 0.2127515822649002,
      "step": 325
    },
    {
      "epoch": 0.18111254851228978,
      "grad_norm": 14.198319435119629,
      "learning_rate": 1.203448275862069e-05,
      "loss": 0.932662353515625,
      "step": 350
    },
    {
      "epoch": 0.18111254851228978,
      "loss_anomaly": 0.23144802451133728,
      "loss_cls": 0.02146485634148121,
      "loss_mitre": 0.5649169087409973,
      "loss_ner": 0.2143765538930893,
      "step": 350
    },
    {
      "epoch": 0.18111254851228978,
      "loss_anomaly": 0.11280352622270584,
      "loss_cls": 0.0037572032306343317,
      "loss_mitre": 0.027937158942222595,
      "loss_ner": 0.2252315729856491,
      "step": 350
    },
    {
      "epoch": 0.19404915912031048,
      "grad_norm": 14.724109649658203,
      "learning_rate": 1.2896551724137932e-05,
      "loss": 0.79138671875,
      "step": 375
    },
    {
      "epoch": 0.19404915912031048,
      "loss_anomaly": 0.15610040724277496,
      "loss_cls": 0.0014543408760800958,
      "loss_mitre": 0.049078959971666336,
      "loss_ner": 0.20558440685272217,
      "step": 375
    },
    {
      "epoch": 0.19404915912031048,
      "loss_anomaly": 0.12666428089141846,
      "loss_cls": 0.00126453151460737,
      "loss_mitre": 0.05539293587207794,
      "loss_ner": 0.21833527088165283,
      "step": 375
    },
    {
      "epoch": 0.2069857697283312,
      "grad_norm": 12.281414985656738,
      "learning_rate": 1.3758620689655173e-05,
      "loss": 0.7504615783691406,
      "step": 400
    },
    {
      "epoch": 0.2069857697283312,
      "loss_anomaly": 0.14674343168735504,
      "loss_cls": 0.01188975851982832,
      "loss_mitre": 0.06474322080612183,
      "loss_ner": 0.1434691846370697,
      "step": 400
    },
    {
      "epoch": 0.2069857697283312,
      "loss_anomaly": 0.17108482122421265,
      "loss_cls": 0.012205853126943111,
      "loss_mitre": 0.6510561108589172,
      "loss_ner": 0.14611703157424927,
      "step": 400
    },
    {
      "epoch": 0.21992238033635186,
      "grad_norm": 7.8789381980896,
      "learning_rate": 1.4620689655172416e-05,
      "loss": 0.5611345672607422,
      "step": 425
    },
    {
      "epoch": 0.21992238033635186,
      "loss_anomaly": 0.11254362761974335,
      "loss_cls": 0.014451020397245884,
      "loss_mitre": 0.022060807794332504,
      "loss_ner": 0.11996780335903168,
      "step": 425
    },
    {
      "epoch": 0.21992238033635186,
      "loss_anomaly": 0.14694508910179138,
      "loss_cls": 0.002252220641821623,
      "loss_mitre": 0.021973488852381706,
      "loss_ner": 0.10807537287473679,
      "step": 425
    },
    {
      "epoch": 0.23285899094437257,
      "grad_norm": 5.659252166748047,
      "learning_rate": 1.548275862068966e-05,
      "loss": 0.5522967147827148,
      "step": 450
    },
    {
      "epoch": 0.23285899094437257,
      "loss_anomaly": 0.10406625270843506,
      "loss_cls": 0.014451662078499794,
      "loss_mitre": 0.018083803355693817,
      "loss_ner": 0.12196429818868637,
      "step": 450
    },
    {
      "epoch": 0.23285899094437257,
      "loss_anomaly": 0.11139430105686188,
      "loss_cls": 0.005909971427172422,
      "loss_mitre": 0.03319530561566353,
      "loss_ner": 0.1200285330414772,
      "step": 450
    },
    {
      "epoch": 0.24579560155239327,
      "grad_norm": 9.285419464111328,
      "learning_rate": 1.63448275862069e-05,
      "loss": 0.4184379959106445,
      "step": 475
    },
    {
      "epoch": 0.24579560155239327,
      "loss_anomaly": 0.10336190462112427,
      "loss_cls": 0.009995656087994576,
      "loss_mitre": 0.020997054874897003,
      "loss_ner": 0.0665409043431282,
      "step": 475
    },
    {
      "epoch": 0.24579560155239327,
      "loss_anomaly": 0.1371651589870453,
      "loss_cls": 0.0009332228801213205,
      "loss_mitre": 0.04225302115082741,
      "loss_ner": 0.08025738596916199,
      "step": 475
    },
    {
      "epoch": 0.258732212160414,
      "grad_norm": 6.152985572814941,
      "learning_rate": 1.7206896551724138e-05,
      "loss": 0.46059120178222657,
      "step": 500
    },
    {
      "epoch": 0.258732212160414,
      "eval_anomaly_mae": 0.1486216638849716,
      "eval_anomaly_mse": 0.04347467073309788,
      "eval_anomaly_n_eval": 27475.0,
      "eval_cls_accuracy": 0.97788836735848,
      "eval_cls_auprc": 0.9867153514489324,
      "eval_cls_f1": 0.9254243753576197,
      "eval_cls_n_eval": 17683.0,
      "eval_cls_precision": 0.8761285662694114,
      "eval_cls_recall": 0.9805982215036378,
      "eval_loss": 0.10236462950706482,
      "eval_mitre_accuracy": 0.962847492855287,
      "eval_mitre_macro_f1": 0.15452932938486336,
      "eval_mitre_n_eval": 3849.0,
      "eval_ner_n_tokens": 666658.0,
      "eval_ner_token_acc": 0.9774577069501903,
      "eval_runtime": 68.5772,
      "eval_samples_per_second": 400.643,
      "eval_steps_per_second": 3.135,
      "step": 500
    },
    {
      "epoch": 0.258732212160414,
      "loss_anomaly": 0.15775607526302338,
      "loss_cls": 0.0035615349188447,
      "loss_mitre": 0.7043123841285706,
      "loss_ner": 0.06773389130830765,
      "step": 500
    },
    {
      "epoch": 0.258732212160414,
      "loss_anomaly": 0.14393946528434753,
      "loss_cls": 0.012784991413354874,
      "loss_mitre": 0.5881390571594238,
      "loss_ner": 0.06087030842900276,
      "step": 500
    },
    {
      "epoch": 0.2716688227684347,
      "grad_norm": 7.114372730255127,
      "learning_rate": 1.806896551724138e-05,
      "loss": 0.4251505661010742,
      "step": 525
    },
    {
      "epoch": 0.2716688227684347,
      "loss_anomaly": 0.10564332455396652,
      "loss_cls": 0.0044641331769526005,
      "loss_mitre": 0.9402504563331604,
      "loss_ner": 0.04252464696764946,
      "step": 525
    },
    {
      "epoch": 0.2716688227684347,
      "loss_anomaly": 0.12212535738945007,
      "loss_cls": 0.00028572589508257806,
      "loss_mitre": 0.02400970831513405,
      "loss_ner": 0.10424104332923889,
      "step": 525
    },
    {
      "epoch": 0.2846054333764554,
      "grad_norm": 5.844174861907959,
      "learning_rate": 1.893103448275862e-05,
      "loss": 0.47771366119384767,
      "step": 550
    },
    {
      "epoch": 0.2846054333764554,
      "loss_anomaly": 0.10037468373775482,
      "loss_cls": 0.011210967786610126,
      "loss_mitre": 0.010874348692595959,
      "loss_ner": 0.06607252359390259,
      "step": 550
    },
    {
      "epoch": 0.2846054333764554,
      "loss_anomaly": 0.10040736198425293,
      "loss_cls": 0.0018024571472778916,
      "loss_mitre": 0.4879980683326721,
      "loss_ner": 0.03655198588967323,
      "step": 550
    },
    {
      "epoch": 0.2975420439844761,
      "grad_norm": 6.809467315673828,
      "learning_rate": 1.9793103448275863e-05,
      "loss": 0.5264387130737305,
      "step": 575
    },
    {
      "epoch": 0.2975420439844761,
      "loss_anomaly": 0.1492130160331726,
      "loss_cls": 0.003011732827872038,
      "loss_mitre": 0.5947777032852173,
      "loss_ner": 0.07491794973611832,
      "step": 575
    },
    {
      "epoch": 0.2975420439844761,
      "loss_anomaly": 0.0969763770699501,
      "loss_cls": 0.011376527138054371,
      "loss_mitre": 0.006226290017366409,
      "loss_ner": 0.05118679255247116,
      "step": 575
    },
    {
      "epoch": 0.31047865459249674,
      "grad_norm": 4.727795124053955,
      "learning_rate": 1.9927189116689022e-05,
      "loss": 0.5454753112792968,
      "step": 600
    },
    {
      "epoch": 0.31047865459249674,
      "loss_anomaly": 0.09024052321910858,
      "loss_cls": 0.0030815002974122763,
      "loss_mitre": 1.0119731426239014,
      "loss_ner": 0.11578906327486038,
      "step": 600
    },
    {
      "epoch": 0.31047865459249674,
      "loss_anomaly": 0.0651533454656601,
      "loss_cls": 0.0025258660316467285,
      "loss_mitre": 0.008180954493582249,
      "loss_ner": 0.043203361332416534,
      "step": 600
    },
    {
      "epoch": 0.32341526520051744,
      "grad_norm": 5.662601470947266,
      "learning_rate": 1.983138532285879e-05,
      "loss": 0.4436736297607422,
      "step": 625
    },
    {
      "epoch": 0.32341526520051744,
      "loss_anomaly": 0.11418907344341278,
      "loss_cls": 0.0064962925389409065,
      "loss_mitre": 0.008304150775074959,
      "loss_ner": 0.052326224744319916,
      "step": 625
    },
    {
      "epoch": 0.32341526520051744,
      "loss_anomaly": 0.05799151584506035,
      "loss_cls": 0.0015510590746998787,
      "loss_mitre": 0.011419427581131458,
      "loss_ner": 0.044967301189899445,
      "step": 625
    },
    {
      "epoch": 0.33635187580853815,
      "grad_norm": 3.9786980152130127,
      "learning_rate": 1.973558152902855e-05,
      "loss": 0.37106273651123045,
      "step": 650
    },
    {
      "epoch": 0.33635187580853815,
      "loss_anomaly": 0.06457138061523438,
      "loss_cls": 0.006690428592264652,
      "loss_mitre": 0.6185184717178345,
      "loss_ner": 0.04909238591790199,
      "step": 650
    },
    {
      "epoch": 0.33635187580853815,
      "loss_anomaly": 0.08697918057441711,
      "loss_cls": 0.005499826278537512,
      "loss_mitre": 0.008673307485878468,
      "loss_ner": 0.019459327682852745,
      "step": 650
    },
    {
      "epoch": 0.34928848641655885,
      "grad_norm": 4.723515033721924,
      "learning_rate": 1.9639777735198315e-05,
      "loss": 0.3766423034667969,
      "step": 675
    },
    {
      "epoch": 0.34928848641655885,
      "loss_anomaly": 0.07555042207241058,
      "loss_cls": 0.005717230029404163,
      "loss_mitre": 0.008249030448496342,
      "loss_ner": 0.0581187829375267,
      "step": 675
    },
    {
      "epoch": 0.34928848641655885,
      "loss_anomaly": 0.07105275243520737,
      "loss_cls": 0.0011884495615959167,
      "loss_mitre": 0.6000533699989319,
      "loss_ner": 0.04121296852827072,
      "step": 675
    },
    {
      "epoch": 0.36222509702457956,
      "grad_norm": 3.1347367763519287,
      "learning_rate": 1.9543973941368078e-05,
      "loss": 0.38748218536376955,
      "step": 700
    },
    {
      "epoch": 0.36222509702457956,
      "loss_anomaly": 0.06983435153961182,
      "loss_cls": 0.002311189891770482,
      "loss_mitre": 0.0036953790113329887,
      "loss_ner": 0.03168926388025284,
      "step": 700
    },
    {
      "epoch": 0.36222509702457956,
      "loss_anomaly": 0.0773167535662651,
      "loss_cls": 0.00393596850335598,
      "loss_mitre": 0.5264779925346375,
      "loss_ner": 0.04743402823805809,
      "step": 700
    },
    {
      "epoch": 0.37516170763260026,
      "grad_norm": 6.316786289215088,
      "learning_rate": 1.9448170147537845e-05,
      "loss": 0.30317699432373046,
      "step": 725
    },
    {
      "epoch": 0.37516170763260026,
      "loss_anomaly": 0.06988630443811417,
      "loss_cls": 0.011228875257074833,
      "loss_mitre": 0.7557423114776611,
      "loss_ner": 0.021567991003394127,
      "step": 725
    },
    {
      "epoch": 0.37516170763260026,
      "loss_anomaly": 0.04522155225276947,
      "loss_cls": 0.001466374029405415,
      "loss_mitre": 0.005125878378748894,
      "loss_ner": 0.017750408500432968,
      "step": 725
    },
    {
      "epoch": 0.38809831824062097,
      "grad_norm": 3.2551321983337402,
      "learning_rate": 1.9352366353707608e-05,
      "loss": 0.2562189483642578,
      "step": 750
    },
    {
      "epoch": 0.38809831824062097,
      "loss_anomaly": 0.1082291305065155,
      "loss_cls": 0.0279394518584013,
      "loss_mitre": 0.005703972652554512,
      "loss_ner": 0.03317469730973244,
      "step": 750
    },
    {
      "epoch": 0.38809831824062097,
      "loss_anomaly": 0.05690450221300125,
      "loss_cls": 0.0002418720832793042,
      "loss_mitre": 0.0040161036886274815,
      "loss_ner": 0.03340652957558632,
      "step": 750
    },
    {
      "epoch": 0.40103492884864167,
      "grad_norm": 3.604860544204712,
      "learning_rate": 1.925656255987737e-05,
      "loss": 0.29800813674926757,
      "step": 775
    },
    {
      "epoch": 0.40103492884864167,
      "loss_anomaly": 0.06834549456834793,
      "loss_cls": 0.000718119612429291,
      "loss_mitre": 0.0031323879957199097,
      "loss_ner": 0.06853349506855011,
      "step": 775
    },
    {
      "epoch": 0.40103492884864167,
      "loss_anomaly": 0.09035581350326538,
      "loss_cls": 0.0019903257489204407,
      "loss_mitre": 0.00418852549046278,
      "loss_ner": 0.025127923116087914,
      "step": 775
    },
    {
      "epoch": 0.4139715394566624,
      "grad_norm": 4.539907455444336,
      "learning_rate": 1.9160758766047138e-05,
      "loss": 0.3676939010620117,
      "step": 800
    },
    {
      "epoch": 0.4139715394566624,
      "loss_anomaly": 0.11512207239866257,
      "loss_cls": 0.0010495882015675306,
      "loss_mitre": 0.7834214568138123,
      "loss_ner": 0.015636181458830833,
      "step": 800
    },
    {
      "epoch": 0.4139715394566624,
      "loss_anomaly": 0.0721721276640892,
      "loss_cls": 0.0013183070113882422,
      "loss_mitre": 0.003540542908012867,
      "loss_ner": 0.01643516682088375,
      "step": 800
    },
    {
      "epoch": 0.4269081500646831,
      "grad_norm": 3.5921435356140137,
      "learning_rate": 1.90649549722169e-05,
      "loss": 0.3933557510375977,
      "step": 825
    },
    {
      "epoch": 0.4269081500646831,
      "loss_anomaly": 0.06489208340644836,
      "loss_cls": 0.010899069719016552,
      "loss_mitre": 0.0028563092928379774,
      "loss_ner": 0.024311313405632973,
      "step": 825
    },
    {
      "epoch": 0.4269081500646831,
      "loss_anomaly": 0.039788976311683655,
      "loss_cls": 0.002522006630897522,
      "loss_mitre": 0.002886980539187789,
      "loss_ner": 0.01573309116065502,
      "step": 825
    },
    {
      "epoch": 0.4398447606727037,
      "grad_norm": 2.6121408939361572,
      "learning_rate": 1.8969151178386667e-05,
      "loss": 0.34168865203857424,
      "step": 850
    },
    {
      "epoch": 0.4398447606727037,
      "loss_anomaly": 0.08011620491743088,
      "loss_cls": 0.0002537903201300651,
      "loss_mitre": 0.00223373482003808,
      "loss_ner": 0.03189283609390259,
      "step": 850
    },
    {
      "epoch": 0.4398447606727037,
      "loss_anomaly": 0.07320950925350189,
      "loss_cls": 0.004216733854264021,
      "loss_mitre": 0.0046006618067622185,
      "loss_ner": 0.022002464160323143,
      "step": 850
    },
    {
      "epoch": 0.45278137128072443,
      "grad_norm": 2.4920225143432617,
      "learning_rate": 1.887334738455643e-05,
      "loss": 0.425952262878418,
      "step": 875
    },
    {
      "epoch": 0.45278137128072443,
      "loss_anomaly": 0.04904380068182945,
      "loss_cls": 0.0013018827885389328,
      "loss_mitre": 0.002484935102984309,
      "loss_ner": 0.024313218891620636,
      "step": 875
    },
    {
      "epoch": 0.45278137128072443,
      "loss_anomaly": 0.05911850929260254,
      "loss_cls": 0.008243133313953876,
      "loss_mitre": 0.007032583002001047,
      "loss_ner": 0.032650746405124664,
      "step": 875
    },
    {
      "epoch": 0.46571798188874514,
      "grad_norm": 1.3299481868743896,
      "learning_rate": 1.8777543590726194e-05,
      "loss": 0.3497086334228516,
      "step": 900
    },
    {
      "epoch": 0.46571798188874514,
      "loss_anomaly": 0.06609110534191132,
      "loss_cls": 0.0005804927204735577,
      "loss_mitre": 0.4866850674152374,
      "loss_ner": 0.023745549842715263,
      "step": 900
    },
    {
      "epoch": 0.46571798188874514,
      "loss_anomaly": 0.05200420320034027,
      "loss_cls": 0.0015868014888837934,
      "loss_mitre": 0.002841484732925892,
      "loss_ner": 0.016115477308630943,
      "step": 900
    },
    {
      "epoch": 0.47865459249676584,
      "grad_norm": 2.1148135662078857,
      "learning_rate": 1.868173979689596e-05,
      "loss": 0.3434376907348633,
      "step": 925
    },
    {
      "epoch": 0.47865459249676584,
      "loss_anomaly": 0.052690114825963974,
      "loss_cls": 0.00013846400543116033,
      "loss_mitre": 1.2001196146011353,
      "loss_ner": 0.025999944657087326,
      "step": 925
    },
    {
      "epoch": 0.47865459249676584,
      "loss_anomaly": 0.07163169980049133,
      "loss_cls": 0.0010361387394368649,
      "loss_mitre": 0.46292874217033386,
      "loss_ner": 0.014140576124191284,
      "step": 925
    },
    {
      "epoch": 0.49159120310478654,
      "grad_norm": 1.5439307689666748,
      "learning_rate": 1.8585936003065723e-05,
      "loss": 0.33875518798828125,
      "step": 950
    },
    {
      "epoch": 0.49159120310478654,
      "loss_anomaly": 0.05600789189338684,
      "loss_cls": 0.009089753031730652,
      "loss_mitre": 1.6100499629974365,
      "loss_ner": 0.02818966470658779,
      "step": 950
    },
    {
      "epoch": 0.49159120310478654,
      "loss_anomaly": 0.04825662449002266,
      "loss_cls": 0.00019881287880707532,
      "loss_mitre": 0.0027828284073621035,
      "loss_ner": 0.019804565235972404,
      "step": 950
    },
    {
      "epoch": 0.5045278137128072,
      "grad_norm": 4.597489833831787,
      "learning_rate": 1.8490132209235487e-05,
      "loss": 0.38815601348876955,
      "step": 975
    },
    {
      "epoch": 0.5045278137128072,
      "loss_anomaly": 0.04038494825363159,
      "loss_cls": 0.0012326519936323166,
      "loss_mitre": 0.001278909738175571,
      "loss_ner": 0.06657722592353821,
      "step": 975
    },
    {
      "epoch": 0.5045278137128072,
      "loss_anomaly": 0.10056827962398529,
      "loss_cls": 0.0014719143509864807,
      "loss_mitre": 0.8068041205406189,
      "loss_ner": 0.023483285680413246,
      "step": 975
    },
    {
      "epoch": 0.517464424320828,
      "grad_norm": 1.45570969581604,
      "learning_rate": 1.839432841540525e-05,
      "loss": 0.2720743179321289,
      "step": 1000
    },
    {
      "epoch": 0.517464424320828,
      "eval_anomaly_mae": 0.10720256794799539,
      "eval_anomaly_mse": 0.02014333687925468,
      "eval_anomaly_n_eval": 27475.0,
      "eval_cls_accuracy": 0.9941751965164282,
      "eval_cls_auprc": 0.9978392823857588,
      "eval_cls_f1": 0.9790095781536581,
      "eval_cls_n_eval": 17683.0,
      "eval_cls_precision": 0.987258528565557,
      "eval_cls_recall": 0.9708973322554567,
      "eval_loss": 0.06316110491752625,
      "eval_mitre_accuracy": 0.962847492855287,
      "eval_mitre_macro_f1": 0.15516269057417065,
      "eval_mitre_n_eval": 3849.0,
      "eval_ner_n_tokens": 666658.0,
      "eval_ner_token_acc": 0.9929859088168146,
      "eval_runtime": 68.9143,
      "eval_samples_per_second": 398.684,
      "eval_steps_per_second": 3.12,
      "step": 1000
    },
    {
      "epoch": 0.517464424320828,
      "loss_anomaly": 0.033122383058071136,
      "loss_cls": 0.00033859116956591606,
      "loss_mitre": 0.0016783454921096563,
      "loss_ner": 0.003214341588318348,
      "step": 1000
    },
    {
      "epoch": 0.517464424320828,
      "loss_anomaly": 0.05769438296556473,
      "loss_cls": 0.001896923640742898,
      "loss_mitre": 0.0014118148246780038,
      "loss_ner": 0.023480521515011787,
      "step": 1000
    },
    {
      "epoch": 0.5304010349288486,
      "grad_norm": 1.7671699523925781,
      "learning_rate": 1.8298524621575016e-05,
      "loss": 0.2527183532714844,
      "step": 1025
    },
    {
      "epoch": 0.5304010349288486,
      "loss_anomaly": 0.044356346130371094,
      "loss_cls": 0.021346019580960274,
      "loss_mitre": 0.0019673698116093874,
      "loss_ner": 0.042252860963344574,
      "step": 1025
    },
    {
      "epoch": 0.5304010349288486,
      "loss_anomaly": 0.05318473279476166,
      "loss_cls": 0.0010298237903043628,
      "loss_mitre": 0.001789235626347363,
      "loss_ner": 0.02958856150507927,
      "step": 1025
    },
    {
      "epoch": 0.5433376455368694,
      "grad_norm": 1.2865092754364014,
      "learning_rate": 1.820272082774478e-05,
      "loss": 0.1818370056152344,
      "step": 1050
    },
    {
      "epoch": 0.5433376455368694,
      "loss_anomaly": 0.055549293756484985,
      "loss_cls": 0.0010960310464724898,
      "loss_mitre": 0.0018367705633863807,
      "loss_ner": 0.018995175138115883,
      "step": 1050
    },
    {
      "epoch": 0.5433376455368694,
      "loss_anomaly": 0.036615341901779175,
      "loss_cls": 0.002259816974401474,
      "loss_mitre": 0.0017252865945920348,
      "loss_ner": 0.0039041393902152777,
      "step": 1050
    },
    {
      "epoch": 0.55627425614489,
      "grad_norm": 2.4620025157928467,
      "learning_rate": 1.8106917033914543e-05,
      "loss": 0.19637855529785156,
      "step": 1075
    },
    {
      "epoch": 0.55627425614489,
      "loss_anomaly": 0.04112130403518677,
      "loss_cls": 0.0035661556757986546,
      "loss_mitre": 0.0013310230569913983,
      "loss_ner": 0.035554736852645874,
      "step": 1075
    },
    {
      "epoch": 0.55627425614489,
      "loss_anomaly": 0.05373847857117653,
      "loss_cls": 0.008511231280863285,
      "loss_mitre": 0.0012156771263107657,
      "loss_ner": 0.02155883051455021,
      "step": 1075
    },
    {
      "epoch": 0.5692108667529108,
      "grad_norm": 1.8406708240509033,
      "learning_rate": 1.801111324008431e-05,
      "loss": 0.2494220733642578,
      "step": 1100
    },
    {
      "epoch": 0.5692108667529108,
      "loss_anomaly": 0.0265185609459877,
      "loss_cls": 0.001267327694222331,
      "loss_mitre": 1.0360641479492188,
      "loss_ner": 0.01629761792719364,
      "step": 1100
    },
    {
      "epoch": 0.5692108667529108,
      "loss_anomaly": 0.03336966782808304,
      "loss_cls": 0.0031769261695444584,
      "loss_mitre": 0.004188026767224073,
      "loss_ner": 0.16942231357097626,
      "step": 1100
    },
    {
      "epoch": 0.5821474773609314,
      "grad_norm": 3.99717378616333,
      "learning_rate": 1.7915309446254073e-05,
      "loss": 0.3627820205688477,
      "step": 1125
    },
    {
      "epoch": 0.5821474773609314,
      "loss_anomaly": 0.040600311011075974,
      "loss_cls": 0.0026269599329680204,
      "loss_mitre": 0.0045071737840771675,
      "loss_ner": 0.022756541147828102,
      "step": 1125
    },
    {
      "epoch": 0.5821474773609314,
      "loss_anomaly": 0.028398018330335617,
      "loss_cls": 0.0027763599064201117,
      "loss_mitre": 0.00328185991384089,
      "loss_ner": 0.032867737114429474,
      "step": 1125
    },
    {
      "epoch": 0.5950840879689522,
      "grad_norm": 3.4986274242401123,
      "learning_rate": 1.7819505652423836e-05,
      "loss": 0.3926167678833008,
      "step": 1150
    },
    {
      "epoch": 0.5950840879689522,
      "loss_anomaly": 0.032711438834667206,
      "loss_cls": 0.0003663312236312777,
      "loss_mitre": 0.0014742615167051554,
      "loss_ner": 0.013649587519466877,
      "step": 1150
    },
    {
      "epoch": 0.5950840879689522,
      "loss_anomaly": 0.044117871671915054,
      "loss_cls": 0.0004688550834544003,
      "loss_mitre": 0.7916901111602783,
      "loss_ner": 0.016949232667684555,
      "step": 1150
    },
    {
      "epoch": 0.6080206985769728,
      "grad_norm": 1.9495267868041992,
      "learning_rate": 1.7723701858593602e-05,
      "loss": 0.3362046432495117,
      "step": 1175
    },
    {
      "epoch": 0.6080206985769728,
      "loss_anomaly": 0.046557825058698654,
      "loss_cls": 0.0009046901832334697,
      "loss_mitre": 0.003380028996616602,
      "loss_ner": 0.01866360753774643,
      "step": 1175
    },
    {
      "epoch": 0.6080206985769728,
      "loss_anomaly": 0.03562621772289276,
      "loss_cls": 0.0023968967143446207,
      "loss_mitre": 0.0014965090667828918,
      "loss_ner": 0.016764942556619644,
      "step": 1175
    },
    {
      "epoch": 0.6209573091849935,
      "grad_norm": 5.321065425872803,
      "learning_rate": 1.7627898064763366e-05,
      "loss": 0.29448991775512695,
      "step": 1200
    },
    {
      "epoch": 0.6209573091849935,
      "loss_anomaly": 0.03304995596408844,
      "loss_cls": 0.0006202247459441423,
      "loss_mitre": 0.001240433193743229,
      "loss_ner": 0.019781434908509254,
      "step": 1200
    },
    {
      "epoch": 0.6209573091849935,
      "loss_anomaly": 0.03305395692586899,
      "loss_cls": 0.0010951264994218946,
      "loss_mitre": 0.0014926099684089422,
      "loss_ner": 0.007148008793592453,
      "step": 1200
    },
    {
      "epoch": 0.6338939197930142,
      "grad_norm": 2.388376474380493,
      "learning_rate": 1.7532094270933132e-05,
      "loss": 0.30932207107543946,
      "step": 1225
    },
    {
      "epoch": 0.6338939197930142,
      "loss_anomaly": 0.024413706734776497,
      "loss_cls": 0.00047862829524092376,
      "loss_mitre": 0.0009499476873315871,
      "loss_ner": 0.019292833283543587,
      "step": 1225
    },
    {
      "epoch": 0.6338939197930142,
      "loss_anomaly": 0.04142192751169205,
      "loss_cls": 0.0007658992544747889,
      "loss_mitre": 0.5820366144180298,
      "loss_ner": 0.005492477677762508,
      "step": 1225
    },
    {
      "epoch": 0.6468305304010349,
      "grad_norm": 2.719923734664917,
      "learning_rate": 1.7436290477102895e-05,
      "loss": 0.4299563217163086,
      "step": 1250
    },
    {
      "epoch": 0.6468305304010349,
      "loss_anomaly": 0.04693226143717766,
      "loss_cls": 0.0006305835559032857,
      "loss_mitre": 0.7845224142074585,
      "loss_ner": 0.023018812760710716,
      "step": 1250
    },
    {
      "epoch": 0.6468305304010349,
      "loss_anomaly": 0.030941192060709,
      "loss_cls": 0.0006185569800436497,
      "loss_mitre": 0.6936325430870056,
      "loss_ner": 0.013628752902150154,
      "step": 1250
    },
    {
      "epoch": 0.6597671410090556,
      "grad_norm": 2.9341540336608887,
      "learning_rate": 1.734048668327266e-05,
      "loss": 0.2818842887878418,
      "step": 1275
    },
    {
      "epoch": 0.6597671410090556,
      "loss_anomaly": 0.03564213588833809,
      "loss_cls": 0.0004033461445942521,
      "loss_mitre": 0.0013078005285933614,
      "loss_ner": 0.018637005239725113,
      "step": 1275
    },
    {
      "epoch": 0.6597671410090556,
      "loss_anomaly": 0.049177657812833786,
      "loss_cls": 0.00011123003059765324,
      "loss_mitre": 0.002049276139587164,
      "loss_ner": 0.008561822585761547,
      "step": 1275
    },
    {
      "epoch": 0.6727037516170763,
      "grad_norm": 2.899069309234619,
      "learning_rate": 1.7244682889442422e-05,
      "loss": 0.3292347717285156,
      "step": 1300
    },
    {
      "epoch": 0.6727037516170763,
      "loss_anomaly": 0.03685444965958595,
      "loss_cls": 0.001188848284073174,
      "loss_mitre": 0.0014924099668860435,
      "loss_ner": 0.0045685116201639175,
      "step": 1300
    },
    {
      "epoch": 0.6727037516170763,
      "loss_anomaly": 0.03293314576148987,
      "loss_cls": 0.00833556056022644,
      "loss_mitre": 0.002210019389167428,
      "loss_ner": 0.004529324825853109,
      "step": 1300
    },
    {
      "epoch": 0.685640362225097,
      "grad_norm": 1.4815462827682495,
      "learning_rate": 1.714887909561219e-05,
      "loss": 0.3548107147216797,
      "step": 1325
    },
    {
      "epoch": 0.685640362225097,
      "loss_anomaly": 0.05004151165485382,
      "loss_cls": 0.00024574712733738124,
      "loss_mitre": 0.0016638465458527207,
      "loss_ner": 0.01460960227996111,
      "step": 1325
    },
    {
      "epoch": 0.685640362225097,
      "loss_anomaly": 0.03705885261297226,
      "loss_cls": 0.004278011154383421,
      "loss_mitre": 0.00241676764562726,
      "loss_ner": 0.009104462340474129,
      "step": 1325
    },
    {
      "epoch": 0.6985769728331177,
      "grad_norm": 1.8607110977172852,
      "learning_rate": 1.705307530178195e-05,
      "loss": 0.32282634735107424,
      "step": 1350
    },
    {
      "epoch": 0.6985769728331177,
      "loss_anomaly": 0.032848287373781204,
      "loss_cls": 0.0006282940157689154,
      "loss_mitre": 0.0014925220748409629,
      "loss_ner": 0.04962236061692238,
      "step": 1350
    },
    {
      "epoch": 0.6985769728331177,
      "loss_anomaly": 0.02458915486931801,
      "loss_cls": 0.0019540516659617424,
      "loss_mitre": 0.0010779064614325762,
      "loss_ner": 0.007461730390787125,
      "step": 1350
    },
    {
      "epoch": 0.7115135834411385,
      "grad_norm": 2.268500566482544,
      "learning_rate": 1.6957271507951715e-05,
      "loss": 0.2535930442810059,
      "step": 1375
    },
    {
      "epoch": 0.7115135834411385,
      "loss_anomaly": 0.02982262708246708,
      "loss_cls": 0.0001926450786413625,
      "loss_mitre": 0.4734683632850647,
      "loss_ner": 0.023081941530108452,
      "step": 1375
    },
    {
      "epoch": 0.7115135834411385,
      "loss_anomaly": 0.024092771112918854,
      "loss_cls": 0.0017024531261995435,
      "loss_mitre": 0.0013937061885371804,
      "loss_ner": 0.021269405260682106,
      "step": 1375
    },
    {
      "epoch": 0.7244501940491591,
      "grad_norm": 1.4661478996276855,
      "learning_rate": 1.686146771412148e-05,
      "loss": 0.28330217361450194,
      "step": 1400
    },
    {
      "epoch": 0.7244501940491591,
      "loss_anomaly": 0.03662571683526039,
      "loss_cls": 0.0001973439211724326,
      "loss_mitre": 0.8232212066650391,
      "loss_ner": 0.024045821279287338,
      "step": 1400
    },
    {
      "epoch": 0.7244501940491591,
      "loss_anomaly": 0.05424869433045387,
      "loss_cls": 0.00038987002335488796,
      "loss_mitre": 0.0016864342615008354,
      "loss_ner": 0.0066833398304879665,
      "step": 1400
    },
    {
      "epoch": 0.7373868046571799,
      "grad_norm": 1.9335036277770996,
      "learning_rate": 1.6765663920291244e-05,
      "loss": 0.2740249443054199,
      "step": 1425
    },
    {
      "epoch": 0.7373868046571799,
      "loss_anomaly": 0.02933870628476143,
      "loss_cls": 0.00021847354946658015,
      "loss_mitre": 0.0010881597409024835,
      "loss_ner": 0.01534744631499052,
      "step": 1425
    },
    {
      "epoch": 0.7373868046571799,
      "loss_anomaly": 0.030352795496582985,
      "loss_cls": 0.0005927593447268009,
      "loss_mitre": 0.0018089774530380964,
      "loss_ner": 0.028235234320163727,
      "step": 1425
    },
    {
      "epoch": 0.7503234152652005,
      "grad_norm": 2.1172397136688232,
      "learning_rate": 1.6669860126461008e-05,
      "loss": 0.3629410171508789,
      "step": 1450
    },
    {
      "epoch": 0.7503234152652005,
      "loss_anomaly": 0.028663773089647293,
      "loss_cls": 0.0004491836589295417,
      "loss_mitre": 0.001362752984277904,
      "loss_ner": 0.0078657241538167,
      "step": 1450
    },
    {
      "epoch": 0.7503234152652005,
      "loss_anomaly": 0.03159458190202713,
      "loss_cls": 0.001696251449175179,
      "loss_mitre": 0.001056863577105105,
      "loss_ner": 0.033637743443250656,
      "step": 1450
    },
    {
      "epoch": 0.7632600258732212,
      "grad_norm": 4.106175899505615,
      "learning_rate": 1.6574056332630774e-05,
      "loss": 0.42166770935058595,
      "step": 1475
    },
    {
      "epoch": 0.7632600258732212,
      "loss_anomaly": 0.014318330213427544,
      "loss_cls": 0.0005142980953678489,
      "loss_mitre": 0.8220781087875366,
      "loss_ner": 0.007649966981261969,
      "step": 1475
    },
    {
      "epoch": 0.7632600258732212,
      "loss_anomaly": 0.03337906301021576,
      "loss_cls": 0.00045864717685617507,
      "loss_mitre": 0.0010977600468322635,
      "loss_ner": 0.021456070244312286,
      "step": 1475
    },
    {
      "epoch": 0.7761966364812419,
      "grad_norm": 2.5052170753479004,
      "learning_rate": 1.6478252538800537e-05,
      "loss": 0.3195821189880371,
      "step": 1500
    },
    {
      "epoch": 0.7761966364812419,
      "eval_anomaly_mae": 0.07873653985813744,
      "eval_anomaly_mse": 0.012156558876394205,
      "eval_anomaly_n_eval": 27475.0,
      "eval_cls_accuracy": 0.99230899734208,
      "eval_cls_auprc": 0.9986360596462154,
      "eval_cls_f1": 0.9731755424063117,
      "eval_cls_n_eval": 17683.0,
      "eval_cls_precision": 0.9503081664098614,
      "eval_cls_recall": 0.9971705739692805,
      "eval_loss": 0.05566314235329628,
      "eval_mitre_accuracy": 0.962847492855287,
      "eval_mitre_macro_f1": 0.1551826622144415,
      "eval_mitre_n_eval": 3849.0,
      "eval_ner_n_tokens": 666658.0,
      "eval_ner_token_acc": 0.9950034350446556,
      "eval_runtime": 68.8284,
      "eval_samples_per_second": 399.181,
      "eval_steps_per_second": 3.124,
      "step": 1500
    },
    {
      "epoch": 0.7761966364812419,
      "loss_anomaly": 0.027138780802488327,
      "loss_cls": 0.00010288995690643787,
      "loss_mitre": 0.0012955592246726155,
      "loss_ner": 0.013659147545695305,
      "step": 1500
    },
    {
      "epoch": 0.7761966364812419,
      "loss_anomaly": 0.037488795816898346,
      "loss_cls": 0.00036084410385228693,
      "loss_mitre": 0.02214754931628704,
      "loss_ner": 0.008589695207774639,
      "step": 1500
    },
    {
      "epoch": 0.7891332470892626,
      "grad_norm": 2.0157363414764404,
      "learning_rate": 1.6382448744970304e-05,
      "loss": 0.25744136810302737,
      "step": 1525
    },
    {
      "epoch": 0.7891332470892626,
      "loss_anomaly": 0.02584865316748619,
      "loss_cls": 0.00013460626360028982,
      "loss_mitre": 0.9819451570510864,
      "loss_ner": 0.008688627742230892,
      "step": 1525
    },
    {
      "epoch": 0.7891332470892626,
      "loss_anomaly": 0.03479987382888794,
      "loss_cls": 5.1172210078220814e-05,
      "loss_mitre": 0.0008438712102361023,
      "loss_ner": 0.010025804862380028,
      "step": 1525
    },
    {
      "epoch": 0.8020698576972833,
      "grad_norm": 1.1862971782684326,
      "learning_rate": 1.6286644951140067e-05,
      "loss": 0.25492258071899415,
      "step": 1550
    },
    {
      "epoch": 0.8020698576972833,
      "loss_anomaly": 0.03114180825650692,
      "loss_cls": 0.00044722165330313146,
      "loss_mitre": 0.0012386329472064972,
      "loss_ner": 0.024881526827812195,
      "step": 1550
    },
    {
      "epoch": 0.8020698576972833,
      "loss_anomaly": 0.021546896547079086,
      "loss_cls": 0.0002450781175866723,
      "loss_mitre": 0.0012907489435747266,
      "loss_ner": 0.00662786653265357,
      "step": 1550
    },
    {
      "epoch": 0.815006468305304,
      "grad_norm": 3.5263750553131104,
      "learning_rate": 1.619084115730983e-05,
      "loss": 0.3250553894042969,
      "step": 1575
    },
    {
      "epoch": 0.815006468305304,
      "loss_anomaly": 0.03475474938750267,
      "loss_cls": 0.00019672553753480315,
      "loss_mitre": 1.2826789617538452,
      "loss_ner": 0.020434493198990822,
      "step": 1575
    },
    {
      "epoch": 0.815006468305304,
      "loss_anomaly": 0.02526925317943096,
      "loss_cls": 0.0013996752677485347,
      "loss_mitre": 0.0012240078067407012,
      "loss_ner": 0.023910274729132652,
      "step": 1575
    },
    {
      "epoch": 0.8279430789133247,
      "grad_norm": 1.783176302909851,
      "learning_rate": 1.6095037363479597e-05,
      "loss": 0.3168270874023438,
      "step": 1600
    },
    {
      "epoch": 0.8279430789133247,
      "loss_anomaly": 0.029444359242916107,
      "loss_cls": 0.0013169805752113461,
      "loss_mitre": 0.0011630244553089142,
      "loss_ner": 0.006933807395398617,
      "step": 1600
    },
    {
      "epoch": 0.8279430789133247,
      "loss_anomaly": 0.03782776743173599,
      "loss_cls": 0.00032637911499477923,
      "loss_mitre": 0.0021163844503462315,
      "loss_ner": 0.019955134019255638,
      "step": 1600
    },
    {
      "epoch": 0.8408796895213454,
      "grad_norm": 1.5878742933273315,
      "learning_rate": 1.599923356964936e-05,
      "loss": 0.2444746971130371,
      "step": 1625
    },
    {
      "epoch": 0.8408796895213454,
      "loss_anomaly": 0.03068961203098297,
      "loss_cls": 0.00035219924757257104,
      "loss_mitre": 0.0009333481430076063,
      "loss_ner": 0.012504921294748783,
      "step": 1625
    },
    {
      "epoch": 0.8408796895213454,
      "loss_anomaly": 0.04046826809644699,
      "loss_cls": 0.0002882872358895838,
      "loss_mitre": 0.0012135659344494343,
      "loss_ner": 0.02691074088215828,
      "step": 1625
    },
    {
      "epoch": 0.8538163001293662,
      "grad_norm": 0.9152362942695618,
      "learning_rate": 1.5903429775819123e-05,
      "loss": 0.17250593185424803,
      "step": 1650
    },
    {
      "epoch": 0.8538163001293662,
      "loss_anomaly": 0.021911904215812683,
      "loss_cls": 9.275071352021769e-05,
      "loss_mitre": 0.0008014110499061644,
      "loss_ner": 0.018309077247977257,
      "step": 1650
    },
    {
      "epoch": 0.8538163001293662,
      "loss_anomaly": 0.025614449754357338,
      "loss_cls": 0.00015435040404554456,
      "loss_mitre": 0.4860636591911316,
      "loss_ner": 0.004262399394065142,
      "step": 1650
    },
    {
      "epoch": 0.8667529107373868,
      "grad_norm": 1.4376814365386963,
      "learning_rate": 1.5807625981988887e-05,
      "loss": 0.2658406639099121,
      "step": 1675
    },
    {
      "epoch": 0.8667529107373868,
      "loss_anomaly": 0.019813990220427513,
      "loss_cls": 0.00124869286082685,
      "loss_mitre": 0.0012147987727075815,
      "loss_ner": 0.02259921468794346,
      "step": 1675
    },
    {
      "epoch": 0.8667529107373868,
      "loss_anomaly": 0.01047268882393837,
      "loss_cls": 4.5353855966823176e-05,
      "loss_mitre": 0.0006397221004590392,
      "loss_ner": 0.020802611485123634,
      "step": 1675
    },
    {
      "epoch": 0.8796895213454075,
      "grad_norm": 1.1629902124404907,
      "learning_rate": 1.5711822188158653e-05,
      "loss": 0.2738294792175293,
      "step": 1700
    },
    {
      "epoch": 0.8796895213454075,
      "loss_anomaly": 0.04004453867673874,
      "loss_cls": 0.0005411506281234324,
      "loss_mitre": 0.43969687819480896,
      "loss_ner": 0.0054477727971971035,
      "step": 1700
    },
    {
      "epoch": 0.8796895213454075,
      "loss_anomaly": 0.026284415274858475,
      "loss_cls": 0.0001304253819398582,
      "loss_mitre": 0.0013733640080317855,
      "loss_ner": 0.040846943855285645,
      "step": 1700
    },
    {
      "epoch": 0.8926261319534282,
      "grad_norm": 1.4011722803115845,
      "learning_rate": 1.5616018394328416e-05,
      "loss": 0.19272537231445314,
      "step": 1725
    },
    {
      "epoch": 0.8926261319534282,
      "loss_anomaly": 0.017832405865192413,
      "loss_cls": 7.61475894250907e-05,
      "loss_mitre": 0.0019670207984745502,
      "loss_ner": 0.011657193303108215,
      "step": 1725
    },
    {
      "epoch": 0.8926261319534282,
      "loss_anomaly": 0.026118949055671692,
      "loss_cls": 0.00028680352261289954,
      "loss_mitre": 0.001415868173353374,
      "loss_ner": 0.005695144180208445,
      "step": 1725
    },
    {
      "epoch": 0.9055627425614489,
      "grad_norm": 2.343963623046875,
      "learning_rate": 1.552021460049818e-05,
      "loss": 0.19371442794799804,
      "step": 1750
    },
    {
      "epoch": 0.9055627425614489,
      "loss_anomaly": 0.048529934138059616,
      "loss_cls": 0.005885741673409939,
      "loss_mitre": 0.0011815240141004324,
      "loss_ner": 0.018601063638925552,
      "step": 1750
    },
    {
      "epoch": 0.9055627425614489,
      "loss_anomaly": 0.031835854053497314,
      "loss_cls": 0.0024055379908531904,
      "loss_mitre": 0.865527331829071,
      "loss_ner": 0.014334994368255138,
      "step": 1750
    },
    {
      "epoch": 0.9184993531694696,
      "grad_norm": 1.3225733041763306,
      "learning_rate": 1.5424410806667943e-05,
      "loss": 0.22574737548828125,
      "step": 1775
    },
    {
      "epoch": 0.9184993531694696,
      "loss_anomaly": 0.020961880683898926,
      "loss_cls": 0.0014616218395531178,
      "loss_mitre": 1.1574831008911133,
      "loss_ner": 0.0012983516789972782,
      "step": 1775
    },
    {
      "epoch": 0.9184993531694696,
      "loss_anomaly": 0.020367208868265152,
      "loss_cls": 0.0005371167790144682,
      "loss_mitre": 0.0007229397888295352,
      "loss_ner": 0.009105700068175793,
      "step": 1775
    },
    {
      "epoch": 0.9314359637774903,
      "grad_norm": 1.574251413345337,
      "learning_rate": 1.532860701283771e-05,
      "loss": 0.18649547576904296,
      "step": 1800
    },
    {
      "epoch": 0.9314359637774903,
      "loss_anomaly": 0.03378242999315262,
      "loss_cls": 0.021547237411141396,
      "loss_mitre": 0.0011016237549483776,
      "loss_ner": 0.01149884145706892,
      "step": 1800
    },
    {
      "epoch": 0.9314359637774903,
      "loss_anomaly": 0.016528617590665817,
      "loss_cls": 0.0007216122467070818,
      "loss_mitre": 1.036590576171875,
      "loss_ner": 0.008952755481004715,
      "step": 1800
    },
    {
      "epoch": 0.944372574385511,
      "grad_norm": 1.4529359340667725,
      "learning_rate": 1.5232803219007474e-05,
      "loss": 0.2877260398864746,
      "step": 1825
    },
    {
      "epoch": 0.944372574385511,
      "loss_anomaly": 0.025467053055763245,
      "loss_cls": 0.00015425081073772162,
      "loss_mitre": 1.4243755340576172,
      "loss_ner": 0.009435012936592102,
      "step": 1825
    },
    {
      "epoch": 0.944372574385511,
      "loss_anomaly": 0.01762721687555313,
      "loss_cls": 0.00016232478083111346,
      "loss_mitre": 0.0007645422010682523,
      "loss_ner": 0.031243233010172844,
      "step": 1825
    },
    {
      "epoch": 0.9573091849935317,
      "grad_norm": 1.1004568338394165,
      "learning_rate": 1.5136999425177239e-05,
      "loss": 0.2271733283996582,
      "step": 1850
    },
    {
      "epoch": 0.9573091849935317,
      "loss_anomaly": 0.03851395845413208,
      "loss_cls": 0.0006107771187089384,
      "loss_mitre": 0.5497031211853027,
      "loss_ner": 0.003301826771348715,
      "step": 1850
    },
    {
      "epoch": 0.9573091849935317,
      "loss_anomaly": 0.020849237218499184,
      "loss_cls": 0.00012773540220223367,
      "loss_mitre": 0.001424551592208445,
      "loss_ner": 0.008700983598828316,
      "step": 1850
    },
    {
      "epoch": 0.9702457956015524,
      "grad_norm": 2.4250588417053223,
      "learning_rate": 1.5041195631347002e-05,
      "loss": 0.28693548202514646,
      "step": 1875
    },
    {
      "epoch": 0.9702457956015524,
      "loss_anomaly": 0.02998732030391693,
      "loss_cls": 0.0002579705324023962,
      "loss_mitre": 0.0009674489265307784,
      "loss_ner": 0.020900435745716095,
      "step": 1875
    },
    {
      "epoch": 0.9702457956015524,
      "loss_anomaly": 0.023745644837617874,
      "loss_cls": 0.001049898797646165,
      "loss_mitre": 0.0006965760258026421,
      "loss_ner": 0.018128128722310066,
      "step": 1875
    },
    {
      "epoch": 0.9831824062095731,
      "grad_norm": 2.139815330505371,
      "learning_rate": 1.4945391837516767e-05,
      "loss": 0.25633460998535157,
      "step": 1900
    },
    {
      "epoch": 0.9831824062095731,
      "loss_anomaly": 0.020074013620615005,
      "loss_cls": 0.0010015229927375913,
      "loss_mitre": 0.0010988382855430245,
      "loss_ner": 0.029696721583604813,
      "step": 1900
    },
    {
      "epoch": 0.9831824062095731,
      "loss_anomaly": 0.021022068336606026,
      "loss_cls": 9.426203178009018e-05,
      "loss_mitre": 1.9992709159851074,
      "loss_ner": 0.009341380558907986,
      "step": 1900
    },
    {
      "epoch": 0.9961190168175937,
      "grad_norm": 1.0862038135528564,
      "learning_rate": 1.4849588043686532e-05,
      "loss": 0.29308744430541994,
      "step": 1925
    },
    {
      "epoch": 0.9961190168175937,
      "loss_anomaly": 0.02043241634964943,
      "loss_cls": 0.00013761580339632928,
      "loss_mitre": 0.0009721170645207167,
      "loss_ner": 0.0015850475756451488,
      "step": 1925
    },
    {
      "epoch": 0.9961190168175937,
      "loss_anomaly": 0.02548939548432827,
      "loss_cls": 0.0006969188107177615,
      "loss_mitre": 0.0006972584524191916,
      "loss_ner": 0.0012235039612278342,
      "step": 1925
    },
    {
      "epoch": 1.008796895213454,
      "grad_norm": 0.8356092572212219,
      "learning_rate": 1.4753784249856295e-05,
      "loss": 0.18306594848632812,
      "step": 1950
    },
    {
      "epoch": 1.008796895213454,
      "loss_anomaly": 0.019225988537073135,
      "loss_cls": 0.005056854337453842,
      "loss_mitre": 0.7886368036270142,
      "loss_ner": 0.008366371504962444,
      "step": 1950
    },
    {
      "epoch": 1.008796895213454,
      "loss_anomaly": 0.013612768612802029,
      "loss_cls": 3.722293331520632e-05,
      "loss_mitre": 1.2579680681228638,
      "loss_ner": 0.0064998227171599865,
      "step": 1950
    },
    {
      "epoch": 1.0217335058214747,
      "grad_norm": 1.7102057933807373,
      "learning_rate": 1.465798045602606e-05,
      "loss": 0.2914961051940918,
      "step": 1975
    },
    {
      "epoch": 1.0217335058214747,
      "loss_anomaly": 0.03318130224943161,
      "loss_cls": 4.4262684241402894e-05,
      "loss_mitre": 0.0008646201458759606,
      "loss_ner": 0.01221438217908144,
      "step": 1975
    },
    {
      "epoch": 1.0217335058214747,
      "loss_anomaly": 0.020027531310915947,
      "loss_cls": 0.0002374051109654829,
      "loss_mitre": 0.0008053589845076203,
      "loss_ner": 0.0074626426212489605,
      "step": 1975
    },
    {
      "epoch": 1.0346701164294956,
      "grad_norm": 0.8470438122749329,
      "learning_rate": 1.4562176662195823e-05,
      "loss": 0.20288766860961915,
      "step": 2000
    },
    {
      "epoch": 1.0346701164294956,
      "eval_anomaly_mae": 0.06504074649096765,
      "eval_anomaly_mse": 0.007545223290811399,
      "eval_anomaly_n_eval": 27475.0,
      "eval_cls_accuracy": 0.9969462195328846,
      "eval_cls_auprc": 0.9993931500712193,
      "eval_cls_f1": 0.9890953150242326,
      "eval_cls_n_eval": 17683.0,
      "eval_cls_precision": 0.9882970137207425,
      "eval_cls_recall": 0.9898949070331448,
      "eval_loss": 0.052999772131443024,
      "eval_mitre_accuracy": 0.962847492855287,
      "eval_mitre_macro_f1": 0.1551826622144415,
      "eval_mitre_n_eval": 3849.0,
      "eval_ner_n_tokens": 666658.0,
      "eval_ner_token_acc": 0.9955014415187398,
      "eval_runtime": 68.9094,
      "eval_samples_per_second": 398.712,
      "eval_steps_per_second": 3.12,
      "step": 2000
    },
    {
      "epoch": 1.0346701164294956,
      "loss_anomaly": 0.023906324058771133,
      "loss_cls": 0.0002251170517411083,
      "loss_mitre": 0.6040520071983337,
      "loss_ner": 0.001748745096847415,
      "step": 2000
    },
    {
      "epoch": 1.0346701164294956,
      "loss_anomaly": 0.035044483840465546,
      "loss_cls": 0.00026185301248915493,
      "loss_mitre": 0.0006540652248077095,
      "loss_ner": 0.007632287219166756,
      "step": 2000
    },
    {
      "epoch": 1.0476067270375162,
      "grad_norm": 1.3731613159179688,
      "learning_rate": 1.4466372868365588e-05,
      "loss": 0.20995426177978516,
      "step": 2025
    },
    {
      "epoch": 1.0476067270375162,
      "loss_anomaly": 0.015156958252191544,
      "loss_cls": 0.0013743344461545348,
      "loss_mitre": 1.2271348237991333,
      "loss_ner": 0.020554693415760994,
      "step": 2025
    },
    {
      "epoch": 1.0476067270375162,
      "loss_anomaly": 0.020127009600400925,
      "loss_cls": 0.0006712321192026138,
      "loss_mitre": 0.001445153378881514,
      "loss_ner": 0.006123651750385761,
      "step": 2025
    },
    {
      "epoch": 1.0605433376455369,
      "grad_norm": 2.081221580505371,
      "learning_rate": 1.4370569074535351e-05,
      "loss": 0.24433412551879882,
      "step": 2050
    },
    {
      "epoch": 1.0605433376455369,
      "loss_anomaly": 0.01878357119858265,
      "loss_cls": 0.00014495113282464445,
      "loss_mitre": 0.5021836161613464,
      "loss_ner": 0.004055038094520569,
      "step": 2050
    },
    {
      "epoch": 1.0605433376455369,
      "loss_anomaly": 0.02561427652835846,
      "loss_cls": 0.00039751073927618563,
      "loss_mitre": 0.0007216735393740237,
      "loss_ner": 0.003903037868440151,
      "step": 2050
    },
    {
      "epoch": 1.0734799482535575,
      "grad_norm": 2.3261566162109375,
      "learning_rate": 1.4274765280705116e-05,
      "loss": 0.28628589630126955,
      "step": 2075
    },
    {
      "epoch": 1.0734799482535575,
      "loss_anomaly": 0.013037241995334625,
      "loss_cls": 7.89217883720994e-05,
      "loss_mitre": 0.0006200690986588597,
      "loss_ner": 0.020654648542404175,
      "step": 2075
    },
    {
      "epoch": 1.0734799482535575,
      "loss_anomaly": 0.0202198289334774,
      "loss_cls": 0.00016387243522331119,
      "loss_mitre": 0.0004837829328607768,
      "loss_ner": 0.005061125382781029,
      "step": 2075
    },
    {
      "epoch": 1.0864165588615782,
      "grad_norm": 1.068611979484558,
      "learning_rate": 1.4178961486874881e-05,
      "loss": 0.23995353698730468,
      "step": 2100
    },
    {
      "epoch": 1.0864165588615782,
      "loss_anomaly": 0.03270808607339859,
      "loss_cls": 0.0001325369521509856,
      "loss_mitre": 0.0006522354087792337,
      "loss_ner": 0.01712031662464142,
      "step": 2100
    },
    {
      "epoch": 1.0864165588615782,
      "loss_anomaly": 0.017478816211223602,
      "loss_cls": 0.00010829779785126448,
      "loss_mitre": 0.00080539146438241,
      "loss_ner": 0.020477144047617912,
      "step": 2100
    },
    {
      "epoch": 1.099353169469599,
      "grad_norm": 2.3407516479492188,
      "learning_rate": 1.4083157693044646e-05,
      "loss": 0.4032162094116211,
      "step": 2125
    },
    {
      "epoch": 1.099353169469599,
      "loss_anomaly": 0.015723805874586105,
      "loss_cls": 0.00016424623026978225,
      "loss_mitre": 0.0005868371808901429,
      "loss_ner": 0.005872505716979504,
      "step": 2125
    },
    {
      "epoch": 1.099353169469599,
      "loss_anomaly": 0.04772692546248436,
      "loss_cls": 0.00011684605851769447,
      "loss_mitre": 0.0006373367505148053,
      "loss_ner": 0.03072967566549778,
      "step": 2125
    },
    {
      "epoch": 1.1122897800776197,
      "grad_norm": 0.481103777885437,
      "learning_rate": 1.3987353899214411e-05,
      "loss": 0.3153705596923828,
      "step": 2150
    },
    {
      "epoch": 1.1122897800776197,
      "loss_anomaly": 0.013974327594041824,
      "loss_cls": 8.167556370608509e-05,
      "loss_mitre": 0.000683383725117892,
      "loss_ner": 0.005771585740149021,
      "step": 2150
    },
    {
      "epoch": 1.1122897800776197,
      "loss_anomaly": 0.01934165321290493,
      "loss_cls": 7.479079067707062e-05,
      "loss_mitre": 0.0005174435209482908,
      "loss_ner": 0.002547829644754529,
      "step": 2150
    },
    {
      "epoch": 1.1252263906856403,
      "grad_norm": 2.3815994262695312,
      "learning_rate": 1.3891550105384174e-05,
      "loss": 0.2900893402099609,
      "step": 2175
    },
    {
      "epoch": 1.1252263906856403,
      "loss_anomaly": 0.022310033440589905,
      "loss_cls": 0.007321539800614119,
      "loss_mitre": 0.0006692165043205023,
      "loss_ner": 0.005127822048962116,
      "step": 2175
    },
    {
      "epoch": 1.1252263906856403,
      "loss_anomaly": 0.02089717984199524,
      "loss_cls": 0.01244731992483139,
      "loss_mitre": 0.0007734185201115906,
      "loss_ner": 0.0038180097471922636,
      "step": 2175
    },
    {
      "epoch": 1.138163001293661,
      "grad_norm": 2.495168685913086,
      "learning_rate": 1.3795746311553939e-05,
      "loss": 0.30621482849121096,
      "step": 2200
    },
    {
      "epoch": 1.138163001293661,
      "loss_anomaly": 0.028657875955104828,
      "loss_cls": 0.0004990220768377185,
      "loss_mitre": 1.1394323110580444,
      "loss_ner": 0.0006616846076212823,
      "step": 2200
    },
    {
      "epoch": 1.138163001293661,
      "loss_anomaly": 0.012738168239593506,
      "loss_cls": 0.0002232497645309195,
      "loss_mitre": 0.8477733731269836,
      "loss_ner": 0.05123792216181755,
      "step": 2200
    },
    {
      "epoch": 1.1510996119016816,
      "grad_norm": 2.551671028137207,
      "learning_rate": 1.3699942517723704e-05,
      "loss": 0.2993073844909668,
      "step": 2225
    },
    {
      "epoch": 1.1510996119016816,
      "loss_anomaly": 0.013496749103069305,
      "loss_cls": 0.0006067177746444941,
      "loss_mitre": 0.0007858956814743578,
      "loss_ner": 0.003732062876224518,
      "step": 2225
    },
    {
      "epoch": 1.1510996119016816,
      "loss_anomaly": 0.014321195892989635,
      "loss_cls": 0.00047141703544184566,
      "loss_mitre": 0.0004643360443878919,
      "loss_ner": 0.01032601110637188,
      "step": 2225
    },
    {
      "epoch": 1.1640362225097025,
      "grad_norm": 1.5278488397598267,
      "learning_rate": 1.3604138723893467e-05,
      "loss": 0.26617944717407227,
      "step": 2250
    },
    {
      "epoch": 1.1640362225097025,
      "loss_anomaly": 0.0365014523267746,
      "loss_cls": 0.0051254816353321075,
      "loss_mitre": 0.0007170307799242437,
      "loss_ner": 0.01677655056118965,
      "step": 2250
    },
    {
      "epoch": 1.1640362225097025,
      "loss_anomaly": 0.01387342531234026,
      "loss_cls": 9.764728747541085e-05,
      "loss_mitre": 0.0007709644851274788,
      "loss_ner": 0.01660926640033722,
      "step": 2250
    },
    {
      "epoch": 1.1769728331177232,
      "grad_norm": 1.8666132688522339,
      "learning_rate": 1.3508334930063232e-05,
      "loss": 0.3025775146484375,
      "step": 2275
    },
    {
      "epoch": 1.1769728331177232,
      "loss_anomaly": 0.013378642499446869,
      "loss_cls": 0.000887310947291553,
      "loss_mitre": 0.0004750054213218391,
      "loss_ner": 0.04510455206036568,
      "step": 2275
    },
    {
      "epoch": 1.1769728331177232,
      "loss_anomaly": 0.02309090830385685,
      "loss_cls": 0.000867229828145355,
      "loss_mitre": 0.5931065678596497,
      "loss_ner": 0.010082912631332874,
      "step": 2275
    },
    {
      "epoch": 1.1899094437257438,
      "grad_norm": 0.5624589323997498,
      "learning_rate": 1.3412531136232995e-05,
      "loss": 0.32593910217285155,
      "step": 2300
    },
    {
      "epoch": 1.1899094437257438,
      "loss_anomaly": 0.019520899280905724,
      "loss_cls": 4.1231287468690425e-05,
      "loss_mitre": 0.0006004556780681014,
      "loss_ner": 0.023955071344971657,
      "step": 2300
    },
    {
      "epoch": 1.1899094437257438,
      "loss_anomaly": 0.019854847341775894,
      "loss_cls": 0.00019567832350730896,
      "loss_mitre": 0.0007013783324509859,
      "loss_ner": 0.016580786556005478,
      "step": 2300
    },
    {
      "epoch": 1.2028460543337645,
      "grad_norm": 1.4393984079360962,
      "learning_rate": 1.331672734240276e-05,
      "loss": 0.3340352249145508,
      "step": 2325
    },
    {
      "epoch": 1.2028460543337645,
      "loss_anomaly": 0.008132830262184143,
      "loss_cls": 0.0003556626325007528,
      "loss_mitre": 0.0004941364750266075,
      "loss_ner": 0.010740389116108418,
      "step": 2325
    },
    {
      "epoch": 1.2028460543337645,
      "loss_anomaly": 0.019706156104803085,
      "loss_cls": 0.0002016889484366402,
      "loss_mitre": 0.00045537855476140976,
      "loss_ner": 0.014079401269555092,
      "step": 2325
    },
    {
      "epoch": 1.2157826649417853,
      "grad_norm": 2.7967543601989746,
      "learning_rate": 1.3220923548572523e-05,
      "loss": 0.36630577087402344,
      "step": 2350
    },
    {
      "epoch": 1.2157826649417853,
      "loss_anomaly": 0.014452074654400349,
      "loss_cls": 0.0020079209934920073,
      "loss_mitre": 0.00036739170900546014,
      "loss_ner": 0.018935551866889,
      "step": 2350
    },
    {
      "epoch": 1.2157826649417853,
      "loss_anomaly": 0.015323217958211899,
      "loss_cls": 0.0015238523483276367,
      "loss_mitre": 0.0006302858819253743,
      "loss_ner": 0.009866571053862572,
      "step": 2350
    },
    {
      "epoch": 1.228719275549806,
      "grad_norm": 0.7481634616851807,
      "learning_rate": 1.3125119754742288e-05,
      "loss": 0.32022560119628907,
      "step": 2375
    },
    {
      "epoch": 1.228719275549806,
      "loss_anomaly": 0.014570397324860096,
      "loss_cls": 0.00024905227473936975,
      "loss_mitre": 0.00041999941458925605,
      "loss_ner": 0.024214105680584908,
      "step": 2375
    },
    {
      "epoch": 1.228719275549806,
      "loss_anomaly": 0.016893548890948296,
      "loss_cls": 0.0001635189400985837,
      "loss_mitre": 1.986352801322937,
      "loss_ner": 0.02190866693854332,
      "step": 2375
    },
    {
      "epoch": 1.2416558861578266,
      "grad_norm": 0.6428819298744202,
      "learning_rate": 1.3029315960912053e-05,
      "loss": 0.32503883361816405,
      "step": 2400
    },
    {
      "epoch": 1.2416558861578266,
      "loss_anomaly": 0.012902333401143551,
      "loss_cls": 0.0006824465817771852,
      "loss_mitre": 0.0006813708459958434,
      "loss_ner": 0.004001901485025883,
      "step": 2400
    },
    {
      "epoch": 1.2416558861578266,
      "loss_anomaly": 0.016256771981716156,
      "loss_cls": 6.013128950144164e-05,
      "loss_mitre": 0.00039899503462947905,
      "loss_ner": 0.001566912280395627,
      "step": 2400
    },
    {
      "epoch": 1.2545924967658473,
      "grad_norm": 0.8010119795799255,
      "learning_rate": 1.2933512167081816e-05,
      "loss": 0.20589805603027345,
      "step": 2425
    },
    {
      "epoch": 1.2545924967658473,
      "loss_anomaly": 0.013803602196276188,
      "loss_cls": 0.0010306210024282336,
      "loss_mitre": 0.0005180782172828913,
      "loss_ner": 0.013143762946128845,
      "step": 2425
    },
    {
      "epoch": 1.2545924967658473,
      "loss_anomaly": 0.013454873114824295,
      "loss_cls": 2.5506282327114604e-05,
      "loss_mitre": 0.00048195908311754465,
      "loss_ner": 0.005375875160098076,
      "step": 2425
    },
    {
      "epoch": 1.2675291073738681,
      "grad_norm": 0.5453146696090698,
      "learning_rate": 1.2837708373251583e-05,
      "loss": 0.36898643493652344,
      "step": 2450
    },
    {
      "epoch": 1.2675291073738681,
      "loss_anomaly": 0.012206099927425385,
      "loss_cls": 7.170513708842918e-05,
      "loss_mitre": 0.0009691754239611328,
      "loss_ner": 0.003211378352716565,
      "step": 2450
    },
    {
      "epoch": 1.2675291073738681,
      "loss_anomaly": 0.019067179411649704,
      "loss_cls": 0.000338495010510087,
      "loss_mitre": 0.0007351937238126993,
      "loss_ner": 0.010844800621271133,
      "step": 2450
    },
    {
      "epoch": 1.2804657179818888,
      "grad_norm": 1.8999277353286743,
      "learning_rate": 1.2741904579421348e-05,
      "loss": 0.18926116943359375,
      "step": 2475
    },
    {
      "epoch": 1.2804657179818888,
      "loss_anomaly": 0.027009334415197372,
      "loss_cls": 0.006458726245909929,
      "loss_mitre": 0.46747109293937683,
      "loss_ner": 0.007208418101072311,
      "step": 2475
    },
    {
      "epoch": 1.2804657179818888,
      "loss_anomaly": 0.020106099545955658,
      "loss_cls": 2.764397686405573e-05,
      "loss_mitre": 0.0005335361347533762,
      "loss_ner": 0.011734112165868282,
      "step": 2475
    },
    {
      "epoch": 1.2934023285899094,
      "grad_norm": 1.1794712543487549,
      "learning_rate": 1.2646100785591111e-05,
      "loss": 0.2153662109375,
      "step": 2500
    },
    {
      "epoch": 1.2934023285899094,
      "eval_anomaly_mae": 0.0669968583572739,
      "eval_anomaly_mse": 0.00819596732849952,
      "eval_anomaly_n_eval": 27475.0,
      "eval_cls_accuracy": 0.9914607249901035,
      "eval_cls_auprc": 0.9991892901356755,
      "eval_cls_f1": 0.9703631010794896,
      "eval_cls_n_eval": 17683.0,
      "eval_cls_precision": 0.9431514689049981,
      "eval_cls_recall": 0.9991915925626516,
      "eval_loss": 0.05480101704597473,
      "eval_mitre_accuracy": 0.962847492855287,
      "eval_mitre_macro_f1": 0.1542247445464511,
      "eval_mitre_n_eval": 3849.0,
      "eval_ner_n_tokens": 666658.0,
      "eval_ner_token_acc": 0.9953349393542116,
      "eval_runtime": 68.8708,
      "eval_samples_per_second": 398.936,
      "eval_steps_per_second": 3.122,
      "step": 2500
    },
    {
      "epoch": 1.2934023285899094,
      "loss_anomaly": 0.010878972709178925,
      "loss_cls": 6.187670805957168e-05,
      "loss_mitre": 0.0003779818653129041,
      "loss_ner": 0.1094539612531662,
      "step": 2500
    },
    {
      "epoch": 1.2934023285899094,
      "loss_anomaly": 0.010625490918755531,
      "loss_cls": 8.77404963830486e-05,
      "loss_mitre": 0.00046886358177289367,
      "loss_ner": 0.030667250975966454,
      "step": 2500
    },
    {
      "epoch": 1.30633893919793,
      "grad_norm": 0.6211164593696594,
      "learning_rate": 1.2550296991760876e-05,
      "loss": 0.1816808319091797,
      "step": 2525
    },
    {
      "epoch": 1.30633893919793,
      "loss_anomaly": 0.015835735946893692,
      "loss_cls": 0.0013944017700850964,
      "loss_mitre": 0.8621689677238464,
      "loss_ner": 0.013868269510567188,
      "step": 2525
    },
    {
      "epoch": 1.30633893919793,
      "loss_anomaly": 0.015190893784165382,
      "loss_cls": 0.00020194222452118993,
      "loss_mitre": 0.0005538045079447329,
      "loss_ner": 0.014073855243623257,
      "step": 2525
    },
    {
      "epoch": 1.3192755498059507,
      "grad_norm": 1.6281105279922485,
      "learning_rate": 1.2454493197930639e-05,
      "loss": 0.2751642417907715,
      "step": 2550
    },
    {
      "epoch": 1.3192755498059507,
      "loss_anomaly": 0.008461305871605873,
      "loss_cls": 9.267262066714466e-05,
      "loss_mitre": 0.0003041632880922407,
      "loss_ner": 0.01069162879139185,
      "step": 2550
    },
    {
      "epoch": 1.3192755498059507,
      "loss_anomaly": 0.015457214787602425,
      "loss_cls": 9.374332148581743e-05,
      "loss_mitre": 0.0003866647311951965,
      "loss_ner": 0.01723729446530342,
      "step": 2550
    },
    {
      "epoch": 1.3322121604139716,
      "grad_norm": 1.6300017833709717,
      "learning_rate": 1.2358689404100404e-05,
      "loss": 0.09837031364440918,
      "step": 2575
    },
    {
      "epoch": 1.3322121604139716,
      "loss_anomaly": 0.022395003587007523,
      "loss_cls": 0.012403501197695732,
      "loss_mitre": 0.7393324971199036,
      "loss_ner": 0.00702168233692646,
      "step": 2575
    },
    {
      "epoch": 1.3322121604139716,
      "loss_anomaly": 0.012582188472151756,
      "loss_cls": 0.0006854962557554245,
      "loss_mitre": 0.0009359424002468586,
      "loss_ner": 0.0054854960180819035,
      "step": 2575
    },
    {
      "epoch": 1.3451487710219923,
      "grad_norm": 0.6836620569229126,
      "learning_rate": 1.2262885610270167e-05,
      "loss": 0.21568716049194336,
      "step": 2600
    },
    {
      "epoch": 1.3451487710219923,
      "loss_anomaly": 0.011938800103962421,
      "loss_cls": 5.816365228383802e-05,
      "loss_mitre": 0.00038957787910476327,
      "loss_ner": 0.012005290016531944,
      "step": 2600
    },
    {
      "epoch": 1.3451487710219923,
      "loss_anomaly": 0.023809347301721573,
      "loss_cls": 0.010898502543568611,
      "loss_mitre": 0.0005312489229254425,
      "loss_ner": 0.009086277335882187,
      "step": 2600
    },
    {
      "epoch": 1.358085381630013,
      "grad_norm": 2.1050238609313965,
      "learning_rate": 1.2167081816439932e-05,
      "loss": 0.2545591163635254,
      "step": 2625
    },
    {
      "epoch": 1.358085381630013,
      "loss_anomaly": 0.01691151037812233,
      "loss_cls": 0.00016709815827198327,
      "loss_mitre": 0.000679396151099354,
      "loss_ner": 0.018727079033851624,
      "step": 2625
    },
    {
      "epoch": 1.358085381630013,
      "loss_anomaly": 0.019649256020784378,
      "loss_cls": 0.00017713509441819042,
      "loss_mitre": 0.0004887788672931492,
      "loss_ner": 0.028297269716858864,
      "step": 2625
    },
    {
      "epoch": 1.3710219922380336,
      "grad_norm": 0.6973246335983276,
      "learning_rate": 1.2071278022609695e-05,
      "loss": 0.2401924514770508,
      "step": 2650
    },
    {
      "epoch": 1.3710219922380336,
      "loss_anomaly": 0.022197779268026352,
      "loss_cls": 0.0001472203148296103,
      "loss_mitre": 0.0005663212505169213,
      "loss_ner": 0.019326871261000633,
      "step": 2650
    },
    {
      "epoch": 1.3710219922380336,
      "loss_anomaly": 0.012809923849999905,
      "loss_cls": 8.435967902187258e-05,
      "loss_mitre": 0.0005093385116197169,
      "loss_ner": 0.036501675844192505,
      "step": 2650
    },
    {
      "epoch": 1.3839586028460542,
      "grad_norm": 1.9963735342025757,
      "learning_rate": 1.197547422877946e-05,
      "loss": 0.24824462890625,
      "step": 2675
    },
    {
      "epoch": 1.3839586028460542,
      "loss_anomaly": 0.01372695341706276,
      "loss_cls": 0.0015133093111217022,
      "loss_mitre": 0.9346805214881897,
      "loss_ner": 0.014929568395018578,
      "step": 2675
    },
    {
      "epoch": 1.3839586028460542,
      "loss_anomaly": 0.01090892031788826,
      "loss_cls": 9.218286140821874e-05,
      "loss_mitre": 0.5927169919013977,
      "loss_ner": 0.020914411172270775,
      "step": 2675
    },
    {
      "epoch": 1.396895213454075,
      "grad_norm": 1.163235068321228,
      "learning_rate": 1.1879670434949225e-05,
      "loss": 0.1828053665161133,
      "step": 2700
    },
    {
      "epoch": 1.396895213454075,
      "loss_anomaly": 0.013695789501070976,
      "loss_cls": 0.001141462242230773,
      "loss_mitre": 0.5726767778396606,
      "loss_ner": 0.014582694508135319,
      "step": 2700
    },
    {
      "epoch": 1.396895213454075,
      "loss_anomaly": 0.014948965981602669,
      "loss_cls": 5.958670226391405e-05,
      "loss_mitre": 1.1178618669509888,
      "loss_ner": 0.01075416523963213,
      "step": 2700
    },
    {
      "epoch": 1.4098318240620957,
      "grad_norm": 0.7179940342903137,
      "learning_rate": 1.1783866641118988e-05,
      "loss": 0.2592370223999023,
      "step": 2725
    },
    {
      "epoch": 1.4098318240620957,
      "loss_anomaly": 0.012018092907965183,
      "loss_cls": 0.0003169737756252289,
      "loss_mitre": 0.8908511400222778,
      "loss_ner": 0.0034455040004104376,
      "step": 2725
    },
    {
      "epoch": 1.4098318240620957,
      "loss_anomaly": 0.019606605172157288,
      "loss_cls": 3.895735426340252e-05,
      "loss_mitre": 0.0006469278014265001,
      "loss_ner": 0.053776778280735016,
      "step": 2725
    },
    {
      "epoch": 1.4227684346701164,
      "grad_norm": 3.051974296569824,
      "learning_rate": 1.1688062847288753e-05,
      "loss": 0.23894590377807617,
      "step": 2750
    },
    {
      "epoch": 1.4227684346701164,
      "loss_anomaly": 0.024621905758976936,
      "loss_cls": 0.00012052075908286497,
      "loss_mitre": 0.001186867943033576,
      "loss_ner": 0.02270055189728737,
      "step": 2750
    },
    {
      "epoch": 1.4227684346701164,
      "loss_anomaly": 0.017941489815711975,
      "loss_cls": 3.711535828188062e-05,
      "loss_mitre": 0.4501696228981018,
      "loss_ner": 0.019470414146780968,
      "step": 2750
    },
    {
      "epoch": 1.4357050452781372,
      "grad_norm": 0.6286669373512268,
      "learning_rate": 1.159225905345852e-05,
      "loss": 0.2963027572631836,
      "step": 2775
    },
    {
      "epoch": 1.4357050452781372,
      "loss_anomaly": 0.01491194311529398,
      "loss_cls": 0.005534338764846325,
      "loss_mitre": 0.0004877503961324692,
      "loss_ner": 0.029695909470319748,
      "step": 2775
    },
    {
      "epoch": 1.4357050452781372,
      "loss_anomaly": 0.01918921247124672,
      "loss_cls": 3.283240948803723e-05,
      "loss_mitre": 0.0006283007678575814,
      "loss_ner": 0.025125499814748764,
      "step": 2775
    },
    {
      "epoch": 1.448641655886158,
      "grad_norm": 2.021233320236206,
      "learning_rate": 1.1496455259628283e-05,
      "loss": 0.27625757217407226,
      "step": 2800
    },
    {
      "epoch": 1.448641655886158,
      "loss_anomaly": 0.014241679571568966,
      "loss_cls": 0.00012262605014257133,
      "loss_mitre": 0.0005230527021922171,
      "loss_ner": 0.003387876320630312,
      "step": 2800
    },
    {
      "epoch": 1.448641655886158,
      "loss_anomaly": 0.022427977994084358,
      "loss_cls": 0.0005792703595943749,
      "loss_mitre": 0.0006910700467415154,
      "loss_ner": 0.018860239535570145,
      "step": 2800
    },
    {
      "epoch": 1.4615782664941785,
      "grad_norm": 1.8107495307922363,
      "learning_rate": 1.1400651465798048e-05,
      "loss": 0.2465883255004883,
      "step": 2825
    },
    {
      "epoch": 1.4615782664941785,
      "loss_anomaly": 0.018636995926499367,
      "loss_cls": 3.577331153792329e-05,
      "loss_mitre": 0.0003679011424537748,
      "loss_ner": 0.011252309195697308,
      "step": 2825
    },
    {
      "epoch": 1.4615782664941785,
      "loss_anomaly": 0.020528730005025864,
      "loss_cls": 0.00010540004586800933,
      "loss_mitre": 0.00027465401217341423,
      "loss_ner": 0.008705786429345608,
      "step": 2825
    },
    {
      "epoch": 1.4745148771021992,
      "grad_norm": 0.5550847053527832,
      "learning_rate": 1.1304847671967811e-05,
      "loss": 0.30955841064453127,
      "step": 2850
    },
    {
      "epoch": 1.4745148771021992,
      "loss_anomaly": 0.023095853626728058,
      "loss_cls": 8.502982382196933e-05,
      "loss_mitre": 0.0007495786994695663,
      "loss_ner": 0.00387753127142787,
      "step": 2850
    },
    {
      "epoch": 1.4745148771021992,
      "loss_anomaly": 0.021517198532819748,
      "loss_cls": 0.0033346391282975674,
      "loss_mitre": 0.0005998152773827314,
      "loss_ner": 0.034884702414274216,
      "step": 2850
    },
    {
      "epoch": 1.4874514877102198,
      "grad_norm": 0.7628108263015747,
      "learning_rate": 1.1209043878137576e-05,
      "loss": 0.24397279739379882,
      "step": 2875
    },
    {
      "epoch": 1.4874514877102198,
      "loss_anomaly": 0.010845046490430832,
      "loss_cls": 0.00016504111408721656,
      "loss_mitre": 0.958946168422699,
      "loss_ner": 0.018389496952295303,
      "step": 2875
    },
    {
      "epoch": 1.4874514877102198,
      "loss_anomaly": 0.011439431458711624,
      "loss_cls": 6.390576163539663e-05,
      "loss_mitre": 0.000311239215079695,
      "loss_ner": 0.002312989206984639,
      "step": 2875
    },
    {
      "epoch": 1.5003880983182407,
      "grad_norm": 1.989557147026062,
      "learning_rate": 1.1113240084307339e-05,
      "loss": 0.3191413879394531,
      "step": 2900
    },
    {
      "epoch": 1.5003880983182407,
      "loss_anomaly": 0.013311603106558323,
      "loss_cls": 0.002338236663490534,
      "loss_mitre": 0.000635758857242763,
      "loss_ner": 0.012487683445215225,
      "step": 2900
    },
    {
      "epoch": 1.5003880983182407,
      "loss_anomaly": 0.027776291593909264,
      "loss_cls": 0.020793914794921875,
      "loss_mitre": 0.6856548190116882,
      "loss_ner": 0.008850504644215107,
      "step": 2900
    },
    {
      "epoch": 1.5133247089262614,
      "grad_norm": 1.5332950353622437,
      "learning_rate": 1.1017436290477104e-05,
      "loss": 0.314085636138916,
      "step": 2925
    },
    {
      "epoch": 1.5133247089262614,
      "loss_anomaly": 0.008963122963905334,
      "loss_cls": 3.940012902603485e-05,
      "loss_mitre": 0.00038699762080796063,
      "loss_ner": 0.047626182436943054,
      "step": 2925
    },
    {
      "epoch": 1.5133247089262614,
      "loss_anomaly": 0.017725735902786255,
      "loss_cls": 0.0007755075348541141,
      "loss_mitre": 0.0004693713563028723,
      "loss_ner": 0.008204883895814419,
      "step": 2925
    },
    {
      "epoch": 1.526261319534282,
      "grad_norm": 1.9319820404052734,
      "learning_rate": 1.0921632496646867e-05,
      "loss": 0.22065746307373046,
      "step": 2950
    },
    {
      "epoch": 1.526261319534282,
      "loss_anomaly": 0.02106974832713604,
      "loss_cls": 3.904648474417627e-05,
      "loss_mitre": 0.0003819903649855405,
      "loss_ner": 0.01714177243411541,
      "step": 2950
    },
    {
      "epoch": 1.526261319534282,
      "loss_anomaly": 0.013488267548382282,
      "loss_cls": 1.980853448912967e-05,
      "loss_mitre": 0.0004373960546217859,
      "loss_ner": 0.007951504550874233,
      "step": 2950
    },
    {
      "epoch": 1.5391979301423027,
      "grad_norm": 2.3831703662872314,
      "learning_rate": 1.0825828702816632e-05,
      "loss": 0.24496219635009767,
      "step": 2975
    },
    {
      "epoch": 1.5391979301423027,
      "loss_anomaly": 0.017378518357872963,
      "loss_cls": 6.283900438575074e-05,
      "loss_mitre": 0.00031495417351834476,
      "loss_ner": 0.015951918438076973,
      "step": 2975
    },
    {
      "epoch": 1.5391979301423027,
      "loss_anomaly": 0.012609847821295261,
      "loss_cls": 0.0001453019940527156,
      "loss_mitre": 0.6801510453224182,
      "loss_ner": 0.00387714640237391,
      "step": 2975
    },
    {
      "epoch": 1.5521345407503233,
      "grad_norm": 2.0736501216888428,
      "learning_rate": 1.0730024908986397e-05,
      "loss": 0.34307525634765623,
      "step": 3000
    },
    {
      "epoch": 1.5521345407503233,
      "eval_anomaly_mae": 0.06065390383229388,
      "eval_anomaly_mse": 0.006440325089716522,
      "eval_anomaly_n_eval": 27475.0,
      "eval_cls_accuracy": 0.9968331165526212,
      "eval_cls_auprc": 0.9994118470112706,
      "eval_cls_f1": 0.988791032826261,
      "eval_cls_n_eval": 17683.0,
      "eval_cls_precision": 0.979381443298969,
      "eval_cls_recall": 0.9983831851253031,
      "eval_loss": 0.053022030740976334,
      "eval_mitre_accuracy": 0.962847492855287,
      "eval_mitre_macro_f1": 0.15515600478901564,
      "eval_mitre_n_eval": 3849.0,
      "eval_ner_n_tokens": 666658.0,
      "eval_ner_token_acc": 0.9951339367411777,
      "eval_runtime": 69.266,
      "eval_samples_per_second": 396.659,
      "eval_steps_per_second": 3.104,
      "step": 3000
    },
    {
      "epoch": 1.5521345407503233,
      "loss_anomaly": 0.009740239009261131,
      "loss_cls": 0.00012566568329930305,
      "loss_mitre": 0.0002837084175553173,
      "loss_ner": 0.022708537057042122,
      "step": 3000
    },
    {
      "epoch": 1.5521345407503233,
      "loss_anomaly": 0.016651296988129616,
      "loss_cls": 0.0036468859761953354,
      "loss_mitre": 0.7794739603996277,
      "loss_ner": 0.008276540786027908,
      "step": 3000
    },
    {
      "epoch": 1.5650711513583442,
      "grad_norm": 1.03823983669281,
      "learning_rate": 1.063422111515616e-05,
      "loss": 0.1437802505493164,
      "step": 3025
    },
    {
      "epoch": 1.5650711513583442,
      "loss_anomaly": 0.016815753653645515,
      "loss_cls": 4.8987672926159576e-05,
      "loss_mitre": 0.0003577231545932591,
      "loss_ner": 0.01522695180028677,
      "step": 3025
    },
    {
      "epoch": 1.5650711513583442,
      "loss_anomaly": 0.017188146710395813,
      "loss_cls": 0.0005204758490435779,
      "loss_mitre": 0.00029559832182712853,
      "loss_ner": 0.08285553008317947,
      "step": 3025
    },
    {
      "epoch": 1.5780077619663648,
      "grad_norm": 0.483877956867218,
      "learning_rate": 1.0538417321325925e-05,
      "loss": 0.2943231391906738,
      "step": 3050
    },
    {
      "epoch": 1.5780077619663648,
      "loss_anomaly": 0.017248032614588737,
      "loss_cls": 0.00030716811306774616,
      "loss_mitre": 0.00023548616445623338,
      "loss_ner": 0.0038062557578086853,
      "step": 3050
    },
    {
      "epoch": 1.5780077619663648,
      "loss_anomaly": 0.01716945692896843,
      "loss_cls": 0.00019824097398668528,
      "loss_mitre": 0.0005544639425352216,
      "loss_ner": 0.007636624854058027,
      "step": 3050
    },
    {
      "epoch": 1.5909443725743855,
      "grad_norm": 0.5525802969932556,
      "learning_rate": 1.0442613527495688e-05,
      "loss": 0.3126355171203613,
      "step": 3075
    },
    {
      "epoch": 1.5909443725743855,
      "loss_anomaly": 0.013125371187925339,
      "loss_cls": 0.00016951029829215258,
      "loss_mitre": 0.8770272731781006,
      "loss_ner": 0.002740446710959077,
      "step": 3075
    },
    {
      "epoch": 1.5909443725743855,
      "loss_anomaly": 0.011183667927980423,
      "loss_cls": 4.190829349681735e-05,
      "loss_mitre": 1.044488787651062,
      "loss_ner": 0.0018307386199012399,
      "step": 3075
    },
    {
      "epoch": 1.6038809831824064,
      "grad_norm": 2.2551937103271484,
      "learning_rate": 1.0346809733665455e-05,
      "loss": 0.34402904510498045,
      "step": 3100
    },
    {
      "epoch": 1.6038809831824064,
      "loss_anomaly": 0.014632594771683216,
      "loss_cls": 5.3048985137138516e-05,
      "loss_mitre": 0.00037217713543213904,
      "loss_ner": 0.007123979274183512,
      "step": 3100
    },
    {
      "epoch": 1.6038809831824064,
      "loss_anomaly": 0.015848295763134956,
      "loss_cls": 3.1169289286481217e-05,
      "loss_mitre": 0.00037824345054104924,
      "loss_ner": 0.006550630554556847,
      "step": 3100
    },
    {
      "epoch": 1.6168175937904268,
      "grad_norm": 1.0736223459243774,
      "learning_rate": 1.025100593983522e-05,
      "loss": 0.18241216659545897,
      "step": 3125
    },
    {
      "epoch": 1.6168175937904268,
      "loss_anomaly": 0.024028368294239044,
      "loss_cls": 5.0393962737871334e-05,
      "loss_mitre": 1.2683638334274292,
      "loss_ner": 0.012304785661399364,
      "step": 3125
    },
    {
      "epoch": 1.6168175937904268,
      "loss_anomaly": 0.01591101475059986,
      "loss_cls": 0.0001066147378878668,
      "loss_mitre": 0.0005610958905890584,
      "loss_ner": 0.0024503623135387897,
      "step": 3125
    },
    {
      "epoch": 1.6297542043984476,
      "grad_norm": 1.5398974418640137,
      "learning_rate": 1.0155202146004983e-05,
      "loss": 0.28943666458129885,
      "step": 3150
    },
    {
      "epoch": 1.6297542043984476,
      "loss_anomaly": 0.016260845586657524,
      "loss_cls": 6.0208440118003637e-05,
      "loss_mitre": 0.00032423785887658596,
      "loss_ner": 0.01831919699907303,
      "step": 3150
    },
    {
      "epoch": 1.6297542043984476,
      "loss_anomaly": 0.010116100311279297,
      "loss_cls": 0.00020345351367723197,
      "loss_mitre": 0.8215476870536804,
      "loss_ner": 0.006925326772034168,
      "step": 3150
    },
    {
      "epoch": 1.6426908150064683,
      "grad_norm": 1.5689847469329834,
      "learning_rate": 1.0059398352174748e-05,
      "loss": 0.3190275192260742,
      "step": 3175
    },
    {
      "epoch": 1.6426908150064683,
      "loss_anomaly": 0.008585051633417606,
      "loss_cls": 0.0003338292008265853,
      "loss_mitre": 0.0002780704526230693,
      "loss_ner": 0.01989084854722023,
      "step": 3175
    },
    {
      "epoch": 1.6426908150064683,
      "loss_anomaly": 0.008763767778873444,
      "loss_cls": 0.00010170751193072647,
      "loss_mitre": 0.0003170716227032244,
      "loss_ner": 0.0030237610917538404,
      "step": 3175
    },
    {
      "epoch": 1.655627425614489,
      "grad_norm": 3.661947727203369,
      "learning_rate": 9.963594558344511e-06,
      "loss": 0.2891384506225586,
      "step": 3200
    },
    {
      "epoch": 1.655627425614489,
      "loss_anomaly": 0.015240545384585857,
      "loss_cls": 3.656553235487081e-05,
      "loss_mitre": 0.0002243490016553551,
      "loss_ner": 0.017177367582917213,
      "step": 3200
    },
    {
      "epoch": 1.655627425614489,
      "loss_anomaly": 0.009892044588923454,
      "loss_cls": 0.0002472090709488839,
      "loss_mitre": 1.4188412427902222,
      "loss_ner": 0.003667959477752447,
      "step": 3200
    },
    {
      "epoch": 1.6685640362225098,
      "grad_norm": 0.6044095754623413,
      "learning_rate": 9.867790764514276e-06,
      "loss": 0.16328802108764648,
      "step": 3225
    },
    {
      "epoch": 1.6685640362225098,
      "loss_anomaly": 0.01665683090686798,
      "loss_cls": 0.00018484868633095175,
      "loss_mitre": 0.6360612511634827,
      "loss_ner": 0.008440269157290459,
      "step": 3225
    },
    {
      "epoch": 1.6685640362225098,
      "loss_anomaly": 0.009340588934719563,
      "loss_cls": 0.000187637735507451,
      "loss_mitre": 0.0002500047557987273,
      "loss_ner": 0.004861086141318083,
      "step": 3225
    },
    {
      "epoch": 1.6815006468305302,
      "grad_norm": 1.8614221811294556,
      "learning_rate": 9.771986970684039e-06,
      "loss": 0.3485657501220703,
      "step": 3250
    },
    {
      "epoch": 1.6815006468305302,
      "loss_anomaly": 0.015193227678537369,
      "loss_cls": 4.1644761950010434e-05,
      "loss_mitre": 0.00035684052272699773,
      "loss_ner": 0.002724610036239028,
      "step": 3250
    },
    {
      "epoch": 1.6815006468305302,
      "loss_anomaly": 0.015739228576421738,
      "loss_cls": 4.071089279022999e-05,
      "loss_mitre": 0.00046605177340097725,
      "loss_ner": 0.005115457344800234,
      "step": 3250
    },
    {
      "epoch": 1.6944372574385511,
      "grad_norm": 0.6039745211601257,
      "learning_rate": 9.676183176853804e-06,
      "loss": 0.3137714958190918,
      "step": 3275
    },
    {
      "epoch": 1.6944372574385511,
      "loss_anomaly": 0.016523994505405426,
      "loss_cls": 0.00035512723843567073,
      "loss_mitre": 0.0003880197473336011,
      "loss_ner": 0.017620591446757317,
      "step": 3275
    },
    {
      "epoch": 1.6944372574385511,
      "loss_anomaly": 0.016961077228188515,
      "loss_cls": 0.0001830564287956804,
      "loss_mitre": 0.0004005334048997611,
      "loss_ner": 0.0060051423497498035,
      "step": 3275
    },
    {
      "epoch": 1.7073738680465718,
      "grad_norm": 1.1061040163040161,
      "learning_rate": 9.580379383023569e-06,
      "loss": 0.22100120544433594,
      "step": 3300
    },
    {
      "epoch": 1.7073738680465718,
      "loss_anomaly": 0.010083070024847984,
      "loss_cls": 0.00029682792956009507,
      "loss_mitre": 1.40492844581604,
      "loss_ner": 0.004322717897593975,
      "step": 3300
    },
    {
      "epoch": 1.7073738680465718,
      "loss_anomaly": 0.019687607884407043,
      "loss_cls": 4.8832625907380134e-05,
      "loss_mitre": 0.00029066193383187056,
      "loss_ner": 0.0083921542391181,
      "step": 3300
    },
    {
      "epoch": 1.7203104786545924,
      "grad_norm": 1.9494675397872925,
      "learning_rate": 9.484575589193334e-06,
      "loss": 0.26554988861083983,
      "step": 3325
    },
    {
      "epoch": 1.7203104786545924,
      "loss_anomaly": 0.013344477862119675,
      "loss_cls": 0.0005308371619321406,
      "loss_mitre": 0.00031528930412605405,
      "loss_ner": 0.023450151085853577,
      "step": 3325
    },
    {
      "epoch": 1.7203104786545924,
      "loss_anomaly": 0.01631994917988777,
      "loss_cls": 0.00016061817586887628,
      "loss_mitre": 0.00035562150878831744,
      "loss_ner": 0.025084493681788445,
      "step": 3325
    },
    {
      "epoch": 1.7332470892626133,
      "grad_norm": 0.5726103782653809,
      "learning_rate": 9.388771795363097e-06,
      "loss": 0.2843345260620117,
      "step": 3350
    },
    {
      "epoch": 1.7332470892626133,
      "loss_anomaly": 0.013320177793502808,
      "loss_cls": 0.0005615715635940433,
      "loss_mitre": 0.00024170574033632874,
      "loss_ner": 0.019564880058169365,
      "step": 3350
    },
    {
      "epoch": 1.7332470892626133,
      "loss_anomaly": 0.019075587391853333,
      "loss_cls": 5.192687240196392e-05,
      "loss_mitre": 0.39112210273742676,
      "loss_ner": 0.00992780551314354,
      "step": 3350
    },
    {
      "epoch": 1.746183699870634,
      "grad_norm": 0.4476933777332306,
      "learning_rate": 9.292968001532862e-06,
      "loss": 0.18291492462158204,
      "step": 3375
    },
    {
      "epoch": 1.746183699870634,
      "loss_anomaly": 0.012598084285855293,
      "loss_cls": 0.0009073443943634629,
      "loss_mitre": 0.8664438724517822,
      "loss_ner": 0.006114356219768524,
      "step": 3375
    },
    {
      "epoch": 1.746183699870634,
      "loss_anomaly": 0.01685905084013939,
      "loss_cls": 0.001568684703670442,
      "loss_mitre": 0.0004991695168428123,
      "loss_ner": 0.03309878334403038,
      "step": 3375
    },
    {
      "epoch": 1.7591203104786546,
      "grad_norm": 0.7106324434280396,
      "learning_rate": 9.197164207702625e-06,
      "loss": 0.30513988494873046,
      "step": 3400
    },
    {
      "epoch": 1.7591203104786546,
      "loss_anomaly": 0.012734104879200459,
      "loss_cls": 0.003784437198191881,
      "loss_mitre": 0.00024524968466721475,
      "loss_ner": 0.00029471245943568647,
      "step": 3400
    },
    {
      "epoch": 1.7591203104786546,
      "loss_anomaly": 0.007143813651055098,
      "loss_cls": 6.995545845711604e-05,
      "loss_mitre": 0.00033021034323610365,
      "loss_ner": 0.02161831595003605,
      "step": 3400
    },
    {
      "epoch": 1.7720569210866755,
      "grad_norm": 0.8984391093254089,
      "learning_rate": 9.10136041387239e-06,
      "loss": 0.28195699691772463,
      "step": 3425
    },
    {
      "epoch": 1.7720569210866755,
      "loss_anomaly": 0.010857810266315937,
      "loss_cls": 1.7781916540116072e-05,
      "loss_mitre": 0.00042408783338032663,
      "loss_ner": 0.007513903547078371,
      "step": 3425
    },
    {
      "epoch": 1.7720569210866755,
      "loss_anomaly": 0.007702332455664873,
      "loss_cls": 0.000801904418040067,
      "loss_mitre": 1.8239386081695557,
      "loss_ner": 0.01817120611667633,
      "step": 3425
    },
    {
      "epoch": 1.7849935316946959,
      "grad_norm": 1.0031296014785767,
      "learning_rate": 9.005556620042155e-06,
      "loss": 0.26656059265136717,
      "step": 3450
    },
    {
      "epoch": 1.7849935316946959,
      "loss_anomaly": 0.01269534882158041,
      "loss_cls": 0.00025101163191720843,
      "loss_mitre": 0.0005213024560362101,
      "loss_ner": 0.0027407542802393436,
      "step": 3450
    },
    {
      "epoch": 1.7849935316946959,
      "loss_anomaly": 0.02052510716021061,
      "loss_cls": 0.0024114700499922037,
      "loss_mitre": 0.6056978702545166,
      "loss_ner": 0.002207239856943488,
      "step": 3450
    },
    {
      "epoch": 1.7979301423027167,
      "grad_norm": 1.0015690326690674,
      "learning_rate": 8.909752826211918e-06,
      "loss": 0.24145267486572267,
      "step": 3475
    },
    {
      "epoch": 1.7979301423027167,
      "loss_anomaly": 0.01354694738984108,
      "loss_cls": 0.0003511975228320807,
      "loss_mitre": 0.6443483829498291,
      "loss_ner": 0.01706799305975437,
      "step": 3475
    },
    {
      "epoch": 1.7979301423027167,
      "loss_anomaly": 0.027390889823436737,
      "loss_cls": 0.001245654420927167,
      "loss_mitre": 0.00036267630639486015,
      "loss_ner": 0.06720304489135742,
      "step": 3475
    },
    {
      "epoch": 1.8108667529107374,
      "grad_norm": 0.9977234601974487,
      "learning_rate": 8.813949032381683e-06,
      "loss": 0.26366716384887695,
      "step": 3500
    },
    {
      "epoch": 1.8108667529107374,
      "eval_anomaly_mae": 0.06021205218655868,
      "eval_anomaly_mse": 0.0062111569792523014,
      "eval_anomaly_n_eval": 27475.0,
      "eval_cls_accuracy": 0.9980206978453883,
      "eval_cls_auprc": 0.99962880084875,
      "eval_cls_f1": 0.9929534930541575,
      "eval_cls_n_eval": 17683.0,
      "eval_cls_precision": 0.9891696750902527,
      "eval_cls_recall": 0.9967663702506063,
      "eval_loss": 0.052463434636592865,
      "eval_mitre_accuracy": 0.962847492855287,
      "eval_mitre_macro_f1": 0.153278580469479,
      "eval_mitre_n_eval": 3849.0,
      "eval_ner_n_tokens": 666658.0,
      "eval_ner_token_acc": 0.9961584500598508,
      "eval_runtime": 68.8381,
      "eval_samples_per_second": 399.125,
      "eval_steps_per_second": 3.123,
      "step": 3500
    },
    {
      "epoch": 1.8108667529107374,
      "loss_anomaly": 0.00882576685398817,
      "loss_cls": 0.0006897655548527837,
      "loss_mitre": 0.0004122978716623038,
      "loss_ner": 0.007296321447938681,
      "step": 3500
    },
    {
      "epoch": 1.8108667529107374,
      "loss_anomaly": 0.015035003423690796,
      "loss_cls": 0.00013635579671245068,
      "loss_mitre": 0.0003419334243517369,
      "loss_ner": 0.00973536353558302,
      "step": 3500
    },
    {
      "epoch": 1.823803363518758,
      "grad_norm": 0.7065778970718384,
      "learning_rate": 8.718145238551448e-06,
      "loss": 0.20447547912597655,
      "step": 3525
    },
    {
      "epoch": 1.823803363518758,
      "loss_anomaly": 0.014240586198866367,
      "loss_cls": 0.000674597395118326,
      "loss_mitre": 1.103667140007019,
      "loss_ner": 0.013945364393293858,
      "step": 3525
    },
    {
      "epoch": 1.823803363518758,
      "loss_anomaly": 0.006432974711060524,
      "loss_cls": 0.0002866593422368169,
      "loss_mitre": 0.0002242319460492581,
      "loss_ner": 0.05720466375350952,
      "step": 3525
    },
    {
      "epoch": 1.836739974126779,
      "grad_norm": 1.7170870304107666,
      "learning_rate": 8.622341444721211e-06,
      "loss": 0.2319007110595703,
      "step": 3550
    },
    {
      "epoch": 1.836739974126779,
      "loss_anomaly": 0.006782154086977243,
      "loss_cls": 6.733499321853742e-05,
      "loss_mitre": 0.0004929343704134226,
      "loss_ner": 0.013774075545370579,
      "step": 3550
    },
    {
      "epoch": 1.836739974126779,
      "loss_anomaly": 0.008127948269248009,
      "loss_cls": 4.2949137423420325e-05,
      "loss_mitre": 0.00029730054666288197,
      "loss_ner": 0.009983094409108162,
      "step": 3550
    },
    {
      "epoch": 1.8496765847347993,
      "grad_norm": 0.8989444375038147,
      "learning_rate": 8.526537650890976e-06,
      "loss": 0.2669964790344238,
      "step": 3575
    },
    {
      "epoch": 1.8496765847347993,
      "loss_anomaly": 0.012247581034898758,
      "loss_cls": 0.000707136990968138,
      "loss_mitre": 0.00022664667631033808,
      "loss_ner": 0.001999597530812025,
      "step": 3575
    },
    {
      "epoch": 1.8496765847347993,
      "loss_anomaly": 0.01648803800344467,
      "loss_cls": 5.1071274356218055e-05,
      "loss_mitre": 0.3762340247631073,
      "loss_ner": 0.005339115858078003,
      "step": 3575
    },
    {
      "epoch": 1.8626131953428202,
      "grad_norm": 2.067859649658203,
      "learning_rate": 8.43073385706074e-06,
      "loss": 0.18086437225341798,
      "step": 3600
    },
    {
      "epoch": 1.8626131953428202,
      "loss_anomaly": 0.010289935395121574,
      "loss_cls": 0.0002096419921144843,
      "loss_mitre": 0.0006814638036303222,
      "loss_ner": 0.014274529181420803,
      "step": 3600
    },
    {
      "epoch": 1.8626131953428202,
      "loss_anomaly": 0.014422830194234848,
      "loss_cls": 8.224547491408885e-05,
      "loss_mitre": 1.197877049446106,
      "loss_ner": 0.003829971421509981,
      "step": 3600
    },
    {
      "epoch": 1.8755498059508409,
      "grad_norm": 1.5562846660614014,
      "learning_rate": 8.334930063230504e-06,
      "loss": 0.34598915100097655,
      "step": 3625
    },
    {
      "epoch": 1.8755498059508409,
      "loss_anomaly": 0.013270292431116104,
      "loss_cls": 7.640532567165792e-05,
      "loss_mitre": 0.7115105986595154,
      "loss_ner": 0.0011268486268818378,
      "step": 3625
    },
    {
      "epoch": 1.8755498059508409,
      "loss_anomaly": 0.013531089760363102,
      "loss_cls": 0.0010494283633306623,
      "loss_mitre": 0.0003532020200509578,
      "loss_ner": 0.0073974537663161755,
      "step": 3625
    },
    {
      "epoch": 1.8884864165588615,
      "grad_norm": 1.83979332447052,
      "learning_rate": 8.239126269400269e-06,
      "loss": 0.27997690200805664,
      "step": 3650
    },
    {
      "epoch": 1.8884864165588615,
      "loss_anomaly": 0.013621575199067593,
      "loss_cls": 8.4779312601313e-05,
      "loss_mitre": 0.0003371862694621086,
      "loss_ner": 0.016381412744522095,
      "step": 3650
    },
    {
      "epoch": 1.8884864165588615,
      "loss_anomaly": 0.011506017297506332,
      "loss_cls": 0.00038539525121450424,
      "loss_mitre": 0.00035716858110390604,
      "loss_ner": 0.002431832952424884,
      "step": 3650
    },
    {
      "epoch": 1.9014230271668824,
      "grad_norm": 1.0744963884353638,
      "learning_rate": 8.143322475570034e-06,
      "loss": 0.21728593826293946,
      "step": 3675
    },
    {
      "epoch": 1.9014230271668824,
      "loss_anomaly": 0.014505377039313316,
      "loss_cls": 6.845614552730694e-05,
      "loss_mitre": 0.00031231340835802257,
      "loss_ner": 0.0031559141352772713,
      "step": 3675
    },
    {
      "epoch": 1.9014230271668824,
      "loss_anomaly": 0.017829257994890213,
      "loss_cls": 0.00014526411541737616,
      "loss_mitre": 0.0003723620029632002,
      "loss_ner": 0.0074340952560305595,
      "step": 3675
    },
    {
      "epoch": 1.9143596377749028,
      "grad_norm": 1.7731951475143433,
      "learning_rate": 8.047518681739799e-06,
      "loss": 0.23735828399658204,
      "step": 3700
    },
    {
      "epoch": 1.9143596377749028,
      "loss_anomaly": 0.00931689515709877,
      "loss_cls": 0.00011358175106579438,
      "loss_mitre": 0.00028601690428331494,
      "loss_ner": 0.018421567976474762,
      "step": 3700
    },
    {
      "epoch": 1.9143596377749028,
      "loss_anomaly": 0.021102124825119972,
      "loss_cls": 6.706844578729942e-05,
      "loss_mitre": 0.00020552454225253314,
      "loss_ner": 0.01248264778405428,
      "step": 3700
    },
    {
      "epoch": 1.9272962483829237,
      "grad_norm": 0.5731003284454346,
      "learning_rate": 7.951714887909562e-06,
      "loss": 0.2937332725524902,
      "step": 3725
    },
    {
      "epoch": 1.9272962483829237,
      "loss_anomaly": 0.013354729861021042,
      "loss_cls": 0.0002184194017900154,
      "loss_mitre": 0.00015485627227462828,
      "loss_ner": 0.007030876819044352,
      "step": 3725
    },
    {
      "epoch": 1.9272962483829237,
      "loss_anomaly": 0.008557142689824104,
      "loss_cls": 8.368914132006466e-05,
      "loss_mitre": 0.6412281394004822,
      "loss_ner": 0.010540354065597057,
      "step": 3725
    },
    {
      "epoch": 1.9402328589909443,
      "grad_norm": 0.608862042427063,
      "learning_rate": 7.855911094079327e-06,
      "loss": 0.19168155670166015,
      "step": 3750
    },
    {
      "epoch": 1.9402328589909443,
      "loss_anomaly": 0.013506601564586163,
      "loss_cls": 3.73452712665312e-05,
      "loss_mitre": 0.4528420567512512,
      "loss_ner": 0.005484382621943951,
      "step": 3750
    },
    {
      "epoch": 1.9402328589909443,
      "loss_anomaly": 0.014446323737502098,
      "loss_cls": 0.00012811571650672704,
      "loss_mitre": 0.561677873134613,
      "loss_ner": 0.005451133940368891,
      "step": 3750
    },
    {
      "epoch": 1.953169469598965,
      "grad_norm": 0.7754667401313782,
      "learning_rate": 7.76010730024909e-06,
      "loss": 0.24718086242675782,
      "step": 3775
    },
    {
      "epoch": 1.953169469598965,
      "loss_anomaly": 0.005721200257539749,
      "loss_cls": 2.146512815670576e-05,
      "loss_mitre": 0.0006499694427475333,
      "loss_ner": 0.009109324775636196,
      "step": 3775
    },
    {
      "epoch": 1.953169469598965,
      "loss_anomaly": 0.007130315061658621,
      "loss_cls": 5.991811121930368e-05,
      "loss_mitre": 0.0003713934274856001,
      "loss_ner": 0.004676891956478357,
      "step": 3775
    },
    {
      "epoch": 1.9661060802069859,
      "grad_norm": 1.0227056741714478,
      "learning_rate": 7.664303506418855e-06,
      "loss": 0.24416637420654297,
      "step": 3800
    },
    {
      "epoch": 1.9661060802069859,
      "loss_anomaly": 0.009808780625462532,
      "loss_cls": 0.00015462306328117847,
      "loss_mitre": 0.00026141590205952525,
      "loss_ner": 0.015597537159919739,
      "step": 3800
    },
    {
      "epoch": 1.9661060802069859,
      "loss_anomaly": 0.010770508088171482,
      "loss_cls": 5.8081222960026935e-05,
      "loss_mitre": 0.0002535905805416405,
      "loss_ner": 0.007329947780817747,
      "step": 3800
    },
    {
      "epoch": 1.9790426908150065,
      "grad_norm": 0.5006385445594788,
      "learning_rate": 7.5684997125886196e-06,
      "loss": 0.23605180740356446,
      "step": 3825
    },
    {
      "epoch": 1.9790426908150065,
      "loss_anomaly": 0.008122812025249004,
      "loss_cls": 3.245268453611061e-05,
      "loss_mitre": 0.0002717509923968464,
      "loss_ner": 0.002587289083749056,
      "step": 3825
    },
    {
      "epoch": 1.9790426908150065,
      "loss_anomaly": 0.015802983194589615,
      "loss_cls": 3.6965404433431104e-05,
      "loss_mitre": 0.00032987023587338626,
      "loss_ner": 0.007811681367456913,
      "step": 3825
    },
    {
      "epoch": 1.9919793014230271,
      "grad_norm": 0.6727224588394165,
      "learning_rate": 7.472695918758384e-06,
      "loss": 0.14334449768066407,
      "step": 3850
    },
    {
      "epoch": 1.9919793014230271,
      "loss_anomaly": 0.00593550456687808,
      "loss_cls": 4.987114152754657e-05,
      "loss_mitre": 0.0003577448951546103,
      "loss_ner": 0.046022433787584305,
      "step": 3850
    },
    {
      "epoch": 1.9919793014230271,
      "loss_anomaly": 0.015977051109075546,
      "loss_cls": 1.7043590560206212e-05,
      "loss_mitre": 0.00035829434636980295,
      "loss_ner": 0.00821262039244175,
      "step": 3850
    },
    {
      "epoch": 2.0046571798188872,
      "grad_norm": 2.1143484115600586,
      "learning_rate": 7.376892124928148e-06,
      "loss": 0.30631084442138673,
      "step": 3875
    },
    {
      "epoch": 2.0046571798188872,
      "loss_anomaly": 0.014530245214700699,
      "loss_cls": 3.6709079722641036e-05,
      "loss_mitre": 0.00018194659787695855,
      "loss_ner": 0.005164771340787411,
      "step": 3875
    },
    {
      "epoch": 2.0046571798188872,
      "loss_anomaly": 0.008557276800274849,
      "loss_cls": 7.766803173581138e-05,
      "loss_mitre": 0.000601294101215899,
      "loss_ner": 0.042008839547634125,
      "step": 3875
    },
    {
      "epoch": 2.017593790426908,
      "grad_norm": 1.3338024616241455,
      "learning_rate": 7.281088331097912e-06,
      "loss": 0.2758142280578613,
      "step": 3900
    },
    {
      "epoch": 2.017593790426908,
      "loss_anomaly": 0.00811825692653656,
      "loss_cls": 2.6154320948990062e-05,
      "loss_mitre": 0.00029385232483036816,
      "loss_ner": 0.007923233322799206,
      "step": 3900
    },
    {
      "epoch": 2.017593790426908,
      "loss_anomaly": 0.020626258105039597,
      "loss_cls": 0.007167462725192308,
      "loss_mitre": 0.0003667330020107329,
      "loss_ner": 0.028289571404457092,
      "step": 3900
    },
    {
      "epoch": 2.030530401034929,
      "grad_norm": 1.8036752939224243,
      "learning_rate": 7.185284537267676e-06,
      "loss": 0.1414434242248535,
      "step": 3925
    },
    {
      "epoch": 2.030530401034929,
      "loss_anomaly": 0.010629957541823387,
      "loss_cls": 0.0001564385020174086,
      "loss_mitre": 0.0002841159584932029,
      "loss_ner": 0.003898436203598976,
      "step": 3925
    },
    {
      "epoch": 2.030530401034929,
      "loss_anomaly": 0.005633784923702478,
      "loss_cls": 3.197791374986991e-05,
      "loss_mitre": 1.7922440767288208,
      "loss_ner": 0.0012041209265589714,
      "step": 3925
    },
    {
      "epoch": 2.0434670116429494,
      "grad_norm": 0.4039155840873718,
      "learning_rate": 7.089480743437441e-06,
      "loss": 0.3203788375854492,
      "step": 3950
    },
    {
      "epoch": 2.0434670116429494,
      "loss_anomaly": 0.007605663500726223,
      "loss_cls": 4.171342879999429e-05,
      "loss_mitre": 0.0003168498515151441,
      "loss_ner": 0.0027098972350358963,
      "step": 3950
    },
    {
      "epoch": 2.0434670116429494,
      "loss_anomaly": 0.016289951279759407,
      "loss_cls": 3.145293521811254e-05,
      "loss_mitre": 0.00033561509917490184,
      "loss_ner": 0.00219352962449193,
      "step": 3950
    },
    {
      "epoch": 2.0564036222509703,
      "grad_norm": 0.9069077968597412,
      "learning_rate": 6.9936769496072055e-06,
      "loss": 0.14510302543640136,
      "step": 3975
    },
    {
      "epoch": 2.0564036222509703,
      "loss_anomaly": 0.00837149377912283,
      "loss_cls": 2.058164136542473e-05,
      "loss_mitre": 1.1937812566757202,
      "loss_ner": 0.010055807419121265,
      "step": 3975
    },
    {
      "epoch": 2.0564036222509703,
      "loss_anomaly": 0.006204293109476566,
      "loss_cls": 5.2155741286696866e-05,
      "loss_mitre": 0.8760230541229248,
      "loss_ner": 0.004079723730683327,
      "step": 3975
    },
    {
      "epoch": 2.069340232858991,
      "grad_norm": 1.2868099212646484,
      "learning_rate": 6.8978731557769695e-06,
      "loss": 0.2508967399597168,
      "step": 4000
    },
    {
      "epoch": 2.069340232858991,
      "eval_anomaly_mae": 0.06127451726669756,
      "eval_anomaly_mse": 0.006398145601838627,
      "eval_anomaly_n_eval": 27475.0,
      "eval_cls_accuracy": 0.9982469038059153,
      "eval_cls_auprc": 0.9995506916181314,
      "eval_cls_f1": 0.9937537779568809,
      "eval_cls_n_eval": 17683.0,
      "eval_cls_precision": 0.9907593411008437,
      "eval_cls_recall": 0.9967663702506063,
      "eval_loss": 0.05204245448112488,
      "eval_mitre_accuracy": 0.962847492855287,
      "eval_mitre_macro_f1": 0.15050848564171734,
      "eval_mitre_n_eval": 3849.0,
      "eval_ner_n_tokens": 666658.0,
      "eval_ner_token_acc": 0.9964059532773926,
      "eval_runtime": 68.8148,
      "eval_samples_per_second": 399.26,
      "eval_steps_per_second": 3.124,
      "step": 4000
    },
    {
      "epoch": 2.069340232858991,
      "loss_anomaly": 0.008320493623614311,
      "loss_cls": 5.5816592066548765e-05,
      "loss_mitre": 0.00027772202156484127,
      "loss_ner": 0.004044220317155123,
      "step": 4000
    },
    {
      "epoch": 2.069340232858991,
      "loss_anomaly": 0.01316654309630394,
      "loss_cls": 6.82731915730983e-05,
      "loss_mitre": 0.00020346511155366898,
      "loss_ner": 0.021167371422052383,
      "step": 4000
    },
    {
      "epoch": 2.0822768434670116,
      "grad_norm": 0.4941268265247345,
      "learning_rate": 6.802069361946734e-06,
      "loss": 0.26878644943237306,
      "step": 4025
    },
    {
      "epoch": 2.0822768434670116,
      "loss_anomaly": 0.013361909426748753,
      "loss_cls": 0.0009903309401124716,
      "loss_mitre": 0.00044405151857063174,
      "loss_ner": 0.009320746175944805,
      "step": 4025
    },
    {
      "epoch": 2.0822768434670116,
      "loss_anomaly": 0.011931534856557846,
      "loss_cls": 0.0010136328637599945,
      "loss_mitre": 0.0005698399618268013,
      "loss_ner": 0.004955674521625042,
      "step": 4025
    },
    {
      "epoch": 2.0952134540750325,
      "grad_norm": 0.7527535557746887,
      "learning_rate": 6.706265568116498e-06,
      "loss": 0.19819692611694337,
      "step": 4050
    },
    {
      "epoch": 2.0952134540750325,
      "loss_anomaly": 0.008233550935983658,
      "loss_cls": 7.985815318534151e-05,
      "loss_mitre": 1.3625037670135498,
      "loss_ner": 0.0035238030832260847,
      "step": 4050
    },
    {
      "epoch": 2.0952134540750325,
      "loss_anomaly": 0.013475269079208374,
      "loss_cls": 2.9816348614986055e-05,
      "loss_mitre": 1.0345561504364014,
      "loss_ner": 0.019785651937127113,
      "step": 4050
    },
    {
      "epoch": 2.108150064683053,
      "grad_norm": 1.974482536315918,
      "learning_rate": 6.610461774286262e-06,
      "loss": 0.2615984916687012,
      "step": 4075
    },
    {
      "epoch": 2.108150064683053,
      "loss_anomaly": 0.01001359149813652,
      "loss_cls": 0.00020026958372909576,
      "loss_mitre": 0.0003851601795759052,
      "loss_ner": 0.028509479016065598,
      "step": 4075
    },
    {
      "epoch": 2.108150064683053,
      "loss_anomaly": 0.013581177219748497,
      "loss_cls": 1.6998355931718834e-05,
      "loss_mitre": 0.0004642099083866924,
      "loss_ner": 0.009371397085487843,
      "step": 4075
    },
    {
      "epoch": 2.1210866752910738,
      "grad_norm": 1.6544978618621826,
      "learning_rate": 6.5146579804560266e-06,
      "loss": 0.2452238082885742,
      "step": 4100
    },
    {
      "epoch": 2.1210866752910738,
      "loss_anomaly": 0.008918270468711853,
      "loss_cls": 8.90135852387175e-05,
      "loss_mitre": 1.0531771183013916,
      "loss_ner": 0.014068137854337692,
      "step": 4100
    },
    {
      "epoch": 2.1210866752910738,
      "loss_anomaly": 0.01156361773610115,
      "loss_cls": 0.0009197311010211706,
      "loss_mitre": 0.9506100416183472,
      "loss_ner": 0.008343671448528767,
      "step": 4100
    },
    {
      "epoch": 2.1340232858990946,
      "grad_norm": 2.7254319190979004,
      "learning_rate": 6.4188541866257914e-06,
      "loss": 0.2847759437561035,
      "step": 4125
    },
    {
      "epoch": 2.1340232858990946,
      "loss_anomaly": 0.015419462695717812,
      "loss_cls": 2.4481003492837772e-05,
      "loss_mitre": 0.4545666575431824,
      "loss_ner": 0.009143486618995667,
      "step": 4125
    },
    {
      "epoch": 2.1340232858990946,
      "loss_anomaly": 0.015284527093172073,
      "loss_cls": 3.61297934432514e-05,
      "loss_mitre": 0.7253503203392029,
      "loss_ner": 0.010132042691111565,
      "step": 4125
    },
    {
      "epoch": 2.146959896507115,
      "grad_norm": 0.8793251514434814,
      "learning_rate": 6.3230503927955555e-06,
      "loss": 0.2482437515258789,
      "step": 4150
    },
    {
      "epoch": 2.146959896507115,
      "loss_anomaly": 0.012415017932653427,
      "loss_cls": 0.00015091794193722308,
      "loss_mitre": 0.00037797275581397116,
      "loss_ner": 0.0013250692281872034,
      "step": 4150
    },
    {
      "epoch": 2.146959896507115,
      "loss_anomaly": 0.009901116602122784,
      "loss_cls": 0.000154787689098157,
      "loss_mitre": 0.802198052406311,
      "loss_ner": 0.004892283584922552,
      "step": 4150
    },
    {
      "epoch": 2.159896507115136,
      "grad_norm": 4.042051315307617,
      "learning_rate": 6.2272465989653195e-06,
      "loss": 0.30551069259643554,
      "step": 4175
    },
    {
      "epoch": 2.159896507115136,
      "loss_anomaly": 0.014676053076982498,
      "loss_cls": 3.72290778614115e-05,
      "loss_mitre": 0.7631111145019531,
      "loss_ner": 0.019290173426270485,
      "step": 4175
    },
    {
      "epoch": 2.159896507115136,
      "loss_anomaly": 0.00494834128767252,
      "loss_cls": 1.4669924894405995e-05,
      "loss_mitre": 0.00021722732344642282,
      "loss_ner": 0.004895164165645838,
      "step": 4175
    },
    {
      "epoch": 2.1728331177231563,
      "grad_norm": 1.7447869777679443,
      "learning_rate": 6.1314428051350836e-06,
      "loss": 0.2973886871337891,
      "step": 4200
    },
    {
      "epoch": 2.1728331177231563,
      "loss_anomaly": 0.011754093691706657,
      "loss_cls": 9.693633910501376e-05,
      "loss_mitre": 0.0003550758701749146,
      "loss_ner": 0.008211388252675533,
      "step": 4200
    },
    {
      "epoch": 2.1728331177231563,
      "loss_anomaly": 0.02241530641913414,
      "loss_cls": 3.8131929613882676e-05,
      "loss_mitre": 0.0003213226154912263,
      "loss_ner": 0.013914806768298149,
      "step": 4200
    },
    {
      "epoch": 2.185769728331177,
      "grad_norm": 2.7444612979888916,
      "learning_rate": 6.035639011304848e-06,
      "loss": 0.22508052825927735,
      "step": 4225
    },
    {
      "epoch": 2.185769728331177,
      "loss_anomaly": 0.011935844086110592,
      "loss_cls": 2.7344509362592362e-05,
      "loss_mitre": 0.0002479364047758281,
      "loss_ner": 0.004520807880908251,
      "step": 4225
    },
    {
      "epoch": 2.185769728331177,
      "loss_anomaly": 0.009402118623256683,
      "loss_cls": 2.9666731279576197e-05,
      "loss_mitre": 0.0003125301154796034,
      "loss_ner": 0.013685286045074463,
      "step": 4225
    },
    {
      "epoch": 2.198706338939198,
      "grad_norm": 0.7307819128036499,
      "learning_rate": 5.9398352174746125e-06,
      "loss": 0.24998779296875,
      "step": 4250
    },
    {
      "epoch": 2.198706338939198,
      "loss_anomaly": 0.00730071309953928,
      "loss_cls": 4.985220220987685e-05,
      "loss_mitre": 0.0002765592944342643,
      "loss_ner": 0.0322604663670063,
      "step": 4250
    },
    {
      "epoch": 2.198706338939198,
      "loss_anomaly": 0.02427421137690544,
      "loss_cls": 5.4206444474402815e-05,
      "loss_mitre": 0.00020796587341465056,
      "loss_ner": 0.009407813660800457,
      "step": 4250
    },
    {
      "epoch": 2.2116429495472185,
      "grad_norm": 2.4722394943237305,
      "learning_rate": 5.8440314236443765e-06,
      "loss": 0.24115915298461915,
      "step": 4275
    },
    {
      "epoch": 2.2116429495472185,
      "loss_anomaly": 0.01095388364046812,
      "loss_cls": 1.7167581972898915e-05,
      "loss_mitre": 0.00027304532704874873,
      "loss_ner": 0.011305732652544975,
      "step": 4275
    },
    {
      "epoch": 2.2116429495472185,
      "loss_anomaly": 0.007901731878519058,
      "loss_cls": 2.313071490789298e-05,
      "loss_mitre": 0.7151867747306824,
      "loss_ner": 0.0020167327020317316,
      "step": 4275
    },
    {
      "epoch": 2.2245795601552394,
      "grad_norm": 1.0491037368774414,
      "learning_rate": 5.7482276298141414e-06,
      "loss": 0.25906829833984374,
      "step": 4300
    },
    {
      "epoch": 2.2245795601552394,
      "loss_anomaly": 0.009591039270162582,
      "loss_cls": 6.342156848404557e-05,
      "loss_mitre": 0.0005639753071591258,
      "loss_ner": 0.004083022475242615,
      "step": 4300
    },
    {
      "epoch": 2.2245795601552394,
      "loss_anomaly": 0.014324353076517582,
      "loss_cls": 6.273038889048621e-05,
      "loss_mitre": 0.0002766890684142709,
      "loss_ner": 0.015092900954186916,
      "step": 4300
    },
    {
      "epoch": 2.23751617076326,
      "grad_norm": 0.782667338848114,
      "learning_rate": 5.6524238359839055e-06,
      "loss": 0.22257757186889648,
      "step": 4325
    },
    {
      "epoch": 2.23751617076326,
      "loss_anomaly": 0.014229313470423222,
      "loss_cls": 0.0024453033693134785,
      "loss_mitre": 0.00020200887229293585,
      "loss_ner": 0.019246965646743774,
      "step": 4325
    },
    {
      "epoch": 2.23751617076326,
      "loss_anomaly": 0.01569352298974991,
      "loss_cls": 5.9996466006850824e-05,
      "loss_mitre": 0.0003446988121140748,
      "loss_ner": 0.011662724427878857,
      "step": 4325
    },
    {
      "epoch": 2.2504527813712807,
      "grad_norm": 0.8641570210456848,
      "learning_rate": 5.5566200421536695e-06,
      "loss": 0.32492546081542967,
      "step": 4350
    },
    {
      "epoch": 2.2504527813712807,
      "loss_anomaly": 0.008642058819532394,
      "loss_cls": 0.0001295106194447726,
      "loss_mitre": 0.9009512066841125,
      "loss_ner": 0.007780010811984539,
      "step": 4350
    },
    {
      "epoch": 2.2504527813712807,
      "loss_anomaly": 0.01018516719341278,
      "loss_cls": 5.694009087164886e-05,
      "loss_mitre": 0.00025365909095853567,
      "loss_ner": 0.006260272115468979,
      "step": 4350
    },
    {
      "epoch": 2.2633893919793016,
      "grad_norm": 0.5171841382980347,
      "learning_rate": 5.4608162483234335e-06,
      "loss": 0.29791507720947263,
      "step": 4375
    },
    {
      "epoch": 2.2633893919793016,
      "loss_anomaly": 0.008683836087584496,
      "loss_cls": 0.00016564899124205112,
      "loss_mitre": 0.0005062950658611953,
      "loss_ner": 0.006627688184380531,
      "step": 4375
    },
    {
      "epoch": 2.2633893919793016,
      "loss_anomaly": 0.010567162185907364,
      "loss_cls": 7.285261381184682e-05,
      "loss_mitre": 0.00031890772515907884,
      "loss_ner": 0.02301815152168274,
      "step": 4375
    },
    {
      "epoch": 2.276326002587322,
      "grad_norm": 1.7277240753173828,
      "learning_rate": 5.3650124544931984e-06,
      "loss": 0.25072494506835935,
      "step": 4400
    },
    {
      "epoch": 2.276326002587322,
      "loss_anomaly": 0.0085634496062994,
      "loss_cls": 9.352247434435412e-05,
      "loss_mitre": 0.00037191357114352286,
      "loss_ner": 0.010336395353078842,
      "step": 4400
    },
    {
      "epoch": 2.276326002587322,
      "loss_anomaly": 0.014634187333285809,
      "loss_cls": 0.00045117284753359854,
      "loss_mitre": 0.00034102165955118835,
      "loss_ner": 0.003089477075263858,
      "step": 4400
    },
    {
      "epoch": 2.289262613195343,
      "grad_norm": 0.9978009462356567,
      "learning_rate": 5.2692086606629625e-06,
      "loss": 0.19572723388671875,
      "step": 4425
    },
    {
      "epoch": 2.289262613195343,
      "loss_anomaly": 0.010576414875686169,
      "loss_cls": 4.297230771044269e-05,
      "loss_mitre": 0.00035436011967249215,
      "loss_ner": 0.008432015776634216,
      "step": 4425
    },
    {
      "epoch": 2.289262613195343,
      "loss_anomaly": 0.00568235432729125,
      "loss_cls": 2.3053580662235618e-05,
      "loss_mitre": 0.0002977067488245666,
      "loss_ner": 0.01351890154182911,
      "step": 4425
    },
    {
      "epoch": 2.3021992238033633,
      "grad_norm": 1.913575291633606,
      "learning_rate": 5.173404866832727e-06,
      "loss": 0.2688895225524902,
      "step": 4450
    },
    {
      "epoch": 2.3021992238033633,
      "loss_anomaly": 0.010919969528913498,
      "loss_cls": 0.0002230466780019924,
      "loss_mitre": 0.00020754276192747056,
      "loss_ner": 0.0016151710879057646,
      "step": 4450
    },
    {
      "epoch": 2.3021992238033633,
      "loss_anomaly": 0.009628540836274624,
      "loss_cls": 3.211016155546531e-05,
      "loss_mitre": 0.0002325730019947514,
      "loss_ner": 0.002014349214732647,
      "step": 4450
    },
    {
      "epoch": 2.315135834411384,
      "grad_norm": 0.6013554334640503,
      "learning_rate": 5.077601073002491e-06,
      "loss": 0.23239572525024413,
      "step": 4475
    },
    {
      "epoch": 2.315135834411384,
      "loss_anomaly": 0.02044932171702385,
      "loss_cls": 4.909139533992857e-05,
      "loss_mitre": 0.00017131958156824112,
      "loss_ner": 0.009649614803493023,
      "step": 4475
    },
    {
      "epoch": 2.315135834411384,
      "loss_anomaly": 0.011505111120641232,
      "loss_cls": 0.0008714436553418636,
      "loss_mitre": 0.558799684047699,
      "loss_ner": 0.0021230652928352356,
      "step": 4475
    },
    {
      "epoch": 2.328072445019405,
      "grad_norm": 1.6690176725387573,
      "learning_rate": 4.9817972791722554e-06,
      "loss": 0.29621395111083987,
      "step": 4500
    },
    {
      "epoch": 2.328072445019405,
      "eval_anomaly_mae": 0.05803012833341167,
      "eval_anomaly_mse": 0.005735053512765403,
      "eval_anomaly_n_eval": 27475.0,
      "eval_cls_accuracy": 0.9985296612565742,
      "eval_cls_auprc": 0.999697336528059,
      "eval_cls_f1": 0.9947538337368845,
      "eval_cls_n_eval": 17683.0,
      "eval_cls_precision": 0.9931506849315068,
      "eval_cls_recall": 0.9963621665319321,
      "eval_loss": 0.052291061729192734,
      "eval_mitre_accuracy": 0.962847492855287,
      "eval_mitre_macro_f1": 0.1542247445464511,
      "eval_mitre_n_eval": 3849.0,
      "eval_ner_n_tokens": 666658.0,
      "eval_ner_token_acc": 0.9961074493968421,
      "eval_runtime": 68.8799,
      "eval_samples_per_second": 398.883,
      "eval_steps_per_second": 3.121,
      "step": 4500
    },
    {
      "epoch": 2.328072445019405,
      "loss_anomaly": 0.010790541768074036,
      "loss_cls": 0.00012380471162032336,
      "loss_mitre": 0.00018757471116259694,
      "loss_ner": 0.004269399214535952,
      "step": 4500
    },
    {
      "epoch": 2.328072445019405,
      "loss_anomaly": 0.00940575823187828,
      "loss_cls": 3.897739588865079e-05,
      "loss_mitre": 0.00022414705017581582,
      "loss_ner": 0.0019454988650977612,
      "step": 4500
    },
    {
      "epoch": 2.3410090556274255,
      "grad_norm": 1.914414644241333,
      "learning_rate": 4.8859934853420195e-06,
      "loss": 0.23826648712158202,
      "step": 4525
    },
    {
      "epoch": 2.3410090556274255,
      "loss_anomaly": 0.013514300808310509,
      "loss_cls": 6.731702160323039e-05,
      "loss_mitre": 0.0003035802219528705,
      "loss_ner": 0.005692583508789539,
      "step": 4525
    },
    {
      "epoch": 2.3410090556274255,
      "loss_anomaly": 0.016534090042114258,
      "loss_cls": 3.888470018864609e-05,
      "loss_mitre": 0.00032683435711078346,
      "loss_ner": 0.012986347079277039,
      "step": 4525
    },
    {
      "epoch": 2.3539456662354463,
      "grad_norm": 0.5982169508934021,
      "learning_rate": 4.790189691511784e-06,
      "loss": 0.24550437927246094,
      "step": 4550
    },
    {
      "epoch": 2.3539456662354463,
      "loss_anomaly": 0.00862367358058691,
      "loss_cls": 4.1718387365108356e-05,
      "loss_mitre": 0.5650574564933777,
      "loss_ner": 0.01365505252033472,
      "step": 4550
    },
    {
      "epoch": 2.3539456662354463,
      "loss_anomaly": 0.010092079639434814,
      "loss_cls": 1.8169983377447352e-05,
      "loss_mitre": 0.0003193278389517218,
      "loss_ner": 0.001256322953850031,
      "step": 4550
    },
    {
      "epoch": 2.366882276843467,
      "grad_norm": 0.7107054591178894,
      "learning_rate": 4.694385897681548e-06,
      "loss": 0.20273223876953125,
      "step": 4575
    },
    {
      "epoch": 2.366882276843467,
      "loss_anomaly": 0.009132372215390205,
      "loss_cls": 2.2383670511771925e-05,
      "loss_mitre": 0.00019276422972325236,
      "loss_ner": 0.002332808682695031,
      "step": 4575
    },
    {
      "epoch": 2.366882276843467,
      "loss_anomaly": 0.009097011759877205,
      "loss_cls": 8.38629639474675e-05,
      "loss_mitre": 0.0002482338750269264,
      "loss_ner": 0.07502219825983047,
      "step": 4575
    },
    {
      "epoch": 2.3798188874514876,
      "grad_norm": 0.867290198802948,
      "learning_rate": 4.5985821038513125e-06,
      "loss": 0.3206205368041992,
      "step": 4600
    },
    {
      "epoch": 2.3798188874514876,
      "loss_anomaly": 0.017011824995279312,
      "loss_cls": 1.7224707335117273e-05,
      "loss_mitre": 0.0003030187217518687,
      "loss_ner": 0.0077579994685947895,
      "step": 4600
    },
    {
      "epoch": 2.3798188874514876,
      "loss_anomaly": 0.007355109788477421,
      "loss_cls": 3.43846004398074e-05,
      "loss_mitre": 0.0002561837318353355,
      "loss_ner": 0.004326194059103727,
      "step": 4600
    },
    {
      "epoch": 2.3927554980595085,
      "grad_norm": 0.7451397180557251,
      "learning_rate": 4.502778310021077e-06,
      "loss": 0.2721388244628906,
      "step": 4625
    },
    {
      "epoch": 2.3927554980595085,
      "loss_anomaly": 0.015124676749110222,
      "loss_cls": 3.089640085818246e-05,
      "loss_mitre": 0.580135703086853,
      "loss_ner": 0.005392693914473057,
      "step": 4625
    },
    {
      "epoch": 2.3927554980595085,
      "loss_anomaly": 0.014807621017098427,
      "loss_cls": 1.9908939066226594e-05,
      "loss_mitre": 0.00023214827524498105,
      "loss_ner": 0.00117432267870754,
      "step": 4625
    },
    {
      "epoch": 2.405692108667529,
      "grad_norm": 0.7073512077331543,
      "learning_rate": 4.406974516190841e-06,
      "loss": 0.22024505615234374,
      "step": 4650
    },
    {
      "epoch": 2.405692108667529,
      "loss_anomaly": 0.006994477938860655,
      "loss_cls": 3.84441482310649e-05,
      "loss_mitre": 0.6854382157325745,
      "loss_ner": 0.010501325130462646,
      "step": 4650
    },
    {
      "epoch": 2.405692108667529,
      "loss_anomaly": 0.010296674445271492,
      "loss_cls": 0.0006492320680990815,
      "loss_mitre": 0.00015519658336415887,
      "loss_ner": 0.012113397940993309,
      "step": 4650
    },
    {
      "epoch": 2.41862871927555,
      "grad_norm": 0.8409565687179565,
      "learning_rate": 4.3111707223606054e-06,
      "loss": 0.17784936904907225,
      "step": 4675
    },
    {
      "epoch": 2.41862871927555,
      "loss_anomaly": 0.009697743691504002,
      "loss_cls": 0.0003591996501199901,
      "loss_mitre": 0.00027057857369072735,
      "loss_ner": 0.010936555452644825,
      "step": 4675
    },
    {
      "epoch": 2.41862871927555,
      "loss_anomaly": 0.01389912236481905,
      "loss_cls": 5.1508475735317916e-05,
      "loss_mitre": 0.00026033088215626776,
      "loss_ner": 0.003531505586579442,
      "step": 4675
    },
    {
      "epoch": 2.4315653298835707,
      "grad_norm": 2.0643138885498047,
      "learning_rate": 4.21536692853037e-06,
      "loss": 0.24683675765991211,
      "step": 4700
    },
    {
      "epoch": 2.4315653298835707,
      "loss_anomaly": 0.00871001835912466,
      "loss_cls": 4.418119715410285e-05,
      "loss_mitre": 0.00013339644647203386,
      "loss_ner": 0.026179444044828415,
      "step": 4700
    },
    {
      "epoch": 2.4315653298835707,
      "loss_anomaly": 0.011772315949201584,
      "loss_cls": 0.00013093982124701142,
      "loss_mitre": 0.504274308681488,
      "loss_ner": 0.010632442310452461,
      "step": 4700
    },
    {
      "epoch": 2.444501940491591,
      "grad_norm": 1.7393274307250977,
      "learning_rate": 4.119563134700134e-06,
      "loss": 0.2740378570556641,
      "step": 4725
    },
    {
      "epoch": 2.444501940491591,
      "loss_anomaly": 0.010508459992706776,
      "loss_cls": 0.00016458086611237377,
      "loss_mitre": 0.0004838503082282841,
      "loss_ner": 0.007062310352921486,
      "step": 4725
    },
    {
      "epoch": 2.444501940491591,
      "loss_anomaly": 0.010079092346131802,
      "loss_cls": 0.00020289007807150483,
      "loss_mitre": 2.2555224895477295,
      "loss_ner": 0.0014108462492004037,
      "step": 4725
    },
    {
      "epoch": 2.457438551099612,
      "grad_norm": 0.6491615772247314,
      "learning_rate": 4.023759340869899e-06,
      "loss": 0.32904457092285155,
      "step": 4750
    },
    {
      "epoch": 2.457438551099612,
      "loss_anomaly": 0.011585722677409649,
      "loss_cls": 2.3262293325387873e-05,
      "loss_mitre": 0.00018831947818398476,
      "loss_ner": 0.009024116210639477,
      "step": 4750
    },
    {
      "epoch": 2.457438551099612,
      "loss_anomaly": 0.006712028756737709,
      "loss_cls": 3.0222256100387312e-05,
      "loss_mitre": 0.0002316298778168857,
      "loss_ner": 0.026120491325855255,
      "step": 4750
    },
    {
      "epoch": 2.470375161707633,
      "grad_norm": 0.4990725815296173,
      "learning_rate": 3.927955547039663e-06,
      "loss": 0.13322944641113282,
      "step": 4775
    },
    {
      "epoch": 2.470375161707633,
      "loss_anomaly": 0.015867847949266434,
      "loss_cls": 0.000102093858004082,
      "loss_mitre": 0.0003824292798526585,
      "loss_ner": 0.015285702422261238,
      "step": 4775
    },
    {
      "epoch": 2.470375161707633,
      "loss_anomaly": 0.015864912420511246,
      "loss_cls": 4.27213562943507e-05,
      "loss_mitre": 0.3888065814971924,
      "loss_ner": 0.012085996568202972,
      "step": 4775
    },
    {
      "epoch": 2.4833117723156533,
      "grad_norm": 1.2361217737197876,
      "learning_rate": 3.832151753209427e-06,
      "loss": 0.1715012550354004,
      "step": 4800
    },
    {
      "epoch": 2.4833117723156533,
      "loss_anomaly": 0.014173504896461964,
      "loss_cls": 3.3614705898799e-05,
      "loss_mitre": 0.00020319409668445587,
      "loss_ner": 0.03238772973418236,
      "step": 4800
    },
    {
      "epoch": 2.4833117723156533,
      "loss_anomaly": 0.005606211721897125,
      "loss_cls": 4.3287855078233406e-05,
      "loss_mitre": 0.795783519744873,
      "loss_ner": 0.0006611908902414143,
      "step": 4800
    },
    {
      "epoch": 2.496248382923674,
      "grad_norm": 1.3072915077209473,
      "learning_rate": 3.736347959379192e-06,
      "loss": 0.32708507537841797,
      "step": 4825
    },
    {
      "epoch": 2.496248382923674,
      "loss_anomaly": 0.011301710270345211,
      "loss_cls": 0.00015663215890526772,
      "loss_mitre": 0.6441166996955872,
      "loss_ner": 0.01618947833776474,
      "step": 4825
    },
    {
      "epoch": 2.496248382923674,
      "loss_anomaly": 0.012924677692353725,
      "loss_cls": 0.00010502592340344563,
      "loss_mitre": 0.00023402234364766628,
      "loss_ner": 0.03210625424981117,
      "step": 4825
    },
    {
      "epoch": 2.5091849935316946,
      "grad_norm": 1.5723249912261963,
      "learning_rate": 3.640544165548956e-06,
      "loss": 0.28765043258666995,
      "step": 4850
    },
    {
      "epoch": 2.5091849935316946,
      "loss_anomaly": 0.015032323077321053,
      "loss_cls": 3.773265052586794e-05,
      "loss_mitre": 0.00030894370866008103,
      "loss_ner": 0.0136774443089962,
      "step": 4850
    },
    {
      "epoch": 2.5091849935316946,
      "loss_anomaly": 0.005834552459418774,
      "loss_cls": 6.606243550777435e-05,
      "loss_mitre": 0.00034626605338416994,
      "loss_ner": 0.049564383924007416,
      "step": 4850
    },
    {
      "epoch": 2.5221216041397154,
      "grad_norm": 0.5278390049934387,
      "learning_rate": 3.5447403717187203e-06,
      "loss": 0.17757431030273438,
      "step": 4875
    },
    {
      "epoch": 2.5221216041397154,
      "loss_anomaly": 0.007380392402410507,
      "loss_cls": 3.2555726647842675e-05,
      "loss_mitre": 0.00018125421775039285,
      "loss_ner": 0.05160900205373764,
      "step": 4875
    },
    {
      "epoch": 2.5221216041397154,
      "loss_anomaly": 0.007491810247302055,
      "loss_cls": 2.7131753085996024e-05,
      "loss_mitre": 1.6812635660171509,
      "loss_ner": 0.005006161518394947,
      "step": 4875
    },
    {
      "epoch": 2.5350582147477363,
      "grad_norm": 0.7485597729682922,
      "learning_rate": 3.4489365778884848e-06,
      "loss": 0.28791305541992185,
      "step": 4900
    },
    {
      "epoch": 2.5350582147477363,
      "loss_anomaly": 0.009711001068353653,
      "loss_cls": 5.4296309826895595e-05,
      "loss_mitre": 0.0001873108121799305,
      "loss_ner": 0.015396644361317158,
      "step": 4900
    },
    {
      "epoch": 2.5350582147477363,
      "loss_anomaly": 0.0066732848063111305,
      "loss_cls": 1.9774768588831648e-05,
      "loss_mitre": 1.0985652208328247,
      "loss_ner": 0.0015381318517029285,
      "step": 4900
    },
    {
      "epoch": 2.5479948253557567,
      "grad_norm": 1.8604272603988647,
      "learning_rate": 3.353132784058249e-06,
      "loss": 0.29027772903442384,
      "step": 4925
    },
    {
      "epoch": 2.5479948253557567,
      "loss_anomaly": 0.01007651537656784,
      "loss_cls": 0.00011728587560355663,
      "loss_mitre": 1.1814435720443726,
      "loss_ner": 0.0029878567438572645,
      "step": 4925
    },
    {
      "epoch": 2.5479948253557567,
      "loss_anomaly": 0.007112786639481783,
      "loss_cls": 0.0009369925828650594,
      "loss_mitre": 0.7003740072250366,
      "loss_ner": 0.02510830946266651,
      "step": 4925
    },
    {
      "epoch": 2.5609314359637776,
      "grad_norm": 1.630570411682129,
      "learning_rate": 3.2573289902280133e-06,
      "loss": 0.34013519287109373,
      "step": 4950
    },
    {
      "epoch": 2.5609314359637776,
      "loss_anomaly": 0.014175206422805786,
      "loss_cls": 0.00022577946947421879,
      "loss_mitre": 0.00018096319399774075,
      "loss_ner": 0.0014615971595048904,
      "step": 4950
    },
    {
      "epoch": 2.5609314359637776,
      "loss_anomaly": 0.005819328129291534,
      "loss_cls": 0.000692427740432322,
      "loss_mitre": 0.0005952088395133615,
      "loss_ner": 0.008300238288939,
      "step": 4950
    },
    {
      "epoch": 2.573868046571798,
      "grad_norm": 0.8284881711006165,
      "learning_rate": 3.1615251963977777e-06,
      "loss": 0.2583424949645996,
      "step": 4975
    },
    {
      "epoch": 2.573868046571798,
      "loss_anomaly": 0.009614307433366776,
      "loss_cls": 2.7847088858834468e-05,
      "loss_mitre": 0.00024705223040655255,
      "loss_ner": 0.00304098566994071,
      "step": 4975
    },
    {
      "epoch": 2.573868046571798,
      "loss_anomaly": 0.013317020609974861,
      "loss_cls": 2.1513213141588494e-05,
      "loss_mitre": 0.00021577050210908055,
      "loss_ner": 0.0011342058423906565,
      "step": 4975
    },
    {
      "epoch": 2.586804657179819,
      "grad_norm": 1.594215750694275,
      "learning_rate": 3.0657214025675418e-06,
      "loss": 0.2293134880065918,
      "step": 5000
    },
    {
      "epoch": 2.586804657179819,
      "eval_anomaly_mae": 0.05762500512998829,
      "eval_anomaly_mse": 0.00577316675363941,
      "eval_anomaly_n_eval": 27475.0,
      "eval_cls_accuracy": 0.9962110501611717,
      "eval_cls_auprc": 0.9996413733171601,
      "eval_cls_f1": 0.9866240766620084,
      "eval_cls_n_eval": 17683.0,
      "eval_cls_precision": 0.9747534516765286,
      "eval_cls_recall": 0.9987873888439773,
      "eval_loss": 0.05282457545399666,
      "eval_mitre_accuracy": 0.962847492855287,
      "eval_mitre_macro_f1": 0.15515265651335824,
      "eval_mitre_n_eval": 3849.0,
      "eval_ner_n_tokens": 666658.0,
      "eval_ner_token_acc": 0.9963819529653886,
      "eval_runtime": 68.8524,
      "eval_samples_per_second": 399.042,
      "eval_steps_per_second": 3.123,
      "step": 5000
    },
    {
      "epoch": 2.586804657179819,
      "loss_anomaly": 0.014123639091849327,
      "loss_cls": 9.365836012875661e-05,
      "loss_mitre": 0.799057126045227,
      "loss_ner": 0.014721494168043137,
      "step": 5000
    },
    {
      "epoch": 2.586804657179819,
      "loss_anomaly": 0.0076553975231945515,
      "loss_cls": 0.00016330275684595108,
      "loss_mitre": 0.00027454772498458624,
      "loss_ner": 0.032210901379585266,
      "step": 5000
    },
    {
      "epoch": 2.5997412677878398,
      "grad_norm": 0.4491652846336365,
      "learning_rate": 2.9699176087373062e-06,
      "loss": 0.27346641540527344,
      "step": 5025
    },
    {
      "epoch": 2.5997412677878398,
      "loss_anomaly": 0.009846086613833904,
      "loss_cls": 7.922141230665147e-05,
      "loss_mitre": 0.8916264772415161,
      "loss_ner": 0.002172709908336401,
      "step": 5025
    },
    {
      "epoch": 2.5997412677878398,
      "loss_anomaly": 0.006744863465428352,
      "loss_cls": 2.6754458303912543e-05,
      "loss_mitre": 0.00031506316736340523,
      "loss_ner": 0.007226115092635155,
      "step": 5025
    },
    {
      "epoch": 2.61267787839586,
      "grad_norm": 0.8242404460906982,
      "learning_rate": 2.8741138149070707e-06,
      "loss": 0.18463075637817383,
      "step": 5050
    },
    {
      "epoch": 2.61267787839586,
      "loss_anomaly": 0.012214062735438347,
      "loss_cls": 4.280828215996735e-05,
      "loss_mitre": 0.0002945391461253166,
      "loss_ner": 0.0028660993557423353,
      "step": 5050
    },
    {
      "epoch": 2.61267787839586,
      "loss_anomaly": 0.006347620859742165,
      "loss_cls": 2.0680852685472928e-05,
      "loss_mitre": 0.8753023743629456,
      "loss_ner": 0.006965992506593466,
      "step": 5050
    },
    {
      "epoch": 2.625614489003881,
      "grad_norm": 0.586241602897644,
      "learning_rate": 2.7783100210768348e-06,
      "loss": 0.3965864944458008,
      "step": 5075
    },
    {
      "epoch": 2.625614489003881,
      "loss_anomaly": 0.009989984333515167,
      "loss_cls": 8.54308673297055e-05,
      "loss_mitre": 0.00027699500788003206,
      "loss_ner": 0.009061187505722046,
      "step": 5075
    },
    {
      "epoch": 2.625614489003881,
      "loss_anomaly": 0.007820794358849525,
      "loss_cls": 2.2355605324264616e-05,
      "loss_mitre": 0.00019417847215663642,
      "loss_ner": 0.014220799319446087,
      "step": 5075
    },
    {
      "epoch": 2.6385510996119015,
      "grad_norm": 0.8540490865707397,
      "learning_rate": 2.6825062272465992e-06,
      "loss": 0.3342641448974609,
      "step": 5100
    },
    {
      "epoch": 2.6385510996119015,
      "loss_anomaly": 0.008233914151787758,
      "loss_cls": 1.7062968254322186e-05,
      "loss_mitre": 1.321924090385437,
      "loss_ner": 0.0013339878059923649,
      "step": 5100
    },
    {
      "epoch": 2.6385510996119015,
      "loss_anomaly": 0.014277070760726929,
      "loss_cls": 0.002032305346801877,
      "loss_mitre": 0.0002420383389107883,
      "loss_ner": 0.013920888304710388,
      "step": 5100
    },
    {
      "epoch": 2.6514877102199224,
      "grad_norm": 0.6986236572265625,
      "learning_rate": 2.5867024334163637e-06,
      "loss": 0.2934440231323242,
      "step": 5125
    },
    {
      "epoch": 2.6514877102199224,
      "loss_anomaly": 0.0154969347640872,
      "loss_cls": 0.00056486512767151,
      "loss_mitre": 0.6184648871421814,
      "loss_ner": 0.0022235726937651634,
      "step": 5125
    },
    {
      "epoch": 2.6514877102199224,
      "loss_anomaly": 0.006998585071414709,
      "loss_cls": 0.001359469722956419,
      "loss_mitre": 0.00014196526899468154,
      "loss_ner": 0.007534559816122055,
      "step": 5125
    },
    {
      "epoch": 2.6644243208279432,
      "grad_norm": 0.5714066624641418,
      "learning_rate": 2.4908986395861277e-06,
      "loss": 0.25185964584350584,
      "step": 5150
    },
    {
      "epoch": 2.6644243208279432,
      "loss_anomaly": 0.00899103470146656,
      "loss_cls": 0.00012971556861884892,
      "loss_mitre": 0.0002184327458962798,
      "loss_ner": 0.01346038468182087,
      "step": 5150
    },
    {
      "epoch": 2.6644243208279432,
      "loss_anomaly": 0.016832886263728142,
      "loss_cls": 2.48481137532508e-05,
      "loss_mitre": 0.4635433554649353,
      "loss_ner": 0.004323191940784454,
      "step": 5150
    },
    {
      "epoch": 2.6773609314359637,
      "grad_norm": 0.2909940779209137,
      "learning_rate": 2.395094845755892e-06,
      "loss": 0.2783392524719238,
      "step": 5175
    },
    {
      "epoch": 2.6773609314359637,
      "loss_anomaly": 0.010287614539265633,
      "loss_cls": 1.43335091706831e-05,
      "loss_mitre": 0.44528040289878845,
      "loss_ner": 0.004506333265453577,
      "step": 5175
    },
    {
      "epoch": 2.6773609314359637,
      "loss_anomaly": 0.009350099600851536,
      "loss_cls": 3.336322333780117e-05,
      "loss_mitre": 1.0143870115280151,
      "loss_ner": 0.005959881469607353,
      "step": 5175
    },
    {
      "epoch": 2.6902975420439845,
      "grad_norm": 2.187404155731201,
      "learning_rate": 2.2992910519256562e-06,
      "loss": 0.270738410949707,
      "step": 5200
    },
    {
      "epoch": 2.6902975420439845,
      "loss_anomaly": 0.01114029623568058,
      "loss_cls": 3.0259863706305623e-05,
      "loss_mitre": 0.00022767388145439327,
      "loss_ner": 0.021200401708483696,
      "step": 5200
    },
    {
      "epoch": 2.6902975420439845,
      "loss_anomaly": 0.010370288975536823,
      "loss_cls": 2.6376661480753683e-05,
      "loss_mitre": 0.00018165900837630033,
      "loss_ner": 0.004314595367759466,
      "step": 5200
    },
    {
      "epoch": 2.703234152652005,
      "grad_norm": 2.057216167449951,
      "learning_rate": 2.2034872580954207e-06,
      "loss": 0.2463442039489746,
      "step": 5225
    },
    {
      "epoch": 2.703234152652005,
      "loss_anomaly": 0.009365225210785866,
      "loss_cls": 1.9027858797926456e-05,
      "loss_mitre": 0.6647012233734131,
      "loss_ner": 0.0029549505561590195,
      "step": 5225
    },
    {
      "epoch": 2.703234152652005,
      "loss_anomaly": 0.009595481678843498,
      "loss_cls": 0.0009715517517179251,
      "loss_mitre": 0.5955434441566467,
      "loss_ner": 0.009333120658993721,
      "step": 5225
    },
    {
      "epoch": 2.716170763260026,
      "grad_norm": 0.791507363319397,
      "learning_rate": 2.107683464265185e-06,
      "loss": 0.2523703193664551,
      "step": 5250
    },
    {
      "epoch": 2.716170763260026,
      "loss_anomaly": 0.008425502106547356,
      "loss_cls": 1.883971162897069e-05,
      "loss_mitre": 0.00020940389367751777,
      "loss_ner": 0.012234054505825043,
      "step": 5250
    },
    {
      "epoch": 2.716170763260026,
      "loss_anomaly": 0.008154422044754028,
      "loss_cls": 2.032423253695015e-05,
      "loss_mitre": 0.00022716510284226388,
      "loss_ner": 0.0031471371185034513,
      "step": 5250
    },
    {
      "epoch": 2.7291073738680467,
      "grad_norm": 1.399855613708496,
      "learning_rate": 2.0118796704349496e-06,
      "loss": 0.18280942916870116,
      "step": 5275
    },
    {
      "epoch": 2.7291073738680467,
      "loss_anomaly": 0.012044312432408333,
      "loss_cls": 9.979008609661832e-06,
      "loss_mitre": 0.00015265573165379465,
      "loss_ner": 0.010115692391991615,
      "step": 5275
    },
    {
      "epoch": 2.7291073738680467,
      "loss_anomaly": 0.012784391641616821,
      "loss_cls": 2.8773059966624714e-05,
      "loss_mitre": 0.00021209687110967934,
      "loss_ner": 0.008900034241378307,
      "step": 5275
    },
    {
      "epoch": 2.742043984476067,
      "grad_norm": 0.79322749376297,
      "learning_rate": 1.9160758766047137e-06,
      "loss": 0.30192766189575193,
      "step": 5300
    },
    {
      "epoch": 2.742043984476067,
      "loss_anomaly": 0.017987564206123352,
      "loss_cls": 2.0804980522370897e-05,
      "loss_mitre": 0.3239697515964508,
      "loss_ner": 0.0024679447524249554,
      "step": 5300
    },
    {
      "epoch": 2.742043984476067,
      "loss_anomaly": 0.010860880836844444,
      "loss_cls": 6.343478162307292e-05,
      "loss_mitre": 0.8190928101539612,
      "loss_ner": 0.003431122750043869,
      "step": 5300
    },
    {
      "epoch": 2.754980595084088,
      "grad_norm": 0.5965147018432617,
      "learning_rate": 1.820272082774478e-06,
      "loss": 0.19757602691650392,
      "step": 5325
    },
    {
      "epoch": 2.754980595084088,
      "loss_anomaly": 0.010920082218945026,
      "loss_cls": 0.00092275720089674,
      "loss_mitre": 0.0002123079902958125,
      "loss_ner": 0.01331285573542118,
      "step": 5325
    },
    {
      "epoch": 2.754980595084088,
      "loss_anomaly": 0.009001597762107849,
      "loss_cls": 2.7923717425437644e-05,
      "loss_mitre": 0.00019920153135899454,
      "loss_ner": 0.013484859839081764,
      "step": 5325
    },
    {
      "epoch": 2.7679172056921084,
      "grad_norm": 2.1338672637939453,
      "learning_rate": 1.7244682889442424e-06,
      "loss": 0.3471563720703125,
      "step": 5350
    },
    {
      "epoch": 2.7679172056921084,
      "loss_anomaly": 0.008374959230422974,
      "loss_cls": 0.00015298837388399988,
      "loss_mitre": 0.6713603734970093,
      "loss_ner": 0.006324976682662964,
      "step": 5350
    },
    {
      "epoch": 2.7679172056921084,
      "loss_anomaly": 0.013801923021674156,
      "loss_cls": 0.00033084210008382797,
      "loss_mitre": 0.00024313453468494117,
      "loss_ner": 0.01427267212420702,
      "step": 5350
    },
    {
      "epoch": 2.7808538163001293,
      "grad_norm": 1.5798860788345337,
      "learning_rate": 1.6286644951140066e-06,
      "loss": 0.2284281349182129,
      "step": 5375
    },
    {
      "epoch": 2.7808538163001293,
      "loss_anomaly": 0.014251915737986565,
      "loss_cls": 0.0004911569412797689,
      "loss_mitre": 0.00024227162066381425,
      "loss_ner": 0.002551252953708172,
      "step": 5375
    },
    {
      "epoch": 2.7808538163001293,
      "loss_anomaly": 0.007541059050709009,
      "loss_cls": 3.760322942980565e-05,
      "loss_mitre": 0.00020070653408765793,
      "loss_ner": 0.008264089934527874,
      "step": 5375
    },
    {
      "epoch": 2.79379042690815,
      "grad_norm": 0.8067485690116882,
      "learning_rate": 1.5328607012837709e-06,
      "loss": 0.19027471542358398,
      "step": 5400
    },
    {
      "epoch": 2.79379042690815,
      "loss_anomaly": 0.018382152542471886,
      "loss_cls": 1.6543761375942267e-05,
      "loss_mitre": 0.00019456323934718966,
      "loss_ner": 0.009598970413208008,
      "step": 5400
    },
    {
      "epoch": 2.79379042690815,
      "loss_anomaly": 0.015049546025693417,
      "loss_cls": 1.6827632862259634e-05,
      "loss_mitre": 0.5477402210235596,
      "loss_ner": 0.0015703728422522545,
      "step": 5400
    },
    {
      "epoch": 2.806727037516171,
      "grad_norm": 1.7402863502502441,
      "learning_rate": 1.4370569074535354e-06,
      "loss": 0.29348251342773435,
      "step": 5425
    },
    {
      "epoch": 2.806727037516171,
      "loss_anomaly": 0.009188555181026459,
      "loss_cls": 1.5698497008997947e-05,
      "loss_mitre": 0.00018457592523191124,
      "loss_ner": 0.0007683912990614772,
      "step": 5425
    },
    {
      "epoch": 2.806727037516171,
      "loss_anomaly": 0.010223882272839546,
      "loss_cls": 0.00019135061302222311,
      "loss_mitre": 0.0001849103282438591,
      "loss_ner": 0.004980142693966627,
      "step": 5425
    },
    {
      "epoch": 2.8196636481241915,
      "grad_norm": 0.5058518648147583,
      "learning_rate": 1.3412531136232996e-06,
      "loss": 0.20126958847045898,
      "step": 5450
    },
    {
      "epoch": 2.8196636481241915,
      "loss_anomaly": 0.00825396180152893,
      "loss_cls": 2.5777373593882658e-05,
      "loss_mitre": 0.00018922121671494097,
      "loss_ner": 0.012451560236513615,
      "step": 5450
    },
    {
      "epoch": 2.8196636481241915,
      "loss_anomaly": 0.009696164168417454,
      "loss_cls": 0.00018279984942637384,
      "loss_mitre": 0.44173702597618103,
      "loss_ner": 0.0029517582152038813,
      "step": 5450
    },
    {
      "epoch": 2.832600258732212,
      "grad_norm": 1.4039902687072754,
      "learning_rate": 1.2454493197930639e-06,
      "loss": 0.25932624816894534,
      "step": 5475
    },
    {
      "epoch": 2.832600258732212,
      "loss_anomaly": 0.008404770866036415,
      "loss_cls": 2.8064258003723808e-05,
      "loss_mitre": 0.00025275410735048354,
      "loss_ner": 0.004676959477365017,
      "step": 5475
    },
    {
      "epoch": 2.832600258732212,
      "loss_anomaly": 0.0073769125156104565,
      "loss_cls": 7.579826342407614e-05,
      "loss_mitre": 0.00018044054741039872,
      "loss_ner": 0.0030400685500353575,
      "step": 5475
    },
    {
      "epoch": 2.8455368693402328,
      "grad_norm": 0.4619896709918976,
      "learning_rate": 1.1496455259628281e-06,
      "loss": 0.22061048507690428,
      "step": 5500
    },
    {
      "epoch": 2.8455368693402328,
      "eval_anomaly_mae": 0.05601488870721534,
      "eval_anomaly_mse": 0.0052261094553052775,
      "eval_anomaly_n_eval": 27475.0,
      "eval_cls_accuracy": 0.9982469038059153,
      "eval_cls_auprc": 0.9997473023444335,
      "eval_cls_f1": 0.9937537779568809,
      "eval_cls_n_eval": 17683.0,
      "eval_cls_precision": 0.9907593411008437,
      "eval_cls_recall": 0.9967663702506063,
      "eval_loss": 0.05265136808156967,
      "eval_mitre_accuracy": 0.962847492855287,
      "eval_mitre_macro_f1": 0.15166314557461924,
      "eval_mitre_n_eval": 3849.0,
      "eval_ner_n_tokens": 666658.0,
      "eval_ner_token_acc": 0.9963294522828797,
      "eval_runtime": 68.8535,
      "eval_samples_per_second": 399.036,
      "eval_steps_per_second": 3.123,
      "step": 5500
    },
    {
      "epoch": 2.8455368693402328,
      "loss_anomaly": 0.01622823253273964,
      "loss_cls": 1.9246870579081587e-05,
      "loss_mitre": 0.00023424334358423948,
      "loss_ner": 0.0025254071224480867,
      "step": 5500
    },
    {
      "epoch": 2.8455368693402328,
      "loss_anomaly": 0.008821530267596245,
      "loss_cls": 3.800423655775376e-05,
      "loss_mitre": 0.0001813665876397863,
      "loss_ner": 0.009001723490655422,
      "step": 5500
    },
    {
      "epoch": 2.8584734799482536,
      "grad_norm": 0.4590380787849426,
      "learning_rate": 1.0538417321325926e-06,
      "loss": 0.15234447479248048,
      "step": 5525
    },
    {
      "epoch": 2.8584734799482536,
      "loss_anomaly": 0.011138280853629112,
      "loss_cls": 8.773666922934353e-05,
      "loss_mitre": 0.00020814232993870974,
      "loss_ner": 0.0033627457451075315,
      "step": 5525
    },
    {
      "epoch": 2.8584734799482536,
      "loss_anomaly": 0.008864171802997589,
      "loss_cls": 4.834592618863098e-05,
      "loss_mitre": 1.6217385530471802,
      "loss_ner": 0.014116772450506687,
      "step": 5525
    },
    {
      "epoch": 2.8714100905562745,
      "grad_norm": 2.5710086822509766,
      "learning_rate": 9.580379383023568e-07,
      "loss": 0.24974967956542968,
      "step": 5550
    },
    {
      "epoch": 2.8714100905562745,
      "loss_anomaly": 0.006814560852944851,
      "loss_cls": 1.8506676497054286e-05,
      "loss_mitre": 0.7971141934394836,
      "loss_ner": 0.023074399679899216,
      "step": 5550
    },
    {
      "epoch": 2.8714100905562745,
      "loss_anomaly": 0.007784162648022175,
      "loss_cls": 2.2390713638742454e-05,
      "loss_mitre": 0.9186891317367554,
      "loss_ner": 0.000934305542614311,
      "step": 5550
    },
    {
      "epoch": 2.884346701164295,
      "grad_norm": 2.5065431594848633,
      "learning_rate": 8.622341444721212e-07,
      "loss": 0.3678630447387695,
      "step": 5575
    },
    {
      "epoch": 2.884346701164295,
      "loss_anomaly": 0.010027755051851273,
      "loss_cls": 1.8293749235454015e-05,
      "loss_mitre": 0.0001559879310661927,
      "loss_ner": 0.001763995736837387,
      "step": 5575
    },
    {
      "epoch": 2.884346701164295,
      "loss_anomaly": 0.009827349334955215,
      "loss_cls": 0.0005358976777642965,
      "loss_mitre": 0.00019292118668090552,
      "loss_ner": 0.007400609087198973,
      "step": 5575
    },
    {
      "epoch": 2.897283311772316,
      "grad_norm": 0.8356830477714539,
      "learning_rate": 7.664303506418854e-07,
      "loss": 0.2538101959228516,
      "step": 5600
    },
    {
      "epoch": 2.897283311772316,
      "loss_anomaly": 0.011536290869116783,
      "loss_cls": 0.0001587838924024254,
      "loss_mitre": 0.00023856121697463095,
      "loss_ner": 0.036318231374025345,
      "step": 5600
    },
    {
      "epoch": 2.897283311772316,
      "loss_anomaly": 0.008266433142125607,
      "loss_cls": 4.586399518302642e-05,
      "loss_mitre": 0.00017365520761813968,
      "loss_ner": 0.018704354763031006,
      "step": 5600
    },
    {
      "epoch": 2.910219922380336,
      "grad_norm": 0.9187011122703552,
      "learning_rate": 6.706265568116498e-07,
      "loss": 0.18651405334472657,
      "step": 5625
    },
    {
      "epoch": 2.910219922380336,
      "loss_anomaly": 0.007268995977938175,
      "loss_cls": 3.090189056820236e-05,
      "loss_mitre": 0.00018393970094621181,
      "loss_ner": 0.008768973872065544,
      "step": 5625
    },
    {
      "epoch": 2.910219922380336,
      "loss_anomaly": 0.01079569198191166,
      "loss_cls": 0.0005609479849226773,
      "loss_mitre": 0.00024498600396327674,
      "loss_ner": 0.07177890837192535,
      "step": 5625
    },
    {
      "epoch": 2.923156532988357,
      "grad_norm": 0.737373411655426,
      "learning_rate": 5.748227629814141e-07,
      "loss": 0.1915519905090332,
      "step": 5650
    },
    {
      "epoch": 2.923156532988357,
      "loss_anomaly": 0.012341979891061783,
      "loss_cls": 0.0009810298215597868,
      "loss_mitre": 0.00020667299395427108,
      "loss_ner": 0.008651599287986755,
      "step": 5650
    },
    {
      "epoch": 2.923156532988357,
      "loss_anomaly": 0.008650192990899086,
      "loss_cls": 2.1586924049188383e-05,
      "loss_mitre": 0.00015385652659460902,
      "loss_ner": 0.0029026165138930082,
      "step": 5650
    },
    {
      "epoch": 2.936093143596378,
      "grad_norm": 1.3863157033920288,
      "learning_rate": 4.790189691511784e-07,
      "loss": 0.23931795120239258,
      "step": 5675
    },
    {
      "epoch": 2.936093143596378,
      "loss_anomaly": 0.008568516001105309,
      "loss_cls": 2.178256181650795e-05,
      "loss_mitre": 0.00016127264825627208,
      "loss_ner": 0.003524213097989559,
      "step": 5675
    },
    {
      "epoch": 2.936093143596378,
      "loss_anomaly": 0.01075021643191576,
      "loss_cls": 3.199388083885424e-05,
      "loss_mitre": 0.00016014168795663863,
      "loss_ner": 0.011834627948701382,
      "step": 5675
    },
    {
      "epoch": 2.9490297542043984,
      "grad_norm": 0.414189875125885,
      "learning_rate": 3.832151753209427e-07,
      "loss": 0.1525564956665039,
      "step": 5700
    },
    {
      "epoch": 2.9490297542043984,
      "loss_anomaly": 0.012188466265797615,
      "loss_cls": 6.671861046925187e-05,
      "loss_mitre": 0.0001780789898475632,
      "loss_ner": 0.00409987336024642,
      "step": 5700
    },
    {
      "epoch": 2.9490297542043984,
      "loss_anomaly": 0.009935321286320686,
      "loss_cls": 1.8401462511974387e-05,
      "loss_mitre": 0.00016090722056105733,
      "loss_ner": 0.018778221681714058,
      "step": 5700
    },
    {
      "epoch": 2.9619663648124193,
      "grad_norm": 0.5477131605148315,
      "learning_rate": 2.8741138149070703e-07,
      "loss": 0.20090675354003906,
      "step": 5725
    },
    {
      "epoch": 2.9619663648124193,
      "loss_anomaly": 0.014383917674422264,
      "loss_cls": 0.007157640065997839,
      "loss_mitre": 0.00019896159938070923,
      "loss_ner": 0.03777958080172539,
      "step": 5725
    },
    {
      "epoch": 2.9619663648124193,
      "loss_anomaly": 0.006888764910399914,
      "loss_cls": 0.0001929187128553167,
      "loss_mitre": 0.00010971269512083381,
      "loss_ner": 0.009115373715758324,
      "step": 5725
    },
    {
      "epoch": 2.9749029754204397,
      "grad_norm": 0.5208523869514465,
      "learning_rate": 1.9160758766047136e-07,
      "loss": 0.2482157325744629,
      "step": 5750
    },
    {
      "epoch": 2.9749029754204397,
      "loss_anomaly": 0.011900864541530609,
      "loss_cls": 0.0005876242648810148,
      "loss_mitre": 0.0001930329599417746,
      "loss_ner": 0.0017161347204819322,
      "step": 5750
    },
    {
      "epoch": 2.9749029754204397,
      "loss_anomaly": 0.010328579694032669,
      "loss_cls": 1.9634106138255447e-05,
      "loss_mitre": 0.00021860451670363545,
      "loss_ner": 0.0033704445231705904,
      "step": 5750
    },
    {
      "epoch": 2.9878395860284606,
      "grad_norm": 0.5997846722602844,
      "learning_rate": 9.580379383023568e-08,
      "loss": 0.23403270721435546,
      "step": 5775
    },
    {
      "epoch": 2.9878395860284606,
      "loss_anomaly": 0.007067985367029905,
      "loss_cls": 2.1628260583383963e-05,
      "loss_mitre": 1.0395928621292114,
      "loss_ner": 0.003586969105526805,
      "step": 5775
    },
    {
      "epoch": 2.9878395860284606,
      "loss_anomaly": 0.012543615885078907,
      "loss_cls": 0.00016412700642831624,
      "loss_mitre": 0.0002789763093460351,
      "loss_ner": 0.0026889871805906296,
      "step": 5775
    },
    {
      "epoch": 3.0,
      "eval_anomaly_mae": 0.05540453448392929,
      "eval_anomaly_mse": 0.005122618450574397,
      "eval_anomaly_n_eval": 27475.0,
      "eval_cls_accuracy": 0.9981338008256517,
      "eval_cls_auprc": 0.9997308030437642,
      "eval_cls_f1": 0.993358824713222,
      "eval_cls_n_eval": 17683.0,
      "eval_cls_precision": 0.9891783567134268,
      "eval_cls_recall": 0.9975747776879548,
      "eval_loss": 0.052331775426864624,
      "eval_mitre_accuracy": 0.962847492855287,
      "eval_mitre_macro_f1": 0.1530833242500602,
      "eval_mitre_n_eval": 3849.0,
      "eval_ner_n_tokens": 666658.0,
      "eval_ner_token_acc": 0.9964644540379025,
      "eval_runtime": 68.9128,
      "eval_samples_per_second": 398.693,
      "eval_steps_per_second": 3.12,
      "step": 5799
    }
  ],
  "logging_steps": 25,
  "max_steps": 5799,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "stateful_callbacks": {
    "EarlyStoppingCallback": {
      "args": {
        "early_stopping_patience": 3,
        "early_stopping_threshold": 0.0
      },
      "attributes": {
        "early_stopping_patience_counter": 1
      }
    },
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 6.718015162601393e+17,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": null
}