diff --git "a/last-checkpoint/trainer_state.json" "b/last-checkpoint/trainer_state.json"
--- "a/last-checkpoint/trainer_state.json"
+++ "b/last-checkpoint/trainer_state.json"
@@ -1,9 +1,9 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.19786307874950534,
+  "epoch": 0.3957261574990107,
   "eval_steps": 5,
-  "global_step": 500,
+  "global_step": 1000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -5224,6 +5224,5206 @@
       "eval_samples_per_second": 5.782,
       "eval_steps_per_second": 0.193,
       "step": 500
+    },
+    {
+      "epoch": 0.19825880490700434,
+      "grad_norm": 1.1264535554417732,
+      "learning_rate": 9.709390616735456e-05,
+      "loss": 0.1603,
+      "step": 501
+    },
+    {
+      "epoch": 0.19865453106450337,
+      "grad_norm": 1.2549995485933023,
+      "learning_rate": 9.707065470215174e-05,
+      "loss": 0.1631,
+      "step": 502
+    },
+    {
+      "epoch": 0.19905025722200237,
+      "grad_norm": 1.4298566453240185,
+      "learning_rate": 9.70473133971038e-05,
+      "loss": 0.2256,
+      "step": 503
+    },
+    {
+      "epoch": 0.1994459833795014,
+      "grad_norm": 1.243183708125135,
+      "learning_rate": 9.702388229676033e-05,
+      "loss": 0.1871,
+      "step": 504
+    },
+    {
+      "epoch": 0.1998417095370004,
+      "grad_norm": 0.8346136506497759,
+      "learning_rate": 9.700036144584237e-05,
+      "loss": 0.157,
+      "step": 505
+    },
+    {
+      "epoch": 0.1998417095370004,
+      "eval_PRM Accuracy": 0.8571428571428571,
+      "eval_PRM F1": 0.9217391304347826,
+      "eval_PRM F1 AUC": 0.5479586817511067,
+      "eval_PRM F1 AUC (fixed)": 0.8603049680275455,
+      "eval_PRM F1 Neg": 0.18181818181818182,
+      "eval_PRM NPV": 0.6666666666666666,
+      "eval_PRM Precision": 0.8617886178861789,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.10526315789473684,
+      "eval_loss": 0.3991861939430237,
+      "eval_runtime": 5.0295,
+      "eval_samples_per_second": 5.965,
+      "eval_steps_per_second": 0.199,
+      "step": 505
+    },
+    {
+      "epoch": 0.2002374356944994,
+      "grad_norm": 1.3900068270293604,
+      "learning_rate": 9.697675088924218e-05,
+      "loss": 0.2165,
+      "step": 506
+    },
+    {
+      "epoch": 0.20063316185199842,
+      "grad_norm": 0.5945249548749268,
+      "learning_rate": 9.695305067202328e-05,
+      "loss": 0.1352,
+      "step": 507
+    },
+    {
+      "epoch": 0.20102888800949742,
+      "grad_norm": 1.043679639075919,
+      "learning_rate": 9.692926083942029e-05,
+      "loss": 0.1934,
+      "step": 508
+    },
+    {
+      "epoch": 0.20142461416699645,
+      "grad_norm": 0.9441783291481128,
+      "learning_rate": 9.690538143683891e-05,
+      "loss": 0.2495,
+      "step": 509
+    },
+    {
+      "epoch": 0.20182034032449545,
+      "grad_norm": 0.7709899617459975,
+      "learning_rate": 9.688141250985574e-05,
+      "loss": 0.2009,
+      "step": 510
+    },
+    {
+      "epoch": 0.20182034032449545,
+      "eval_PRM Accuracy": 0.8650793650793651,
+      "eval_PRM F1": 0.9230769230769231,
+      "eval_PRM F1 AUC": 0.6608460403344811,
+      "eval_PRM F1 AUC (fixed)": 0.8499754058042301,
+      "eval_PRM F1 Neg": 0.45161290322580644,
+      "eval_PRM NPV": 0.5833333333333334,
+      "eval_PRM Precision": 0.8947368421052632,
+      "eval_PRM Recall": 0.9532710280373832,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.32386067509651184,
+      "eval_runtime": 5.0065,
+      "eval_samples_per_second": 5.992,
+      "eval_steps_per_second": 0.2,
+      "step": 510
+    },
+    {
+      "epoch": 0.20221606648199447,
+      "grad_norm": 1.1223110605792201,
+      "learning_rate": 9.68573541042183e-05,
+      "loss": 0.298,
+      "step": 511
+    },
+    {
+      "epoch": 0.20261179263949347,
+      "grad_norm": 1.4007344018582633,
+      "learning_rate": 9.683320626584486e-05,
+      "loss": 0.2643,
+      "step": 512
+    },
+    {
+      "epoch": 0.20300751879699247,
+      "grad_norm": 1.7682189070436682,
+      "learning_rate": 9.680896904082439e-05,
+      "loss": 0.2133,
+      "step": 513
+    },
+    {
+      "epoch": 0.2034032449544915,
+      "grad_norm": 1.036671216282991,
+      "learning_rate": 9.678464247541648e-05,
+      "loss": 0.1837,
+      "step": 514
+    },
+    {
+      "epoch": 0.2037989711119905,
+      "grad_norm": 0.695210891373307,
+      "learning_rate": 9.67602266160512e-05,
+      "loss": 0.1752,
+      "step": 515
+    },
+    {
+      "epoch": 0.2037989711119905,
+      "eval_PRM Accuracy": 0.8650793650793651,
+      "eval_PRM F1": 0.9230769230769231,
+      "eval_PRM F1 AUC": 0.6608460403344811,
+      "eval_PRM F1 AUC (fixed)": 0.867437284800787,
+      "eval_PRM F1 Neg": 0.45161290322580644,
+      "eval_PRM NPV": 0.5833333333333334,
+      "eval_PRM Precision": 0.8947368421052632,
+      "eval_PRM Recall": 0.9532710280373832,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.33037108182907104,
+      "eval_runtime": 5.3133,
+      "eval_samples_per_second": 5.646,
+      "eval_steps_per_second": 0.188,
+      "step": 515
+    },
+    {
+      "epoch": 0.20419469726948952,
+      "grad_norm": 1.0365457684608814,
+      "learning_rate": 9.673572150932909e-05,
+      "loss": 0.2064,
+      "step": 516
+    },
+    {
+      "epoch": 0.20459042342698852,
+      "grad_norm": 1.4331670555504556,
+      "learning_rate": 9.6711127202021e-05,
+      "loss": 0.3918,
+      "step": 517
+    },
+    {
+      "epoch": 0.20498614958448755,
+      "grad_norm": 1.4725826262605737,
+      "learning_rate": 9.668644374106805e-05,
+      "loss": 0.2939,
+      "step": 518
+    },
+    {
+      "epoch": 0.20538187574198655,
+      "grad_norm": 0.7531585558002998,
+      "learning_rate": 9.666167117358149e-05,
+      "loss": 0.1518,
+      "step": 519
+    },
+    {
+      "epoch": 0.20577760189948555,
+      "grad_norm": 1.1173060271029716,
+      "learning_rate": 9.663680954684268e-05,
+      "loss": 0.1819,
+      "step": 520
+    },
+    {
+      "epoch": 0.20577760189948555,
+      "eval_PRM Accuracy": 0.873015873015873,
+      "eval_PRM F1": 0.9285714285714286,
+      "eval_PRM F1 AUC": 0.6438760452533202,
+      "eval_PRM F1 AUC (fixed)": 0.8718642400393507,
+      "eval_PRM F1 Neg": 0.42857142857142855,
+      "eval_PRM NPV": 0.6666666666666666,
+      "eval_PRM Precision": 0.8888888888888888,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.3157894736842105,
+      "eval_loss": 0.3360026180744171,
+      "eval_runtime": 5.1335,
+      "eval_samples_per_second": 5.844,
+      "eval_steps_per_second": 0.195,
+      "step": 520
+    },
+    {
+      "epoch": 0.20617332805698457,
+      "grad_norm": 1.2406857590155362,
+      "learning_rate": 9.661185890830293e-05,
+      "loss": 0.1799,
+      "step": 521
+    },
+    {
+      "epoch": 0.20656905421448357,
+      "grad_norm": 0.8502138580876508,
+      "learning_rate": 9.658681930558345e-05,
+      "loss": 0.1954,
+      "step": 522
+    },
+    {
+      "epoch": 0.2069647803719826,
+      "grad_norm": 0.8615146218402908,
+      "learning_rate": 9.656169078647526e-05,
+      "loss": 0.224,
+      "step": 523
+    },
+    {
+      "epoch": 0.2073605065294816,
+      "grad_norm": 1.0325978121139723,
+      "learning_rate": 9.653647339893905e-05,
+      "loss": 0.2473,
+      "step": 524
+    },
+    {
+      "epoch": 0.2077562326869806,
+      "grad_norm": 0.8077612731881857,
+      "learning_rate": 9.651116719110517e-05,
+      "loss": 0.1659,
+      "step": 525
+    },
+    {
+      "epoch": 0.2077562326869806,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.9321266968325792,
+      "eval_PRM F1 AUC": 0.6918347270044269,
+      "eval_PRM F1 AUC (fixed)": 0.8644859813084113,
+      "eval_PRM F1 Neg": 0.5161290322580645,
+      "eval_PRM NPV": 0.6666666666666666,
+      "eval_PRM Precision": 0.9035087719298246,
+      "eval_PRM Recall": 0.9626168224299065,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.3466796875,
+      "eval_runtime": 5.1442,
+      "eval_samples_per_second": 5.832,
+      "eval_steps_per_second": 0.194,
+      "step": 525
+    },
+    {
+      "epoch": 0.20815195884447962,
+      "grad_norm": 1.0624343543580033,
+      "learning_rate": 9.648577221127346e-05,
+      "loss": 0.1215,
+      "step": 526
+    },
+    {
+      "epoch": 0.20854768500197862,
+      "grad_norm": 5.168173680614494,
+      "learning_rate": 9.646028850791325e-05,
+      "loss": 0.1992,
+      "step": 527
+    },
+    {
+      "epoch": 0.20894341115947765,
+      "grad_norm": 1.8652195410740853,
+      "learning_rate": 9.64347161296631e-05,
+      "loss": 0.2237,
+      "step": 528
+    },
+    {
+      "epoch": 0.20933913731697665,
+      "grad_norm": 0.7799574378963159,
+      "learning_rate": 9.640905512533091e-05,
+      "loss": 0.2083,
+      "step": 529
+    },
+    {
+      "epoch": 0.20973486347447567,
+      "grad_norm": 0.9037284373512658,
+      "learning_rate": 9.638330554389374e-05,
+      "loss": 0.1608,
+      "step": 530
+    },
+    {
+      "epoch": 0.20973486347447567,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.8644859813084113,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.3483235538005829,
+      "eval_runtime": 5.0089,
+      "eval_samples_per_second": 5.989,
+      "eval_steps_per_second": 0.2,
+      "step": 530
+    },
+    {
+      "epoch": 0.21013058963197467,
+      "grad_norm": 0.9656568325957725,
+      "learning_rate": 9.635746743449763e-05,
+      "loss": 0.1943,
+      "step": 531
+    },
+    {
+      "epoch": 0.21052631578947367,
+      "grad_norm": 1.1758744585978005,
+      "learning_rate": 9.633154084645766e-05,
+      "loss": 0.1958,
+      "step": 532
+    },
+    {
+      "epoch": 0.2109220419469727,
+      "grad_norm": 1.1040540916105812,
+      "learning_rate": 9.630552582925772e-05,
+      "loss": 0.2008,
+      "step": 533
+    },
+    {
+      "epoch": 0.2113177681044717,
+      "grad_norm": 0.9935287884699731,
+      "learning_rate": 9.627942243255055e-05,
+      "loss": 0.1666,
+      "step": 534
+    },
+    {
+      "epoch": 0.21171349426197072,
+      "grad_norm": 0.972737886868156,
+      "learning_rate": 9.625323070615751e-05,
+      "loss": 0.1853,
+      "step": 535
+    },
+    {
+      "epoch": 0.21171349426197072,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.8703885882931629,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.30195313692092896,
+      "eval_runtime": 4.9376,
+      "eval_samples_per_second": 6.076,
+      "eval_steps_per_second": 0.203,
+      "step": 535
+    },
+    {
+      "epoch": 0.21210922041946972,
+      "grad_norm": 0.7467895741053363,
+      "learning_rate": 9.622695070006855e-05,
+      "loss": 0.134,
+      "step": 536
+    },
+    {
+      "epoch": 0.21250494657696875,
+      "grad_norm": 0.9403120886314019,
+      "learning_rate": 9.620058246444218e-05,
+      "loss": 0.2194,
+      "step": 537
+    },
+    {
+      "epoch": 0.21290067273446775,
+      "grad_norm": 1.2576960118400111,
+      "learning_rate": 9.617412604960523e-05,
+      "loss": 0.2212,
+      "step": 538
+    },
+    {
+      "epoch": 0.21329639889196675,
+      "grad_norm": 0.8308266600414512,
+      "learning_rate": 9.614758150605286e-05,
+      "loss": 0.1599,
+      "step": 539
+    },
+    {
+      "epoch": 0.21369212504946578,
+      "grad_norm": 0.960188799817658,
+      "learning_rate": 9.612094888444845e-05,
+      "loss": 0.1733,
+      "step": 540
+    },
+    {
+      "epoch": 0.21369212504946578,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9375,
+      "eval_PRM F1 AUC": 0.6748647319232661,
+      "eval_PRM F1 AUC (fixed)": 0.8607968519429415,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.7777777777777778,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.3046875,
+      "eval_runtime": 5.0892,
+      "eval_samples_per_second": 5.895,
+      "eval_steps_per_second": 0.196,
+      "step": 540
+    },
+    {
+      "epoch": 0.21408785120696477,
+      "grad_norm": 0.8821226244088822,
+      "learning_rate": 9.609422823562345e-05,
+      "loss": 0.1838,
+      "step": 541
+    },
+    {
+      "epoch": 0.2144835773644638,
+      "grad_norm": 0.8779396786356636,
+      "learning_rate": 9.606741961057736e-05,
+      "loss": 0.1994,
+      "step": 542
+    },
+    {
+      "epoch": 0.2148793035219628,
+      "grad_norm": 0.9278291992515524,
+      "learning_rate": 9.604052306047755e-05,
+      "loss": 0.1424,
+      "step": 543
+    },
+    {
+      "epoch": 0.2152750296794618,
+      "grad_norm": 1.3601463696375726,
+      "learning_rate": 9.601353863665925e-05,
+      "loss": 0.2202,
+      "step": 544
+    },
+    {
+      "epoch": 0.21567075583696083,
+      "grad_norm": 0.9571474831287845,
+      "learning_rate": 9.598646639062538e-05,
+      "loss": 0.2449,
+      "step": 545
+    },
+    {
+      "epoch": 0.21567075583696083,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.8630103295622233,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.3026692569255829,
+      "eval_runtime": 4.9598,
+      "eval_samples_per_second": 6.049,
+      "eval_steps_per_second": 0.202,
+      "step": 545
+    },
+    {
+      "epoch": 0.21606648199445982,
+      "grad_norm": 0.6974800822103099,
+      "learning_rate": 9.595930637404649e-05,
+      "loss": 0.1742,
+      "step": 546
+    },
+    {
+      "epoch": 0.21646220815195885,
+      "grad_norm": 0.7413282402564647,
+      "learning_rate": 9.593205863876062e-05,
+      "loss": 0.17,
+      "step": 547
+    },
+    {
+      "epoch": 0.21685793430945785,
+      "grad_norm": 1.0502428203770906,
+      "learning_rate": 9.590472323677328e-05,
+      "loss": 0.1984,
+      "step": 548
+    },
+    {
+      "epoch": 0.21725366046695688,
+      "grad_norm": 1.0691413471866484,
+      "learning_rate": 9.58773002202573e-05,
+      "loss": 0.1281,
+      "step": 549
+    },
+    {
+      "epoch": 0.21764938662445588,
+      "grad_norm": 0.6566165275714569,
+      "learning_rate": 9.584978964155266e-05,
+      "loss": 0.1253,
+      "step": 550
+    },
+    {
+      "epoch": 0.21764938662445588,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.8605509099852434,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.3195638060569763,
+      "eval_runtime": 5.1351,
+      "eval_samples_per_second": 5.842,
+      "eval_steps_per_second": 0.195,
+      "step": 550
+    },
+    {
+      "epoch": 0.21804511278195488,
+      "grad_norm": 0.9558714705740771,
+      "learning_rate": 9.582219155316656e-05,
+      "loss": 0.1737,
+      "step": 551
+    },
+    {
+      "epoch": 0.2184408389394539,
+      "grad_norm": 0.9211184932049711,
+      "learning_rate": 9.579450600777314e-05,
+      "loss": 0.1874,
+      "step": 552
+    },
+    {
+      "epoch": 0.2188365650969529,
+      "grad_norm": 0.9822808185532677,
+      "learning_rate": 9.576673305821353e-05,
+      "loss": 0.2077,
+      "step": 553
+    },
+    {
+      "epoch": 0.21923229125445193,
+      "grad_norm": 1.0456582618812262,
+      "learning_rate": 9.573887275749564e-05,
+      "loss": 0.1653,
+      "step": 554
+    },
+    {
+      "epoch": 0.21962801741195093,
+      "grad_norm": 2.7426170342114937,
+      "learning_rate": 9.57109251587941e-05,
+      "loss": 0.3692,
+      "step": 555
+    },
+    {
+      "epoch": 0.21962801741195093,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.8644859813084111,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.3179687559604645,
+      "eval_runtime": 4.8311,
+      "eval_samples_per_second": 6.21,
+      "eval_steps_per_second": 0.207,
+      "step": 555
+    },
+    {
+      "epoch": 0.22002374356944995,
+      "grad_norm": 0.9584841161621672,
+      "learning_rate": 9.568289031545017e-05,
+      "loss": 0.2137,
+      "step": 556
+    },
+    {
+      "epoch": 0.22041946972694895,
+      "grad_norm": 0.902570061001305,
+      "learning_rate": 9.565476828097163e-05,
+      "loss": 0.1766,
+      "step": 557
+    },
+    {
+      "epoch": 0.22081519588444795,
+      "grad_norm": 0.6781830371649948,
+      "learning_rate": 9.562655910903267e-05,
+      "loss": 0.1375,
+      "step": 558
+    },
+    {
+      "epoch": 0.22121092204194698,
+      "grad_norm": 0.7032453896001626,
+      "learning_rate": 9.55982628534738e-05,
+      "loss": 0.1655,
+      "step": 559
+    },
+    {
+      "epoch": 0.22160664819944598,
+      "grad_norm": 1.2747164883201627,
+      "learning_rate": 9.55698795683017e-05,
+      "loss": 0.1696,
+      "step": 560
+    },
+    {
+      "epoch": 0.22160664819944598,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.8664535169699951,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.3035481870174408,
+      "eval_runtime": 4.7243,
+      "eval_samples_per_second": 6.35,
+      "eval_steps_per_second": 0.212,
+      "step": 560
+    },
+    {
+      "epoch": 0.222002374356945,
+      "grad_norm": 0.6865135453027532,
+      "learning_rate": 9.554140930768922e-05,
+      "loss": 0.1609,
+      "step": 561
+    },
+    {
+      "epoch": 0.222398100514444,
+      "grad_norm": 1.1115197727130854,
+      "learning_rate": 9.551285212597516e-05,
+      "loss": 0.1638,
+      "step": 562
+    },
+    {
+      "epoch": 0.222793826671943,
+      "grad_norm": 0.6641373048947078,
+      "learning_rate": 9.548420807766425e-05,
+      "loss": 0.1424,
+      "step": 563
+    },
+    {
+      "epoch": 0.22318955282944203,
+      "grad_norm": 1.4565119763313692,
+      "learning_rate": 9.545547721742698e-05,
+      "loss": 0.2122,
+      "step": 564
+    },
+    {
+      "epoch": 0.22358527898694103,
+      "grad_norm": 1.174160384376405,
+      "learning_rate": 9.542665960009959e-05,
+      "loss": 0.1643,
+      "step": 565
+    },
+    {
+      "epoch": 0.22358527898694103,
+      "eval_PRM Accuracy": 0.8650793650793651,
+      "eval_PRM F1": 0.925764192139738,
+      "eval_PRM F1 AUC": 0.5742744712247909,
+      "eval_PRM F1 AUC (fixed)": 0.8792424987702901,
+      "eval_PRM F1 Neg": 0.2608695652173913,
+      "eval_PRM NPV": 0.75,
+      "eval_PRM Precision": 0.8688524590163934,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.15789473684210525,
+      "eval_loss": 0.36243489384651184,
+      "eval_runtime": 5.0113,
+      "eval_samples_per_second": 5.986,
+      "eval_steps_per_second": 0.2,
+      "step": 565
+    },
+    {
+      "epoch": 0.22398100514444005,
+      "grad_norm": 1.1951671765700973,
+      "learning_rate": 9.539775528068384e-05,
+      "loss": 0.2291,
+      "step": 566
+    },
+    {
+      "epoch": 0.22437673130193905,
+      "grad_norm": 1.0704835132881214,
+      "learning_rate": 9.536876431434703e-05,
+      "loss": 0.2534,
+      "step": 567
+    },
+    {
+      "epoch": 0.22477245745943808,
+      "grad_norm": 1.1412400220486025,
+      "learning_rate": 9.533968675642178e-05,
+      "loss": 0.2867,
+      "step": 568
+    },
+    {
+      "epoch": 0.22516818361693708,
+      "grad_norm": 0.8719591915839799,
+      "learning_rate": 9.531052266240601e-05,
+      "loss": 0.1715,
+      "step": 569
+    },
+    {
+      "epoch": 0.22556390977443608,
+      "grad_norm": 0.6894055076847694,
+      "learning_rate": 9.528127208796282e-05,
+      "loss": 0.1572,
+      "step": 570
+    },
+    {
+      "epoch": 0.22556390977443608,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9375,
+      "eval_PRM F1 AUC": 0.6748647319232661,
+      "eval_PRM F1 AUC (fixed)": 0.8748155435317265,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.7777777777777778,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.31067708134651184,
+      "eval_runtime": 5.3395,
+      "eval_samples_per_second": 5.619,
+      "eval_steps_per_second": 0.187,
+      "step": 570
+    },
+    {
+      "epoch": 0.2259596359319351,
+      "grad_norm": 0.9686660329013871,
+      "learning_rate": 9.525193508892034e-05,
+      "loss": 0.1939,
+      "step": 571
+    },
+    {
+      "epoch": 0.2263553620894341,
+      "grad_norm": 0.8864659755188832,
+      "learning_rate": 9.522251172127166e-05,
+      "loss": 0.1876,
+      "step": 572
+    },
+    {
+      "epoch": 0.22675108824693313,
+      "grad_norm": 0.7742166668044917,
+      "learning_rate": 9.519300204117469e-05,
+      "loss": 0.1888,
+      "step": 573
+    },
+    {
+      "epoch": 0.22714681440443213,
+      "grad_norm": 0.8043464652581477,
+      "learning_rate": 9.516340610495215e-05,
+      "loss": 0.2561,
+      "step": 574
+    },
+    {
+      "epoch": 0.22754254056193116,
+      "grad_norm": 0.6464540788353937,
+      "learning_rate": 9.513372396909133e-05,
+      "loss": 0.1456,
+      "step": 575
+    },
+    {
+      "epoch": 0.22754254056193116,
+      "eval_PRM Accuracy": 0.873015873015873,
+      "eval_PRM F1": 0.9298245614035088,
+      "eval_PRM F1 AUC": 0.6005902606984752,
+      "eval_PRM F1 AUC (fixed)": 0.8701426463354648,
+      "eval_PRM F1 Neg": 0.3333333333333333,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.8760330578512396,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.21052631578947367,
+      "eval_loss": 0.3508463501930237,
+      "eval_runtime": 5.0857,
+      "eval_samples_per_second": 5.899,
+      "eval_steps_per_second": 0.197,
+      "step": 575
+    },
+    {
+      "epoch": 0.22793826671943015,
+      "grad_norm": 0.5592506160676732,
+      "learning_rate": 9.510395569024404e-05,
+      "loss": 0.1366,
+      "step": 576
+    },
+    {
+      "epoch": 0.22833399287692915,
+      "grad_norm": 0.9384888980101656,
+      "learning_rate": 9.507410132522652e-05,
+      "loss": 0.1857,
+      "step": 577
+    },
+    {
+      "epoch": 0.22872971903442818,
+      "grad_norm": 0.9045955332067419,
+      "learning_rate": 9.50441609310193e-05,
+      "loss": 0.2088,
+      "step": 578
+    },
+    {
+      "epoch": 0.22912544519192718,
+      "grad_norm": 1.2276700932407885,
+      "learning_rate": 9.501413456476717e-05,
+      "loss": 0.2593,
+      "step": 579
+    },
+    {
+      "epoch": 0.2295211713494262,
+      "grad_norm": 0.7015735002553686,
+      "learning_rate": 9.498402228377892e-05,
+      "loss": 0.1231,
+      "step": 580
+    },
+    {
+      "epoch": 0.2295211713494262,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.868666994589277,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.3221028745174408,
+      "eval_runtime": 5.0205,
+      "eval_samples_per_second": 5.975,
+      "eval_steps_per_second": 0.199,
+      "step": 580
+    },
+    {
+      "epoch": 0.2299168975069252,
+      "grad_norm": 1.1283779128227227,
+      "learning_rate": 9.495382414552737e-05,
+      "loss": 0.1915,
+      "step": 581
+    },
+    {
+      "epoch": 0.23031262366442423,
+      "grad_norm": 0.7567864418022118,
+      "learning_rate": 9.492354020764919e-05,
+      "loss": 0.137,
+      "step": 582
+    },
+    {
+      "epoch": 0.23070834982192323,
+      "grad_norm": 1.0400915649584914,
+      "learning_rate": 9.489317052794481e-05,
+      "loss": 0.1729,
+      "step": 583
+    },
+    {
+      "epoch": 0.23110407597942223,
+      "grad_norm": 1.0751961953781162,
+      "learning_rate": 9.486271516437832e-05,
+      "loss": 0.2185,
+      "step": 584
+    },
+    {
+      "epoch": 0.23149980213692126,
+      "grad_norm": 1.16836913547842,
+      "learning_rate": 9.483217417507734e-05,
+      "loss": 0.1408,
+      "step": 585
+    },
+    {
+      "epoch": 0.23149980213692126,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9375,
+      "eval_PRM F1 AUC": 0.6748647319232661,
+      "eval_PRM F1 AUC (fixed)": 0.8745696015740285,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.7777777777777778,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.31787109375,
+      "eval_runtime": 4.9632,
+      "eval_samples_per_second": 6.045,
+      "eval_steps_per_second": 0.201,
+      "step": 585
+    },
+    {
+      "epoch": 0.23189552829442026,
+      "grad_norm": 1.128841484632027,
+      "learning_rate": 9.480154761833293e-05,
+      "loss": 0.2409,
+      "step": 586
+    },
+    {
+      "epoch": 0.23229125445191928,
+      "grad_norm": 1.0484734719659377,
+      "learning_rate": 9.477083555259943e-05,
+      "loss": 0.2563,
+      "step": 587
+    },
+    {
+      "epoch": 0.23268698060941828,
+      "grad_norm": 1.1994550998647855,
+      "learning_rate": 9.474003803649441e-05,
+      "loss": 0.2347,
+      "step": 588
+    },
+    {
+      "epoch": 0.23308270676691728,
+      "grad_norm": 1.4231828524881203,
+      "learning_rate": 9.470915512879852e-05,
+      "loss": 0.2093,
+      "step": 589
+    },
+    {
+      "epoch": 0.2334784329244163,
+      "grad_norm": 0.9663546927104574,
+      "learning_rate": 9.467818688845544e-05,
+      "loss": 0.2118,
+      "step": 590
+    },
+    {
+      "epoch": 0.2334784329244163,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.8735858337432366,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.29755860567092896,
+      "eval_runtime": 5.2713,
+      "eval_samples_per_second": 5.691,
+      "eval_steps_per_second": 0.19,
+      "step": 590
+    },
+    {
+      "epoch": 0.2338741590819153,
+      "grad_norm": 0.907891643306674,
+      "learning_rate": 9.464713337457163e-05,
+      "loss": 0.1468,
+      "step": 591
+    },
+    {
+      "epoch": 0.23426988523941433,
+      "grad_norm": 0.9215718171161111,
+      "learning_rate": 9.461599464641638e-05,
+      "loss": 0.1335,
+      "step": 592
+    },
+    {
+      "epoch": 0.23466561139691333,
+      "grad_norm": 1.5082953660735747,
+      "learning_rate": 9.458477076342157e-05,
+      "loss": 0.3044,
+      "step": 593
+    },
+    {
+      "epoch": 0.23506133755441236,
+      "grad_norm": 0.6348812107302226,
+      "learning_rate": 9.455346178518164e-05,
+      "loss": 0.2146,
+      "step": 594
+    },
+    {
+      "epoch": 0.23545706371191136,
+      "grad_norm": 1.122679016046897,
+      "learning_rate": 9.452206777145342e-05,
+      "loss": 0.2414,
+      "step": 595
+    },
+    {
+      "epoch": 0.23545706371191136,
+      "eval_PRM Accuracy": 0.8650793650793651,
+      "eval_PRM F1": 0.925764192139738,
+      "eval_PRM F1 AUC": 0.5742744712247909,
+      "eval_PRM F1 AUC (fixed)": 0.8622725036891293,
+      "eval_PRM F1 Neg": 0.2608695652173913,
+      "eval_PRM NPV": 0.75,
+      "eval_PRM Precision": 0.8688524590163934,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.15789473684210525,
+      "eval_loss": 0.33574217557907104,
+      "eval_runtime": 5.1043,
+      "eval_samples_per_second": 5.877,
+      "eval_steps_per_second": 0.196,
+      "step": 595
+    },
+    {
+      "epoch": 0.23585278986941036,
+      "grad_norm": 0.7336962804636922,
+      "learning_rate": 9.44905887821561e-05,
+      "loss": 0.1946,
+      "step": 596
+    },
+    {
+      "epoch": 0.23624851602690938,
+      "grad_norm": 1.5076773027459303,
+      "learning_rate": 9.445902487737095e-05,
+      "loss": 0.2643,
+      "step": 597
+    },
+    {
+      "epoch": 0.23664424218440838,
+      "grad_norm": 0.9990565626793124,
+      "learning_rate": 9.442737611734141e-05,
+      "loss": 0.2163,
+      "step": 598
+    },
+    {
+      "epoch": 0.2370399683419074,
+      "grad_norm": 0.7677618013103612,
+      "learning_rate": 9.439564256247281e-05,
+      "loss": 0.237,
+      "step": 599
+    },
+    {
+      "epoch": 0.2374356944994064,
+      "grad_norm": 0.8864752202385261,
+      "learning_rate": 9.436382427333237e-05,
+      "loss": 0.2222,
+      "step": 600
+    },
+    {
+      "epoch": 0.2374356944994064,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.9321266968325792,
+      "eval_PRM F1 AUC": 0.6918347270044269,
+      "eval_PRM F1 AUC (fixed)": 0.852680767338908,
+      "eval_PRM F1 Neg": 0.5161290322580645,
+      "eval_PRM NPV": 0.6666666666666666,
+      "eval_PRM Precision": 0.9035087719298246,
+      "eval_PRM Recall": 0.9626168224299065,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.29798176884651184,
+      "eval_runtime": 4.819,
+      "eval_samples_per_second": 6.225,
+      "eval_steps_per_second": 0.208,
+      "step": 600
+    },
+    {
+      "epoch": 0.23783142065690543,
+      "grad_norm": 0.843227711539601,
+      "learning_rate": 9.4331921310649e-05,
+      "loss": 0.2179,
+      "step": 601
+    },
+    {
+      "epoch": 0.23822714681440443,
+      "grad_norm": 0.9023599319655619,
+      "learning_rate": 9.429993373531326e-05,
+      "loss": 0.2213,
+      "step": 602
+    },
+    {
+      "epoch": 0.23862287297190343,
+      "grad_norm": 0.5349869383171925,
+      "learning_rate": 9.426786160837713e-05,
+      "loss": 0.1216,
+      "step": 603
+    },
+    {
+      "epoch": 0.23901859912940246,
+      "grad_norm": 0.6550727960830575,
+      "learning_rate": 9.423570499105403e-05,
+      "loss": 0.1914,
+      "step": 604
+    },
+    {
+      "epoch": 0.23941432528690146,
+      "grad_norm": 0.6256497404073587,
+      "learning_rate": 9.420346394471864e-05,
+      "loss": 0.1665,
+      "step": 605
+    },
+    {
+      "epoch": 0.23941432528690146,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.8460403344810624,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.36315104365348816,
+      "eval_runtime": 4.9856,
+      "eval_samples_per_second": 6.017,
+      "eval_steps_per_second": 0.201,
+      "step": 605
+    },
+    {
+      "epoch": 0.23981005144440048,
+      "grad_norm": 0.6183163375383558,
+      "learning_rate": 9.417113853090675e-05,
+      "loss": 0.137,
+      "step": 606
+    },
+    {
+      "epoch": 0.24020577760189948,
+      "grad_norm": 0.8027547947526502,
+      "learning_rate": 9.413872881131518e-05,
+      "loss": 0.1754,
+      "step": 607
+    },
+    {
+      "epoch": 0.24060150375939848,
+      "grad_norm": 0.671864441942157,
+      "learning_rate": 9.410623484780168e-05,
+      "loss": 0.1527,
+      "step": 608
+    },
+    {
+      "epoch": 0.2409972299168975,
+      "grad_norm": 1.1548928413207609,
+      "learning_rate": 9.407365670238479e-05,
+      "loss": 0.2031,
+      "step": 609
+    },
+    {
+      "epoch": 0.2413929560743965,
+      "grad_norm": 1.3152261646506562,
+      "learning_rate": 9.404099443724368e-05,
+      "loss": 0.2406,
+      "step": 610
+    },
+    {
+      "epoch": 0.2413929560743965,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.8430890309886866,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.38823240995407104,
+      "eval_runtime": 5.115,
+      "eval_samples_per_second": 5.865,
+      "eval_steps_per_second": 0.196,
+      "step": 610
+    },
+    {
+      "epoch": 0.24178868223189554,
+      "grad_norm": 1.8054059950134695,
+      "learning_rate": 9.400824811471811e-05,
+      "loss": 0.2868,
+      "step": 611
+    },
+    {
+      "epoch": 0.24218440838939453,
+      "grad_norm": 1.483288223294385,
+      "learning_rate": 9.397541779730827e-05,
+      "loss": 0.2897,
+      "step": 612
+    },
+    {
+      "epoch": 0.24258013454689356,
+      "grad_norm": 0.8151158082075218,
+      "learning_rate": 9.394250354767467e-05,
+      "loss": 0.2069,
+      "step": 613
+    },
+    {
+      "epoch": 0.24297586070439256,
+      "grad_norm": 0.9609947966439928,
+      "learning_rate": 9.390950542863797e-05,
+      "loss": 0.2153,
+      "step": 614
+    },
+    {
+      "epoch": 0.24337158686189156,
+      "grad_norm": 1.0979707104056993,
+      "learning_rate": 9.387642350317894e-05,
+      "loss": 0.2064,
+      "step": 615
+    },
+    {
+      "epoch": 0.24337158686189156,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.9321266968325792,
+      "eval_PRM F1 AUC": 0.6918347270044269,
+      "eval_PRM F1 AUC (fixed)": 0.8428430890309887,
+      "eval_PRM F1 Neg": 0.5161290322580645,
+      "eval_PRM NPV": 0.6666666666666666,
+      "eval_PRM Precision": 0.9035087719298246,
+      "eval_PRM Recall": 0.9626168224299065,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.3037109375,
+      "eval_runtime": 4.9687,
+      "eval_samples_per_second": 6.038,
+      "eval_steps_per_second": 0.201,
+      "step": 615
+    },
+    {
+      "epoch": 0.24376731301939059,
+      "grad_norm": 0.8619863219027731,
+      "learning_rate": 9.384325783443832e-05,
+      "loss": 0.1569,
+      "step": 616
+    },
+    {
+      "epoch": 0.24416303917688958,
+      "grad_norm": 0.8612757328902113,
+      "learning_rate": 9.381000848571666e-05,
+      "loss": 0.2054,
+      "step": 617
+    },
+    {
+      "epoch": 0.2445587653343886,
+      "grad_norm": 1.1283574341271685,
+      "learning_rate": 9.377667552047423e-05,
+      "loss": 0.1545,
+      "step": 618
+    },
+    {
+      "epoch": 0.2449544914918876,
+      "grad_norm": 0.7739505180687086,
+      "learning_rate": 9.374325900233088e-05,
+      "loss": 0.1364,
+      "step": 619
+    },
+    {
+      "epoch": 0.24535021764938664,
+      "grad_norm": 0.9957171774689368,
+      "learning_rate": 9.370975899506593e-05,
+      "loss": 0.1954,
+      "step": 620
+    },
+    {
+      "epoch": 0.24535021764938664,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.8438268568617807,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.32962238788604736,
+      "eval_runtime": 5.0662,
+      "eval_samples_per_second": 5.922,
+      "eval_steps_per_second": 0.197,
+      "step": 620
+    },
+    {
+      "epoch": 0.24574594380688564,
+      "grad_norm": 1.4168595803088824,
+      "learning_rate": 9.367617556261808e-05,
+      "loss": 0.21,
+      "step": 621
+    },
+    {
+      "epoch": 0.24614166996438464,
+      "grad_norm": 0.9480516220046619,
+      "learning_rate": 9.364250876908522e-05,
+      "loss": 0.1425,
+      "step": 622
+    },
+    {
+      "epoch": 0.24653739612188366,
+      "grad_norm": 0.8553451927402599,
+      "learning_rate": 9.360875867872437e-05,
+      "loss": 0.1364,
+      "step": 623
+    },
+    {
+      "epoch": 0.24693312227938266,
+      "grad_norm": 2.176078009465126,
+      "learning_rate": 9.357492535595151e-05,
+      "loss": 0.2277,
+      "step": 624
+    },
+    {
+      "epoch": 0.2473288484368817,
+      "grad_norm": 1.1193803787713226,
+      "learning_rate": 9.354100886534152e-05,
+      "loss": 0.1802,
+      "step": 625
+    },
+    {
+      "epoch": 0.2473288484368817,
+      "eval_PRM Accuracy": 0.9047619047619048,
+      "eval_PRM F1": 0.9464285714285714,
+      "eval_PRM F1 AUC": 0.7058534185932119,
+      "eval_PRM F1 AUC (fixed)": 0.8524348253812101,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.8888888888888888,
+      "eval_PRM Precision": 0.905982905982906,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.34028321504592896,
+      "eval_runtime": 4.8865,
+      "eval_samples_per_second": 6.139,
+      "eval_steps_per_second": 0.205,
+      "step": 625
+    },
+    {
+      "epoch": 0.2477245745943807,
+      "grad_norm": 1.2630058603031382,
+      "learning_rate": 9.350700927162794e-05,
+      "loss": 0.1891,
+      "step": 626
+    },
+    {
+      "epoch": 0.24812030075187969,
+      "grad_norm": 1.1552702223291675,
+      "learning_rate": 9.347292663970301e-05,
+      "loss": 0.1836,
+      "step": 627
+    },
+    {
+      "epoch": 0.2485160269093787,
+      "grad_norm": 1.033400900315006,
+      "learning_rate": 9.34387610346174e-05,
+      "loss": 0.1597,
+      "step": 628
+    },
+    {
+      "epoch": 0.2489117530668777,
+      "grad_norm": 1.068934167801897,
+      "learning_rate": 9.340451252158015e-05,
+      "loss": 0.1603,
+      "step": 629
+    },
+    {
+      "epoch": 0.24930747922437674,
+      "grad_norm": 0.6160709854200137,
+      "learning_rate": 9.337018116595855e-05,
+      "loss": 0.1179,
+      "step": 630
+    },
+    {
+      "epoch": 0.24930747922437674,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.8553861288735858,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.3403157591819763,
+      "eval_runtime": 5.114,
+      "eval_samples_per_second": 5.866,
+      "eval_steps_per_second": 0.196,
+      "step": 630
+    },
+    {
+      "epoch": 0.24970320538187574,
+      "grad_norm": 0.7854758690178264,
+      "learning_rate": 9.333576703327803e-05,
+      "loss": 0.1222,
+      "step": 631
+    },
+    {
+      "epoch": 0.25009893153937474,
+      "grad_norm": 0.9830013903638098,
+      "learning_rate": 9.330127018922194e-05,
+      "loss": 0.1573,
+      "step": 632
+    },
+    {
+      "epoch": 0.2504946576968738,
+      "grad_norm": 0.8726529687641472,
+      "learning_rate": 9.326669069963156e-05,
+      "loss": 0.1647,
+      "step": 633
+    },
+    {
+      "epoch": 0.2508903838543728,
+      "grad_norm": 1.3151345227205198,
+      "learning_rate": 9.32320286305059e-05,
+      "loss": 0.2453,
+      "step": 634
+    },
+    {
+      "epoch": 0.2512861100118718,
+      "grad_norm": 0.7023228968713068,
+      "learning_rate": 9.319728404800157e-05,
+      "loss": 0.1018,
+      "step": 635
+    },
+    {
+      "epoch": 0.2512861100118718,
+      "eval_PRM Accuracy": 0.9047619047619048,
+      "eval_PRM F1": 0.9464285714285714,
+      "eval_PRM F1 AUC": 0.7058534185932119,
+      "eval_PRM F1 AUC (fixed)": 0.8598130841121495,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.8888888888888888,
+      "eval_PRM Precision": 0.905982905982906,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.3390462100505829,
+      "eval_runtime": 4.9693,
+      "eval_samples_per_second": 6.037,
+      "eval_steps_per_second": 0.201,
+      "step": 635
+    },
+    {
+      "epoch": 0.2516818361693708,
+      "grad_norm": 0.7983733224023781,
+      "learning_rate": 9.316245701843266e-05,
+      "loss": 0.1775,
+      "step": 636
+    },
+    {
+      "epoch": 0.2520775623268698,
+      "grad_norm": 1.4342366435295932,
+      "learning_rate": 9.312754760827061e-05,
+      "loss": 0.1809,
+      "step": 637
+    },
+    {
+      "epoch": 0.25247328848436884,
+      "grad_norm": 0.9750865337714575,
+      "learning_rate": 9.309255588414412e-05,
+      "loss": 0.1549,
+      "step": 638
+    },
+    {
+      "epoch": 0.25286901464186784,
+      "grad_norm": 0.9562341987446537,
+      "learning_rate": 9.305748191283898e-05,
+      "loss": 0.1988,
+      "step": 639
+    },
+    {
+      "epoch": 0.25326474079936684,
+      "grad_norm": 1.0368633553672248,
+      "learning_rate": 9.302232576129797e-05,
+      "loss": 0.118,
+      "step": 640
+    },
+    {
+      "epoch": 0.25326474079936684,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.9321266968325792,
+      "eval_PRM F1 AUC": 0.6918347270044269,
+      "eval_PRM F1 AUC (fixed)": 0.8625184456468273,
+      "eval_PRM F1 Neg": 0.5161290322580645,
+      "eval_PRM NPV": 0.6666666666666666,
+      "eval_PRM Precision": 0.9035087719298246,
+      "eval_PRM Recall": 0.9626168224299065,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.30970051884651184,
+      "eval_runtime": 5.1568,
+      "eval_samples_per_second": 5.818,
+      "eval_steps_per_second": 0.194,
+      "step": 640
+    },
+    {
+      "epoch": 0.25366046695686584,
+      "grad_norm": 0.8142380529929587,
+      "learning_rate": 9.29870874966207e-05,
+      "loss": 0.1667,
+      "step": 641
+    },
+    {
+      "epoch": 0.25405619311436484,
+      "grad_norm": 0.8805785694244057,
+      "learning_rate": 9.295176718606355e-05,
+      "loss": 0.1907,
+      "step": 642
+    },
+    {
+      "epoch": 0.2544519192718639,
+      "grad_norm": 0.9212718889915351,
+      "learning_rate": 9.291636489703943e-05,
+      "loss": 0.1987,
+      "step": 643
+    },
+    {
+      "epoch": 0.2548476454293629,
+      "grad_norm": 0.9962379537993588,
+      "learning_rate": 9.288088069711774e-05,
+      "loss": 0.2247,
+      "step": 644
+    },
+    {
+      "epoch": 0.2552433715868619,
+      "grad_norm": 1.169838474412042,
+      "learning_rate": 9.284531465402424e-05,
+      "loss": 0.2164,
+      "step": 645
+    },
+    {
+      "epoch": 0.2552433715868619,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.8632562715199213,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.3142741024494171,
+      "eval_runtime": 5.0218,
+      "eval_samples_per_second": 5.974,
+      "eval_steps_per_second": 0.199,
+      "step": 645
+    },
+    {
+      "epoch": 0.2556390977443609,
+      "grad_norm": 0.9185654433989892,
+      "learning_rate": 9.280966683564088e-05,
+      "loss": 0.2482,
+      "step": 646
+    },
+    {
+      "epoch": 0.2560348239018599,
+      "grad_norm": 1.3013352290446198,
+      "learning_rate": 9.277393731000568e-05,
+      "loss": 0.285,
+      "step": 647
+    },
+    {
+      "epoch": 0.25643055005935894,
+      "grad_norm": 0.763502671129575,
+      "learning_rate": 9.273812614531261e-05,
+      "loss": 0.1462,
+      "step": 648
+    },
+    {
+      "epoch": 0.25682627621685794,
+      "grad_norm": 0.8298355780986364,
+      "learning_rate": 9.270223340991147e-05,
+      "loss": 0.1856,
+      "step": 649
+    },
+    {
+      "epoch": 0.25722200237435694,
+      "grad_norm": 0.8506194786434039,
+      "learning_rate": 9.266625917230774e-05,
+      "loss": 0.1553,
+      "step": 650
+    },
+    {
+      "epoch": 0.25722200237435694,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.9321266968325792,
+      "eval_PRM F1 AUC": 0.6918347270044269,
+      "eval_PRM F1 AUC (fixed)": 0.8637481554353172,
+      "eval_PRM F1 Neg": 0.5161290322580645,
+      "eval_PRM NPV": 0.6666666666666666,
+      "eval_PRM Precision": 0.9035087719298246,
+      "eval_PRM Recall": 0.9626168224299065,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.28649088740348816,
+      "eval_runtime": 5.1222,
+      "eval_samples_per_second": 5.857,
+      "eval_steps_per_second": 0.195,
+      "step": 650
+    },
+    {
+      "epoch": 0.25761772853185594,
+      "grad_norm": 1.22543292326297,
+      "learning_rate": 9.263020350116247e-05,
+      "loss": 0.2312,
+      "step": 651
+    },
+    {
+      "epoch": 0.258013454689355,
+      "grad_norm": 0.7237414074956429,
+      "learning_rate": 9.259406646529209e-05,
+      "loss": 0.1754,
+      "step": 652
+    },
+    {
+      "epoch": 0.258409180846854,
+      "grad_norm": 1.0021829845702939,
+      "learning_rate": 9.255784813366837e-05,
+      "loss": 0.1796,
+      "step": 653
+    },
+    {
+      "epoch": 0.258804907004353,
+      "grad_norm": 0.8674056319285317,
+      "learning_rate": 9.252154857541825e-05,
+      "loss": 0.1464,
+      "step": 654
+    },
+    {
+      "epoch": 0.259200633161852,
+      "grad_norm": 1.1397401398742075,
+      "learning_rate": 9.248516785982364e-05,
+      "loss": 0.2552,
+      "step": 655
+    },
+    {
+      "epoch": 0.259200633161852,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.8679291687161831,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.3284505307674408,
+      "eval_runtime": 5.0817,
+      "eval_samples_per_second": 5.904,
+      "eval_steps_per_second": 0.197,
+      "step": 655
+    },
+    {
+      "epoch": 0.259596359319351,
+      "grad_norm": 0.9211209637559996,
+      "learning_rate": 9.244870605632142e-05,
+      "loss": 0.2238,
+      "step": 656
+    },
+    {
+      "epoch": 0.25999208547685004,
+      "grad_norm": 1.2824954013546763,
+      "learning_rate": 9.24121632345032e-05,
+      "loss": 0.2338,
+      "step": 657
+    },
+    {
+      "epoch": 0.26038781163434904,
+      "grad_norm": 1.0663396352881847,
+      "learning_rate": 9.237553946411519e-05,
+      "loss": 0.2045,
+      "step": 658
+    },
+    {
+      "epoch": 0.26078353779184804,
+      "grad_norm": 0.8001839653771903,
+      "learning_rate": 9.233883481505817e-05,
+      "loss": 0.2267,
+      "step": 659
+    },
+    {
+      "epoch": 0.26117926394934704,
+      "grad_norm": 0.7498852961610969,
+      "learning_rate": 9.230204935738725e-05,
+      "loss": 0.172,
+      "step": 660
+    },
+    {
+      "epoch": 0.26117926394934704,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9375,
+      "eval_PRM F1 AUC": 0.6748647319232661,
+      "eval_PRM F1 AUC (fixed)": 0.8647319232661091,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.7777777777777778,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.2878580689430237,
+      "eval_runtime": 5.1742,
+      "eval_samples_per_second": 5.798,
+      "eval_steps_per_second": 0.193,
+      "step": 660
+    },
+    {
+      "epoch": 0.26157499010684604,
+      "grad_norm": 1.0755131951508787,
+      "learning_rate": 9.226518316131176e-05,
+      "loss": 0.1972,
+      "step": 661
+    },
+    {
+      "epoch": 0.2619707162643451,
+      "grad_norm": 0.6873731378290169,
+      "learning_rate": 9.222823629719516e-05,
+      "loss": 0.1689,
+      "step": 662
+    },
+    {
+      "epoch": 0.2623664424218441,
+      "grad_norm": 0.6996977850198727,
+      "learning_rate": 9.219120883555486e-05,
+      "loss": 0.1789,
+      "step": 663
+    },
+    {
+      "epoch": 0.2627621685793431,
+      "grad_norm": 0.8330640110587905,
+      "learning_rate": 9.215410084706206e-05,
+      "loss": 0.1788,
+      "step": 664
+    },
+    {
+      "epoch": 0.2631578947368421,
+      "grad_norm": 0.6960659280191442,
+      "learning_rate": 9.211691240254173e-05,
+      "loss": 0.1782,
+      "step": 665
+    },
+    {
+      "epoch": 0.2631578947368421,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.8735858337432366,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.30374348163604736,
+      "eval_runtime": 4.9925,
+      "eval_samples_per_second": 6.009,
+      "eval_steps_per_second": 0.2,
+      "step": 665
+    },
+    {
+      "epoch": 0.2635536208943411,
+      "grad_norm": 0.636245322646594,
+      "learning_rate": 9.207964357297235e-05,
+      "loss": 0.1565,
+      "step": 666
+    },
+    {
+      "epoch": 0.26394934705184014,
+      "grad_norm": 0.7137372039946666,
+      "learning_rate": 9.204229442948585e-05,
+      "loss": 0.1796,
+      "step": 667
+    },
+    {
+      "epoch": 0.26434507320933914,
+      "grad_norm": 1.0119871222857901,
+      "learning_rate": 9.20048650433674e-05,
+      "loss": 0.2521,
+      "step": 668
+    },
+    {
+      "epoch": 0.26474079936683814,
+      "grad_norm": 0.810789129823208,
+      "learning_rate": 9.196735548605541e-05,
+      "loss": 0.1693,
+      "step": 669
+    },
+    {
+      "epoch": 0.26513652552433714,
+      "grad_norm": 1.4243984555870466,
+      "learning_rate": 9.19297658291412e-05,
+      "loss": 0.1752,
+      "step": 670
+    },
+    {
+      "epoch": 0.26513652552433714,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.8812100344318741,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.30504557490348816,
+      "eval_runtime": 5.1828,
+      "eval_samples_per_second": 5.788,
+      "eval_steps_per_second": 0.193,
+      "step": 670
+    },
+    {
+      "epoch": 0.2655322516818362,
+      "grad_norm": 0.6983153438067933,
+      "learning_rate": 9.189209614436906e-05,
+      "loss": 0.1612,
+      "step": 671
+    },
+    {
+      "epoch": 0.2659279778393352,
+      "grad_norm": 0.6844516702460982,
+      "learning_rate": 9.185434650363596e-05,
+      "loss": 0.1085,
+      "step": 672
+    },
+    {
+      "epoch": 0.2663237039968342,
+      "grad_norm": 0.9623071331666166,
+      "learning_rate": 9.181651697899152e-05,
+      "loss": 0.1929,
+      "step": 673
+    },
+    {
+      "epoch": 0.2667194301543332,
+      "grad_norm": 0.9791661878102461,
+      "learning_rate": 9.177860764263779e-05,
+      "loss": 0.1841,
+      "step": 674
+    },
+    {
+      "epoch": 0.2671151563118322,
+      "grad_norm": 1.0575489340859652,
+      "learning_rate": 9.174061856692919e-05,
+      "loss": 0.1889,
+      "step": 675
+    },
+    {
+      "epoch": 0.2671151563118322,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.9327354260089686,
+      "eval_PRM F1 AUC": 0.6701918347270044,
+      "eval_PRM F1 AUC (fixed)": 0.8888342351205116,
+      "eval_PRM F1 Neg": 0.4827586206896552,
+      "eval_PRM NPV": 0.7,
+      "eval_PRM Precision": 0.896551724137931,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.30569660663604736,
+      "eval_runtime": 5.2772,
+      "eval_samples_per_second": 5.685,
+      "eval_steps_per_second": 0.189,
+      "step": 675
+    },
+    {
+      "epoch": 0.26751088246933125,
+      "grad_norm": 0.9711032560120395,
+      "learning_rate": 9.17025498243723e-05,
+      "loss": 0.1877,
+      "step": 676
+    },
+    {
+      "epoch": 0.26790660862683024,
+      "grad_norm": 0.871454567740767,
+      "learning_rate": 9.166440148762576e-05,
+      "loss": 0.1415,
+      "step": 677
+    },
+    {
+      "epoch": 0.26830233478432924,
+      "grad_norm": 1.0870337171879758,
+      "learning_rate": 9.162617362950015e-05,
+      "loss": 0.2121,
+      "step": 678
+    },
+    {
+      "epoch": 0.26869806094182824,
+      "grad_norm": 1.0263145249551762,
+      "learning_rate": 9.158786632295776e-05,
+      "loss": 0.2044,
+      "step": 679
+    },
+    {
+      "epoch": 0.26909378709932724,
+      "grad_norm": 1.5736627103678087,
+      "learning_rate": 9.15494796411126e-05,
+      "loss": 0.2348,
+      "step": 680
+    },
+    {
+      "epoch": 0.26909378709932724,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.883177570093458,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.294921875,
+      "eval_runtime": 4.8969,
+      "eval_samples_per_second": 6.126,
+      "eval_steps_per_second": 0.204,
+      "step": 680
+    },
+    {
+      "epoch": 0.2694895132568263,
+      "grad_norm": 0.9592710551250099,
+      "learning_rate": 9.151101365723013e-05,
+      "loss": 0.1873,
+      "step": 681
+    },
+    {
+      "epoch": 0.2698852394143253,
+      "grad_norm": 1.3107103493867434,
+      "learning_rate": 9.147246844472716e-05,
+      "loss": 0.1817,
+      "step": 682
+    },
+    {
+      "epoch": 0.2702809655718243,
+      "grad_norm": 0.7689651641593306,
+      "learning_rate": 9.143384407717175e-05,
+      "loss": 0.141,
+      "step": 683
+    },
+    {
+      "epoch": 0.2706766917293233,
+      "grad_norm": 0.9300396795313893,
+      "learning_rate": 9.1395140628283e-05,
+      "loss": 0.1624,
+      "step": 684
+    },
+    {
+      "epoch": 0.2710724178868223,
+      "grad_norm": 1.1186614704956632,
+      "learning_rate": 9.1356358171931e-05,
+      "loss": 0.1516,
+      "step": 685
+    },
+    {
+      "epoch": 0.2710724178868223,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.9327354260089686,
+      "eval_PRM F1 AUC": 0.6701918347270044,
+      "eval_PRM F1 AUC (fixed)": 0.8785046728971962,
+      "eval_PRM F1 Neg": 0.4827586206896552,
+      "eval_PRM NPV": 0.7,
+      "eval_PRM Precision": 0.896551724137931,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.3026367127895355,
+      "eval_runtime": 5.1819,
+      "eval_samples_per_second": 5.789,
+      "eval_steps_per_second": 0.193,
+      "step": 685
+    },
+    {
+      "epoch": 0.27146814404432135,
+      "grad_norm": 1.254189960241517,
+      "learning_rate": 9.131749678213657e-05,
+      "loss": 0.2783,
+      "step": 686
+    },
+    {
+      "epoch": 0.27186387020182035,
+      "grad_norm": 1.0113966691212009,
+      "learning_rate": 9.127855653307123e-05,
+      "loss": 0.2462,
+      "step": 687
+    },
+    {
+      "epoch": 0.27225959635931934,
+      "grad_norm": 0.8125794573843638,
+      "learning_rate": 9.123953749905697e-05,
+      "loss": 0.1396,
+      "step": 688
+    },
+    {
+      "epoch": 0.27265532251681834,
+      "grad_norm": 0.9128450199165946,
+      "learning_rate": 9.12004397545662e-05,
+      "loss": 0.1608,
+      "step": 689
+    },
+    {
+      "epoch": 0.2730510486743174,
+      "grad_norm": 0.5624585862098771,
+      "learning_rate": 9.116126337422152e-05,
+      "loss": 0.1436,
+      "step": 690
+    },
+    {
+      "epoch": 0.2730510486743174,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.9327354260089686,
+      "eval_PRM F1 AUC": 0.6701918347270044,
+      "eval_PRM F1 AUC (fixed)": 0.8669454008853911,
+      "eval_PRM F1 Neg": 0.4827586206896552,
+      "eval_PRM NPV": 0.7,
+      "eval_PRM Precision": 0.896551724137931,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.2922200560569763,
+      "eval_runtime": 5.3417,
+      "eval_samples_per_second": 5.616,
+      "eval_steps_per_second": 0.187,
+      "step": 690
+    },
+    {
+      "epoch": 0.2734467748318164,
+      "grad_norm": 0.7591065889823867,
+      "learning_rate": 9.112200843279565e-05,
+      "loss": 0.2158,
+      "step": 691
+    },
+    {
+      "epoch": 0.2738425009893154,
+      "grad_norm": 0.9354338454813161,
+      "learning_rate": 9.108267500521121e-05,
+      "loss": 0.2026,
+      "step": 692
+    },
+    {
+      "epoch": 0.2742382271468144,
+      "grad_norm": 0.7268205891367698,
+      "learning_rate": 9.104326316654067e-05,
+      "loss": 0.1468,
+      "step": 693
+    },
+    {
+      "epoch": 0.2746339533043134,
+      "grad_norm": 1.4335552146734205,
+      "learning_rate": 9.10037729920061e-05,
+      "loss": 0.223,
+      "step": 694
+    },
+    {
+      "epoch": 0.27502967946181245,
+      "grad_norm": 0.7555211334022334,
+      "learning_rate": 9.096420455697914e-05,
+      "loss": 0.1987,
+      "step": 695
+    },
+    {
+      "epoch": 0.27502967946181245,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.8627643876045253,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.2944987118244171,
+      "eval_runtime": 5.047,
+      "eval_samples_per_second": 5.944,
+      "eval_steps_per_second": 0.198,
+      "step": 695
+    },
+    {
+      "epoch": 0.27542540561931145,
+      "grad_norm": 0.8189189658622638,
+      "learning_rate": 9.092455793698075e-05,
+      "loss": 0.1846,
+      "step": 696
+    },
+    {
+      "epoch": 0.27582113177681045,
+      "grad_norm": 0.6793424965135729,
+      "learning_rate": 9.088483320768115e-05,
+      "loss": 0.1841,
+      "step": 697
+    },
+    {
+      "epoch": 0.27621685793430945,
+      "grad_norm": 1.1346566755573384,
+      "learning_rate": 9.08450304448996e-05,
+      "loss": 0.2525,
+      "step": 698
+    },
+    {
+      "epoch": 0.27661258409180844,
+      "grad_norm": 0.6534103036603515,
+      "learning_rate": 9.080514972460439e-05,
+      "loss": 0.1462,
+      "step": 699
+    },
+    {
+      "epoch": 0.2770083102493075,
+      "grad_norm": 0.7628441572243372,
+      "learning_rate": 9.076519112291246e-05,
+      "loss": 0.1739,
+      "step": 700
+    },
+    {
+      "epoch": 0.2770083102493075,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.8750614854894245,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.2897135317325592,
+      "eval_runtime": 5.2861,
+      "eval_samples_per_second": 5.675,
+      "eval_steps_per_second": 0.189,
+      "step": 700
+    },
+    {
+      "epoch": 0.2774040364068065,
+      "grad_norm": 0.6607457838568496,
+      "learning_rate": 9.07251547160895e-05,
+      "loss": 0.1313,
+      "step": 701
+    },
+    {
+      "epoch": 0.2777997625643055,
+      "grad_norm": 0.8533872530467852,
+      "learning_rate": 9.06850405805497e-05,
+      "loss": 0.1595,
+      "step": 702
+    },
+    {
+      "epoch": 0.2781954887218045,
+      "grad_norm": 0.6476126920331056,
+      "learning_rate": 9.064484879285555e-05,
+      "loss": 0.1456,
+      "step": 703
+    },
+    {
+      "epoch": 0.2785912148793035,
+      "grad_norm": 0.982952385932065,
+      "learning_rate": 9.060457942971776e-05,
+      "loss": 0.2164,
+      "step": 704
+    },
+    {
+      "epoch": 0.27898694103680255,
+      "grad_norm": 0.7667149787248076,
+      "learning_rate": 9.056423256799513e-05,
+      "loss": 0.1383,
+      "step": 705
+    },
+    {
+      "epoch": 0.27898694103680255,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.883669454008854,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.2916015684604645,
+      "eval_runtime": 4.9834,
+      "eval_samples_per_second": 6.02,
+      "eval_steps_per_second": 0.201,
+      "step": 705
+    },
+    {
+      "epoch": 0.27938266719430155,
+      "grad_norm": 0.9769531041036345,
+      "learning_rate": 9.052380828469436e-05,
+      "loss": 0.1637,
+      "step": 706
+    },
+    {
+      "epoch": 0.27977839335180055,
+      "grad_norm": 0.9214654827890637,
+      "learning_rate": 9.048330665696993e-05,
+      "loss": 0.1894,
+      "step": 707
+    },
+    {
+      "epoch": 0.28017411950929955,
+      "grad_norm": 1.02055901185936,
+      "learning_rate": 9.04427277621239e-05,
+      "loss": 0.2138,
+      "step": 708
+    },
+    {
+      "epoch": 0.2805698456667986,
+      "grad_norm": 1.638505514806321,
+      "learning_rate": 9.040207167760586e-05,
+      "loss": 0.293,
+      "step": 709
+    },
+    {
+      "epoch": 0.2809655718242976,
+      "grad_norm": 0.5103129313867601,
+      "learning_rate": 9.036133848101269e-05,
+      "loss": 0.1144,
+      "step": 710
+    },
+    {
+      "epoch": 0.2809655718242976,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.883423512051156,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.271484375,
+      "eval_runtime": 5.0799,
+      "eval_samples_per_second": 5.906,
+      "eval_steps_per_second": 0.197,
+      "step": 710
+    },
+    {
+      "epoch": 0.2813612979817966,
+      "grad_norm": 0.851892128891332,
+      "learning_rate": 9.032052825008845e-05,
+      "loss": 0.1526,
+      "step": 711
+    },
+    {
+      "epoch": 0.2817570241392956,
+      "grad_norm": 0.8483666128644993,
+      "learning_rate": 9.027964106272423e-05,
+      "loss": 0.1492,
+      "step": 712
+    },
+    {
+      "epoch": 0.2821527502967946,
+      "grad_norm": 0.7632322196341024,
+      "learning_rate": 9.023867699695804e-05,
+      "loss": 0.1646,
+      "step": 713
+    },
+    {
+      "epoch": 0.28254847645429365,
+      "grad_norm": 0.7612194930893317,
+      "learning_rate": 9.019763613097455e-05,
+      "loss": 0.1688,
+      "step": 714
+    },
+    {
+      "epoch": 0.28294420261179265,
+      "grad_norm": 1.1141932691697012,
+      "learning_rate": 9.015651854310506e-05,
+      "loss": 0.2438,
+      "step": 715
+    },
+    {
+      "epoch": 0.28294420261179265,
+      "eval_PRM Accuracy": 0.9047619047619048,
+      "eval_PRM F1": 0.9464285714285714,
+      "eval_PRM F1 AUC": 0.7058534185932119,
+      "eval_PRM F1 AUC (fixed)": 0.8876045253320216,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.8888888888888888,
+      "eval_PRM Precision": 0.905982905982906,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.29472655057907104,
+      "eval_runtime": 4.9817,
+      "eval_samples_per_second": 6.022,
+      "eval_steps_per_second": 0.201,
+      "step": 715
+    },
+    {
+      "epoch": 0.28333992876929165,
+      "grad_norm": 1.1587755741290395,
+      "learning_rate": 9.011532431182729e-05,
+      "loss": 0.1748,
+      "step": 716
+    },
+    {
+      "epoch": 0.28373565492679065,
+      "grad_norm": 1.5392021948756802,
+      "learning_rate": 9.007405351576524e-05,
+      "loss": 0.3444,
+      "step": 717
+    },
+    {
+      "epoch": 0.28413138108428965,
+      "grad_norm": 0.7938171697497031,
+      "learning_rate": 9.003270623368905e-05,
+      "loss": 0.1653,
+      "step": 718
+    },
+    {
+      "epoch": 0.2845271072417887,
+      "grad_norm": 1.0671784206576023,
+      "learning_rate": 8.999128254451486e-05,
+      "loss": 0.2085,
+      "step": 719
+    },
+    {
+      "epoch": 0.2849228333992877,
+      "grad_norm": 0.8571675690584678,
+      "learning_rate": 8.99497825273046e-05,
+      "loss": 0.2005,
+      "step": 720
+    },
+    {
+      "epoch": 0.2849228333992877,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.8772749631087065,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.27314454317092896,
+      "eval_runtime": 5.1322,
+      "eval_samples_per_second": 5.845,
+      "eval_steps_per_second": 0.195,
+      "step": 720
+    },
+    {
+      "epoch": 0.2853185595567867,
+      "grad_norm": 0.7020104268585275,
+      "learning_rate": 8.990820626126589e-05,
+      "loss": 0.1178,
+      "step": 721
+    },
+    {
+      "epoch": 0.2857142857142857,
+      "grad_norm": 0.551517656772806,
+      "learning_rate": 8.986655382575192e-05,
+      "loss": 0.1306,
+      "step": 722
+    },
+    {
+      "epoch": 0.2861100118717847,
+      "grad_norm": 0.835937866856043,
+      "learning_rate": 8.982482530026122e-05,
+      "loss": 0.1967,
+      "step": 723
+    },
+    {
+      "epoch": 0.28650573802928375,
+      "grad_norm": 0.8368102251324648,
+      "learning_rate": 8.978302076443754e-05,
+      "loss": 0.1727,
+      "step": 724
+    },
+    {
+      "epoch": 0.28690146418678275,
+      "grad_norm": 0.8660458425103343,
+      "learning_rate": 8.974114029806976e-05,
+      "loss": 0.2213,
+      "step": 725
+    },
+    {
+      "epoch": 0.28690146418678275,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.9321266968325792,
+      "eval_PRM F1 AUC": 0.6918347270044269,
+      "eval_PRM F1 AUC (fixed)": 0.8856369896704377,
+      "eval_PRM F1 Neg": 0.5161290322580645,
+      "eval_PRM NPV": 0.6666666666666666,
+      "eval_PRM Precision": 0.9035087719298246,
+      "eval_PRM Recall": 0.9626168224299065,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2822916805744171,
+      "eval_runtime": 5.1756,
+      "eval_samples_per_second": 5.796,
+      "eval_steps_per_second": 0.193,
+      "step": 725
+    },
+    {
+      "epoch": 0.28729719034428175,
+      "grad_norm": 1.4395936296050706,
+      "learning_rate": 8.969918398109162e-05,
+      "loss": 0.2175,
+      "step": 726
+    },
+    {
+      "epoch": 0.28769291650178075,
+      "grad_norm": 0.8392540153410275,
+      "learning_rate": 8.965715189358164e-05,
+      "loss": 0.2052,
+      "step": 727
+    },
+    {
+      "epoch": 0.2880886426592798,
+      "grad_norm": 0.8498842269373648,
+      "learning_rate": 8.9615044115763e-05,
+      "loss": 0.23,
+      "step": 728
+    },
+    {
+      "epoch": 0.2884843688167788,
+      "grad_norm": 1.1484022806367529,
+      "learning_rate": 8.95728607280033e-05,
+      "loss": 0.1833,
+      "step": 729
+    },
+    {
+      "epoch": 0.2888800949742778,
+      "grad_norm": 0.6436287635235969,
+      "learning_rate": 8.953060181081447e-05,
+      "loss": 0.1945,
+      "step": 730
+    },
+    {
+      "epoch": 0.2888800949742778,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.9327354260089686,
+      "eval_PRM F1 AUC": 0.6701918347270044,
+      "eval_PRM F1 AUC (fixed)": 0.8917855386128873,
+      "eval_PRM F1 Neg": 0.4827586206896552,
+      "eval_PRM NPV": 0.7,
+      "eval_PRM Precision": 0.896551724137931,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.31298828125,
+      "eval_runtime": 5.1871,
+      "eval_samples_per_second": 5.784,
+      "eval_steps_per_second": 0.193,
+      "step": 730
+    },
+    {
+      "epoch": 0.2892758211317768,
+      "grad_norm": 0.7279521368670537,
+      "learning_rate": 8.948826744485258e-05,
+      "loss": 0.1433,
+      "step": 731
+    },
+    {
+      "epoch": 0.2896715472892758,
+      "grad_norm": 0.9415593375937553,
+      "learning_rate": 8.944585771091773e-05,
+      "loss": 0.1826,
+      "step": 732
+    },
+    {
+      "epoch": 0.29006727344677485,
+      "grad_norm": 0.8353676957085796,
+      "learning_rate": 8.940337268995385e-05,
+      "loss": 0.1902,
+      "step": 733
+    },
+    {
+      "epoch": 0.29046299960427385,
+      "grad_norm": 1.1591437479608018,
+      "learning_rate": 8.936081246304855e-05,
+      "loss": 0.2265,
+      "step": 734
+    },
+    {
+      "epoch": 0.29085872576177285,
+      "grad_norm": 1.0589753650535267,
+      "learning_rate": 8.931817711143302e-05,
+      "loss": 0.2601,
+      "step": 735
+    },
+    {
+      "epoch": 0.29085872576177285,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.9327354260089686,
+      "eval_PRM F1 AUC": 0.6701918347270044,
+      "eval_PRM F1 AUC (fixed)": 0.8922774225282833,
+      "eval_PRM F1 Neg": 0.4827586206896552,
+      "eval_PRM NPV": 0.7,
+      "eval_PRM Precision": 0.896551724137931,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.29293620586395264,
+      "eval_runtime": 5.1837,
+      "eval_samples_per_second": 5.787,
+      "eval_steps_per_second": 0.193,
+      "step": 735
+    },
+    {
+      "epoch": 0.29125445191927185,
+      "grad_norm": 0.9466532951110602,
+      "learning_rate": 8.92754667164818e-05,
+      "loss": 0.269,
+      "step": 736
+    },
+    {
+      "epoch": 0.29165017807677085,
+      "grad_norm": 0.5546897117290948,
+      "learning_rate": 8.923268135971267e-05,
+      "loss": 0.1516,
+      "step": 737
+    },
+    {
+      "epoch": 0.2920459042342699,
+      "grad_norm": 0.726795549085925,
+      "learning_rate": 8.91898211227865e-05,
+      "loss": 0.1738,
+      "step": 738
+    },
+    {
+      "epoch": 0.2924416303917689,
+      "grad_norm": 0.8758479495103402,
+      "learning_rate": 8.914688608750701e-05,
+      "loss": 0.1995,
+      "step": 739
+    },
+    {
+      "epoch": 0.2928373565492679,
+      "grad_norm": 0.9058601441832761,
+      "learning_rate": 8.91038763358208e-05,
+      "loss": 0.2906,
+      "step": 740
+    },
+    {
+      "epoch": 0.2928373565492679,
+      "eval_PRM Accuracy": 0.8650793650793651,
+      "eval_PRM F1": 0.9251101321585903,
+      "eval_PRM F1 AUC": 0.5959173635022135,
+      "eval_PRM F1 AUC (fixed)": 0.8883423512051157,
+      "eval_PRM F1 Neg": 0.32,
+      "eval_PRM NPV": 0.6666666666666666,
+      "eval_PRM Precision": 0.875,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.21052631578947367,
+      "eval_loss": 0.2775065004825592,
+      "eval_runtime": 5.1279,
+      "eval_samples_per_second": 5.85,
+      "eval_steps_per_second": 0.195,
+      "step": 740
+    },
+    {
+      "epoch": 0.2932330827067669,
+      "grad_norm": 0.9063094013514543,
+      "learning_rate": 8.906079194981695e-05,
+      "loss": 0.1671,
+      "step": 741
+    },
+    {
+      "epoch": 0.29362880886426596,
+      "grad_norm": 0.6651368106596527,
+      "learning_rate": 8.901763301172708e-05,
+      "loss": 0.1513,
+      "step": 742
+    },
+    {
+      "epoch": 0.29402453502176495,
+      "grad_norm": 0.5973154384515552,
+      "learning_rate": 8.897439960392507e-05,
+      "loss": 0.1293,
+      "step": 743
+    },
+    {
+      "epoch": 0.29442026117926395,
+      "grad_norm": 0.8955840947808993,
+      "learning_rate": 8.89310918089269e-05,
+      "loss": 0.2392,
+      "step": 744
+    },
+    {
+      "epoch": 0.29481598733676295,
+      "grad_norm": 0.6206958893144313,
+      "learning_rate": 8.888770970939057e-05,
+      "loss": 0.1787,
+      "step": 745
+    },
+    {
+      "epoch": 0.29481598733676295,
+      "eval_PRM Accuracy": 0.873015873015873,
+      "eval_PRM F1": 0.9298245614035088,
+      "eval_PRM F1 AUC": 0.6005902606984752,
+      "eval_PRM F1 AUC (fixed)": 0.8910477127397934,
+      "eval_PRM F1 Neg": 0.3333333333333333,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.8760330578512396,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.21052631578947367,
+      "eval_loss": 0.30009764432907104,
+      "eval_runtime": 4.8482,
+      "eval_samples_per_second": 6.188,
+      "eval_steps_per_second": 0.206,
+      "step": 745
+    },
+    {
+      "epoch": 0.29521171349426195,
+      "grad_norm": 0.6786888424779063,
+      "learning_rate": 8.88442533881159e-05,
+      "loss": 0.1603,
+      "step": 746
+    },
+    {
+      "epoch": 0.295607439651761,
+      "grad_norm": 0.9414471796153763,
+      "learning_rate": 8.880072292804435e-05,
+      "loss": 0.1964,
+      "step": 747
+    },
+    {
+      "epoch": 0.29600316580926,
+      "grad_norm": 0.7837874338689773,
+      "learning_rate": 8.875711841225888e-05,
+      "loss": 0.2305,
+      "step": 748
+    },
+    {
+      "epoch": 0.296398891966759,
+      "grad_norm": 1.2912999325498091,
+      "learning_rate": 8.871343992398383e-05,
+      "loss": 0.2515,
+      "step": 749
+    },
+    {
+      "epoch": 0.296794618124258,
+      "grad_norm": 0.9026489779277271,
+      "learning_rate": 8.866968754658466e-05,
+      "loss": 0.206,
+      "step": 750
+    },
+    {
+      "epoch": 0.296794618124258,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.8893261190359075,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.26956379413604736,
+      "eval_runtime": 5.3776,
+      "eval_samples_per_second": 5.579,
+      "eval_steps_per_second": 0.186,
+      "step": 750
+    },
+    {
+      "epoch": 0.297190344281757,
+      "grad_norm": 0.7696797543394589,
+      "learning_rate": 8.862586136356794e-05,
+      "loss": 0.1674,
+      "step": 751
+    },
+    {
+      "epoch": 0.29758607043925606,
+      "grad_norm": 0.9602999971402979,
+      "learning_rate": 8.858196145858104e-05,
+      "loss": 0.1898,
+      "step": 752
+    },
+    {
+      "epoch": 0.29798179659675506,
+      "grad_norm": 0.8828895719430124,
+      "learning_rate": 8.853798791541204e-05,
+      "loss": 0.2273,
+      "step": 753
+    },
+    {
+      "epoch": 0.29837752275425405,
+      "grad_norm": 0.9976380413908069,
+      "learning_rate": 8.849394081798962e-05,
+      "loss": 0.1945,
+      "step": 754
+    },
+    {
+      "epoch": 0.29877324891175305,
+      "grad_norm": 0.8342263917602064,
+      "learning_rate": 8.844982025038279e-05,
+      "loss": 0.2335,
+      "step": 755
+    },
+    {
+      "epoch": 0.29877324891175305,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9375,
+      "eval_PRM F1 AUC": 0.6748647319232661,
+      "eval_PRM F1 AUC (fixed)": 0.8893261190359075,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.7777777777777778,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.2806640565395355,
+      "eval_runtime": 4.8594,
+      "eval_samples_per_second": 6.174,
+      "eval_steps_per_second": 0.206,
+      "step": 755
+    },
+    {
+      "epoch": 0.29916897506925205,
+      "grad_norm": 0.9151882315488452,
+      "learning_rate": 8.84056262968008e-05,
+      "loss": 0.2161,
+      "step": 756
+    },
+    {
+      "epoch": 0.2995647012267511,
+      "grad_norm": 0.8459616664970567,
+      "learning_rate": 8.836135904159302e-05,
+      "loss": 0.1541,
+      "step": 757
+    },
+    {
+      "epoch": 0.2999604273842501,
+      "grad_norm": 0.6746795344350985,
+      "learning_rate": 8.831701856924864e-05,
+      "loss": 0.162,
+      "step": 758
+    },
+    {
+      "epoch": 0.3003561535417491,
+      "grad_norm": 0.7654457054114777,
+      "learning_rate": 8.827260496439662e-05,
+      "loss": 0.1826,
+      "step": 759
+    },
+    {
+      "epoch": 0.3007518796992481,
+      "grad_norm": 1.1923245027585447,
+      "learning_rate": 8.822811831180555e-05,
+      "loss": 0.2223,
+      "step": 760
+    },
+    {
+      "epoch": 0.3007518796992481,
+      "eval_PRM Accuracy": 0.8650793650793651,
+      "eval_PRM F1": 0.925764192139738,
+      "eval_PRM F1 AUC": 0.5742744712247909,
+      "eval_PRM F1 AUC (fixed)": 0.8925233644859814,
+      "eval_PRM F1 Neg": 0.2608695652173913,
+      "eval_PRM NPV": 0.75,
+      "eval_PRM Precision": 0.8688524590163934,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.15789473684210525,
+      "eval_loss": 0.3451497256755829,
+      "eval_runtime": 5.0736,
+      "eval_samples_per_second": 5.913,
+      "eval_steps_per_second": 0.197,
+      "step": 760
+    },
+    {
+      "epoch": 0.30114760585674716,
+      "grad_norm": 1.315315242050645,
+      "learning_rate": 8.818355869638339e-05,
+      "loss": 0.2256,
+      "step": 761
+    },
+    {
+      "epoch": 0.30154333201424616,
+      "grad_norm": 0.7026193936703187,
+      "learning_rate": 8.81389262031774e-05,
+      "loss": 0.2259,
+      "step": 762
+    },
+    {
+      "epoch": 0.30193905817174516,
+      "grad_norm": 0.7378162205814667,
+      "learning_rate": 8.809422091737387e-05,
+      "loss": 0.1867,
+      "step": 763
+    },
+    {
+      "epoch": 0.30233478432924416,
+      "grad_norm": 0.83139474139224,
+      "learning_rate": 8.804944292429807e-05,
+      "loss": 0.2439,
+      "step": 764
+    },
+    {
+      "epoch": 0.30273051048674315,
+      "grad_norm": 0.6124547932922225,
+      "learning_rate": 8.800459230941405e-05,
+      "loss": 0.1712,
+      "step": 765
+    },
+    {
+      "epoch": 0.30273051048674315,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9380530973451328,
+      "eval_PRM F1 AUC": 0.6532218396458436,
+      "eval_PRM F1 AUC (fixed)": 0.8812100344318741,
+      "eval_PRM F1 Neg": 0.46153846153846156,
+      "eval_PRM NPV": 0.8571428571428571,
+      "eval_PRM Precision": 0.8907563025210085,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3157894736842105,
+      "eval_loss": 0.27991536259651184,
+      "eval_runtime": 5.2088,
+      "eval_samples_per_second": 5.759,
+      "eval_steps_per_second": 0.192,
+      "step": 765
+    },
+    {
+      "epoch": 0.3031262366442422,
+      "grad_norm": 0.8159197300922938,
+      "learning_rate": 8.795966915832443e-05,
+      "loss": 0.182,
+      "step": 766
+    },
+    {
+      "epoch": 0.3035219628017412,
+      "grad_norm": 0.7908198253381381,
+      "learning_rate": 8.791467355677031e-05,
+      "loss": 0.2122,
+      "step": 767
+    },
+    {
+      "epoch": 0.3039176889592402,
+      "grad_norm": 0.7234911177006348,
+      "learning_rate": 8.786960559063105e-05,
+      "loss": 0.222,
+      "step": 768
+    },
+    {
+      "epoch": 0.3043134151167392,
+      "grad_norm": 0.7952637819088901,
+      "learning_rate": 8.782446534592413e-05,
+      "loss": 0.2116,
+      "step": 769
+    },
+    {
+      "epoch": 0.3047091412742382,
+      "grad_norm": 0.9612974289352592,
+      "learning_rate": 8.777925290880496e-05,
+      "loss": 0.1918,
+      "step": 770
+    },
+    {
+      "epoch": 0.3047091412742382,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9380530973451328,
+      "eval_PRM F1 AUC": 0.6532218396458436,
+      "eval_PRM F1 AUC (fixed)": 0.8804722085587801,
+      "eval_PRM F1 Neg": 0.46153846153846156,
+      "eval_PRM NPV": 0.8571428571428571,
+      "eval_PRM Precision": 0.8907563025210085,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3157894736842105,
+      "eval_loss": 0.28177082538604736,
+      "eval_runtime": 4.9519,
+      "eval_samples_per_second": 6.058,
+      "eval_steps_per_second": 0.202,
+      "step": 770
+    },
+    {
+      "epoch": 0.30510486743173726,
+      "grad_norm": 0.9087054032483611,
+      "learning_rate": 8.773396836556679e-05,
+      "loss": 0.2429,
+      "step": 771
+    },
+    {
+      "epoch": 0.30550059358923626,
+      "grad_norm": 0.7447575951133999,
+      "learning_rate": 8.768861180264045e-05,
+      "loss": 0.2014,
+      "step": 772
+    },
+    {
+      "epoch": 0.30589631974673526,
+      "grad_norm": 0.9069651617022895,
+      "learning_rate": 8.764318330659424e-05,
+      "loss": 0.2401,
+      "step": 773
+    },
+    {
+      "epoch": 0.30629204590423426,
+      "grad_norm": 0.9521867301096485,
+      "learning_rate": 8.759768296413376e-05,
+      "loss": 0.1792,
+      "step": 774
+    },
+    {
+      "epoch": 0.30668777206173325,
+      "grad_norm": 0.6072237551406509,
+      "learning_rate": 8.755211086210172e-05,
+      "loss": 0.1562,
+      "step": 775
+    },
+    {
+      "epoch": 0.30668777206173325,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9380530973451328,
+      "eval_PRM F1 AUC": 0.6532218396458436,
+      "eval_PRM F1 AUC (fixed)": 0.8799803246433842,
+      "eval_PRM F1 Neg": 0.46153846153846156,
+      "eval_PRM NPV": 0.8571428571428571,
+      "eval_PRM Precision": 0.8907563025210085,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3157894736842105,
+      "eval_loss": 0.306640625,
+      "eval_runtime": 5.0011,
+      "eval_samples_per_second": 5.999,
+      "eval_steps_per_second": 0.2,
+      "step": 775
+    },
+    {
+      "epoch": 0.3070834982192323,
+      "grad_norm": 0.8792892279462375,
+      "learning_rate": 8.750646708747781e-05,
+      "loss": 0.1507,
+      "step": 776
+    },
+    {
+      "epoch": 0.3074792243767313,
+      "grad_norm": 0.6571602790522012,
+      "learning_rate": 8.746075172737851e-05,
+      "loss": 0.1421,
+      "step": 777
+    },
+    {
+      "epoch": 0.3078749505342303,
+      "grad_norm": 1.0283840524578438,
+      "learning_rate": 8.741496486905691e-05,
+      "loss": 0.1422,
+      "step": 778
+    },
+    {
+      "epoch": 0.3082706766917293,
+      "grad_norm": 0.6998294955241987,
+      "learning_rate": 8.736910659990261e-05,
+      "loss": 0.1464,
+      "step": 779
+    },
+    {
+      "epoch": 0.30866640284922836,
+      "grad_norm": 1.2720310775742825,
+      "learning_rate": 8.732317700744146e-05,
+      "loss": 0.2327,
+      "step": 780
+    },
+    {
+      "epoch": 0.30866640284922836,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.8900639449090014,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2822265625,
+      "eval_runtime": 5.103,
+      "eval_samples_per_second": 5.879,
+      "eval_steps_per_second": 0.196,
+      "step": 780
+    },
+    {
+      "epoch": 0.30906212900672736,
+      "grad_norm": 0.7342901789664035,
+      "learning_rate": 8.727717617933544e-05,
+      "loss": 0.1264,
+      "step": 781
+    },
+    {
+      "epoch": 0.30945785516422636,
+      "grad_norm": 1.1028064366961383,
+      "learning_rate": 8.723110420338251e-05,
+      "loss": 0.2212,
+      "step": 782
+    },
+    {
+      "epoch": 0.30985358132172536,
+      "grad_norm": 1.2893229878786714,
+      "learning_rate": 8.718496116751644e-05,
+      "loss": 0.1484,
+      "step": 783
+    },
+    {
+      "epoch": 0.31024930747922436,
+      "grad_norm": 0.5643435084286054,
+      "learning_rate": 8.71387471598066e-05,
+      "loss": 0.0959,
+      "step": 784
+    },
+    {
+      "epoch": 0.3106450336367234,
+      "grad_norm": 0.7583173597008048,
+      "learning_rate": 8.709246226845782e-05,
+      "loss": 0.1048,
+      "step": 785
+    },
+    {
+      "epoch": 0.3106450336367234,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.8999016232169208,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.28942057490348816,
+      "eval_runtime": 5.0981,
+      "eval_samples_per_second": 5.885,
+      "eval_steps_per_second": 0.196,
+      "step": 785
+    },
+    {
+      "epoch": 0.3110407597942224,
+      "grad_norm": 0.8073987997530575,
+      "learning_rate": 8.704610658181021e-05,
+      "loss": 0.1581,
+      "step": 786
+    },
+    {
+      "epoch": 0.3114364859517214,
+      "grad_norm": 1.114560450506898,
+      "learning_rate": 8.699968018833904e-05,
+      "loss": 0.1419,
+      "step": 787
+    },
+    {
+      "epoch": 0.3118322121092204,
+      "grad_norm": 0.7316494749476418,
+      "learning_rate": 8.69531831766545e-05,
+      "loss": 0.0942,
+      "step": 788
+    },
+    {
+      "epoch": 0.3122279382667194,
+      "grad_norm": 2.010956156016771,
+      "learning_rate": 8.690661563550156e-05,
+      "loss": 0.2824,
+      "step": 789
+    },
+    {
+      "epoch": 0.31262366442421846,
+      "grad_norm": 1.153707937935504,
+      "learning_rate": 8.68599776537598e-05,
+      "loss": 0.2088,
+      "step": 790
+    },
+    {
+      "epoch": 0.31262366442421846,
+      "eval_PRM Accuracy": 0.873015873015873,
+      "eval_PRM F1": 0.9285714285714286,
+      "eval_PRM F1 AUC": 0.6438760452533202,
+      "eval_PRM F1 AUC (fixed)": 0.8991637973438268,
+      "eval_PRM F1 Neg": 0.42857142857142855,
+      "eval_PRM NPV": 0.6666666666666666,
+      "eval_PRM Precision": 0.8888888888888888,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.3157894736842105,
+      "eval_loss": 0.30595701932907104,
+      "eval_runtime": 4.7699,
+      "eval_samples_per_second": 6.289,
+      "eval_steps_per_second": 0.21,
+      "step": 790
+    },
+    {
+      "epoch": 0.31301939058171746,
+      "grad_norm": 1.1484907644773685,
+      "learning_rate": 8.68132693204433e-05,
+      "loss": 0.2056,
+      "step": 791
+    },
+    {
+      "epoch": 0.31341511673921646,
+      "grad_norm": 1.6777980586879817,
+      "learning_rate": 8.676649072470034e-05,
+      "loss": 0.1962,
+      "step": 792
+    },
+    {
+      "epoch": 0.31381084289671546,
+      "grad_norm": 0.7395973147724139,
+      "learning_rate": 8.671964195581336e-05,
+      "loss": 0.1298,
+      "step": 793
+    },
+    {
+      "epoch": 0.31420656905421446,
+      "grad_norm": 0.9643682491142119,
+      "learning_rate": 8.66727231031987e-05,
+      "loss": 0.171,
+      "step": 794
+    },
+    {
+      "epoch": 0.3146022952117135,
+      "grad_norm": 0.8899614793760491,
+      "learning_rate": 8.662573425640645e-05,
+      "loss": 0.165,
+      "step": 795
+    },
+    {
+      "epoch": 0.3146022952117135,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.8962124938514511,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2797200381755829,
+      "eval_runtime": 4.8426,
+      "eval_samples_per_second": 6.195,
+      "eval_steps_per_second": 0.207,
+      "step": 795
+    },
+    {
+      "epoch": 0.3149980213692125,
+      "grad_norm": 0.9323404131120511,
+      "learning_rate": 8.657867550512033e-05,
+      "loss": 0.1627,
+      "step": 796
+    },
+    {
+      "epoch": 0.3153937475267115,
+      "grad_norm": 0.7969863305269954,
+      "learning_rate": 8.65315469391575e-05,
+      "loss": 0.1364,
+      "step": 797
+    },
+    {
+      "epoch": 0.3157894736842105,
+      "grad_norm": 0.6073649742768036,
+      "learning_rate": 8.64843486484683e-05,
+      "loss": 0.1061,
+      "step": 798
+    },
+    {
+      "epoch": 0.31618519984170956,
+      "grad_norm": 0.9190067904070243,
+      "learning_rate": 8.643708072313618e-05,
+      "loss": 0.2136,
+      "step": 799
+    },
+    {
+      "epoch": 0.31658092599920856,
+      "grad_norm": 0.5106188430854406,
+      "learning_rate": 8.63897432533775e-05,
+      "loss": 0.1453,
+      "step": 800
+    },
+    {
+      "epoch": 0.31658092599920856,
+      "eval_PRM Accuracy": 0.873015873015873,
+      "eval_PRM F1": 0.9292035398230089,
+      "eval_PRM F1 AUC": 0.6222331529758978,
+      "eval_PRM F1 AUC (fixed)": 0.8942449581898672,
+      "eval_PRM F1 Neg": 0.38461538461538464,
+      "eval_PRM NPV": 0.7142857142857143,
+      "eval_PRM Precision": 0.8823529411764706,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.2631578947368421,
+      "eval_loss": 0.3200846314430237,
+      "eval_runtime": 5.0855,
+      "eval_samples_per_second": 5.899,
+      "eval_steps_per_second": 0.197,
+      "step": 800
+    },
+    {
+      "epoch": 0.31697665215670756,
+      "grad_norm": 0.7289283720156537,
+      "learning_rate": 8.634233632954139e-05,
+      "loss": 0.1527,
+      "step": 801
+    },
+    {
+      "epoch": 0.31737237831420656,
+      "grad_norm": 0.8570109277701821,
+      "learning_rate": 8.629486004210945e-05,
+      "loss": 0.1973,
+      "step": 802
+    },
+    {
+      "epoch": 0.31776810447170556,
+      "grad_norm": 0.7857443785523225,
+      "learning_rate": 8.624731448169576e-05,
+      "loss": 0.1507,
+      "step": 803
+    },
+    {
+      "epoch": 0.3181638306292046,
+      "grad_norm": 0.9553421553523828,
+      "learning_rate": 8.619969973904655e-05,
+      "loss": 0.2024,
+      "step": 804
+    },
+    {
+      "epoch": 0.3185595567867036,
+      "grad_norm": 0.6153155522590844,
+      "learning_rate": 8.615201590504017e-05,
+      "loss": 0.1317,
+      "step": 805
+    },
+    {
+      "epoch": 0.3185595567867036,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9375,
+      "eval_PRM F1 AUC": 0.6748647319232661,
+      "eval_PRM F1 AUC (fixed)": 0.8866207575012297,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.7777777777777778,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.3003906309604645,
+      "eval_runtime": 5.1121,
+      "eval_samples_per_second": 5.868,
+      "eval_steps_per_second": 0.196,
+      "step": 805
+    },
+    {
+      "epoch": 0.3189552829442026,
+      "grad_norm": 0.7682299096727117,
+      "learning_rate": 8.610426307068674e-05,
+      "loss": 0.1388,
+      "step": 806
+    },
+    {
+      "epoch": 0.3193510091017016,
+      "grad_norm": 0.8666234152885707,
+      "learning_rate": 8.605644132712814e-05,
+      "loss": 0.2311,
+      "step": 807
+    },
+    {
+      "epoch": 0.3197467352592006,
+      "grad_norm": 1.5250090898838904,
+      "learning_rate": 8.600855076563776e-05,
+      "loss": 0.2397,
+      "step": 808
+    },
+    {
+      "epoch": 0.32014246141669966,
+      "grad_norm": 1.045702461189082,
+      "learning_rate": 8.596059147762034e-05,
+      "loss": 0.1861,
+      "step": 809
+    },
+    {
+      "epoch": 0.32053818757419866,
+      "grad_norm": 0.8749524500579257,
+      "learning_rate": 8.591256355461176e-05,
+      "loss": 0.1504,
+      "step": 810
+    },
+    {
+      "epoch": 0.32053818757419866,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.8863748155435317,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2663411498069763,
+      "eval_runtime": 5.168,
+      "eval_samples_per_second": 5.805,
+      "eval_steps_per_second": 0.193,
+      "step": 810
+    },
+    {
+      "epoch": 0.32093391373169766,
+      "grad_norm": 1.383279312317896,
+      "learning_rate": 8.586446708827896e-05,
+      "loss": 0.1709,
+      "step": 811
+    },
+    {
+      "epoch": 0.32132963988919666,
+      "grad_norm": 0.8403605155927515,
+      "learning_rate": 8.581630217041963e-05,
+      "loss": 0.1964,
+      "step": 812
+    },
+    {
+      "epoch": 0.32172536604669566,
+      "grad_norm": 1.0912036401913052,
+      "learning_rate": 8.576806889296216e-05,
+      "loss": 0.1711,
+      "step": 813
+    },
+    {
+      "epoch": 0.3221210922041947,
+      "grad_norm": 1.7141470559355845,
+      "learning_rate": 8.57197673479654e-05,
+      "loss": 0.2641,
+      "step": 814
+    },
+    {
+      "epoch": 0.3225168183616937,
+      "grad_norm": 1.0645503918252057,
+      "learning_rate": 8.56713976276185e-05,
+      "loss": 0.2157,
+      "step": 815
+    },
+    {
+      "epoch": 0.3225168183616937,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.9035907525823906,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2689778506755829,
+      "eval_runtime": 5.0017,
+      "eval_samples_per_second": 5.998,
+      "eval_steps_per_second": 0.2,
+      "step": 815
+    },
+    {
+      "epoch": 0.3229125445191927,
+      "grad_norm": 0.7009662996692828,
+      "learning_rate": 8.562295982424069e-05,
+      "loss": 0.14,
+      "step": 816
+    },
+    {
+      "epoch": 0.3233082706766917,
+      "grad_norm": 1.241855373078175,
+      "learning_rate": 8.557445403028122e-05,
+      "loss": 0.1907,
+      "step": 817
+    },
+    {
+      "epoch": 0.32370399683419077,
+      "grad_norm": 0.6956449753509186,
+      "learning_rate": 8.552588033831905e-05,
+      "loss": 0.1316,
+      "step": 818
+    },
+    {
+      "epoch": 0.32409972299168976,
+      "grad_norm": 1.2506381497847148,
+      "learning_rate": 8.547723884106274e-05,
+      "loss": 0.2073,
+      "step": 819
+    },
+    {
+      "epoch": 0.32449544914918876,
+      "grad_norm": 0.7410411089502383,
+      "learning_rate": 8.542852963135029e-05,
+      "loss": 0.1695,
+      "step": 820
+    },
+    {
+      "epoch": 0.32449544914918876,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9375,
+      "eval_PRM F1 AUC": 0.6748647319232661,
+      "eval_PRM F1 AUC (fixed)": 0.9043285784554845,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.7777777777777778,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.29212239384651184,
+      "eval_runtime": 5.1898,
+      "eval_samples_per_second": 5.781,
+      "eval_steps_per_second": 0.193,
+      "step": 820
+    },
+    {
+      "epoch": 0.32489117530668776,
+      "grad_norm": 1.0926492341151244,
+      "learning_rate": 8.537975280214889e-05,
+      "loss": 0.1957,
+      "step": 821
+    },
+    {
+      "epoch": 0.32528690146418676,
+      "grad_norm": 1.135039526246786,
+      "learning_rate": 8.533090844655482e-05,
+      "loss": 0.1414,
+      "step": 822
+    },
+    {
+      "epoch": 0.3256826276216858,
+      "grad_norm": 1.3071292433628887,
+      "learning_rate": 8.528199665779328e-05,
+      "loss": 0.1876,
+      "step": 823
+    },
+    {
+      "epoch": 0.3260783537791848,
+      "grad_norm": 0.8116672832951006,
+      "learning_rate": 8.523301752921811e-05,
+      "loss": 0.1551,
+      "step": 824
+    },
+    {
+      "epoch": 0.3264740799366838,
+      "grad_norm": 1.027783869631157,
+      "learning_rate": 8.518397115431169e-05,
+      "loss": 0.2153,
+      "step": 825
+    },
+    {
+      "epoch": 0.3264740799366838,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.9001475651746188,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2613281309604645,
+      "eval_runtime": 5.3164,
+      "eval_samples_per_second": 5.643,
+      "eval_steps_per_second": 0.188,
+      "step": 825
+    },
+    {
+      "epoch": 0.3268698060941828,
+      "grad_norm": 0.8780206436295659,
+      "learning_rate": 8.513485762668476e-05,
+      "loss": 0.2199,
+      "step": 826
+    },
+    {
+      "epoch": 0.3272655322516818,
+      "grad_norm": 0.9425617394206817,
+      "learning_rate": 8.508567704007627e-05,
+      "loss": 0.145,
+      "step": 827
+    },
+    {
+      "epoch": 0.32766125840918087,
+      "grad_norm": 1.2828138498832955,
+      "learning_rate": 8.503642948835305e-05,
+      "loss": 0.2381,
+      "step": 828
+    },
+    {
+      "epoch": 0.32805698456667987,
+      "grad_norm": 0.8888925013939754,
+      "learning_rate": 8.498711506550983e-05,
+      "loss": 0.1869,
+      "step": 829
+    },
+    {
+      "epoch": 0.32845271072417886,
+      "grad_norm": 1.0940759254085846,
+      "learning_rate": 8.493773386566899e-05,
+      "loss": 0.2316,
+      "step": 830
+    },
+    {
+      "epoch": 0.32845271072417886,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.897196261682243,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2684895694255829,
+      "eval_runtime": 5.1585,
+      "eval_samples_per_second": 5.816,
+      "eval_steps_per_second": 0.194,
+      "step": 830
+    },
+    {
+      "epoch": 0.32884843688167786,
+      "grad_norm": 0.7567948350209694,
+      "learning_rate": 8.488828598308028e-05,
+      "loss": 0.1438,
+      "step": 831
+    },
+    {
+      "epoch": 0.32924416303917686,
+      "grad_norm": 0.7678359729015216,
+      "learning_rate": 8.483877151212077e-05,
+      "loss": 0.1517,
+      "step": 832
+    },
+    {
+      "epoch": 0.3296398891966759,
+      "grad_norm": 0.9923361241209745,
+      "learning_rate": 8.47891905472946e-05,
+      "loss": 0.1844,
+      "step": 833
+    },
+    {
+      "epoch": 0.3300356153541749,
+      "grad_norm": 0.7085662804763562,
+      "learning_rate": 8.473954318323287e-05,
+      "loss": 0.1897,
+      "step": 834
+    },
+    {
+      "epoch": 0.3304313415116739,
+      "grad_norm": 0.7147096103898686,
+      "learning_rate": 8.468982951469333e-05,
+      "loss": 0.1607,
+      "step": 835
+    },
+    {
+      "epoch": 0.3304313415116739,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.8944909001475652,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.2911783754825592,
+      "eval_runtime": 5.1004,
+      "eval_samples_per_second": 5.882,
+      "eval_steps_per_second": 0.196,
+      "step": 835
+    },
+    {
+      "epoch": 0.3308270676691729,
+      "grad_norm": 0.6596445924702422,
+      "learning_rate": 8.464004963656037e-05,
+      "loss": 0.1603,
+      "step": 836
+    },
+    {
+      "epoch": 0.33122279382667197,
+      "grad_norm": 0.5971853090797905,
+      "learning_rate": 8.45902036438447e-05,
+      "loss": 0.1523,
+      "step": 837
+    },
+    {
+      "epoch": 0.33161851998417097,
+      "grad_norm": 0.6966287567958265,
+      "learning_rate": 8.454029163168317e-05,
+      "loss": 0.1741,
+      "step": 838
+    },
+    {
+      "epoch": 0.33201424614166997,
+      "grad_norm": 0.6851643180686253,
+      "learning_rate": 8.449031369533876e-05,
+      "loss": 0.1682,
+      "step": 839
+    },
+    {
+      "epoch": 0.33240997229916897,
+      "grad_norm": 0.7036682208218882,
+      "learning_rate": 8.444026993020017e-05,
+      "loss": 0.2282,
+      "step": 840
+    },
+    {
+      "epoch": 0.33240997229916897,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.8947368421052633,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.28925782442092896,
+      "eval_runtime": 5.0929,
+      "eval_samples_per_second": 5.891,
+      "eval_steps_per_second": 0.196,
+      "step": 840
+    },
+    {
+      "epoch": 0.33280569845666796,
+      "grad_norm": 0.6675456484859625,
+      "learning_rate": 8.439016043178176e-05,
+      "loss": 0.1559,
+      "step": 841
+    },
+    {
+      "epoch": 0.333201424614167,
+      "grad_norm": 0.7725292485843828,
+      "learning_rate": 8.433998529572338e-05,
+      "loss": 0.1756,
+      "step": 842
+    },
+    {
+      "epoch": 0.333597150771666,
+      "grad_norm": 0.5651744758485415,
+      "learning_rate": 8.428974461779014e-05,
+      "loss": 0.1308,
+      "step": 843
+    },
+    {
+      "epoch": 0.333992876929165,
+      "grad_norm": 0.8938081414973864,
+      "learning_rate": 8.423943849387223e-05,
+      "loss": 0.1983,
+      "step": 844
+    },
+    {
+      "epoch": 0.334388603086664,
+      "grad_norm": 1.083079267779291,
+      "learning_rate": 8.418906701998477e-05,
+      "loss": 0.2044,
+      "step": 845
+    },
+    {
+      "epoch": 0.334388603086664,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.9038366945400886,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.27307942509651184,
+      "eval_runtime": 5.0396,
+      "eval_samples_per_second": 5.953,
+      "eval_steps_per_second": 0.198,
+      "step": 845
+    },
+    {
+      "epoch": 0.334784329244163,
+      "grad_norm": 1.237028516692739,
+      "learning_rate": 8.413863029226762e-05,
+      "loss": 0.2041,
+      "step": 846
+    },
+    {
+      "epoch": 0.33518005540166207,
+      "grad_norm": 0.7362456243045792,
+      "learning_rate": 8.408812840698517e-05,
+      "loss": 0.1298,
+      "step": 847
+    },
+    {
+      "epoch": 0.33557578155916107,
+      "grad_norm": 0.8836264510234058,
+      "learning_rate": 8.403756146052617e-05,
+      "loss": 0.2198,
+      "step": 848
+    },
+    {
+      "epoch": 0.33597150771666007,
+      "grad_norm": 0.5770378079109552,
+      "learning_rate": 8.398692954940352e-05,
+      "loss": 0.1073,
+      "step": 849
+    },
+    {
+      "epoch": 0.33636723387415907,
+      "grad_norm": 1.4507493057396252,
+      "learning_rate": 8.393623277025415e-05,
+      "loss": 0.2093,
+      "step": 850
+    },
+    {
+      "epoch": 0.33636723387415907,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.9011313330054107,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.27347004413604736,
+      "eval_runtime": 5.0973,
+      "eval_samples_per_second": 5.885,
+      "eval_steps_per_second": 0.196,
+      "step": 850
+    },
+    {
+      "epoch": 0.33676296003165807,
+      "grad_norm": 0.8132004482782825,
+      "learning_rate": 8.388547121983881e-05,
+      "loss": 0.1809,
+      "step": 851
+    },
+    {
+      "epoch": 0.3371586861891571,
+      "grad_norm": 0.813328882882879,
+      "learning_rate": 8.383464499504183e-05,
+      "loss": 0.1313,
+      "step": 852
+    },
+    {
+      "epoch": 0.3375544123466561,
+      "grad_norm": 0.8381291582903194,
+      "learning_rate": 8.378375419287099e-05,
+      "loss": 0.1769,
+      "step": 853
+    },
+    {
+      "epoch": 0.3379501385041551,
+      "grad_norm": 0.6462416856495375,
+      "learning_rate": 8.373279891045735e-05,
+      "loss": 0.1444,
+      "step": 854
+    },
+    {
+      "epoch": 0.3383458646616541,
+      "grad_norm": 0.6221755845055602,
+      "learning_rate": 8.368177924505504e-05,
+      "loss": 0.1343,
+      "step": 855
+    },
+    {
+      "epoch": 0.3383458646616541,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.9021151008362027,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2757161557674408,
+      "eval_runtime": 4.9916,
+      "eval_samples_per_second": 6.01,
+      "eval_steps_per_second": 0.2,
+      "step": 855
+    },
+    {
+      "epoch": 0.33874159081915317,
+      "grad_norm": 0.7442770131734728,
+      "learning_rate": 8.363069529404102e-05,
+      "loss": 0.1787,
+      "step": 856
+    },
+    {
+      "epoch": 0.33913731697665217,
+      "grad_norm": 0.8298459969984038,
+      "learning_rate": 8.357954715491498e-05,
+      "loss": 0.1956,
+      "step": 857
+    },
+    {
+      "epoch": 0.33953304313415117,
+      "grad_norm": 0.8822403855159155,
+      "learning_rate": 8.352833492529914e-05,
+      "loss": 0.1922,
+      "step": 858
+    },
+    {
+      "epoch": 0.33992876929165017,
+      "grad_norm": 0.9979725148915223,
+      "learning_rate": 8.3477058702938e-05,
+      "loss": 0.1866,
+      "step": 859
+    },
+    {
+      "epoch": 0.34032449544914917,
+      "grad_norm": 0.6889519291640593,
+      "learning_rate": 8.342571858569826e-05,
+      "loss": 0.1687,
+      "step": 860
+    },
+    {
+      "epoch": 0.34032449544914917,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.9008853910477127,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2668619751930237,
+      "eval_runtime": 4.8792,
+      "eval_samples_per_second": 6.149,
+      "eval_steps_per_second": 0.205,
+      "step": 860
+    },
+    {
+      "epoch": 0.3407202216066482,
+      "grad_norm": 0.8316984538300345,
+      "learning_rate": 8.337431467156851e-05,
+      "loss": 0.1574,
+      "step": 861
+    },
+    {
+      "epoch": 0.3411159477641472,
+      "grad_norm": 1.7234635973632797,
+      "learning_rate": 8.332284705865914e-05,
+      "loss": 0.2331,
+      "step": 862
+    },
+    {
+      "epoch": 0.3415116739216462,
+      "grad_norm": 0.9132787253556621,
+      "learning_rate": 8.327131584520207e-05,
+      "loss": 0.1717,
+      "step": 863
+    },
+    {
+      "epoch": 0.3419074000791452,
+      "grad_norm": 1.1689814264870746,
+      "learning_rate": 8.321972112955068e-05,
+      "loss": 0.2204,
+      "step": 864
+    },
+    {
+      "epoch": 0.3423031262366442,
+      "grad_norm": 0.668992889648423,
+      "learning_rate": 8.31680630101795e-05,
+      "loss": 0.1289,
+      "step": 865
+    },
+    {
+      "epoch": 0.3423031262366442,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.8991637973438269,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2648111879825592,
+      "eval_runtime": 5.1813,
+      "eval_samples_per_second": 5.79,
+      "eval_steps_per_second": 0.193,
+      "step": 865
+    },
+    {
+      "epoch": 0.34269885239414327,
+      "grad_norm": 0.9898982801677739,
+      "learning_rate": 8.311634158568408e-05,
+      "loss": 0.2062,
+      "step": 866
+    },
+    {
+      "epoch": 0.34309457855164227,
+      "grad_norm": 0.6630768646564437,
+      "learning_rate": 8.306455695478081e-05,
+      "loss": 0.141,
+      "step": 867
+    },
+    {
+      "epoch": 0.34349030470914127,
+      "grad_norm": 1.3165642093943424,
+      "learning_rate": 8.301270921630673e-05,
+      "loss": 0.2335,
+      "step": 868
+    },
+    {
+      "epoch": 0.34388603086664027,
+      "grad_norm": 0.9142797355976449,
+      "learning_rate": 8.296079846921927e-05,
+      "loss": 0.153,
+      "step": 869
+    },
+    {
+      "epoch": 0.3442817570241393,
+      "grad_norm": 0.9216845264807406,
+      "learning_rate": 8.290882481259618e-05,
+      "loss": 0.1797,
+      "step": 870
+    },
+    {
+      "epoch": 0.3442817570241393,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.9321266968325792,
+      "eval_PRM F1 AUC": 0.6918347270044269,
+      "eval_PRM F1 AUC (fixed)": 0.8984259714707329,
+      "eval_PRM F1 Neg": 0.5161290322580645,
+      "eval_PRM NPV": 0.6666666666666666,
+      "eval_PRM Precision": 0.9035087719298246,
+      "eval_PRM Recall": 0.9626168224299065,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.26598307490348816,
+      "eval_runtime": 5.1912,
+      "eval_samples_per_second": 5.779,
+      "eval_steps_per_second": 0.193,
+      "step": 870
+    },
+    {
+      "epoch": 0.3446774831816383,
+      "grad_norm": 0.8651769085395622,
+      "learning_rate": 8.285678834563524e-05,
+      "loss": 0.155,
+      "step": 871
+    },
+    {
+      "epoch": 0.3450732093391373,
+      "grad_norm": 0.9518289213674653,
+      "learning_rate": 8.280468916765415e-05,
+      "loss": 0.1035,
+      "step": 872
+    },
+    {
+      "epoch": 0.3454689354966363,
+      "grad_norm": 1.087527989751469,
+      "learning_rate": 8.275252737809028e-05,
+      "loss": 0.1987,
+      "step": 873
+    },
+    {
+      "epoch": 0.3458646616541353,
+      "grad_norm": 0.8612811027947653,
+      "learning_rate": 8.270030307650048e-05,
+      "loss": 0.1976,
+      "step": 874
+    },
+    {
+      "epoch": 0.3462603878116344,
+      "grad_norm": 0.8266407382637784,
+      "learning_rate": 8.264801636256094e-05,
+      "loss": 0.1474,
+      "step": 875
+    },
+    {
+      "epoch": 0.3462603878116344,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.9038366945400885,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2710612118244171,
+      "eval_runtime": 5.1806,
+      "eval_samples_per_second": 5.791,
+      "eval_steps_per_second": 0.193,
+      "step": 875
+    },
+    {
+      "epoch": 0.3466561139691334,
+      "grad_norm": 1.2250109101178033,
+      "learning_rate": 8.259566733606696e-05,
+      "loss": 0.2165,
+      "step": 876
+    },
+    {
+      "epoch": 0.34705184012663237,
+      "grad_norm": 0.8999170721823463,
+      "learning_rate": 8.25432560969328e-05,
+      "loss": 0.1934,
+      "step": 877
+    },
+    {
+      "epoch": 0.34744756628413137,
+      "grad_norm": 1.2885561911684744,
+      "learning_rate": 8.24907827451914e-05,
+      "loss": 0.2555,
+      "step": 878
+    },
+    {
+      "epoch": 0.34784329244163037,
+      "grad_norm": 1.2109685924497229,
+      "learning_rate": 8.243824738099431e-05,
+      "loss": 0.2004,
+      "step": 879
+    },
+    {
+      "epoch": 0.3482390185991294,
+      "grad_norm": 1.298688275406524,
+      "learning_rate": 8.238565010461138e-05,
+      "loss": 0.2294,
+      "step": 880
+    },
+    {
+      "epoch": 0.3482390185991294,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.9055582882439744,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2764648497104645,
+      "eval_runtime": 5.1409,
+      "eval_samples_per_second": 5.836,
+      "eval_steps_per_second": 0.195,
+      "step": 880
+    },
+    {
+      "epoch": 0.3486347447566284,
+      "grad_norm": 1.1408633103873533,
+      "learning_rate": 8.23329910164307e-05,
+      "loss": 0.2087,
+      "step": 881
+    },
+    {
+      "epoch": 0.3490304709141274,
+      "grad_norm": 1.1464720838539775,
+      "learning_rate": 8.228027021695827e-05,
+      "loss": 0.2062,
+      "step": 882
+    },
+    {
+      "epoch": 0.3494261970716264,
+      "grad_norm": 0.7193282757881734,
+      "learning_rate": 8.22274878068179e-05,
+      "loss": 0.153,
+      "step": 883
+    },
+    {
+      "epoch": 0.3498219232291254,
+      "grad_norm": 1.0106511876092554,
+      "learning_rate": 8.2174643886751e-05,
+      "loss": 0.2208,
+      "step": 884
+    },
+    {
+      "epoch": 0.3502176493866245,
+      "grad_norm": 0.9517391729236356,
+      "learning_rate": 8.212173855761636e-05,
+      "loss": 0.2146,
+      "step": 885
+    },
+    {
+      "epoch": 0.3502176493866245,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9369369369369369,
+      "eval_PRM F1 AUC": 0.6965076242006887,
+      "eval_PRM F1 AUC (fixed)": 0.9080177078209543,
+      "eval_PRM F1 Neg": 0.5333333333333333,
+      "eval_PRM NPV": 0.7272727272727273,
+      "eval_PRM Precision": 0.9043478260869565,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.25589191913604736,
+      "eval_runtime": 4.7861,
+      "eval_samples_per_second": 6.268,
+      "eval_steps_per_second": 0.209,
+      "step": 885
+    },
+    {
+      "epoch": 0.3506133755441235,
+      "grad_norm": 0.8364283541995906,
+      "learning_rate": 8.206877192038995e-05,
+      "loss": 0.1736,
+      "step": 886
+    },
+    {
+      "epoch": 0.3510091017016225,
+      "grad_norm": 0.9185753485607663,
+      "learning_rate": 8.201574407616483e-05,
+      "loss": 0.2102,
+      "step": 887
+    },
+    {
+      "epoch": 0.35140482785912147,
+      "grad_norm": 0.7018609574067978,
+      "learning_rate": 8.196265512615081e-05,
+      "loss": 0.1564,
+      "step": 888
+    },
+    {
+      "epoch": 0.3518005540166205,
+      "grad_norm": 1.1693031030621928,
+      "learning_rate": 8.190950517167437e-05,
+      "loss": 0.221,
+      "step": 889
+    },
+    {
+      "epoch": 0.3521962801741195,
+      "grad_norm": 0.6653103853980346,
+      "learning_rate": 8.18562943141784e-05,
+      "loss": 0.1346,
+      "step": 890
+    },
+    {
+      "epoch": 0.3521962801741195,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.9067879980324643,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.2787109315395355,
+      "eval_runtime": 4.8288,
+      "eval_samples_per_second": 6.213,
+      "eval_steps_per_second": 0.207,
+      "step": 890
+    },
+    {
+      "epoch": 0.3525920063316185,
+      "grad_norm": 0.7886924397386983,
+      "learning_rate": 8.180302265522206e-05,
+      "loss": 0.208,
+      "step": 891
+    },
+    {
+      "epoch": 0.3529877324891175,
+      "grad_norm": 0.7837903777469855,
+      "learning_rate": 8.174969029648052e-05,
+      "loss": 0.1589,
+      "step": 892
+    },
+    {
+      "epoch": 0.3533834586466165,
+      "grad_norm": 0.9005986478126117,
+      "learning_rate": 8.169629733974482e-05,
+      "loss": 0.1926,
+      "step": 893
+    },
+    {
+      "epoch": 0.3537791848041156,
+      "grad_norm": 0.8139863060862231,
+      "learning_rate": 8.164284388692166e-05,
+      "loss": 0.1706,
+      "step": 894
+    },
+    {
+      "epoch": 0.3541749109616146,
+      "grad_norm": 0.6170196410524107,
+      "learning_rate": 8.158933004003319e-05,
+      "loss": 0.1529,
+      "step": 895
+    },
+    {
+      "epoch": 0.3541749109616146,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.9070339399901624,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.28422850370407104,
+      "eval_runtime": 5.3817,
+      "eval_samples_per_second": 5.574,
+      "eval_steps_per_second": 0.186,
+      "step": 895
+    },
+    {
+      "epoch": 0.3545706371191136,
+      "grad_norm": 0.7476950704069479,
+      "learning_rate": 8.153575590121686e-05,
+      "loss": 0.1748,
+      "step": 896
+    },
+    {
+      "epoch": 0.3549663632766126,
+      "grad_norm": 0.8524765376473925,
+      "learning_rate": 8.148212157272517e-05,
+      "loss": 0.1749,
+      "step": 897
+    },
+    {
+      "epoch": 0.35536208943411157,
+      "grad_norm": 0.7538529262656093,
+      "learning_rate": 8.142842715692548e-05,
+      "loss": 0.1635,
+      "step": 898
+    },
+    {
+      "epoch": 0.3557578155916106,
+      "grad_norm": 0.8730864342562548,
+      "learning_rate": 8.137467275629988e-05,
+      "loss": 0.1708,
+      "step": 899
+    },
+    {
+      "epoch": 0.3561535417491096,
+      "grad_norm": 0.5626621536324666,
+      "learning_rate": 8.132085847344493e-05,
+      "loss": 0.1412,
+      "step": 900
+    },
+    {
+      "epoch": 0.3561535417491096,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9380530973451328,
+      "eval_PRM F1 AUC": 0.6532218396458436,
+      "eval_PRM F1 AUC (fixed)": 0.9026069847515986,
+      "eval_PRM F1 Neg": 0.46153846153846156,
+      "eval_PRM NPV": 0.8571428571428571,
+      "eval_PRM Precision": 0.8907563025210085,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3157894736842105,
+      "eval_loss": 0.2946940064430237,
+      "eval_runtime": 4.9661,
+      "eval_samples_per_second": 6.041,
+      "eval_steps_per_second": 0.201,
+      "step": 900
+    },
+    {
+      "epoch": 0.3565492679066086,
+      "grad_norm": 0.7534223585165879,
+      "learning_rate": 8.126698441107146e-05,
+      "loss": 0.178,
+      "step": 901
+    },
+    {
+      "epoch": 0.3569449940641076,
+      "grad_norm": 0.8770347695832404,
+      "learning_rate": 8.121305067200442e-05,
+      "loss": 0.1336,
+      "step": 902
+    },
+    {
+      "epoch": 0.3573407202216066,
+      "grad_norm": 0.895664182084698,
+      "learning_rate": 8.115905735918268e-05,
+      "loss": 0.1736,
+      "step": 903
+    },
+    {
+      "epoch": 0.3577364463791057,
+      "grad_norm": 0.8759322368558397,
+      "learning_rate": 8.110500457565873e-05,
+      "loss": 0.1802,
+      "step": 904
+    },
+    {
+      "epoch": 0.3581321725366047,
+      "grad_norm": 1.0048722558881418,
+      "learning_rate": 8.105089242459866e-05,
+      "loss": 0.147,
+      "step": 905
+    },
+    {
+      "epoch": 0.3581321725366047,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.8947368421052632,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.2684570252895355,
+      "eval_runtime": 5.1537,
+      "eval_samples_per_second": 5.821,
+      "eval_steps_per_second": 0.194,
+      "step": 905
+    },
+    {
+      "epoch": 0.3585278986941037,
+      "grad_norm": 0.8539026591098068,
+      "learning_rate": 8.099672100928184e-05,
+      "loss": 0.2033,
+      "step": 906
+    },
+    {
+      "epoch": 0.3589236248516027,
+      "grad_norm": 1.3646544143578732,
+      "learning_rate": 8.094249043310073e-05,
+      "loss": 0.2432,
+      "step": 907
+    },
+    {
+      "epoch": 0.35931935100910173,
+      "grad_norm": 0.8815465613362381,
+      "learning_rate": 8.088820079956074e-05,
+      "loss": 0.1462,
+      "step": 908
+    },
+    {
+      "epoch": 0.3597150771666007,
+      "grad_norm": 0.8491536790059588,
+      "learning_rate": 8.083385221227997e-05,
+      "loss": 0.1465,
+      "step": 909
+    },
+    {
+      "epoch": 0.3601108033240997,
+      "grad_norm": 0.8261044063135888,
+      "learning_rate": 8.077944477498905e-05,
+      "loss": 0.154,
+      "step": 910
+    },
+    {
+      "epoch": 0.3601108033240997,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.8937530742744713,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.26663410663604736,
+      "eval_runtime": 5.2415,
+      "eval_samples_per_second": 5.724,
+      "eval_steps_per_second": 0.191,
+      "step": 910
+    },
+    {
+      "epoch": 0.3605065294815987,
+      "grad_norm": 0.9567908722972123,
+      "learning_rate": 8.072497859153091e-05,
+      "loss": 0.1667,
+      "step": 911
+    },
+    {
+      "epoch": 0.3609022556390977,
+      "grad_norm": 1.1472506022125033,
+      "learning_rate": 8.067045376586066e-05,
+      "loss": 0.2597,
+      "step": 912
+    },
+    {
+      "epoch": 0.3612979817965968,
+      "grad_norm": 1.0171193652971011,
+      "learning_rate": 8.061587040204528e-05,
+      "loss": 0.1249,
+      "step": 913
+    },
+    {
+      "epoch": 0.3616937079540958,
+      "grad_norm": 1.3550690920076445,
+      "learning_rate": 8.056122860426352e-05,
+      "loss": 0.2233,
+      "step": 914
+    },
+    {
+      "epoch": 0.3620894341115948,
+      "grad_norm": 0.8560245996279463,
+      "learning_rate": 8.050652847680562e-05,
+      "loss": 0.181,
+      "step": 915
+    },
+    {
+      "epoch": 0.3620894341115948,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.933920704845815,
+      "eval_PRM F1 AUC": 0.6269060501721594,
+      "eval_PRM F1 AUC (fixed)": 0.8935071323167733,
+      "eval_PRM F1 Neg": 0.4,
+      "eval_PRM NPV": 0.8333333333333334,
+      "eval_PRM Precision": 0.8833333333333333,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.2631578947368421,
+      "eval_loss": 0.31640625,
+      "eval_runtime": 5.1641,
+      "eval_samples_per_second": 5.809,
+      "eval_steps_per_second": 0.194,
+      "step": 915
+    },
+    {
+      "epoch": 0.3624851602690938,
+      "grad_norm": 1.34936282084295,
+      "learning_rate": 8.045177012407316e-05,
+      "loss": 0.1816,
+      "step": 916
+    },
+    {
+      "epoch": 0.3628808864265928,
+      "grad_norm": 1.0256281829092346,
+      "learning_rate": 8.039695365057887e-05,
+      "loss": 0.1906,
+      "step": 917
+    },
+    {
+      "epoch": 0.36327661258409183,
+      "grad_norm": 1.1601798823326257,
+      "learning_rate": 8.034207916094638e-05,
+      "loss": 0.2272,
+      "step": 918
+    },
+    {
+      "epoch": 0.36367233874159083,
+      "grad_norm": 0.8806699686211412,
+      "learning_rate": 8.028714675991006e-05,
+      "loss": 0.1842,
+      "step": 919
+    },
+    {
+      "epoch": 0.3640680648990898,
+      "grad_norm": 1.5865857329484852,
+      "learning_rate": 8.023215655231488e-05,
+      "loss": 0.292,
+      "step": 920
+    },
+    {
+      "epoch": 0.3640680648990898,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.8915395966551894,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2671875059604645,
+      "eval_runtime": 4.946,
+      "eval_samples_per_second": 6.065,
+      "eval_steps_per_second": 0.202,
+      "step": 920
+    },
+    {
+      "epoch": 0.3644637910565888,
+      "grad_norm": 0.6705389484948697,
+      "learning_rate": 8.017710864311599e-05,
+      "loss": 0.1569,
+      "step": 921
+    },
+    {
+      "epoch": 0.3648595172140878,
+      "grad_norm": 1.0123403523097558,
+      "learning_rate": 8.012200313737881e-05,
+      "loss": 0.1477,
+      "step": 922
+    },
+    {
+      "epoch": 0.3652552433715869,
+      "grad_norm": 1.0947664605424559,
+      "learning_rate": 8.006684014027862e-05,
+      "loss": 0.247,
+      "step": 923
+    },
+    {
+      "epoch": 0.3656509695290859,
+      "grad_norm": 0.6980648657843986,
+      "learning_rate": 8.001161975710045e-05,
+      "loss": 0.128,
+      "step": 924
+    },
+    {
+      "epoch": 0.3660466956865849,
+      "grad_norm": 0.8033232131722131,
+      "learning_rate": 7.995634209323886e-05,
+      "loss": 0.221,
+      "step": 925
+    },
+    {
+      "epoch": 0.3660466956865849,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9422222222222222,
+      "eval_PRM F1 AUC": 0.6795376291195278,
+      "eval_PRM F1 AUC (fixed)": 0.896458435809149,
+      "eval_PRM F1 Neg": 0.5185185185185185,
+      "eval_PRM NPV": 0.875,
+      "eval_PRM Precision": 0.8983050847457628,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.26611328125,
+      "eval_runtime": 5.1601,
+      "eval_samples_per_second": 5.814,
+      "eval_steps_per_second": 0.194,
+      "step": 925
+    },
+    {
+      "epoch": 0.3664424218440839,
+      "grad_norm": 0.7763032809624354,
+      "learning_rate": 7.990100725419771e-05,
+      "loss": 0.1454,
+      "step": 926
+    },
+    {
+      "epoch": 0.36683814800158293,
+      "grad_norm": 1.1357204737601847,
+      "learning_rate": 7.984561534559003e-05,
+      "loss": 0.1841,
+      "step": 927
+    },
+    {
+      "epoch": 0.36723387415908193,
+      "grad_norm": 0.8015236100480437,
+      "learning_rate": 7.979016647313774e-05,
+      "loss": 0.1554,
+      "step": 928
+    },
+    {
+      "epoch": 0.36762960031658093,
+      "grad_norm": 0.8381641470220285,
+      "learning_rate": 7.97346607426715e-05,
+      "loss": 0.1334,
+      "step": 929
+    },
+    {
+      "epoch": 0.36802532647407993,
+      "grad_norm": 0.7889919595847641,
+      "learning_rate": 7.96790982601305e-05,
+      "loss": 0.1616,
+      "step": 930
+    },
+    {
+      "epoch": 0.36802532647407993,
+      "eval_PRM Accuracy": 0.873015873015873,
+      "eval_PRM F1": 0.9298245614035088,
+      "eval_PRM F1 AUC": 0.6005902606984752,
+      "eval_PRM F1 AUC (fixed)": 0.9050664043285784,
+      "eval_PRM F1 Neg": 0.3333333333333333,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.8760330578512396,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.21052631578947367,
+      "eval_loss": 0.32275390625,
+      "eval_runtime": 5.2216,
+      "eval_samples_per_second": 5.745,
+      "eval_steps_per_second": 0.192,
+      "step": 930
+    },
+    {
+      "epoch": 0.3684210526315789,
+      "grad_norm": 1.2137875075708977,
+      "learning_rate": 7.962347913156218e-05,
+      "loss": 0.1983,
+      "step": 931
+    },
+    {
+      "epoch": 0.368816778789078,
+      "grad_norm": 1.0212924006651691,
+      "learning_rate": 7.956780346312218e-05,
+      "loss": 0.2174,
+      "step": 932
+    },
+    {
+      "epoch": 0.369212504946577,
+      "grad_norm": 1.2280396662918818,
+      "learning_rate": 7.951207136107401e-05,
+      "loss": 0.2034,
+      "step": 933
+    },
+    {
+      "epoch": 0.369608231104076,
+      "grad_norm": 1.4206643849057659,
+      "learning_rate": 7.945628293178891e-05,
+      "loss": 0.2465,
+      "step": 934
+    },
+    {
+      "epoch": 0.370003957261575,
+      "grad_norm": 0.8171085263301563,
+      "learning_rate": 7.940043828174562e-05,
+      "loss": 0.1313,
+      "step": 935
+    },
+    {
+      "epoch": 0.370003957261575,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9411764705882353,
+      "eval_PRM F1 AUC": 0.7228234136743729,
+      "eval_PRM F1 AUC (fixed)": 0.9124446630595179,
+      "eval_PRM F1 Neg": 0.5806451612903226,
+      "eval_PRM NPV": 0.75,
+      "eval_PRM Precision": 0.9122807017543859,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.47368421052631576,
+      "eval_loss": 0.24602864682674408,
+      "eval_runtime": 5.1448,
+      "eval_samples_per_second": 5.831,
+      "eval_steps_per_second": 0.194,
+      "step": 935
+    },
+    {
+      "epoch": 0.370399683419074,
+      "grad_norm": 1.4021018408112313,
+      "learning_rate": 7.934453751753017e-05,
+      "loss": 0.1737,
+      "step": 936
+    },
+    {
+      "epoch": 0.37079540957657303,
+      "grad_norm": 1.8067271031971475,
+      "learning_rate": 7.928858074583569e-05,
+      "loss": 0.2164,
+      "step": 937
+    },
+    {
+      "epoch": 0.37119113573407203,
+      "grad_norm": 1.264366749628032,
+      "learning_rate": 7.923256807346224e-05,
+      "loss": 0.1756,
+      "step": 938
+    },
+    {
+      "epoch": 0.37158686189157103,
+      "grad_norm": 1.2378286555676006,
+      "learning_rate": 7.917649960731655e-05,
+      "loss": 0.1816,
+      "step": 939
+    },
+    {
+      "epoch": 0.37198258804907003,
+      "grad_norm": 0.7591404751997808,
+      "learning_rate": 7.912037545441182e-05,
+      "loss": 0.1421,
+      "step": 940
+    },
+    {
+      "epoch": 0.37198258804907003,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9375,
+      "eval_PRM F1 AUC": 0.6748647319232661,
+      "eval_PRM F1 AUC (fixed)": 0.9146581406787998,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.7777777777777778,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.24915364384651184,
+      "eval_runtime": 5.0582,
+      "eval_samples_per_second": 5.931,
+      "eval_steps_per_second": 0.198,
+      "step": 940
+    },
+    {
+      "epoch": 0.372378314206569,
+      "grad_norm": 0.9935140380300687,
+      "learning_rate": 7.906419572186756e-05,
+      "loss": 0.1339,
+      "step": 941
+    },
+    {
+      "epoch": 0.3727740403640681,
+      "grad_norm": 1.1058985799495225,
+      "learning_rate": 7.900796051690937e-05,
+      "loss": 0.2501,
+      "step": 942
+    },
+    {
+      "epoch": 0.3731697665215671,
+      "grad_norm": 0.7937010098446688,
+      "learning_rate": 7.895166994686869e-05,
+      "loss": 0.1948,
+      "step": 943
+    },
+    {
+      "epoch": 0.3735654926790661,
+      "grad_norm": 1.0455700003601043,
+      "learning_rate": 7.889532411918267e-05,
+      "loss": 0.1841,
+      "step": 944
+    },
+    {
+      "epoch": 0.3739612188365651,
+      "grad_norm": 0.9725602260528471,
+      "learning_rate": 7.88389231413939e-05,
+      "loss": 0.1684,
+      "step": 945
+    },
+    {
+      "epoch": 0.3739612188365651,
+      "eval_PRM Accuracy": 0.8809523809523809,
+      "eval_PRM F1": 0.933920704845815,
+      "eval_PRM F1 AUC": 0.6269060501721594,
+      "eval_PRM F1 AUC (fixed)": 0.9158878504672897,
+      "eval_PRM F1 Neg": 0.4,
+      "eval_PRM NPV": 0.8333333333333334,
+      "eval_PRM Precision": 0.8833333333333333,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.2631578947368421,
+      "eval_loss": 0.2781575620174408,
+      "eval_runtime": 5.0131,
+      "eval_samples_per_second": 5.984,
+      "eval_steps_per_second": 0.199,
+      "step": 945
+    },
+    {
+      "epoch": 0.37435694499406413,
+      "grad_norm": 1.078264910314406,
+      "learning_rate": 7.878246712115022e-05,
+      "loss": 0.2273,
+      "step": 946
+    },
+    {
+      "epoch": 0.37475267115156313,
+      "grad_norm": 1.0811931291171262,
+      "learning_rate": 7.872595616620458e-05,
+      "loss": 0.2198,
+      "step": 947
+    },
+    {
+      "epoch": 0.37514839730906213,
+      "grad_norm": 1.1226757702102015,
+      "learning_rate": 7.86693903844147e-05,
+      "loss": 0.2247,
+      "step": 948
+    },
+    {
+      "epoch": 0.37554412346656113,
+      "grad_norm": 1.0826415708689405,
+      "learning_rate": 7.861276988374302e-05,
+      "loss": 0.1979,
+      "step": 949
+    },
+    {
+      "epoch": 0.37593984962406013,
+      "grad_norm": 1.1862370372748008,
+      "learning_rate": 7.855609477225635e-05,
+      "loss": 0.2086,
+      "step": 950
+    },
+    {
+      "epoch": 0.37593984962406013,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9375,
+      "eval_PRM F1 AUC": 0.6748647319232661,
+      "eval_PRM F1 AUC (fixed)": 0.9173635022134777,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.7777777777777778,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.2525065243244171,
+      "eval_runtime": 4.9205,
+      "eval_samples_per_second": 6.097,
+      "eval_steps_per_second": 0.203,
+      "step": 950
+    },
+    {
+      "epoch": 0.3763355757815592,
+      "grad_norm": 0.8486565289453545,
+      "learning_rate": 7.849936515812578e-05,
+      "loss": 0.2056,
+      "step": 951
+    },
+    {
+      "epoch": 0.3767313019390582,
+      "grad_norm": 0.7210638719103278,
+      "learning_rate": 7.844258114962642e-05,
+      "loss": 0.1255,
+      "step": 952
+    },
+    {
+      "epoch": 0.3771270280965572,
+      "grad_norm": 0.6569444304353967,
+      "learning_rate": 7.838574285513716e-05,
+      "loss": 0.1598,
+      "step": 953
+    },
+    {
+      "epoch": 0.3775227542540562,
+      "grad_norm": 0.9795940508799222,
+      "learning_rate": 7.832885038314053e-05,
+      "loss": 0.117,
+      "step": 954
+    },
+    {
+      "epoch": 0.3779184804115552,
+      "grad_norm": 0.9137163087822452,
+      "learning_rate": 7.82719038422225e-05,
+      "loss": 0.1754,
+      "step": 955
+    },
+    {
+      "epoch": 0.3779184804115552,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9375,
+      "eval_PRM F1 AUC": 0.6748647319232661,
+      "eval_PRM F1 AUC (fixed)": 0.9168716182980816,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.7777777777777778,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.2566080689430237,
+      "eval_runtime": 4.9491,
+      "eval_samples_per_second": 6.062,
+      "eval_steps_per_second": 0.202,
+      "step": 955
+    },
+    {
+      "epoch": 0.37831420656905423,
+      "grad_norm": 1.3946163286152813,
+      "learning_rate": 7.821490334107216e-05,
+      "loss": 0.1732,
+      "step": 956
+    },
+    {
+      "epoch": 0.37870993272655323,
+      "grad_norm": 0.8957317042914218,
+      "learning_rate": 7.815784898848163e-05,
+      "loss": 0.2181,
+      "step": 957
+    },
+    {
+      "epoch": 0.37910565888405223,
+      "grad_norm": 1.2690009363253354,
+      "learning_rate": 7.810074089334581e-05,
+      "loss": 0.2025,
+      "step": 958
+    },
+    {
+      "epoch": 0.37950138504155123,
+      "grad_norm": 0.5657547196893011,
+      "learning_rate": 7.804357916466216e-05,
+      "loss": 0.0974,
+      "step": 959
+    },
+    {
+      "epoch": 0.37989711119905023,
+      "grad_norm": 0.9762911393922666,
+      "learning_rate": 7.798636391153056e-05,
+      "loss": 0.2395,
+      "step": 960
+    },
+    {
+      "epoch": 0.37989711119905023,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9380530973451328,
+      "eval_PRM F1 AUC": 0.6532218396458436,
+      "eval_PRM F1 AUC (fixed)": 0.9109690113133301,
+      "eval_PRM F1 Neg": 0.46153846153846156,
+      "eval_PRM NPV": 0.8571428571428571,
+      "eval_PRM Precision": 0.8907563025210085,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.3157894736842105,
+      "eval_loss": 0.2858072817325592,
+      "eval_runtime": 5.1336,
+      "eval_samples_per_second": 5.844,
+      "eval_steps_per_second": 0.195,
+      "step": 960
+    },
+    {
+      "epoch": 0.3802928373565493,
+      "grad_norm": 1.047448753651798,
+      "learning_rate": 7.792909524315298e-05,
+      "loss": 0.22,
+      "step": 961
+    },
+    {
+      "epoch": 0.3806885635140483,
+      "grad_norm": 1.1084392876329912,
+      "learning_rate": 7.787177326883336e-05,
+      "loss": 0.1425,
+      "step": 962
+    },
+    {
+      "epoch": 0.3810842896715473,
+      "grad_norm": 1.350540964010863,
+      "learning_rate": 7.781439809797743e-05,
+      "loss": 0.2144,
+      "step": 963
+    },
+    {
+      "epoch": 0.3814800158290463,
+      "grad_norm": 1.0413842349607318,
+      "learning_rate": 7.775696984009236e-05,
+      "loss": 0.1588,
+      "step": 964
+    },
+    {
+      "epoch": 0.38187574198654534,
+      "grad_norm": 0.956582675892353,
+      "learning_rate": 7.769948860478669e-05,
+      "loss": 0.1631,
+      "step": 965
+    },
+    {
+      "epoch": 0.38187574198654534,
+      "eval_PRM Accuracy": 0.9047619047619048,
+      "eval_PRM F1": 0.9464285714285714,
+      "eval_PRM F1 AUC": 0.7058534185932119,
+      "eval_PRM F1 AUC (fixed)": 0.911706837186424,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.8888888888888888,
+      "eval_PRM Precision": 0.905982905982906,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2758137881755829,
+      "eval_runtime": 4.7485,
+      "eval_samples_per_second": 6.318,
+      "eval_steps_per_second": 0.211,
+      "step": 965
+    },
+    {
+      "epoch": 0.38227146814404434,
+      "grad_norm": 0.8330337149079056,
+      "learning_rate": 7.764195450177011e-05,
+      "loss": 0.1409,
+      "step": 966
+    },
+    {
+      "epoch": 0.38266719430154333,
+      "grad_norm": 0.9879636926532511,
+      "learning_rate": 7.758436764085315e-05,
+      "loss": 0.187,
+      "step": 967
+    },
+    {
+      "epoch": 0.38306292045904233,
+      "grad_norm": 0.8126722501498516,
+      "learning_rate": 7.752672813194707e-05,
+      "loss": 0.1648,
+      "step": 968
+    },
+    {
+      "epoch": 0.38345864661654133,
+      "grad_norm": 0.9812482162780913,
+      "learning_rate": 7.746903608506362e-05,
+      "loss": 0.1817,
+      "step": 969
+    },
+    {
+      "epoch": 0.3838543727740404,
+      "grad_norm": 0.9625135680327048,
+      "learning_rate": 7.74112916103148e-05,
+      "loss": 0.1587,
+      "step": 970
+    },
+    {
+      "epoch": 0.3838543727740404,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9411764705882353,
+      "eval_PRM F1 AUC": 0.7228234136743729,
+      "eval_PRM F1 AUC (fixed)": 0.9163797343826856,
+      "eval_PRM F1 Neg": 0.5806451612903226,
+      "eval_PRM NPV": 0.75,
+      "eval_PRM Precision": 0.9122807017543859,
+      "eval_PRM Recall": 0.9719626168224299,
+      "eval_PRM Specificty": 0.47368421052631576,
+      "eval_loss": 0.24947915971279144,
+      "eval_runtime": 5.0624,
+      "eval_samples_per_second": 5.926,
+      "eval_steps_per_second": 0.198,
+      "step": 970
+    },
+    {
+      "epoch": 0.3842500989315394,
+      "grad_norm": 0.7805689586454879,
+      "learning_rate": 7.73534948179127e-05,
+      "loss": 0.1242,
+      "step": 971
+    },
+    {
+      "epoch": 0.3846458250890384,
+      "grad_norm": 1.1125621100367127,
+      "learning_rate": 7.729564581816923e-05,
+      "loss": 0.1905,
+      "step": 972
+    },
+    {
+      "epoch": 0.3850415512465374,
+      "grad_norm": 1.0460512273276006,
+      "learning_rate": 7.723774472149601e-05,
+      "loss": 0.1768,
+      "step": 973
+    },
+    {
+      "epoch": 0.3854372774040364,
+      "grad_norm": 0.9058358879564835,
+      "learning_rate": 7.717979163840401e-05,
+      "loss": 0.1469,
+      "step": 974
+    },
+    {
+      "epoch": 0.38583300356153544,
+      "grad_norm": 1.077238475230232,
+      "learning_rate": 7.712178667950346e-05,
+      "loss": 0.1348,
+      "step": 975
+    },
+    {
+      "epoch": 0.38583300356153544,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.9181013280865715,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.25013020634651184,
+      "eval_runtime": 4.8996,
+      "eval_samples_per_second": 6.123,
+      "eval_steps_per_second": 0.204,
+      "step": 975
+    },
+    {
+      "epoch": 0.38622872971903444,
+      "grad_norm": 1.0799868694732282,
+      "learning_rate": 7.706372995550361e-05,
+      "loss": 0.1533,
+      "step": 976
+    },
+    {
+      "epoch": 0.38662445587653343,
+      "grad_norm": 1.2633713733016638,
+      "learning_rate": 7.700562157721254e-05,
+      "loss": 0.2576,
+      "step": 977
+    },
+    {
+      "epoch": 0.38702018203403243,
+      "grad_norm": 1.069770180047322,
+      "learning_rate": 7.694746165553682e-05,
+      "loss": 0.1861,
+      "step": 978
+    },
+    {
+      "epoch": 0.38741590819153143,
+      "grad_norm": 0.8497199090648525,
+      "learning_rate": 7.688925030148149e-05,
+      "loss": 0.1686,
+      "step": 979
+    },
+    {
+      "epoch": 0.3878116343490305,
+      "grad_norm": 1.4547279595576492,
+      "learning_rate": 7.683098762614971e-05,
+      "loss": 0.2087,
+      "step": 980
+    },
+    {
+      "epoch": 0.3878116343490305,
+      "eval_PRM Accuracy": 0.9047619047619048,
+      "eval_PRM F1": 0.9464285714285714,
+      "eval_PRM F1 AUC": 0.7058534185932119,
+      "eval_PRM F1 AUC (fixed)": 0.9153959665518938,
+      "eval_PRM F1 Neg": 0.5714285714285714,
+      "eval_PRM NPV": 0.8888888888888888,
+      "eval_PRM Precision": 0.905982905982906,
+      "eval_PRM Recall": 0.9906542056074766,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.2621419131755829,
+      "eval_runtime": 5.0453,
+      "eval_samples_per_second": 5.946,
+      "eval_steps_per_second": 0.198,
+      "step": 980
+    },
+    {
+      "epoch": 0.3882073605065295,
+      "grad_norm": 1.059626672976495,
+      "learning_rate": 7.677267374074262e-05,
+      "loss": 0.1883,
+      "step": 981
+    },
+    {
+      "epoch": 0.3886030866640285,
+      "grad_norm": 0.9351305620279964,
+      "learning_rate": 7.671430875655907e-05,
+      "loss": 0.1755,
+      "step": 982
+    },
+    {
+      "epoch": 0.3889988128215275,
+      "grad_norm": 0.4610310550817402,
+      "learning_rate": 7.665589278499547e-05,
+      "loss": 0.0865,
+      "step": 983
+    },
+    {
+      "epoch": 0.38939453897902654,
+      "grad_norm": 1.5068778942952485,
+      "learning_rate": 7.659742593754551e-05,
+      "loss": 0.3118,
+      "step": 984
+    },
+    {
+      "epoch": 0.38979026513652554,
+      "grad_norm": 1.2931522160932827,
+      "learning_rate": 7.653890832580002e-05,
+      "loss": 0.1822,
+      "step": 985
+    },
+    {
+      "epoch": 0.38979026513652554,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.9082636497786523,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.24667969346046448,
+      "eval_runtime": 5.548,
+      "eval_samples_per_second": 5.407,
+      "eval_steps_per_second": 0.18,
+      "step": 985
+    },
+    {
+      "epoch": 0.39018599129402454,
+      "grad_norm": 0.6926202905142572,
+      "learning_rate": 7.648034006144667e-05,
+      "loss": 0.1558,
+      "step": 986
+    },
+    {
+      "epoch": 0.39058171745152354,
+      "grad_norm": 0.6636695070528967,
+      "learning_rate": 7.642172125626986e-05,
+      "loss": 0.1385,
+      "step": 987
+    },
+    {
+      "epoch": 0.39097744360902253,
+      "grad_norm": 0.8028049784237185,
+      "learning_rate": 7.636305202215041e-05,
+      "loss": 0.1265,
+      "step": 988
+    },
+    {
+      "epoch": 0.3913731697665216,
+      "grad_norm": 0.7024509893389485,
+      "learning_rate": 7.630433247106543e-05,
+      "loss": 0.1471,
+      "step": 989
+    },
+    {
+      "epoch": 0.3917688959240206,
+      "grad_norm": 0.8983355345926901,
+      "learning_rate": 7.624556271508805e-05,
+      "loss": 0.2198,
+      "step": 990
+    },
+    {
+      "epoch": 0.3917688959240206,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.9109690113133301,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.244140625,
+      "eval_runtime": 5.052,
+      "eval_samples_per_second": 5.938,
+      "eval_steps_per_second": 0.198,
+      "step": 990
+    },
+    {
+      "epoch": 0.3921646220815196,
+      "grad_norm": 1.0022856692403896,
+      "learning_rate": 7.61867428663872e-05,
+      "loss": 0.2446,
+      "step": 991
+    },
+    {
+      "epoch": 0.3925603482390186,
+      "grad_norm": 0.6454129825230569,
+      "learning_rate": 7.612787303722744e-05,
+      "loss": 0.1265,
+      "step": 992
+    },
+    {
+      "epoch": 0.3929560743965176,
+      "grad_norm": 0.8218286001550995,
+      "learning_rate": 7.606895333996871e-05,
+      "loss": 0.1633,
+      "step": 993
+    },
+    {
+      "epoch": 0.39335180055401664,
+      "grad_norm": 1.7572035348522699,
+      "learning_rate": 7.600998388706615e-05,
+      "loss": 0.1339,
+      "step": 994
+    },
+    {
+      "epoch": 0.39374752671151564,
+      "grad_norm": 0.7835985377610045,
+      "learning_rate": 7.595096479106985e-05,
+      "loss": 0.2012,
+      "step": 995
+    },
+    {
+      "epoch": 0.39374752671151564,
+      "eval_PRM Accuracy": 0.8968253968253969,
+      "eval_PRM F1": 0.9417040358744395,
+      "eval_PRM F1 AUC": 0.7011805213969503,
+      "eval_PRM F1 AUC (fixed)": 0.9139203148057058,
+      "eval_PRM F1 Neg": 0.5517241379310345,
+      "eval_PRM NPV": 0.8,
+      "eval_PRM Precision": 0.9051724137931034,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.42105263157894735,
+      "eval_loss": 0.24492187798023224,
+      "eval_runtime": 5.2828,
+      "eval_samples_per_second": 5.679,
+      "eval_steps_per_second": 0.189,
+      "step": 995
+    },
+    {
+      "epoch": 0.39414325286901464,
+      "grad_norm": 1.0611986223039003,
+      "learning_rate": 7.589189616462465e-05,
+      "loss": 0.1715,
+      "step": 996
+    },
+    {
+      "epoch": 0.39453897902651364,
+      "grad_norm": 1.2109979184538513,
+      "learning_rate": 7.583277812046993e-05,
+      "loss": 0.2168,
+      "step": 997
+    },
+    {
+      "epoch": 0.3949347051840127,
+      "grad_norm": 0.9158287931741367,
+      "learning_rate": 7.577361077143939e-05,
+      "loss": 0.1774,
+      "step": 998
+    },
+    {
+      "epoch": 0.3953304313415117,
+      "grad_norm": 0.8044180426512219,
+      "learning_rate": 7.571439423046079e-05,
+      "loss": 0.1577,
+      "step": 999
+    },
+    {
+      "epoch": 0.3957261574990107,
+      "grad_norm": 1.0638141821573956,
+      "learning_rate": 7.565512861055589e-05,
+      "loss": 0.1678,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3957261574990107,
+      "eval_PRM Accuracy": 0.8888888888888888,
+      "eval_PRM F1": 0.9375,
+      "eval_PRM F1 AUC": 0.6748647319232661,
+      "eval_PRM F1 AUC (fixed)": 0.9131824889326119,
+      "eval_PRM F1 Neg": 0.5,
+      "eval_PRM NPV": 0.7777777777777778,
+      "eval_PRM Precision": 0.8974358974358975,
+      "eval_PRM Recall": 0.9813084112149533,
+      "eval_PRM Specificty": 0.3684210526315789,
+      "eval_loss": 0.2607421875,
+      "eval_runtime": 5.2004,
+      "eval_samples_per_second": 5.769,
+      "eval_steps_per_second": 0.192,
+      "step": 1000
     }
   ],
   "logging_steps": 1,
@@ -5243,7 +10443,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 754390401613824.0,
+  "total_flos": 1512256190087168.0,
   "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null